AGARD-CP-577

10
dL
u
Q
ff
< ADVISORY GROUP FOR AEROSPACE RESEARCH & DEVELOPMENT
<
7 RUE ANCELLE, 92200 NEUILLY-SUR-SEINE, FRANCE

AGARD CONFERENCE PROCEEDINGS 577

Flight Simulation —
Where are the Challenges?
(Simulation de vol — quels sont les defis?)

Copies of papers presented at the Flight Vehicle Integration Panel Symposium,

held in Braunschweig, Germany, from 22-25 May 1995.

19960715 mi

m NORTH ATLANTIC TREATY ORGANIZATION

(DISTRIBUTION &l'ATEWtlbW J§
in i r " ■ ii

Approved for public release;

Distribution Unlimited
i»

Published April 1996

Distribution and Availability on Back Cover

mm im Hg iBgP&.tl^
 AGARD-CP-577

ADVISORY GROUP FOR AEROSPACE RESEARCH & DEVELOPMENT

7 RUE ANCELLE, 92200 NEUILLY-SUR-SEINE, FRANCE

AGARD CONFERENCE PROCEEDINGS 577

Flight Simulation —
Where are the Challenges?
(Simulation de vol — quels sont les defis?)

Copies of papers presented at the Flight Vehicle Integration Panel Symposium,

held in Braunschweig, Germany, from 22-25 May 1995.

North Atlantic Treaty Organization

Organisation du Tratte de l'Atlantique Nord
 The Mission of AGARD

According to its Charter, the mission of AGARD is to bring together the leading personalities of the NATO nations in the
fields of science and technology relating to aerospace for the following purposes:

— Recommending effective ways for the member nations to use their research and development capabilities for the
common benefit of the NATO community;

— Providing scientific and technical advice and assistance to the Military Committee in the field of aerospace research
and development (with particular regard to its military application);

— Continuously stimulating advances in the aerospace sciences relevant to strengthening the common defence posture;

— Improving the co-operation among member nations in aerospace research and development;

— Exchange of scientific and technical information;

— Providing assistance to member nations for the purpose of increasing their scientific and technical potential;

— Rendering scientific and technical assistance, as requested, to other NATO bodies and to member nations in
connection with research and development problems in the aerospace field.

The highest authority within AGARD is the National Delegates Board consisting of officially appointed senior
representatives from each member nation. The mission of AGARD is carried out through the Panels which are composed of
experts appointed by the National Delegates, the Consultant and Exchange Programme and the Aerospace Applications
Studies Programme. The results of AGARD work are reported to the member nations and the NATO Authorities through the
AGARD series of publications of which this is one.

Participation in AGARD activities is by invitation only and is normally limited to citizens of the NATO nations.

The content of this publication has been reproduced

directly from material supplied by AGARD or the authors.

Published April 1996

Copyright © AGARD 1996

All Rights Reserved

ISBN 92-836-0024-X

Printed by Canada Communication Group

45 Sacre-Cceur Blvd., Hull (Quebec), Canada K1A 0S7
 Flight Simulation — Where are the Challenges?
(AGARD CP-577)

Executive Summary
Effective flight simulation is an important capability for NATO nations, and it will become even more
important in the face of reductions in defence budgets. The principal objectives of this symposium were
to review the state of the art in flight simulation, and to identify key areas where additional research and
development are needed. Presentations dealt with simulation used for engineering and for training, both
on the ground and in flight.
The symposium met its objectives. The papers contained in the Proceedings should be valuable to
anyone currently:
— considering the implications of procuring and applying simulators;
— developing simulators for training and engineering;
— using simulators for engineering;
— doing basic research in simulation.
The Keynote speech, by Major General G.K. Muellner, USAF, provided a dramatic stimulus to the
symposium by describing how advanced modelling and simulation methods are being used in the Joint
Advanced Strike Technology (JAST) programme to achieve an affordable weapon system.
Technical papers in the symposium deal with visual and motion cueing, advances in modelling,
simulation in design and development for both rotorcraft and fixed wing aircraft and their systems,
simulation in training, in-flight simulation, and the future applications of networked simulators. Many
of the papers were based on current military programmes: Eurocopter Tiger, Comanche, Osprey, Rafale
and the C-160, and included the use of piloted simulation in the evaluation process of bids for an attack
helicopter.
A significant theme emerged, to use simulators linked up via networking technology to expand the
scope of man-in-the-loop simulation into battlefield environments using a mixture of simulated and real
mission equipment. Cost savings in development have been achieved as a result.
Weaknesses in current flight simulation technology were identified in the area of visual displays, in
particular in achieving wider fields of view and higher resolution; in ground handling models; and in
verification and validation. A new approach to identifying and defining the features required in
simulated visual scenes was outlined; it would be useful to extend this new approach by further research
in NATO nations. There is still considerable debate on how much motion the pilot needs to 'feel', and
in what circumstances. This may well lead to a future study project in the Flight Vehicle Integration
Panel.
The most interesting trend apparent throughout the symposium, from the Keynote onwards, was the
concept that future simulators should be designed and procured for a wide spectrum of uses. A
simulator created to design air-to-air missiles should also be used to test missiles, and it should be
suitable for integration into a training system. It is even conceived, for example, that a simulator testing
infra-red suppression systems in France could be flown against a simulator developing heat-seeking
missiles in Canada, as part of a distributed simulation network exercise testing NATO's air C31. Such
multi-echelon usage will reduce costs and improve the realism of the synthetic environment being
simulated.
 Simulation de vol — quels sont les defis?
(AGARD CP-577)

Synthese
La competence en matiere de simulation du vol est d'une grande importance pour 1'ensemble des pays
membres de l'OTAN et le deviendra de plus en plus eu egard ä la reduction generalisee des budgets de
defense. Ce symposium a fait le point de l'etat de l'art dans le domaine de la simulation du vol, afin
d'identifier d'eventuels points faibles pour lesquels un effort supplemental de recherche et developpement
serait necessaire. Les presentations ont porte sur la simulation pour ingenierie et pour l'entrainement, tant au
sol qu'en vol.
Le symposium a atteint ses objectifs. Les communications figurant dans le compte-rendu de conference
doivent interesser tous ceux qui:
— envisagent l'achat et la mise en oeuvre de simulateurs
— travaillent au developpement de simulateurs pour l'entrainement et l'ingenierie
— utilisent des simulateurs aux fins de l'ingenierie
Le discours d'Ouvertüre, prononce par le Major-General G.K. Muellner USAF, a ete un stimulant opportun
pour le symposium, notamment par la description de 1'application de methodes avancees de modelisation et
de simulation dans le cadre du Programme conjoint des technologies de frappe avancees (JAST), qui a pour
objet la fabrication d'un Systeme d'armes abordable.
Les communications techniques du symposium ont porte sur la Stimulation visuelle et motrice, les derniers
progres en modelisation, la simulation dans la conception et le developpement d'aeronefs ä voilure fixe et ä
voilure tournante et de leurs systemes, la simulation et l'entrainement, la simulation en vol et les
applications futures des simulateurs en reseau. Bon nombre des communications, structurees autour de
programmes militaires actuellement en cours (Eurocopter Tiger, Comanche, Osprey, Rafale et C-160),
traitaient de l'emploi de la simulation pilotee lors du depouillement des soumissions relatives ä un
helicoptere d'attaque.
Le symposium a identifie le theme important de l'utilisation de simulateurs relies entre eux par les
techniques de reseaux, afin d'elargir le champ de simulation de «l'homme dans le boucle» dans les
environnements du champ de bataille, par l'association d'equipements operationnels reels et simules. II en
est resulte des economies de coüts de developpement.
Un certain nombre de lacunes ont ete identifiers dans le domaine des technologies de la simulation du vol,
en particulier en ce qui concerne les visualisations, ou il s'agit d'obtenir des champs de vision plus larges et
une meilleure resolution, les modeles des manoeuvres au sol, la verification et la validation. Une nouvelle
approche du probleme de 1'identification et la definition des caracteristiques topographiques demandees
pour la simulation de sequences visuelles a ete exposee. Le symposium a considere que ce sujet devait faire
l'objet de nouveaux programmes de recherche ä realiser par les pays membres de l'OTAN.
II existe toujours un debat important sur la question de la quantite de mouvement qui doit etre percue par le
pilote et, dans quelles circonstances? L'interet exprime pour ce sujet pourrait se concretiser sous la forme
d'un projet d'etude pour le Panel FVP.
La tendance la plus interessante, qui s'est manifestee tout au long du symposium, du discours d'ouverture
jusqu'ä la fin, a ete la notion de la conception et l'achat des futurs simulateurs offrant une large gamme
d'applications. Un simulateur etudie pour la conception de missiles air-air devra aussi servir pour les essais
et devra etre integrable dans un programme d'entrainement. II est meme envisage, par exemple, qu'un
simulateur d'essais d'un Systeme de suppression infrarouge developpe en France puisse etre compare ä un
autre au Canada travaillant sur le developpement de missiles ä autodirecteur infrarouge dans le cadre d'un
exercice de simulation d'un reseau reparti pour les essais des moyens C3i air de l'OTAN. De telles
applications multi-echelon permettront de reduire les coüts et d'ameliorer le realisme de l'environnement
synthetique simule.
 Contents

Page
Executive Summary iii
Synthese iv
Preface/Acknowledgements vii
Flight Vehicle Integration Panel viii
Reference
Technical Evaluation Report T
by L.D. Reid
Keynote Address K
by MGen G.K. Muellner
Historical Review of Piloted Simulation at NASA Ames 1
by S.B. Anderson
Simulated Visual Scenes — Which are the Critical Cues? 2
by H.M. Mclntyre and M.E.C. Roberts
Visual Scenes for Battlefield Helicopter Operations: Evaluation of Requirements and how to 3
specify them
by C.D.B. Deighton and A.A. Woodfield
Visual System Operational Evaluation 4
by J.E. Brown, Lt.Col. D.R. Poe, TJ. Lincourt and lLt. MJ. Leos
Yaw Motion Cues in Helicopter Simulation 5
by J.A. Schroeder and W.W. Johnson
Achieving High-Fidelity Motion Cues in Flight Simulation 6
by S.K. Advani and J.A. Mulder
APOGEE: A Breakthrough in Synthetic Image Generation 7
by J.-C. Chauvin
The RTSS Image Generation System 8
by K. Alvermann, S. Graeber, J.W.L.J. Mager and M.H. Smit
Ship Airwakes — A new Generic Model for Piloted Simulation 9
by A.A. Woodfield and B.N. Tomlinson
A Dynamic Challenge: Helicopter/Ship Interface Simulation — Development, Integration 10
and Application
by Lt. S.J. Täte
Simulation of Rotor Blade Element Turbulence 11
by R.E. McFarland and K. Duisenberg
C-160 Ground Handling Model Update using Taxi Test Data 12
by D. Fischenberg and W. Mönnich
Tactical Environment Servers 13
by F.D. Heran
UK Attack Helicopter Flying Qualities: The Role of Piloted Simulation Evaluation in 14
Supporting the Procurement Decision Making Process
by M.T. Charlton, G.D. Padfield and J.T. Green
Paper cancelled 15
The use of Piloted Simulation for Civil Tilt-Rotor Integrated Cockpit Design 16
by W.A. Decker, R.C. Simmons and G.E. Tucker
Simulation Applications in V-22 and RAH-66 Design and Development 17
by R.E. Sheffield and L.E. Lalicker
La simulation pilotee pour la conception et le developpement des appareils ä voilure 18
tournantes. L'experience d'EUROCOPTER et les tendances pour le futur:
Piloted Simulation in Rotorcraft Design & Development. EUROCOPTER Experience and Future
Trends
by P. Rollet
Technological and Economical Limits Experienced for R&D Mission Simulation 19
by P. Castoldi, M. Allocca, S.Lo Presti and M. Trifoglietti
Utilisation de la simulation pour la conception du SNA du Rafale 20
by R. Goussault and M. Leclere
ARC Segment Attitude Reference — ASAR 21
by W. Fuchs and G. Fischer
Issues in the Flight Clearance of Vehicle Management Systems 22
by DJ. Diston and B.R.C. Weiler
The Dutch National Simulation Facility: Advancements in Simulator Technology and 23
Application
by H.A.J.M. Offerman
Simulateur multicible pilote pour le developpement des conduites de tir air/air 24
by J.-E. Chevillot
Paper cancelled 25
Printed version not available 26
Issues in the Development of Training Analysis Methodologies 27
by T. Simpson
Delayed Pilot Response in Windshear 28
by G. Schänzer and J. Krüger
ASRA — A New Tool for In-Flight Simulation. Current Status and Actuation Studies 29
by J.M. Morgan and S.W. Baillie
Research Applications and Capabilities of the NASA/Army Rotorcraft Aircrew Systems 30
Concepts Airborne Laboratory (RASCAL)
by E.W. Aiken, R.A. Jacobsen and W.S. Hindson
ATTAS & ATTHeS In-Flight Simulators 31
by J.J. Buchholz, J.-M. Bauschat, K.-U. Hahn and H.J. Pausder
Variable Stability In-Flight Simulator Test Aircraft (VISTA) 32
by K. Buehler, P. Reynolds, S. Markman and G. Hellmann
Experiences acquises lors de la mise en oeuvre des protocoles SID avec des simulations dejä 33
existantes
by C. Crassous de Medeuil
High Fidelity, Mobile, Net Workable Trainers. The Trainers of the Future? 34
by C.C. Miller, D. Perdue and S. Davis
Simulation, Distributed Simulation and Synthetic Environments 35
by C.E. Adolph and J. Thorpe
The Synthetic Environment — The Ultimate Defence Simulation? 36
by P. Beckett
Estimating the Training Effectiveness of Interactive Air Combat Simulation 37
by W.L. Waag and H.H. Bell
 Preface

Effective Flight Simulation is an important capability for NATO nations, and it will become even more important in the face
of reductions in defence budgets. This symposium reviewed the state of the art in flight simulation, in order to identify
weaknesses where additional research and development are needed. Presentations dealt with simulation used for engineering
and for training, both on the ground and in flight.

Papers on fixed and rotary wing, as well as tilt-rotor, aircraft are included. Much of the material was oriented on military
applications: Eurocopter, Comanche, Osprey, Rafale, and the C-160. Simulation was addressed both for individual systems
and for distributive networks.

Acknowledgements

The Flight Vehicle Integration Panel wishes to express its thanks to the National Authorities of Germany for the invitation to
hold this meeting in their country as well as to D.L.R. and D.G.L.R. for the facilities and personnel which made the meeting
possible.
 Flight Vehicle Integration Panel

Chairman: Dipl.-Ing. Horst WUENNENBERG

Mgr. of Flight Physics & Predesign
DORNIER Luftfahrt GmbH
P.O. Box 1103
D-82230 Wessling
Germany

TECHNICAL PROGRAMME COMMITTEE

Mr. B. TOMLINSON Dr. D. McERLEAN

Flight Dynamics & Simulation Dept. 1215 Jefferson Davis Blvd.
Right Systems Department Bldg. CG3, Suite 406
Defence Research Agency Arlington, VA 22202
Bedford MK41 6AE United States
United Kingdom

Mr. B.B. BLAKE

Director, Research & Technology
BOEING Defense & Space Group
Helicopters Division
P.O. Box 16858, Mail Stop P32-10
Philadelphia, PA 19142-0858
United States

HOST COORDINATOR

Dr.-Ing. Peter HAMEL

Director, Institut für Flugmechanik
D.L.R.
Postfach 3267
38022 Braunschweig
Germany

PANEL EXECUTIVE

John B. WHEATLEY, LTC, USA

Mail from Europe: Mail from USA and Canada:

LTC J.B. WHEATLEY AGARD-NATO/FVP
AGARD-OTAN/FVP PSC 116
7, rue Ancelle APO AE 09777
92200 Neuilly-sur-Seine
France
 T-1

Technical Evaluation Report

by

L. D. Reid
University of Toronto Institute for Aerospace Studies
4925 Dufferin Street
North York, Ontario, Canada
M3H 5T6

1. INTRODUCTION
now the UK Royal Navy. He outlined his vision for
Flight simulation continues to be an area of expanding achieving an affordable weapon system through the
applications and technical growth. This has been effective application of modelling and simulation,
driven to a considerable extent by demonstrated cost which he described as "the dominant factor" in
benefits that accrue from its use. Many of the papers achieving the program's goals. His vision was based
presented at this symposium touched upon military cost on four key areas. First, use modelling and simulation
savings associated with simulator use. In addition, to support joint government and industry studies of
increasingly groups of simulators are being networked possible trade-offs between cost and performance, and
together in order to expand their application into large- thus define an affordable requirement. Second, expand
scale military operations in order to study in a cost the role of simulation in the design and development
effective manner an expanding repertoire of battle process accomplishing as much as possible in the
scenarios. virtual world before building hardware, creating
software or planning assembly. Third, use expensive
The AGARD symposium "Flight Simulation — Where flight and ground tests only as validation points for
Are the Challenges?" addressed a wide range of topics results obtained through modelling and simulation.
covering the development and application of flight Fourth, design all models, databases and simulation
simulators. Most of the papers either directly dealt tools to be re-usable at all stages of the system's life
with military topics or could be applied to military cycle, including evaluation, training and force
aspects of flight simulation. deployment. He illustrated this vision by describing
how the JAST program, in what is truly a dramatic
The symposium, held in Braunschweig, Germany, from expansion of the role of modelling and simulation, is
May 22-25,1995, was divided into the following seven exploiting these concepts to achieve effectiveness,
sessions: Visual and motion cueing and requirements, supportability and, above all, affordability.
Advances in modelling, Simulation in design and
development — rotorcraft, Simulation in design and Setting the Scene: S. B. Anderson
development — fixed wing and systems, Simulation in This presentation was intended to set the scene for the
training, In-flight simulation: its uses and benefits, and symposium by giving an historical perspective based
Future applications. on the speaker's extensive flight simulation experience
at NASA Ames Research Center over the years. By
2. PRELIMINARY SPEAKERS viewing the challenges and solutions from the past
perhaps we will get a clearer picture of where we are
Keynote Speaker: MGen G. K. Muellner currently headed. Flight simulation at Ames began in
The keynote speech was provided by Major General G. the late 1940's and has progressed over the years to the
K. Muellner of the United States Air Force, who has current version of the vertical motion simulator (VMS)
had an illustrious career in military flight and was well and the in-flight simulation using a UH-60A Black
qualified to lead off the symposium. His challenge to Hawk helicopter. A very complete chronicle was
the audience was based on his experience as Program presented along with the lessons learned at each stage.
Director of the Joint Advanced Strike Technology For example, they found that it was possible to obtain
Program (JAST). This program is focussed on useful answers to problems from even crude
developing the next generation strike aircraft weapon simulators. The speaker passed on a number of
system for the US Navy, Marine Corps, Air Force and suggestions gained from experience:
T-2

(1) the task often dictates the level of motion fidelity interesting, the session lacked cohesion.
required.
(2) vertical motion should not be compromised. Session V(a) — " Simulation in Training." This session
(3) visuals with a wide field-of-view are often needed. began with a detailed description of the German
(4) in-flight tests will often be needed to accurately experience with the Tornado low level training
define flight criteria. simulator. This was followed by two very useful
(5) understanding the human element of the system is papers on the structuring and analysis of training
an important challenge. programs. The final paper dealt with pilot response to
wind shear on the landing approach as measured in
3. SYMPOSIUM ORGANIZATION flight simulator studies. This highlighted an
unexplained tendency for pilots to respond late to the
The symposium was divided into seven sessions onset of wind shear effects.
containing 34 technical papers. It was concluded with
a round-table forum which addressed the questions Session V(b) — "In-flight Simulation - Its Uses and
"What challenges have been identified?" and "What Benefits." This was a very cohesive session with all the
issues for the future have emerged?" major players in the in-flight simulator business
represented. The papers tended to be up-to-date
The general layout of the symposium was as follows: descriptions of the various facilities. Some examples
of the application of in-flight simulation were included
Session I — "Visual and Motion Cueing and to demonstrate the capability of such devices. An
Requirements." This topic is central to the flight interesting addition to this session was an informal
simulation field. Because it is a very broad subject area report, during the general discussion period, on Russian
the papers in this session could only hope to touch in-flight simulation over the years.
upon a few current points of interest. This was done
with the emphasis on visual scene content requirements Session VI - "Future Applications." The papers in this
and display resolution. Some of the latest thinking on session tended to emphasize networking of simulators
motion systems rounded out the session. and a simulation environment that goes beyond the
flight simulator itself. One paper on mobile training
Session II—"Advances in Modelling" Three of these simulators pointed the way towards smaller, low-cost
papers related to helicopter modelling with an emphasis trainers in the future.
on helicopter/ship interface simulation. One paper
dealt with the nagging problem of modelling aircraft The total participation at the symposium was indicative
ground handling. The final paper touched on of a successful meeting with 152 in attendance, of
modelling the tactical environment. whom 46 were authors, 40 panel members and 66
observers. The complete range of simulation
Session III— "Simulation in Design and Development community members was represented and formal
- Rotorcraft." The first paper in this session dealt with papers from six nations were presented.
the fairly new concept of employing simulations of
candidate helicopters in the aircraft selection process. 4. TECHNICAL EVALUATION
Two papers dealt with the application of simulators in
the development of tilt-rotor aircraft. To round out the Session I — Visual and Motion Cueing and
session, the use of simulation in the design of more Requirements
conventional helicopters was described. The first paper of this session was a tutorial on visual
information. This was followed logically by a practical
Session IV—"Simulation in Design and Development - paper on how to develop a visual data base. The next
Fixed Wing and Systems." This session contained a paper fitted in nicely by describing the on-site
collection of papers spanning a broad range of evaluation of several simulator visual systems. The
simulator applications. These included aircraft next two papers dealt with the application and design of
development, HUD symbology development, a multi- simulator motion systems. The final two papers were
target simulator, flight clearance of vehicle descriptions of recently developed image generation
management systems and a description of the NLR systems.
simulation facility. Although all the papers were
 T-3

The tutorial presented in Paper 2 began by cautioning necessary in helicopter simulators. The experiment
the visual data base designer to be careful when took place in the NASA Ames VMS configured as an
deleting objects because we cannot presently predict AH-64 helicopter. The tasks involved both small and
which ones are necessary for a specific task. The paper large yaw amplitudes under four different motion
then went on to list the common features of static and constraints. Minimal motion washout was employed.
dynamic scenes. The authors then dealt with simulator The conclusion of the experiment was that the yaw
hardware considerations and pointed out that it was not rotational motion degree-of-freedom may not be
possible at the present time to optimize the visual necessary in the simulation of hovering flight. This
display system's performance in all areas fact can then be used to free up extra travel in other
simultaneously. A special discussion was dedicated to degrees-of-freedom in such motion systems as the
the simulation of rotary wing vehicles. The paper popular hexapod configuration.
ended with the challenge to make improvements
through a list of research and development topics. The analysis of the motion system hydraulic actuator
Although the authors did not deal with the potential loop described in Paper 6 was intended to show how
benefits of helmet-mounted displays, they presented an modern control designs can be combined with modern
excellent overview. actuators to minimize time delay and increase
bandwidth. Although no reasoned case was made
Paper 3 described the development and application of proving that this step is essential to cure an identified
a technique for identifying what features should be flaw in current simulators, it does make sense to do the
presented in simulated visual scenes for helicopter best you can when building a state-of-the-art research
battlefield operations. The first step in the process is to simulator. We can look forward to its demonstrated
identify and prioritize flying activities in the battlefield. effectiveness when the new simulator is completed.
Next, the relative importance of visual scene
components must be determined. Finally a relational Paper 7 described a new image generator product
database is generated that can be used by scene employing advanced Z-buffer technology (AZtec) for
designers when selecting objects to include. A hidden surface elimination and anti-aliasing. This
practical process for carrying out these steps was process speeds up image processing by eliminating the
developed and described in the paper. The authors need to process transparent faces at the end of each
carried out field trials to demonstrate the process and computation cycle. A general description of the
then implemented simulator trials. The process was APOGEE image generator was given in rather broad
deemed to be successful. The authors concluded their terms.
paper with seven challenges to advance their work to
the next level of understanding. The paper was quite The Real-Time Simulation System (RTSS) as
complete and covered a broad range of work. developed within the European ESPRIT project
HAMLET was outlined in Paper 8. From the
The following presentation (Paper 4) was unique in description it would appear that this image generator is
that it reported on the evaluation of simulator visual similar in performance to current low end hardware.
displays by the user community, that is, operational Its strong feature is expandability and thus it may find
fighter pilots. Three different visual display applications where this is essential.
technologies were covered: projector/dome, rear
projection faceted, and fiber optic helmet-mounted Session II — Advances in Modelling
display. They employed 26 tactical tasks in a In this session the first two presentations dealt with the
subjective evaluation of training effectiveness. They simulation of helicopters landing on small ships. The
found that all the visual systems could be used to train next paper dealt with turbulent inputs to simulator rotor
some but not all of the evaluation tasks. They models and thus could be tied in with the first two
concluded that current systems do not provide adequate papers. This was followed by a description of the
contrast, resolution and brightness to allow for air development of a ground handling model. The final
model discrimination. paper in this session dealt with generating friendly and
enemy aircraft in the simulation of a tactical
The motion cueing experiment described in Paper 5 environment.
was well conceived and executed. Its purpose was to
determine whether the yawing motion channel was Paper 9 tackled one aspect of the simulation of
T-4

helicopters landing on small ships. The helipad being rotor. Filtered white noise (using the Dryden spectrum)
at the stern of these ships is subject to the ship airwake is employed and it is spread along a line between the
created by the upwind hull and superstructure. This points (the onset line) using a process called Gaussian
effect must be created in the simulator because it interpolation. The frozen flow assumption is then
impacts directly on the pilot's ability to operate his applied, allowing the turbulent field to sweep over the
helicopter while over the deck. A modular airwake rotor. When flown in the VMS in a UH-60A
model has been developed based largely on published simulation, pilots responded favourably.
data and models describing wind effects around
buildings. The model is modular in form and includes: Paper 12 outlined the generation of a ground handling
corner flow, lee rotor flow, profile drag flow, and model for a simulation of the C-160 transport aircraft.
inflow/outflow. A suitable turbulence model has not The process included physical modelling and actual
yet been developed. The resulting model has been taxi trials. A parameter identification process was used
tested against wind tunnel data and tested in the as part of the development (but not described in detail).
Advanced Flight Simulator at DRA Bedford. In the Of interest was the use of an impulse input to simulate
latter case pilot acceptance was favourable. The the complex spin-up of the wheels on landing. No
authors suggested that further improvements will be mention was made of how the model handled holding
made shortly. This model shows great promise. with brakes against engine thrust. Overall, the process
described appears to reflect current practice in this area.
The next presentation (Paper 10) covered the broad
spectrum of topics related to simulating the operation The last paper in this session (Paper 13) differed from
of helicopters from small ships. The previous paper's the rest in that it described the computer generation of
airwake work was included as one item of interest. The friendly aircraft and enemy aircraft in a battle
author expanded on the needs involving visual cues, simulation scenario. The author discussed the
motion cues, the aircraft mathematical model, the ship requirements, the algorithm and the technology needed
airwake and turbulence representation, and ship motion to implement this approach to generating a tactical
modelling. Tying in with the Session I material, it was environment. He concluded by noting that it was very
pointed out that in the visual display area there is a difficult to completely remove human intervention
need for good scene texturing and high resolution, from this process.
while good motion cues (including realistic heave cues)
can lead to more realistic workloads for the pilot. Session III — Simulation in Design and
Some example results from simulation projects Development - Rotorcraft
involving handling quality ratings and performance The first paper of this session described the use of
were presented. This work has indicated that the ADS- flight simulation as part of the evaluation process in the
33 Level 1/Level 2 boundaries need to be modified for procurement of an attack helicopter. The next two
helicopter/ship operations. The paper ends with a list presentations dealt with the contribution of flight
of challenges for future developments. Overall, the simulation to the development of tilt-rotor aircraft (the
paper indicated a very active and effective program in third paper also included the RAH-66 helicopter). The
this area. final paper in this session involved the use of flight
simulation in the development of an attack helicopter.
The problem of generating turbulence inputs for
helicopter rotor models was addressed in Paper 11. The use of piloted flight simulators to evaluate
Here the problem was to develop a technique that could proposed aircraft during the procurement decision-
be implemented in real-time and still produce results making process is a fairly new concept. Paper 14
acceptable to the simulator pilot. The proposed described the UK experience in applying this technique
technique is called Simulation of Rotor Blade Element to three candidate attack helicopters. The required
Turbulence (SORBET) and it was tested on the blade helicopter data were supplied by the bidders and the
element rotor model found in the helicopter code simulation was carried out in the Advanced Flight
GenHel. Tests demonstrated that current articulated Simulator configured with a Lynx cockpit. Handling
and semi-rigid rotors are insensitive to in-plane qualities and agility were assessed by the evaluation
turbulence and hence only the vertical component need pilots following ADS-33. The pilots were pleased with
be generated. SORBET generates two independent the process and the results. The paper concluded with a
turbulence fields located at two points upwind of the summary of important simulator features needed for
 T-5

this kind of work. The process described in this paper reference to simulation. Paper 23 was a facility
appears to be effective and likely to become the description of the Dutch National Simulation Facility,
standard means for evaluating aircraft during and the last paper was a description of an air-to-air
procurement. As noted by the audience, once an weapon system simulator.
aircraft has been accepted through this process it will
be up to the manufacturer to prove that the aircraft The authors of Paper 19 described their experiences
(when finally built) meets the performance of the using fixed-base flight simulators during the design and
simulator. development of the AMX (air-to-ground mission) and
the EF2000 (air-to-air mission). This type of paper can
Paper 15 - withdrawn. be very useful to an audience of simulator people
because we can all learn from others' experiences. The
Paper 16 described the use of the NASA Ames VMS authors described some of the problems encountered
facility in a study of design requirements related to a and how they solved them (just as the author of Paper 1
commercial tilt-rotor transport aircraft. Five piloted did). For example, attempts to simulate refueling failed
simulator experiments were carried out aimed at due to the absence of motion cues and a stereoscopic
cockpit design questions and terminal area operations display.
(both normal and emergency). Sample results were
provided of handling qualities results. As part of the Paper 20 outlined the use of flight simulation during
project the authors also considered airport facilities the Rafale development. An important aspect of this
designs directed at tilt-rotor operations. prograrffwas the controlling of costs. The paper
presented the background of the simulation and test
Both the V-22 tilt-rotor and the RAH-66 helicopter program carried out to obtain defence approval of the
were involved in the work described in Paper 17. The aircraft configuration. The program represented a
authors indicated that both constructive simulation (off- successful cooperation between government and
line) and virtual simulation (piloted, ground-based) industry.
played a significant role in the development phases. Of
note was the fact that (following the new trend The problems associated with the generation of HUD
indicated in several symposium papers) in the V-22 symbology during aggressive maneuvers was the
program a triple lab tie in link was established among: genesis of Paper 21. In particular the pitch and roll
(1) the flight control laboratory, (2) the flight simulator attitude symbology used on conventional HUD's is
laboratory, and (3) the mission computer/avionics difficult to interpret under these conditions. The
software laboratory. It was found that this resulted in authors described an Arc Segment Attitude Reference
fewer flight tests and lower overall costs. The authors (ASAR) symbology that eliminates this problem. The
see a strong future for distributed interactive simulation concept was first demonstrated during simulator trials
and virtual prototyping. in 1987. The paper traced the development of the
ASAR to the final version established in 1994. The
Paper 18 documented the use of flight simulation by paper demonstrated a very successful interplay between
EUROCOPTER in developing their helicopters over simulator tests and flight tests in developing new
the past ten years. Their SPHERE simulator is being display technology.
used in the development of the NH-90 helicopter.
They also employ simulators to study handling The author of Paper 22 presented an excellent tutorial
qualities and the man/machine interface. on vehicle management system concepts. This
included the concept of flight clearance. He also
Session IV — Simulation in Design and pointed out that in the future, simulation will play an
Development - Fixed Wing and Systems important role in this process.
To some extent this session became a catch-all for a
number of papers on diverse topics. The first two Paper 23 described the new Dutch National Simulation
papers described the authors' experiences in employing Facility (NSF). It is currently involved in the F-16
simulation during the development of fighter aircraft. mid-life update program. Its capabilities will be
The next paper concerned the design of HUD extended to helicopter simulation later in 1995 to
symbology for aggressive 3-D maneuvers. Paper 22 support work on the CH-47D and AH-64D.
discussed vehicle management systems with some
T-6

Paper 24 outlined the architecture of a piloted multi- combat mission training. He pointed out that most of
target simulator for air-to-air weapon system this training is conducted at the operational unit. This,
development. It is capable of generating many possible however, is becoming more difficult and expensive and
targets while creating a realistic operational tactical they are looking to DIS as a partial solution. Training
environment. There can be many piloted simulator effectiveness can be assessed by: evaluating simulator
stations on the net. It was found that pilots are more fidelity, gathering subjective opinions, measuring
motivated when engaging a colleague rather than a within simulator performance improvement, and
computer-generated opponent. The system is DIS performing transfer of training studies. He describes
compatible. the results from tests carried out at the unit level and
results from applying the Multi-Service Distributed
Session V(a) — Simulation in Training Training Testbed. It would appear that simulator
The first paper in this session was an excellent networking has become an effective training tool.
summary of the efforts made to create a simulator for
low level fighter mission training. The next two Paper 28 described the use of two simulators to study
presentations described the analysis and evaluation of pilot response to windshear during the landing
simulator training programs. The final paper dealt with approach. It was demonstrated that the display of
pilot response to windshear. pertinent data to the pilot could be used to eliminate a
potentially fatal tendency on the part of pilots to delay
Paper 25 — withdrawn. (by about 30 seconds) the response necessary to
successfully fly out of danger.
The German Air Force, underpressure from
environmental considerations, initiated a program to Session V(b) — In-FIight Simulation - Its Uses and
replace live low level training by flight simulator Benefits
training. This presentation (Paper 26) was a clear The four papers in this session represent the state-of-
description of how they attempted to achieve this goal the-art of in-flight simulation. They were mainly
with an upgraded Tornado simulator. To the original facility descriptions although the third paper included a
simulator they added a six degrees-of-freedom motion long sequence of brief descriptions of research carried
system, a G seat, a G suit and a fiber optic helmet- out on their simulators. The session was rounded out
mounted display system. During the evaluation phase by a description during the discussion period of in-
both objective and subjective evaluations were carried flight simulation in Russia.
out. Unfortunately the initial results were negative in
that training effectiveness could not be proven despite Paper 29 described a new Canadian facility called
the fact that the pilots were extremely impressed by the ASRA based on a Bell 412 HP helicopter. It will be
simulator. Before the system problems could be converted to a fly-by-wire in-flight simulator in the
rectified the program was cancelled. This task is near future. The presentation included a list of steps
obviously very difficult to simulate but perhaps the required for this conversion. For safety and
problems could have been overcome if they had been performance, it has been decided to employ a two
allowed more time. actuator compound actuator in each of the pitch and
roll channels. The author has verified the performance
Paper 27 was a tutorial on how to structure a of such a design through computer modelling of the
simulator-based training program. This is perhaps an system. By building on their past experience with their
area of expertise that is often not properly employed Bell 205 facility it appears that the new system with its
when we develop training devices. Some common increased agility will be a welcome addition to the in-
training myths were clarified and many useful ideas flight simulator community.
were presented. The authors closed by reminding us
that the objective of the training process is to produce a The US-developed RASCAL is based on a modified
change in people. We must also ensure that the UH-60 helicopter. Paper 30 described the work in
training objectives are clearly understood and achieved. progress aimed at completing this facility. The
presentation emphasized the helmet-mounted display
Perhaps one of the most important tasks in the training and research flight control system aspects of this
simulator environment is the assessment of training development. Its future application will be directed
effectiveness. The author of Paper 37 concentrated on towards the development of integrated control, sensor
 T-7

and display technology in order to increase the before DIS fully matures.
capability and safety for all-weather, low-level
rotorcraft operations. As with the ASRA of Paper 29, The presentation of Paper 34 described the
the schedule has been arranged to allow the conduct of development of a mobile AH-1W aircrew procedures
research programs during the development period. The trainer. This appears to be a direct response to the
presentation included lists of test programs to date, and generally stated goal of reduced costs. The system
subsystems to be included. This facility will features: reduced physical size, COTS computers, and
incorporate several unique features that will make it an full-size system performance. The paper lists nine
extremely capable research tool. design requirements for the ideal Small Mobile Trainer.
Technical details are given concerning the computer
The German ATTAS and ATTHeS in-flight simulators and the visual system. An aircraft-based system was
have had a long history of research activity. Paper 31 then proposed as a further means of reducing costs. In
reviewed the main features of these facilities and this scenario a ground-based operational aircraft
briefly described several past research projects. The becomes the heart of the system, an idea that has
ATTAS (based on a VFW 614 aircraft) has been used proven useful in the past for research facilities.
for: flight controls development, reconfiguration
following system damage, A3XX in-flight simulation, Paper 35 was a survey on the topic of advanced
and small airliners controls refinement and flight distributed simulation (ADS). As a review, the
envelope protection. ATTHeS (based on a BO 105 presentation reiterated many of the topics brought out
helicopter) has been used for: bandwidth criteria by previous speakers. For example, the trend towards
evaluation in support of ADS-33, rotor decoupling the increasing use of system integration labs was
study, Lynx in-flight simulation, and vision-based pointed out. You can mix and match real and virtual
hovering under adverse conditions. The facilities have players in an ADS environment to achieve cost savings.
made significant contributions in several areas. The author pointed out the differences in simulator
requirements between test and training functions. The
Paper 32 described the US VISTA in-flight simulator ADS challenges were listed along with the key related
based on AF-16D aircraft. They have just completed activities and developments. This was a good summary
Phase II of their development, test and evaluation, and paper and was well-placed in the program.
the authors reported on the test results. It was pointed
out that in this time of shrinking budgets, the VISTA The final Paper 36 described a concept termed the
project depends on minimizing costs in order to Synthetic Environment. The author defined this to be
survive. They plan to branch out into non-traditional "a computing based environment which allows the risk
missions such as acting as technology demonstrators reduction of defence issues to take place through the
and test beds. Efforts will also be made to find new interaction of people, models and live equipment". The
customers (e.g., foreign civil organizations). It would levels in this structure range from "customer defined
appear that they have an excellent product and a geopolitical scenarios" down through "simulations and
progressive attitude which should bring success in the live equipment" until reaching "systems support".
coming years. Thus flight simulation is only one cog in the wheel.
The flight simulation element would be used within a
Session VI — Future Applications DIS protocol where it was noted that as the number of
The first three papers of this final session dealt with DIS participants increase, technical limitations (e.g.,
distributed simulation topics with the second paper also transmission bandwidth) cause the fidelity to decrease.
describing a mobile flight simulator. The final paper The users of the synthetic environment would carry
discussed the concept of the synthetic environment, of out: forces training and mission rehearsal, mission
which flight simulation is a part. planning, and defence policy analysis. The message
appears to be that we in flight simulation should be
The author of Paper 33 presented a discussion of how prepared to participate in much larger organizational
to implement distributed interactive simulation. He structures than we have in the past.
talked about problems of timing and how they could be
solved. The point was made that how problems are Round-Table Forum
handled is scenario-dependent. It would appear that A lively exchange of ideas took place during the round-
there are still a number of difficulties to be overcome table forum held at the end of the symposium. Several
T-8

observations and topics brought out in discussions at 6. Additional Areas Requiring Further Development
the end of each session have been included below. — Here a number of areas were pinpointed. Ground
handling is still in need of better models. As well,
1. The Trade-Off Between Motion and Visual Cues — more work should be done on high angle-of-attack
This is always a hot topic for discussion at any aerodynamic models, external loads, elasticity, air-to-
simulator conference and the present symposium was air refueling and unusual flight attitudes.
no exception. The discussion began with the statement
that motion systems were too expensive and all that is 7. Russian In-Flight Simulators — In an informal
really needed is a good visual display. This was presentation following Session V(b) the Russians
countered by several rebuttals. In the end the reported on the development and application of in-
consensus appeared to be that there are scenarios where flight simulation in their country. Their simulators
motion is required and those where it is not. have spanned the same range of types and tasks as
those in the West starting with VTOL in the 1950's and
2. Visual System Development — In this instance, the continuing up to the current SU-27 in-flight simulator
need for improved visual displays was expressed. The of today. They are open for business with the West and
desire for resolution approaching eye limits, wide field- would encourage joint research programs.
of-view, and full field-of-regard was expressed. The
air-to-air trainer group felt that improved resolution 5. CONCLUSIONS
was essential while scene content could be reduced.
On the other hand, those involved in helicopter • The current state of flight simulation and its future
simulation felt that for the hovering task, increased course were well documented by the 35 technical
scene content (microstructure) was needed. It was papers of this symposium.
agreed that improving visual simulator cues would be a
long term task. Ending on a positive note, the opinion • With an attendance of 152 it was clearly a
was expressed that in the next five to seven years successful meeting and it covered a relevant topic.
improved visuals might allow the simulation of 95% of
all flying tasks. • There was a heavy emphasis on military problems
and topics. This was present to a greater extent than
3. Simulation Use During Aircraft Development — has been the norm in the past.
On this topic it was claimed that a good development
simulator can save the cost of one development aircraft. • Networking of simulators is obviously increasing
In this era of reduced budgets this certainly appears and remains an active topic for future developments.
attractive. For this to be fully exploited it was pointed
out that there is a need to speed up the updating of A number of areas of flight simulation technology
simulator properties during aircraft development are still seen as needing further research and
projects. development: visual displays, ground handling models,
verification and validation.
4. Distributed Interactive Simulation — Here the
consensus was that DIS should be nourished by • There is an increasing emphasis on cost reduction.
government support (as in the UK and US) and The ability of flight simulation to achieve this is one of
encouraged to grow. its strengths.

5. Fidelity — It was pointed out that the required Several papers and the keynote address indicated
level of fidelity depends on the task under study. In that flight simulation must often fit into an environment
addition more time should be devoted to improving the of models and software aimed at achieving a number of
operational fidelity of our training facilities. In a broad goals.
related area, it was felt that more effort should be
placed on verification and validation of simulators, 6. RECOMMENDATIONS
including operations near the edges of the flight
envelope. Researchers in the visual display area should
 T-9

continue to document the visual needs for the complete those involved.
range of tasks demanded from flight simulators. The
goal should be the ability to select the necessary set of • The current five-year cycle for flight simulation
features before undertaking a flight simulator task. symposia seems to be about right. Another one should
be planned for the year 2000.
• The time may be right for an AGARD Working
Group to assess the state-of-the-art in the field of • The integration of flight simulation into the
simulator networking. It appears that significant architecture of broader development environments
progress is being made in this area in several countries. should be encouraged and be a significant element in
A summarizing document could be quite useful for all the next AGARD Flight Simulation Symposium.
 K-l

Modeling and Flight Simulation—Tools to Produce Affordable Weapon Systems

MGen George K. Muellner, USAF
JAST Program Director
Crystal Square 4
1745 Jefferson Davis Highway
Arlington, Virginia, USA 22202-3251

My challenge to you this morning is to look beyond optimizing

system design for lethality, survivability, and supportability and
to become a major factor in assuring that our weapon systems are
also affordable.
Specifically, your challenge is to structure our modeling and flight
simulation tools to achieve a balanced design that produces all
four of the necessary characteristics of:
Joint Advanced Strike ■ Lethality
Technology Program ■ Survivability
■ Supportability
VISION
■ Affordability
"A Joint Services Team creating the building
blocks for affordable, successful development Several years ago, one of the leaders of the US aerospace indus-
of next generation strike weapon systems." try, Norm Augustine, coauthored a book called The Defense Revo-
lution.
Good morning ladies and gentlemen ! I am very pleased to In this work he described some of the changes that had to occur if
have this opportunity to kick off this important forum and the military and the defense industries were to accommodate to
help focus your thoughts for the next several days on some- the changing world environment. In this book he identified one
thing other than the beautiful country and the great food. of the key challenges as stopping what he called runaway
techflation.
My primary objective will be to challenge you with what I believe
are the necessary areas of change and growth that must occur in
the modeling and flight simulation disciplines, if we are to re- AUGUSTINE'S LAW - TECHFLATION
main a major element in achieving system effectiveness.
Let me first describe to you what the Joint Advanced Strike Tech-
nology Program is all about and why we have such a strong inter-
est in your success.
We are focused on developing the next generation of strike war-
fare weapon systems for the US Navy, Marine Corps, Air Force
and also, now, the Royal Navy. The program was generated when
a number of significant Service modernization programs were can-
celed due to being unaffordable. The Services and our Depart-
ment of Defense generated the JAST Program to look for a more
affordable way of meeting the Services' strike warfare needs. It
was this search for improved affordability that increased our in- 1900 1950 2000 2050 2100 2150

terest in modeling and simulation playing a broader role in weapon YEAR OF INITIAL OPERATION OF AIRCRAFT
system development and operation.

For years, modeling and flight simulation activities have become Techflation means that, while every new generation of weapon sys-
increasingly important, if not essential, tools in designing, test-
tem is far more capable then its predecessor, it also costs consid-
ing, and operating military systems.
erably more. His prediction is that, by the year 2054, it will take
These tools were first focused on the tasks of optimizing the le- the entire U.S. Defense budget to buy one tactical aircraft.
thality or survivability of the weapon system. Later, modeling and
Clearly, this is an unacceptable trend and must be corrected if our
simulation became critical to achieving better supportability in our
nations are to afford the systems necessary to support our national
weapon systems.
security policies.

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?",
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
K-2

We, in the Joint Advanced Strike Technology Program, believe M & S SUPPORT FOR THE ENTIRE LIFE CYCLE
that the modeling and flight simulation community can be the
dominant factor in controlling techflation and in helping our Ser-
Warfighters
vices achieve affordable weapon systems. ■ Developers

For the next few minutes, I will highlight what we believe to be

the vision for achieving weapon system affordability through the
effective application of modeling and simulation. r^v^W^Ä^/ Synthetic Environments
liiiMi
Synthetic _ virtual Design», virtu« Prototyping, Syntnetlc_
MODELING, SIMULATION, & ANALYSIS Ba«iefleWs"~*^ Factories, Manufacturing Processes.... ...

VISION
Concept Develop- Testa
Understand-
ing the need - Exploration - ment _ Evaluation - Production Operation

Exploit modeling and simulation 1

to create an affordable weapon t EMPLOYMENT
I
system by:
To accomplish these goals, we in the JAST Program have found
• Supporting early common (government and that you must begin with an objective of fielding modeling and
industry), joint cost-performance trades simulation tools to provide life-cycle support for the weapon sys-
. Enabling a lower cost design, development, and tem.
evaluation process :
• Providing warfighters early system experience You must involve the warfighters, the developers, the testers, the
• Enabling more effective system training and cost analysts, and many other communities in the initial design of
employment throughout the program life-cycle the modeling and simulation architecture and standards. And once
you get buy-in on those standards—you must enforce them!

This chart highlights the primary areas where we believe model-

ing and simulation can aid in achieving affordable weapon sys- JAST PROGRAM MS&A
tems. MODELING HIERARCHY
First you must help the warfighter generate an affordable require- FACES OF THE JAST PYRAMID
ment, that is, a requirement based upon the warfighter understand-
ing their needs—not just their wants—and also an understanding
of the cost of the various alternative solutions for satisfying their
needs. Our weapon systems must represent a better balance of
performance and cost, as do most commercial products. You can
help the warfighter with achieving that balance early in the pro-
cess.
Next, we must expand the role of simulation in the design and
development process. We must try to accomplish as much as pos-
sible in the virtual world before we build hardware or software or
lay out an assembly process.
Next, the developmental and operational test and evaluation pro-
cess must use flight and ground tests to verify our systems models,
and only conduct expensive flight or ground tests as necessary to In the JAST Program, we began with a modeling hierarchy that
validate the system models. was designed to address effectiveness, supportability, and
Finally, we must design our models, databases, and simulation affordability. For each of these areas we chose core models and
benchmarked their characteristics and the associated databases
tools to transition to providing crew-in-the-loop and hardware-in-
across all government and industry participants. From the begin-
the-loop applications to support system evaluation, training, and
ning, we forced trades between effectiveness and supportability
force employment.
characteristics to achieve affordability. These trades required the
From the beginning, these tools must be designed to support dis- modeling environment to develop embedded cost estimating tech-
tributed activities. Geographically disbursed operations with de- niques.
ployed forces and with coalition partners are becoming the normal
mode of operation. Clearly, our modeling and flight simulation The next chart illustrates our approach to the effectiveness mod-
tools must adapt to this new environment. This is equally true eling. The JAST Program employs six core models to support en-
when most major development programs involve contractor teams gagement, mission, and campaign level trades and evaluation.
that are often international in scope. At the engineering level, besides the normal system and subsystem
models, we are creating a virtual avionics prototype that will al-
 K-3

EFFECTIVENESS MODELING HIERARCHY Our objective in employing this constructive and interactive digi-
MOOS, MOEs. MOP«
tal simulation is to create a balanced, affordable design that meets
used throughoutthoi
program

Measures of
■ Force Effectiveness Operational Outcome AFFORDABILITY MODELING HIERARCHY
• Force Survlvabllity (MOO)
* Campaign Outcomes MOO.. v::;...unp
used throughout the
• Mission Effectiveness/ Measures of program
Lethality/Survivability
Effectiveness (MOE)

♦
• Mission Outcomes
Measures of
• Force Coat-Effectiveness Operational Outcome
* System effectiveness • Life Cycle Cost (MOO)
* Vehicle capabilities Measures of
• Integration of sub-systems
Performance
• System performance (MOP) ' Mission Cost-Effectiveness Measures of
• Sub-system definition Effectiveness (MOE)
• 2001+ Service force elements • Cost Goals
■ 2010 rhreBt laydowm / target base ; • Development Cost
• Production Cost
> Operations & Support Measures of
Cost Performance
low us to design and develop the avionics system "virtually" to (MOP)
allow deferral of hardware assembly as long as possible. This al-
lows better exploitation of the latest technology available and pro-
vides a more open competitive environment. It will also provide
the tools and models to transition into our crew-in-the-loop flight the warfighters' needs. But as this chart depicts, the large payoff
simulation activities. is not just in reducing the cost of development, it is in reducing
the production and life cycle costs.
An important thread of continuity through our modeling environ-
ment is created by the constant set of measures of merit. These
measures were jointly established by the warfighters and the de-
velopers to reflect the necessary characteristics of the system and AFFORDABILITY OBJECTIVE
will be used throughout development and test and then transition
into training objectives.

SUPPORTABIÜTY MODELING HIERARCHY LETHALITY

SURVIVABILITY
SUPPORTABILITY

CAMPAIGN
• Availability
- SWS Robustness
- Logistics Level of Effort- CAPABILITY COST
Sortie Generation
• Thunder load planning
MISSION
• Maint/supply trans concepts
• Logistics footprint AFFORDABILITY
• Organization structure
• Intertheater/intratheater A determination that the life cycle cost of an acquisition
lift level of effort program is in consonance with the long range investment
SYSTEIWENGINEERING and force structure plans
• MTTR
• MTBF
• MTBM
As we now take the weapon system concepts into the detailed de-
sign and flight demonstration phase, we are finding that the vir-
tual tools we have developed for the tradeoff and design tasks
In a similar fashion we identified and benchmarked models for
supportability and deployability. In this area we employ three
supportability and two deployability models. These models not MODELING & FLIGHT SIMULATION PROCESS
only address tradeoffs in reliability and maintainability with sys-
tem characteristics, but also evaluate the impact on supporting
INTERACTIVE
airlift and tanker resources. ENVIRONMENT
CONSTRUCTIVE SIMULATION
QFD SIMULATION
For each of these models, as was the case for the effectiveness ANALYSIS A E>
models, we have undertaken the task of joint validation, verifi-
cation, and accreditation so that senior decision-makers from the
Services and the Department of Defense will accept the results.
In the area of affordability modeling, we have taken the process
all the way down to the commodity or process level. We found
this necessary to ensure that we could effectively employ integrated
cost and design databases, and employ cost as an independent
variable from the beginning of the system design process.
K-4

provide the basis for our crew-station design and flight-control de- At this time, I would be happy to answer a few questions. As I do
velopmental activities and eventually will lead to training and mis- that, this chart identifies the most effective way of getting addi-
sion rehearsal capabilities. tional information on the JAST Program or for providing us your
ideas.
It is this type of life-cycle focus that must be employed as we de-
velop these tools. We cannot afford to build modeling and simu- Thank you for your attention.
lation capability that is limited to unique program phases.

Affordable flight test also results from the effective coupling of

modeling and simulation and ground or flight demonstrations.
Rather than having the modeling compliment the flight test, we
are using the flight or other hardware-in-the-loop testing test to
validate the system models.
This approach shows promise for significantly reducing the amount
of test resources and test time and, therefore, significantly reduc-
ing the cost of these program phases.
For further information,
From the very beginning, our modeling, simulation and analysis contact the JAST Program Office
process has been designed to function in a distributed fashion. This on the World-Wide Web (Internet)
was necessary to effectively exploit existing resources and also
allowed us to keep all the key agencies involved and participat- http://www.jast.mil
ing in the process.
e-mail JAST program questions to:
This network will transition to be the backbone for a distributed
training and exercise capability. This capability has provided us mailbox© ntrprs.jast.mil
an early opportunity for the warfighters to evaluate various op-
erational concepts and to develop tactics interactively with the
weapon system development.

Unfortunately, our time is limited and you have so many great pre-
sentations on the schedule that I must wrap up. Let me summa-
rize with these three key points:

SUMMARY

. AFFORDABILITY IS AN ESSENTIAL CHARACTERISTIC OF

MODERN WEAPON SYSTEMS
. MODELING AND FLIGHT SIMULATION CAN PROVIDE FOR
EFFECTIVE COST VS PERFORMANCE TRADES AND A
MORE EFFECTIVE AND EFFICIENT WEAPON SYSTEM
ACQUISITION
. SIMULATION ENVIRONMENT MUST BE DESIGNED FOR
THE PROGRAM LIFE-CYCLE AND SUPPORT DISTRIBUTED,
MULTI-LEVEL OPERATIONS

First, affordability is an absolutely essential characteristic of our

new weapon systems. Affordability will only occur when we pay
as much attention to it as we do the other essential performance
characteristics of our weapon systems.
Modeling and flight simulation can be an important tool in achiev-
ing a balanced, affordable design and in supporting that design
throughout its lifetime. However, this will only occur if we evolve
our modeling and simulation tools to treat cost and affordability
as a major program objective.
As you can see from this brief description, the JAST Program is
dependent on this community to deliver the tools to achieve le-
thal, survivable, supportable, and affordable weapon systems. We
actively solicit your help in working this most difficult challenge.
 1-1

Historical Review of Piloted Simulation at NASA Ames

Seth B. Anderson

NASA Ames Research Center

Mail Stop 262-1
Moffett Field, CA 94035-1000 U.S.A.

ABSTRACT DISCUSSION
This paper traces the conception and development of in-flight
and ground based simulators at NASA Ames Research Center, Description and Research Projects
starting in 1947 and continuing to the early 1990's. Problems
with their development and operation and how limitations were Link "Blue Box" procedures trainer The first piloted simulator
handled are recounted. Advances needed in simulator equip- (fig. 1) at Ames was one that introduced the era of flight
ment to improve performance and fidelity to gain pilot accep- simulation (1929) and served as an 1FR procedures trainer
tance are discussed. The uses of these simulators in various during WWII. It had 3 degrees of angular motion and only a
aircraft research and development programs and their basic 1FR (needle, ball and airspeed) visual system. Attempts to
importance to aircraft design and flight testing are reviewed. improve the drive system which used pneumatic valves, bellows,
Challenges remaining include a better understanding of the and actuators originally created to make organ music, proved
tradeoff between motion cues and visual cues, the importance of futile and no research results were obtained. It was relegated to
simulation sophistication when examining aircraft with marginal become a museum artifact, but remains operational to show how
handling qualities characteristics, and the continuing need for far simulation has come.
upgrading simulation technology as more complex problems are
encountered. Additional research is needed to understand the
human behavior aspect in the pilot/simulator system.

INTRODUCTION
Both ground based and in-flight research simulators have proven
to be valuable assets to aid design and development of a wide
variety of aircraft. At the NASA Ames Research Center, the
development and use of piloted flight simulators which started in
the late 1940s has continued with an improved range of
capabilities. Early on, the need for improved realism both in
terms of cockpit motion cues and visual displays was recognized
as more definitive answers were required for more complex
problems. As a result, simulators were designed and developed
to provide two, three, five, and finally six degrees of motion
freedom, and visual displays increased in sophistication to
provide out-the-window, wide field of view scenes approaching
real life quality. In-flight simulation which started in 1947 with
a WWII F6F aircraft has utilized a variety of aircraft including
fixed wing, V/STOL, and rotorcraft configurations. They Rudimentary fixed base The first application of piloted
functioned as new flight research tools to study old and new simulators for aeronautical research originated in the NACA
problems with varying degrees of success. time period (1955), as part of a study to develop criteria for
selecting carrier landing approach speeds (ref. 1). This
One may ask about the future for research simulators at the turn simulator (fig.2) was very rudimentary, consisting of a control
of the century. A partial answer may be found by examining the stick geared linearly to aircraft lift coefficient (no stick force
past to learn what was needed to continue advancement in gradient) through a first-order time constant, a throttle geared
simulator technology. It is convenient to review the history of linearly to thrust, a stall-warning audible buzzer, a visual pres-
piloted simulator development at NASA Ames where simulation entation of an airspeed indicator and a CRT showing altitude
has developed into a major national research facility. About a (horizontal line) above a ground reference with a shorter
dozen ground based and an equal number of in-flight simulators horizontal line to indicate vertical acceleration. These visual
have been developed at Ames, each having a limited life span. references were selected to give information on control of
The purpose of this paper is to review the history of piloted altitude, the primary reason given by pilots for selecting the
flight simulation to focus attention on the challenges remaining minimum airspeed in most carrier approach landings. An analog
for continued improvement. computer was used to model basic flight characteristics for each
of three carrier type aircraft.
The scope includes a brief description of several select ground
based and in-flight simulators used in research studies at NASA
Ames Research Center, examples of results obtained, and a
review of the problems (and solutions) encountered along the
way.

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?'
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
1-2

Initially, the pilots were skeptical that the crude simulator

environment would provide any degree of correlation with
flight. However, after suitable adjustments were made to such
items as stick gain (CT_ per §e), throttle response, and thrust
margin, reasonable correlation with flight was achieved for
approach speeds (within 3 knots) for several Navy carrier
aircraft. However, because of the use of a "generalized" math
model, lack of cockpit realism and lack of associated motion
response, the pilot could not "identify" flying a particular
aircraft, thereby restricting correlation to only those conditions
where altitude control was the primary factor for selecting
minimum approach speed. The challenge that remained was to
provide the sophistication needed to determine minimum These early studies showed that good correlation with flight test
approach speed when other factors such as adverse stability and results was restricted to areas where handling qualities were
control characteristics were dominant reasons for choosing satisfactory. Correlation with flight results was poor when
approach speed. exploring areas of low damping and poor stability or when
investigating control characteristics conducive to pilot induced
In part because of the lack of motion cues, a form of oscillations (PIO). Although the F-86 cockpit added realism, the
atmospheric disturbance was needed to provide flight path lack of motion cues and an adequate visual scene tended to
excursions similar to those encountered in actual approaches produce results which were too conservative.
since this factor directly impacts ability to control altitude and
therefore selection of approach speed. Although the pilots were NE-2 simulator The need to improve accuracy in establishing
less than enthusiastic about the credibility of their answers, this control system requirements by using motion cues resulted in
early study showed the potential for using piloted simulators to the first motion research simulator (fig.4) introduced in 1959.
study important aircraft operational problems. It was called the NE-2 ("any two") because its gimbal arrange-
ment provided mounting for roll and pitch, pitch and yaw, or
F-86AID cockpit simulator Recognizing the need to improve yaw and roll motion. Approximately + 45 deg angular motion
cockpit realism, a fixed base simulator (fig.3)was made was provided by an electric motor. An analog computer
operational in 1958 using of the forward half of an F-86A/D provided outputs for angular position, rate, and acceleration.
airplane with the cockpit interior and controls closely resembling The cockpit controls were similar to those of an actual airplane.
those of the actual airplane. The cockpit could be covered with Control requirements for a VTOL hover task were studied
an opaque canopy to isolate the test pilot from outside distrac- (ref. 4) with an open cockpit view of the outside world. In
tions or left open to provide a real-world view. The only cockpit studies of lateral control requirements for fighter-type aircraft,
instrument was a 5" oscilloscope, which was used to display the cockpit was covered, and a CRT and a limited number of
attitude and a target for a tracking task in a study of the effects panel instruments provided aircraft state information.
of longitudinal control-system dynamics (ref. 2), and bank angle
for a study of lateral control requirements for fighter-type
aircraft (ref. 3). The drive signals to the oscilloscope were
provided by an analog computer which solved the equations of
motion for the airplane.
 1-3

Electromechanical servo systems driven by d.c. electrical motors

provided rotary degrees of freedom, +90° in roll, ±40° in pitch,
Although use of this first motion simulator was short lived, +70° in yaw. Translational motion was available in two direc-
several tests of historical interest indicated the importance of tions, +1.75 ft. vertically and unlimited sideward movement by
providing motion feedback. For example, an engine failure for rotating the centrifuge arm. A single-place cockpit was
the YF-12 aircraft in cruise (Mach 3-) showed the need for equipped with a stick control, two throttles, and conventional
motion cues to develop an operationally acceptable control flight instruments. In the three angular degrees of freedom
system to avoid catastrophic directional divergence. In another mode, an external view of the real world was used. The cockpit
case, a test pilot from the U.K. was given an opportunity to was covered with an opaque canopy for conditions where side
sample the hover control characteristics of the X-14 VTOL acceleration cues were required (ref. 5). In 1969, a televised
aircraft, before starting flight tests of the British Harrier VTOL visual display was installed in the cab. General purpose analog
aircraft. The motion and visual cues provided in the practice computers were used to generate the commands for the motion
runs in the simulator were inadequate since when flying the real system and instruments (ref. 6).
aircraft lateral drift was not eliminated before touchdown,
severely damaging the X-14. The 5DOFS was used to study handling qualities requirements
for supersonic transports (refs. 5, 7, and 8). Two historical
The motion cues did provide more accurate pilot opinion ratings contributions are of interest. One was its use in the Gemini
of control power and damping requirements for hover operation spacecraft program to subject astronauts to vibrations with the
of VTOL aircraft using an open cockpit visual scene. The pilots same frequencies and amplitudes as those that had been
were able to use the roll angular acceleration cues to define measured on the Titan launch system. These frequencies were
satisfactory levels of lateral control power; however, large in the critical range (3 to 10 Hz), which might debilitate astro-
amplitude roll angles were disorienting because of unrealistic nauts during liftoff. The other historical contribution was a non-
side forces imposed on the pilot. Two axes motion feedback aerospace use of the simulator (1968) where a bullet fragment
provided better correlation with flight particularly for simulation in a patient's brain was repositioned by centrifugal forces to a
of VTOL hovering when poor stability and low damping aircraft "safe" location.
characteristics prevailed. For studying fighter aircraft maneu-
vering control requirements, the motion characteristics were The 5DOFS centrifuge design did not provide a significant
inadequate restricting studies to simple one axis tracking tasks advance in the state-of-the-art for aircraft research studies
where control cross coupling was essentially non-existent. primarily because of motion deficiencies. (The 5DOFS was
Multi-axis motion feedback including vertical and side force dismantled in 1975.) Providing only a small amount of vertical
cues remained as an important challenge for the future. travel was ineffective since strong (disorienting) washouts were
required to avoid hitting stops. In addition, since the cab was
Five degree of freedom simulator This simulator (5DOFS) not counter balanced, dynamic response was non-linear;
(fig. 5) introduced in 1960 was designed for use as a piloted downward acceleration was fast and vice versa; therefore this
centrifuge test facility. It had a three-gimbaled cab mounted motion was not used in research studies (ref. 7). Attempts to
on a vertical track which, in turn, was mounted on the end of an accurately reproduce and sustain side accelerations resulted in
30 ft. arm (ref. 5). A circular rail on the floor supported the large velocities of the centrifuge arm which were objectionable
vertical track assembly. Two interchangeable single place cabs to the pilot because of vibration (noise) and centrifugal accel -
were available: one, representing a conventional aircraft had the eration effects. The use of washout to minimize the adverse
pilot facing radially outward, whereas the other, representing the centrifugal effects met with only limited success. In addition,
pilot seated in a spacecraft, was mounted facing radially inward. dynamic response was poor due to cable stretch. Use of this
In the former case, changes in side acceleration at the cockpit facility for centrifuge research studies was not undertaken
could be simulated by angular acceleration of the centrifuge arm. because of large amplitude vibration levels associated with the
rail support system and safety concerns when tangential
velocities of 60 mph were reached at 5 "g" acceleration. A
1-4

challenge remained to provide non-disorienting translational conventional cockpit instrument display, normal flight controls
motion cues in a yet to be defined geometric area. (with control forces provided by springs and dampers), and a
general purpose analog computer programmed with math model
Basic simulator with external horizon Another fixed base equations of six degrees of motion freedom.
simulator, housed in a 20 ft diameter dome was used briefly in
1961 to investigate the value of extensive peripheral vision cues. An important addition was an out-the-window view of a runway
The cockpit was similar to that of a F-86E airplane, with con- generated by a Dalto Visual System developed for use with
ventional rudder pedals, center stick, and flight instruments. The training simulators. It consisted of a television camera servo-
simulator had a novel visual system employing a two-axis servo- driven in three angular degrees of freedom, and vertical and
driven (roll and pitch) motion picture projector, controlled by an lateral displacement relative to a runway model installed on a
analog computer, to project a moving artificial horizon onto a continuously moving belt. The resulting television scene,
20-foot-diameter hemispherical screen. To the pilot, the image projected onto a screen mounted about 10 feet forward of the
appeared to be a brightly illuminated layer of clouds several simulator cockpit, showed the approach lighting and runway as
thousand feet below the cockpit. Another projection method seen in a landing at dusk with hazy, one-half-mile visibility.
used an approach and touchdown scene obtained from a 16mm
camera located in the cockpit of a fighter aircraft. This simulator was used for several handling qualities studies
including a supersonic transport configuration in landing
The simulator was used in a study of the effect of lateral - approach (refs. 7, 10, andl 1) and for lateral-directional evalu-
directional control coupling for supersonic and hypersonic ations of a large STOL transport airplane (refs. 12 and 13)
aircraft (ref. 9). Although the wide field of view provided a Although the addition of the Dalto visual system to a fixed base
good bank angle reference, limitations of the visual scene simulator had a beneficial effect on pilot evaluations of handling
content and lack of motion cues discouraged extensive use of qualities, all pilots had difficulty making consistent landings,
this simulator. The strong visual cues supplied by the wide field due in part to the lack of motion cues which normally provide
of view horizon projector in the absence of motion cues were useful lead information. To help the pilot "calibrate" the
disorienting when large amplitude roll excursions were used. simulator for the landing task, an opportunity to "fly" a known
When the visual system was used for approach and landings aircraft proved to be beneficial. Better correlation with flight
with the preprogrammed movie scene, the results were only of was obtained after the simulator cockpit mechanical control
subjective interest since no matter how poorly the pilot flew characteristics were modified to provide control centering and
the aircraft on approach, a perfect landing always resulted. low friction so the pilot could obtain a "feel" of the aircraft in
Projecting a usable real world scene over the large area of the the absence of motion feedback. The deficiencies of this visual
dome required considerably more light intensity than was system (ref. 14) adversely affected precision of landing perfor-
available at that time. mance due in part to poor resolution inherent in black and white
television projection and other undesirable features including
Transport landing-approach sim-ulator In part because of the mechanical backlash, noisy drive signals, unsteady video
lack of success with the motion system of the 5DOFS, and the performance, and lags in the camera drives. Lateral/directional
immediate need for research results in landing approach for handing qualities evaluations were compromised particularly for
transports, another fixed based simulator was developed (1961). STOL aircraft which were maneuvered more extensively by the
This simulator (fig.6) had a fixed-base transport-type cockpit, a restricted peripheral visual cues inherent in the TV system.

Height control test apparatus (HiConTA) In the same time

Runway »imutator An a (09 conipu!«r
period (1961), a somewhat unique (single degree-of-freedom)
simulator (fig. 7) was built for the primary task of studying
height control requirements for VTOL aircraft. Attached to the
structural framework of the 40x80-ft Wind Tunnel, it provided a
relatively large vertical travel (100 ft) and a true outside world
visual scene. A two-place helicopter cab was suitably equipped
for several studies; helicopter cockpit controls for VTOL
aircraft, transport type controls for a jet transport control
problem study, and a conventional stick for a low level, high
speed terrain following program (ref. 15). In each case, cockpit
panel instruments provided the pilot with essential (but meager)
information for each of the test situations. In addition, for a
landing performance study (ref. 16), a TV system displayed a
night runway scene on a black and white 16 in. TV monitor.
Slmuiatof cab R«ord«

Fiptn 6. Tr«iipnrt Ijuuttns-ApjmMrhSimiiUt«

 1-5

with the cab because of the adverse effect of added weight on

dynamic response. The HiConTA was dismantled about 1984.

Moving base transport simulator Motivated by the need for

improved accuracy in developing handling qualities criteria for
STOL aircraft, another simulator (fig. 8) became operational in
1963. The motion system used three linear hydraulic servo
actuators, which operated differentially or in synchronization to
provide roll angles in the range of ±9° , pitch angles from -6° to
+14°, and a small amount of vertical travel (2 ft.). A transport
cab was used with conventional seating for two pilots, instru -
mentation, and controls which were hydraulically powered for
variation of control system parameters. Initially a scene of the
approach lighting and runway was projected onto a screen
outside the cab by the Dalto visual system. Later, the Redifon
color TV system was implemented. The drive signals for the
motion system, instruments, and visual system were generated
by an analog computer.

The vertical motion system was driven in response to the analog

computed cockpit vertical acceleration modified by high-pass
filtering (wash out) to constrain the cab to acceptable excursion
limits. A lead network was required to reduce lag in the electric
motor drive system. Accelerations of + 2g and maximum
velocities of + 20 ft/sec were possible. The frequency response
was essentially flat out to 1 HZ.
This facility was used for a variety of studies, including several
The single degree of freedom (heave) motion was sufficient for carrier landing aircraft, handling qualities requirements of a
the specialized VTOL height control task, and good correlation STOL seaplane (ref. 19), STOL aircraft (ref.20) and 21), and a
with flight was achieved (ref. 17). Landing performance special purpose counterinsurgency (COIN) aircraft. The very
measurements (touchdown velocities) correlated well with flight limited amount of vertical motion (2 ft.), was not usable in
(ref. 16); however, considerable pilot adaptation (landing prac- STOL landing approach studies because of adverse dynamics
tices) was required to achieve consistent results in part due to response characteristics associated with large wash out require-
lack of other motion cues and limited visual cues from the TV ments. When the roll motion was programmed to provide a
display. This was helped by flying a "known" aircraft. The 1-to-l ratio of input bank angle to cab motion, large bank angles
vertical dynamics (frequency and amplitude) inherent in the typical of STOL operations could not be used because the pilot
cable/drive system realistically simulated structural cockpit objected to an unrealistic side force. Also, without yaw capa-
motion for transport jet aircraft when used in studies to deter- bility, roll /yaw crosscoupling problems could not be studied.
mine reasons for loss of flight path control when flying in large The addition of the Redifon color system was a strong factor in
scale atmospheric turbulence, (ref. 18). improving pilot acceptance of handling qualities results and
compensated to some degree for the lack of yaw and heave
For landing performance studies, only 40 ft (+ 20 ft) of vertical motion. This simulator had a long history of successful use and
travel was needed to obtain desired fidelity using acceptable was finally dismantled circa 1984.
"wash out" frequencies, and ± 20 ft/sec vertical velocity
capability. Motion system noises (track rumble) detracted from All-axes motion generator Based primarily on the need to
simulation fidelity giving false velocity cues. The amount of provide more definitive control system design values for VTOL
available vertical travel was larger than necessary, with adverse aircraft, the first six-degree-of-freedom motion simulator at
implications on safety, dynamic response (cable stretch), and Ames (fig. 9) was put into operation in 1964. The single place
maintenance. Improved sophistication in terms of visual cab could move + 45° about the pitch, roll, and yaw axes and
displays or additional degrees of freedom were not possible
1-6

translate + 9 feet in three orthogonal directions. This simulator sophisticated systems such as velocity command control.
was one of the first to use equilibrators, instead of a counter- Although the outside view provided flight-like realism, the pilots
weight to help offset gravitational effects and thereby improve still felt the confining effect of the inner walls of the building.
vertical dynamic performance. A conventional center stick, Providing six degrees of motion freedom did not in itself
rudder pedals, and a fighter-type throttle quadrant were provided guarantee high utilization of this simulator. The single place cab
An analog computer system generated the drive signals for the and low quality cockpit visual display available at that time,
motion system and instruments. discouraged more extensive use. In addition, for large lateral
maneuvers the required wash out accelerations were larger than
the pilot's input values, and were disorienting. Safety require-
ments to limit translational velocities caused frequent shut down
of the drive system and inconvenient reset procedures for start
up. The suspended heavy utility cable introduced undesirable
(spurious) cab motion. The use of equilibrators successfully
improved vertical dynamic response and were used on
subsequent motion simulators.

Flight simulator for advanced aircraft (FSAA) Past experience

in simulation of large aircraft had pointed out a strong need for
multi crew cabs with more extensive translational motion
capabilities. Accordingly, the FSAA (fig. 10) became opera-
tional in 1969 (refs. 24, 25, and 26). This six degree of freedom
motion system, using DC velocity servo drives, provided an
extended lateral travel (+ 40 ft), vertical travel (+ 4 ft) and fore
and aft (+ 3-1/2 ft) motion. Two tractors moved the simulator
laterally, with supporting wheels taking the principal vertical
loads. The lateral carriage carried a vertical platform raised and
lowered by three continuous ball screws. To unburden the
Operational safety was a primary concern. A shock absorber
screw drive, equilibrators were used to float the vertical
system was developed to avoid excessive loads at the extremities
platform. On top of the vertical platform was the longitudinal
of travel. Several large orange colored balls were appropriately
motion system which used a single ball screw to drive the
located in space to warn the pilot of impending travel limits. In
cockpit and gimbal systems fore and aft. Relatively conventional
addition, the electric drive system would automatically shut off
chain-drive gimbals provided three angular motions. For
if the pilot commanded too large translational velocities when
operational safety, the position, velocity, and acceleration of
approaching limits of travel. On a few occasions, the cockpit
each motion system axis were constrained by an array of
restraint system (pilot safety belts) proved valuable when the
electrical limiters and ultimately by mechanical buffers. The
cab was inadvertently driven hard-over to the extremities of all
analog computer initially used (ref. 27) had so many elements
six degrees of freedom.
that the program could not be quickly and thoroughly checked.
The simulator validated its intended design features by providing
the real-life sensations of VTOL hover operation (ref. 22).
Hangar doors could be opened for a natural outdoor scene, and
VTOL hover tasks could be conducted without motion washout
and the results showed good correlation with flight. The
opportunity to evaluate advanced control systems was valuable
in preparing for a cooperative flight test program of the German
DO-31 VTOL jet transport aircraft. Translational control
requirements for lateral positioning in hover, (ref. 23) defined
the magnitude of side acceleration needed to translate without
banking, which correlated well with flight tests in the X-14B
VTOL aircraft. Of historical interest, this simulator was used by
the Apollo Lunar Landing pilots, starting with Neil Armstrong,
to simulate lunar landing dynamics (ref. 9).

Only a limited amount of testing was conducted with a hooded

cab using a 16" CRT Redifon TV system. Later, the simulator
was modified to permit repositioning the cab to a nominal 90°
pitch-up angle (pilot was lying on the back) for vertical attitude
take off and landing (VATOL) concepts.

The travel within an 18 ft cube VTOL limited control

evaluations to a hover maneuver task. More extensive trans -
lation motion than available was desired for evaluating more
 1-7

A compensation circuit (ref. 25) was required to reduce the seat fighter aircraft. Cockpit controls for various aircraft
"turn-around-bump", a disturbance which could occur every configurations with realistic force-feel cues are provided by a
time a motion drive reversed direction in the lateral and vertical programmable McFadden control loader system. A collimating
drives. mirror/beam-splitter system is used with an Evans and
Sutherland (E&S) CT5A CIG system. Four speakers mounted in
The FSAA was a major step up in simulator sophistication the cab are needed to reproduce a variety of sounds associated
offering for the first time a 3 place cab with full motion with different types of aircraft and helicopters.
including a uniquely large lateral travel range which provided
more realistic side force cues associated with engine failure for
transports where the pilot's position is far forward of the CG.
and for turn entry maneuvers peculiar to STOL aircraft. A wide
range of notable aircraft research and development programs
used the FSAA including Supersonic Transport (Concord), many
V/STOL aircraft, and the Space Shuttle Orbiter. (ref. 28).

Although the FSAA proved quite useful in several important

research programs, it had basic limitations which would
eventually limit its service life. The need to provide good
motion (travel) harmony was not fully appreciated in the design
of the FSAA. Less lateral travel would have been acceptable, but
considerably more vertical travel was needed. The extensive
lateral travel was a constant source of maintenance. Investi-
gations found that the FSAA had limitations in matching flight
performance in control of flight path near touchdown (ref. 29),
indicating that available vertical motion cues were inadequate to
provide high-fidelity simulation of flare and touchdown. The
fore and aft travel (+ 3.5 ft.) was adequate for conventional
aircraft simulation; however more travel was needed for
simulation of low speed maneuvers with VTOL aircraft and
rotorcraft. The realism of the simulation was compromised by a
velocity-related audible noise in the lateral drive motion system
(refs.25 and 26) and required amplification of aural cues of the
engine to mask this disturbance. The Redifon TV visual system,
although providing real life color cues and a well appreciated
go-around capability was limited in diversity for multi task
operation and by the narrow field of view. The FSAA was
dismantled circa 1985. The first use of the VMS was a Space Shuttle approach and
landing study .. The VMS has continued to be used by
Vertical motion simulator Building on the lessons learned from Astronauts to examine important operational characteristics.
the FSAA operation, an improved simulator (VMS) (fig.l 1), The XV-15 tilt rotor aircraft design, flight experiments, and
began operation in 1980 with a better match in motion harmony . handling qualities (ref. 32) were carried out using the VMS.
Compared to the FSAA, it had extended vertical travel and less Other major research and development programs include the
than half the amount of lateral motion. In 1987, a major UH-60 Black Hawk Helicopter, LHX Helicopter, RASCAL,
upgrade to the VMS (refs. 30 and 31) provided enhanced motion YAV8B Harrier, STOVL Fighter Attack Aircraft, Oblique
capabilities and visual cues. The VMS (ref. 33) utilizes a large Wing, Automated Nap-Of-The-Earth Rotorcraft Guidance,
aluminum platform supported from below by two pneumatic Helicopter Terrain-Following Terrain-Avoidance System,
equilibrators, each of which is driven vertically (± 22 ft) by four VTOL Flight Controls and Display Concepts for Shipboard
dc servo motors. The lateral travel (± 15 ft) is provided by a Operations, Computer Aided Guidance for Low-Altitude
carriage moved across the platform by four d.c. servo motors. A Helicopter Flight.
hydraulically driven four degrees-of-freedom motion system
provides angular-rate and acceleration capabilities and longi - The ICAB feature provided increased utility and efficiency in
tudinal displacement (± 3 ft). A programmable vibration conducting research on a wide variety of vehicle concepts.
generator system, was needed for high frequency, low amplitude The ability to reconfigure, check out, and use a cab for fixed-
accelerations, characteristic of helicopter vibrations. base simulation in a separate development area has proved
invaluable. The advent of digital computers and in-house
Four interchangeable cabs (ICABs) are available to simulate a development of remote input/output interfaces have reduced
variety of aerospace vehicles. A three-window two-seat side-by- greatly the number of electrical and data leads connected to the
side configuration is used for simulating transport aircraft or the cab thus improving reliability and reducing maintenance. There
Space Shuttle, two four-window configurations are used is a continuing need for increasing computational capacity to
primarily for simulating helicopters and V/STOL aircraft, and a model more advanced vehicle/rotor configurations, advanced
three-window continuous wraparound configuration for single rotorcraft on-board systems, and integrated flight/propulsion
control. The replacement of the TV terrain model visual system aircraft interfaces and flight control devices, is designed to
with CIGs was needed for a much improved field of view, permit virtually unlimited flexibility in information presentation,
clarity, and scene content and detail. Lead compensation and command and control by the aircrew. Such flexibility
(ref.33) was required to remove inherent transport lags. permits the simulation of operations that may be possible with
advanced aircraft and air traffic control concepts and equipment
The VMS with existing motion capabilities and improved visual of the future. The ATC system simulator provides a realistic
system has received a high degree of pilot acceptance for the ATC environment, including communications systems as they
majority of test programs. However it still lacks the ability to impact crew performance.
accurately simulate all the conditions conducive to PIO or the
complex aero/structural/control interface for the VLA transport. The Boeing 747-400 flight simulator is equipped with
In addition, pilot criticism of motion roughness and noise, programmable flight displays that can be easily modified to
occasional occurrences of "simulator sickness," and reference to create displays aimed at enhancing flight crew situational
lack of depth perception remain as other important challenges awareness. The simulator also has a fully digital control loading
for improvement. system, a six degree-of-freedom motion system, a digital sound
and aural cues system and a fully integrated autoflight system
The crew-vehicle systems research facility (CVSRF), A major which provides aircraft guidance and control. It is also equipped
change in simulator equipment occurred in the mid 1980's. with a weather radar system simulation and a Flight Safety
(ref.32). Previous simulators were designed to study aircraft International VITAL Vile visual system. The visual system can
stability and control problems and define boundaries for depict out the window scenes in either day, dusk, night, or
satisfactory control. As aircraft control systems grew in com- twilight modes. The visual, weather radar, and motion systems
plexity so did the human/machine interface and the need for are tightly coupled, simulating weather effects with a high
specialized simulation research tools. This equipment degree of realism. The Boeing 747-400 simulator provides all
(figure 12) was designed primarily to study human factors in modes of airplane operation from cockpit preflight to parking
aviation safety, (ref. 33) The CVSRF allows scientists to study and shutdown at destination. The simulator flight crew
how errors are mad, as well as the effects of automation, compartment is a fully detailed replica of a current airline
advanced instrumentation, and other factors, such as fatigue, on cockpit.
human performance in aircraft. The facility is used to analyze
performance characteristics of flight crews; formulate principles Like the Boeing 747-400 simulator, the ACFS is also equipped
and design criteria for future aviation environments; evaluate with a six degree-of-freedom motion system and is equipped
new and contemporary air traffic control procedures; and with a Link-Miles Image II visual system. This unique research
develop new training and simulation techniques required by the tool is a generic commercial transport aircraft simulator
continued technical evolution of flight systems. The facility employing many advanced flight systems as well as features
includes a Boeing 747-400 flight simulator, the Advanced existing in the newest aircraft being built today. Among its
Concepts Flight Simulator (ACFS), and a simulated Air Traffic advanced flight systems, the ACFS includes touch sensitive
Control (ATC) system simulator. Both flight simulators are electronic checklists, advanced graphical flight displays, aircraft
capable of performing full-mission simulations. systems schematics, a flight management system, and a spatial-
ized aural warning and communications system. In addition, the
ACFS utilizes sidestick controllers for pitch and roll control
axes. The ACFS generic aircraft was formulated and sized on
the basis of projected user needs beyond the year 2000.

Some typical research programs that have been conducted at the

CVSRF include Complex Air Traffic Control Procedures to
evaluated the advantages and benefits of advanced curved
approach and departure procedures (using the Microwave
Landing System) for high density aircraft operations; the Data
Link Project which demonstrated the feasibility of displaying
real-time weather data; Advanced Glass Cockpit Design study,
which evaluated the usefulness of electronic checklists for
coping with onboard malfunctions in highly automated aircraft;
Triple Parallel Runways Experiments, in cooperation with the
FAA which evaluated the capability of multiple (triple and
quadruple) parallel runways to increase airport capacity in a safe
and acceptable manner. Other simulations supported by the
CVSRF in the near term include support of the NASA High
Speed Research (HSR) and Advanced Subsonic Technology
The Boeing 747-400 simulator represents a current technology
(AST) focused programs, as well as support of FAA programs
state-of-the-art glass cockpit aircraft in an environment that
including the study of Terminal Area Productivity and Air
ensures that aircrew behavior in simulated flights is repre-
Traffic Management.
sentative of actual flight operations. In contrast, the ACFS,
configured with multiple electronic displays, advanced crew-
 1-9

A remaining challenge in the use of this facility is how to

evaluate simulation results in terms of human behavior and
limitations in the total human/machine interface.

The Future Of Ground Based Research Simulators

Based on past experience, future development of piloted
research simulators at Ames will hinge for the most part on
improvements in its subsystems. It is not expected that any
major changes in motion travel will be made, however reducing
time delays of the motion system response will be needed for
solving more complex problems with visual/motion/model
mismatches. Improvements in the cab environment will
continue with better acoustic engineering and better lighting
realism. Also expected is the greater use of the "glass cockpit",
with all-digital displays and voice interactive systems in
simulation of advanced aircraft. A needed advance in tech-
nology will come from the use of later generation CIGs to VSC studies in the USA started using a Grumman F6F-3
improve realism of flight. In addition, advances in computer propeller-driven, low-mid wing, single-place, carrier based
technology will allow an increase in sophistication and com- fighter aircraft in 1947 (figure 13). The need for developing a
plexity of simulation models. Other future developments will VSC capability originated because a "new" carrier based aircraft,
reflect the need to simulate advances made in the cockpits of a Ryan FR-1 Fireball had questionable "effective dihedral" (roll
future transport aircraft to allow the pilot's view to be aug- due to sideslip) such that in the landing configuration, it was not
mented by a combination of sensor data, previously stored data, possible to pick up a low wing by normal use of the rudder. A
and real-time data received through aircraft data links to create "brute force" approach was taken to answer the question by
"virtual reality", meaning that the pilot's view need not be tied building and flight testing three aircraft, with 7°, 9° and 11° of
to the pilot's eye location. geometric dihedral. An Ames engineer viewing the three
modified FR-1 aircraft side-by-side on the flight ramp said
In-Flight Simulation "there must be a better way"-the result was the variable stability
The use of actual aircraft as in-flight simulators started at NACA F6F aircraft (ref 36). In addition to its use to establish more
Ames in 1947, several years before ground base simulators were credible lateral/directional handling qualities specifications, it
developed. The use for in-flight simulation was driven by the served to help develop the lateral/directional stability and control
need to examine and define specific handling qualities related systems for a variety of advanced aircraft including the Douglas
primarily to dynamic (oscillatory) aircraft behavior. Because the X-3, Martin B-57, Convair B-58, Lockheed F-104, Vought F8V,
principal use was to study a range of stability derivatives with Grumman F-l IF, and Cessna T-37 aircraft.
varying degrees of control authority, the name "Variable
Stability and Control" (VSC) aircraft was coined. The net result As tests continued, modifications were made to improve
was the ability to respond to the pilot's commands as would an versatility as a research tool. After several years, directional
aircraft of widely different aerodynamic and geometric features. (rudder) control was added to broaden studies of lateral/
The degree of sophistication has increased to where the aircraft directional characteristics. In addition, the effects of atmo-
can "simulate" a full mission scenario with advanced guidance, spheric turbulence was added by driving the ailerons and rudder
helmet mounted displays, autonomous systems to relieve pilot with a mechanical random input. This provided "consistent"
workload, along with computers programmed to model the approximations of rough air, which helped the pilot in evaluating
dynamics of different type aircraft. These capabilities provide to dynamic response characteristics. A complex differential-
a greater degree a more versatile in-flight simulator, that years control linkage and a remote driven aerodynamic lab system was
ago with vacuum tube circuitry, one could only dream about. developed to prevent the pilot from detecting variable stability
inputs. This was not completely successful and remained a
Over the last 50 years Ames has developed and used over challenge over the life time of F6F use. Since there was no
12 aircraft with varying amounts of in-flight simulation capa- safety pilot to take over control in the event of a control
bility. It is beyond the scope of this paper to review each of malfunction, studies of landing approach characteristics had to
these in detail. Emphasis is placed on those vehicles that were be conducted at altitude.
dedicated to conduct VSC research using visiting pilots to refine
handling qualities requirements. Interesting highlights of flight
test programs are reviewed with primary emphasis placed on
challenges encountered and remaining for the various research
aircraft used.
1-10

The results obtained from the F6F-3 tests showed that the was so overwhelming that very little VSC flight research was
VSC concept had tremendous potential to provide solutions to achieved with this aircraft. A notable exception was a control
handling qualities problems of current and future aircraft sensitivity study for an "unknown" very high performance
designs. The challenge to improve the F6F VSC concept for aircraft, which later proved to be the Lockheed SR-71 Mach 3
higher performance aircraft led to the modification of a North design. The sole reported research (ref- 39) was an evaluation of
American (NAA) F-86A swept wing fighter (ref.37) in the use of DLC for in-flight refueling, which had proved to be
1955.(fig 14.) Initially, only a limited (rudder only) variable demanding for the F100. The DLC capability was achieved by a
stability system was used. This provided not only a more blend of symmetrical aileron deflection and horizontal tail
modern cockpit with an improved mechanical (boosted) control deflection. The DLC use showed a 20 percent improvement in
system, but also allowed research to be extended to much higher reducing vertical error.
speeds and higher values of Dutch roll frequency. The rudder
servo system installed in the F-86A was an electro-hydraulic
type that provided variations in yawing-movement derivatives
including Cnp, Cnr, and Cn§a. Tests flight conditions covered
speed ranges from 0.60 to 0.80 Mach No., and 10,000 to
35,000 ft.. Characteristics of a wide variety of advanced aircraft
were simulated (ref.35).

The next challenge was to add variable lateral control features

(1958) using an F-86E (ref. 38) which had an all-movable
"flying tail". This provided an improvement in simulating rough
air in that combinations of stabilizer and symmetrical aileron
deflections gave lift components which felt more realistic. The
overall system provided a capability to investigate a larger range
of lateral/directional dynamics than available with the F6F-3.
Results with this more representative fighter aircraft (ref-38)
indicated that more lenient damping requirements were
acceptable for the emergency (dampers-failed) condition.
The first VTOL aircraft used in VSC research studies was the
Another step-up in VSC capability used a F-100C aircraft, Bell X-14A (1960), a single place (open cockpit) twin engine
(1960), providing three degree-of-rotary freedom and one vehicle using a deflected jet principle (fig. 16). Two turbojet
translational (vertical) axis in a fighter aircraft designed for GE YJ85 engines located forward in the fuselage and a
super sonic flight, (fig. 15) The pitch variable stability system moveable cascade thrust diverter system provided VTOL
required special control authority limits to insure flight safety capability. Control in hover was obtained from reaction jets
concerns in use of the all-moving horizontal tail. The primary located at the wing tips and tail. To provide VSC, additional
challenge was to design the system such that if the pilot selected electric servo-driven reaction nozzles were to which provide
a high gain setting, the control authority was reduced and vise- variations in rotary damping and control power using a
versa. Similar to previous installations, the pilot control inputs response-feedback technique. The pilot could set the magnitude
were introduced "fly-by-wire" fashion in a response feedback and sign of various input signals in flight. An error detection
VSC system to drive the hydraulic valves for the control system circuit was included in the VSC electronics to monitor signals
actuators. The servo amplifiers were 400-Hz units which commanding the servomotors to avoid erroneous reaction nozzle
through a phase relationship governed the signs of the response positions.
feedback gains. The challenge to maintain this electrical system
 1-11

Control power and damping characteristics were evaluated in

hover to establish boundaries between satisfactory, unsatis-
factory, and unacceptable regions to aid design of future VTOL
concepts. One of the first challenge was to eliminate inherent
pitch/yaw coupling caused by gyroscopic torque of the X-14
co-rotating jet engines. In addition, for specialized tests, a
hydraulically powered vane was mounted in the exhaust flow to
provide a fourth-degree-of freedom-lateral translational control.
Finally, after about 10 years, a model-following system was
added to allow evaluation of attitude stabilized control modes.

The X-14 in-flight simulator was a prolific source of VTOL

*"".' . .'. .
research results spanning over 20 years of test flight operation.
Along the way, it was involved in a number of mishaps
including dual engine flame-out in landing approach, ingestion
of grass turf in hover, and lift off to hover without bleed air for
the reaction nozzles. Over two dozen test pilots flew a variety of The flight experiments conducted made extensive use of the
missions ranging from pre check out for hover of the British explicit model-following control system designed for near-hover
PI 127 (Harrier) to simulated lunar landing profiles flown by maneuvering. The effects of high-order dynamics on the
Neil Armstrong. It provided for the first time, a more accurate bandwidth of helicopter flight control systems was investigated
definition of pitch, roll, and yaw hover control and rotary which included control and display law design and landing
damping requirements in a realistic environment of jet exhaust qualities studies. These investigations sought design goals and
ground effects. Comparison of in-flight and ground based handling qualities issues associated with the symbol drive laws
simulation results indicated that pilots were willing to accept for hovering displays such as that used in the AH-64 Apache
smaller control power values in-flight when hovering out of Pilot Night Vision System, and validated advanced, integrated
ground effect. control-and-display laws for blind vertical landings as applied to
a VTOL fighter.
In summing up, there were limitations inherent in the X-14
powered lift concept which limited its versatility as a research The CH-47B proved to be a versatile research tool with
tool. First, tests to define control criteria for STOL operation 450 research hours and was flown by many visiting pilots for
was not feasible because of unsteady aircraft behavior at 10 years. It had limitations, however that prevented its use for
transition flight speeds (60 knots) due to outward flow of the studies of certain critical areas such as high-bandwidth flight
exhaust gases when the thrust vectoring vanes were partially control and agility/maneuverability due to system delays in the
deflected. Another was a safety issue associated with the need digital flight control system and by potential destabilization of
to fly close to the ground in hover. The X-14 was normally the rotor modes. In addition, the actuator rate monitors limited
operated at a hover altitude of 12-15 ft.; high enough to avoid aggressive maneuvering and nuisance disengagements were
hot gas ingestion and low enough to be able to survive an engine common. Finally, the basic CH-47B airframe had drag
failure. For various reasons, loss of lateral control resulted in characteristics which limited its ability to correctly simulate
"hard" landings where the pilot was unable to completely other VTOL concepts such as a tilt wing.
eliminated lateral drift before ground contact occurred. Finally,
accurate definition of minimum acceptable control boundaries
(P.R. 6 1/2) for landing is traditionally difficult because of the
potential loss of flight path control and tests are usually made at ^jM^^ggBagsK
m
altitude. For VTOL aircraft, tests must be made close to the
ground such that translational velocities are apparent to the pilot. ??*?'■ ,pf ■* '1^'dk ääO*- >3äW "'
The ability to safely conduct in-flight simulation tests in hover
remains a challenge particularly for single place aircraft.

Research on handling qualities criteria for rotorcraft using

helicopter in-flight simulation started at Ames in 1977. A
CH-47B (fig. 17) helicopter equipped with full-authority,
electro-hydraulic actuators in each of the four control axes and
with motions transmitted to the flight control system through
rotary clutches (ref. 40). In-line triply-redundant, digital control
rate monitors were set to disengage the research system when-
ever excessive control rates were commanded for safety
purposes. In the event of a failure to disengage, the clutches
could be slipped by the safety pilot. The research system was
qualified for use up to 120 knots and for hover operation to
The use of in-flight simulation is continuing at NASA Ames
touchdown.
using a UH-60A Black Hawk helicopter (fig. 18) described in
1-12

detail in Session V (b). The research system includes a full - Ground Simulator Studies. NASA MEMO 10-1-58A,
authority, programmable, fault-tolerant/fail-safe, fly-by-wire Oct. 1958.
flight control system and a real-time obstacle detection and
avoidance system which will generate low- altitude guidance 3. Creer, Brent Y.; Stewart, John D.; Merrick, Robert B.; and
commands to the pilot on a wide field-of-view, color helmet- Drinkwater, Fred J., Ill: A Pilot Opinion Study of Lateral
mounted display. Judging from current funding and flight Control Requirements for Fighter-Type Aircraft. NASA
program restrictions, there appears to be the end of an era for in- MEMO 1-29-59A, Mar. 1959.
flight simulation at Ames. In part, this is a natural consequence
4. Faye, Alan E., Jr.: Attitude Control Requirements for
of advancement in other testing technologies and the fact that
Hovering Determined Through the Use of a Piloted Flight
VSC aircraft at Ames had fulfilled a highly specialized role in
Simulator. NASA TN D-792, April 1961.
flight testing in a timely matter. They were unique in capability
to define handling qualities/pilot rated boundaries and for their 5. White, Maurice D.; Vomaske, Richard F.; McNeill,
use for researching problem areas not anticipated when first Walter E.; and Cooper, George E.: A Preliminary Study of
developed. Their strong virtue from the pilots' standpoint was Handling-Qualities Requirements of Supersonic Transports
that they flew like real aircraft, in contrast to the make-believe in High-Speed Cruising Flight Using Piloted Simulators.
imitations encountered in early ground based simulators. NASA TN D-1888, May 1963.

CONCLUDING REMARKS 6. Dusterberry, John C; and Barnett, Robert M.: A

A brief historical review of the development and use of piloted Multipurpose Research Laboratory for Flight Simulation.
research simulators at NASA Ames has shown that the specific ASME paper no. 63-AHGT-87, April 1964.
task and related maneuvering requirements dictated the level of
motion system sophistication required. When conducting tests 7. White, Maurice D.; Bray, Richard S.; and Cooper, George
to determine selection of carrier approach speeds with an aircraft E.: Some Design Problems of Supersonic Transports as
that had satisfactory stability and control characteristics, a fixed Identified in Piloted-Simulator Studies. Presented to 3rd
base simulator was adequate. When evaluating flight path ICAS Congress, Stockholm, Sweden, Aug.-Sept. 1962,
control for a vehicle with stability and control limitations such as pp. 1043-1064.
the space shuttle, a very sophisticated all axes motion system is
required to achieve meaningful results. The type and amount of 8. White, Maurice D., and Cooper, George E.: A Piloted
motion needed to evaluate aircraft control also depends on the Simulation Study of Operational Aspects of the Stall Pitch-
task. For example, in a VTOL hover task, initial angular control Up. NASA TN D-4071, July 1967.
response is sufficient. For tasks where translational positioning
is critical such as engine failure during take off, a large amount 9. Vomaske, Richard F.; Sadoff, Melvin; and Drinkwater,
of lateral translation is required. Experience has indicated that Fred J., Ill: The Effect of Lateral-Directional Control
Coupling on Pilot Control of an Airplane as Determined
in a trade off with all six degrees of freedom, vertical motion
in Flight and in a Fixed-Base Flight Simulator. NASA
should not be compromised. The degree of motion system
sophistication is also influenced by the type of visual display. TND-1141,Nov. 1961.
In general, TV displays were adequate only when limited
10. White, Maurice D.; Sadoff, Melvin; Bray, Richard S., and
maneuvering was required. CIG displays with a wide field of
Cooper, George E.: Assessment of Critical Problem Areas
view are required to define control requirements for translational
of the Supersonic Transport by Means of Piloted Simu-
tasks. lators. Aerospace Engineering, vol. 21, no. 5, May 1962,
pp. 12-21.
In-flight simulation played a timely role in defining critical
handling qualities criteria primarily influenced by dynamic 11. Bray, Richard S.: A Piloted Simulator Study of Longi -
(oscillatory) response characteristics. Continued use will be in tudinal Handling Qualities of Supersonic Transports in the
highly specialized roles such as the need to more accurately Landing Maneuver. NASA TN D-2251, April 1964.
define criteria for coupled guidance, control, and display
systems for real mission tasks that heavily involve pilot/machine 12. Cooper, George E.: The Use of Piloted Flight Simulators in
interface. Understanding (interpreting) human functioning and Take-Off and Landing Research. AGARD Report 430,
interaction in the human/simulator system remains an important Jan. 1963.
challenge for continuing and future piloted simulation research.
13. Quigley, Hervey C; and Lawson, Herbert F., Jr.: Simulator
Study of the Lateral-Directional Handling Qualities of a
REFERENCES Large Four-Propellered STOL Transport Airplane. NASA
1. White, Maurice D.; and Drinkwater,Fred J., Ill: TND-1773, May 1963.
A Comparison of CarrierApproach Speeds as Determined
From Flight Tests and From Pilot-Operated Simulator 14. Bray, Richard S.: Experiences with Visual Simulation in
Studies. NACA RM A57D30, Junel957. Landing and Take-Off Research. AGARD Simulation for
Aerospace Research. AGARDograph 99, Feb. 1964,
2. Sadoff, Melvin: The Effects of Longitudinal Control- pp. 75-91.
System Dynamics on Pilot Opinion and Response
Characteristics as Determined from Flight Tests and from
 1-13

15. Stinnett, Glen W.: Piloted Simulator Studies of New 29. Bray, Richard S.; Drinkwater, Fred J., Ill; and Fry, Emmett
Aircraft Missions. AGARD Simulation for Aerospace B.: The Influence of Motion on the Effectiveness of Flight
Research. AGARDograph 99, Feb. 1964, pp. 113-129. Simulators in Training Maneuvers. Proceedings of NASA
Safety and Operating Problems Conference, Hampton,
16. Bray, Richard S.: Vertical Motion Requirements for Virginia, May 4-6, 1971, NASA SP-270, vol. 1,
Landing Simulation. NASA TM X-62,236. Feb. 1973. pp. 207-220.

17. Gerdes, Ronald M.: A Piloted Motion Simulator 30. Statler, Irving C; and Deel, Arlin: The Role of the
Investigation of VTOL Height-Control Requirements. Research Simulator in the Systems Development of
NASA TN D-2451, Aug. 1964. Rotorcraft. NASA TM 81276, Mar. 1981.

18. Sadoff, Melvin; and Bray, Richard S.: Summary of NASA 31. Cook, Anthony M.: Simulation World Moves up to
Research on Jet Transport Control Problems in Severe V/STOVL. Aerospace America, Nov. 1985, pp. 46-48.
Turbulence. J. of Aircraft, vol. 3, no. 3, May-June 1966,
pp. 193-200. 32. Alderete, Thomas S.: Rotorcraft Simulation at Ames.
Proceedings of the Conference on Aerosp&ce Simulation,
19. Holzhauser, Curt A.; Innis, Robert C; and Vomaske, San Diego, CA, Feb. 1984, pp. 39-49.
Richard F.: A Flight and Simulator Study of the Handling
Qualities of a Deflected Slipstream STOL Seaplane having 33. McFarland, Richard E.: Transport Delay Compensation
Four Propellers and Boundary-Layer Control. NASA for Computer-Generated Imagery Systems. NASA
TN D-2966, Sept. 1965. TM 100084, Jan. 1988.

20. Quigley, Hervey C; Innis, Robert C; Vomaske, Richard F.; 34. Styles, Frederick J.: Research and Technology Annual
and Ratcliff, Jack W.: A Flight and Simulator Study of Report 1984: Man Vehicle Systems Research Facility.
Directional Augmentation Criteria for a Four-Propellered NASA TM 86662, 1984.
STOL Airplane. NASA TN D-3909, May 1967.
35. Shiner, Robert J.: Research and Technology Annual Report
21. Quigley, Hervey C: Simulation Techniques for the Study 1992: Research and Development of Glass Cockpit
of V/STOL Problems. AGARD Simulation for Aerospace Technology. NASA TM 103996, 1992.
Research. AGARDograph 99, Feb. 1964, pp. 93-112.
36. Kauffnam, William M., Liddell, Jr., Charles J., Smith,
22. Greif, Richard K.; Fry, Emmett B.; Gossett, Terrrence D.; Allan Smith, and Van Dyke, Jr., Rudolph D.: National
and Gerdes, Ronald M.: Simulator Investigations of Advisory Committee for Aeronautics. An Apparatus For
Various Control Systems for VTOL Aircraft. Conference Varying Effective Dihedral In Flight with Application To
on V/STOVL and STOL Aircraft, NASA SP-116, April, A Study Of Tolerable Dihedral On A Conventional Fighter
1966, pp. 249-267. Airplane. Report 948, 1949.

23. Fry, Emmett B.; Greif, Richard K.; and Gerdes, Ronald M.: 37. McNeill, Walter E.: A Summary of Results Obtained
Use of a Six-Degrees-Of-Freedom Motion Simulator for During Flight Simulation Of Several Aircraft Prototypes
VTOL Hovering Tasks. NASA TN D-5383, Aug. 1969. With Variable-Stability Airplanes. NACA RM A56C08,
May 25, 1956.
24. White, Maurice D.; and Dusterberry, John C: The
Simulator to Match the Transports to Come. Astronautics 38. McNeill, Walter E. and Vomaske, Richard F.: A Flight
and Aeronautics, Sept. 1969, pp. 54-60. Investigation To Determine The Lateral Oscillatory
Damping Acceptable For An Airplane In The Landing
25. Zuccaro, Joseph J.: The Flight Simulator for Advanced Approach. NASA Memo 12-10-58A, February 1959.
Aircraft - A New Aeronautical Research Tool. AIAA
paper no. 70-359, Mar. 1970. 39. McNeill, Walter E.; and Gerdes, Ronald M.; and Innis,
Robert C; and Ratcliff, Jack D.: A Flight Study Of The
26. Bray, Richard S.: Initial Operating Experience with an Use Of Direct-Life-Control Flaps To Improve Station
Aircraft Simulator Having Extensive Lateral Motion. Keeping During In-Flight Refueling. NASA TM X-2936,
NASA TM X-62,155, May 1972. October 1973.

27. Barnett, Robert M.: Computer Requirements for Manned 40. Aiken, Edwin W., Hindson, William S., Lebacqz, J. Victor,
Aerospace Research Simulation Facilities. AGARD Denery, Dallas G., and Eshow, Michelle M.: Rotorcraft
Simulation for Aerospace Research. AGARDograph 99, In-Flight Simulation Research at NASA Ames Research
Feb. 1964, pp. 61-73. Center: A Review of the 80's and Plans for the 90's.
NASA TM 103873, August 1991.
28. Dusterberry, John C; and White, Maurice D.: The
Development and Use of Large-Motion Simulator Systems
in Aeronautical Research and Development. In 50 Years of
Flight Simulation: Proceedings of the Conference, London,
England, April 23-25, 1979, Session 2, pp. 1-16.
 2-1

SIMULATED VISUAL SCENES - WHICH ARE THE CRITICAL CUES?

H.M. M'Intyre
M.E.C. Roberts
Thomson Training & Simulation Ltd
Gatwick Road
Crawley
West Sussex, RH10 2RL, UK

SUMMARY the simulator. Training may be a matter of

practising existing skills or learning new skills for
Research has shown that pilots can extract transfer to the real world. In the case of aircraft
information from relatively impoverished visual simulation these may be handling skills for the
scenes. However, performance of a variety of novice or tactical/mission skills for more
simulated flight tasks improves with greater scene experienced pilots. In this paper we are concerned
complexity. Simulator visual systems cannot only with aircraft handling skills which depend on
replicate the real world. Further, it is not possible the use of external visual information, not those
to optimise visual system performance in all areas which rely on utilisation of instrument displays
simultaneously. Some improvements in flight (either head up or head down).
simulator visual cueing will come inevitably, as
technology advances. Others present a research This paper considers the question of how we can
challenge, particularly where the likely effects of define which cues are critical to the success of a
missing, contradictory or distorted information are simulated visual scene.
not fully understood. These include:
A logical approach may be specified as follows:
The luminance dynamic range of the display;
this is far less than that encountered in reality. Identify visual judgments pilots need to make
Relative luminances between objects cannot be when performing essential tasks.
maintained. Luminance variations with range
will therefore be distorted. Maintaining Identify links to real world characteristics
accurate colour ratios at low luminances is which supply the information needed to make
also difficult. By careful mapping, detection these judgments. (This requires an
ranges could be adjusted to be nominally understanding of the principles of how such
accurate under specific conditions but not judgments are made and the perceptual
continuously accurate. The implications need mechanisms which underpin the visual system
to be considered carefully. response. It would be necessary to take into
account research carried out into both static
The simulation of night scenes, with some and dynamic visual perception).
illuminated areas may require the
simultaneous mixing of 2 or 3 models in the Review current simulator capabilities and
same scene, creating unusual data base limitations as these will constrain present cue
management demands. This requires further representation and the ability to match specific
investigation. cues to specific task requirements. These
also represent areas for improvement/research.
Distance judgments may be observed to be
inaccurate in the simulator. To prevent this The apparent simplicity of this approach, masks a
leading to degraded simulator performance and number of inherent complications. There are gaps
deficiencies in training it may be possible to in our current knowledge, as reflected in our
compensate for the absence of some cues by incomplete understanding of human perceptual
enhancing the effect of others. Further processes, of how real world information is used
investigation is required to establish whether and the consequences of a reduced fidelity
such compensation is truly possible, to what environment in the simulator.
degree it enhances simulator effectiveness and
to identify associated costs. 2. IDENTIFICATION OF VISUAL
JUDGMENTS
1. INTRODUCTION The intention would be to encompass all the
different types of judgment pilots make while
"Critical" cues are defined here as those necessary flying.
and sufficient to enable training to take place in

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?'
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
2-2

2.1 Subjective Approach speed and altitude from a simple grid pattern on a
One option is to ask pilots themselves about the flat terrain surface (Ref 1). However, performance
judgments they need to make when performing of a variety of simulated flight tasks improves with
essential tasks within a specified environment. greater complexity and the addition of specific
They could even be asked to identify the cues they features. The most highly rated scenes are those
would use to help them make these judgments. representing discontinuous distributions, including
This approach is attractive because it could be used regions with "considerable vertical and horizontal
to take into account the needs of the user in extent." Scenes such as this are typical of the
deriving a visual database specification. For each natural world e.g. groups of trees and/or large
mission, or part of a mission, an analysis can be buildings separated by fairly large spaces. The
carried out, following a structured approach; for work of Lintern (Ref 2), indicates that subjects
example task analysis based on a knowledge trained with a simplified scene never subsequently
elicitation exercise carried out with experienced achieve the same high level of performance as
pilots and instructors. those trained with a relatively detailed landscape.
This suggests that a realistic visual environment is
Caution should be exercised as it is by no means important, at least for the initial acquisition of
certain that even "expert" aviators are capable of appropriate skills. Caution should therefore be
identifying all aspects of visual cueing that they exercised in deciding what may be omitted.
use. This is because the skills being exercised,
such as judging distances involve perceptual Many aspects of the visual environment, both static
processes which operate at an unconscious level. and dynamic, contain potential information. For
These are in fact universal human skills which example, static cues, such as texture gradients,
have become adapted to the flight environment, in colour contrasts and luminance contrasts, may be
which orientation with respect to external objects used to identify objects and terrain features (Ref
can be much more varied than in a conventional 3). Equally, dynamic cues, such as motion
earthbound setting. The adaptation process is parallax and optic flow can fulfil the same
mediated through experiential learning, it is not function. The optic flow pattern (Ref 4) can be
explicitly taught and the operation of such skills is used to estimate the direction, rate of self motion
not consciously monitored. and orientation of a vehicle

2.2 Objective Approach For distance and closure information, again both
An alternative to conducting interviews with static and dynamic cues may be of importance. As
individual pilots is to refer to previous findings in an illustration of the complexity involved in such
the research literature. When moving through the perceptual processes some of the mechanisms of
terrain, in good visibility, pilots have been shown depth/distance perception are discussed in more
to use visual cues to: detail below. Static cues include both monocular
phenomena ( such as object size and surface
identify terrain features texture) and also binocular cues (such as
determine rate and direction of movement convergence and stereopsis). Motion parallax and
determine rate of closure texture changes provide important dynamic
determine orientation distance cues.
determine distance from obstacles
determine height above terrain 3.1 Static Depth/Distance Cues - Monocular
Object size: Absolute distance judgments are
The characteristics associated with appropriate probably dependent on size and perspective i.e.
sources of cueing information may be predictable relative size on the retina. Familiarity with the
in some cases but the actual environmental cue object and its likely size is important. Pilots
used will depend on the task and the specific would, or at least could, be consciously aware of
visual environment. Further, individual pilots may using this cue. Anecdotal evidence suggests that
choose idiosyncratic features within that visual Army helicopter pilots make use of trees to help
environment to aid them. There may be a number them visualise long range distance. They have a
of possible cues and an individual pilot is not notion of the height of a "standard tree." This can
necessarily dependent on any particular one, nor do lead to errors if appearances are deceptive, for
they all need to be present. This adds to the example if there are stunted trees. However, the
difficulty of determining which cues are the level of detail visible will also vary with range
essential ones. and can be used to confirm an initial distance
judgment. The smallest area that can be
3. SOURCES OF INFORMATION AND discriminated as distinct from its background
VISUAL PERCEPTUAL MECHANISMS varies considerably with range. At 20 ft, an
insulator on one of the crossbars of a telegraph
Pilots have been shown to be able to extract pole may be discernible. At 4000 ft the observer
information, even from relatively impoverished would be looking for the whole pole.
visual scenes, for example detecting changes in
 2-3

Texture: Texture gradient provides information on delineating boundaries between surfaces.

the relative distance of locations on a surface and
on the shape and slant of the surface itself. Abrupt Which perceptual mechanism is of greatest
change in texture indicates separate objects. importance in providing the observer with
information may vary according to circumstances.
3.2 Static Depth/Distance cues • Binocular For example, at slow speeds motion parallax
Convergence: The closer an object of interest is information is weak. Distance judgments would
to an observer the more the direction of gaze of depend mostly on size, relative size and stereopsis.
both eyes will be turned inwards, towards the nose, At longer range, size is more sensitive as a range
i.e. the greater the degree of convergence of the discriminator, than stereopsis. There is however a
eyes. Cues to distance are derived from feedback cross over at a range of about 10m when range
on the degree of contraction in the relevant discernibility due to stereopsis becomes more
controlling muscles of the eyes. At the same time sensitive than that of size (Ref 5, Ref 6).
the lens in each eye accommodates by becoming
thicker in order to ensure the image is in focus. Assuming cue requirements can be established
The degree of accommodation can also be these will need to be translated into simulator
interpreted as a cue to distance. requirements. Some critical issues which limit the
simulator's performance need to be considered.
Stereopsis: Due to the horizontal separation These are discussed in section 4 below.
between the two eyes, each sees a slightly different
view of the external world. There is overlap 4. SIMULATOR CONSIDERATIONS
between the two visual fields which provides
stereoscopic information. If the eyes are focused A number of factors influence the simulator's
on an object, the image of this object is said to fall ability to represent visual cues adequately. These
on corresponding points on each retina. At the include:
same time the images of an object which is either
nearer or more distant will fall on disparate points level of scene detail
on the two retinas. The amount of disparity will display brightness
depend on the distance between the non fixated field of view (FOV)
object and the fixation point. A fairly precise resolution
appreciation of relative distances and depth ambient light levels
relationships between objects can be derived from flicker (display refresh rate)
the degree of retinal disparity present. smoothness of apparent motion
contrast brightness
Binocular cues to distance will only work at colour
relatively short distances, say within 30 ft of the accommodation micropsia (Ref 7, Ref 8)
observer with stereopsis providing information lack of true stereopsis
out to longer distances than convergence (Ref 5).
Simulator visual systems cannot replicate the real
3.3 Dynamic Cues to Depth/Distance world. Further, it is not possible to optimise visual
Motion parallax: This allows perception and system performance in all areas simultaneously.
judgment of distance based on differential rates of Compromises need to be made but should be such
displacement of retinal projections corresponding that the simulator configuration will represent
to objects. Assuming that an observer is moving important cues well. Task analysis and knowledge
through the environment, those static objects elicitation exercises can help to identify some of
outside the direction of movement which are closer the important cues, on a task relevant basis.
to the observer move more quickly over the retina However task analysis does not yield the effects of
than do more distant objects. This is an important missing, contradictory or distorted information
cue to the relative distance of objects. which may be encountered in the simulator. In
these cases reference has to be made to previous
Size: For objects in or near the direction of research findings, to consider the potential
movement approach causes an increase in size at deleterious effects on aircrew performance,
the retina and moving away causes a decrease in workload and behaviour.
size. Rate of change of the monocular depth cue
of size provides range and speed information as the The factors listed above affect: perceived realism
greater the distance from the eye the slower the and visual search behaviour (Ref 9) range of
retinal velocity. detection, recognition and identification (Ref 9, Ref
10, Ref 11), reaction times (Ref 11) and
Texture: Dynamic changes in the texture density height/speed judgments (Ref 12). For example
of objects also indicates movement in depth. object recognition is more difficult at low levels of
Small scale variations in terrain contour can be illumination and visual acuity is a function of
very salient (Ref 1). It is suggested that optical display luminance, contrast, resolution and level of
flow discontinuities may play an important role in scene detail. Limitations in display resolution and
2-4

update rates will affect the quality of smooth Night scenes are generally monochrome to the
apparent motion and therefore reduce the viewer, apart from lights and illuminated areas.
effectiveness of motion parallax as a cue to relative The relative luminance of surfaces will change in
distances (Ref 12). transitioning from daylight to night (Ref 14)
requiring different visual colour and contrast
Apart from scene detail, which depends on the models (ie. the polygon shape will tend to remain
simulator's image generation capabilities these the same but different contrasts are required). If
issues are related to the display and are largely there are illuminated areas in the night scene then
interdependent. For example increasing the FOV for a particular area a different, daylight model
of a visual channel degrades resolution, brightness may be required. This is further complicated by
and contrast. Improvements in display brightness, Night Vision Goggle (NVG) viewing when relative
fov, resolution, ambient light levels, flicker and luminances will again change. Examples from
smoothness of apparent motion will be seen as Howard and Burnidge (Ref 14) are: red sandstone
visual systems (Computer Image Generators, appears lighter than dry boulders in daylight but
(CIGs) and displays) develop. On the other hand appears darker than dry boulders in moonlight;
the ability to provide appropriate contrast, to green grass appears darker than gravel in daylight
counteract the influence of accommodation and moonlight but lighter when using NVGs. Yet
micropsia and to provide true stereopsis are areas another contrast model will be required. Thus there
which require research to establish if and how may be a simultaneous mixture of 2 or 3 models in
improvements would be possible. Each of these the same scene requiring a form of data base
three areas will be considered in turn. management based on conditions, rather than the
conventional reliance on position and heading
4.1 Contrast alone. The relative discrimination between these
Contrast, both brightness and colour contrasts, surfaces will also be a function of the display and
affect; distance judgements at all ranges, the human observer (Ref 13) and emphasises the
discernibility of cues, realism and the observed contrast mapping problem mentioned earlier.
quality of the image (Ref 13).
NVGs require stringent controls on lighting in the
One major problem is that the luminance dynamic simulator cockpit, for example reflections should
range of the display device is far less than that be kept down to a minimum, as the goggles are
encountered in reality. Hence the computed sensitive to any stray light, including non visible
luminances need to be mapped onto a display light. Consideration of such factors as non
luminance which has a smaller dynamic range. visible light do not normally concern display
This reduction in dynamic range means that manufacturers. If NVGs are to be used in the
relative luminances between objects cannot be simulator there may be increased expense in
maintained in the simulator so that luminance selecting suitable displays.
variations with range will be distorted and, more
seriously, detection ranges will be in error. By 4.2 Accommodation Micropsia
careful mapping, taking into account the luminance As already stated absolute distance judgments are
perception of the user and the spatial frequencies probably dependent on object size and also on
of the object (Ref 13), detection ranges can be perspective i.e. relative size on the retina. However
adjusted to be nominally accurate under specific this information can be deceptive. When the
conditions but not continuously accurate. Hence it accommodation and convergence of the eyes are
may be important to designate conditions in the increased i.e. when there is a move inwards to
exercise when accuracy is paramount and accept focus on a position closer to the observer this has
discrepancies outside these conditions. an effect on the apparent size of other objects in
the visual scene. An object positioned beyond the
Colour adds another problem to that above: the point of focus appears to be reduced in size (Ref
mapping also has to allow the tracking of colour 8). This phenomenon is commonly known as
from high luminance to low luminance in the accommodation micropsia.
daylight scene. The problem is to maintain accurate
colour ratios at low luminances and maintain The accommodation micropsia phenomenon is not
realistic differentiation of cues from the however simply due to an error in size at the retina
background, although the human ability to (Ref 7). The change in size of retinal image
distinguish colour diminishes as the size of objects attributable to accommodation is virtually
reduces. There is yet another added complication negligible compared to the subjective perceived
in that the user often thinks that computer reduction in object size. The underlying
generated scenes are more realistic when the scene mechanisms for this phenomenon are not
is actually richer in colour than in real life (Ref understood. It seems that there is some
14). Which then is most applicable to training - relationship between distance judgments and visual
accurate representation or what the user thinks is accommodation distance but it is not a simple one.
more realistic?
 2-5

When not viewing the scene directly (Ref 8), as in NOE flight, in which the pilot operates close to the
a simulator, with the image presented through a ground, is particularly demanding. This can mean
display system and at a fixed distance it is known flying in and among buildings, trees, bushes and
that there are likely to be misjudgments of other terrain features. This allows the pilot to take
distance. Whether this is in part attributable to the advantage of both natural and manmade objects to
observer's accommodative state is not known. The screen him from enemy detection systems,
effect on the eyes accommodation when confronted whether visual, optical or electronic. Airspeed
with a simulator visual system with its fixed image and altitude will vary depending on terrain,
distance has not been measured. It is known that weather, ambient light and presence/location of
accommodation tends to drift towards a natural threats but generally the flight will be low ( at 20
resting position unless forced to do otherwise by ft or less) and slow (< 40 kts) (Ref 16).
the acuity demands of the visual scene. This
resting position varies from one individual to The margin for error is very small. The pilot
another but tends to be at a fairly close distance therefore needs to have a good awareness of the
( approx. 1 metre for the normal eye). This vehicle's physical dimensions and performance
response would be expected to real world scenes capabilities and to be continuously and accurately
lacking in information for example at night or in estimating present and projected position relative to
fog, leading to so called "empty field myopia." the immediate environment. There is high
Distance judgments would be likely to be affected. dependency on visual judgment to be able to
A simulator scene lacking in texture might avoid hazards at close proximity such as trees,
produce the same effect. The fixed distance of wires etc. Specifically there is a need to be aware
the simulated image means a lack of demand for of current position relative to these, i.e. proximity,
changes in accommodation, whatever the level of height and aspect relative to those which present
scene detail. This might itself produce biased the greatest potential danger to the aircraft.
distance judgments in a similar fashion.
Distance judgments are of paramount importance.
What is the solution? A simple adjustment to At the slow speeds typical of NOE flying, these
image magnification would presumably not be are mainly based on size, relative size and
sufficient to correct for accommodation micropsia stereopsis (retinal disparity). Information from
as the size of the image at the retina seems not to motion parallax is weak (Ref 5, Ref 15). This
be directly linked to judgments of object size. An contrasts with low level fixed wing flight in
alternative would be to try to overwhelm the false which, owing to the high speeds involved motion
size impression by providing stronger alternative parallax is a strong cue and stereopsis is probably
cues in the form of scene detail and texture. not important as objects remain in the stereoscopic
Another option is to have the image distance of the zone for only a short time. For the helicopter pilot
display device match that of the observed object as close range judgments are probably enhanced by
nearly as possible; requiring a collimated display scene detail and texture. Approach speed
for long distance judgements and non-collimated judgements, i.e. judgments of motion in depth rely
for short distance judgements. on rate of change of size and may use rate of
change in retinal disparity.
4.3 Stereopsis
Lack of true stereopsis has effects on range, height 6. ROTARY WING SIMULATION ■ SPECIAL
and speed judgements. These are more critical in CONSIDERATIONS
rotary wing flight than for fixed wing due to the
closer ranges involved and the slower flight The above considerations present a considerable
speeds (Ref 5, Ref 15). As rotary wing flight and challenge to achieving appropriate simulation for
its simulation requirements are very different from rotary wing aircraft. Close range distance
those of fixed wing it is perhaps necessary to judgments are known to be difficult in the
consider some of the more relevant aspects of simulator at the slow speeds generally required by
these differences in greater detail. helicopter NOE flying. At slower speeds human
sensitivity to motion parallax is reduced. Display
resolution limitations mean that even less
5. ROTARY WING AIRCRAFT - SPECIAL stimulation is available to this visual mechanism in
CONSIDERATIONS the simulator. In the real world stereopsis may
compensate for the deficiency. The judgment of
Rotary wing aircraft possess a number of unique distances of less than 30 ft. probably normally
characteristics; such as the range of manoeuvres relies on some stereopsis information but this is
they carry out. These include stationary hover, not represented in conventional simulators.
short accurate stop/start manoeuvres and flying Distance and approach speed judgments are
slowly, at low level, in and around obstacles. therefore likely to be degraded. The situation is
These place unique demands on the pilot's made worse by the fixed distance of the image
perceptual abilities and therefore on any attempt to which provides a false stereopsis cue and which
simulate the pilot's perceptual environment. degrades size judgments.
2-6

It is probably not practical to provide suitable Research is needed in areas where the human
stereoscopic displays (Ref 15). It is not however responses to information, lack of information or
acceptable that equipment limitations should contradictory information are not yet fully
present pilots with perceptual deficits which can understood. The questions to be considered
lead to degraded performance in the simulator and include:
deficiencies in training. As humans generally
have a variety of perceptual mechanisms which can whether short range judgements at slow speeds
provide them with required information it may be require compensation for lack of stereopsis
possible to compensate for the absence of some information in the simulator; whether such
cues by enhancing the effect of others. In this compensation is possible and if so implications
case, compensation for lack of motion parallax and for both effectiveness and costs
stereopsis cues may be possible to some extent,
by providing finer detail, as range reduces, to effect of fixed image distance on
strengthen relative size and perspective cues (Ref accommodative state of the eyes and
17). Relative size cues provide greater sensitivity possibilities of compensation for
than absolute size. Further investigation is required accommodation micropsia
to establish whether such compensation is truly
possible, to what degree it enhances simulator the matching of colours and contrasts for
effectiveness and also to identify the associated objects to be recognizable at specific ranges
costs. and possible tuning of the matching for
specific tasks.
There may however be undesirable consequences
associated with this solution which requires the whether specific data bases may be required
CIG to generate fine detail at close ranges. NOE for specific tasks.
flight, which typically involves rapid dashes for
cover and moving above close terrain already tests whether simultaneous mixture of contrast
the capabilities of the CIG and its data base models is required for simulation of night time
management system. Rapid changes in the scene and/or NVG operations
mean that the CIG needs to be able to switch its
computing power quickly to show new detail as it 8 REFERENCES
comes into view. If the level of detail increases,
the demands on data storage and computing power 1. Kleiss, JA. "Perceptual dimensions of visual
will rise. This will affect the iteration (update) rate scenes relevant for simulating low altitude flight."
of the computations. It is likely that this rate AL-TR-1992-0011, 1992.
would have to slow down to prevent information
being lost. This has serious implications because 2. Lintern, G et al. "Content, variety and
the iteration rate indirectly controls the display augmentation of simulated visual scenes for
flicker rate and hence the quality and acceptability teaching air-to-ground attack." Human Factors, 29
of the display. The iteration rate also affects the (1), Feb 1987
quality of the smooth motion: the higher the
iteration rate, the smoother the motion. There is 3. Stevens, K. "Computational analysis:
thus no simple solution as any changes made in Implications for visual simulation of terrain." In
one part of the simulation environment are likely W. Richards and K. Dismukes (Eds.) Vision
to have ramifications for a number of others. Research for Flight Simulation. Washington D.C.:
National Academy Press, 1982, pp 38-64
7. CONCLUSIONS.
4. Regan, D.M. "Visual sensory aspects of
The potential areas for improving visual cueing in simulation." In W. Richards and K. Dismukes
flight simulators can be divided into two broad (Eds.) Vision Research for Flight Simulation.
categories, those which can be accomplished Washington D.C.: National Academy Press, 1982,
through development, as the technology advances pp 65-71
and those which present a research challenge.
5. Gold, T. "The limits of stereopsis for depth
Areas where development is likely to produce perception in dynamic visual situations." In
significant improvements in the future include: Proceedings of 1972 SID Symposium, 150-151, 6-
8 June 1972.
increased CIG detail
field of view 6. Sexton S., Burbridge R., Roberts M.E.C.
resolution "Satisfactory visual motion cueing for
display brightness helicopter/VSTOL simulation." RAe.Soc.
reduced levels of flicker Conference: "Progress in Helicopter/VSTOL
enhanced smooth motion Aircraft Simulation," May 1990, pp 12.1-12.19.
 2-7

7. Smith G., Meehan J.W., Day R.H. "The effect

of accommodation on retinal image size." Human
Factors, 34(3), June 1992, pp 289-301.

8. Roscoe S.N. "Aviation Psychology." The Iowa

State University Press, 1980, pp 97-107.

9. Crane P.M. "Evaluation of two wide-field-of-

view display systems for air-combat "training.
Journal of SID, 2/1, 1994, pp 59-61.

10. Johnson J. "Analysis of image forming

systems." Image Intensifier Conference, US Army,
Fort Belvoir, Virginia, 6-7 October, 1958, pp 249-
273.

11. Leibowitz H.W. and Owens D.A. "Can outdoor

activities be carried out during civil twilight?"
Applied Optics, 30 (24), 20 August 1991, pp
3501-3503.

12. Roberts M.E.C. and P.M.Murray. "Optical

flow: the key to the integration of visual and
vestibular motion cueing. " in R.Ae.Soc.,
International Conference Proa, "Advances in Flight
Simulation - Visual and Motion Systems," 29, 30
April, 1986.

13. Carlson C.R. "Economic display design."

Information Display, May 1988, pp 16-19.

14. Howard CM. and Burnidge J.A. "Colors in

natural landscapes. " J.Soc. Information Display, 2
(1), April, 1994, pp 47-55.

15. Cross P. and Roberts M.E.C". Helicopter

simulation visual cueing requirements and
solutions." Conf. on Helicopter Simulation Proc.
F.A.A., April 17, 18, 1984.

16. Hart, S.G. "Helicopter human factors." In E.

Weiner & D. Nagel (Eds.) "Human Factors in
Aviation." New York: Academic Press, 1987

17. Meehan J.W. and Triggs T.J. "Magnification

effects with imaging displays depend on scene
content and viewing conditions." Human Factors,
30, 1988, pp 487-494.
 3-1

Visual scenes for battlefield helicopter operations:

Evaluation of requirements and how to specify them.

CDBDeighton,PhD AAWoodfield,FRAeS
Applied Psychology Department Woodfield Aviation Research
College of Aeronautics 9 Colworth Road
Cranfield University Sharnbrook
Bedford MK43 OAL Bedford MK44 1ET
United Kingdom United Kingdom

ABSTRACT the effectiveness of the simulation of appropriate visual

The structured specification and evaluation of simulated scenes.
visual scenes for training pilots in tasks such as Despite continuing advances in the capabilities of real-
Battlefield Helicopter operations is required to ensure time visual scene generation and display systems,
that cost-effective flight simulator training is being current visual scenes provided for battlefield operations
achieved. Historically this has been a difficult process are, at best, disappointing and, more often, not able to
because of a lack of consistent and prioritised provide adequate support for training, other than
descriptions of different visual flight activities within procedural activities where a visual scene is a
the intended training missions; limited evaluations of the background feature. This is partly because of
contributions and relative importance of different scene limitations in current image generation and display
features within appropriate operational scenes; and a systems, but mainly because of the lack of adequate
lack of any structured way of collating and presenting reference data to assist with the specification, design and
such information to those who specify and design visual evaluation of visual scene systems for specific tasks.
scene databases for flight training simulators. A research
programme fimded by the UK Defence Research It does not seem unreasonable to say that the design aim
Agency was conducted in collaboration with Cranfield should be to achieve a realistic visual scene. However,
University, UK to address these challenges. Specific it is currently not possible to achieve this aim, and it
aims were to identify and prioritise visual flight will probably never be totally possible. Attempts to
activities in the context of a battlefield nap-of-the-earth increase realism are mostly beneficial to the
(NoE) attack mission; to determine the relative performance of mission tasks, but the scene database
importance of visual scene components (e.g. woods, designer will always have to make compromises in the
farm buildings, roads, livestock) to the achievement of representation of the real world and the all important
visual flight activities; and to assess the impact of details that are needed to perform a visual mission. In
removing or adding information to the scene upon the making these decisions the designer needs detailed
achievement of visual flight activities in a mission information about the specific tasks to be simulated and
simulator. The purpose of the paper is twofold: firstly to which visual features are necessary for the successful
describe the human factors procedures and techniques performance of these tasks. Currently such information
used to address these aims, and secondly to describe the is very limited; it often does not relate to specific tasks,
proposed structure of a computer based relational and there are no indications of the relative importance to
database to integrate the results of the programme. the performance of a task of particular types of features
Selected findings from the study are used to highlight or the individual visual elements that are present in a
the structure of the database. feature. The designer uses his skills and expertise to
create appropriate visual scenes and the end result will
1. INTRODUCTION often look quite good at initial acceptance, but, if it is
Simulation of Battlefield Helicopter operations can be a not satisfactory, major changes would be long and costly
major contributor to training programmes and, with the to make at this stage Even if a visual scene database
provision of appropriate geographic scene databases, can looks quite good it is very often found to be
also provide the capability to rehearse missions into disappointingly inadequate in service, because the
previously unfamiliar areas. Battlefield Helicopter customers and the designers have not been able to
operations depend to a very large degree on interpreting identify and prioritise visual features that are important
the visual scene either directly or through Electro- to the tasks that are to be simulated.
Optical systems such as Night Vision Goggles (NVG) or
Forward Looking Infra-Red (FLIR) sensor displays. It
follows that the effectiveness of Helicopter simulators
for battlefield operations will be critically dependent on

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
3-2

The fundamental problems in specifying visual scene component and any comments will be subjective
content for battlefield helicopter operations are opinions, which probably will reflect the expected
importance of the component without any real
a. Lack of consistent and prioritised descriptions
knowledge of how other scene components may
of the different Visual Flight activities, e.g.
alleviate the effects of the loss The effects of removing
control of speed, height and attitude, that are
a component can only be identified by evaluating the
parts of the main tasks within the missions that
effects of its removal in a simulation of the task.
are to be simulated
Simulation evaluations of the influence of important
b. Lack of evaluation in appropriate operational
scene components on mission task performance have
scenarios of the contributions and relative
been completed as part of this study programme
importance of different visual scene objects,
(ref. 12), which has been conducted on behalf of the
textures, etc. to the success of the various
Defence Research Agency This paper is presented in
Visual Flight activities
two parts. In part one the evaluation procedures used to
c. Lack of a structured way of presenting the specify the minimum scene content requirements of
information relevant to the previous two visual simulation systems to support the training of NoE
problems so that designers can identify and helicopter operations are reported. Part two describes a
then apply relevant knowledge to the design of way of structuring the information so that is can be
the overall visual system accessed readily by personnel involved in the design,
In addition, current visual display systems fall short of evaluation and procurement of visual simulator systems.
the real world because
2. EVALUATION PROCEDURES
a. Display resolution is not as good as a pilot's The research objective was the development of a
eyes specification for the procurement of visual simulation
b. Scene brightness and contrast are much less systems to support the training and rehearsal of Nap-of-
than normal daylight (although they can be the-Earth flight in battlefield helicopters. The
appropriate for NVG operations) philosophy adopted was that the information contained
in a visual simulation scene needed to be sufficient to
c. Two-dimensional displays lead to compromises support the student training requirements as specified by
in presenting three-dimensional effects. the instructor. Accordingly, the scene content needed to
It is difficult to overcome these fundamental problems be sufficient to enable the demonstration of appropriate
in a simple way Previous studies have either visual cueing strategies by instructors (e.g. backdrop
concentrated on understanding the contributions of technique) and subsequent rehearsal by students. The
simple visual scene components either in isolation or objective was investigated in three phases which will be
together with a 'manageable' number of other simple described in the following sections.
elements, or, in other cases, they have tried to
summarise the requirements of pilots in a general 2.1 Phase One: Preliminary studies.
descriptive document based on questionnaires and In the first phase observational flights in battlefield
discussions (refs. 1-11) Both types of study have helicopters and structured interviews with aircrew lead
contributed significantly to the body of knowledge about to the definition of the mission tasks; training goals and
visual scene elements and pilots requirements However specification of a real world area to be modelled. A
they have not overcome the fundamental problems battlefield operation consisting of 8 mission tasks, which
because purchasers and designers cannot readily extract were slope landing, transition, tactical flight, spot turn,
and prioritise information relevant to particular quickstep, missile firing, transit and low level
operational tasks and Visual Flight activities One of the navigation was chosen to represent a typical range of
important features of this present study is the visual activities. The training goals were defined as
presentation of visual scene information in a framework precision hover, transition, low level manoeuvre and
of priorities for Visual Flight activities within particular forward flight. The set of flight activities which needed
mission tasks, and the development of a structured way to be mastered to achieve these tasks were the control of
of presenting the knowledge gained during the study. heading, track, ground position, attitude, height, speed,
flight path, range; the avoidance of obstacles; rate of
One of the main problems in determining the relative turn and knowledge of map position.
importance of scene components in the performance of a
flying task is finding out the effects on the task of In addition, as part of this phase a set of 12 visual cueing
removing a single component Pilots have experience of strategies were identified, which were relevant to the
the effects of losing large numbers of scene components achievement of the flight activities. These strategies
in operations over water, snow or sand, but much less were termed recognisable features; vertical backdrop;
experience of the effects of removing a few components lateral backdrop; surface shading; natural texture; man-
from a richer scene In most cases a pilot will not have made object texture; perspective; streaming; ground
any experience of the effects of losing a single rush; colour; line features; and the coincidence of
objects.
 3-3

2.2 Phase Two: Survey research complete and was followed by a debriefing when pilots
The preliminary research led to the development of a were invited to discuss their responses and to comment
survey questionnaire to identify the relative importance on the content and structure of the form.
of each flight activity to the performance of a given
mission; the relative importance of each visual cueing 2.2.3 Hierarchy construction.
strategy to the achievement of the eight flight activities; Group means and standard deviations were calculated
and the information in the real world scene used to for each flight activity and visual cueing strategy.
support the implementation of each visual cueing Separate hierarchies were derived for flight activities
strategy. and visual cues and contained those items which
received a group mean rating of between 5 and 7 (rather
2.2.1 Questionnaire structure and content. important, important or vital). Numerical rankings were
The questionnaire consisted of three sections which assigned separately to each flight activity and visual
were semi-structured and designed for completion in the cueing strategy to represent the relative position of the
presence of the researcher following a 20 minute item within the relevant hierarchy. Tied rankings were
briefing. In the opening section participants were assigned where identical mean ratings of importance
advised of the purpose of the exercise and requested to were achieved for two or more items. As an example of
provide brief details of their flying experience (i.e. this hierarchical organisation Table 1 summarises the
course attending or instructing; status; and flying hours). relative importance of the visual cueing strategies to the
The purpose of section two was to determine the 11 flight activities investigated.
importance of the flight activities to the performance of Information derived from both hierarchies and pilot
each mission tasks. Pilots were provided with the comments was used to define the parameter set of visual
mission task descriptions and requested to evaluate the cueing strategies, features and elements to be
importance of the flight control activities to the overall manipulated during the flight simulation studies
performance of each mission task. This evaluation was conducted in phase three.
undertaken by each pilot for both poor and rich visual
cueing environment. For the purposes of this study a 2.3 Phase three: Flight simulation studies.
poor cueing environment was defined as undulating Two flight simulation studies were conducted to assess
terrain with poor visual cues (e.g. desert, NVG) and a the impact upon pilot performance workload and
rich visual environment as undulating terrain with rich perceived scene content sufficiency of removing or
visual cues (e.g. rural wooded countryside). Importance adding information to a computer generated operational
ratings were made using a 7 point Likert scale with scene. The facilities, parameter sets and methodologies
descriptors 'irrelevant'; 'unimportant'; 'rather applied during each study will be described in the
unimportant'; 'in between'; 'rather important'; 'important' following sections.
and 'vital'.
Section three was designed to quantify the importance of 2.3.1 Simulation facilities.
each visual cueing strategy to the maintenance of each Both trials used Defence Research Agency flight
flight control activity. Similar to section 2 a separate simulation facilities. In the first trial the facility used
evaluation was made for a poor and a rich visual comprised a single seat attack helicopter cockpit with
environment and evaluations returned using the 7 point 200 degree horizontal field of view and 40 degree
Likert scale. vertical field of view presented on 5 monitors mounted
on a 3 degree of freedom (pitch, roll and heave motion
In additional for each visual cue and flight control platform). The visual simulation was presented using
activity pilots were requested, where appropriate, to Silicon Graphics Reality Engines (SGRE) and was based
record examples of real world features which supported upon a rich visual cueing environment regularly used by
the use of a particular visual cue. the Army Air Corps Centre to train battlefield helicopter
operations. The second simulation trial used a single
2.2.2 Procedure. seat helicopter cockpit with 140 degree horizontal field
The questionnaire was pre-tested and approved by the of view and 48 degree vertical field of view mounted on
point of contact at the Army Air Corps Centre Middle a 5 degrees of freedom (pitch, roll, heave, sway, roll,
Wallop and Central Flying School, RAF Shawbury. yaw) motion platform. Differences in the visual
Twenty-one rotary wing instructors from these simulation facilities meant that the scene database for
squadrons took part in the survey (8 RAF and 13 Army manipulation during study 2 needed to be translated
Air Corps). Prior to completing the questionnaire from SGRE to a Thomson Link-Miles IMAGE 600PT.
participants were given a 20 minute briefing outlining This process was not purely a physical one, e.g. ground
the purpose of the study and describing how to record and surface texture patterns needed to be selected from
responses using the 7 point Likert scale. Pilots attended the IMAGE texture pattern libraries to match the
the briefing in groups of at least 5 and completed the granularity and detail of those presented using SGRE.
form in the presence of the research team who were on The extent to which this process produced a baseline
hand to clarify completion procedures. The IMAGE 600PT database which conveyed the same
questionnaire took approximately 45 minutes to
3-4

Table 1: Relevance of Visual Cueing strategies to Visual Flight activities.

visual cueing information as the SGRE database was iv. Verbal commentary of the visual cueing
tested to be satisfactory in a direct comparison on the strategies employed
same simulator. v. Judgement of height and speed at random
2.3.2 Trial design and Parameter sets. points during the mission
A description of the scene content parameters vi. Primary task performance
manipulated during each trial are summarised in Tables The debriefing questionnaire and accompanying scene
2 and 3. The trial design and task sequences used each content sufficiency scale was developed specifically for
of the parametric scene changes within each task. A the trial. The questionnaire was divided into two
repeated measures design was used for both trials sections which were completed for each evaluation task.
whereby eight rotary wing instructors performed a In the first section a matrix of 12 visual cueing strategies
mission on four separate occasions. The mission and up to five flight activities presented. Pilots were
consisted of eight tasks and a maximum of two requested to evaluate the sufficiency of the computer
parameters with two levels were manipulated in five of generated image scene to support the use of the visual
the task training areas. This resulted in a full factorial cueing strategies to control each flight activity.
design and minimised the number of conditions to be Evaluations were made for each strategy and flight
flown by each pilot to four. To vary task order and scene activity using the six point Scene Content Sufficiency
familiarity effects, four start points along the mission scale presented in Figure 1. This scale consisted of two
were defined and the order of these four sequences main categories labelled sufficient and insufficient
varied between pilots. which were further divided into 6 levels ranging from
2.3.3 Methodology. excellent (rating 1) to very poor (rating 6). Each level
The following subjective and objective measures were was described by the extent to which the scene was
used to evaluate the sufficiency of the scene to support deficient and the potential impact of the deficiency upon
the performance of the mission tasks on four separate the pilots visual cueing strategy (no deficiencies to
major deficiencies) and the training requirement. Pilots
occasions.
were instructed to return a zero rating if a given strategy
i. Scene content sufficiency scale was not relevant to the achievement of a particular flight
ii. Debriefing questionnaire activity. In section two the visual cueing strategies were
once again listed and pilots requested to describe the
iii. NASA Task Load Index workload rating scale features in the computer generated image scene which
were used to support the strategy.
 3-5

Table 2. Scene content parameters manipulated during simulation trial 1.

Task Area Parameter title Description

Slope Landing Rotor Downwash Cone projecting from the base of the aircraft

Surface shading Directional source of light on the ground surface

enhancing ridgelines

NoE tactical flight Fogging Distance of 6km.

Outer tree layer Line of trees presented 30 feet from the main wood block
and modelled as a two dimensional surface with
transparency between tree trunks and foliage.

Spot turn Outer tree layer As per NoE area

Known size objects Tractor and oak tree placed near entrance to confined
area. Two tractors placed inside area.

Quickstep Shading As per slope landing area

Known size objects Livestock, farm machinery, stationary road vehicle.

Missile Firing Positional cues Foliage projecting horizontally from the outer tree layer

Outer tree layer As per the spot turn area.

Table 3 . Scene content parameters manipulated during simulation trial 2

Task Area Parameter title Description

Slope Landing Wood texture detail Presentation of phototexture patterning on the sides of
woods.

Line features Tractor tracks

NoE tactical flight Wood texture detail As per slope landing area

Outer tree layer Row of individually modelled three dimensional trees,

spaced 70m apart and 30m from the main wood block &
modelled one third the height of the main wood block.

Spot turn Outer tree layer As per NoE tactical flight

Positioning cues Three dimensional unique trees placed against the wood
block.

Quickstep Line features Tractor tracks as per slope landing area

Objects with vertical Hay bales and bushes.

extent

Missile Firing Outer tree layer Individual line of trees against wood block with height
extending beyond background block.

Unique canopy features Solid triangular blocks placed on the surface of the rear
wood block canopy.
 3-6

SUFFICIENT Excellent No Deficiencies

Scene content is Scene content contains all the 1

sufficient to exercise the information required to exercise
relevant Visual Cueing relevant VCS's to achieve the task
Strategies (VCS's) to training requirement.
support the training of
the flight activity or task
objective.

Very Good Some Deficiencies, but not in the

primary visual features required to
exercise relevant VCS's to achieve the 2
training requirement.

Good Some Deficiencies in primary visual

features required to exercise relevant
VCS's but task training requirement 3
remains achievable.

INSUFFICIENT Fair Deficiencies in primary visual features

required to exercise relevant VCS's.
Scene content is Task could be achieved by the trainee 4
insufficient to exercise with a significant amount of additional
the relevant Visual effort and skill.
Cueing Strategies to
support the training of
the flight activity or task
objective.

Poor Major deficiencies in the primary

visual features required to exercise
relevant VCS's. Task could only be 5
achieved by the trainee through the
use of unrepresentative VCS's which
would be unacceptable for training.

Very Poor Major deficiencies in the visual

features. Task is not achievable.
6

Figure 1. Scene content sufficiency scale

and in group two 2,4,1,3. Each pilot was given a 30
2.3.4 Procedure.
minute practice session during which he became
Eight rotary wing instructors from the Army Air Corps
Training centre at Middle Wallop and the Central flying familiarised with the handling of the simulation and the
mission route. One mission sequence was flown during
school at RAF Shawbury took part in the trial. Six of
a 30 minute continuous sortie. Navigational information
the eight instructors took part in both simulation studies.
On arrival pilots were briefed on the purpose of the trial, was provided by a hand held map and complemented by
handling of the simulation facilities, safety procedures verbal directions from the trials team who were able to
and trials design, methodology and procedures. A video interrogate a 'God-eye' view of the database. This
recording taken of the outside world whilst flying the information was not provided for the low level and
mission in the real world was presented to the pilots to transit navigation tasks. At pre-specified points during
allow practice in the use of the evaluation procedures. each task flight parameters were data-logged to provide
Pilots were then randomly assigned to one of two an indication of primary task performance and flight
groups. In group one the sequence order was 1,3,2,4 instruments were frozen and a judgement of height and
 3-7

speed returned by the pilot. Throughout the task the consists of area, group features, objects and elements.
pilot provided a verbal commentary' of the types of The area is a portion of the database that contains a
strategies and scene content information used to achieve limited and specific set of features and objective e.g. an
the task parameters. The scene content questionnaire urban area contains houses, industrial units, roads but no
and NASA workload scale were completed in the farms or woodlands. Group features denote a collection
debriefing room following the completion of each sortie. of objects that by their arrangement form a recognisable
feature, e.g. a farm site will have a farmhouse, yard,
3. SPECIFICATION OF REQUIREMENTS outbuildings, pond. Objects are the basic unit of a
A wealth of information was collated from the activities visual scene, which is a recognisable feature that can
described in part one of this paper and the briefest exist on its own, e.g. a tree, a house, a car, or a person.
summary of results encompassed 30 pages of the Finally, an element is an identifiable component of an
technical report. This is lengthy and cumbersome to 'object' but does not usually exist on its own e.g. a door
scene modellers who wish to make expedient decisions or stone texture. An additional visual scene component
regarding the types of features which must be included covers environment aspects such as reduced visibility
in a visual simulation to support specific types of and wind. These are not objects but can influence visual
activities. Furthermore, as new knowledge is obtained cues available from the scene.
and the Services wish to consider other missions the
problem of information retrieval will be exacerbated. A 3.1.2 Mission structure.
solution to the data management problem is the A similar hierarchy can be produced to describe the
construction of a relational database which incorporates components that make-up a mission and this is shown on
structured descriptions of both the visual scene and the the right of the diagram. A mission is a complete
mission together with importance ratings. Information operational sortie with particular objectives, which can
held in such a form can be incorporated in a way that is be described as a series of tasks such as hovering, nap of
independent of particular mission scenarios so that the earth flying, low speed manoeuvring and within each
knowledge can be applied directly to a much wider task there are a number of visual flight activities such as
range of missions than that used for the simulation trials. height control, ground position control, attitude control
In addition to providing a structured format to hold data, etc.
such a database can also hold standard forms of objects 3.1.3 Linkage.
for use in automatic database generation tools, together
The general link between the mission factors and the
with references to standard object descriptions such as
visual scene occurs at the level of visual flight activities
those used for the world wide Digital Land Mass Survey
where there are a set of visual cueing strategies that link
(DLMS) cultural features held in DFAD data files and
visual flight activities directly to elements and objects in
those in distributed interactive simulation (DIS)
standards.
In the second part of this paper a
Area
proposed form for a relational Description
database to provide a structured 1
dataset for mission based visual 1
scenes is described and examples
of its use are provided from a
T
Group
demonstration database containing * F eatures
some of the knowledge from the
Cue
present programme. Importance
Component
Reference,
3.1 Description of the structure DFAD, DIS
for data
A general structure for a relational
database to link visual scene
characteristics and missions is
presented in Figure 2. In des-
cribing components of a visual Image
Group
scene it is essential to have a Feature Suitability
hierarchy of features with Cues
consistent names.

3.1.1 Structure for visual scenes.

On the left hand side of the
diagram the hierarchical structure
for visual scenes is presented and Figure 2. General Database Structure
3-8

Table 2. Highest Importance ratings for VCS's in a Battlefield attack mission

Visual Cueing strategy Importance Task Activity

Coincidence of objects 9.32 Nap of the earth Map position

Recognisable features 9.00 Nap of the earth Map position

Vertical backdrop 9.00 Missile firing Attitude

Streaming 8.96 Nap of the earth Vertical flight path

Vertical backdrop 8.96 Low speed manoeuvring Height agl

Ground rush 8.65 Nap of the earth Ground Speed

Lateral backdrop 8.65 Low speed manoeuvring Ground position

Lateral backdrop 8.65 Quick stop Ground position

Line features 8.65 Missile firing Heading

Line features 8.65 Nap of the earth Track

Man-made object texture 8.65 Low speed manoeuvring Ground position

Man-made object texture 8.65 Low speed manoeuvring Height agl

Man-made object texture 8.65 Quick stop Ground position

Natural texture 8.65 Low speed manoeuvring Ground position

Natural texture 8.65 Low speed manoeuvring Height agl

Natural texture 8.65 Quick stop Ground position

Streaming 8.65 Nap of the earth Track

the visual scene. It is these visual cueing strategies that As an example of Visual Cueing strategy priorities,
link the requirements of the Service Operator who consider two missions
defines and knows the Operational Scenario to those of a. A full battlefield attack mission
the Visual Scene Database Designer who defines and is
knowledgeable about the Modelling Scenario. b. A transit navigation mission that might be a
mission for training emergency procedures
4. CALCULATING PRIORITIES FOR MISSION For the battlefield attack mission the preliminary dataset
AND SCENE COMPONENTS identifies 28 combinations with an importance level
On the mission side of the database numerical rankings greater than 8 with the highest importance reaching
can be given for the importance of visual cueing 9.32. These 28 combinations include 9 of the 12 VCS's
strategies to visual flight activities; for the importance of (counting both general and local recognisable objects as
visual flight activities to tasks and for the precision a single VCS during operations). The large number of
needed for a task within a particular mission. An overall conditions with importance greater than 8 illustrates the
relative importance value to the mission can then be importance of visual scene information to the mission.
derived for a combination of visual cueing strategy, The VCS's that are not present with an importance
visual flight activity and task by multiplying the greater than 8 are Perspective, Surface Shading and
numerical values for each of the three combinations and Colour. Surface Shading and Colour were not rated very-
taking the cubed root of the resulting number. relevant to Visual Flight activities in
Table 1 and this reduces their importance to any
mission. Perspective was rated very relevant to Range
 3-9

and Obstacle Avoidance Visual Table 3. Highest Importance ratings for VCS's in a Transit Navigation
Flight activities, but is less mission
important for these activities
than the Recognisable Features
VCS. The 17 conditions with the
Visual Cueing Strategy Importance Flight Activity
highest importance ratings are
listed in Table 2
Coincidence of objects 7.94 Map position
In contrast, if the training
mission is restricted to transit Recognisable features 7.66 Map position
navigation then there are no
conditions with importance Ground rush 7.11 Ground Speed
levels for VCS's that are above
8. The 18 conditions with levels Line features 7.11 Track
of importance greater than 5
from the preliminary dataset are Recognisable features 7.11 Range
listed in Table 3.
Streaming 7.11 Track
The lower levels of importance
compared with the battlefield Streaming 6.54 Ground Speed
attack mission reflect the
reduced importance of visual Man-made object texture 6.21 Range
cues for this mission. Although
these cues are clearly still Natural texture 6.21 Range
necessary, a training mission
may not be compromised if the Perspective 6.21 Range
quality of visual cues is reduced
compared with those needed for Colour 5.85 Range
a Battlefield attack mission.
Recognisable features 5.85 Ground Speed
The only 2 VCS's not
represented in the conditions Streaming 5.85 Vertical flight path
with importance levels greater
than 5 are Surface shading and Ground rush 5.65 Track
Lateral backdrop. Neither of
these VCS's are particularly Vertical backdrop 5.65 Attitude
relevant to Visual Flight
activities that are important to Colour 5.31 Map position
this mission.
Man-made object texture 5.31 Map position
It is interesting and an important
point to note that Coincidence of Natural texture 5.31 Map position
objects and Recognisable
features are the two most
important VCS's for both missions and these occur for combination of flight and simulator training that will be
the Map position Flight activity. This reflects the used for the complete training activity. The types of
importance to the success of the mission of recognising visual scene components that will correspond to the
where the map location of the helicopter is throughout specified missions and image suitability can be obtained
the mission. Pilots are very critical of visual scenes on from the dataset and the purchasers can modify any of
simulators that do not satisfy these VCS's and this these specification features until they are satisfied that
confirms their importance. they are suitable for the intended purpose.

5. PRACTICAL APPLICATIONS OF THE The supplier can then use the specified mission types,
STRUCTURED DATASET area types and level of image suitability to obtain
Structured datasets of information relating visual scene information on specific image detail, including standard
details to mission task priorities can be used by both library features, that are appropriate to meet the
purchasers and suppliers of visual scene generation specification. This will provide a detailed specification
systems for specific flight training simulators. The that can be used to design and build a visual scene
purchaser can specify types of mission and their database and image display system that can be an
associated geographic or generic types of area that are to acceptable form of compliance with the specification. If
be practised on the simulator, and a level of image there is significant spare capability in an Image
suitability that will be acceptable for the expected Generation system proposed by the manufacturer then
3-10

they may choose to enhance the level of image information from the dataset held in the
suitability or add optional mission types which will database.
provide useful enhancements for the purchasers. (This • Generate a battlefield helicopter attack mission
is much more worthwhile than just improving 'realism' visual scene dataset using the completed
at random.) A manufacturer may also choose to offer a dataset. Put all relevant data for a Battlefield
cheaper system that meets the basic specification with Attack mission into the completed database,
less spare capability. including all the data from the present study.
The same structured dataset can also be used by the Verify all the mission based ratings with
purchasers and supplier to define a mutually agreed appropriate Service staff such as the helicopter
acceptance test schedule for the complete image instructors used in the present study.
generation system. • Validation of scene content during training
effectiveness studies. The impact of the
6. CONCLUSIONS minimum scene content specification upon
The aim of the study was to specify the minimum scene training effectiveness is a major challenge
content requirements of visual simulation systems to confronting the community and must be
support the training of battlefield helicopter operations
addressed.
The human factors methods and procedures
implemented during the study provided a successful way • Extend dataset to encompass other helicopter
of collecting the information required to address this training missions. Where appropriate personnel
aim. In particular through semi-structured interviews, conducting visual simulation studies should be
questionnaires and piloted simulation trials both the encouraged to implement the battery of
relative importance of visual cueing strategies to the techniques developed during the study
battlefield helicopter attack mission and the alongside trials specific questionnaires.
effectiveness of many of the major visual scene Additional piloted simulation studies for the
components in supporting visual cueing strategies were few remaining visual cues and Visual Cueing
established. It is recommended that these procedures strategies should be considered.
should form the core of a battery of subjective, • Usability assessment of the content and
objective, physiological and behavioural techniques to structure of the relational database. A
be used in the development, testing and acceptance of complete dataset preferably extended to cover a
rotary and fixed wing visual simulation systems for wide range of helicopter missions should be
rotary and fixed wing operations. evaluated by MoD, a supplier and scene
A structured way of collecting and presenting the modeller to test its usability. This evaluation
information gained from the study using a computer would be expected to identify some necessary
based relational database was outlined. It is concluded changes to the standard queries and perhaps to
that the use of a relational data as a repository for the the structure and types of data in the database.
knowledge gained from this and subsequent studies is • Contribute to the development of international
essential to providing answers to questions from standards for visual scene description to
customers, suppliers, scene modellers about visual improve capabilities for Distributed Interactive
scenes for particular training missions. Simulation exercises. The concept of this
visual scene/mission database and the details in
7. WHERE ARE THE CHALLENGES ?
a complete dataset will provide unique inputs to
The research reported in this paper has contributed
assist in defining international standards in this
significantly to ways of evaluating and specifying visual
field. The concept should be introduced into
scenes for battlefield helicopter operations and has
these discussions as soon as possible and this
identified several additional challenges for the
should be followed by demonstration of the
community. completed dataset.
• Complete and document the design of the visual
• Provision of visual scene/mission datasets for
scene/mission database. Appropriate actions
other types of aircraft, e.g. fast jet low altitude
should be taken to ensure that the design of the
missions. Database design should be adapted to
database is completed and functionally tested.
provide a dataset for other missions where
This should include the design of standard
visual scenes are important, e.g. fast jet low-
forms for entering data, more complete
altitude missions. Existing data on visual scene
definitions of the descriptions of the various
effectiveness that are relevant to such missions
terms used for Visual Cueing strategies, Visual
should be put into the dataset and simulation
Flight activities, etc. and for the description of
trials conducted to evaluate any important
visual scene components and also development
visual scene components and Visual Cueing
of some standard queries to extract appropriate
strategies where data is not available.
 3-11

8. REFERENCES 11 ACM Nicholson. "Visual system for Harrier

and Lynx Flight Simulators: A study of the
E Rinalducci. "Visual cues in the simulation of
requirements. Final Report.", Marconi
low level flight", Georgia Institute of
Command and Control Systems Ltd. Report
Technology, Atlanta, USA, 1983
No. ST/86/8509 (MoD Contract No. A85a/420 -
D McCormick, T Smith, F Lewandowski, Annex B), UK. 1986
F Preskar. E Martin, "Low Altitude Database
12 A A Woodfield, C D B Deighton, "Simulating
Development Evaluation and Research
Visual Scenes for Training and Mision
(LADDER)", Proceedings of the 5th
Rehearsal of Battlefield Helicopter Operations:
Interservice/Industry Training Equipment
Specification, design and Evaluation", Applied
Conference, 1983
Psychology Department, College of
M L Fineberg, J D Dressel, "The effects of Aeronautics, Cranfield University, UK, March
ambient illumination, aircraft velocity and 1995
feedback on absolute distance judgements
during simulated night nap-of-the-earth flight".
Proceedings of Human Factors 21st Annual
Meeting, 1977

E Martin, E Rinalducci, "Low-level flight

simulation: Vertical cues", Air Force Systems
Command, Brooks Air Force Base, Texas,
USA, 1983

J De Maio, R Brooks, "Assessment of simulator

visual cueing effectiveness by psychophysical
techniques", Proceedings of the 4th
Interservice/Industry Training Equipment
Conference, 1982

E Rinalducci, J De Maio, M J Patterson,

R Brooks, "Psychophysical assessment of
simulator visual displays", Proceedings of the
2nd Symposium on Aviation Psychology.
Columbus, Ohio. USA. 1983

J De Maio, R Brooks, "Perception of altitude in

the low and medium ranges", Proceedings of
the 3rd Symposium on Aviation Psychology.
Columbus, Ohio, USA, 1985

J De Maio, E Rinalducci, R Brooks,

J Brunderman, "Visual cueing effectiveness
comparison of perception and flying
performance", 27th meeting of the Human
Factors Society, Norfolk, Virginia, USA, 1983

P M Murray, D R Campbell, P Cross,

M E Roberts, "A feasibility study into visual
systems for low level flight. Final Report
Volume II - Nap-of-the-earth flight",
Rediffusion Simulation Ltd. Report No. RSL-
RD-83-5 (MoD Contract No. A85a/392 - Annex
1), UK, 1983

10 M Simpson, "Task and cue analysis for the

Lynx. Final Report.", Singer Link Miles Ltd,
Report No. LMR 1369 (MoD Contract No.
A85b/687), UK, 1989
 4-1

VISUAL SYSTEM OPERATIONAL EVALUATION

James E. Brown
Lt Col Don R. Poe
Timothy J. Lincourt
ILt Melissa J. Leos
Training Systems Product Group
Wright-Patterson Air Force Base OH 45433-7111
USA

SUMMARY
This paper presents the results of an operational evaluation of 1. INTRODUCTION
the training effectiveness of three different visual technologies. Ground-based simulator training for tactical fighter aircrews is
Purpose of the program was to determine (1) trainability of limited by lack of adequate visual display systems. Efforts to
low altitude tasks on available visual display technology; (2) develop visual systems with the capability to provide useful
demonstrate current visual simulation technology to users; (3) tactics training have met with limited success. A major
get feedback from those users to help define future visual requirement of tactical visual systems is that the display must
requirements; and (4) provide information and data to support have an instantaneously large field-of -view (FOV) both
future simulation acquisition decisions. Three visual horizontally and vertically. This requirement has been
simulation display technologies at three different sites were difficult for industry to meet and still provide resolution and
evaluated: (1) a dome display with head tracked area-of- brightness that is adequate to realistically train tactical flying
interest, (2) a rear-projection display, and (3) a fiber optic tasks. Other constraints have been in the area of data base size
helmet mounted display. A team of highly experienced F-16C and detail. Fighter aircraft rapidly traverse long ranges in a
and F-15E instructor pilots evaluated each of the three display very short time span. This places major emphasis upon data
technologies. Three evaluation missions were flown by each base development and image generation. They also operate at
pilot. At the completion of each evaluation mission, extensive altitudes ranging from the surface to thirty or forty thousand
questionnaires were completed and debriefings were feet. The fighter pilot needs to be able to recognize objects
conducted to rate the training capability of the visual system such as another F-16 with sufficient detail to visually identify
for each task. Results are presented for each of the three other aircraft at realistic tactical ranges, assess aspect angle of
display systems evaluated. The results are expressed in terms another aircraft, fly tactical formation, and identify ground
of tasks that were rated trainable and those tasks that were not vehicles, roads and bridges. This wide range of requirements
trainable in the system. has made it difficult for industry to develop a display system
adequate to meet the full range of fighter training
ABBREVIATIONS requirements.
AB Air Base
ACC Air Combat Command In the past, the Air Force has undertaken operational
AOI Area-of-Interest evaluations to determine if new advances in visual system
DART Display for Advanced Research and Training technology provide capability to train tactical flying tasks.
DB Dive Bomb Among these efforts were Project 2235, Air-to-Ground Visual
DMPI Desired Munitions Point of Impact Simulation Demonstration (1976), Simulator Systems
DTOS Dive Toss Comparative Evaluations (1977, 1979), and F-15 Limited
ESIG Evans and Sutherland Image Generator Field of View Visual System Training Effectiveness
FOHMD Fiber-Optic Helmet Mounted Display Evaluation (1984). The general findings of these efforts
FOV Field of View indicated that existing visual systems could train some but not
HAS High Angle Strafe all critical tactical flying tasks.
HD High Dive
HMD Helmet Mounted Display In March 1989, the United States Air Force In Europe
HUD Heads-Up Display (USAFE) requested assistance from Air Force Materiel
LAB Low Angle Bomb Command to meet its low altitude training needs for the 1990s.
LALD Low Angle Low Drag USAFE aircrews were limited to train at altitudes no less than
LAS Low Angle Strafe 250 feet and at airspeeds no more than 475-550 knots.
LATS Low Altitude Training System However, pilots indicated that in time of war, threat conditions
MTT Multi-Task Trainer might require them to fly at altitudes as low as 100 feet. A
NR Not Rated training systems requirements analysis (TSRA) was conducted
OFT Operational Flight Trainer based upon the F-16C and F-15E weapon systems.
RWR Radar Warning Receiver Recommendations based upon the analysis indicated that
SAM Surface-to-Air Missile modern visual systems have the potential to significantly
TSRA Training System Requirements Analysis enhance available tactical aircraft training and may assist in
USAFE United States Air Force in Europe slowing down the loss of critical low altitude flying skills that
VLD Visual Level Delivery are not frequently practiced due to range or safety constraints
WSO Weapon System Officer (1991). The consensus of most engineering and training

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?'
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
4-2

experts involved in the analysis was that current image evaluate only the FOHMD technology at Site 3 and an F-4
generation technology was adequate to support low altitude FOHMD at Neuberg AB, GE since these sites were the only
training but that image display technology needs further sites with a WSO visual system. Due to operational duty
assessment. To verify adequacy of image display technology commitments, not all pilot evaluators participated in every
to support low altitude training, this operational evaluation, visual system evaluation. The number of evaluators at any
using highly experienced aircrews, was conducted. one site varied between five and seven instructor pilots.

2. SCOPE AND LIMITING FACTORS Even though visual displays were the focus of the evaluation,
The evaluation method used was an evaluation team composed and fidelity was not evaluated per se, it was recognized that
of eight tactical fighter instructor pilots and two instructor cockpit differences, image generation and data base
Weapon Systems Operators (WSOs), with current F-16C, F- capabilities impacted training capability ratings.
15E, and F-l 11 background, to evaluate three different image
display technologies located at three different sites. The image Only subjective aircrew data was gathered during the
display technologies evaluated were: (1) a two-channel head- operational evaluation. Objective data such as bombing
tracked dome display located at Evans and Sutherland scores, hits, etc., was not available at all sites. Engineering
Corporation, Salt Lake City, UT; (2) a rear-projection data was furnished by the site organization and verified by
mosaiced display, the Display for Advanced Research and evaluation team engineering personnel. The evaluation did not
Training (DART)* and a smaller display version (mini- assess specific hardware. Therefore operational suitability
DART)* located at Armstrong Laboratory, Williams Gateway issues were not evaluated although general availability of
Airport, AZ; and (3) a fiber-optic helmet mounted display systems was noted.
(FOHMD) located at CAE Stolberg, Stolberg GE. An
engineering assessment for each device configuration was 3. METHOD OF ACCOMPLISHMENT
conducted to verify the current visual systems display and The evaluation team was composed of a pool of eight fighter
image generator attributes. pilots and three weapon system operators (WSOs) specifically
* Similar display technologies selected for their fighter experience and training background.
All pilots and WSOs had extensive fighter experience in such
The focus of the evaluation was the training capability of weapon systems such as F-4, A-10, F-15C, F-15E, and F-16.
various image display technologies in the high threat low- Two of the pilots had combat experience in the Middle East
altitude environment. The evaluation was conducted over a Conflict and two pilots were Fighter Weapons School
seven month period beginning in January 1993 and ending in graduates. During the evaluation period, one evaluation pilot
July 1993. This evaluation was conducted at commercial and was selected for the Thunderbird Demonstration Team.
government flight simulator facilities. Evaluations were Average flying time for pilots was 1989 hours. Average
performed based upon facility availability and were often instructor pilot time was 1039 hours. Average flying time for
separated in time by as much as two months. It was WSOs was 2213 hours. Average instructor time for WSOs
recognized that experimental order effects, i.e., order in which was 983 hours.
a display technology was evaluated, would be present but had
to be accepted as part of the evaluation. The evaluation was Training for the team was provided in visual system
an operational evaluation and was not structured as an technology and the evaluation process. Classroom instruction
experimental comparison. No transfer of training comparisons and demonstration training on visual system technology were
were made although evaluation aircrews were asked to rate the given to evaluation pilots and WSOs for recognition of visual
training capability of the visual display system under features related to visual displays, visual image generation, and
evaluation. Each system was evaluated to assess the tasks that data base. This training enabled the evaluators to assess the
were rated trainable. interrelationships of visual system components and to focus on
the image displays for training capability ratings.
Due to large differences in weapon system components and
performance capability, it was not possible to compare one The same general evaluation methodology was used at each
visual system to another. Nor was a comparison intended. site with minor differences as required by the system
Rather, the intent of this evaluation was to rate the capability configuration and data base availability.
of a given display technology to support training of a selected
set of twenty-six tactical tasks. The evaluation was conducted at each site for a one week
period. Four generic tactical missions were constructed to
The limited availability of operational instructor pilots and permit each pilot to evaluate twenty-six selected tactical flying
instructor WSOs and the limited availability of the flight tasks. See Table 1. Not all tasks were flown on each mission
simulation facilities meant that the evaluation at each site had but the missions were developed to permit each of the tasks to
to be structured to be conducted and completed within one be evaluated at least once during the evaluation missions if the
week. The evaluation team was composed of a small, highly site equipment and data base permitted. Each mission required
trained group of tactical fighter instructor pilots and approximately 1 hour to complete. The first mission was a
instructor WSOs. Evaluators were given training on familiarization mission. The next three missions were
recognition of visual features related to displays, image evaluation missions designed to evaluate air-to-surface tasks
generation, and data base. Evaluation WSOs were used to such as, tactical formation, low altitude evasive maneuvering,
 4-3

and low altitude air-to-surface weapon employment. The pilot mission, evaluators were briefed on the specific mission to be
tasks for the missions were based upon results from the flown.
USAFE Low-Altitude Training System (LATS) Requirements
Analysis (1991) and modified by Air Combat Command Pilot and WSO evaluators flew each mission and then rated the
tasks flown for training capability using a five point rating
# Pilot Tasks scale. With the exception of the familiarization mission, all
1 Tanker rendezvous evaluators completed an extensive questionnaire after each
mission. This questionnaire was used as the basis for an
2 Tactical formation from fingertip
evaluator debrief to explain ratings. The ratings and
3 Tactical formation above 500 feet questionnaire data were later entered into computer data files
4 Combat descent that were used as an aid to data analysis. The debriefings were
5 Tactical formation below 500 feet tape recorded and used as additional references to interpret the
6 Single ship low level data.
7 Visual low level navigate to initial point
The engineering data collection was conducted as follows.
8 Mutual support/lookout in various tactical formations
Prior to going to the evaluation site, the engineering specialist
9 Detect visual threats requested specific engineering data from the site
10 Detect electronic threats representatives. Two days prior to the team arrival, the
11 Terrain Masking (direct/indirect) engineering specialist visited the site and began collecting and
12 Individual/formation threat reactions
verifying the data provided. The specialist remained on site
throughout the evaluation and was responsible to track down
13 Tactical instruments cross check
limitations of the system and problems noted by the evaluation
14 Visual target acquisition/identification pilots and WSOs during their missions.
15 Coordinated tactical attack
16 Low altitude weapon delivery (LAS, LALD, LAB.VLD) The percentage of acceptable or better ratings for each task
17 High altitude weapon delivery (HAS, HD, DB, DTOS) was compared to the criteria to arrive at assessments. The
measures of effectiveness were the evaluation pilot or WSOs
18 Reform after tactical attack
subjective ratings and the evaluation team's assessment of the
19 Target reattack capability of each system to train pilots or WSOs in an
20 Aircraft battle damage check operational training environment. The criteria were that at
21 Low altitude intercept least 80 percent of the evaluators must rate each task a 3 or
22 AIM-9 employment better (first criterion) or the sub-objective must receive an
overall acceptable assessment by the evaluation team (second
23 Low altitude air-to-air gun employment
criterion).
24 Flight lead responsibilities
25 Wingman responsibilities The intent of the engineering evaluation was to document
26 Situational awareness of tactical situations visual system performance as it existed at the time of the
operational evaluation, not to assess the performance of each
Table 1 List of Pilot Evaluation Tasks visual system. Using the pilot's assessment of the visual
system's capabilities to train the identified tasks, and the
For Site 3 only, four missions were developed to permit each engineering evaluation documented performance, systems
WSO to evaluate twenty-two selected tactical tasks. These engineers will be better able to translate the user's task
missions were the same length as the missions for evaluation requirements into a visual system performance specification.
pilots and were integrated into the missions for the pilot
evaluators. The first mission was a familiarization mission; 4. RESULTS
the next three missions were evaluation missions. The WSO
tasks were identified as high value tactical flying tasks by Air Site 1: Two-Channel Head-Slaved Area of Interest (AOI)
Combat Command (ACC) and were correlated to tasks that Dome/ESIG 3000.
would be performed by the pilots during missions at Site 3.
Evaluation Cockpit
Approximately two days prior to arrival of the evaluation The evaluation cockpit was a generic aircraft cockpit mounted
team, the lead evaluation pilot and WSO would visit each site on a pedestal within a 24 foot diameter dome. The cockpit had
to prepare for the evaluation. They would develop and fly the instruments to represent standard aircraft flight parameters
four generic missions to make adjustments in the flight profiles such as attitude, altitude, airspeed, vertical velocity and,
for the missions based upon limitations of the hardware and engine instruments. It had a control stick and throttle controls
database capabilities of each site. that were representative of a generic fighter, and permitted
control of flaps, speed brakes, and weapons release. Landing
Upon arrival of the team, evaluators were given a briefing on gear control was located on the panel. The cockpit had some
the system at the site, the missions to be flown, evaluation weapons radar and fire control capability but not a fully
procedures, and general plans for the week. Prior to each capable integrated system found in current fighter aircraft.
4-4

There was no Head-Up Display (HUD) and tactical instrument Tactical formation above 500 feet ®
cross-check was all accomplished inside the cockpit. Sound Combat descent ®
and motion were not simulated. Flight performance of the Tactical formation below 500 feet
simulated aircraft was representative of a generic high
®
Single ship low level ®
performance jet fighter aircraft (although not necessarily an F-
16orF-15E. Visual low level navigate to initial point ®
Mutual support/lookout in various tactical formations ®
Image Display Detect visual threats ®
The two-channel image display provided one channel for a Detect electronic threats NR
head tracked background image with the other channel
Terrain Masking (direct/indirect) •
inserted, centered as an area of interest (AOI) scene. The AOI
was superimposed on the background scene directly in front of Individual/formation threat reactions ®
the pilot's head. Both scenes were projected on the 24 foot Tactical instruments cross check ®
dome by light valve projectors mounted above and behind the Visual target acquisition/identification •
cockpit. Coordinated tactical attack NR
Low altitude weapon delivery (LAS, LALD, LAB.VLD) ®
Image Generator
High altitude weapon delivery (HAS, HD, DB, DTOS) •
The image generator was a two channel ESIG 3000 built by
Evans and Sutherland Corporation. Reform after tactical attack NR
Target reattack •
Data Base Aircraft battle damage check NR
The data base was representative of the Hunter-Liggett range Low altitude intercept
in the USA (not correlated to real world).
®
AIM-9 employment NR
Low altitude air-to-air gun employment NR
Discussion
From the list of twenty-six tasks that were initially identified, Flight lead responsibilities NR
eighteen tasks were able to be evaluated during the missions Wingman responsibilities ®
flown. Situational awareness of tactical situations ®
Tanker rendezvous was eliminated by team consensus since ® - Did Not Meet Criteria
% - Met Criteria
this task was not supportable by the simulation capability, is
normally performed frequently during normal operational NR - Not Rated
training missions, and would have required unacceptable
additional mission time to accomplish. Seven tasks were not Table 2 Tasks Rated Trainable and Not
able to be performed due to simulator system limitations and Trainable
were not evaluated. It was believed that all of are correctable
with improvements to the simulation system. However, the Fourteen tasks did not meet the criterion for various reasons.
tasks would still need to be evaluated for training capability. Tactical formation from fingertip, tactical formation above 500
feet, and tactical formation below 500 feet did not meet the
Of the remaining eighteen tasks, the Two-Channel Dome AOI criterion for the following reasons: Unrealistic head motion
visual system was evaluated as being capable of supporting was required to place the high resolution AOI onto the
operational training for four tasks. These four tasks received wingman to check his position. Pilots indicated that this is not
a rating of 3 or greater for at least 80% of the task ratings by a normal flying habit. Rather they indicated fighter pilots
the evaluation pilots. The four tasks that met the criterion normally move their eyes to pick up the target rather than
were: Task #11, terrain masking (89%); Task #14, visual make large head movements. Another frequently mentioned
target acquisition/identification (80%); Task #17, high problem in performing these tasks was the target aircraft
altitude weapon delivery (HAS, HD, DB, DTOS) (93%); and suddenly changed color to a bright green as distance increased
Task #19, target reattack (95%). Pilot comments indicated to approximately 1200 feet to 1500 feet.
areas of improvements for these tasks even though they were
rated acceptable (three or greater). Most frequently cited areas Combat descent did not meet the criterion due to difficulty in
of improvement were: (1) wider AOI inset area and determining altitude visually. Several pilots indicated the
background scene and (2) improved resolution in background ground appeared to be out of focus, however, buildings
scene outside the AOL The number of tasks rated trainable appeared to be in focus. Several indicated that other than the
and not trainable are shown in Table 2. horizon, there were no adequate indications they were
approaching the ground. Normal ground rush was not
provided in the visual simulation. Small objects appeared to
Two Channel Head Tracked Area of Interest Dome be blurry at longer ranges. Pilots also indicated accentuated
Tanker rendezvous NR head movements "to steer the AOI" were also required to
perform this task. Again, the large head movements were not
Tactical formation from fingertip ®
 4-5

considered natural and impacted normal tactical cross-check Wingman responsibilities did not meet the criterion due to the
pacing. exaggerated head movements required to place the AOI on
desired visual areas and the popping in of the high contrast
Single ship low level did not meet the criterion due to green aircraft between 1200 and 1500 feet.
exaggerated head movement to move the AOI to the desired
visual area, lack of adequate detail to maintain low altitude Situational awareness of tactical situations did not meet the
without reference to instruments, limited field of view and low criterion primarily due to the lack of other than visual cues.
resolution in the periphery, and lack of sufficient detail in data
base (particularly forested areas). Additional FOV and Resolution
Evaluation pilots indicated that they believed a wider field of
Visual low level navigate to initial point did not meet the view for both the inset AOI and the background scene was
criterion. Frequently cited problems in performing this task required. Most pilots reported that seeing the black edges of
were the exaggerated head movements required to move the the background scene was very distracting.
high resolution AOI inset and lack of resolution in the
peripheral (background) scene. Target Projector
The pilots indicated the target projector produced a green
Mutual support/lookout in various tactical formations did not image that was too bright. Pilots indicated that at longer
meet the criterion due to the exaggerated head movement to ranges, the contrast of a real world aircraft is not as good as
steer the AOI and the target projector providing an aircraft the current simulated aircraft.
image that was too easy to see.
Generic Fighter Cockpit
Detect visual threats did not meet the criterion because; (1) Evaluation pilots indicated the lack of a fighter cockpit and
the aircraft threats were too easy to see due to their bright HUD adversely impacted the results of this study. The generic
green color, (2) the surface to air missile (SAM) threats were fighter cockpit at Site 1 did not permit the normal instrument
not realistic due to visual display system resolution, lack of cross check. It did not have a weapons system radar or RWR.
smoke and fire trails, and (3) unrealistic head movements were Most importantly, it was missing a HUD.
required to detect threats.
Image Generator and Data Base
Individual/formation threat reactions did not meet the criterion (Size and optimization for Air to Surface) The pilots reported
due to a combination of factors. Unrealistic head movements the data base needed to have more cultural and geographic
were required to detect visual threats. Since there was no detail available. The pilots pointed out that at low altitude they
Radar Warning Receiver (RWR) indicator in the cockpit, this can see numerous features such as people, vegetation, trees,
additional piece of information was missing to assist the pilots vehicles, animals, etc.; and all of these objects give them a
to determine threat location and type. sense of altitude.

Tactical instruments cross check did not meet the criterion due Head-tracking
to lack of adequate instrumentation to perform the cross-check. Head-tracking forced the pilots to use exaggerated head
The cockpit did not have a HUD, RWR indicator, or radar movements to visually acquire a target or visual object. Head-
display. Without these systems, the task could not be tracking required the pilots to steer the AOI to the area they
realistically performed. wanted to see. This resulted not only in changes to normal
visual habit patterns but also caused them to fixate on the
Low altitude weapon delivery (LAS, LALD, LAB, VLD) did visual object longer than normal.
not meet the criterion because of the head movement required
to steer the high resolution AOI and the pilots indicated there Site 2: Display for Advanced Research and Training
was a noticeable lag in the visual system. The limited high (DART)/Compu-Scene IVA
resolution AOI forced unnatural scan patterns. The majority
of the pilots reported the resolution was adequate for target Evaluation Cockpit
identification and desired munitions point of impact (DMPI) The evaluation cockpit was an F-15C Mission Tactics Trainer
selection. However, most pilots indicated the lack of (MTT) developed by Armstrong Laboratory. The cockpit
peripheral cues and the limits in the high resolution AOI made replicated the physical and functional controls and displays of
scanning unnatural and detracted from weapons delivery. the F-15C weapon system. The cockpit provided an integrated
They reported having a less than full peripheral visual display weapons radar and fire control system and had a projected
made aircraft attitude and pop up attacks more difficult to HUD. Flight performance of the simulated aircraft was
evaluate. representative of the F- 15C weapon system. The cockpit was
situated inside the display.
Low altitude intercept did not meet the criterion primarily
because the visual phase of a low altitude intercept requires
radar work, switchology, geometry analysis, and threat
analysis using both visual and avionics systems.
4-6

Image Display display screen were often mistaken for air-to-air threats; and
The image display was a nine-channel mosaiced rear- (4) the on/off switching of the visual display channels located
projection flat screen wrap-around display that provided a total at the 6 O'clock position of the display needs to be improved
FOV of 360 degrees horizontal by 110 degrees vertical. so that it is less noticeable and distracting. Table 2 presents
the tasks rated trainable and not trainable.
Image Generation
A General Electric Compu-Scerie IVA furnished the image
generation for the evaluation. Display for Advanced Research and Training (DART)
Tanker rendezvous NR
Data Base Tactical formation from fingertip ®
The data bases used in the evaluation were representative of a
Tactical formation above 500 feet ®
portion of the state of Washington and Germany.
Combat descent •
Site 2: Mini-Display for Advanced Research and Training Tactical formation below 500 feet ®
(Mini-DART)/Compu-Scene IVA Single ship low level •
Visual low level navigate to initial point •
Evaluation Cockpit Mutual support/lookout in various tactical formations
The evaluation cockpit was an F-15C Mission Tactics Trainer
®
(MTT) developed by Armstrong Laboratory. See the Detect visual threats ®
description for the DART. Detect electronic threats •
Terrain Masking (direct/indirect) ®
Image Display Individual/formation threat reactions ®
The image display was an eight-channel mosaiced rear- Tactical instruments cross check
projection flat screen wrap-around display that provided a total
•
Visual target acquisition/identification ®
FOV of 360 degrees horizontal by 110 degrees vertical. This
display had a smaller footprint than the DART. Also, the rear Coordinated tactical attack ®
quadrant display was a single flat screen that was lowered into Low altitude weapon delivery (LAS, LALD, LAB.VLD) •
place after the pilot had entered the cockpit. High altitude weapon delivery (HAS, HD, DB, DTOS) ®
Reform after tactical attack ®
Image Generation
Target reattack •
Same as the DART.
Aircraft battle damage check NR
Data Base Low altitude intercept ®
Same as the DART. AIM-9 employment •
Low altitude air-to-air gun employment •
Discussion
Flight lead responsibilities ®
Twenty four tasks were evaluated in missions on the DART
and Mini-DART. Wingman responsibilities ®
Situational awareness of tactical situations •
Tanker rendezvous was not evaluated at Site 2 because the
® - Did Not Meet Criteria
evaluation team believed that this task was not supportable by 0 - Met Criteria
the simulation capability and is normally performed frequently
NR - Not Rated
during normal operational training missions and does not need
to be taught in the simulator. Aircraft battle damage check
was not performed because of the unavailability of an aircraft Table 2. Tasks Rated Trainable and Not
model to fly the task upon. This task requires close Trainable on the DART.
maneuvering in relation to another aircraft.
Tactical formation from fingertip, tactical formation above 500
Of the twenty-four remaining tasks, the DART visual system feet, and tactical formation below 500 feet did not meet the
was evaluated as being capable of supporting operational criterion for the following reasons: Pilots reported that they
training for ten tasks. These ten tasks received a rating of 3 or could not get close enough to fly reasonable fingertip
formation. The majority of the pilots reported there was not
greater for at least 80% of the task ratings by the evaluation
enough detail to fly fingertip formation. Several pilots
pilots.
reported the 10 O'clock and 2 O'clock display screens
Although the tasks shown were rated acceptable for training appeared a little blurry compared to the front display screen
capability, pilots comments indicated areas of improvements. and this contributed to the difficulty in accomplishing the task.
Most frequently cited areas were: (1) objects abruptly
Mutual support/lookout in various tactical formations and
appeared on the display; (2) pilots had to rely more upon the
detect visual threats did not meet the criterion. Pilots reported
radar altimeter and HUD to obtain altitude information; (3)
the resolution outside 3000 to 5000 feet was not adequate to
due to inadequate resolution of the display, blemishes on the
 4-7

perform the task without using more radio calls and other
sensor information not normally required in actual flight The differences in the tasks rated trainable and not trainable
conditions. are as follows. Task #6, Single ship low level was not rated
as trainable on the Mini-DART where it had been rated as
Terrain masking (direct/indirect) did not meet the criterion trainable on the DART. Task #16, Low altitude weapon
because pilots said there was insufficient scene detail to delivery (LAS, LALD, LAB, VLD) was rated not trainable on
provide them with altitude cues. the Mini-DART where it had been rated as trainable on the
DART. Pilot comments indicated that the close proximity of
Individual/formation threat reactions did not meet the the display screens in the Mini-DART may have been the
criterion. There was a lack of sufficient resolution outside the cause of lower ratings. Task #21, Low altitude intercept was
3000 to 5000 foot range. rated trainable on the Mini-DART but not trainable on the
DART (71 %) (Since the percentages do not differ by much
Visual target acquisition/identification did not meet the (80% vs 71%) this change may be a statistical anomaly
criterion. The lack of resolution outside 4000 feet made target resulting from slight rating shifts).
identification occur too late. Pilots said target identification
was satisfactory inside of this range. Resolution
Pilots indicated the resolution of the DART and Mini-DART
Coordinated tactical attack did not meet the criterion due to of was not adequate at ranges outside 3000 to 5000 feet. Lack of
the lack of resolution. Pilots said if one of the aircraft adequate resolution impacted their ratings on tasks involving
responded to a threat, then returning to tactical formation formation flight, threat identification, and target acquisition .
became a major problem due to the difficulty of reacquiring
wingman to compensate for a lack visual cues. Brightness
Pilots commented that the brightness for both the DART and
High altitude weapon delivery (HAS, HD, DB, DTOS) did not Mini-DART was very good.
meet the criterion. Targets appeared in the scene too late and
lacked sufficient detail. Image Generator and Data Base
Several of the pilot comments and their ratings reflected
Reform after tactical attack did not meet the criterion because limitations of the Compu-Scene IVA image generator and the
of the lack of adequate resolution outside the 3000 to 5000 data base. At the time of the evaluation, Site 2 was performing
foot range. Pilots said the visual cues were too poor for good an air-to-air evaluation and both the image generator and the
tactical formation rejoins. They reported that because of the data base had been optimized for air-to-air operations. The
lack of visual resolution, they had to talk to their wingman to lack of detail in some of the ground targets may have been due
get within visual range (3000 to 5000 feet). Pilots indicated to optimization of the data base for air-to-air operations
the visual display was not adequate to get back together
outside this range. Head-tracking
The majority of evaluation pilots reported that the head-
Low altitude intercept did not meet the criterion. The target tracking for the DART and Mini-DART was adequate.
aircraft did not show up at the expected distances with the However, several pilots commented the on/off switching of the
proper clarity. display scene for the aft display panel was noticeable and
distracting.
Flight lead responsibilities and wingman responsibilities did
not meet the criterion. Low resolution of the display forced Site 3: Fiber Optic Helmet Mounted Display/ESIG-1000.
the pilots to rely abnormally upon instrument cross checks to
analyze closure and range due to degraded visual cues. Pilots Evaluation Cockpit
reported lack of resolution and detail of the lead aircraft or The evaluation cockpit was an Operational Flight Trainer
wingman aircraft would not to permit realistic tactical (OFT) for the German Tornado Weapon System. It accurately
formation work. replicated the physical and functional controls, flight systems,
and flight characteristics of the Tornado aircraft. The cockpit
Results for the Mini-DART at Site 2 indicated the visual provided an integrated weapons radar and fire control system
system was evaluated as being capable of supporting training and had a HUD. Fully integrated motion simulation was
for eight of the twenty-four tasks (see Table 3-7). These eight provided by (1) a six-post 60 inch stroke platform motion
tasks received a rating of 3 or better for at least 80% of the base, (2) a g-seat, and (3) a g-suit.
task ratings by the evaluation pilots. For the most part, the
results on the Mini-DART parallel those of the DART since Image Display
they are relatively similar technology. The Mini-DART is The 3-channel fiber optic helmet mounted display (FOHMD)
slightly smaller and the rear display arrangement differs from was fitted onto the pilot's and WSO's head after they had
the DART in that a single display panel is used rather than entered the cockpit.
several display panels. There were some differences between
the trainable tasks on the DART and the tasks rated trainable
on the Mini-DART.
4-8

Image Generation High altitude weapon delivery (HAS, HD, DB, DTOS) •
An Evans and Sutherland ESIG-1000 provided the image Reform after tactical attack ®
generation for the evaluation. Target reattack •
Aircraft battle damage check •
Data Base
The data base used in the evaluation was an accurate Low altitude intercept ®
representation of south central Germany. AIM-9 employment •
Low altitude air-to-air gun employment •
Discussion - Pilots Flight lead responsibilities ®
Evaluation pilots flew twenty-four tasks from the list of
Wingman responsibilities ®
twenty-six tasks that were initially identified.
Situational awareness of tactical situations ®
Tanker rendezvous was eliminated by team consensus since ® - Did Not Meet Criteria
this task was not supportable by the simulation capability and Q - Met Criteria
is normally performed during normal operational fighter NR - Not Rated
training missions. Detect electronic threats was not evaluated
in order to avoid security classification issues with the Table 3. Tasks Rated Trainable and Not
Tornado radar and RWR system. Trainable for the FOHMD.

Of the remaining twenty-four tasks, the FOHMD visual system Fifteen tasks did not meet the criterion for various reasons.
was evaluated as being capable of supporting operational
training for nine tasks. These nine tasks received a rating of 3 Tactical formation from fingertip, tactical formation above 500
or greater for all task ratings by the evaluation pilots. Pilots feet, and tactical formation below 500 feet did not meet the
who evaluated the system commented very favorably on the criterion for the following reasons: Pilots indicated the area of
brightness of the display. Several areas of improvements were interest (AOI) was too small and the background imagery on
identified for the above tasks; (1) the ability to make head the display lacked enough resolution to keep track of wing
movement left or right and up to check high 12 o'clock is too men while avoiding the ground and attempting to perform an
restricted by the fiber optic cables; (2) instrument cross- instrument cross check.
checking in the cockpit was too time consuming due to the low
visibility of instruments within the cockpit; transitioning from Mutual support/lookout in various tactical formations did not
outside to inside the cockpit was very difficult particularly if meet the criterion for the following reasons: Pilots said the
the display information was on the attack radar display; (3) bulkiness and difficulty of moving the helmet due to
blemishes in the image display caused by broken fiber optic restrictions imposed by the fiber optic cables made it too
bundles were distracting since pilots had to differentiate difficult to check the 6 o'clock or high 12 o'clock position.
whether it was a threat aircraft or a blemish; and (4) the eye Broken fiber optic bundles also degraded the display. The
tracked area of interest needed to be larger. Table 3 presents pilots also reported that several times during their missions,
the tasks that were rated trainable and not trainable. the opposite eye from the direction that they were looking
would black out when they moved their eyes to the extreme
left, right, or up: this was extremely distracting and forced
Fiber Optic Helmet Mounted Display (FOHMD) them to change their normal visual scan technique.
Tanker rendezvous NR
Tactical formation from fingertip ® Detect visual threats, individual/formation threat reactions, and
Tactical formation above 500 feet ® visual target acquisition/identification did not meet the
criterion for the following reasons: Pilot comments indicated
Combat descent • it was almost impossible to see a threat if it was not in the high
Tactical formation below 500 feet ® resolution AOL Visual pick-up of the threat at long ranges
Single ship low level • was too difficult.
Visual low level navigate to initial point •
Mutual support/lookout in various tactical formations ® Terrain masking (direct/indirect) did not meet the criterion for
the following reasons: Pilot interviews indicated the texturing
Detect visual threats ® of the ground needed improvement to provide better low
Detect electronic threats NR
altitude cues (this problem is a limitation of the image
Terrain Masking (direct/indirect) ® generator but manifests itself in this task).
Individual/formation threat reactions ®
Tactical instruments cross check ® Tactical instruments cross check did not meet the criterion for
Visual target acquisition/identification
the following reasons: Pilots indicated they could not easily
® look through the helmet mounted display optics into the
Coordinated tactical attack ® cockpit and quickly gather the needed information.
Low altitude weapon delivery (LAS, LALD, LAB.VLD) •
 4-9

Coordinated tactical attack and reform after tactical attack did Coordinated tactical attack f
not meet the criterion for the following reasons: Pilots Low altitude weapon delivery (Radar/EO) •
reported that their visual search patterns were significantly
High altitude weapon delivery (Radar/EO) •
changed; they had to look directly at their wingman in order to
see them whereas in actual flight they would use their Reform after tactical attack ®
peripheral vision more. Target reattack •
Aircraft battle damage check •
Low altitude intercept did not meet the criterion: Pilots Low altitude intercept ®
reported pure visual intercepts were generally not possible due
Situational awareness of tactical situations ®
to short range visual pick-up.
Direct tactical formation ®
Flight lead responsibilities did not meet the criterion for the Direct individual/formation threat reactions ®
following reasons. Pilots indicated that their visual search Direct target attack/reattack •
patterns were changed due to the lack of resolution in the Direct reform ®
background image of the FOHMD. To confirm the wing man's Direct egress
position, pilots reported they had to stare at them with the AOI •
longer than normal. ® - Did Not Meet Criteria
0 - Met Criteria
Wingman responsibilities did not meet the criterion for the NR - Not Rated
following reasons: Pilots stated lack of detail of the flight
lead at ranges greater than 10,000 feet adversely affected their Table 4. Tasks Rated Trainable and Not
rating. Trainable for the FOHMD by WSOs.

Situational awareness of tactical situations did not meet the Ten tasks did not meet the criterion for various reasons.
criterion for the following reasons: The majority of evaluation
pilots reported that they were unable to maintain situation Mutual support/lookout in various tactical formations did not
awareness adequately using the FOHMD due to difficulty in meet the criterion for the following reasons: The WSOs
moving the headgear to check 6 o'clock and the broken fiber indicated that they were unable to perform this task adequately
optic bundles that forced them to take more time to since they were unable to check their six o'clock position for
differentiate between threat aircraft and blemishes. threats. The difficulty of moving the FOHMD due to the
routing of the fiber optic cables was the primary problem.
Discussions - WSOs
Twenty-two tasks were identified for possible evaluation by Detect visual threats did not meet the criterion for the
the WSO evaluators. Of these tasks, twenty-one tasks were following reasons: WSOs indicated that ratings of this task
able to be evaluated during the WSO missions using the were impacted by the difficulty of moving the FOHMD due to
FOHMD. the routing of the fiber optic cables. They indicated that their
normal scan patterns were impacted; they could not check
Detect electronic threats was not evaluated. Lack of time to high 6 o'clock due to the visual system limits.
train WSOs to proficiency in these systems and desire to avoid
security classification issues precluded meaningful evaluation Tactical instruments cross check did not meet the criterion for
of the task. the following reasons: This task was difficult to perform due
to the difficulty of looking through the FOHMD into the
Of the remaining twenty-one tasks, the FOHMD visual system cockpit. Even with back lighting, the instruments were
was evaluated as being capable of supporting operational difficult to see and quickly get information.
WSO training for eleven tasks. These eleven tasks received a
rating of 3 or greater for all ratings of the task by the WSOs. Visual target acquisition/identification did not meet the
See Table 4. criterion for the following reason: The visual target could not
be acquired and identified at normal ranges. WSOs had to
FOHMD - WSO Tasks wait until the visual target was closer than usual and this was
Single ship low level £ not acceptable.
Visual low level navigate to initial point •
Reform after tactical attack did not meet the criterion for the
Mutual support/lookout in various tactical ® following reasons: WSOs stated that they were unable to see
Intraflight coordination/communication • wingman at normal reform ranges.
Intercockpit coordination/communication •
Detect visual threats ® Low altitude intercept did not meet the criterion for the
Detect electronic threats NR following reasons: WSOs indicated that there was not enough
detail and contrast for threat aircraft to give early acquisition
Tactical instruments cross check ® and identification. This made the task of intercept difficult to
Visual target acquisition/identification ® perform.
4-10

Situational awareness of tactical situations did not meet the Head Movement Limitations
criterion for the following reasons: WSOs stated that they A major negative criticism of the display concerned the
rated this task low because they were unable to spot bandits restrictions to normal head movement with the FOHMD. The
(threat aircraft) at normal ranges. routing of the fiber optic bundles restricted head movement
and required extra effort on the part of the user when looking
Direct tactical formation did not meet the criterion for the aft of the three-nine line. Due to their size and location, pilots
following reasons: WSOs indicated that they had to work too reported difficulty in checking 6 o'clock and at the high 12
hard to keep visual contact with flight leads or wingman. o'clock position. A subsequest visit to Neuberg AB, GE
Much of this they said was due to the low contrast of the visual indicated that these restrictions has been removed through re-
image of other formation aircraft in the high resolution AOL routing the fiber optic cables.

Direct individual/formation threat reactions did not meet the Fiber Optic Bundle Wear
criterion for the following reasons: Both WSOs indicated that Over the period of the evaluation, movement of the fiber optic
they were unable to scan an area visually as rapidly as they cables resulted in breakage of the individual fiber optic
normally do in actual flight. This was due to the low contrast bundles. The result was a darkening of the displayed image.
of aircraft against the sky. This made it much more difficult to
detect and then identify an aircraft. Motion Simulation
The integrated motion systems at Site 3 were representative of
Direct reform did not meet the criterion for the following the latest state of the art in motion systems. Pilots and WSOs
reason: Both WSOs indicated that the low visual contrast of commented that the combination of the six-degree of freedom
the lead aircraft against the sky made it too difficult to platform motion system, g-seat, and g-suit provided very
determine which way lead was heading. realistic motion cueing. The consensus was that the motion
simulation was outstanding; however its impact to training is
Resolution unknown.
Pilots and WSOs indicated that the resolution of the AOI was
very good but that the resolution in the area outside the AOI 5. CONCLUSIONS
needed improvement. Throughout the visual evaluations, the importance of area of
coverage, scene content, and scene detail of the visual
Brightness presentations for operational training continued to surface.
Pilots commented very favorably upon the brightness of the The evaluation team believes that realistic low altitude
display. However, display brightness was degraded as fiber operational fighter training requires large visual data bases,
optic bundles were broken during the evaluation. with realistic threat modeling, high scene content, and high
scene detail.
Tornado Fighter Cockpit
(Transition to a wing-sweep aircraft) Pilots and WSOs did not To accommodate realistic threat modeling, scene content, and
have any problem transitioning to flying the Tornado fighter accurate data bases requires a high end image generator.
simulator. Although none of the pilots had flown an aircraft Image generators presenting data bases for tactical training
with adjustable wing sweep, the transition to the Tornado was must be able to rapidly process large quantities of data in a
accomplished with minimum difficulty. timely manner so that the aircrew member can realistically see
it.
Image Generator and Data Base
The Pilots and WSOs liked the Germany data base. The The team believes that evaluations of this type should be
majority of the evaluation aircrews had flown over similar accomplished using senior instructor pilots or WSOs. Highly
areas of Germany during deployments or while stationed in experienced instructor pilots and instructor WSOs are
Europe. experienced in teaching flying principles to new students as
well as experienced aircrew members returning to flying.
Head-tracking and Eye Tracking They know what it takes to train the full range of students and
The majority of the evaluation missions were flown using eye- can comprehensively express training issues to other
tracked AOL Pilots liked this feature when it was used. They disciplines so that they are understood.
commented that the AOI needs to be made larger and there
needs to be better resolution in the background channel of the Overall, the evaluation team believed that texturing in the
visual display. displayed image can provide helpful cues for low altitude
flight. However, a single level of texturing does not provide
Eye Tracking and Calibration Process adequate cues required by pilots for low altitude flight using
The procedures needed to provide eye tracking required that visual information. Texturing should become sharper in focus
each pilot and WSO evaluator have a special custom fitted as the range from the textured object decreases. Multiple
helmet liner. It also required a special eye tracking alignment levels of texturing are a possible solution and should be
process be conducted in the cockpit after the pilot or WSO evaluated.
had strapped in.
 4-11

Based upon the three evaluations, the evaluation team believes

that single ship air-to-ground tasks are trainable now with the
right combination of database, image generator, and matched
visual display system.

The evaluation team does not believe that current systems

provide adequate resolution, contrast, and brightness to allow
for dynamic air model discrimination. Dynamic air models
refer to anything (another aircraft, SAM, air-to-air missile,
flack, bullets, etc.) in the airspace that the pilot or WSO must
assess it's attitude, range, etc.

REFERENCES
1. Hutton, D.P., Capt, Burke, D.K., Capt., Englehart, J.D.,
Wilson, J.M. Jr, Romaglia, F.J., & Schneider, AJ. (1976).
Project 2235, Air-to-Ground Visual Simulation
Demonstration: Final Report. Volumes 1 & 2, Aeronautical
Systems Division, Simulator SPO, Wright-Patterson AFB, OH.

2. Rivers, H.A. LtCol. (1977). Final Report: Simulator

Comparative Evaluation. Tactical Air Command, USAF
Tactical Air Warfare Center, Eglin AFB, FL.

3. Konopatzke, D. L., Capt. & Van Arsdall, R. S. (1979).

Final Report: Follow-On Simulator Comparative Evaluation.
Tactical Air Command, USAF Tactical Air Warfare Center,
Eglin AFB, FL.

4. O'Neal, M.E., III, LtCol. (1984). F-15 Limited Field of

View Visual System Trainine Effectiveness Evaluation:
Special Project. Final Report. Tactical Air Command, USAF
Tactical Air Warfare Center, Eglin AFB, FL.
 5-1

Yaw Motion Cues in Helicopter Simulation

Jeffery A. Schroeder
Walter W. Johnson
NASA Ames Research Center
Mail Stop 262-3
Moffett Field, California, USA 94035-1000

1. SUMMARY 5r cockpit pedal position, in

A piloted simulation that examined the effects of yaw motion motion filter damping ratio, nondimensional
C
cues on pilot-vehicle performance, pilot workload, and pilot V vehicle model yaw angle, rad
motion perception was conducted on the NASA Ames vehicle model yaw angle rate, rad/sec
Vertical Motion Simulator. The vehicle model that was used
V
represented an AH-64 helicopter. Three tasks were performed v vehicle model yaw acceleration, rad/sec2
in which only combinations of vehicle yaw and vertical dis- Vcom motion system yaw acceleration command, rad/sec'
placement were allowed. The commands issued to the motion Vp yaw acceleration at pilot station, rad/sec2
platform were modified to present the following four motion
Ysens pilot sensed yaw acceleration, rad/sec2
configurations for a pilot located forward of the center of rota-
tion: 1) only the linear translations, 2) only the angular rota- Vsim motion system yaw acceleration, rad/sec2
CO frequency, rad/sec
tion, 3) both the linear translations and the angular rotation,
and 4) no motion. The objective data indicated that pilot-ve- com motion filter natural frequency, rad/sec
hicle performance was reduced and the necessary control ac-
tivity increased when linear motion was removed; however, 3. INTRODUCTION
The question of whether or not a yaw rotational motion de-
the lack of angular rotation did not result in a measured
gree-of-freedom is necessary in flight simulation is important
degradation for almost all cases. Also, pilots provided sub-
to government regulators, since they specify the simulator
jective assessments of their compensation required, the mo- motions necessary for desired levels of simulator sophistica-
tion fidelity, and their judgment of whether or not linear or tion (Ref. 1). It is also important to simulator designers and
rotational cockpit motion was present. Ratings of compensa- operators, who build and/or configure how the motion system
tion and fidelity were affected only by linear acceleration, and actuator displacements should be apportioned into the motion
the rotational motion had no significant impact. Also, when degrees-of-freedom.
only linear motion was present, pilots typically reported the
presence of rotation. Thus, linear acceleration cues, not yaw Currently, the Federal Aviation Administration specifies two
rotational cues, appear necessary to simulate hovering flight. levels of motion fidelity for helicopter flight simulators: full
6-degrees-of-freedom motion, and 3-degrees-of-freedom mo-
tion (Ref. 1). For the latter, the assumed three degrees of
2. LIST OF SYMBOLS
freedom are pitch, roll, and vertical. If degrees of freedom
longitudinal acceleration at pilot station, ft/sec2 different from these three are selected, these new selections
a must be qualified on a case-by-case basis. The selection of
v lateral acceleration at pilot station, ft/sec2 the pitch, roll, and vertical axes is reasonable and based upon
cm. vehicle center of mass experience, but evidence to support these axes or any set of
F(x,y) F ratio with x numerator degrees-of-freedom and y axes is lacking.
denominator degrees-of-freedom
gravitational acceleration, ft/sec2 Differences of opinion exist on the relative importance of ro-
h altitude of cm., ft tational versus linear motion cues in the selection of critical
h vertical acceleration of cm., ft/sec2 axes. For example, the abstract of Reference 2 states that, for
n tracking, "translational motion cues appear to be generally
con motion system vertical acceleration command,
less important than rotational ones, although linear motion can
ft/sec2 be significant in special situations." In contrast, the conclud-
"sim motion system vertical acceleration, ft/sec2 ing remarks of Reference 3 state, "For large aircraft, due to
K motion filter gain, nondimensional size and to the basic nature of their maneuvering dynamics,
Ky the cockpit lateral acceleration cues appear to be much more
vestibular model gain important than roll acceleration cues. There was the indica-
rms root mean square tion that this observation might be extended to the generaliza-
s complex variable, 1/sec tion that, in each plane of motion, the linear cues are much
motion system longitudinal acceleration command, more valuable than the rotational cues."
ft/sec2
xp Most flight simulation studies focusing on the interactions be-
pilot longitudinal displacement from cm., ft
2
tween rotational and linear acceleration cues have been per-
*sim motion system longitudinal acceleration, ft/sec formed in the roll axis (Ref. 4). Roll-axis interactions are es-
Ycom motion system lateral acceleration command, ft/sec2 pecially important for simulating turn coordination. For the
Ysens pilot sensed lateral acceleration, ft/sec2 yaw axis, work examining the usefulness of the associated
cues has been sparse and not conclusive. Meiry, in the first
Vsim motion system lateral acceleration, ft/sec2 detailed investigation into the effects of yaw motion, found
Z[ state for dynamic inflow approximation that its addition was shown to be beneficial, as the study indi-
°c cockpit collective position, in cated a reduction in operator-time-delay of 100 msec, with a
concomitant improvement in performance (Ref. 5).

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?"
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
 5-2

erations were typically 0.05 g's, which are more than double
In contrast, two recent simulation studies that examined a pi- the above maximum threshold of perception.
lot's ability to perform hovering flight tasks with a representa-
tive vehicle model found little or no effect of yaw on pilot The tasks discussed later will illustrate that the rotational and
performance or opinion (Refs. 6 and 7). However, in those lateral accelerations that were generated were well above the
studies, the pilots were intentionally located at the vehicle's published thresholds and at frequencies within the bandwidth
center of rotation to study the effects of yaw motion only; of the dynamic models in eqns. 1 and 2.
thus, the typical accompanying lateral and longitudinal accel-
erations were absent by design. Those results suggested that 4. EXPERIMENT DESCRIPTION
simulator yaw motion may not be necessary for the tasks that
simulate hovering flight. 4.1 Tasks
NASA test pilots flew three tasks that represented a broad
The purpose of the study described here is to extend this pre- class of situations in which both linear and rotational motion
vious work and determine what components of motion are cues may be useful in flight simulation. Task 1 was a small-
important to include in flight simulation. Specifically, linear amplitude command task that allowed for full math model
accelerations induced by yaw rotations, when the pilot is not motion to be provided by the motion system in the horizontal
at the center of rotation, were physically introduced to deter- plane. Task 2 was a large-amplitude command task that did
mine if the results from References 6 and 7 may be general- not allow full math-model motion to be presented (simulator
ized into a hypothesis that yaw rotational simulator cab mo- cab rotational and longitudinal translational limits would have
tion has no significant effect. If so, then a savings might re- been exceeded), but was accompanied by strong rotational vi-
sult from a reduced level of complexity required in the design, sual cues. Task 3 was a disturbance rejection task, which also
development, and operation of flight simulators. allowed full math-model motion to be provided by the motion
system, but with the pilot also controlling vehicle altitude.
This paper is organized as follows. First, a brief discussion of
human motion sensing characteristics is given in order to Task 1:15-deg yaw capture. For the first task, the pilot con-
place the magnitude of the accelerations provided in this ex- trolled the vehicle only in the yaw axis and was required to
periment in context. Second, the tasks and the experimental rapidly acquire a north heading from 15-deg yaw offsets to
apparatus are described. Then, the measured pilot-vehicle either the east or west. This task allowed for full math-model
performance, pilot control activity, pilot compensation, mo- motion to be represented by the motion system in all axes
tion fidelity, and motion reports, are presented and discussed. (rotational, lateral, and longitudinal). An aircraft plan view,
with the pilot's simulated position relative to the fixed cm., is
3.1 Human Lateral-Directional Motion Sensing shown in Fig. 1. The desired pilot-vehicle performance for
Perception of motion arises from a complex, and incompletely the task was to rapidly capture and stay within +/-1 deg about
understood, interaction of visual, tactile, and vestibular stim- north with two overshoots or less. This 2 deg range was visu-
uli. Of the two non-visual stimuli, the vestibular system has ally demarcated by the sides of a vertical pole, shown in his
received the most attention. forward field-of-view in Fig. 2. Pilots performed six total
captures for each configuration, alternating between initial
In an extensive review, Peters (Ref. 8) found the following west and east directions. The repositionings from north to the
model, taken from Meiry (Ref. 2), to be the most reasonable initial east or west directions were not part of the task.
characterization of the subjective perception of rotational ac-
celeration about a vertical axis due to an excitation of the Task 2:180-hover turn. The second task was a 180-deg pedal
semi-circular canals in the inner ear: turn over a runway performed in 10 sec. The pilot again con-
trolled the aircraft with the pedals only, and the position of the
10s cm. remained fixed. This maneuver was taken from the cur-
(1) rent U.S. Army rotary-wing design standard (Ref. 9) and, with
Vs (s + 0.1)(s + 10)
one proviso, is representative of a handling qualities maneu-
ver performed for acceptance of military helicopters. In the
Rotational acceleration thresholds of perception were also ex- military acceptance maneuver, the pilot controls all six de-
amined extensively in Reference 8. Of the 14 studies re- grees of freedom rather than one. This maneuver did not al-
viewed, a vertical axis rotational acceleration threshold range low full math-model motion, since the simulator cab cannot
between 0.035 and 2.0 deg/sec2 was found. Given those rotate 180 degs. As a result, attenuated motion was used (see
thresholds, pilots should have perceived the rotational accel- Section 4.4). Desired performance was to stabilize the com-
erations in the study described herein, as the maximum rota- plete turn within +/- 3 degs and within 10 sees. Pilots per-
tional accelerations were typically 20 deg/sec2, thus exceed- formed six 180-deg turns, always turning over the same side
ing the maximum threshold above by an order of magnitude. of the runway to keep the visual scene consistent for the set of
turns. Figure 3 shows the visual scene from the starting posi-
Dynamics of the utricles in the inner ear, which sense linear tion.
acceleration, are given in Reference 2 as:
Task 3: yaw regulation The third task required the pilot to
yiCvs perform a rapid 9-ft climb while attempting to maintain a
sens y constant heading. This disturbance rejection task was chal-
(2)
ysim (s + 0.1)(s + 1.5) lenging, because collective lever movement in the unaug-
mented AH-64 model results in a substantial yawing moment
The cutoff frequency of 1.5 rad/sec suggests that high-fre- disturbance (due to engine torque) that must be countered
quency accelerations must be sensed by the tactile mecha- (rejected) by the pilot with pedal inputs. This task allowed
nisms in the body and not the vestibular system. Reference 8 full vertical and yaw motion, and the pilot now had to control
reviewed horizontal acceleration thresholds from seven these two axes simultaneously. Desired performance was for
sources, and found that they ranged between 0.002 and 0.023 the pilot to acquire the new height as quickly as possible
g's. In the study described herein, the maximum lateral accel- while keeping the heading within +/-1 deg north. The same
 5-3

visual scene was presented as in Task 1 (Fig. 2), but with the
scene also indicating height variations as the vehicle model ^™(s)«e-°J4! (10)
changed altitude.
The Evans and Sutherland CT5 A visual system was used to
4.2 Vehicle Dynamics
The math model represented an unaugmented AH-64 Apache provide the out-the-window visual cues. The visual field-of-
helicopter in hover, which had been identified from flight test view is shown in Fig. 5, and it had a math-model-to-visual-
data and subsequently validated by several AH-64 pilots (Ref. image-generation delay of 86 msec (Ref. 14). The visual cues
10). Equation 3 provides the vehicle dynamics for the yaw presented to the pilot did not vary with the motion filter con-
and vertical degrees-of-freedom. figurations (see Section 4.4) and were always those of the
math model. These cues represented his physical offset of 4.5
ft forward of the simulated vehicle cm.
-0.270 0.000 o.ooo" v"
0.000 -0.122 -118. h Conventional pedals and a left-hand collective lever were
used. The pedals had a travel of +/- 2.7 in, a breakout force of
0.000 0.000 -12.9 Zl 3.0 lb, a force gradient of 3 lb/in, and a damping ratio of 0.5.
(3)
The collective had a travel of +/- 5 in, had no force gradient,
0.494 0.266
and the friction was adjustable by the pilot.
0.000 14.6
0.000 1.000 All cockpit instruments were disabled, which made the visual
scene and motion system cues the only primary cues available
All other vehicle states were kinematically related to the to the pilot. Rotor and transmission noises were present to
above dynamics. So, in effect, the vehicle cm. was con- mask the motion system noise. Six NASA Ames test pilots
strained to remain on a vertical axis fixed in space for all participated in Task 1, and five of the same six participated in
tasks. While the tail rotor in an actual helicopter causes both tasks 2 and 3. All pilots had extensive rotorcraft flight and
a side force and a moment about the cm., only the moment simulation experience.
was represented in this experiment, due to the fixed cm.
These vehicle constraints were introduced to simplify the 4.4 Motion System Configurations
number of motion sensations that had to be interpreted by the Four motion system configurations were examined for each
pilot. No coordination of the gravity vector was required, as it task: 1) rotational and linear motion, 2) linear without rota-
remained fixed relative to the pilot. No atmospheric turbu- tional motion, 3) rotational without linear motion, and 4) no
lence was present in any of the tasks. The collective lever motion. Figure 6 illustrates, in a plan view, the simulator cab
was used for Task 3 only. motion for these configurations for Task 1 (+/- 15 deg heading
turns). In the Linear+Rotation case, the cab rotates and trans-
The pilot was located 4.5 ft forward of the simulated cm., lates as if it were placed on the end of a 4.5 ft vector rotating
which is representative of the AH-64 pilot location. Thus for in the horizontal plane. In the Linear case, the pilot always
this case, math model rotational accelerations are accompa- points in the same direction, as the cab translates in x and y.
nied by lateral accelerations at the pilot station, and rotational In the Rotation case, the cab rotates but does not translate.
rates are accompanied by longitudinal accelerations at the pi- Finally, in the Motionless case, the cab does not move.
lot station. That is, the accelerations at the pilot station in this
When either linear motion or rotational motion was present
experiment are:
for Tasks 1 and 3, it was the full motion calculated by the ve-
aXp=-xpV■2 (4) hicle math model. That is, the cockpit provided the full accel-
a (5) erations that the math model calculated and that the visual
yP=xpV scene provided (along with the effective motion delays in
VP=V (6) eqns 7-10). This statement was true, except for the longitudi-
nal motion provided by the Linear motion configuration. This
43 Simulator And Cockpit was because for yaw turns about a point, the longitudinal ac-
The NASA Ames Vertical Motion Simulator (VMS) was used celeration at the pilot's station is always negative (centripetal
for this experiment (Ref. 11). The mainframe-computer cycle acceleration in eqn. 4). These accelerations, if integrated
time was 25 msec A cutaway view of the motion system and twice to motion system position commands, would cause
the experiment's software-command position, velocity, and continual longitudinal cab movement aft for this motion con-
acceleration limits are shown in Fig. 4. Within the listed lim- figuration, which results in the simulator cab eventually ex-
its, the dynamic performance of the simulator depends on the ceeding its available longitudinal displacement. Thus, a sec-
axis. Using frequency response testing techniques (Ref. 12), ond order, high-pass filter was used in the longitudinal axis so
the dynamics of the yaw rotational, longitudinal, lateral, and that the cab would return to its initial position in the steady
vertical axes were fitted with equivalent time delays (that is, state. This type filter is typically used in flight simulation,
the phase response was approximated as a pure time delay) as and it had the form of:
shown below (eqns 8 and 9 are taken from Reference 13): Ks2
-(s) = 2 (11)
¥s -(s) = „-0.13s (7) s +2Ccoms + <
Vc
As described earlier, Task 2 did not allow full motion. Thus,
-(s) = e -0.17s a high-pass filter of the same form in Eqn. 11 was used in all
(8)
axes. The values of K and com were empirically selected to
use as much cockpit motion as available (Fig. 4). For Task 3,
-(s)-e- (9) the vertical motion was always the full math model vertical
motion, even in the "Motionless" condition. That is, "Motion-
less" for Task 3 refers to the simulator cab being motionless in
 5-4

the horizontal plane. Table 1 lists K and com for each tested marginally significant here (F(l,4)=5.58, p=0.077). The ef-
configuration in each axis (a configuration with K=l and fects of rotational and linear motion did not interact in this
measure (i.e., were statistically independent).
com=1.0E-5 rad/sec effectively makes the filter in equation 11
unity for the tasks). The filter damping ratio (£) was 0.7 for Control rate. Figure 10 illustrates the nns cockpit control
all configurations. (pedal) rate for the four configurations. Often, this measure is
associated with pilot workload, with more control rate being
4.5 Procedure generally indicative of more pilot lead compensation required
Pilots were asked to rate the overall level of compensation re- for a given task. The analysis of variance for these data
quired for a task using the following descriptors: not-a-factor, shows when linear motion is added that the decrease in pedal
minimal, moderate, considerable, extensive, and maximum- rate is statistically significant (F( 1,4)= 18.53, p=0.013). No
tolerable. For analysis, these adjectives were given interval significant differences were noted when rotational motion was
numerical values from 0 to 5. Next, the pilots rated the mo- added, and rotational and linear motion effects did not inter-
tion fidelity according to the following three categories: 1) act.
Low Fidelity - motion cueing differences from actual flight
were noticeable and objectionable, 2) Medium Fidelity - mo- Compensation ratings. Figure 11 shows the means and stan-
tion cueing differences from actual flight were perceptible, dard deviations of the compensation required, as rated by the
but not objectionable, and 3) High Fidelity - motion cues were pilots, for the four motion conditions. When linear motion
close to those of actual flight. These definitions were taken was added, the compensation required significantly decreased
and slightly modified from Reference 15. These subjective from considerable to moderate compensation (F(l,5)=6.83,
ratings were given numerical values 0 to 2, respectively. p=0.047) while no significant differences were found for the
Third, pilots were asked to report whether or not they felt any addition of rotational motion. Rotational and linear motion
motion in the lateral or rotational axis, or both. A zero was did not interact for pilot compensation. These subjective pilot
assigned if they did not feel motion in a given axis, and a one opinions are consistent with the measured control rate differ-
was assigned if they felt motion. Each of the four configura- ences just discussed. That is, the addition of linear motion re-
tions was flown four times in a random sequence by each pi- duced needed control activity, which in turn is generally re-
lot. lated to the pilot compensation required.

5. RESULTS Fidelity ratings. Similar results occurred for the pilots' rating
The experimental design above is a two-factor fully within of motion fidelity, as shown in Fig. 12. When linear motion
subjects factorial experiment (Ref. 16), with the two indepen- was added, the motion fidelity rating improved (F(l,5)=7.74,
dent variables being linear and rotational motion. The combi- p=0.039). The fidelity increased from low-to-medium to
nation of the two levels (motion present and motion absent) medium-to-high, on average. Again, no reliable differences
within each independent variable results in four configura- were noted for the addition of rotational motion (the mean
tions for each task. An analysis of variance on the data taken improvement shown when rotational motion was added with-
for each task was performed, with the observed significance out linear motion was not statistically significant). Rotational
levels (p-values) given below. The quantity F(x,y) is the ratio and linear motion did not interact in the fidelity ratings.
of the data variations between configurations to the variations
within the configurations. The p-values represent the proba- Lateral motion reports. Figure 13 depicts the mean percentage
bility of making an error in stating that a difference exists of the time pilots reported lateral motion to be present for the
based on the experimental results, when no difference actually four motion configurations. Statistically, the two factors of
exists. Typically, differences are deemed significant for rotational and linear motion interacted (F(l,5)=30.6,
p<0.05 (5 chances in 100 of making an error). p=0.003). Lateral motion was reported an average of 85% of
the time when it was present, and the addition of rotational
5.1 Task 1: 15-deg Yaw Capture motion did not increase lateral motion reports (actually, it de-
creased lateral motion reports from 91% to 79%). On the
Full motion versus no-motion. Figure 7 shows a representa- other hand, while lateral motion was never reported in the no
tive time history of several key variables for Task 1 for the motion condition, it was reported nearly 50% of the time
Linear+Rotation motion condition. Peak yaw rates (not when only rotational motion was present.
shown) for this run were consistently on the order of 10
deg/sec. Figure 8 shows the same variables for the same pilot, The influences of purely rotational cues on lateral motion re-
but with the Motionless configuration. When compared to the ports could be due to the pilots sensing some true lateral ac-
full-motion case in Fig. 7, this motionless case results in more celeration. The pilot's design eye point was less than 0.5 ft
yaw overshoots, higher math model accelerations, and larger forward of the motion system's rotation point. It is possible
control inputs. These general degradations are typical when that, depending on the variation in pilots' posture, this small
all motion is removed and are consistent with previous data offset may have resulted in their vestibular system registering
taken for the vertical axis (Ref. 7). a linear acceleration. The maximum rotational accelerations
for the likely worst case (0.5 ft offset and a 20 deg/sec^ yaw
Pilot-vehicle stability Figure 9 depicts, for the four motion accelerations, see Fig. 7) results in a 0.005-g linear accelera-
conditions, the means and standard deviations of the number tion. This acceleration is small but perhaps just within a pi-
of times pilots overshot the +/- 1 deg heading point about lot's threshold (see Section 3.1).
north. For instance, when no rotational and no linear motion
was present (Motionless), the mean number of overshoots Rotational motion reports. Pilot reporting of rotational mo-
outside the +/-1 deg point was 11 per run across the pilots. tion, shown in Fig. 14, was also affected by an interaction
This measure is generally indicative of the level of damping between actual rotational and linear motion (F(l,5)=10.4,
(or stability) in the pilot-vehicle system. The analysis of vari- p=0.023). Rotational motion was reported 30% of the time
ance for these results show that when linear motion was when no motion was present at all. The reporting of rotational
added, the decrease in the number of overshoots was statisti- motion increased dramatically to 87% when any motion, rota-
cally significant (F(L4)=9.16, p=0.039). The decrease in tional, linear, or both was given. Apparently, when combined
overshoots with the addition of rotational motion was
 5-5

with visual cues, the linear motion enhances the onset of vec-
tion. Lateral motion reports. Figure 21 illustrates the percentage of
time pilots reported lateral motion being present. There were
To summarize the results for this task, linear motion was significantly more reports of lateral motion when it was pre-
clearly the most important motion variable. Linear motion sent (70%), than when it was not (5%) (F(l,4)=14.8,
improved pilot-vehicle performance, lowered control activity, p=0.018). There was no significant impact of rotational mo-
lowered pÜot compensation, improved pilot impression of tion on lateral motion reports, nor was there any significant
motion fidelity, and caused pilots to believe that rotational rotational motion and lateral motion interaction (unlike Task
motion was present when it was not. Also, from the accelera- !)•
tion magnitudes involved, it is likely that it was the lateral
component of the two linear motions that was effective. The Rotational motion reports. In the reporting of rotational mo-
addition of rotational motion showed no statistically signifi- tion, the rotational and linear motion factors interacted (Fig.
cant improvement, with the exception of its effect on reporting 22) (F(l,4)=20.0, p=0.011). Rotational motion was reported
when lateral motion is present, as possibly explained earlier. 90% of the time when linear motion was present, both when
rotational motion was actually present and when it was absent.
5.2 Task 2: 180-deg Hover Turn Only when linear motion was absent did the presence of rota-
tional motion lead to increased reports of rotational motion.
Full motion versus no-motion. Figure 15 is a representative When no motion was presented, rotational reports occurred
time history of several key variables for the Linear+Rotation 27% of the time on average, but increased to 65% of the time
motion condition in Task 2. Peak math model and visual yaw when only an actual rotational motion was added.
rates for this turn were consistently on the order of 50 deg/sec
(not shown). These rates were, of course, attenuated by the To summarize the results for this task, linear (and therefore
motion system (Table 1) so that it remained within its dis- lateral) motion is again the key motion variable. Its addition
placement constraints. Figure 16 shows the same variables reduced control activity, improved motion fidelity, and led to
for the same pilot for the Motionless configuration. An in- the belief that rotational motion was also present when it was
crease in yaw overshoots is noted, which is evident in the not present. These results are similar to those of Task 1, ex-
displacements, rates, accelerations, and control inputs. These cept for the interesting result that the addition of rotational
trends are consistent with Task 1. motion degraded performance slightly in Task 2.

Pilot-vehicle stability. Figure 17 depicts, for the four motion 53 Task 3: Yaw Regulation
conditions, the means and standard deviations of number of
times pilots overshot the +/- 3 deg heading point about the Full motion versus no-motion. Figure 23 depicts key variables
runway centerline during the 180 deg turns. These results in a sample run for the Linear+Rotation condition in Task 3.
show that when linear motion was added, the decrease in the The peak yaw rate for this run was 7.5 deg/sec (not shown).
number of overshoots was marginally significant statistically The peak yaw accelerations for this task were similar to those
(F(l,4>=5.40, p=0.081). Interestingly, in this case, the addi- of Task 1, but the rms accelerations were slightly higher in
tion of rotational motion made the performance worse, and Task 3 than Task 1 (5.67 deg/sec2 versus 4.21 deg/sec2, re-
this result was statistically significant (F( 1,4)= 13.26, spectively). The amount of visual rotation was less in this dis-
p=0.022). Rotational and linear motion did not interact in this turbance rejection task than in the command task of Task 1.
measure. These results are not easily explained; however, it Figure 24 illustrates the same variables for the same pilot for
must be remembered that this task did not have full math the Motionless configuration Slightly more persistent accel-
model motion. So, a possibility is that some false cueing in eration overshoots are present when motion is removed.
rotation, due to the motion filter and its selected parameters,
had a negative impact on performance in this case. On the Pilot-vehicle stability. Figure 25 depicts, for the four motion
other hand, the same false cueing would also be present at the conditions, the means and standard deviations of number of
same time in the lateral axis, since the same filter was used times pilots had an excursion outside the +/-1 deg heading
and the motion cues are proportionally related (eqn. 5). about north per run. These results show that when linear mo-
tion was added, the decrease in the number of overshoots was
Control rate. Figure 18 illustrates the rms cockpit control statistically significant (F(l,4)=8.06, p=0.047). The addition
(pedal) rate for this task. The analysis of variance for these of rotational motion did not yield a significant difference.
data indicated that the decrease in pedal rate was statistically The effects of rotational and linear motion did not interact in
significant for the two cases when linear motion was added this measure.
(F(l,4)=l 1.69, p=0.027). No significant differences were
noted when rotational motion was added, and linear and rota- Control rate. Figure 26 illustrates the cockpit control (pedal)
tional motion effects did not interact. rate for the four configurations. Unlike the previous two
tasks, the addition of linear motion did not significantly re-
Compensation ratings. Average pilot compensation required duce the rms pedal rate. However, the addition of rotational
for this task is shown in Figure 19. Large variations in pilot motion actually increased pedal rate (F( 1,4)= 18.74, p=0.012).
opinion resulted. No statistically significant differences were The rotational and linear motion effects did not interact. So,
noted; based upon the variation in the data, one cannot say this is another case where the addition of rotational motion
that the motion configurations affected the subjective com- made matters worse; however, as Fig. 26 shows, the percent-
pensation required. However, the trends in Fig. 15 follow age increase in control rate was not dramatic.
those in Task 1 (Fig. 11).
Compensation ratings. Average pilot compensation ratings
Fidelity ratings. Figure 20 shows the mean motion fidelity rat- are shown in Fig. 27. The improvement in the ratings for the
ings for Task 2. Here motion fidelity was significantly higher linear motion conditions, relative to those in the rotational
when linear motion was present (F(l,4)=47.9, p=0.002), motion conditions, was marginally significant (F(l,4)=6.38,
while the presence of rotational motion did not affect rated fi- p=0.065). The addition of rotational motion resulted in no
delity. Rotational and linear effects did not interact.
 5-6

statistical difference in compensation. The effects of rota- These results lead to the conclusion that for hover flight simu-
tional and linear motion did not interact. lation, one should make every attempt to represent the linear
lateral acceleration cue, as it is important to all the key aspects
Fidelity ratings. The same result occurred for rated fidelity, of pilot-vehicle performance and pilot opinion. In contrast,
which is presented in Fig. 28. The addition of linear motion the presence of rotational motion about the vertical axis does
resulted in an improvement in fidelity ratings that was not seem to provide significant, if any, value.
marginally significant (F(l,4)=6.15, p=0.068). The addition
of rotational motion again made no difference. The effects of Thus, the yaw rotational motion degree of freedom in hover-
rotational and linear motion were statistically independent. ing flight simulation may not be necessary. The combination
of lateral motion with a compelling visual scene may be all
Lateral motion reports. Figure 29 illustrates the percentage of that is needed to make pilots perceive that physical yaw rota-
the time pilots reported lateral motion present. The addition tional motion is present.
of linear motion significantly increased the reports of lateral
motion (F(l,4)=12.1, p=0.025). Interestingly, the addition of 7. REFERENCES
rotational motion led to a marginally significant decrease in 1. Helicopter Simulator Qualification (DRAFT), Federal
the reports of lateral motion (F(l,4)=5.4, p=0.08). Aviation Administration, Advisory Circular AC-120-xx,
July, 1992.
Rotational motion reports. Figure 30 illustrates the percentage
of the time pilots reported rotational motion was present. No 2. Stapleford, R.L., Peters, R.A., and Alex, F.R., "Exper-
significant effects were found, with rotation being reported an iments and a Model for Pilot Dynamics with Visual and
average of 73% of the time, independent of the motion con- Motion Inputs," NASA CR-1325, May, 1969.
figuration.
3. Bray, R.S., "Initial Operating Experience with an Air-
Summarizing the results of this task does not reveal differ- craft Simulator Having Extensive Lateral Motion,"
ences from Tasks 1 or 2. Linear (lateral) motion was again NASA TM X-62155, May, 1972.
the dominant variable. Its addition improved pilot-vehicle
performance; although its effect upon rated pilot compensa- 4. Jex, H.R., Magdaleno, R.E., and Junker, A.M., "Roll
tion and fidelity was less reliable than in the other two tasks. Tracking Effects of G-Vector Tilt and Various Types of
The only statistically significant effect of rotation was a dele- Motion Washout," NASA CP-2060,1978, pp. 463-502.
terious effect on control activity.
5. Meiry, J.L., "The Vestibular System and Human Dyn-
5.4 Combined Results amics Space Orientation," NASA CR-628, Oct., 1966.
Table 2 presents a summarizes the effects of the presence of
lateral or yaw rotational motion on the six measurables for the 6. Schroeder, J. A., "Simulation Motion Effects on Single
three tasks. All of the improvements in the measures occur Axis Compensatory Tracking," AIAA Flight Simulation
for the presence of lateral motion except for one. In that sin- Technologies Conference, A Collection of Technical
gle case, a marginally significant effect was noted in pilot- Papers, Monterey, CA, Aug., 1993, pp. 202-213.
vehicle stability for the addition of yaw rotation. Statistically
significant degradations in the measures occurred only for the 7. Schroeder, J. A., "Evaluation of Simulation Fidelity
addition of yaw motion. Interactions between lateral and yaw Criteria in the Vertical and Directional Axes," 49th
rotational motion occurred only in the lateral and rotational Annual Forum Proceedings of the American Helicopter
motion perception reporting. However, in Tasks 1, 2, and 3, Society, St. Louis, MO, May, 1993, pp. 419-437.
pilots reported (on average) rotational motion being present
87%, 90%, and 80% of the time when only lateral motion was 8. Peters, R.A., "Dynamics of the Vestibular System and
present, respectively. Their Relation to Motion Perception, Spatial
Disorientation, and Elusions," NASA CR-1309, Apr.,
6. CONCLUSIONS 1969.
Three tasks were performed with a high-fidelity AH-64 heli-
copter hover mathematical model in the NASA Ames Vertical 9. Airworthiness Design Standard, Handling Qualities,
Motion Simulator to determine the importance of yaw rota- Rotary Wing, U.S. Army AVSCOM ADS-33D, July,
tional motion. Three tasks were selected to cover a range of 1994.
situations often encountered in flight simulation. Pilot opin-
ions of the compensation required to perform the task, the 10. Schroeder, J.A., Tischler, M.B., Watson, D.C., and
overall motion fidelity, and their ability to sense motion in all Eshow, M.M., "Identification and Simulation Evaluation
axes were collected. of a Combat Helicopter in Hover," AIAA Journal of
Guidance, Control, and Dynamics, Vol. 18, No. 1, 1995,
The results indicate that for all tasks the addition of linear pp. 31-38.
motion improved pilot-vehicle performance, reduced pilot
compensation, improved pilot subjective impression of fi- 11. Danek, G.L., "Vertical Motion Simulator Familiariza-
delity, and usually reduced control activity. Furthermore, the tion Guide," NASA TM 103923, May, 1993.
addition of lateral motion gave the pilots a strong sensation of
both rotational and lateral motion being present. 12. Tischler, M.B. and Cauffman, M.G., "Frequency-
Response Method for Rotorcraft System Identification
On the other hand, the addition of rotational motion did not with Applications to the BO-105 Helicopter," American
contribute to the improvement of pilot-vehicle performance, Helicopter Society 46th Annual Forum, Washington,
control activity, subjectively rated compensation, or subjec- DC, May, 1990.
tively rated motion fidelity. The presence of rotational motion
led to a weaker sensation of rotational and lateral motion. 13. Mitchell, D.G. and Hart, DC, "Effects of Simulator
Motion and Visual Characteristics on Rotorcraft
Handling Qualities Evaluations," American Helicopter
 5-7

Society Conference on Piloting Vertical Flight Aircraft,

San Francisco, CA, Jan. 1993.

14. McFarland, R.E., "Transport Delay Compensation for

Computer-Generated Imagery Systems," NASA TM-
100084, Jan. 1988.

15. Sinacori, J.B., "The Determination of Some Require-

ments for a Helicopter Flight Research Simulation
Facility," NASACR-152066, Sept., 1977.

16. Mendenhall, W. and Sincich, T, Statistics for

Engineering and the Sciences, Dellen Publishing
Company, San Francisco, 1992, pp. 653-657.
 Table 1 - Motion filter quantities for all tasks

Rotation Lateral Longitudinal Vertical

Task CO,m CO m K CO m CO m
(rad/sec) (rad/sec) (rad/sec) (rad/sec)

1 - linear+rot.1.00 1.0E-5 1.00 1 .OE-5 1.00 1 .OE-5 0.00 1 .OE-5

1 - linear 0.00 1.0E-5 1.00 1 .OE-5 1.00 0.2 0.00 1.0E-5
1 - rot. 1.00 1 .OE-5 0.00 1.0E-5 0.00 1.0E-5 0.00 1.0E-5
motionless 0.00 1.0E-5 0.00 1.0E-5 0.00 1.0E-5 0.00 1.0E-5
linear+rot.0.35 0.55 0.35 0.55 0.35 0.55 0.00 1.0E-5
linear 0.00 0.55 0.35 0.55 0.35 0.55 0.00 1.0E-5
rot. 0.35 0.55 0.00 0.55 0.00 0.55 0.00 1.0E-5
motionless 0.00 0.55 0.00 0.55 0.00 0.55 0.00 1.0E-5
linear+rot. 1.00 1.0E-5 1.00 1 .OE-5 1.00 1.0E-5 1.00 1.0E-5
linear 0.00 1.0E-5 1.00 1 .OE-5 1.00 0.2 1.00 1.0E-5
rot. 1.00 1.0E-5 0.00 1 .OE-5 0.00 1.0E-5 1.00 1.0E-5
motionless 0.00 1.0E-5 0.00 1 .OE-5 0.00 1.0E-5 1.00 1.0E-5

Table 2 - Summary of results

Lat/Rot
Lateral Rotational Interaction
Task: 1 2 3 1 2 3 1 2 3
Measure
Pilot-vehicle stability + + + + - 0
Control rate + + 0 0 0-
Compensation + 0 + 0 0 0
Fidelity + + + 0 0 0
Lateral reporting + + 0 - X
Rotational reporting 0 o X X

+: Significant improvement; Marginal improvement; -: Significant degradation

O: No effect; x: Interaction

o ,_o.
\15|15fY

Pilot

V*-5tf
Y

Figure 1 - Task 1 plan view. Figure 2 - Visual scene for tasks 1 and 3.
 5-9

Figure 3 - Visual scene for task 2 Figure 5 - Visual field-of-view

u£©0 e©^
Linear+Rot. Linear

Rot.
©
Motionless

Figure 6 - Simulator cockpit motion for motion

configurations in Task 1.

Figure 4 - Vertical Motion Simulator

5-10

20 40 60 80 100 20 40 60 80 100
Time (sec) Time (sec)
Figure 7 - Task 1, Lin.+Rot. motion Figure 8 - Task 1, No motion
 5-11

High ■
■

Linear
*—*
.gMed. T

LJ- 1
1
T No linear
I
i
*■ -

I i
LOW
1
1

No rotation Rotation No rotation Rotation

Fig. 9 - Measured Pert., Task 1 Fig. 12 - Motion fidelity, Task 1

Yes '
>—-__Linear
c
'•c
o 1
Q.
(D - !
No linear
2 —■
i
CD
(0 i
.—
No
. 1

No rotation Rotation No rotation Rotation

Fig. 10 - Control rate, Task 1 Fig. 13 - Lat. Motion Perception, Task 1

Yes
c
'•e
o
Q.
0)
DC
CO
>
No

No rotation Rotation No rotation Rotation

Fig. 11 - Pilot Comp., Task 1 Fig. 14 - Rot. Motion Perception, Task 1
5-12

20 40 60 40 60
Time (sec) Time (sec)
Figure 15 - Task 2, Lin.+Rot. motion Figure 16 - Task 2, No motion
 5-13

10
T High •
CO 1
o 1
o Linear
sz
co
I 1 >*, c)
i—
I . -X
0 No linear - - '
>
O I . 1 igMed. ■

O
¥- " Linear__—-—< 1
1 il
i
O T
1 T No linear - X
1 x- - 1
Low 1

-2
No rotation Rotation No rotation Rotation
Fig. 17 - Measured Pert., Task 2 Fig. 20 - Motion fidelity, Task 2

T
I T
Yes ■

I 1
1.5 X- -
No linear c
-X

T I "•c Linear
o
Q.
To (1 1 CD
J: 1 1 Linear DC
1

CO
Lateral

TJ
CD
Q.
W0.5 No linear
z:
o

oc 1 1

0
No rotation Rotation No rotation Rotation
Fig. 18 - Control rate, Task 2 Fig. 21 - Lat. Motion Perception, Task 2

Ext.
Yes Iinear
T ( () i
c O)
L i
_o I c
Cons. | ■c
CO
o J<
CO
c i No linear - X
a. J - " I
CD CD
Q. J I DC I „ - l
£ Mod. 5 x
o
Ü
Linear 1
I
CO
>-
k" "No linear
i
i

• No i ■

Min. ■

No rotation Rotation No rotation Rotation

Fig. 19 - Pilot Comp., Task 2 Fig. 22 - Rot. Motion Perception, Task 2
5-14

10

c» 5
0

C
TJ
>/|((fV^^
CO
0
I -5

-10

. ,
Ü
0
CO
20 -
I
^3
0
w I J
0
TJ

Ü
Ü
CO
mmm in
if
^

i
5-20
CO

1——

1
IT
I,t ,.
10 o Mm ftm m*Wfm •
-1

0 50 100 150 100

Time (sec) Time (sec)
Figure 23 - Task 3, Lin.+Rot. motion Figure 24 - Task 3, No motion
 5-15

High

cy Linear

■—
X- H
® Med. T
E x- -
1 No linear

■
1
Low ■

No rotation Rotation No rotation Rotation

Fig. 25 - Measured Pert., Task 3 Fig. 28 - Motion fidelity, Task 3

1.5 1

o i Yes
CD
-x- -

1
c
£, 1 1 No linear _-_; ■c
o
■
T
CD
CO i
Q.
CD
_JJnear
1
Linear . i
CC 1
CO "cö 1
1
8 0.5 £ T
00 ■ 1
c/> No linear ■ *

1
No 1
1

No rotation Rotation No rotation Rotation

Fig. 26 - Control rate, Task 3 Fig. 29 - Lat. Motion Perception, Task 3

—
1
Ext. 1 Yes
i
T
i i
No linear
c cj i Linear ,
o i l c
■Jo Cons. i__ — c
'■c
o
X--
i
X

w Q. No linear
c 1 Linear CD . i
CD
CC i
1
| Mod. ■

CO
i
1
ü
No •
Min.

No rotation Rotation No rotation Rotation

Fig. 27 - Pilot Comp., Task 3 Fig. 30 - Rot. Motion Perception, Task 3
 6-1

Achieving High-Fidelity Motion Cues in

Flight Simulation
S.K. Advani and J.A. Mulder

SIMONA International Research Centre

Delft University of Technology
Kluyverweg 1, 2629 HS Delft, The Netherlands
tel.+31 15 78 1395 • fax+31 15 78 6480 • E-mail: [email protected]

Abstract Nomenclature i

The simulation of high-bandwidth manual flight Fext external force

control tasks dictate the use of simulator motion Kq position error gain
systems for the reproduction of motion cues on the K4p pressure diff'l feedback gain
vestibular and neuromuscular-mechanical arm- Ko(s) control signal gain (time varying)
manipulator system. The reproduction of these Kvel velocity feedback gain
cues, with particular emphasis on the lowest M mass of actuator load
possible time delay, is necessary in human Pos. fb. position feedback
perception research, experimental flight control Press, fb. pressure feedback
system development studies, as well as in routine qo demanded actuator position
flight training. Perfect reproduction of motion cues q measured actuator position
with ground-based flight simulators is principally s Laplace operator
impossible due to the kinematic limitations u valve input signal
inherent to the motion system. Washout filters AP pressure differential
minimize these effects. The dynamic e error
characteristics of the motion system however lead *o valve supply flow
to two types of control errors: short-term, due to
finite oil stiffness and line dynamics, as well as Introduction
limited control valve bandwidth, and long-term, The essential purpose of piloted flight simulators is
due to complexity of the non-linear motion system to reproduce the physical and environmental cues
dynamics, making compensation of unwanted as closely as possible to those encountered in
parasitic errors difficult. This paper will review flight. These cues result from control inputs
techniques which increase the performance of six- exercised by the pilot, and are also due to external
degrees-of-freedom hydraulic motion systems for disturbances acting upon the vehicle. The pilot-
flight simulators. Application of pressure-feedback vehicle interaction forms a closed loop in which the
actuator control increases the robustness of the pilot receives information about the state, and
motion system dynamics, hence decreasing the change in state of the vehicle, through the sensory
short-term errors. The long-term control errors are organs. The visual information is obtained through
addressed by a (separate) robust, multi-variable the foveal and central vision, while motion cues
motion system controller which provides control are derived from the visual and the vestibular
signals to the platform with knowledge of the system. These stimuli are processed by the brain,
system state and its inherent properties. The total which can be thought of as to compare all the
mass and the vertical location of the centre of signals to an internally-developed model. In order
gravity of the platform influence the time delay for the response of two "similar" systems to match
(phase lag) and fidelity of the motion system. (hence, implying the fidelity of the simulation), the
These properties also limit any such improvements input signals must correspond with the aircraft
due to design changes in software or hydraulic specific forces and angular accelerations1 at the
hardware. As a result of these studies, a pilot head position.
fundamentally unique light-weight motion platform
design is proposed, making extensive use of In human control tasks, visual information requires
advanced composite materials. This is called the longer processing time than the vestibular cues. In
SIMONA Research Simulator. With the other words, the onset of motion is primarily
improvements made to motion cue quality, detected by the vestibular system, while long-term
fundamental research into human perception changes in motion are sensed predominately by
processes in human perception research, and the eyes 2. This requires that the initial response of
experimental flight control system development the simulator motion system matches that of the
work, will not be influenced by parasitic motions. vehicle being simulated. This "onset cue" often

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?"
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
6-2

includes the high-frequency components. In an the vehicle dynamics. The pilot position
ideal situation, the latency of the simulator above experiences frequencies through the audible
that of the vehicle dynamics should be close to range, however for the purposes of motion
zero, over the entire frequency range of the systems, one should try to address a cutoff
vehicle motion. This applies for the presentation of frequency of approximately 15 Hz for most
motion cues in particular. In practice, this is not vehicles. It can be shown from flight tests that
easily achieved, mainly due to the limitations in the aircraft encounter excitations with frequency
mechanical hardware and motion control software components in excess of 15 Hz. These can lead to
which induce the sensation of motions to the pilots. sharp accelerations at the pilot position.
Host processor speed and vehicle model
complexity limit the update rate of the state In order to realize higher performance (and hence
vectors. This report will however focus on the higher fidelity) than current motion systems, a
motion system alone and suggest means to number of issues must be addressed. The
improve its properties. frequency response of the simulator motion
platform can be shown from experimental
Current simulator requirements, such as those measurements5, as well as from dynamic models
established by the United States Federal Aviation which simulate the hydraulic, mechanical and
Administration stipulate that the motion response electrical control feedback system properties. The
should precede all others, and that the total latency latter has the advantage of being able to indicate
not exceed 150 milliseconds3 (for Level-D training the parameters to which the motion quality is most
simulators). It is the aim of the present research to sensitive.
significantly reduce this figure, and maintain a very
high level of fidelity with the required responses. A study to investigate the sensitivity of the moving
platform to variations in payload mass properties
At the Delft University of Technology, the was conducted6,7. A complete non-linear simulation
International Centre for Research in Simulation, model of a synergistic six-degrees-of-freedom
Motion and Navigation Technologies SIMONA4 is motion system8 was used to evaluate the effects of
investigating the fundamentals of motion system platform inertia on the total dynamic behaviour9,10.
dynamics, and proposing approaches to reduce This model includes all hydraulic and mechanical
latency and increase the bandwidth of current properties of the motion system, as well as inertial
systems. A number of critical areas are addressed, properties of the payload mass and the actuators.
which will be reviewed in this paper, namely: All model elements are developed as discrete
systems which are free of internal feedback
• Motion platform dynamics as a function of couplings, similar to the construction of the actual
mass properties simulator hardware. These are also joined in a
data flow architecture representing the mechanical
• Actuator control laws and shortcomings in high system structure.
frequency control
The major components of the motion system in
• Platform control laws this simulation model include the mechanical
system, including the moving and fixed platforms,
hydraulic system, including the power supply
• Platform design concepts to reduce inertial
system, servo valves, actuator hydraulics, and
properties
electronic system, including the computers,
actuator controllers, transducers, filters and safety
systems
Motion System Frequency Response and its
Value on Simulation
A functional diagram of the simulation math model
It is fortunate that for control tasks the specific
is shown in Figure 1.
forces and angular accelerations need be
presented as motion cues to the pilot (rather than Dynamic Simulations
absolute magnitudes of acceleration)2. Due to the
The model described above, once validated with
high-frequency nature of the human vestibular
reasonable reliability8, was used to assess the
sensors however, the quality of the responses,
response of the motion platform. Figure 2 shows
namely their temporal characteristics such as
the response of a simulator platform resulting from
delay and phase, should be reproduced with little
a 0.5 m/s2 step acceleration in the surge (X)
deviation from reality. The critical requirement of a
direction, while the total platform mass is varied
motion system is to provide a given acceleration at
between 2000 and 18000 kilograms. (Note that a
the pilot position such that the onset of this motion
typical full flight simulator may have a mass up to
is presented with a minimal delay. This should be
18000 kg). The mass distribution is assumed
made possible over a wide frequency range,
uniform and throughout a sphere with a radius of
preferably beyond those fundamentally required for
 6-3

3.0 m. In the figure shown, the centre of gravity is stroke actuators due to transmission line effects. In
established at the centroid of the upper gimbal most flight simulator motion systems, long-stroke
plane. Since the total load acting on the actuators (1 m or more) actuators are incorporated.
is a function of the payload mass, this result
implies the control system requirements. A High-performance actuation will also require that the
significant decrease in the system natural effects of parameter uncertainties and system
frequency, and increase in damping, as would be variations be minimized through the application of
expected, results from increasing payload. Most robust control. Some practical examples of these
notably, the onset of the motion is delayed; as phenomena are:
payloads increase, the desired acceleration level is
delayed significantly. • dynamic variations in actuator loads throughout
the operational envelope (including dynamic
A second set of results shows the influence of the coupling due to the presence of large cable
centre of gravity location on the surge response. In bundles)
this case, a fixed mass of 2000 kg is assumed for
the platform. Variations in the centre of gravity, • variations in static loads (hardware
which in surge acceleration inputs influence the configurations, number of on-board occupants)
dynamic contribution load on the actuators, yield
the results in Figure 3. Note that the onset • internal system variations: leakage flows associated
acceleration delay is not significantly affected by with hydrostatic bearings, oil temperature
these variations. The centre of gravity vertical fluctuations, degraded (contaminated) servo valve
location does, however, cause a simultaneous performance, contaminated oil
decrease in both the damping and the natural
frequency. This rate of decrease appears, from this With increased performance, robustness in a motion
study, to be reasonably linear and can be system can also reduce its maintenance time.
attributed to the virtual mass (and not the real
mass), caused by the dynamic load on the Platform Control of Long-Term Errors
actuators. Unlike a true increase in the gross The compensation of inertial effects related to the
moving mass, the increase in the vertical offset in motion system dynamics is also necessary.
the centre of gravity increases a moment coupling Whereas the actuator-level controller compensates
effect, hence increasing the load on the actuators. for each individual actuator, a higher-level system
controller should compensate for the entire
Figure 4, shows the platform response to a platform system dynamics. This may involve a
rotational acceleration in pitch. The same is shown combination of dynamic modelling, prediction, and
for roll motions in Figure 5. Again, a lower mass feed-forward of the system response, and the
yields more favourable properties in terms of the communication of the required platform position to
initial response and natural frequency. the actuator-level controllers.

As one would expect, the above results suggest Inertial Properties of the Motion Platform
that the motion system response is indeed Finally, the platform inertial properties should
sensitive to the load on the actuators. Any form of afford a minimal level of required compensation
control should take into account that these loads from the control system. It can be shown that the
may vary not only throughout the envelope, and as dynamic load variations on actuators can vary by
a result of variations in the platform mass an order of magnitude13, and these variations can
properties. As a result, the actuator control, limit the eventual performance of the entire
platform control and the design of the motion system. Therefore, a practical means of reducing
platform itself should be carefully considered. This the effective actuator loads should be sought. The
is further explained in the following discussion. eventual solution should also provide a platform
with a stiffness greater than or equal to the control
Actuator Control of Short-Term Errors system frequency range, in order that the
High-performance motion system actuators must maximum performance is achieved. Note also that
provide the required acceleration regardless of the actuators themselves should have inherent
their actual loads and internal properties. This natural frequencies in this range.
demand increases the required performance of the
actuator controller in order that wide frequency From the above results (Figures 2 to 5),
ranges can be presented. The control strategy reductions in the platform mass and vertical
therefore shifts towards high-frequency force location of the centre of mass can better the
control loops, i.e. providing pressure control, rather inherent characteristics of the motion system. The
than the conventional mid-frequency position- system natural frequency in particular appears to
based servo control loops11,1 . These high benefit directly from these properties.
frequencies become increasingly relevant in long-
6-4

Realizing High-Fidelity Motion Cues some combinations when aiming for high-
The Delft University of Technology is investing performance control, particularly when long-stroke
major efforts to improve and demonstrate motion actuators are involved*. First, consider that in most
system performance in its new SIMONA Research hydraulic servo control applications, a combination
Simulator. The SIMONA program will, among of classical feedback plus a damping loop,
other issues, address all areas of the motion employing either pressure or acceleration
system design and control, to develop new feedback is used. Highly reliable pressure
strategies which will significantly improve time difference transducers (PDT) are usually preferred
response characteristics. The remaining discussion over more expensive acceleration sensors. For
in this paper will present the proposed scheme, practical purposes, PDT's are usually placed close
review recent developments, and suggest the to the valve. The actuator system model
advantages to this approach at improving the developed by Schothorst show the stability
response of a research simulator. problem encountered in an open-loop" frequency
response, Figure 7. Experiments by the same
The SIMONA approach to high-performance author have reproduced this phenomena, and it is
motion systems involves a multi-level approach suggested that a combination of transmission line
where the problem is initially defined by the type of dynamics (i.e. their resonance peaks) and the 180°
control necessary. The motion control problem phase shift associated with the valve dynamics
begins after the motion washout signals are cause the pressure feedback problems. This can
generated by the host computer. These yield the be shown by removing the valve dynamics
desired state of the simulator actuators, and contribution from the model, and is shown by the
request this response from the Motion System dashed line in Figure 7.
Controller, as shown in Figure 6.
Schothorst et al 14 propose that the solution to
Multivariable Six-Degrees-of-Freedom Motion achieving stability in high-performance systems
System Controller lies in a combination of the following factors:
This high-level control system, with inherent
knowledge of the motion platform and hydraulic • Choice of the most suitable feedback signal,
drive system properties, knows then the current such as the pressure difference at the actuator
state of the platform, and predicts the future state orifices, rather than at the servo valve (thereby
prior to generatinga command signal, thereby eliminating the transmission line dynamics
allowing refinement to the actuator-level input within this critical feedback signal).
signals.
• Choice of the servo valve, since the valve
The hydraulic actuators in fact become "white boxes", properties limit the allowable pressure
with their state of operation being fully known by the 6 feedback gains.
DOF controller. Thus, a general controller is
developed to account for the reflected masses • Application of short transmission lines, to
encountered on each motor, and which compensates maintain the resonance modes in the very high
for all instantaneous interactive loads. frequency ranges.

A first approach will investigate linear and nonlinear • Advanced control design for pressure feedback
multivariable strategies for motion platform control to loops, such as digital filters for the feedback
compensate for platform loads and dynamics. signals. Robust control can provide tradeoffs
between stability robustness and performance.
In a second approach, while the simulator is in use, a
parallel model of its dynamics is run in real-time. This Robustness in actuator control loops
model is driven by the same signals which drive the Traditionally, hydraulic servo control systems have
motion system actuators, and measures the actual incorporated proportional position feedback loops.
load on the motion system. Feed-forward control These are coupled with the desired positions,
implemented in the parallel model can then be which are double integrations of the desired
applied to the motion platform, providing a accelerations resulting from unbalanced forces in
deterministic outcome the vehicle dynamics filtered through the motion
washout algorithms. Note that a platform-to-
Robust Actuator Controller actuator axis transformation resolves the
Improving the total performance of the motion kinematics of the synergistic 6 DOF motion
system requires also that the actuator-level
controller be addressed. During an analysis into An actuator with a length of 1.0 m was used in the experiments
the impact of transmission line effects coupled with described)
hydraulic control system dynamics, Schothorst14 **
shows that severe stability problems may occur in Drifting of the actuator, due to its integration nature was
countered by applying a manual offset.
 6-5

system. The motion control then becomes a SISO including the moving platform. In the next section,
position control system for each actuator, and is a novel and practical solution to light-weight,
usually kept an analog circuit. The standard highly-rigid moving platforms will be presented.
hydraulic servo concept is shown in Figure 8. Note
that the feedback relies on the position signal q, High-Performance Motion Platforms
and the acceleration-related signal KAP. No velocity In order to maximize the motion performance of
signal is required. This technique cannot however the entire simulator, it would be wrongful to
compensate for interaction forces. It has also been disregard the natural properties of the motion
shown that this concept is poor in robustness and platform, the payload-carrying element. A motion
has difficulty in compensating for variations in platform must sustain the occupants and
payload mass. equipment, and transfer the loads from the motion
system actuators to the simulator cab. The
A more recent approach to servo control16 is occupants may include the pilot(s), instructors or
increasingly applied, especially in robotic control. experiment controller, and observers. The
In this concept, shown in Figure 9, the high-gain equipment payload will usually consist of the
pressure feedback is used to achieve pressure aircraft or aircraft-like instruments, displays, the
tracking, while position tracking is obtained from controls and loading devices, electronic equipment
the outer loop feedback. Load decoupling can be required for the simulation, computer systems, and
achieved through pressure feedback, while velocity the visual display system. Modern simulators often
signals are used for stability. In practice, there is employ wide-angle collimating display systems
an additional need for accurate velocity signals, which use off-axis projectors mounted above the
accomplished by integrating the force cab, presenting the image onto a spherical-section
(acceleration) signal. mirror, and via a rear projection screen. A result of
this solution is a high contribution to the mass
The advantages to pressure control over position moment of inertia, while requiring a stiff mounting
control are significant. Figure 10 shows a structure.
comparison of the closed loop frequency
responses for the two systems, obtained from Traditionally, simulators are constructed in a
experimental analyses. The two columns show layered fashion, beginning with a fairly rigid
measured position and acceleration responses. platform structure mounted atop the motion system
Note the lower break frequency associated with the gimbals. This platform supports then the cockpit
position feedback configuration (solid lines), and replica with its interior, instruments, displays and
the larger phase difference. This is especially true controls. Behind the cockpit, an instructor cabin is
in the acceleration signals, those most significant placed where the instructor station and observers'
to simulator operations. chairs are mounted. Above this cabin, the display
system projectors are mounted. The mirror
When the load is doubled (from 750 kg to 1500 structure is mounted directly to the platform at floor
kg), a similar phenomena is apparent, however the height.
differences between the position and pressure-
coupled systems are even greater (Figure 11). The SIMONA Research Simulator (SRS), however,
is not intended for training; therefore a number of
Clearly, the pressure feedback configuration shows the on-board systems can be simplified or
significant robustness in load variations as neglected, reducing the total volume and payload
opposed to the position-controlled system. In the mass requirements.
pressure feedback arrangement, the actuator
behaves as a "force generator" via control by In efforts to reduce significantly the moment of
adequate feedback of differential pressure & inertia of the platform, a solution has been
velocity. Note that this robustness is due only to developed whereby the internal volume
the control system configuration, while even further requirements are first defined, the constraints of
improvements could be achieved through the the motion and visual display system established
application of true robust control techniques. (so that no structure blocks the light beams or
interferes with the motion system in any actuator
These results suggest that the pressure feedback position), and a structural shell placed around the
concept is also well suited to compensate for occupants and internal equipment. An emphasis
interaction forces, and can allow a flexible trade-off was placed on reducing the mass while
between position and pressure feedback control, maintaining sufficient stiffness of the entire
since it represents an extension of the latter. It system. Finite element analyses showed that a
does however require computation of the velocity minimum natural frequency of 17.5 Hz could be
signal, and is sensitive to high-frequency dynamics achieved for the required configuration; a higher
of the moving payload. This last constraint stiffness requiring increasingly more material,
stipulates high rigidity in the entire motion system, thereby raising the mass as well as cost. Lowering
6-6

the centre of gravity was also considered a crucial can be increased through the use of high-stiffness
design objective. The result of the aforementioned light-weight materials.
exercise is a highly-rigid light-weight platform and
a simulator which is capable of generating sharp Research in the SIMONA programme will continue
accelerations. The resulting natural frequencies of to refine these areas, and also consider new
the platform structural dynamics exceed those of approaches to motion washout.
the rigid and relevant aeroelastic aircraft
dynamics. Furthermore, the small cockpit will Acknowledgments
place the pilot very close to the centroid of the The Netherlands Ministry of Education, Science
upper gimbals of the motion system. In this way, and Culture, and the SIMONA Industrial and
maximum use of the pure pitch and roll capabilities Scientific Partners are thanked for their financial
of the motion system are provided to the pilot. It is support of the SIMONA programme. The authors
also possible to create a large window free of would also like to thank ir. G. van Schothorst and ir
reinforcing frames (Figure 12), while still meeting P.C. Teerhuis (Faculty of Mechanical & Marine
the aforementioned structural stiffness Engineering) for their contribution to this paper.
requirements. This allows a variety of cockpit
interiors to be installed, and cosmetic window References
frames inserted if necessary. 1. Hosman, R.J.A.W. and Steen, H. van der, "False Cue
Detection Thresholds in Right Simulation". AIAA-93-35-CP.
From AIAA Flight Simulation Technologies Conference,
The aramid/carbon fibre cockpit is equipped with Monterey, August, 1993.
carbon-fibre outrigger rods which accept the visual
display system structure. A projector platform 2. Vaart, J.C. van der, "Modelling of Perception and Action in
supports the CRT projectors, while a framework Compensatory Manual Control Tasks". Ph.D. thesis, Delft
construction supports both the back projection University of Technology, December 1992.
screen and the mirror chamber. SIMONA is
3. Advisory Circular 120-40C, United States Department of
considering light-weight alternatives to the Transportation, Federal Aviation Administration, Draft,
standard aluminized mylar mirror due to mirror January, 1995
resonance and mass concerns.
4. Advani, S.K., "The Development of SIMONA: A Simulator
When complete, the cockpit of the SRS will allow Facility for Advanced Research into Simulation Techniques,
Motion. System Control and Navigation Systems
fundamental research into human perception
Technologies". AIAA-93-3574-CP. From AiAA Right
processes, interactions with displays, flight control Simulation Technologies Conference, Monterey, August, 1993.
concepts, and flight dynamics models. Note that
with the high performance expected in this motion 5. Grant, P.R., "Motion Characteristics of the UTIAS Right
system, the human perception systems will not be Research Simulator Motion Base". UTIAS Report number 261,
polluted parasitic errors, thereby yielding new CN ISSN 0082-5263, July 1986.
opportunities for future work in this area.
6. Advani, S.K., and Verbeek, R.J., "The Influence of Platform
Inertial Properties on Simulator Motion System Performance".
The SRS integrated system is shown in Figure 13. AIAA-94-3418-CP. From AIAA Flight Simulation Technologies
The development of this simulator, a major Conference, Scottsdale, August, 1994.
technical challenge, has in itself been an exercise
in applying all facets of simulator technology, 7. Svartstrom, A., "The Influence of Platform Inertial Properties
on Simulator Motion System Performance in Pitch and Roll".
combined with new applications of materials,
SIMONA Report SR-95-MP1-001. Delft University of
control theories and design integration. Technology, February, 1995.

Conclusions 8. Kuit, J.J., "A Nonlinear Simulation Model of a Six-Degrees-

Achieving high-fidelity motion cues in flight of-Freedom Motion System". Delft University Thesis Report,
simulators involves re-addressing all major areas July 1991. Kuit, J.J., "A Nonlinear Simulation Model of a Six-
Degrees-of-Freedom Motion System". Delft University Thesis
of both the hardware and the control system. The
Report, July 1991.
approach described in this paper will lead to
significant improvements in motion cue 9. Verbeek, R.J., "Static and Dynamic Design Considerations
capabilities, especially in the high-frequency for a Six-Degrees-of-Freedom Synergistic Motion System".
regimes. Pressure-based control systems can Delft University Thesis Report, August 1994.
compensate for load variations and, later, robust
10. Huurdeman, R.J. "Development of SIREMOT: A Nonlinear
control applied to the actuators can overcome 6-DOF Motion System Model in SIMULINK". Delft University
internal system variations. The multivariable six- Thesis Report, March 1995.
degrees-of-freedom control system will ensure that
the actuator controllers are fed with signals taking 11. Merrit, "Hydraulic Control Systems", Wiley & Sons, 1967.
into account predictions of the system state.
Moreover, the SRS platform design philosophy 12. Viersma, T.J., Analysis, Synthesis and Design of Hydraulic
Servo-Systems and Pipelines". Delft University of Technology,
suggests that the total motion system performance Faculty of Mechanical Engineering, 2nd Edition, 1990.
 6-7

13. Levi, R.W., and Hayashigawa, L, "Specification 15. Schothorst, G.van, "Inner Loop Control of a Hydraulic
Considerations for a Small Motion-Base". From AIAA Flight Actuator: Possibilities and Limitations". From Proceedings of
Simulation Technologies Conference, Atlanta, 1988. 14th Benelux Meeting, Houthalen, March 1995.

14. Schothorst, G. van, Teerhuis, P.C., Weiden, A.J.J. van 16. Heintze, J., and Weiden, A.J.J. van der, "Inner Loop
der, "Stability Analysis of a Hydraulic Servo-System Including Design and Analysis of Hydraulic Actuator, with Application to
Transmission Line Effects". From Proceedings of International Control". From Proceedings of 4th IFAC Symposium on
Conference on Automation, Robotics and Computer Vision, Robotic Control, Capri, Italy, September 1994.
Singapore, November 1994.

MOST

aat point«

MOTION
DRIVE
COMPUTER

I aal pol nil

T
TRANSDUCER CONTROLLER

■
.
plat« dlaplacamanla
■«vovahr* control algnala p|« on valorttlM

■■ r
eyllnd« pnaaurva HYDRAULIC
SYSTEM

platon fore

actualon
axtanalona MECHANICAL
vaJocHIa* SYSTEM

Figure 1. Simulation Model Structure [8]

Suige-acceleration slepresponso with varying c o.g. (mass*2000 kg)

Surge-acceleration stepresponsa with varying mass (c o g= 0.00) i r 1
' I I ; selpoint signal
"selpoint signal'' ZCOG 0.00m —-
"2000kg- - ZCOG-0 50m

i\\ ... ■4000kg-

-8000kg'
■12CO0kg- - - -
'18000kg' - - .
A 1/ Z COG 0.75 m
ZCOG-1.00m ---

/7 U 'y'i\
-•• Irf Y 7 \\\ ./"v ..... \ v'i \ VY
- ." "--'..';'-•„ :...., ! ! •: i *■
\ H-T- '

i
!:'■"!.''■ : '•( '■ •■' \

| 04 h.i :"\ \: ./ \. / 'v'' /j j V.

j; ;',-' ■ / ■ '•••'...-' \ i
- •
0.2 - \.J ..;
f;i : '"■ |
0.1 -
f " !
' ' •
Time |sec]

Figure 2. Motion System Response in Surge Figure 3. Motion System Response in Surge Acceleration,
Acceleration, Variations in Platform Mass Shown | Variations in Platform Centre of Gravity Vertical Offset
Shown [9]
6-8

! 1 !— l ' !
'setpoirl 5=n;1ar —
"selpoint signal!" -2000kg- ■- ■
"2000kg? —- "40C0kg"

- i
"4000kg?
"8000kg" -- ■ _,_ .__,—.,. ■-^^•Mmf-s-rK;
.::■•■■ "lzoodRg""- -
"12000kg? —- ■ - - --lBOOOkg"
"18000kg? ---

u
0}

a
:■'■■■■'■■■■■'"""

0)
-a

CO 4

o
DC

Time [sec]
Time [sec]

Figure 4. Motion System Response in Pitch Figure 5. Motion System Response in Roll
Acceleration, Variations in Platform Mass Shown [6] Acceleration, Variations in Platform Mass Shown

Simulation Program
6 D.O.F
Aeroplane Motions
Motion Drive Laws
6 D.O.F
Desired Platform Motions
(& Motion System State)
SRS Motion Control
Actuator controls &
measurements
!^
i 4—J

< !<
Mechanical coupling Actuator loads &
displacements
Motion Platform

Figure 6. SIMONA Motion System Control Concept [15]

 6-9

Measurements 1O-MAR--9£
1 2 -
9 _
6 -
,'^7:~ >SN
3

0 ft \ ''S
-3 ) 7 s
W' ' -/ /

-6
_JA>- '>^
-9
r ^^^ _-^
-1 2 - Xs => Delta P

-1 5 -, , , I ,,,!,, , , , 1 ■ ,,!,,, I , .
-15 -10 O 5 10 15 20
Real

Figure 7. Nyquist Plot of Open-Loop Control Gain, with

and without Valve Dynamics [15]

KAP
AP
Valve Actuator * >
1 u + O0
.H ^q
J r
+^
'"
* Kc(s)
► Tr. Line + q >
q .p-
Dynamics Load

Figure 8. Position-Based Servo Control Concept [15]

AP
Valve Actuator <* >
APd, '"£AP u + ®0 <j .
► Kc(s) +
f * KAP +' i
Tr. Line
q
Dynamics Load h
►•

• - r+
T F1ext
q
Kycl ^
K K
q 4
i
Re1
i i,
q
+ -
;4d+ " Q outer loop

Figure 9. Pressure-Based Servo Control Concept [15]

6-10

g(*)
Qd(s) Qd(s)

M = 75Q kg 1O—MAR—95
_ ' i

1
<-> 1 1 ' \
: \ ' \
\ ' '
■Ü

- .1 .1
a
:— Pos.fb.

A
E — Pos.fb.
< — Press.fb _ Press.fb. \
100

1 1 o 1 O 1 OO
Frequency [Hz] Frequency [Hz]

Figure 10. Comparison of Position and Pressure-Based Actuator Control; M=750 kg [15]

q(s) q(s)
Qd(s) Qd(s)

M = 1500 kg 1O-MAR-95
\
_-- i =r—\ ---■*' \
\
\
\
: \ \
I
1,
; \ 1.

;—. pos.fb. \
1
1

"--•Press.fb. \ |

1o
VII 1111

1 oo

-
i

-
\i

-300 -
— Pos.fb.
-400
; - -■Press.f b.
MM'
-500 •■ • ' t

1 O 1 00 1 1o 1 OO
Frequency [Hz] Frequency [Hz]

Figure 11. Comparison of Position and Pressure-Based Actuator Control; M=1500 kg [15]
 6-11

Figure 12. Composite Materials Motion Platform/Flight-Deck of

SI MONA Research Simulator

Tl Oelft

I:

Figure 13. Integration of SIMONA Research Simulator

 7-1

APOGEE : A BREAKTHROUGH IN SYNTHETIC IMAGE GENERATION

Jean-Claude Chauvin
Chief engineer
Sogitec Electronique
4 rue Marcel-Monge, Immeuble Nobel
92158 Suresnes Cedex
France

SUMMARY RESUME
Though polygon and pixel capacity remain significant Bien que les performances en polygones et en pixels
parameters of real time image generators, the visual demeurent des parametres significatifs des generateurs
aspect and the content's quality of the images take d'images temps reel, la qualite du contenu des images
more and more importance with their utilization in prend une importance croissante avec leur utilisation
NOE flight of helicopters, or combined arms en simulation de vol tactique d'helicoptere, ou
simulation and interoperability context. In this field, d'operations interarmes. Dans ce domaine, APOGEE
APOGEE represents a breakthrough in real time image represente une percee dans la generation d'images de
quality and complex scene content restitution. qualite et la restitution de scenes complexes.
An improved management of the database guarantees Une gestion optimisee de la base de donnees assure la
the best restitution of the scene content accounting for meilleure restitution du contenu de scene, compte tenu
the specific polygon and pixel capacity of one system. des capacites particulieres d'un Systeme.
AZtec*, the new algorithm developed by Sogitec for AZtec*, le nouvel algorithme developpe par Sogitec
hidden part elimination and antialiasing, allows for pour le traitement de l'aliasing et des parties cachees,
processing of opaque, transparent, or lighting polygon permet la restitution de surfaces opaques, transparentes
participation, in any number and in any order. A nice ou eclairantes, en nombre et en ordre quelconques. Un
rendering of complex scenes, regarding the lighting rendu realiste, particulierement au niveau de
and shading aspect, is obtained without the sorting l'eclairement et de l'ombrage, est obtenu sans les tris
constraints encountered in similar systems, to process obligatoires, dans les systemes similaires, pour traiter
transparent faces at the end of the computation cycle. les parties transparentes en fin de cycle.
On the whole, the optimization and the integration of L'integration et l'optimisation ä tous les etages du
all stages of the APOGEE image computer, provide the calculateur d'images APOGEE procurent les
best figures on all image generation criteria, and offer meilleures caracteristiques pour tous les criteres de la
modularity in terms of performance and image generation d'images, avec une grande modularite en
definition. performance et en definition d'image.

1. INTRODUCTION - Image Quality: To obtain the best training efficiency,

the synthetic nature of the images should be forgotten
The simulation domain generates severe constraints for
by the trainees. The scene representation must be
the design of real time Image Generation systems:
realistic, without distracting effects which degrade the
- Time constraint: In man in the loop simulation, the training effectiveness.
lowest transport delay is required, with update rates of
The way these constraints have been dealt with in
60 Hz, or higher. A dedicated architecture must be
APOGEE are described in the followings.
designed to satisfy this major requirement.
- Database Coverage: Today's mission training requires 2. ARCHITECTURE
to fly over huge territories, exceeding one million
square kilometer. This implies an efficient database Most of the Image Computer performance is directly
organization and management accounting for related to the architecture and the way the different
geographic coordinates.. computations are sequenced. The APOGEE system is
- Scene Content: Very detailed scenes must be composed of two sub-systems:
represented to simulate complex combined operations, -the Scene Manager, a VME-based computer witha
and when no explicit data is available in the database, real-time operating system,
artificial features must be generated to maintain -the Image Computer, a dedicated architecture
movement and position cues. developed by Sogitec, with proprietary busses.

* AZtec: Advanced Z-buffer technology. Sogitec proprietary algorithm.

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?"
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
7-2

Both are synchronized on the video rate. Each of them The Extraction Processor and the Geometry Processor
takes one frame cycle to perform its tasks. are based on the same computer unit (CGP board).
- The Extraction Processor is composed of one CGP
APOGEE architecture board equipped with additional memory which
supports the local database.
from the. Scene Video
- The Geometry Processor is a parallel structure of up
Simulator Manager output to eight CGP boards. Due to the high throughput of the
polygon bus and the protocol used, there is no derating
on computation power with the number of boards.
simulation first cycle second cycle display
- The Pixel processor is a parallel pool of up to eight
sync sync sync
dedicated CIP boards. One board is composed of four
pixel processor cells in parallel, and contains twenty
An image is then displayed two cycles after new
ASIC of four different types.
position parameters transmitted by the simulation host
computer. -The Video Processor is dedicated to convert the
content of the image memory into video signals. It is
2.1 The Scene Manager also used to process the image for edge matching in
multichannel projection, to generate operational
The Scene Manager performs several tasks among symbology, and to manage calligraphic lightpoints, if
them are the interface with the simulation host any.
computer, the management of the database, the control
of the image computer, and other simulation support
APOGEE
functions. from the
Scene Manager Image Computer
- Interface: The Scene Manager ensures the interface
with the simulation host computer, the transformation
of simulation data into image parameters (ownship and
moving models position and attitude, weather effects,
time of day control, special effects, etc.), and the
control of the image computer. The Scene Manager
sends back IG and scene status to the simulator. sync sync

- Database management: The Scene Manager supports The design of these processors provides the better
the storage of the visual database and geospecific performance, and the derating in the pixel processing
phototexture patterns, as well as auxiliary databases. At capacity versus polygon count which is usually
the mission initialization, the necessary part of the encountered on such systems is therefore suppressed.
database is loaded in the Image Computer. A real time
paging of the terrain database is then performed
accounting for the movement of the ownship. 3. DATABASE
Large territory coverage requires an optimization of the
- Mission support functions: Though not really in the
database storage and access, along with a real-time
concern of the image generation, the Scene Manager
performs several mission support functions such as paging process.
height above terrain determination, terrain following
for terrestrial objects, collision detection for the 3.1 Terrain database
ownship and the moving models, multiple point range The terrain is described as a Triangle Irregular
finding and intervisibility. Using auxiliary databases, Network (TIN) with vertices coordinates and associated
the Scene Manager benefits from the knowledge of the attributes The TIN is preferred to the regular mesh
scene context and terrain characteristics. which, due to the uniform density of polygons,
generates a higher mean-error of position and altitude
2.2 The Image Computer for the same volume of data, and can only provide a
coarse representation of the relief. For load control
The Image Computer is composed of four processors,
purpose, the database is stored under several Levels Of
the Extraction processor, the Geometry processor, the
Detail (LOD) descriptions.
Pixel processor and the Video processor.
The terrain is textured using satellite photographs.
An image is computed in one frame period, based on
Fine levels of detail will preferably be covered with
parameters transmitted in the previous cycle, and it is
aerial photography, giving a lower step per texel.
displayed in following period. Three processors (the
Texture pattern are stored in the image processor and a
Extraction processor, the Geometry processor and the
real-time paging of texture is performed accounting for
Pixel processor) participate in the computation of one
the movement of the ownship, in the same way as for
image, while the previous image is displayed by the
terrain geometry.
Video processor.
 7-3

3.2 Object database For each object, the appropriate LOD is selected
accounting for the field of view, the distance to the
Objects are not explicitly described in the terrain observer, the angular offset in the field of view (for
database. Only their identifier and position, as well as wide field of view configuration with AOI), and some
some layout parameters, are stored in this database. criteria like operational interest (an ordinary tree or
Furthermore, the object database volume is greatly house will be sacrificed earlier than a designated
reduced in so far as each object instance costs only few target).
words instead of the full object description.
Objects pop-up is avoided using size and transparency
This gives a local database capacity of up to 650 000 blending. A blending between two intermediate LODs
terrain polygons and 350 000 object instances, that are can also be done, but is time consuming. In fact, the
stored in the Extraction processor memory. best result is obtained modeling each object LOD in
For the real-time object's instantiation, an object such a way that, in nominal conditions, the detail
library is available in the Extraction processor. This modification is unperceptive at the transition.
allows rapid and easy database creation and
modification, particularly regarding the object density. 4.2 Terrain interpolation
For operational purposes, objects can be put on the The terrain database is stored by rectangular block
terrain at the time of the simulation initialization. entities. Depending on their range to the observer, and
operational considerations, each block of database is
3.3 Objects library extracted with the optimum LOD.
The explicit description of objects is done in the object In order to avoid abrupt transitions between successive
library. Eight LOD per object can be created, and LOD representations, a couple of coordinates is
stored in this library. Up to 4096 classes of objects are associated to each vertex in one level of detail
managed, with their levels of detail and an alternate description: the actual coordinate in the current LOD,
representation, for object's destruction purpose. and the corresponding coordinate of the same vertex in
the following LOD.
An interpolation is then computed on the coordinates
of vertices located between two higher and lower range
limits around the LOD transition range. The transition
between the different representations is then smooth,
and unperceptive in the nominal condition.

4.3 Animation
Independently of moving models, which are controlled
from the host computer with six to nine degrees of
freedom, objects in the database can be endowed with
individual dynamic characteristics.
- Self Animated Models: Each object in the database
Objects Library can be composed of several parts, some of them being
fixed, others being animated along predefined degrees
4. SCENE CONTENT of freedom. This allows the animation of some objects
(radar antenna, or windmill for example) without any
If a perfect rendering quality provides a visual comfort, requirement to the host computer or to the scene
the way the content of the scene and the interest of manager.
features are managed provides the "intelligence" of the
image. - Multiple Representation Objects: Some objects can
All is done in APOGEE to control the operational be modeled with up to 256 representations. Two kinds
quality of the scene content. of animation are possible using this capability:
automatic cycling on the different representations, or
sequential selection, on scene manager initiative, in a
4.1 Level of Detail Management
predefined list of representations (scenario). Troop
According to the distance from which the features are maneuvers, or dynamic object's destruction, for
observed, their description can greatly differ. This is example, can be simulated in this way.
used to define levels of detail for terrain and for
objects, in order to optimize the polygon capacity of 5. POLYGON PROCESSING
the image computer.
From the database to the input of the pixel processor,
The choice of a level of detail representation is done, the image generation is based on a polygonal
for all features, according to their visual aperture and description issued from the Extraction processor, then
operational parameters.
7-4

transformed in the Geometry processor.. Light points determines the useful part of the screen covered by
are computed in the same way. each polygon.

5.1 Extraction 5.3 Light points

Depending on the view point and the field of view, the For night scene, or landing aid simulation, for
visible blocks of terrain description, and object's example, the geometry processor generates lightpoint
instances, are extracted from the local database descriptors that are transmitted to the Pixel processor.
accounting for the above mentioned LOD criteria.
Two kinds of lightpoints are generated by APOGEE:
Selected terrain features and objects are extracted in
the direct visibility order to optimize the Pixel -Raster lightpoints, that are processed in the same
processor efficiency. way as the polygons. Their intensity is computed
accounting for the distance to the observer and a
The heritage of polygons is associated to the specific attenuation table. In the pixel processor, they
descriptors using a tag in which several fields are rendered as textured polygons the size of which is
corresponding to the origin of polygons (terrain/object, growing with the intensity in order to compensate for
object type, and specific properties) are concatenated the lack of video intensity range.
with the object number. This tag will be used later on
in the pixel processor to help the AZtec process. - Calligraphic lightpoints, that are processed in the
same way, down to the occultation stage where they are
only checked for occultation, without modifying the
image content.
In both cases, lightpoints can be omnidirectional or
not. They can have colored lobes to simulate some
landing aids like VASI.

6. PIXEL PROCESSING
The Pixel processor generates color attributes per pixel,
and resolves the occultation between polygons.
24 bit TAG The pixel processor consists in three main stages: the
Segmentation processor, the Rendering processor and
Tag generation the Occultation processor. A parallel architecture has
At the output of the extraction process, packets of been designed with two objectives: modularity in term
vertices, gathered per object or terrain block, are then of image resolution and in term of image complexity
transmitted to the geometry processor. and quality.
The pixel processing is then done by screen zones of
5.2 Geometry processing 64 x 64 pixels. The number of installed CIP boards is
One of the available CGP boards handles one packet at variable depending of the update rate, the image
a time, and performs the usual 3D geometry resolution and pixel processing requirements.Up to
computations -Change of coordinates, Clipping (if eight CIP can be combined.
any), Perspective projection - on each vertex. In the Each CIP board is composed of four identical
same time, the luminance is computed accounting for processing cells. Depending on the image definition
the normal vector and the light sources parameters. and the level of performance, a processing cell will
Vertices parameters are then known in the projection have a variable number of screen zones to process, with
(post-perspective) space. The polygon descriptors are a maximum of 32 zones per cell. Moreover, each
composed in a format dedicated to the rendering by the 64 x 64 zone can be processed by several cells in order
pixel processor, the pixel attributes (color, luminance, to average the computation load.
etc.) being described by the coefficients of their
equation in the projection plane. 6.1 Segmentation Process
In the pixel processor, the attributes per pixel will be The first step of the pixel process is to determine which
obtained applying the coordinates (X,Y) of each pixel pixels in the screen belong to the polygon. This is done
in the parameter equations. This eliminates the usual by the impact and the segmentation functions.
computation errors due to the bi-linear interpolation of - Impact: The impact function is located on each CIP
the parameters in the polygon. board. Using the bounding box parameters, it
The equation coefficients of polygon edges are determines which CIP board(s), and which processing
transmitted, associated to the polygon descriptor, along cell(s) in a board, has to deal with each polygon
with bounding box parameters. This bounding box will descriptor. An impact table is adapted to a Pixel
be used, in the the pixel processor, to speed up the processor configuration and an image resolution.
impact function and the segmentation process which
 7-5

Several CIP boards can process the same polygon, and The sub-pixel mask participation is based on its center
on each CIP, four cells can work on the same polygon, position relatively to the polygon edges, and is
depending on its trace on the screen space. corrected accounting for the area covered by the faxel.
- Meta-Zbuffer: During the segmentation process,
Processedi-^-^V hidden metapixels are eliminated from the dataflow by
polygon V^ / ( plus
bounding box using a Z-buffer function at the meta-pixel level. This
will be detained later in the occultation paragraph.
- screen
:reen;zone
6.2 Rendering
CARD 1, CELL 2
The rendering process is devoted to the computation of
CARD 2, CELL 4
the color components of the faxels. Two main types of
CARD 1, CELL 1
:■:■:■:■:■:■ Uü^
rendering are processed: color and luminance
CARD 1, CELL 3 smoothshading (Gouraud shading), and texturation
CARD 3, CELL 3 with luminance shading.
CARD 2, CELL 1
- Smooth shading: Four component equations are
processed (L, R G, B), in which the X,Y coordinates
Screen Impact function of the faxel sampling points are applied.
-Segmentation: All the cells concerned by one A scene attenuation factor is transmitted apart in
polygon perform a recursive segmentation of this floating format, the color components being computed
polygon into metapixels (blocks of 16x16, 8x8 pixels, at the maximum level of illumination.
etc.), down to the sub-pixel mask per fragment of
polygon in a pixel. In the followings, the polygonal This attenuation factor is used in the last stage of the
participation to one pixel will be called "Faxel. pixel process, and permits the correct rendering of
night/dusk scenes with light lobes illumination.
- Texturation: Full color photographic texture can be
applied on all polygons.
Patterns can be used with up to 1024 x 1024 texels
definition. As they are processed in normalized
Nth level array coordinates, texture patterns can be laid out regardless
their original definition.
Full Meta pixel -> to the Rendering stage
Texture patterns are described in full color mode
D
Secant Metapbcel -> next segmentation level
Empty Meta pixel -> to be discarded
put Resulting
Faxel Mask
(R G, B) plus transparency, and a global transparency
factor can be applied on a textured polygon.
Edge by edge Segmentation process Modulation patterns can also be applied. Photographic
The polygon segmentation is done, edge after edge, texture is processed using the MTP-map algorithm.
using the edge equations. Full metapixels are directly In the texturation process, three component equations
transmitted to the rendering stage, while edge (L, Tx, Ty) are computed. The L parameter is used for
metapixels are segmented again to generate lower size the smooth shading. Texture coordinates (Tx, Ty)
full metapixels or edge metapixels or, at the last step, permit to obtain, after inverse perspective computation,
faxel masks of 4 x 4 sub-pixels. the coordinates of the processed faxel in the original
texture space. The texture number is then concatenated
Hierarchical Segmentation to these coordinates to address the texture memory of
from 64 x 64 pixels Screen-Zone
to 4 x 4 Sub-pixel Mask which the color components R G, B and a
Processed transparency parameter are extracted.
polygon
-MBP-map: Texture being rendered with MTP-map,
the texture pattern are stored under several decreasing
4x4 J—'->. resolution maps obtained by successive lowpass
Metapixel \ filtering. A gradient of texture address is computed per
and so on, pixel, in order to determine which optimum texture
down to the
subpixel mask
couple of maps must be addressed. The fractional part
of this gradient is used to interpolate between the two
4*4 selected maps. This interpolation avoids the visible
subpixeb transition band when only one level is used at a time. A
better rendering is obtained with a bi-linear
16x16 interpolation in each map, using the fractional part of
Metapixel the texture addresses, before interpolating between the
two selected maps.
7-6

-Sharpening: When the higher resolution map is This sort is not a heavy constraint in so far as it is
used, and the texture map is largely oversampled, the limited to objects, and not polygons. The AZtec pixel
image has a blur aspect. A sharpening function is then processing load is thus greatly reduced. This
engaged to render the boundaries of the texels, in order improvement is particularly appreciated in sensor
to restore attitude and movement cues. The sharpening configurations with narrow field of view, where very
level is a programmable parameter. high depth complexity can be encountered when
several objects intersect the line of sight, and occupy
- Microtexture: The above conditions can be easily most of the screen area.
obtained when flying nap of the earth, or running on
the ground. Along with sharpening, better cues are 6.3.21 AZtec
obtained by adding higher detailed microtexture. The
microtexture is a generic pattern, matched with the AZtec is the most significant improvement in image
original ground pattern, which gives a gain in the generation. Derived from the well-known Z-buffer
range of texture use. The 10 MP-map levels can then algorithm, with A-buffer improvements, AZtec can be
be virtually extended to 15 or 20 levels. qualified as a "multiple Z-buffer with adaptive
blending". Based on the management of several layers
6.3 Occultation of 4x4 sub-pixel mask, area and color parameters per
pixel, it takes advantage of the tag generated at the
A major problem in the Computer Image Generation is Extraction.
the hidden parts elimination. The well-known Z-buffer
algorithm, largely used in most image generators, has AZtec technology is not only a solution to the
been criticized for its inherent difficulty to perform occultation problem, but it participates to the final
antialiasing, and its sensitivity to depth complexity. rendering of the image, particularly in the lighting and
Two functions have been implemented in APOGEE: shadowing effects, and allows for nice rendering of any
a Meta-Zbuffer at the Segmentation level, and AZtec, a number of opaque, transparent, or lighting polygon
Sogitec proprietary algorithm. participations, in any order.

Polygon descriptors
from the Geometry Processor

Segmentation META Z-buffer

processor MEMORY

Rendering processor ATTENUATION Tables

I
Texturation processor TEXTURE Memory

I
MULTI-LEVEL
Z-buffer MEMORY One pixel location One pixel location
AZtec processor Without fusion With isofusion
Double-buffer
IMAGE MEMORY Multilevel Z-buffer
Meta Z-buffer In fact, AZtec runs with a eight level stack of faxels,
regardless their processing order or whether they are
6.3.1 Meta-Zbuffer opaque, transparent, shadows or lighting part of
polygons.
The efficiency of the Z-buffer is improved using a
Meta-Zbuffer algorithm in the image processor. Obviously, all faxels that are hidden by a full opaque
faxel (16 opaque sub-pixels) are eliminated, in the
The Meta-Zbuffer consists in the management of an same manner as with ordinary Z-buffer. Only faxels
early Z-buffer in the Segmentation process. It starts at that have at least one visible sub-pixel (opaque or
the 16 x 16 metapixel level. The Meta-Zbuffer runs transparent) remain active in the stack.
similarly to the Newell algorithm, and it is maintained
in the different stages of the segmentation process, Complex situations with a concentration of details plus
down to the pixel level: During the segmentation steps, two or three transparent layers for atmospheric effect
all metapixels that are farther than previously stored or special effects could lead to an overload of the eight
metapixels are rejected from the data stream. levels stack. To avoid such cases of overload, a fusion
process has been implemented in which two levels can
The Meta-Zbuffer function is all the more efficient that be merged into a single level.
polygons are presented to the segmentation process in
the visibility order. This is guaranteed by the extraction Three types of fusion are performed in AZtec: Isofusion
processor where features are sorted. Multitexture fusion, and Forced fusion.
 7-7

- Isofusion: In this case, two faxeis are merged into a If the stack is full when a faxel has to be processed
single faxel accounting for several conditions: with no isofusion capability, Forced fusion is
performed. Accounting for the depth test and
. The two faxels must have the same TAG occultation process, the incoming faxel may be
. They must have non overlapping part discarded or inserted in the stack, in which case the 7th
. The color distance between the two faxels must be and the 8th faxels in the Z-hierarchy are considered.
lower than a predefined threshold One of these two faxels is discarded, or merged with
the other, depending on whether they are opaque or
transparents. In all cases, the best compromise is
adopted.
Same TAG Forced fusion is a remedy to deal with extremely
complicated situations, with a conglomeration of
AND objects the edges of which interfere in a pixel.
Generally, this occurs far from the observer, and the
1 Non overlapping
jq approximation on the final color has no significant
i?w
^ Masks ^ I effect in the result. The main problem is to avoid brute
rejection of faxels. The forced fusion is efficient in
fc local, and not much likely to occur, accumulation of
AND transparent faces. In this case, transparent polygons are
generally overriding the rest of the polygons, and
C1#C2 therefore are not affected by forced fusion.
Similar color
6.3.3 Lighting
Isofusion criteria In the final step of the AZtec process, the content of the
multilevel Z-buffer memory is converted into color
The isofusion helps to maintain the stack under the components accounting for the optical path. The 4x4
maximum number of levels without any compromise sub-pixel masks are combined with their opaque or
on the image quality. The last condition ensures that transparent attributes.
two faxels with contrasted colors are not simply
merged, though belonging to the same object, in which At this level, true lighting conditions are computed.
case a color error could occur if one of the two faxels The rendering of shadows, or the lighting due to
had to be hidden later. headlights, landing lights or searchlights, are applied
to the faxels hit by the light sources. The color of faxels
-Multitexture fusion: This is a particular case of is computed in full daylight condition throughout the
isofusion where the two masks may overlap. Color are rendering pipeline. Dusk or night attenuation is
blended in the ratio assigned to each texture. Any applied at the last stage of AZtec, and modified with
number of textures can be applied to one polygon, the the lighting faxels, the intensity of which have been
result taking only one level in the stack. determined using the true 3D distance and a dedicated
- Forced-fusion: When none of the above mentioned attenuation table.
fusion process can be applied, the incoming faxel is Several chromatic sources can be processed. Their
stored in the a new stack level. One must remark that illumination is added, per color component, on the
transparent faces occupy one level each as long as the pixel basis; giving the capability to restitute colored
image computation is not complete. light effects according to the color of the features (only
red and green objects will be highlighted by a yellow
9th Faxel 8th and 7th levels
light, for example).
fusionned The final color components are stored in a double
8th level buffer image memory.
7th level
7. VIDEO PROCESSING
6th level
5th level 7.1 Video output
4th level The basic function of the Video processor is to convert
3rd level the content of the image memory into analogic video
signals, according to the selected video standard. The
2nd level ^$ image memory content is generally not used straight
1st level out. Some operations can be done on the raw image.
Forced fusion A contour mask can be applied in order to facilitate
multichannel projection integration. In the same time,
a color selective attenuation can be applied for Their efficiency through haze, clouds, or some partially
multi-channel projector/screen harmonization. The transparent objects, though targets are not visible in the
color components are finally modified using a look-up final image, can be simulated. Double echo laser range
table for compensating variable display Gamut. finder can thus be simulated.

7.2 Symbology
8. CONCLUSION
Symbology is often useful in some simulation
Optimization have been done, at all stages of the
applications. It is mainly composed of various text and
APOGEE image generator, which provide the best
vectors, together with application specific symbols.
performance on all Image Generation criteria, and in a
This symbology is generated by the video processor,
broad range of applications. An efficient scene
and inserted in the video signal under fifteen
management, together with improved rendering
programmable colors. algorithms, participate in the generation of realistic
images, with a great content complexity and a high
7.3 Calligraphic light points level of quality.
Calligraphic lightpoints are not displayed in the While most of the deficiencies of the original Z-buffer
standard video signal. The occultation status of the algorithm have been definitely suppressed, appropriate
calligraphic lightpoints is read in the multilevel trade-off have been made to allow hardware
Z-buffer memory by the Video processor in order to implementation of a A-buffer like algorithm,
display only the visible lightpoints. A dedicated improving performance together with image quality.
interface board is added to the Video processor to
transmit the lightpoint data according to the different Significant breakthroughs in real time image
multimode (raster/calligraphic) projectors. generation have been made by introducing new
features such as multiple, true pixel range, colored
7.4 Sensor Simulation Aid light lobes and multiple transparency, without adding
sorting constraints as encountered in other systems.
The image reading capability is used in the simulation
of thermal imagery systems (detection and tracking of The final product offers modularity in terms of
contrasted area). As the pixel history remains available performance and image quality. Only four types of
in the multilevel Z-buffer memory at the end of the boards are used for all configurations, providing a
image computation, it is used in the simulation of some great simplicity of maintenance, very few spare boards
systems (multiple echo laser range finding). and therefore, a low life cycle cost.
 The RTSS Image Generation System
K. Alvermann and S. Graeber
Deutsche Forschungsanstalt für Luft- und Raumfahrt
Institute of Flight Mechanics
Lilienthalplatz 7
38108 Braunschweig
Germany
E-Mail: alvermann, [email protected]

J.W.L.J. Mager and M.H. Smit

TNO Physics and Electronics Laboratory
Oude Waalsdorperweg 63
2509 JG The Hague
The Netherlands
E-Mail: Mager, [email protected]

1. SUMMARY
Main market demands for the visual system of a simula-
tor are photorealism and low latency time. RTSS, a gen-
eral purpose image generation module developed within
the European ESPRIT project HAMLET, can meet these
demands through the use of High Performance Com-
puting technology. This technology provides the needed
communication and computing power. Moreover, by us-
ing parallel processing, the whole system is scalable, i.e., Figure 1: The RTSS Concept
the same software and hardware design can be used for
small, cheap systems, as well as for high-end view sim-
ulations. This allows an easy adaptation to the user's RTSS has to generate photo-realistic images in real-time.
To realize this, the underlying hardware has to provide a
needs.
high computation power. Moreover, the data bandwidth
RTSS also includes an object and scenario editor imple-
needed to communicate the large data flows processed by
mented on a work station, as well as filters to other object
the RTSS places severe demands on the communication
data standards.
power of the underlying hardware.
This paper will give an introduction to the soft- and hard- RTSS is designed to be scalable. If the demands of an ap-
ware design of RTSS. It will then present the features of plication are rather low, a small (and, therefore, cheaper)
the system as well as the interfaces: the filters to import system can be used. If the demands get higher, the RTSS
external model data and the interfaces to the simulation can be expanded to fulfil the growing requirements. Re-
system itself. quirements are the image rate and the latency time, as
well as features like shadow generation, anti-aliasing, and
the size of the database.
2. RTSS
RTSS is divided into an off-line and an on-line part. The
2.1 Overview off-line part consists of an interactive object and scenario
Within the context of the European ESPRIT Project editor and filters to input data from known geometric
6290 HAMLET, TNO Physics and Electronics Labora- formats such as the MultiGen Flight format. The objects
tory (TNO-FEL, NL), Deutsche Forschungsanstalt für and scenarios are stored in a database in a file system.
Luft- und Raumfahrt (DLR, D), and Constructiones The on-line part inputs this database, i.e., data describ-
Aeronauticas (CASA, E) developed the Real-Time Sim- ing the geometry and appearance of objects, light pa-
ulation System (RTSS). rameters, image parameters, etc. RTSS then outputs the
The work in the HAMLET project is done within the corresponding image. As a second task, RTSS contains
framework of the ESPRIT program and partly funded a collision detection module. This module checks the
by the Commission of the European Communities. The moving objects for collisions and reports them back to
following companies form the HAMLET consortium: the application. During run-time, the application can
AEG (D), CAP Gemini (F), CASA (E), DLR (D), change almost all parameters, e.g., the position and ori-
HITEC (GR), INESC (P), INMOS (GB), Parsytec (D), entation of all objects and of the camera, light positions
Gabriel (GR), TNO (NL), and TU Munich (D). and parameters, etc.
The RTSS is a general purpose image generation mod- The interface to the application is a bi-directional chan-
ule. RTSS can be used by simulators, e.g., flight simula- nel. The application sends commands (parameters,
tors (see Fig. 1), and other man-in-the-loop applications movement data, object changes, etc.) to the RTSS and
which require visual feedback. receives in turn the results of the collision detection.

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?'
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
 8-2

2.2 Requirements and Architecture application and the database. It converts the object data
Using RTSS in man-in-the-loop applications leads to the into primitives (points, lines, faces, and shadow faces)
following main requirements which are sent to the Image Generation subsystem.
• real-time generation of images, The Image Generation subsystem (see chapter 5) pro-
• an ergonomic image rate, duces the image from the primitives supplied by Simula-
• a low latency time, and tion Execution. Textures are mapped onto the appropri-
• high resolution, photo-realistic images. ate polygons, the polygons are scanned line by line, and
the lines are rendered into an image buffer. Shadows are
Since no restrictions in the movement of objects are ac-
added and anti-aliasing techniques are used for the final
ceptable, the images cannot be pre-calculated but must
image.
be calculated in real-time. To guarantee a smooth move-
The on-line part of the RTSS is designed to be scalable.
ment of objects in the image, the image rate must be
Computational intensive processes (e.g., database han-
about 20-30 images per second (depending on the dy-
dling, scan conversion, texture mapping, shadow gener-
namics of the application). A low latency time is dictated
ation) are done in parallel. The number of parallel pro-
when RTSS is used in man-in-the-loop applications. The
cesses is variable and determines the power of the system.
time between the command of a user in the application
and the display of the corresponding image, i.e., the la-
tency time, should be below 100 milliseconds to prevent
motion sickness or severe timing differences between the Scenario Simulation Image
Creation -* Execution -> Generation
simulator and the real thing. The image must be realistic
enough to serve as a visual orientation. It should contain A A
shadows as a visual clue for the position of objects, a high
resolution, and anti-aliasing techniques should be used to V V V
suppress edged lines and edged object boundaries. Display
User t Application
The above requirements led to the following specifications Device
of the RTSS. For the off-line definition and construction
of objects, RTSS provides: Figure 2: RTSS Architecture
• an interactive, three dimensional scenario and object
editor; 3. SCENARIO CREATION
• import of external model data (MultiGen, AutoCAD, Before a simulation can be executed, the scenario of this
etc.); simulation has to be defined using the scenario creation
• static and dynamic objects with several levels of de- tools. These tools provide means to import object models
tail; from external modellers, to edit the characteristics of the
• hierarchic grouping of objects; objects contained therein, and to edit the scenario itself.
• objects are built up from points, lines, and planar Scenario creation is done off-line on a host computer.
polygons;
• polygons can be textured; 3.1 Import Filters
• multiple dynamic directional light sources. The import filters provided in RTSS enable a user to
During run time RTSS reports: import geometric object models created by existing com-
mercial modellers. The imported model data is converted
• collisions between objects on polygon level.
into the internal format used by RTSS keeping as much of
The following features are used for the resulting image: the semantics of the model as possible. Currently, RTSS
• light reflection and emission using multiple light sour- supports models made using MultiGen in the Flight for-
ces and several lighting models; mat as well as AutoCAD.
• shadow generation;
3.2 Object Editor
• anti-aliasing and depth cueing (fog simulation).
The user can interactively modify the visual character-
The performance specifications are:
istics of an object using a built-in object editor. This
• true colour (i.e., 24bit) images at high resolutions; editor consists of two parts: a material editor and a tex-
• 25 images per second; ture editor. The material editor can be used to combine
• a latency time of 80 milliseconds. several characteristics of materials. A material defines
transparency, lighting type (emitter or reflector), shad-
2.3 Architecture
ing type (Flat, Gouraud, or Phong), and reflection coef-
RTSS is decomposed into three subsystems, correspond- ficients for ambient, diffuse, and specular lighting. In the
ing roughly to the three companies involved in its devel- texture editor, the user can change the material of each
opment: the Scenario Creation (CASA), the Simulation surface of an object and apply textures to these surfaces.
Execution (DLR), and the Image Generation (TNO) sub-
system (see Fig. 2.) 3.3 Scenario Editor
The Scenario Creation subsystem (see chapter 3) is the Using the scenario editor, the user finally specifies the
off-line part of the RTSS and includes the scenario editor, scenario contents. It contains the initial position of the
the object editor, and the external model data import elements (assemblies, cameras, and light sources) in the
module. The resulting scenarios are stored in files to be scene. An assembly is a hierarchically organized set of ob-
loaded by the on-line part of the RTSS. jects with fixed relative positions. For an assembly, the
The Simulation Execution subsystem (see chapter 4) user can specify whether it is shadow casting and visible
loads the scenario, handles the communication with the or not and its collision detection list. During simulation,
 8-3

the assembly is tested for collision with all assemblies 4.2 Collision Detection
identified in this list. The user can define the type of a The Collision Detection module is an optional part of
light source (ambient or directional) and the values cor- the RTSS. It detects overlapping geometries of objects on
responding to this type. For a camera, a user can define polygon level, i.e., the module reports polygons of differ-
its aperture angle, view direction, and clipping planes. ent objects that intersect each other. For each assembly
Most of the properties of the elements can be changed it can be specified with which other assemblies collisions
during run time. Therefore, the scenario editor enables are checked. This can be used to reduce computation to
a user to specify whether during simulation changing the the interesting objects (one space ship making contact
properties is allowed (element type is dynamic) or disal- with another) and to avoid unnecessary collision reports
lowed (element type is static). Apart from the properties which will occur in any case (a car moving on the street).
of the elements, the scenario also contains global informa- Collision Detection is decoupled from the rest of the sys-
tion, so-called session parameters. These include param- tem (for input as well as output), i.e., the reports to the
eters for the lighting model, screen resolution, depth cue- application are not coupled to the frame rate.
ing, background colour, initial camera, and settings for
switching on/off edge anti-aliasing, texture anti-aliasing 4.3 Database
and shadow generation. The depth cueing parameters as The whole database is distributed over a number of pro-
well as the background colour can be changed during run cessors, the Geometry Handlers. During a frame the Vis-
time. While defining the scenario, the user can inspect ibility Check issues object identifiers of visible objects
the scene using a 3-D viewer. This shows the scene from and shadow casting objects to the respective Geometry
a global view point or from the view point of the initial Handler. Objects not visible in the image or not cast-
camera. In this scene icons indicate the position of light ing a visible shadow are discarded as early as possible.
points and cameras. The Geometry Handler converts the object into primi-
tives (points, lines, faces, and shadow faces), transforms
4. SIMULATION EXECUTION these primitives to the correct coordinate system, and
The Simulation Execution subsystem is responsible for clips them against the view volume. The lighting model
the interface to the application and the handling of the is evaluated for vertices and faces. For shadow casting
database. The individual modules of the subsystem are objects shadow volumes are calculated and decomposed
shown in Fig. 3. into shadow faces. All primitives are collected in a buffer
and broadcasted over a bus to the Image Generation sub-
system upon a system synchronization signal. For details
of the algorithms see [1].
> r > The number of the Geometry Handlers is a parameter of
Communication Matrix Visibility the system design. Systems using small databases need
Handler -> Calculation -> Check only one or two Geometry Handlers, while systems using
> 1 big databases can use many. The number is not limited

>
Collision
1
Geometry
>/
Geometry
by software or hardware restrictions.

Detection Handler Handler Pixel Pixel Pixel

A 1 1 k Processor Processor Processor
I 1-

Figure 3: Structure of Simulation Execution

4.1 Interface to Application

Colour Graphics ^ Display
Display ? Device
The interface to the application is a bi-directional chan-
nel. At the begin of a run, the application sends the name
of the scenario to be used, which is then loaded from files Figure 4: Structure of Image Generation
(thus, not through the application). During run time,
the application sends commands. Commands are: 5. IMAGE GENERATION
• parameter changes: lights, background colour, depth 5.1 Operating Principle
cueing; The output image of RTSS is divided into a number of
• camera changes: position, orientation, parameters; horizontal scan lines, each of which is divided into a num-
• movement data: position, orientation, and scale of ob- ber of pixels, thus forming a rectangular raster. The Im-
jects; age Generation process is responsible for the determina-
• object data: colour and material of objects; tion of the colour of every individual pixel. This is based
• collision detection: which object pairs are checked for upon the render primitives, which are the output of the
collisions. Simulation Execution subsystem (see chapter 4). First,
Commands have to be issued in a specific protocol. There all the render primitives are rendered in the frame buffer.
is no restriction on the timing of the commands, since Then all the shadow face descriptions are processed into
the reception is decoupled from the rest of the system. a separate shadow buffer. Finally, the frame and shadow
Therefore, they may be sent at any time. buffer are merged into the final image buffer. If there are
RTSS in turn sends back the result of the collision detec- no shadows present in the scenario, the last two stages
tion. Collision detection, and thus the feedback, can be are skipped and the frame buffer is considered to be the
switched off it it is not needed. image buffer.
8-4

5.2 Parallel Processing a PowerPC connected to a T425 transputer for communi-

The computations involved in Image Generation are the cation. However, the bandwidth needed to transport the
most time consuming of RTSS. Therefore, an efficient primitives and the image cannot be met by transputer
way of decomposing this process into parallel tasks is re- links. Therefore, a special bus system called Transputer
quired. The decomposition of Image Generation is shown Image Processing (TIP) bus is used.
in Fig. 4
6.1 Processors
The output image is divided across the processors execut-
The processor board TPM-MPC is equipped with a Pow-
ing Image Generation (these processors are called Pixel
erPC 601 processor running at 80 MHz. An INMOS T425
Processors) in a scan line interleaved fashion. Suppose
transputer running at 30 MHz is used for communication
the number of pixel processors to be N. Pixel processor
along the 4 transputer links operating at 20 Mbit/s. The
i (with i < N) 'owns' scan line i and then i + N and so
transputer is equipped with 4 Mb of local memory and
on. This way, every pixel processor has roughly the same
shares 16 Mb of memory with the PowerPC. Using the
number of scan lines (and therefore pixels) to compute,
transputer links, these boards can be connected in any
thus guaranteeing a uniform workload.
kind of network. The PowerPC and the transputer are
The primitives are received via a bus to which all Pixel programmed in C using the PowerTools which are an ex-
Processors, as well as the Geometry Handlers, are con- tension of the INMOS C toolset.
nected. After every Pixel Processor has finished, the
partial images are gathered in the video memory of the 6.2 Communication
Colour Graphics Display for display. This gathering is The high communication demands required by the com-
done across the same bus. The number of Pixel Proces- munication of the primitives (see section 4.3) and the
sors can be adapted to the required performance of the image (see section 5.2) cannot be met by using the trans-
system. It is not limited by software or hardware restric- puter links (about 2 Mb/s): a 512 x 512 true colour im-
tions. age with 25 images/s needs a bandwidth of 25 Mb/s; the
respective 1024 x 1024 images needs 100 Mb/s. Addi-
5.3 Algorithms tionally, the primitives are broadcasted over the bus.
Visible priority of occluding primitives (also called Hid- The demands are satisfied using the TIP-Bus. This bus
den Surface Elimination, HSE) is treated using the z- provides the hardware to transfer data from the local
buffer algorithm. Almost all major visual systems use memory of one processor to that of another. The bus is
this (or adaptations of it) for HSE. The algorithm has 32 bit wide and has a peak bandwidth of 120 Mb/s. The
the advantages that it is bus architecture allows point-to-point communication as
• Independent of the order in which the primitives are well as data broadcasting and gathering (which.is needed
treated. This is of particular importance in the RTSS to distribute the primitives to all Pixel Processors and to
case, because a global priority ordering of the prim- gather the partial images on the Display Processor). If
itives (which would be needed otherwise) can not be necessary, the bus can be divided into several segments
performed by the Geometry Handlers (when imple- working in parallel.
mented on more than one processor). Therefore, this The TIP-MPC boards are equipped with a PowerPC 601
ordering would have to be executed by the Pixel Pro- operating at 80 MHz, an INMOS T425 transputer run-
cessors, which is are already heavily loaded with work. ning at 25 MHz as the bus controller, 16 Mb of shared
• Capable of handling all sorts of render primitives. memory, and 2 Mb of video memory, which is the in-
To prevent disturbing noise in the image, two forms of terface to the bus. Any number of these boards can be
anti-aliasing have been applied: connected to one bus.
• Lines and the edges of polygons are smoothened using Special interface boards are available to connect the TIP-
the a-buffer algorithm, described in [2]. This tech- Bus to display systems or cameras. RTSS uses the Colour
nique filters the well known 'staircase jaggies'. We Graphics Display. The CGD is equipped with an INMOS
have chosen this method from a number of alterna- T805 transputer and a video chip to drive a display de-
tives, because it is suited for a z-buffer set-up, only vice.
uses extra computing power for filtering of the edges, The CGD is also programmed in C using the PowerTools.
is independent of the sub-pixel resolution used, and The TIP-Bus can either be programmed directly or using
does not require a major amount of extra memory. a special language called TIP-Set.
• A technique called mipmapping is used to filter the in-
side of textures. Almost all commercial visual systems 7. REFERENCES
which incorporate texture anti-aliasing use mipmap-
1. Foley, J.D.; van Dam, A.; Feiner, S.K.; Hughes, J.F.,
ping. Refer to [3] for details on this technique.
"Computer Graphics", Addison-Wesley Publishing
Company, Second Edition, 1990.
6. HARDWARE 2. Carpenter, L., "The A-buffer, An Antialiased Hid-
The demands for the on-line part in terms of calculation den Surface Method", Computer Graphics Vol. 18,
power, communication bandwidth, and scalability, are No. 3, July 1984 (SIGGRAPH 84).
satisfied by using parallel processing. To guarantee along 3. Williams, L., "Pyramidal Parametrics", Computer
life cycle and easy expandibility the hardware developed Graphics Vol. 17, No. 3, July 1983 (SIGGRAPH 83).
inside the HAMLET project was used. The processor is
 9-1

Ship Airwakes - A new Generic Model for Piloted Simulation

A A Woodfield B N Tomlinson
Woodfield Aviation Research Flight Dynamics & Simulation Department
9 Colworth Road Defence Research Agency
Sharnbrook Bedford MK41 6AE "
Bedford MK44 1ET United Kingdom
United Kingdom

SUMMARY effects such as the magnitude of a downdraught leaving

The disturbed airwake in the lee of the superstructure of insufficient power available to control descent safely,
ships with an aft flight deck causes significant control the crosswind being so strong that there is insufficient
problems for helicopters operating during strong winds. roll cyclic or rudder control to maintain the proper
Providing adequate data to describe these airwakes for heading relative to the ship, and rapid changes of wind
use in flight simulators has been a difficult task from in any direction over small distances that can make it
either wind tunnel tests or theoretical airflow estimates. difficult to achieve the necessary precision of control
In this report an empirical mathematical model is needed for landing safely on a small moving deck.
developed based on the main types of airflow pattern Thus for any combination of ship and helicopter types
present in airwakes. The model uses a single set of there are Ship Helicopter Operating Limits (SHOLs) that
modelling parameters to produce airwake data for any specify the safe boundaries of Wind over Deck (WoD)
shape of ship with an aft flight deck and for any wind strength and direction. These have implications for
direction. Assessment of the model in a piloted flight operational effectiveness because the need to provide
simulator has been very favourable with the main acceptable WoD strength and direction for helicopter
criticism being the absence of turbulence, which is operations can restrict the freedom of the ship in speed
available in the model but could not be used because of and direction in moderate or strong natural winds.
difficulties in simulating helicopter responses at that
time. Comparison of airwake estimates with a set of Much effort both now, and over many years past, is
boundary layer wind tunnel results shows good directed to
agreement in the complex flow with the wind from a) improve prediction of SHOLs to aid in ship and
starboard at 30°. and suggests some areas where helicopter design
adjustments to the model could improve the comparison
for other wind directions. b) provide simulator training for adverse WoD
(and ship motion) situations.
The model has highlighted several interesting features of
airwakes, including probable improvements in airwakes Testing with ship models in wind tunnels has increased
if hangar tops can have porous net fences around the knowledge of airwake conditions, and attempts are
edges, the effects of gradients in vertical airflow being made to use the increasing power of
velocities on control of a helicopter's height, and the Computational Fluid Dynamics (CFD) to generate
need to include response to flow gradients in helicopter computer models of the airflow around ships. However
aerodynamic models for simulation of operations in the complex shapes and separation of airflow around
airwakes. many corners make it very difficult to apply the
principles of CFD and generate realistic computer
It is recommended that the airwake model should be models. Several papers in the proceedings of the
validated against full scale and wind tunnel data on AGARD Flight Mechanics Panel Symposium on
other shapes of ship, and, after any necessary 'Aircraft Ship Operations' held in 19911 addressed
adjustments to the model, it should be used for a trial operational and modelling issues relating to helicopter
installation on a suitable existing helicopter training operations from small ships.
flight simulator, prior to introduction into general use
for R & D in helicopter/ship research programmes and One of the most powerful tools for developing our
for helicopter training simulators. understanding of the relationship between complex
airwakes and their effects on a helicopter would be
1. INTRODUCTION piloted flight simulators2. However, a simulator can only
The disturbed airwake in the lee of a ship superstructure be used to study such effects if it has a realistic model of
passes directly over the aft landing areas used by the airwake and can satisfactorily represent the
helicopters on small ships, and there are also updraughts behaviour of the helicopter. It is also essential that the
over the edges of the flight deck. In moderate and strong airwake model can provide instantaneous local wind
wind conditions from some directions these airwake information at the helicopter position in real-time, which
disturbances can be sufficient to make it unsafe to land in practice means it must be calculated within a few
or take-off. The main problems arise from airwake milliseconds.

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?"
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
9-2

One way of providing such data would be to tabulate of the helicopter, such as downwind of the hangar edges
measured model data (or full scale data if available) and in the vicinity of the ship. Definition of these gradients
interpolate between the data points to the actual may require spacings of around lm.
helicopter location. Unfortunately, model data is
generally obtained at relatively widely spaced points,
and flow visualisation demonstrates that there are
complex changes in flow between these points that will
not be adequately represented by simple interpolation.
Also, such data is totally specific to the configuration
tested and cannot be readily adapted to account for
changes in ship structure. Eventually it may be possible
to generate data maps from CFD calculations, but in the
meantime there is a need for some form of mathematical
airwake model that can represent the main flow
characteristics in the airwake in a way that can readily
account for significant structural configuration changes
or even different ship types. Any empirical model will
need to be validated against wind tunnel test results and
full scale data. It will then provide a firm basis for
studies of SHOLs. provide a representative environment
for studies of operational developments such as landing Figure 1: Merlin helicopter on the aft flight deck of a
aids, and also provide a model for use on training Frigate
simulators.
Such a mathematical model has been developed for the When considering effects at this scale and greater, it is
Defence Research Agency with funding from the UK reasonable to look at empirical modelling based on the
MoD Defence Research package 25c, and initial datasets physical characteristics of the flow in order to represent
have been tested in a piloted simulation study. This the required conditions. This empirical modelling can be
report describes the model; presents some typical based on observed forms of airflow around individual
examples of airflow datasets; presents data on the body shapes and the sizes of the parts of the body
behaviour of a simulated helicopter in the airwake; and geometry that dictate the scale and shape of the airflow.
describes a future work programme to exploit this From this it will be possible to calculate flow
airwake model in training and research simulators. disturbances for more complex body shapes.
Ship structure consists of combinations of different body
2. MODELLING ISSUES shapes, e.g. hull and hangar, on a ground plane (the sea).
Airflow around and downstream of bluff objects is Figure 1. Thus it is necessary to combine airflow effects
extremely difficult to model because the air cannot turn from different body shapes to obtain overall flow
abruptly and separates from surfaces. In addition, it is disturbances. In general such airflow effects can only be
also difficult to produce fully representative data from added together in a simple way if the effects of one body-
wind tunnel tests because of scale effects (mainly component have largely disappeared before the effects
Reynolds number effects), physical difficulties in of the next body component will become significant.
representing some important small features at model Fortunately this is a reasonable approximation for many
scale, and the sensitivity of airflow disturbances to shear typical ship structures in the region of a flight deck
and turbulence present in the wind before it encounters because of the size of open area needed for aircraft
the ship. A concise summary of the problems and the operations.
current state of measuring techniques and knowledge is
given by Garry3. An empirical approach to modelling using these
assumptions is likely to provide good estimates of the
When establishing SHOLs, both the effects on main features in airwake disturbances once it has been
helicopter performance of average local airflow matched against appropriate full scale and wind tunnel
increments across the rotor, and also the effects of rates experimental measurements. It should also be effective
of change of airflow with distance (airflow gradients) on in estimating the effects of changes in geometry as long
the control of helicopter position, need to be considered. as the basic assumption that effects are additive is still
The spacing between datapoint locations must be close appropriate, which is a reasonable assumption for the
enough to define average values, which requires present generations of ships with flight decks.
horizontal spacing at intervals of approximately rotor
radius, i.e. approximately 5m for typical helicopters that The present model is developed for disturbances aft
operate from small ships. Spacing must also be close from the forward edge of an aft flight deck (often the aft
enough to define gradients wherever velocity changes face of a hangar) and extends aft until all significant
and gradients are both large enough to affect the control flow disturbances that might be felt during an approach
 9-3

to the ship have dissipated. The principles could be but only reduced (unless the disturbance is greater than
applied to flat top ships (aircraft carriers) but some the WoD).
details and values of primary constants used in the Ideally, models would be developed from the results of
modules would be different. tests to identify the effects of different shapes on the
3. DESCRIPTION OF THE MODEL
3.1 General principles
A basic ship geometry with a hangar in front of an aft
flight deck (Figure 2) is used to describe the structure
that causes flow disturbances. It is assumed that effects
from other superstructure further forward than the
hangar will generate turbulence and a momentum loss
but no significant discrete flow disturbances with a
wavelength of 10m or more. This assumption should be
examined for specific ships, but is reasonable in most
cases. The model includes effects when hangar sides are
not vertical, which is quite common; also included are
effects of any fence(s) on the roof of the hangar, and of
folded stanchions and safety netting around the flight
deck.
All aircraft, including helicopters, are sensitive to
changes in airflow, and particularly changes in vertical Figure 2: Basic ship structure
airflow. The significance of airflow patterns is often
best illustrated by plotting significant disturbances
to airflow as incremental changes rather than the
overall airflow. This is particularly true for vertical
components of airflow, where a few knots downflow 6 -

require significant increases in power to prevent a

helicopter descending (lkn downflow « lOOft/min).
As an example Figure 3 shows vector plots in a 2
vertical plane for total vectors and disturbance
vectors normalised by the WoD, where the
maximum disturbance is 30% of the WoD which is
horizontal. If this WoD was 30kn then the peak
updraught would be 9kn (« 900ft/min), which is
very significant, but it does not seem particularly
noticeable in the total vector plot. Similarly,
changes in the horizontal wind are not very obvious
in the total vector plot, but they are clear in the plot Disturbance vectors
of the disturbance vectors. Thus all vector plots in
this report will represent disturbance vectors. The
model calculates disturbances rather than total
velocities.
It is important to remember that a disturbance
velocity vector must be added to the free stream
(WoD) vector to obtain the actual local wind vector. ;
As a simple example there is no horizontal
component of flow along the deck at the aft face of
the hangar as air cannot pass through the solid 0 L.
surface of the aft face. Thus for WoD from the bow
the disturbance velocity at the aft face of the hangar
will be the same as the WoD but in the opposite -8-6-4-2 02468
direction. The total horizontal wind velocity at that x
point is then zero, and the disturbance vector is
equal and opposite to the free stream WoD vector. Total vectors
In general a disturbance vector showing a horizontal
component in the opposite direction to the WoD Figure 3: Comparison of vector maps for Disturbance and Total
does not mean that the horizontal wind is reversed Wind vectors
9-4

airflow downstream of relevant bluff bodies. However,

there are apparently no such data for ship shapes. Whilst
the majority of tests of the effects of airflow on
buildings are concerned with the loads on the building,
there is some limited data, which concentrates on
airflow on the ground around buildings where the main
interest is to avoid blowing people over. One of the most
useful references is 'Wind Environment around
buildings' by Penwarden & Wise4 where some
parametric data is presented. Airwake models have been
developed for this present programme using the above Figure 4: Separated corner flow
reference as a guide in combination with interpretations corner the wind forms a vortex that increases in
of characteristic flow patterns past ships and other bluff strength along the edge and then translates
objects shown in test results for specific shapes. downstream with the wind and decays slowly
It is assumed, as is normal practice, that all velocity (Figure 5). This flow is Vortex corner flow.
increments are proportional to the magnitude of the
Wind-over-Deck (WoD) and functions of the relative WoD
direction of the WoD to the ship. This relationship is
likely to be correct for moderate and strong winds. In
light winds there may be some changes in airflow
patterns but overall disturbances will be small because
of the low wind strength and so will the effects on
Figure 5: Vortex corner flow
helicopters. As a consequence it will be reasonable to
use estimates based on strong wind effects for training b) Lee rotor flow, which describes the horseshoe
or estimation of SHOLs. rotor (vortex) found in the lee of bluff bodies
The coordinate system for incremental disturbance (Figure 6).
velocities is fixed in the ship and moves with the ship
for linear translational movements (forward, lateral
translation and vertical heave) and heading changes at WoD
the origin, but remains fixed relative to the earth in
attitude, i.e. the axes do not move with the ship in pitch
or roll. The origin is placed on the flight deck midway
across the aft face of the hangar for calculating airflow
disturbances. Operationally it may be more appropriate
to use a system with an origin transferred to the centre
Lee rotor
of the landing circle.
Three basic flow patterns are used to estimate mean
airflow disturbances plus a further pattern to satisfy the
need for solid boundaries at the vertical faces of the
hangar and the hull. The basic patterns are Figure 6: Lee rotor flow
a) Corner flow, which describes flow up to and c) Profile drag flow, which describes the loss of
around a sharp corner of a face that is across momentum (velocity) in the wake of bodies and
the main flow, e.g. up to the side of the hull and is a measure of the profile drag of the body.
over the edge of the flight deck for a beam This flow pattern can be introduced
wind. This flow can be divided into two types downstream of any separated flow regions
depending on the angle between the wind and (Figure 7).
the line of the edge of the corner. If the wind is The pattern needed to satisfy the need for solid
almost perpendicular, like a beam wind to the boundaries is called Inflow and is the flow into an inlet
flight deck edge, then the flow is Separated at the solid downstream face such that the velocity
corner flow, (Figure 4) where there is a through the face is equal and opposite to the component
stagnant region just beyond the corner and the of the WoD in that direction. This flow pattern is needed
airflow reattaches to the surface some distance to obtain zero flow through solid faces, and because
downstream. If the wind is at a relatively neither the Lee rotor or the Profile drag flows produce
shallow glancing angle to the line of the edge perpendicular flow at downstream faces. Similarly a
of the corner, like a wind from 30° interacting negative Inflow (Outflow) is needed for upstream faces.
with the fore-and-aft top edge of the hangar,
then on passing over the windward end of the
 9-5

At periods below about 0.2s the aircraft response will

not usually be significant in terms of flight path
deviations although some accelerations may still be felt.
Fluctuations with a period between about Is and 3s are
particularly important when trying to control an aircraft
precisely. For a helicopter during landing or take-off
these periods will be associated with turbulence
wavelengths of about 10 - 100m.
It is very difficult to obtain information on the
distribution of turbulence energy at different and
relevant wavelengths for ship airwakes. Usually the only
Figure 7: Profile drag flow data available are Standard Deviations, which contain
no information on the distribution of energy. For the
In addition, turbulence is added to any existing free
purposes of testing this model a simple approach has
stream turbulence and applied as time varying
been taken where a level of Standard Deviation is
fluctuations to the mean disturbances.
applied to an existing turbulence model. The Standard
3.2 Selecting flow pattern modules Deviation is varied in proportion to the total incremental
Different flow pattern modules have to be selected velocity from the mean flow velocity model. In this
according to the general direction of the WoD and the way, turbulence will be greatest in regions where
different shapes of the hangar and hull. When a small incremental flow disturbances are largest.
change in WoD direction requires a change in the type In practice the ship will be operating in a particular wind
of flow module then it is possible that there may be a condition and there will be natural turbulence present
noticeable step in the total flow vectors, which is not that is related to the free stream wind strength, sea
likely to be present in real life. To avoid such steps the roughness and atmospheric stability. It seems
incoming and outgoing modules are overlapped and appropriate to include some element of the natural
linearly ramped up and down respectively over a small turbulence in the total turbulence and this is done by
(10°) range of direction. This overlap is not applied if taking the Root Mean Square (rms) of the natural and
the output of each module would be zero at the ship-generated turbulence. Ship-generated turbulence is
changeover point. In practice this means that overlap is proportional to the WoD strength, and this strength will
applied around the mid point of each quadrant, i.e. WoD depend on the combination of the natural wind and the
directions of-135°. -45°. 45° and 135°. speed and direction of the ship.
A full list of the flow pattern modules applied to each of A suggested value for the Standard Deviation related to
the faces and corners is given in Table 1. In general, a velocity increments is
separate version of a module is required for each of the
faces or corners that it has to serve because of the 0.4 x (Total local velocity increment),
different dimensions and orientation with respect to the i.e. total rms turbulence is given by
ship of each feature. This would imply creating a total
of 39 different modules. However, every Outflow rms [Free stream rms turbulenceT
module is equal and opposite to an Inflow module, and
the general symmetry of the ship makes it possible to turbulence hlO.4 x (Total local velocity increment)T
use single modules for a Port or Starboard face or corner Unfortunately it has not yet been possible to verify this
with suitable changes of sign. With these simplifications value, or the simplified approach to turbulence
the model could be based on 16 modules. representation, because the available helicopter
3.3 Turbulence simulation model and turbulence models did not
The flow pattern modules determine mean flow velocity produce helicopter disturbances that were recognisable
changes, but an important airflow ingredient in aircraft as normal responses by a helicopter in flight through
operations with ships is the large variation of local wind turbulence. Improvements in the simulation of
velocity with time, which is turbulence. The effect of helicopter responses to turbulence are being actively
turbulence on an aircraft depends on the frequency of pursued. An appropriate response to turbulence is an
the turbulence, which can be a function of the essential ingredient in the simulation of ship operations
turbulence wavelength and the aircraft airspeed as well with a representative airwake.
as a function of time. Very approximately an aircraft
and pilot will need to respond to wind fluctuations that
have a period greater than about 2-3s to avoid
significant flight path deviations. For fluctuations with
periods from about Is down to about 0.2s the effects will
be seen as turbulence to which they may try to respond.
9-6

30
i i OH
1 1
OH

o o 0 O
•3
U
00 00 00 00 1
S ° 1

I 1 1
OH OH
t/2

o o 0 O

HJ 1 1 1 1

o
OS
1
OH OH OH OH

(—( o 0 M

_1 1 CO f CO

H-rf m ►J OH
u Tf
4> 1
i. OH
1
CM OH CM
■o
c
o
Q
o
i.
1—I o 0
£
■w O
s i-J 1 1

0
5^ o
■a
•- 1
OH OH
1
OH OH

■«
es
>o
5«
■o
s
o 1—( o M
et
■3
4>
•o
£ 1 HJ
1
e in
es
CO ■
b> 1
OH OH
1
OH OH
v
CS
t:
3
on o o
i.
o
H-

•a 1 00 I-J 00
0
o
s ON
5«
S
1
OH
i 1
OH
1
> _,
•-01
« o o o 0
C.
£
o 1 1 1 1

fc "E m
f>
©
!■* EM l-H
1 1 1 1 £
OH OH 0
0)
A
cu
es
H o o o 1—i 0
<IJ

G0 CO
e
0 0
TJ 61)
00 00 00 CO <LI
1 1
is SH
X 0

o 1 s o 1 6
<L> 00
i0 TH

>
OH O OH O 0
00
<D
-4—' es
CO CO 0
00 ü T3
Ö
Ml
■a CO
a)
W a
■o
ces
<L>

a>
u
es CJ

S.
••
■a J3
es s H
M D
B CQ
4) es "3
> pd n J £
 9-7

WoD Direction

1
Ship Dimensions ► Parameter Dataset -* Flow Module constants

Vortex Separated Profile

Corner Lee Rotor Drag
Corner Outflow
Flow Flow Flow
Flow

Corner Lee Rotor Profile Drag

Inflow Outflow
Flow Flow Flow
Dataset Dataset
Dataset Dataset Dataset

Ship
Airwake
Dataset

Figure 8: Airwake model structure

4. RESULTS FROM THE AIRWAKE MODEL
Unit vector
4.1 Overall model structure
The overall model structure is shown in Figure 8.
Flow Module constants are determined during
verification of the model and then remain fixed for
all ship dimensions on ships with aft flight decks
and for all WoD directions. Ship dimensions are
defined once for a given class of ship. Details of the
required dimensions are given in Table 2. The final
input variable is WoD. Each of the 16 individual
modules then calculate flow velocity increments in
axes related to the part of the structure to which it
refers. When all of the modules for a particular type
of flow have been calculated then the velocity
increments are converted to ship axes and added
together in a dataset for that type of flow. Finally all
the 5 datasets for the different flow types are added W„
together to obtain the overall ship airwake dataset.
An example of wind increment data in a transverse Figure 9: Airwake increment vectors with WoD from Port
vertical plane through the landing circle for WoD 30°. Transverse plane.
from Port 30° is shown in Figure 9.
9-8

Table 2: Ship dimensions

Svmboi Description Example
value
Flight Deck Wd Datum Length. Beam width of the deck at the aft face of the 15.5m
& Hull hangar.
244m
Ld Flight deck length
W, Width at stern end of the flight deck, or effective width using 12.0m
an average deck taper angle.
Hd Height of flight deck above mean sea level 4.8m
d Offset of hangar centreline to starboard of the flight deck -1.0m
(ship) centre line
yd + 90 Included angle between the flight deck and the upper portion 90.0°
of the side of the hull
y.+90 Included angle between the flight deck and the upper portion
of the stern
*
Hangar Wh Width of the hangar top 8.8m
Lhs Length of the hangar on the starboard side 18.9m
Lhp Length of the hangar on the port side 12.6m
Hh Hangar height above the flight deck 5. 8m
F Hangar roof fence position forward of the aft end of the 7.9m
hangar, starboard
0Fs Hangar starboard roof fence angle aft of perpendicular to the 30.0°
edge
F Hangar roof fence position forward of the aft end of the 4.0m
hangar, port
9F Hangar port roof fence angle aft of perpendicular to the edge 0.0°

Y h + 90 Included angle between the hangar roof and hangar side 96.9°

* Wh= 14.1m for the 'Full Width Hangar' test case

„™ over the ship and experience has shown that these
5. COMPARISON WITH WIND TUNNEL TEST measureme J cm identify many major features of ^
DATA
, ... disturbed airflow. Thus initial validation of the model
It is essential that an empirical model such as this is ^ some adjustment of parameters have been made
validated against as wide a range of measured data as w_nd ^ measurements from a modei 0f a
possible. Measurements on and around real ships with ^ frf 5 Data from Ref 5 have been maiysed t0
aft flight decks would be most appropriate, but data is ^^ incremental velocity ratios ^ &e same axis
limited to a few measurements from flight decks using m as ^ girwake modd described m Ms Kporl,
masts. Measurements with model ships in wind tunnels ^ calculations of ^ow increment ratios have been
are much more extensive, although there is often only madg for ^ sWp ^^0^ of ^ modei used in Ref.
limited representation of small yet significant features 5 ^^ ^ typically 41 measuring points for a
such as safety netting and stanchions around the flight icular configuration with the wind tunnel model,
deck and many items of equipment mounted externally which .g relativdy extensive coverage. However it is
on all parts of a ship. Thus in general wind tunnel coverage when compared with 1287 points from
measurements will not be completely representative of ^ ^^ modd wm M& number of points ^
actual flow around a ship. horizontal spacing in the airwake model is typically
However, wind tunnel measurements are the most about 3m at full scale, and in the wind tunnel
extensive data source for airflow in locations that are not measurements it is 6 - 9m. Calculation points have been
 9-9

adjusted to coincide with locations used in the wind the airwake model, there is good agreement between
tunnel to help with comparison between data. This has wind tunnel data and the calculated airflow increments
resulted in some wider than usual gaps in data in the in vertical planes at the landing circle (transverse plane)
fore-and-aft direction, which are not typical of the and in the approach plane to port, Figure 10. The
calculated data used for simulation studies. horizontal components are very similar throughout. The
In complex flow conditions with a WoD from Starboard vertical components are very similar except in the
30°, where all the different types of flow are present in approach plane abeam of the landing circle where the
calculated vertical velocities are significantly greater
downwards than the model
Unit vector data. The vortex flow from
x/Wd=1.48 \|/ = 30° the hangar roof is clearly
Calculated seen in the model data in the
Wind tunnel transverse plane and is in the
same location as the vortex
in the calculated data. The
greater downflow in the cal-
1.5 culated data in the approach
plane may be a result of too
strong an inflow contribution
and too much reduction in
W, vertical component from
separated corner flow above
0.5 the maximum streamline.
Data comparison in a hori-
zontal plane, Figure 11,
0 ___
confirms that horizontal
components are in close
agreement at all points of
-0.5
-3 -2.5 -1.5 -0.5 0.5
comparison including abeam
of the landing circle at
W, : 1.48.
'Wrf
a) Transverse plane
6. PRELIMINARY
Unit vector EVALUATION IN A
c HELICOPTER FLIGHT
y/Wd = -1.3 y = 30 Calculated SIMULATOR
Wind tunnel One of the important uses
for an airwake model will be
in flight simulators for
training and for R & D
1.5 studies of helicopter oper-
ations with ships. Thus part
of this development pro-
wJ gramme included trials with
a preliminary form of the
model using a helicopter
simulation on the DRA,
Bedford, Advanced Flight
Simulator. This was the first
study of an airwake model
—VT with a helicopter simulation
model and the aim was to
see if the main disturbances
experienced in helicopter
b) Approach plane operations with a frigate
type of ship are represented
Figure 10: Comparison of estimates with wind tunnel data for WoD from Starboard and that the size of
30° disturbance is similar to that
 9-10

the time produced an

; j j ] ; unrepresentative response
from the helicopter in roll
and lacked the expected
:
! ! ! z/Wd=0.43 vu = 30° hard-edged effects in
j normal acceleration that
are experienced in flight
: | through turbulence. In
. Unit vector these circumstances it was
/ / i Calculated decided not to include
W, —-! —- turbulence with the
h> Wind tunnel airwake model for these
!
initial trials. Turbulence is
• ■

/
an important part of an
'1

" airwake and will need to be

- included in simulator trials
- as soon as basic problems
.
1 , , , /

of simulating helicopter
.... . . . . i
responses to turbulence are
3 2 1 () 1 overcome.
y
w Four WoD directions were
selected for the trial to
Figure 11: Comparison of estimates with wind tunnel data for WoD from Starboard represent distinctly diffe-
30°. Horizontal plane rent flow situations, and
strength was set at a level
experienced in real operations. A simulation of a that would be expected to cause significant handling
Westland Lynx helicopter was used for the trial. The difficulties without preventing operations, viz. a 40kn
airwake model was Version D. where the main features wind on the bow and at starboard and port 30°, and a
were represented to a standard that was acceptable for 30kn wind from the starboard beam.
an initial simulation study. Development continued
during and after the trial and the latest version described 6.2 Airwake datasets for simulation
in this report is Version F. Datasets for use in the simulation were provided as
tables of each of the wind increment ratios in beam and
6.1 Trial conditions vertical co-ordinates, and stepping aft in increments.
The trial used the Advanced Flight Simulator at DRA, Beam co-ordinates for this trial started just outside the
Bedford with a good representation of the handling
and performance of a Westland Lynx helicopter 0.6
"Zx."' ,-. "--,.„-,•...„..
with autostabilisers engaged. The computer Collective .
generated visual scene provided a representation 0.4

of a frigate with dimensions very similar (although

0.2
not identical) to those provided by the DRA for
modelling airwake flow increments. The v/V
differences in dimensions were not significant for
W/V .^
this initial study where the emphasis was on -0.2
•v-

comparing the main effects of the airwake on ,..».»•'•"'

S
helicopters. -0.4

It was decided not to include ship motion so that /

-0.6
the trials pilot, who was a qualified Royal Navy
Test Pilot with recent practical experience of u/V /
-0.8
operating a Lynx from a frigate in bad weather,
_„..--''"
could look for expected (and unexpected!)
disturbances without the additional workload from
operating with a heaving, pitching and rolling -1.2
j
ship. 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Height/Datum width, z/Wd
It had been hoped to include turbulence with the
airwake model, but an initial investigation showed
that the combination of the helicopter model and Figure 12: Vertical descent behind the hangar with 40kn WoD
the representation of natural turbulence in use at from ahead.
 9-11

starboard edge of the deck and went Port Deck edge Starboard
to 2.5 beam units (39m) to port of
0.6
the hangar centre line. Vertical co- Collective m ■ :
ordinates started at deck level and ►' ~\s \ ■
0.4 r'""S /Hvi.
went up to 1.6 beam units (25m), Ä
0.3
Z/W
H
V
„^-*-?r
and aft co-ordinates started at 0.3 K„ ^•—ss "——o-^-"* i ■ i - V
beam units (5m) aft of the hangar aft o 0.2
~\
'S v/V „"--""" /'>^ a [.
face and went to 2.7 units (42m). K 0.1
^Kwt™.*««.. J™«™H™a>™
,- >C''* y -\ 7r~^
te,^»»™«^
;

Spacing was 0.2 units aft and

*"'-\ u/V / T -~^~~x—«™
vertically (3m) and 0.3 units across ■0.1
w/V
the beam (4.7m). (Equal spacing *-"--,-.. ~x. .**
-~s
V / ' ■
m

was used in this preliminary model, W-— *•—--.„,-""'

but it is possible, and will often be -1.2 -0.6 -0.4 -0.2 0 0.2 0.4 0.6
Lateral position/Datum width, yANd
preferable, to use unequal spacing so
that data are closely spaced in
regions were they are changing Figure 13: Lateral translation from starboard to port at a low height with
rapidly with distance, and wider 40kn WoD from starboard 30°.
spaced where changes are slower.)
ahead are shown in Figure 12 plotted against the vertical
To ensure that there were no significant disturbances height in beam units. The collective increases from 40%
outside the modelling envelope, additional points were to 55% as the helicopter descends into the lee of the
added to produce values of zero increment and zero hangar and the downflow, w/V, increases from 0.1 to
gradient at the port, aft and top edges of the envelope. 0.22.
For this initial trial the data values at the forward and
starboard edges, and at the flight deck height in the A plot of the same parameters and height against lateral
modelling envelope were extended forward, to starboard position in beam units, Figure 13, for a lateral
and down to the sea. This ensured that there was no translation from starboard to port at 1 beam unit aft of
sudden change when moving into these areas, although the hangar and at a low height in a 40kn WoD from
it was not intended to use these regions during the study. starboard 30° shows a more complex variation of
airflow increments. In particular the vortex from the
With these additional boundary points there were a total hangar roof produces rapid changes between lateral
of 2688 locations with three airflow disturbance positions of -0.3 and -0.7 beam units. Rapid changes in
increment ratios at each point. Values at intermediate downflow from the vortex cause the brief bob-up of 0.1
points were calculated by linear interpolation and total beam units at around -0.4 beam units just before
wind components were then obtained by adding the reaching the port edge of the deck.
appropriate component of WoD. Finally, these total
wind components were transformed from the ship based During the lateral translation, Figure 13, the vertical
axes to helicopter axes, which are the axes of the flow is an upflow of 0.06 over the starboard side and
equations of motion. centre of the deck. Just before reaching the port edge of
the deck there is a brief region where the vertical airflow
6.3 Trial results falls to -0.05 and then rises again to +0.05 before falling
Tests at each WoD condition included general flying in rapidly to -0.22.
the regions of the normal approach path and over
the flight deck, and then specific recorded z/w
a
manoeuvres such as climbs and descents and 0.4
lateral translations at constant height. The pilot
was asked to comment on the effects that he
experienced and the control corrections he had to
make, and to compare these with his real life
experiences. Overall he was pleased with the
representation of airflow conditions over and
around the ship. Changes in collective control due
to downflow and due to reductions in airspeed
were as expected both in the places where the
changes occurred and in the magnitude of the
changes.
The variation of airflow increment ratios and
collective control position during a vertical
Figure 14: Unstable vertical flow region during lateral
descent from above the hangar over the centreline translation.
and about 0.9 beam units aft in a 40kn WoD from
9-12

The bob-up occurs in a small

region where vertical flow
becomes more positive with
increasing height, which makes
the region statically unstable.
This can be seen from the
difference between the vertical
component of the airflow at
z/y = 0.2 and 0.4 in Figure
14, where there is a greater
downdraught at a height of
a) Version D
: 0.2 than at 0.4 in the
W,*
region around * -0.4 ..---j'
/\ s-
^
-■~"""
~~ ■
Improvements in the Airflow „--*--• '
—'■""'"

'"sfc: r~ .„-■"'*
r »— —*—u/v
Model between Version D, Ht^ r-~^s««^-™^..^ wit1 i~ ^K: -C5 ^■--■&"" /^-fl 5 ----S.— w/V
Figure 15a, which was used ~*-\^ '■*-•», -~-:<---w;V. z=0.2

during the simulation trials, and \

Version F, Figure 15b, described ^\
in this report increases the """----^
effects of the inflow component
of the flow, but there is still a
small region around the port b) Version F
edge of the deck at
Figure 15: Flow increment ratios vs lateral position with WOD from
« -0.4 in Figure 15b, Starboard 30° at x/Wd = 1.1 and z/Wd = 0.4
/wd
which will tend to produce a change in fore-and-aft wind increments, and too small a
brief bob-up of similar size. This region is on the difference in these increments between the lee of the
upwind side of the vortex cornerflow from the starboard hangar and regions that are more exposed to the free
hangar roof edge. At smaller WoD angles from stream wind.
starboard or port the bob-up region will be over the
deck, which may increase handling difficulties in the Analysis of data from the simulation trials indicates a
landing area although the strength of the vortex will change of about 1° nose down in pitch attitude between
decrease at smaller WoD angles and reduce the level of about lOkn airspeed and 40kn, which is similar to
changes present in flight. Thus the drag of the helicopter
static instability.
is not a cause of the problem.
There were three aspects of the behaviour of the
helicopter in the simulated airwake where the pilot The change of pitch attitude with airspeed is
wished to see improvements approximately proportional to the square of the airspeed
and the airspeed is already around zero in the lee of the
i) representation of turbulence, which is an hangar. Thus it is unlikely that changes in the fore-and-
essential part of ship airwakes and helicopter aft airflow increment will affect pitch attitudes, and it
handling problems, but had to be removed from would seem likely that this problem is associated with
the airwake model because the simulation could the rate of change of airspeed with position. This rate of
not adequately represent the behaviour of the change could be too gradual either because of the
helicopter in turbulence, airflow model construction or because calculation points
ii) changes in mean airflow in the model appeared are too widely spaced and smooth out rapid changes.
too gradual, particularly around the edges of the It is also possible that the effects of gradients in airflow
hangar across the main rotor are significant and these were not
iii) changes in pitch attitude in moving from the lee included in the simulated helicopter. Gradients of
of the hangar into free stream were too small, vertical airflow could be particularly significant in
or perhaps too gradual. (It was too easy to producing roll and pitch moments. Such gradients will
maintain fore-and-aft position.) also be present in turbulence.
The first aspect is not a problem with the airwake 6.4 Implications of airwake model developments
model. The last aspect could be due to errors in the since the simulation trials
simulation of the drag of the helicopter at speeds below There are significant changes between Version D of the
50kn, or a combination of drag errors, a too gradual model that was used to calculate datasets for the
 9-13

simulation trial and the current Version F. None of the and compares with the wind tunnel increases in
changes is expected to affect the main conclusions of the crossflow angle.
pilot from the initial simulation, but the significant Thus the main effects of the wider hangar are to move
increase in airflow increments in some locations may the vortex from the hangar roof edge to starboard as the
result in new comments. A comparison between the two edge is further to starboard, which increases the region
versions for WoD from starboard 30° is shown in Figure of static instability from the vertical airflow component,
15. The maximum downflow is almost identical and at and to widen the region where significant crossflows and
the same location for both versions, but detailed loss of fore-and-aft flow occur.
variations in downflow are different. The maximum
crossflow and loss of fore-and-aft flow are about twice 8. DISCUSSION
the magnitude in Version F. The unexpected small In this section some important implications for the
changes in crossflow and fore-and-aft flow between helicopter/ship dynamic interface are identified, and this
y
/., = -2.2 and 2.5 in Version D has disappeared is followed by discussion of some implications for future
/wd wind tunnel model tests and for modelling of helicopter
in Version F. flight behaviour.
Helicopters are very sensitive to vertical flow changes,
8.1 Important implications arising from Airwake
are sensitive to crossflow. and are less sensitive to fore-
estimation model data
and-aft flow at low airspeeds. Thus effects on a
helicopter from the greater crossflow and fore-and-aft This study of contributions from particular ingredients in
air flow patterns provides insight into some significant
flow may not be as dramatic as the differences in
features of ship airwakes that may not previously have
airflow suggest.
been obvious. Two such features are the significance of
7. AIRFLOW INCREMENTS WITH A FULL rates of change of airflow increments with position in
WIDTH HANGAR the wake (airflow gradients) and vortex corner flow.
Increasing the width of the hangar in the wind tunnel
8.1.1 A iiflow gradients
model tests of Ref. 5 significantly increased the
If the rate of change of an airflow increment with its
downflow and crossflow angles measured over the deck.
direction is positive then pilots will experience
Airflow increments have been calculated for the ship
difficulties in maintaining position in that axis. This is
example described in Table 2 but with the width of the
hangar increased to occupy the full width of the ship, particularly true in the vertical direction. If the vertical
gradient is positive then any upward movement of a
i.e. a width of 14.1m compared with 8.8m. The airflow
helicopter will produce an upward increment of the
increment ratios at */yy = 1.1 and z/^ = 0.4 shown vertical airflow and the pilot will require a larger change
in Figure 16 for a full width hangar and a WoD from in collective to stop his movement than to start. This
starboard 30° can be compared with those in Figure 15b. will make it more difficult to control the height of a
helicopter. Similar effects will occur in fore-and-aft and
The region with statically unstable vertical flow lateral directions and affect the control of horizontal
conditions is much wider and twice the level of position.
instability. It is also much nearer to the landing point,
Because gradients are important it becomes necessary to
y/
which is at 0 . This would be expected to make locate measuring points much closer together in regions
Wri
with changing gradients. If spacing of data points is too
it significantly more difficult to control height over the
port side of the deck than it is in the case with the large then the apparent gradient is reduced. This could
be the reason for the pilot's comment that changes were
narrower hangar. The magnitude of the downflow over
too gradual in regions such as downwind of hangar
the port side of the deck has
^
also increased a little and the
loss of fore-and-aft ^
K *——-_ —«-,
component of airflow is a „ _..»--'''""' ..- :i:
■■'<&,,
U k.
little larger; these will V .- -"4
combine to produce a greater ^--~ r-:*-—
--"■s----^-.: .-. . -ig.;. A
'i-.f-K .-_. _. - 1
rr' ( »-- / K
-0 5 ,*-- —*—v/V

downward flow angle as - -i ■* / /

-- -- -
//^1 —;-:~-w/V,z=0 2

noticed in the wind tunnel ^"■*---. >■

tests. However it should be ^"'*---.

noted that it is the actual *" ——v^fc.-—"^

downflow that is significant

to the helicopter performance
rather than the flow angle. Figure 16: Full width hangar: Flow increment ratios vs lateral position with WoD
Crossflow over the port side from Starboard 30° at x/Wd = 1.1 and z/Wd = 0.4
of the deck is also stronger
9-14

comers. Spacing of around 1 - 2m would be more 8.3 Helicopter modelling

appropriate in regions with rapidly changing gradients. The issues of concern are
• effects from gradients of airflow across the main
8.1.2 Vortex Corner flow rotor, particularly gradients in vertical flow
Vortex Comer flow produces significant localised
increments in airflow and gradients. It would be better • effects of turbulence
for helicopter operations if it did not occur. In practice it • use of single point or distributed airflow inputs to
is not always clear whether a vortex is present from a helicopter model
measurements over a flight deck. However, it is
relatively easy to inhibit the formation of a vortex by • 'ground' effects
using netting similar to the safety netting around the
8.3.1 Ail flow gradients
flight deck as a fence around the top of the hangar and
Gradients of vertical airflow across the rotor can be as
attached very close to the edge. The fence can be the
much as 30% of the magnitude of the WoD. see Figure
normal height and it is essential that the netting is an
15b. This will produce the same order of forces and
open mesh otherwise it can act as a solid fence and
moments on the helicopter in 40kn WoD as a typical
generate a vortex from the top of the fence. The model
rotor rolling at about 207s. This is likely to generate a
tested in Ref. 5 included a case representing a canvas
significant rolling (or pitching) moment and its
fence around the top of the hangar and this generated a
importance should be discussed with experts in
full vortex.
helicopter flight dynamics. The airflow model contains
A netting fence will also reduce the local wind on the adequate data to define flow gradients across a rotor
top of the hangar, which will improve the environment either by direct calculations of airflow increments at
for anyone who has to be in that area. different parts of the rotor during simulation, or by
Because flow from a vortex is a significant ingredient in providing gradient information based on an appropriate
an airwake, it is important to establish whether it is a rotor diameter as discrete values in addition to the
feature of a particular ship. This can only be established velocity components provided from the current version
from measurements on the full scale ship, as its of the airwake model. Advantages and disadvantages of
existence will depend on detailed features around the the two methods of obtaining gradients are discussed in
comer between the side and top of the hangar. detail later in this section when considering the use of
single point or distributed airflow inputs to a helicopter
8.2 Wind tunnel model testing model.
Development of the airwake estimating model has
highlighted some features of ship model testing that 8.3.2 Turbulence
should be considered when planning further wind tunnel Modelling the response of helicopters to turbulence is a
tests. Such tests will remain an important part of complex subject and there is not yet a consensus on how
programmes to measure likely ship airwake it should be achieved. It is an essential ingredient in any
environments and study the implications of changes in simulation of helicopter operations in ship airwakes, and
geometry on airwakes. Because of the importance of it is important to find ways of adequately representing
vortices in airwakes it is essential that detailed features the main effects of turbulence on a pilot's performance
such as safety netting and stanchions, which will inhibit of operational tasks. In general terms, pilots are able to
the formation of vortices, are represented on models and respond to turbulence components with frequencies less
have appropriate levels of porosity. Also, measuring than about 0.5 - 1Hz and it is turbulence at these
points should be much closer together in regions of frequencies that can cause significant disturbances to the
strong shear, because of the significance of gradients of position of a helicopter. Higher frequency turbulence
flow increments, which can be particularly large in these usually has little effect on a helicopter's position but can
regions. Positions as close as 1 or 2m full scale should cause an increase in overall pilot workload. A frequency
be considered in some critical regions where strong of 1Hz in 30kn WoD corresponds to a wavelength of
shear regions occur near the expected flight path of about 15m, which is of the same order as the main rotor
helicopter main rotors. diameter for many helicopters that operate from small
ships. A distance of 15m is also larger than many of the
In addition, it would help to define appropriate main features such as hangar height, freeboard, and
turbulence levels for simulation if power spectra could hangar width, that produce turbulence in an airwake.
be obtained down to wavelengths of around 20% of the Thus it would be expected that the contribution to
deck width (about. 2 - 3m full scale). Overall rms data is turbulence from an airwake will be mainly at high
not sufficient as it requires assumptions about the form frequency. Disturbances induced as a helicopter moves
of the power spectrum to determine levels of turbulence through the mean steady state airwake may be the mam
at wavelengths that affect the behaviour of a helicopter. cause of low frequency perturbations.
In these circumstances it seems likely that appropriate
turbulence could be generated in an airwake model as a
 9-15

spectrum of force and moment increments determined helicopter rotor model where local forces and moments
from accelerometer measurements on appropriate types can be calculated. Thus it is important to establish
of helicopters during actual ship landings in known whether there will be sufficient differences between
WoD conditions. It may not be necessary to calculate using airflow calculations distributed along the rotor
helicopter responses directly during simulation. Such a blade or using values of the mean airflow at the centre
random spectrum could be varied with local airflow of the rotor and a gradient at the same point.
increments to provide the typical increase in turbulence In general, for real-time simulation there would seem to
in regions of disturbed flow. In principle, turbulence is be little justification for adding the complexity of
greatest in regions with high shear (velocity gradient) as calculating airflow increments across the rotor, because
well as regions where flow increments are large, and it the forces and moments generated by airflow gradients
may be desirable to make turbulence intensity a function are likely to be less significant than changes in forces
of both the size of the flow increment and its gradient if and moments due to the mean airflow and control inputs
evidence from instrumented helicopter operations, and by the pilot and stabilisation system, and because of
piloted simulation trials, suggests that this further some uncertainties about the precision of airflow
refinement is important. calculations and the methods used to calculate the
Some low frequency turbulence will be present when effects of non-linear variations of the airflow field.
operating in natural wind turbulence outside the more It is also important to note that gradient information can
disturbed areas of an airwake. This should be included be used with solid disk models that cannot use
in flight simulators to avoid an unrepresentative step in distributed airflow data, because only mean disc forces
workload between natural wind and airwake turbulence. and moments are calculated. Thus it is recommended
The appropriateness of using a random spectrum of that helicopter aerodynamic models should be adapted
forces and moments varying with airflow increments to accept both mean and gradient data from airflow (and
could be assessed as soon as some suitable level for turbulence) models.
these disturbances can be determined from instrumented
operations. 8.3.4 'Ground' ejfects
The flow produced by a helicopter's rotor is deflected
8.3.3 Single point or Distributed Aitjlow Inputs by nearby surfaces such as the ground or solid vertical
Most real-time flight simulators calculate aerodynamic walls and affects the local flow at the rotor. During
forces and moments from airflow conditions at a single landings over open solid ground the changes in moments
point on the aircraft, which is usually the centre of and forces induced by recirculating flow are known as
gravity or a nearby reference point. This is very ground effect. The presence of parts of a ship will have
satisfactory for flight in undisturbed air without any similar effects, although the flow pattern will be more
airflow gradients. When airflow gradients are present complex because the surfaces on a ship are of a similar
that are significant over distances that are similar to the size to the rotor. Only parts of the flow will impinge on
main dimensions of the aircraft, then it becomes particular surfaces depending on the relative position
necessary to consider introducing forces and moments and attitude of the helicopter to those surfaces. These
due to gradients either by 'ground' effects are a function of the helicopter and the
ship. To a first order they can be assumed to be
• defining gradients at the single point and using
independent of the WoD because velocities in the rotor
appropriate derivatives of forces and moments, or
downwash are much greater than the WoD.
• calculating local forces and moments on different
It is important to include the main effects from
parts of the aircraft and integrating these to
helicopter downwash to simulate all aspects of ship
obtain total forces.
operations. The adequacy of any representation of
It is more rigorous to calculate local forces and 'ground effect' can be assessed by simulating operations
moments, but it introduces a very significant increase in in zero WoD.
the scale of computation required to determine
aerodynamic forces and moments. It is important to 9. FUTURE DEVELOPMENTS
establish whether there are significant advantages in An empirical modular airwake model that can estimate
using distributed calculations; otherwise there is no the airflow in the wake of ships with an aft flight deck
justification (or need) to accept the accompanying large for different ship geometries and directions of WoD has
increase in computation requirements in a real-time been developed and demonstrated in a piloted flight
process. simulator. The model has been compared with results
from wind tunnel tests for a ship with an aft flight deck
In the case of helicopter simulation, solid disc or more
at several WoD directions and shows only small
complex blade element rotor models are both similarly
differences.
affected in principle by decisions on using point or
distributed airflow inputs. However, blade element This is a very promising development and this section
models require a major increase in computing capability describes further stages of development that could lead
for real-time calculations and this is the only type of to the implementation of the model for SHOL
9-16

evaluations and test support, and on flight training necessary ship motion, helicopter response and interface
simulators. modelling on a suitable current helicopter flight
simulator and then install a full set of airwake data. This
9.1 Model validation and development installation process followed by a brief evaluation of
Comparison of the airwake model with wind tunnel data simulated operations from a ship will clarify details of
has only used test results from a single programme. the required form of airwake datasets for
Wind tunnel data is not totally reliable as an indication implementation on training simulators.
of full scale airflow patterns and it is important to
compare airwake estimates with full scale data and other 10. CONCLUSIONS
boundary layer wind tunnel tests before investing too An empirical mathematical model has been developed
much further effort in improving the estimating model. to estimate airflow disturbances around and downstream
of the flight deck on any ship with an aft flight deck and
Tests with the airwake model to date have also only for any direction of Wind over Deck (WoD).
been with WoD directions from forward directions
round to the starboard beam. It is necessary to validate Data from the model has been used in an initial
the model for winds from aft of the beam. assessment on a piloted flight simulator and pilot
response has been very favourable. All the major airflow
There is thus a need to complete the iterative cycle of deviations features expected with WoD from ahead,
validation and model development to demonstrate the starboard and port 30°, and from the starboard beam
suitability of the model to represent airflow for a wide were satisfactorily represented. Unfavourable comments
range of ships and all directions of WoD. Some possible were expressed about the lack of turbulence, which is
improvements to the model have been indicated from available in the model, but could not be used because
the single comparison with wind tunnel data and futher the simulated helicopter model at that time did not
comparisons will show whether these improvements are respond appropriately to simulated turbulence, even in
appropriate or whether they reflect differences that are free air. The relatively smooth and gradual transition
peculiar to the particular wind tunnel tests. between conditions in the lee of the hangar and out into
9.2 Piloted simulation and demonstration free stream was also unrepresentative of actual ship
Piloted simulation is another important ingredient in operations.
validation, particularly when considering SHOL Piloted simulation demonstrated the significance of
evaluation and simulated flight training where the views regions of static instability where airflow increments
of pilots are an essential part of the process. To date the increase with distance, particularly for vertical airflow
model has been evaluated on a flight simulator in a (updraughts or downdraughts). There are tendencies to
controlled, but somewhat artificial condition, with no overcontrol in these unstable airflow regions, and such a
ship motion or air turbulence. This was very helpful for region can occur over the deck with WoD from a
initial assessments of the model and now that stage has forward quarter, and wide hangars increase the problem
been completed satisfactorily it is important to include area.
airwake modelling together with ship motion and the Detailed comparison of data from the airwake model
presence of appropriate air turbulence. This will then with results from wind tunnel tests on a ship with an aft
represent the usual environment experienced by pilots flight deck show a good comparison in the complex
during operations from ships. airflow produced with WoD from starboard 30°. Some
This simulation will depend on the provision of differences are apparent in comparisons at other WoD
simulation capabilites for ship motion and helicopter directions, which, although sufficient to warrant some
response to airwake gradients and turbulence, as well as further modifications to the airwake model, would not
provision of airwake data including gradients. be large enough to influence the pilot's assessment in
The initial model represents sufficient of the main the simulator.
characteristics of an airwake to make it suitable for Development and use of the airwake model has
demonstrating to potential users such as the MoD and highlighted several points that are important in reducing
Simulator Manufacturers. Such demonstrations will be the effects of airwakes on existing and future ships,
esential to ensure a smooth and rapid transfer of providing data for airwake simulation from computer
estimated airwake data into specific flight training estimates or wind tunnel tests, and simulating helicopter
simulators, and it would speed this transfer if these response to airwakes and turbulence. These points are
demonstrations take place as early as possible. • Provision of a porous net fence around the edge
9.3 Implementation of airwake data on a flight of the top of the hangar should prevent the
training simulator formation of vortex flow from the side edges of
The final stage before any routine use of airwake data the hangar for WoD within about 45° from ahead.
from this estimating model should be a trial installation This vortex provides a significant contribution to
and evaluation on a selected helicopter flight training undesirable flow velocity gradients near and over
simulator. It will be important to implement all the the flight deck. Thus eliminating this type of flow
 9-17

should reduce the effects of the airwake on a

helicopter. R E Whitbread. P G G O'Neill & V G Quincey.
• It has been shown that gradients in ainvake 'Project No. 35019. A study of the wind flow
velocities can produce regions of instability and over the Flight Deck of a typical Frigate',
this means that airflow data must be provided at British Maritime Technology Ltd.. June 1986
close spacing to define gradients where they are
significant. Spacing of around 1 - 2m full scale is
needed in such areas, which is significantly closer
than the more normal spacing of about a
helicopter rotor radius (about 5m)
• The presence of gradients as large as 30% of the
magnitude of the WoD in vertical flow changes
across a typical rotor diameter suggests that there
will be significant roll or pitch moments
generated on a helicopter. Helicopter simulator
models need to be modified to respond to such
gradients, and it is recommended that airwake
models should provide gradient data so that
suitable inputs can be provided to both simple
disc rotor aerodynamic models and blade element
models without undue computing overheads.
• It is noted that helicopter models must include
appropriate ground effect simulation to represent
the effects of proximity to solid surfaces such as
the deck and the vertical face of the hangar.
The empirical airwake model is very promising as an
economical and effective means of exploring the effects
of ship geometry on Ship/Helicopter Operating Limits,
and providing a form of airwake data suitable for use in
piloted flight simulators for R & D studies of helicopter
operations from ships and for pilot training. Its validity
should be tested against a wider range of ship ainvake
data and the model should be developed to provide
datasets for a full range of WoD directions on various
ships for use in existing and future helicopter training
simulators.

11 REFERENCES
1 'Aircraft Ship Operations', AGARD Flight
Mechanics Panel Symposium, Seville, Spain,
May 1991, AGARD-CP-509, November 1991
2 Lt. S J Täte RN. 'A dynamic challenge:
helicopter/ship interface simulation
development, integration and application'.
Flight Simulation - Where are the Challenges,
AGARD Flight Vehicle Integration Panel
Symposium, Braunschweig. Gennany, May
1995
3 E A Garry, 'Aerodynamic Aspects of the
Helicopter/Ship Interface - A Literature
survey'. Institute for Aerospace Reasearch,
NRC, Canada, IAR-CR-27, October 1994
4 AD Penwarden & A F E Wise, 'Wind
Environment around buildings', Building
Research Establishment Report. 1975
 10-1

A dynamic challenge: helicopter/ship interface simulation

development, integration and application

Lieutenant S J Täte Royal Navy

Flight Dynamics and Simulation Department
Defence Research Agency
Bedford, MK41 6AE
United Kingdom

Summary exacting. It follows that high fidelity piloted .simulation

of helicopters at the ship interface is a major challenge.
Piloted simulation has a potentially major role to play in Not only is it necessary to simulate the aircraft, but also
training, test and evaluation and research in support of the vessel and the environmental conditions that affect
the helicopter/ship dynamic interface. The combination both. Consequently, although piloted simulation has
of visual/motion cueing and vehicle/environment been used extensively in many fixed wing and rotary
modelling problems makes the helicopter/ship dynamic wing applications, it has not been in widespread use for
interface one of the most challenging in the simulation of the helicopter/ship interface environment.
aerospace vehicles. The exacting fidelity requirements
of this application have hindered the use of simulation in The DRA at Bedford is currently engaged in an
this area until recently. This paper reviews the major integrated research programme to examine
fidelity issues of helicopter/ship interface simulation and helicopter/ship operations and provide methods of
describes some of the findings of a survey carried out of increasing operational limits in respect of more severe
current simulators to attempt to identify areas that require sea states, relative wind and visibility conditions. There
improvement. The cost and operational benefits of the are three elements to the helicopter/ship work;
use of high fidelity piloted simulation in this field are establishing helicopter handling qualities criteria,
outlined. Current Defence Research Agency (DRA) developing automatic approach guidance aud providing
research work directed at improving models of the an enhanced cueing environment aud improving
natural environment is described, as well as the thrust environment models for use in pilot training simulators.
that future work will take. The uses of the high fidelity The work has been carried out under three separate, but
helicopter/ship dynamic interface simulation on the DRA interdependent and integrated, research programmes. All
Advanced Flight Simulator (AFS) in support of other three rely heavily on piloted simulation as a major
research packages is described. This work covers the research tool and demand improvements in the fidelity of
establishment of handling qualities criteria for maritime the simulation.
helicopters and the use of simulation to develop
improved pilot visual cues for deck operations. Use of This paper deals with the broad aspects of the uses of
simulation has allowed rapid development at relatively helicopter/ship dynamic interface simulation, the key
low cost and at much reduced risk. problems faced in improving fidelity aud where the DRA
effort has been focused. Consideration has also been
given to the uses of dynamic interface simulation at
Introduction DRA, as well as a review of future work in this area.

The operation of helicopters from ships, particularly The requirements of a simulation can differ markedly
small ships, presents a demanding task for both the depending on whether it is to be used for training,
aircraft and the pilot. Simulation of rotorcraft is also development, test and evaluation, or research.
Experience at DRA in the research field has been
enhanced by collaboration with other countries, notably
the US, in understanding the needs for test aud
Presented at AGARD FVP Symposium 'Flight simulation:
evaluation work. Specific research work has also been
Wfiere are the challenges?', Braunschweig, Germany,
carried out, as will be described in this paper, to identify
22-25 May 1995. ® British Crown copyright 1995/DRA
the current capabilities and future needs of the training
-published with the permission of the Controller of Her
community.
Majesty's Stationery Office.

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?"
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
10-2
In the training environment there are two areas in which
Benefits of success simulation can be beneficial. The first of these is in
reducing costs. Provision of aircraft flying hours and
The benefits of achieving high fidelity piloted simulation ships for training is expensive. The second is in
of the helicopter/ship dynamic interface are wide- improving the training given to pilots. If students can be
ranging. There are, as noted above, three key areas that taught procedures and how the aircraft will behave, even
can be supported: in the worst expected operational conditions, then pilots
who have newly completed training would be more
• Research operationally effective. It also means that more effective
use can be made of actual sea training time. Currently,
• Test and evaluation tasks there are probably only two or three training simulators
in the world that are capable of being used for effective
• Pilot training for ship operations deck operations training.

Obviously, the detailed needs for each of these areas are The effectiveness of the training provided by the
different, as are the direct benefits. However, in broad simulator will only be fully evident when student pilots
terms the benefits are based around the following: move from carrying out simulated deck operations to
going to sea with the real aircraft. The true measure of
• No need to provide costly aircraft or ships success is the point in the training programme to which
pilots can move having completed the simulator training.
• Precise, repeatable environmental conditions Ideally, one deck landing in the simulator would equal
one in the real world, whatever the conditions. This is
• Repeated attempts/replays of particular tasks not possible with current technology. However, if 10
landings in the real world could be equalled with, say,
• Carry out training and research that would 14 in the simulator there would still be significant
otherwise be difficult, impossible or too risky positive training value.
physically or financially
A preliminary worthwhile aim for naval helicopter
• Use of simulators as a scratch pad for new ideas training simulators would be that, following simulator
training, student pilots be able to carry out their first real
• No flight safety issues deck landing sorties at sea, in benign conditions, without
intervention from the instructor. Initial procedures and
From a research perspective, simulation offers the the behaviour of the aircraft at the ship interface would
opportunity to try ideas and concepts without the need have been effectively demonstrated on the simulator.
for costly and possibly risky flight trials. Indeed, This baseline capability could then be expanded to
simulation may be the only practical way to carry out include, eventually, the full ship/helicopter operating
some work, such as aircraft control system evaluations. envelope. In parallel with this, the research and test and
Development can take place using the simulator as a evaluation communities should be seeking to produce
prototyping tool. Concepts can quickly be modified or simulations that will allow some ship/helicopter
rejected on the basis of their performance in the compatibility testing and research work to be carried out.
simulator.

In the test and evaluation field the drive is to be able to Where are the problems
carry out helicopter/ship compatibility testing and
evaluate new aircraft, ships or systems against their General
specifications. Currently, compatibility testing in the UK
for one aircraft/ship combination requires an Over the last 10 years hundreds of hours of piloted
instrumented aircraft and ship for 3-5 weeks and 350 simulation have been expended in the US, Canada aud
plus deck landings are carried out. This process is Europe in efforts to revise the handling qualities
vulnerable to weather and serviceability. If the right requirements for battlefield helicopters. Iu the same lime
weather conditions are not found then a restrictive set of frame very few piloted simulation experiments addressing
operating limits can result. Simulation could be used to handling qualities at the helicopter/ship dynamic interface
clear an initial operating envelope and determine possible have been reported (references 1 and 2). There has been
critical areas which would then be investigated through no significant use of piloted simulation of the dynamic
flight trials. This would save on the resources required interface in test and evaluation programmes, with the
and improve the chances of a good initial operating notable exception of the US Navy V-22 programme
envelope. (reference 3). The 1991 AGARD conference dealing with
 10-3

aircraft ship operations, although covering many survey was limited in its objectives and in the
modelling issues, contained no papers concerned directly information that it could provide. However, it gave
with piloted simulation (reference 5). This reinforces the insight into the views of operators and their perceptions
fact that adequate modelling is one of the keystones of of the devices with which they work. These insights are
successful simulation. summarised below under various headings.

Perhaps this lack of dynamic interface simulation usage As part of the process of assessing what is currently
should not be surprising. Work conducted by the US possible, the DRA also recently carried out an evaluation
Navy (reference 1) using the NASA Ames Vertical of the latest high fidelity pilot training device to come
Motion Simulator (VMS) attempted to define Level 1/2 into use. This was the US Coast Guard IIII-60J Jayhawk
Cooper-Harper handling qualities rating (HQR) facility, which began to be used for pilot training at the
boundaries (reference 5) for all control axes, in various end of 1994. This was invaluable (reference 8). The
sea states, for a SH-60 class aircraft operating to non- specification for this device determined that no untried
aviation ships. In the event, no Level 1 configuration technology should be incorporated. Consequently, much
could be established at or beyond the sea state 3 baseline. effort was expended on refining available technology and
The impression given in reference 1 is that the visual cue achieving good integration. This was a successful
deficiencies in the VMS (at that time) were a significant approach. Good cueing, special attention to scene
influence on pilot handling ratings. This result is content issues, and the fact that a specially instrumented
entirely consistent with more general findings on aircraft was used to collect data for the aircraft model
research simulators in the mid-1980's. Reference 8 has produced a simulator with the ability to be used for
states that for nap-of-the-earth helicopter mission task some deck operations training. This evaluation
elements in the early 1980's, pilots typically rated the demonstrated what is possible with 1990 technology.
VMS between 1.5 and 2.5 HQR points worse than in
flight; again Level 1 ratings were hard to achieve. By Validation of a simulation is always difficult, and is a
the late 198G"s improvements to the VMS visuals, motion key area of concern. However, the ship environment has
system and transport delays had narrowed the gap particular features that cause problems. The environment
between flight and ground-based simulation to about 1 around the flight deck of a ship is difficult to quantify.
HQR point. The work reported in reference 1 is very For example, taking wind or turbulence measurements
useful because it demonstrates the significant impact that down the final approach path would be very difficult.
deficiencies in the simulation can have on the validity of The air wake environment over the flight deck can be
an experiment or in pilot training. It also highlights captured to some degree, but not the effect the aircraft
critical areas where attention must be paid to achieve has on it as it is inserted in the flow field. Validation is
fidelity when simulating the helicopter/ship dynamic a key issue that will require more attention.
interface - the importance of visual texture and associated
depth perception and the degree of realism of ship air Visual system
wake and turbulence modelling, for example.
The tasks that are to be performed using the simulator
It is not an easy task to identify all the shortfalls in state- have a strong impact on the specification of the visual
of-the-art dynamic interface simulation capability. system. For deck operations the pilot faces problems
Fidelity of different components must be kept in balance which are peculiar to this environment:
with each other. For example, a highly capable ship air
wake and turbulence model may require the use of a • Cues can be very limited, particularly at night
comprehensive aeroelastic blade element model for the
main rotor to be fully effective. Inter-relations such as • In degraded visual conditions the ship itself may
these require careful consideration before committing to provide the only closure rate and positioning
expensive research and development work. cues

The DRA conducted a survey of a number of dynamic • Pilots use the sea surface to assist in estimating
interface simulators (reference 7). The aim of the survey drift and to predict ship motion
was to assist in indicating where the primary deficiencies
lay in providing an adequate simulation of the This means that scene content, peripheral cueing, and
helicopter/ship dynamic interface, and then redirect adequate modelling of the sea surface are all of particular
research effort if necessary. The simulators surveyed importance.
were mainly training devices, but the V-22 simulation on
the US Navy Manned Flight Simulator (MFS) was Simulators with modern display systems provide adequate
included. The ages, and consequently technologies, of fields of view (typically 200 degrees in azimuth, biased
the training devices ranged from 1979 to 1991. The to the right hand seat, and at least 40 degrees in
10-4

elevalion). A primary concern is the lack of downward Aircraft mathematical model

field of view to the right for conducting precision
hovering alongside and over the flight deck and for the None of the simulators surveyed were considered by
landing task. The field of view issue is one which can their operators faithfully to reproduce the handling
largely be solved by providing a state-of-the-art image qualities of the real aircraft. This had led to a number
generation and display system. For an adequate field of of problems in use of the devices for particular tasks.
view to be achieved (past the beam position on the pilot's For deck operation simulation to be meaningful it will be
side of the aircraft and downwards at least 35 degrees) necessary to employ models able to react accurately to
there are several options; using a wide display, a motion- ship air wake and turbulence, as well as possibly having
mounted dome, area of interest displays, or by some main rotor/tail rotor interaction capability. There
employing helmet-mounted visuals. All of these have is no doubt that many current simulators suffer from a
advantages and disadvantages ranging from cost to added lack of adequate aircraft data for the model.
weight on the motion system - which also comes down Traditionally, in the UK, data packages have been (Ir-
to a cost issue. responsibility of the simulator manufacturer and this has
lead to poor or inadequate data being used in modelling
Lack of scene detail was the most frequently noted work. The US Navy sets standards for data packages
problem with image systems. Lack of texturing and that it requires from the prime contractor of the system
inadequate resolution were also noted as difficulties. being procured (reference 10). This method ensures that
Phototexturing can make a large positive impact on scene sufficient and relevant data for modelling requirements is
content and the number of polygons available per image available.
channel continues to rise. One principal requirement for
deck operations was for visual modelling of the sea Ship air wake and turbulence modelling
surface to accurately reflect the actual sea state.
Ship air wake and turbulence modelling plays a major
Motion system role in performing deck operations. Until recently, little
attention was given to the provision of realistic air wake
Many current simulators lack performance in cueing for and turbulence models. This was because other
low speed manoeuvres, particularly in the heave axis. deficiencies, such as the visual system, meant that deck
Significantly, when operating at night, or in degraded operations could not be conducted effectively in a
visual conditions, pilots become more reliant on accurate simulator. The position of the flight deck and the close
motion cueing to be able to complete the task proximity of superstructure means a turbulent
successfully in the simulator. As much of the dynamic environment with many updraughts, dowudraughts and
interface simulation requirement lies in these types of vortices. These impinge on the aircraft causing both low
operations, adequate motion cueing is crucial. and high frequency effects for which the pilot has to
compensate, so increasing workload and influencing pilot
Tests have indicated that the removal of motion cueing control strategy. The relative wind speed and direction
will have little effect on task performance in moderately around the flight deck and superstructure decides Ihe
demanding tasks. However, Ihe pilot's workload in degree of disturbance. An additional consideration is the
achieving that performance increases markedly. This effect of hot gases from the ships exhausts.
indicates that, without motion cueing, a pilot will exhaust
his spare capacity at an earlier point (reference 9). Most current simulators have crude, random, air wake
Vibration can also be a significant cue to the pilot. and turbulence modelling. Probably the best model
currently available is that developed by the US Navy and
Most current simulators use a standard six-degree of in use on the V-22 simulation in the MFS (reference 11).
freedom motion platform. Generally, the motion systems A version of this model is used in the US Coast Guard
are used for all flight tasks, although motion sickness of HH-60J simulator. A key issue with these turbulence
some users was noted. Motion cueing was considered to models is the way in which they are applied to the
be adequate, although heave cues were identified as a aircraft. In all cases this was through the aircraft centre
weak area. of gravity. More complex models would require a more
effective interface with the aircraft.
An interesting result from the survey was that some users
considered that the primary task of motion cueing is to A major question concerns the fidelity of simulation
reduce the effects of motion sickness. It was felt that required. From the training effectiveness standpoint the
few had a real grasp of the significance of motion cueing level of fidelity will depend on the level of training
in assisting the pilot to adopt the correct control strategy required. The number of different cases (ic
for a particular task. combinations of ship types with relative wind strength
and direction) that are required to be simulated can
 10-5

become large. Test and evaluation uses may require the National Research Council of Canada (reference 13)
generation of a large number of cases for ship/aircraft listed many source papers on the subject. Many of the
compatibility testing. papers referenced in this paper contain large reference
lists themselves.
Ship motion modelling
DRA air wake modelling work
Ship motion also plays an important part in the launch
and recovery tasks. Pilots must take-off and land inside The aim of the DRA work has been to develop a generic
the determined limits for the aircraft. In a simulator this model of the flow environment over and around the flight
means that a realistic presentation is important. The deck of a frigate-sized ship, suitable for use iu piloted
wind direction and speed, the direction that the sea is simulations. The model is related to certain
running and the speed of the ship are all significant characteristic features of the ship's geometry iu order
factors. As with air wake and turbulence effects, this that different ship types can be represented. The details
area has received little specific attention as deck landing of the work have been reported at reference 14 and this
was not part of the usual suite of tasks for naval rotary work should be consulted for the technical detail of the
wing training simulators. research. A brief outline is given below.

Most current simulators either provide simple harmonic The wind blowing over the flight deck and superstructure
motion driven models or crude models using a small of a ship induces a complex three-dimensional flow. The
number of superimposed sine waves. Behaviour solution adopted in the DRA model has been to identify
representing several ship types is generally available. basic flow types and perform off-line calculations to
These models are generally not suitable for conducting produce a suitable three-dimensional wind field. The
any form of realistic training. This is due to the basic flow types include:
complex nature of the wave environment (reference 12)
and to the fact that the environments that the ship • Corner flow - the flow towards and around a
operates in can vary markedly. corner when the approaching flow is directed
mainly towards a steep face
With ship motion modelling, as with turbulence
modelling, a significant question concerns the variety of • Lee rotor - a horseshoe vortex that forms iu the
conditions to be simulated. A small number of cases lee of a bluff body (figure 1)
lends itself to time histories stored in memory and
replayed as required, given that the data is scalable over • Edge vortex - corners of objects at a shallow
a significant number of sea states. Large numbers of angle to the incident wind over deck will
ship types and cases would dictate the need for some generate a vortex starting at the upwind corner
form of on-line modelling. and growing as it moves downwind (figure 2)

• Profile drag - au object in a wiud stream causes

Current DRA research a reduction in momentum in its wake

Current DRA research in the dynamic interface area is Inflow - necessary to 'balance the books' and
concentrated on three topics; ship air wake and ensure that there is no air flow through any aft-
turbulence modelling, ship motion modelling, and a facing solid surface such as the hangar end
continuing assessment of what is being achieved by state-
of-the-art training simulators - this allows constant
review of research objectives. In hand with this work
there are other research activities in progress that will WoD
benefit the programme directly or indirectly. These
include aircraft mathematical modelling activities and
work on helicopter motion cueing.

Ship air wake and turbulence modelling

General

This is a large subject and it is receiving much attention

in the DRA and other nations. A recent literature survey
carried out by the Institute of Aerospace Research of the Figure 1 - Lee rotor
10-6

speed along, across and normal to the forc/afl axis of the

ship. These increments are independent of actual wind
WoD
speed and are scaled by the wind over deck, or more
strictly by the datum wind over deck obtained by
combining vectorially the free stream wind speed with
the ship's speed.

The vehicle model used for the trials was the DRA
Helisim model (reference 17), configured as a Lynx
Figure 2 - Edge vortex helicopter with full automatic flight control system.
Helisim defines a generic uou-liuear helicopter
Each of these flow types has a characteristic form mathematical model with the main rolor represented as
determined by the speed and direction of the incident air a disc. Disturbances due to winds and turbulence are fed
flow and the geometry of the ship and its superstructure. to the model in three axes through the aircraft centre of
The dominant flow type will change with wind direction. gravity. The resulting forces are then transferred to the
Knowledge of the fundamental behaviour of flow round rotor hub within the model.
bluff bodies has been combined with data from wind
tunnel tests (references 15 and 16) and with flow Eight relative wind cases have been generated to date,
visualisations recorded on video to set overall with four wind directions and two wind speed cases
dimensions, in both extent and strength, of each flow provided. Figure 3a shows the flow vectors at a point
type. Contributions from each type have then been one ship beam width to port of the hangar centre-line for
combined in a linear manner and with suitable blending a relative wind case of 30 degrees on the starboard bow
to achieve the desired results. With further information, (green 30). Figure 3b shows the same wind case in the
perhaps from full-scale measurements, it would be cross-deck plane at 1.1 beam widths aft of the hangar
possible to adjust the contribution of each of the face, approximately the position of the rotor hub during
constituent flow types. lauding. These figures clearly show the changes in mean
flow velocity that the aircraft will encounter around and
The aircraft is strongly affected by the variations in local over the flight deck.
wind with helicopter position, rather than the absolute
value. Furthermore, as the presence of the ship will
change the local wind flow, the model is expressed solely
in terms of incremental velocity vectors which, when
scaled by and added to the datum wind over deck, give
the total local wind components.

A simplified ship geometry has been assumed, in which

the rear portion of the hull has a flat flight deck and a \ i I : \:?r
near-rectangular hangar. Provision is made for fence tethfrv/-
structures on top of the hangar and for the slightly obtuse pi» ^ /—; ■»

angles used between surfaces to reduce radar reflections. Ä* 7—7-

The model is designed primarily to provide appropriate -0.5 0 0.5 1 1.5 2 2.5 3 3.5 4

flow conditions for positions aft of the hangar face.

Forward of this, the superstructure becomes more Figure 3a - Vector plot of wind variation, fore/aft view
complex. Flow in this region is represented only 3

approximately, to ensure that there are no unrealistic

disturbances as the helicopter flies through. Specific
geometric parameters are used appropriate to the ship in

For a given wind direction, appropriate flow elements

are selected and combined. The results are expressed as
a table of wind vector components in three dimensions.
The position of the helicopter relative to the flight deck
is input to the model, normalised by a characteristic
length given by the ship's beam. Output from the air -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1
wake model is expressed as a spatial distribution of three
non-dimensional incremental components of the wind Figure 3b - Vector plot of wind variation, lateral
 10-7

The model performed well in tests, with the aircraft questionable and ship time is very expensive. A version
experiencing the gross, low frequency effects that the of the air wake model may be developed in the future to
pilot expected (reference 18). These included updraughts conduct an AFS assessment of data collected using US
generated by the up-wind flight deck edge, masking equipment on a UK frigate. Laser anemometers could be
effects in the lee of the superstructure and the used, which can gather data way from the flight deck and
downdraught experienced from the roof of the hangar more rapidly than an instrumented niast. Computational
during the descent to land. Evaluation pilots have fluid dynamics may be another method, as has been used
commented that effects were representative of the real by the US Navy (reference 21).
world, producing correspondent increases in pilot
workload. This has demonstrated that the concept of this Other considerations
model is valid and should be developed further.
None of the models currently in use take account of the
The DRA model currently suffers from a lack of effects of ship motion on the air wake and turbulence
appropriate turbulence modelling. The current flow fields. In high sea states the motion of the
turbulence model in use on the AFS was originally superstructure will cause distortions in the flow.
derived for fixed wing work. In trials with the air wake However, currently it is not even clear whether the
model it was very ineffective in producing appropriate effects of ship motion are significant. Some work has
reactions of the aircraft. It is unclear whether this was been carried out that may indicate the levels of influence
due to problems with the turbulence model itself, or with (reference 22), but further work will be required to
the reaction of the aircraft mathematical model. Work is adequately define the problem.
under way to attempt to define the problem and produce
a new or revised model, as appropriate, for Both the US and UK models use large amounts of data
demonstration on the AFS in 1996. in look-up tables to represent the steady state flow field.
While this may be acceptable if only a limited number of
This research work concentrated on a proof of concept, cases are needed problems may arise with on-line data
generic approach. Further work will be required to storage if larger numbers of cases are required.
investigate whether specific data will be required for
particular ship types. Certainly, in the training Evaluations on the AFS have indicated that aircraft cues
environment it is likely that, unless a vessel has other than the pure reaction of the aircraft are important
particular features that produce unusual flow patterns, a in simulating air wake and turbulence effects. The cues
generic model may suit several similar ship types. provided by vibration of the aircraft were noted as being
However, little work has been conducted in this area. conspicuous by their absence. These effects could
One study (reference 19) indicated significant differences include reaction of the main and tail rotors to vortices
between a generic model and air wake tests carried out passing through them and to the influences of main
at sea on an a real vessel. For ship compatibility testing rotor/tail rotor interactions. These findings have
more type-specific work would be required with detail prompted work in this area to be included in the future
data for particular ship type being made available. DRA programme.
Indeed, there has recently been an approach from UK
MoD to carry out such a helicopter/ship air wake and Future DRA work
turbulence assessment to assist in ship design work for a
new class of vessel. The aim of the work would have The future programme dealing with ship air wake and
been 1o attempt to reduce aircraft operating difficulties turbulence modelling at DRA plans to:
due to the effects of ship air wake and turbulence.
Unfortunately, the technology is not yet mature enough • Integrate the model with an aeroelastic blade
to reduce the risks of taking on such work to a sufficient element aircraft model (in 1995).
level. However, it is likely that this type of work will be
undertaken in future by the DRA using the AFS. • Develop more relative wind speed and direction
cases (up from the current 8).
Wind data
• Examine the possible use of real ship data
Wind tunnel testing can be an expensive way of gathered at sea to provide the steady state flow
gathering the necessary data for inclusion in these pattern.
models. There are alternative methods of generating the
necessary information. One would be to conduct testing • Develop and integrate an improved, or if
at sea on actual ships. Some countries have used an necessary, new turbulence model, depending on
instrumented mast mounted on the flight deck for this the findings of work to be carried out this year.
purpose (reference 20). However, results have been
10-8

The work oil turbulence will be assisted by a flight trial Ship motion modelling
usiug a heavily instrumented Lynx helicopter (the DRA's
ALYCAT Lynx, ZD559) to investigate the reaction of General
the aircraft in flows behind bluff structures. The trial
will also record pilot control strategies and reactions in As with ship air wake and turbulence effects, ship
flight in these regions. Current plans call for a number motion is a key task driver in the ability of the pilot to
of wind cases to be evaluated. A simulated ship carry out a successful launch or recovery. Ship motion
environment will be created by use of containers is principally dependant on the sea state and the hull
assembled on an airfield to represent the approximate form of the ship (reference 12). 'Sea state' includes
dimensions of a frigate flight deck and hangar. The such parameters as wave direction, significant wave
aircraft is equipped with a full suite of instrumentation height and other dynamic characteristics such as modal
including pressure and strain-gauge fitted main rotor period and will be influenced by the sea area being
blades. The results from this work will be applied to the traversed and the season of the year. The way the ship
air wake, turbulence and aircraft mathematical modelling responds to seas will be governed by the natural
programmes. behaviour of the vessel and by such variables as ship
speed and heading with respect to the waves.
An area not currently being addressed in the DIM.
simulator programmes is that of the influence of the A 'real wave' environment can be complex because it
ships hot gas exhausts on the aircraft. There are can consist of a long wavelength swell, induced by a
currently no plans in the DRA programme to address wind that has been blowing several hundred kilometres
this, but some work that may be of future use is being away, in combination with a locally generated wave
undertaken in the US. Again, the aircraft simulation system induced by the local wind. The underlying
community is faced with not being able to quantify the direction of movement of each main component can be
extent of the influences of a particular feature on the quite unrelated.
fidelity of the simulation or the difficulty of the piloting
task. Although a wave environment may have been generated
by a complex set of events, waves at a given point and
The provision of on-line real-time models, as available a given time can appear to be like a series of
for ship motion modelling, would be a desirable aim. superimposed sine waves. The resulting ship motion has
Once validated, these models would not require the a similar sinusoidal appearance (see figure 4). Some of
collection or generation of the necessary steady flow and the key properties of ship motion, which a simulation
turbulence data. However, developing such models, with should reproduce, are the characteristic amplitudes of
sufficient fidelity, and in real time, is some way iu the movement, the frequency content and the appearance of
future, if it were to be possible in a cost-effective lulls when the ship experiences a period of low motion
manner at all. The technical challenges and cost of the activity. It is these lulls, or quiescent periods, that
necessary computing power may see this remain an provide opportunities for the pilot to lauuch or recover
unobtainable goal. the aircraft. Consequently, it is important that they
appear in any simulation of ship motion.
Validation is always going to be a difficult issue with the
air wake and turbulence work. One suggested course
would be to fly a fully instrumented aircraft to a ship in SHIP HOTION SEA STATE 5

trials at sea. The same runs would then be performed in

the simulator using an aircraft model configured
appropriately and with all other models set up to provide
the conditions found at sea. All other considerations
aside, the pilot control strategy and the reaction of the
aircraft could then be compared with those found iu the
flight trial. However, this would be an expensive
exercise and the results, due to other simulation
limitations if nothing else, would likely be ambiguous.
Validation questions remain to be adequately addressed.
Another problem that none of the current modelling work
addresses is the influence on the flow field generated by
the insertion of the aircraft, and in particular the main TIME (SECS)

rotor. Distortions are bound to occur - but are they

significant? Questions such as these remain to be
addressed. Figure 4 - Example of typical ship motion trace
 10-9

There are two methods of providing a representation of models can be configured for particular ship types. The
ship motion: models in use on the MFS and HII-60J simulators have
been viewed and appeared to produce good results.
• By use of an on-line real time computer model These models use 16 sine-waves, although current nou
real-time models can use up to 40.
• Employment of time history data
It is intended to fully evaluate the US Navy ship motion
The latter method is practicable as long as a large model on the AFS during 1995. An evaluation will then
number of cases are not required. DRA work has used be made of the level of fidelity of the model and its
time history data, produced using a non-real time suitability for particular tasks.
computer model, replayed in real time for simulator
trials. A number of cases were generated and these can Aircraft mathematical models
be used depending on the needs of a particular trial. The
data is produced for sea state 5.5, but can be scaled As with other aspects of simulation discussed here, the
between sea states 4 and 7, as the basic characteristics of field of aircraft mathematical modelling of helicopters is
the wave forms remain the same over this range. undergoing constant evolution. Many of the new
However, only one case is stored for use at any one features becoming available are not revolutionary. It is
time. The data covers 20 minutes of motion which at the the ready availability of faster and cheaper computing
end replays via a smoothed loop. The start point of the power that is allowing more complex models to be run
data is varied for every run carried out during a in real time. There are many features that are, or soon
simulation sortie so that the pilot does not become will be, available with current technology:
familiar with a particular portion of the data. Data
provides ship's roll, pitch, heave, yaw and sway axes. • An individual blade clement model for the
rotor. This would include modelling of flapping
The DRA currently holds samples of data for all Royal dynamics and lead/lag, torsion and rotor speed
Navy aviation-capable ships. This method is attractive effects. Stall and Mach effects could also be
because it provides an easy solution. The major incorporated.
limitation is that a large spread of cases would require a
significant amount of memory storage space. With a • Incorporation of forces and moments acting on
simulation sample rate of 50 times per second and the fuselage, fin and tailplane for an appropriate
assuming a 20 minute time history is the minimum range of pitch and sideslip incidence.
acceptable, then 60000 data points would be required for Modelling of rotor wake effects will be possible
each axis. This is a lot of data. It may not be necessary in the near future.
to sample the time history at the full data rate of 50
times per second. Linear interpolation techniques may • Blade element tail rotor modelling should soon
allow a lower sample rate, of perhaps 10 times per be available, as will main rotor/tail rotor
second. However, this has yet to be evaluated on the interaction effects. These have au important
simulator. Pilot acceptance of the current method of bearing on the operational envelope of some
representing ship motion on the AFS has been high. No helicopters.
adverse comments on ship motion have been recorded
and, whilst there has been little unprompted praise, this As noted above, DRA efforts include running the current
usually indicates that the simulation is adequate for the air wake and new or improved turbulence models with au
task being performed. aeroelastic blade element main rotor model this year.

It would also be possible to use real ship time history As has been suggested, the aircraft mathematical model
data. This would probably only be required for specific has a profound influence on the effectiveness of the
applications, but could be useful. Real data is available simulation of air wake ad turbulence effects. Currently,
in limited quantities for a few conditions and could be the outputs of models are generally applied either at the
used if necessary. The same problems apply with the aircraft centre of gravity or at the main rotor hub. As
use of this data as with computer-generated data in terms the aircraft models become more complex and of higher
of flexibility, storage space and a limited number of fidelity air wake and turbulence models will need to
available conditions. evolve to utilise the additional fidelity available. Having
the aircraft model react to the other models at several
On-line, real time models have been implemented by key points on the fuselage, as well as across the main
other research simulators and for some training devices. rotor may be appropriate. Inclusion of the tail rotor, and
Generally these models employ many sine waves the reactions between main and tail rotors will also be a
superimposed to provide the necessary motions. These consideration.
10-10

Visual cues In terms of scene content, phototexluiiug can provide

improved cueing. Although providing more 'realism' to
Achieving adequate visual cueing for the helicopter/ship the picture and assisting in providing positioning
interface environment requires a high performance visual information, phototexturing is not the whole auswer.
system. To achieve meaningful simulation of deck Individual modelling of key features on the flight deck
operations the visual system will become a major cost and superstructure is still necessary. Modelling of a
driver of the device. Research into this area for the dynamic sea surface is also a highly desirable feature.
dynamic interface will begin this year. The aim of the Without this the picture can seem very artificial in even
research will be to identify the key elements of visual moderate sea states. The ship has motion, but the sea
modelling on which effort should be concentrated when surface remains flat. In higher sea states, height cueing
modelling a vessel. can become difficult. Accurate modelling of ship-
mounted visual aids is also essential for training devices.
What is known is that a large field of view is required,
both to provide the pilot with a sufficient view of the Research at DRA Bedford is concentrating on scene
vessel, but also to maintain adequate levels of peripheral content issues at this stage. There are plans to conduct
cueing. A high scene content is also important. The eye-tracking experiments, both in the simulator and
level of cueing provided by a vessel at sea is somewhat during ship trials, to identify the key features the pilot
less than might be available in a confined area on laud. uses during the final approach and lauding tasks. Sonic
Consequently the pilot has to rely on a limited number of work on pilot eye movements during deck landings was
cues to stabilise and position the aircraft over the flight carried out by the Department of National Defence of
deck. The sea surface also provides the pilot with cues Canada some years ago in support of visual aids work
as to height and ship motion. (reference 23). This showed that pilots concentrate ou
very specific areas of the ship superstructure during the
Currently, a phototextured Type 23 frigate model is used landing task. If these are identified for Royal Navy ships
on the AFS (figure 5). This visual scene has been it should be possible to concentrate modelling efforts ou
displayed on three cockpit monitors giving a azimuth these areas and so improve the level of realistic cueing.
field of view of 120 degrees and a maximum of 48
degrees in elevation. This has proved to be limiting in Motion cueing
some respects due to the lack of azimuth cover to the
right and downward field of view to the right. Pilots Motion cueing is an important element of high fidelity
have had to compensate for these deficiencies by helicopter simulator. The level of moliou cueing has an
modifying the way they carry out the landing task. impact on the workload the pilot experiences in
Obviously, this is not desirable. Enhancements to the attempting to carry out a task to a particular performance
visual system will be in place soon that will increase the requirement. No helicopter/ship interface specific
azimuth field of view to 210 degrees and the elevation research work is being conducted at present. However,
field of view to the right to 66 degrees. These other work carried out by the DRA (reference 9) has
improvements should have a significant positive impact indicated the benefits of dynamic seats to provide
on the level of cueing available ou the AFS. additional, and in particular, heave cues. Further trials
evaluating the use of a multi-axis dynamic seat for
helicopter tasks will take place this year using the AFS.

Use of the AFS for dynamic interface research

General

The AFS at DRA Bedford provides a high fidelity

general purpose research tool with a high degree of
flexibility to enable tailoring for a wide range of fixed
and rotary wing applications. The simulator cau be
configured to the needs of a particular task by selecting
hardware and software options. The key facilities
include motion and visual systems, cockpit modules, pilot
controls and primary flight instrumentation. These are
complimented by the capability to build software models
of the aircraft, its systems and the natural environment in
Figure 5 - AFS visual model of Type 23 frigate which it is operating. The facility's foremost asset is the
 10-11

Large Motion System which provides a unique and Navy deck approach and lauding profile. This uieant
highly capable motion cueing system (figure 6). placing the aircraft on a 3 degree glide slope and a 165
Notably, the high displacements, velocities and degree radial from the port bow. Pilots were instructed
accelerations shown can be achieved simultaneously in all to fly a standard approach to the 'port wait' position. At
five axes. More details can be found in reference 24. this point the aircraft is brought to a hover alongside the
flight deck where the main rotor should be clear of the
ships side and the aircraft in the correct fore/aft position.
The aircraft is then held in this position until a suitable
quiescent period in ship motion is identified. The
aircraft is then manoeuvred over the lauding point and a
landing made when the pilot is satisfied with positioning
Pitch
and deck motion.
Roll

For assessment purposes this profile was divided into two

Surge Sway MTEs, as follows:

a. Approach to and maintenance of a steady hover

alongside the flight deck.
PERFORMANCE ENVELOPE
Axis Displacement Velocity Acceleration b. Sidestep manoeuvre, positioning over the flight
Pitch ± 0-5 Rad . i 0-5 Rad/s i 2 Rad/s2
deck and lauding.
Roll 0-5 Rad l-O Rad/s 3 Rad/s2
Yaw 0-5 Rad 0-S Rad/s 1-5 Rad/s*
Sway In the simulator the task was altered by presenting the
(Surge)
4m 2-5 m/s 5m/s2(o-5g)
Heave 5m 3 m/s 2
10m/s (l-0g) pilot with various combinations of night and day visual
conditions and using different sea states.
Dims.in mtrs.

Assessment

Success in completing the task was measured not only

against the pilot perception of his performance, but also
by setting task performance parameters. These
parameters were monitored and recorded and, in the
simulator, were available to the pilot during debriefing.
The parameters were structured to reflect the Cooper-
Figure 6 - DRA AFS Large Motion System Harper handling qualities rating (I1QR) scale (figure 7
and reference 5) for pilot subjective assessment.
Two areas are being studied: handling qualities for Consequently, task performance was divided into bauds
helicopters involved in ship operations and pilot visual of 'desired' (Level 1), 'adequate' (Level 2) and
cueing aids for ship operations. 'unacceptable' (Level 3). These ratings were included in
a post-run debrief questionnaire. This was backed by a
Tlie task post-sortie questionnaire. Task performance parameters
included accuracy of landing and vertical velocity at
All tasks were assessed using, as much as possible, the touchdown, time spent in the hover over the flight deck
same test and assessment methodologies. This assisted and the number and severity of overtorques.
in being able to compare results, as well as providing
pilots with continuity and reducing the overheads in Handling qualities
running trials.
The handling qualities work (refereuce 2) aims to
Task design was based on a mission task element (MTE) establish handling qualities criteria for the operation of
concept promoted in combat helicopter flying qualities helicopters from ships. This will provide the MoD with
research, using the methodology of the US Army a comprehensive document to assist iu the specification
Aeronautical Design Standard 33 (ADS-33) (reference and assessment of future maritime rotorcraft, maintaining
25). This is the generally accepted methodology for MoD's 'intelligent customer' status. The work is
military helicopter handling qualities. A MTE represents looking at the full range of control response types as well
a discrete task, representative of a realistic mission phase as the impact of key control parameters. These
with defined initial and terminal conditions. The mission parameters include roll and pitch baudwidths, control
phase under investigation in these trials was the Royal powers and the influence of vertical damping and thrust
10-12

AIRCRAFT DEMANDS ON THE PILOT IN SELECTED PILOT

ADEQUACY FOR SELECTED TASK OR
TASK OR REQUIRED OPERATION» RATING

i
REQUIRED OPERATION' CHARACTERISTICS

Excellent Pilot compensation not a far

Highly desirable desired performance
LEVEL 1
Good Pilot compensation not a factor for
Negligible deficiencies desired performance
Fair—Some mildly Minimal pilot compensation required for
unpleasant deficiencies desired performance
— 354 -
Minor but annoying Desired performance requires moderate
deficiencies pilot compensation
Deficiencies
Moderately objectionable Adequate performance requires
warrant
deficiencies considerable pilot compensation
LEVEL 2
improvement
Very objectionable but Adequate performance requires extensi
tolerable deficiencies pilot compensation __^
-6V4-
Adequate performance not attainable with
Major deficiencies maximum tolerable pilot compensation.
Deficiencies
Controllability not in question LEVEL 3
Considerable pilot compensation is required
■ require
Major deficiencies
for control
improvement — B'A
Intense pilot compensation is required to
Major dc'icienciej
retain control

No^llmprovement Control will be lost dui

Maior deficiencies
iandatory required operation

'Definition of required OD*UI'on ifivo !i dftignilion ol flight phiic §nd/oi

Pilot decisions lubphjiel wilh »ceofnpjny.ng cond'l
Cooofi H»rper fiel NASA TNO-5153

Figure 7 - Cooper-Harper handling qualities rating scale

margin ou task performance, Other features being Figure 8 shows the lauding scatter for two of the
investigated include: roll/pitch bandwidth configurations trialled. A tighter
lateral pattern is evident for the higher bandwidth case,
• The impact of integrating controls and displays as would be expected. Up to a point, higher bandwidth
by providing piloting information on a helmet- provides more precise control and allows the pilot to
mounted display improve task performance. Pilots were consistently
concerned about fore/aft cueing and this is reflected iu
• Investigation of the use of trim systems the results. The large fore/aft spread is probably due to
deficiencies in peripheral cueing and the pilot not being
• The impact on task performance of limit able to see the deck edge. Although this was recognised
protection systems as a limitation in the simulator, fore/aft positioning is
also a major concern in the real world.
• The influence of providing a limited authority
carefree handling system to allow the pilot to
more fully utilise the performance of the vehicle
\Referenced to deck lock grid centred 1
BLowbandwidth 1
O High bandwidth 1
15
The work has involved the use of the DRA Conceptual
■
Simulation Model (described in reference 22 and 26). 1 ^
This model was originally developed to permit the ■
exploration of helicopter handling qualities in a 1 0.5 _
systematic fashion. Characteristics in any axis can be o
■
changed without impact on the response in other axes 8. ° 0"
1
►.
and cross-coupling can be entirely eliminated if required.
Response types can be selected and a wide variety of 1 -0.5
Ü.

parameters can be altered to assess their impact ou o

-1
handling qualities. o
I I 1- 1

Work so far has concentrated on large helicopter -15 -1 -0.5 0 0.5 1 15

Lateral position (m)
operations from small ships. Roll and pitch bandwidth
were altered, together with control power, using sea state
Figure 8 - Landing scatter comparison for high and low-
to vary task difficulty. Heave axis evaluations have used
bandwidth pitch and roll configurations
damping and thrust margin as the variables, again using
sea state to vary task difficulty. The impact of providing
the pilot with an audio torque warning has also been The pilot ratings returned for the landing task for roll
investigated. bandwidth are shown in figure 9. The results clearly
show that, although Level 1 performance was possible at
 10-13

the higher baudwidths lested, this was only possible in Visual aids
sea states up to three. From the results it could be
hypothesised that au increase in bandwidth to 3.5 rad/sec Current visual cueing aids ou small ships of the Royal
would give Level 1 performance in sea states up to five. Navy were generally developed in the early days of
helicopter operations with small helicopters and there has
been little evolution over the years. The introduction of
ADS33C AOS33C
Level 1/2 Level 1/2 larger aircraft for small ship operations, such as the Sea
general track
a
MTEs King and, in the not too distant future, the 1:11101
7 ^ Sea state 0 Merlin, have highlighted the shortcomings in visual
f 6 O Sea state 3 cueing during the final part of the approach to the ship
VI
O Sea state 4 and over the flight deck, particularly when conditions are
T3 5
▼ Sea state 5 poor. If full benefit is to be drawn from the investment
o-

.14
"D ^c,,,. in the highly capable Merlin, then improved pilot visual
Suggested
S ; cues for ship operations will be necessary. The visual
■<=
X
3
-^
^ performance
aids work has investigated how to provide these
6
2
^>
improved pilot visual cues through developing patterns of
1
1
12 3'
electro-luminescent panels in place of traditional white
Roll bandwidth (rad/sec) floodlighting. These panels can be placed around the
flight deck and superstructure and provide the pilot with
Figure 9 - Mean HQRs for roll bandwidth showing all familiar patterns (figure 10). Geometric positioning aids
configurations and sea states have also been developed to assist the pilot iu positioning
the aircraft more accurately over the flight deck. Iu
The results have clearly shown that the lauding task is hand with this, work has also been carried out to develop
the critical mission task element. There are also an indication system which shows the position of the
indications lhat the landing task closely corresponds to aircraft deck lock probe relative to the centre of the ship
the ADS-33 Level 1/Level 2 boundary for a task with a deck lock grid (figure 11). The simulator has been used
significant tracking element. It also indicates that, not only to develop display formats, but also to develop
particularly for the vertical axis, specific ship operation strategies for use of the device. This work is covered in
mission task elements and task performance parameters reference 27.
are required. These initial results suggest that the ADS-
33 Level 1/Level 2 boundaries need modification for For this work the simulator has been used as a versatile
helicopter/ship operations. and easily changed system to try out new ideas, modify
them or reject them. It is then used as a development
aid

Figure 10 - ELP pattern shown installed on Type 23 frigate during ship trials
 10-14

aid to move concepts forward towards flight trials. • Assessment of deck handling and recovery
Finally, the simulator can be used in direct support of systems
flight trials and as something of a training aid so that
pilots know what to expect when flying to a real ship • Further in the future it may be possible to use
with these aids for the first time. Without the simulator the simulator to provide advice to ship designers
the development of solutions would have taken longer on such items as air wake and turbulence
and involved considerably more financial and physical reduction methods and ship stabilisation systems
risk. There is no doubt that it would also have been
more expensive.
£>9
r [""" ] i ■=g= •-J3
ra 8 X
X
L_|-r T-{-J 1i S 7
! i . ;
1 1
_ 1 1 1« X v
i
i
i
1
1
1
1
+ 1
I
1
J«^
£s T
-+---T- c 4 - X ,, X X -X--X--*--
i * 1 1
! ! _ ir *■ J,
|._J L_4_J 4= C
ro
-C
3
2 - X X
i 1 L !
non-HR HPI assisted
" T
L_+_x
T 1
T_^_^
■ Day/sea state 0 0 Night/sea state 0
A Day/sea state 4 V Night/sea state 4
1 1 i i
■-+---- i i
i 1
' 1 Figure 12 - Mean HQRs for landing MTE with and
i .-+-- + - i
i 1 » 1 i i without hover position indication (HPI) system
II iii
__+.__!_ L_41

L L !
Referenced to deck lock grid centre|[
i T 7 1
I-- + -H T- + -H

I
It III

1 1 i i
1
I
1
1
1
■~*
1
+-
1
i i ^
1
1 1 • 1 £ 1h
1 ~i +-
1
1 1 1 1
i
i
i
i 9
BNon-HPI

M
1 1 i i
1 1 i 1--+--)
1 l_ I 1 =5= OHPIassisted
Figure 11 - Hover position display, showing examples of
display conventions, the central area being the deck lock £o -1h
grid.

-2-10 1
Figure 12 shows a comparison of the Cooper-Harper Lateral position (m)
ratings for landings conducted with and without the
assistance of the hover position indication system. The Figure 13 - Scatter for landings with and without hover
plot shows lhat pilot ratings for landings with or without position indication (HPI) system
the indication system were very similar. The lauding
scatters for the same landings are shown in figure 13. It
is clear that there was a decrease in scatter for those Limitations
landings where the pilot had use of the indication system.
This improvement in performance was balanced by pilots The AFS dynamic interface simulation does have
seeing a marginal increase in workload. limitations. The deficiencies have been covered in most
part above. Of particular concern in the AFS were the
Future work may include: field of view, as well as provision of an effective air
wake and turbulence model. The nou-dyuauiic sea
• Development of autopilot control laws for a surface provided by the visual system is a limitation
differential GPS-driven automatic ship approach when attempting to conduct studies at high sea states.
guidance system. Although already in flight test The motion cueing will never be perfect, although it cau
the simulator will be used to support be very good and the AFS probably provides one of the
development work best motion systems in the world.
 10-15

Having noted these deficiencies it should be remembered • Effective helicopter/ship dynamic interface
that real aircraft also have deficiencies. While we need simulation could provide significant cost and
to strive for improvements in the simulation, the trends operational benefits in pilot training, test aud
for the results were considered to be realistic. Ship trials evaluation and research
have shown that results are always better in the real
world than in the simulator. Consequently, it may be • Adequate aircraft data is vital to the success of
safe to consider that the use of the simulator docs not the simulation of the flyiug qualities of the
lead to over confidence - if a task can be accomplished aircraft
in the simulator where it is more difficult because of the
deficiencies, then it will be safe to perform the task iu • An effective dynamic interface simulation
the real world. Indeed, from the helicopter/ship trials capability has been built up on the DRA AFS.
we have seen that for identical runs in the AFS and the providing a valuable research aud development
real world, with broadly similar aircraft types and tool for work iu this area
weather conditions the pilot rating poiuts are typically
about 0.5-1.0 greater in the simulator. This is far better • International collaboration continues to be a
than could be achieved a few years ago. Continual valuable element in progressing dynamic
improvement in the fidelity of the dynamic interface interface simulation capability.
simulation available on the AFS is sought. The
improvements planned for the AFS visual system have The key challenges that remain may include:
already been outlined. A head-down glass cockpit is also
being fitted this year. • Ship air wake and turbulence is oue of the
major task drivers and a key shortfall in
dynamic interface simulation. Much research
Collaboration work is being conducted to produce realistic
models which will cue pilots to the correct
The dynamic interface research programme has benefited control strategies
from an international collaboration between the UK and
the US, Canada and Australia. The collaboration is a • Improvements in aircraft mathematical
vehicle for exchanging expertise, information and modelling will be necessary to take advantage of
models, as well as promoting common data formats and improvements in other areas of the simulation
sharing the results of relevant research. Work is also
being undertaken into developing common pilot rating • Improved motion cueing, possibly through the
scales and dynamic interface test methodologies. There provision of multi-axis dynamic seats
have also been regular exchanges o pilots and engineers
for trials. • Validation is an important issue that demands
significant further work

Conclusions
References
This paper has presented an overview of the current
work being undertaken by the DRA to improve piloted 1 Jewell W F, Clement W F, Johns J B, "Realtime
simulation of the helicopter/ship dynamic interface. piloted investigation of helicopter Hying qualities
Work undertaken to determine the shortfalls of current during approach aud lauding on uouaviation ships",
simulators has also been reviewed, as has other current in "AIAA 86-0490", 24,h AIAA Aerospace Science
research work being carried out to improve Conference, Reno, Nevada, January 1986.
helicopter/ship operating limits using piloted simulation.
The following conclusions are indicated: 2 Täte S J Lt RN, Padfield G D, "Simulating flying
qualities at the helicopter/ship dynamic interface", in
• Helicopter/ship dynamic interface simulation is "American Helicopter Society Forum 50",
demanding and capability adequate for Washington DC, May 1994.
worthwhile research and pilot traiuiug is only
just becoming available 3 VanderVliet G, "The test, evaluation, development
and use of a Manned Flight Simulator to support
• It can be difficult to identify the influence of Navy developmental testing of the V-22 Osprey",
one element of the simulation on another and to November 1992.
determine the impact on overall fidelity of each
element
10-16

4 "Aircraft ship operations"; AGARD Conference 17 Padfield G D, "A theoretical model of helicopter
Proceedings 509, Seville, Spain, 1991. flight mechanics for application to piloted
simulation", RAE Technical Report TR81048, April
5 Cooper G E, Harper R P, "The use of pilot rating in 1981.
the evaluation of aircraft handling qualities", NASA
TN D-5153, April 1969. 18 Tomlinson B N, Täte S J Lt RN, "Initial evaluation
of a new ship air wake model for use in piloted
6 Condon G W, "Simulation of Nap-of-the-Earth simulation", DRA Customer Report (unpublished),
Flight in Helicopters", in "Computer Aided System March 1994.
Design and Simulation", AGARD CP473, 1990.
19 Erm L P, "A preliminary study of the airwake
7 Täte S J Lt RN, "A review of current Naval model used in an existing SH-60B/FFG-7
helicopter training simulators and implications for helicopter/ship simulation programme", DSTO
further technological developments", DRA Customer Australia Report No DSTO-TR-0015, May 1994.
Report (unpublished), November 1993.
20 Burgess M S, "Ships airwake measurement system
8 Tale S J Lt RN, Cheyne S Lt Cdr RN, "Evaluation (SAMS): development of instrumentation and testing
of Naval helicopter pilot training simulators - a aboard US Navy ships", NAWC AD Report No
review of current capabilities and procurement RW-88R-93, December 1993.
issues", DRA Customer Report (unpublished),
December 1994. 21 Tai T C, Carico D, "Simulation of DD-963 ship
airwake by Navier-Stokes method", in "AIAA 24th
9 White A D, "G-seat heave motion for improved Fluid dynamics conference", AIAA 93-3002, July
handling in helicopter simulators", RAE Tech 1993.
Memo FM33, 1989 (Also presented at the AIAA
Flight Simulation Technology Conference, 1989). 22 Arney A M, "FFG-7 ship motion and airwake trial:
Part II - Removal of ship motion effects from
10 Galloway R T, Cannaday R L, "Guidelines for measured airwake data", DSTO Australia Report No
source data requirements for US Navy aircrew DSTO-TR-0093, October 1994.
training devices', NTSC report 211/1-91, September
1991. 23 Beach A, "Eye movements of Sea King pilots during
deslroyer deck landings at night", Defence and Civil
11 Täte S J Lt RN, "Report on visit to NAWC AD Institute of Environmental Medicine Report No 84-
Dynamic Interface Department 1-5 March 1993", R-54, December 1984.
DIM. Working Paper FSB WP(93)030, March 1993.
24 Padfield G D, Charlton M T, Kimberley A M,
12 Lloyd A R J M, "Seakeeping: ship behaviour in "Helicopter flying qualities in critical mission task
rough weather", Ellis Horwood, 1989. elements. Initial experience with the DRA Bedford
Large Motion Simulator", in "Eighteenth European
13 Garry E A, "Aerodynamic aspects of the Rotorcraft Forum", September 1992.
helicopter/ship interface - a literature survey", NRC
Canada Report No 32166, October 1994. 25 Anon, "Aeronautical Design Standard (ADS) 33C -
Handling Qualities for Military Helicopters", US
14 Woodfield A A, Tomliuson B N, "Ship airwakes - Army AVSCOM, St Louis, 1989.
a new generic model for piloted simulation",
AGARD Flight Vehicle Integration Panel symposium 26 Buckingham S L, Padfield G D, "Piloted simulations
on flight simulation, May 1995. to explore helicopter advanced control systems",
RAE Technical Report TR 86022, 1986.
15 Whitbread R E, O'Neill P G, Quincey V G," 'A
study of the wind flow over the flight deck of a 27 Tale S J Lt RN, "The development and application
typical frigate", BMT Project 35019, June 1986. of novel visual aids to increase operational limits at
the helicopter/ship dynamic interface", in "American
16 Johnson R, Whitbread R E, "Assessment of airflow Helicopter Society Forum 51", Fort Worth, Texas,
quality over naval flight decks", BMT Project May 1995.
35019 Final Report, September 1987.
 11-1

Simulation of Rotor Blade Element Turbulence

R. E. McFarland and Ken Duisenberg*

NASA Arnes Research Center

Mail Stop 243-5
Moffett Field, CA 94035-1000 USA

SUMMARY included in SORBET, but rather occurred as induced responses

A turbulence model has been developed for blade-element from the superposition of vertical components at the blade
helicopter simulation. This model, called Simulation of Rotor stations.
Blade Element Turbulence (SORBET), uses an innovative
temporal and geometrical distribution algorithm that preserves The complete SORBET-Black Hawk model, including all the
the statistical characteristics of the turbulence spectra over the requisite motion and visual communication software, was exe-
rotor disc, while providing velocity components in real time to cuted on a single processor (VAX 9000) with a cycle time of
each of five blade-element stations along each of four blades. 12 milliseconds. Cyclic phenomena are generally experienced in
blade-element rotor models. The introduction of turbulence did
An initial investigation of SORBET has been performed using a not amplify these effects.
piloted, motion-based simulation of the Sikorsky UH60A Black
Hawk. Although only the vertical component of stochastic For the initial proof-of-concept simulation, only the vertical
turbulence was used in this investigation, vertical turbulence component of turbulence was used. Although pilots recognized
components induce vehicle responses in all translational and the ab sence of the horizontal disturbances under certain flight
rotational degrees of freedom of the helicopter. conditions, their comments were generally quite favorable
concerning the vehicle performance in turbulence. As opposed
The single-degree-of-freedom configuration of SORBET was to the conventional body-fixed approach to turbulence modeling,
compared to a conventional full 6-degrees-of-freedom baseline SORBET has, at least incrementally, improved realism.
configuration, where translational velocity inputs are super-
imposed at the vehicle center of gravity, and rotational velocity THE SORBET MODEL
inputs are created from filters that approximate the immersion For implementation into the GENHEL UH60A mathematical
rate into the turbulent field. For high-speed flight the vehicle model of Reference 1, an heuristic model of turbulence
responses were satisfactory for both models. Test pilots could (SORBET) was developed that preserves the Gaussian statistical
not distinguish differences between the baseline configuration characteristics of turbulence filters over the rotor disc. This
and SORBET. In low-speed flight the baseline configuration model avoids the large computational and storage requirements
received criticism for its high frequency content, whereas the of recent investigations into rotorcraft turbulence models
SORBET model elicited favorable pilot opinion. (Refs. 2, 3), and imposes a modest computational overhead.
SORBET was designed to create and propagate all three trans-
For this helicopter, which has fully articulated blades, results lational components of MIL SPEC turbulence (Ref. 4) in real
from SORBET show that vehicle responses to turbulent blade- time. The turbulence velocity components are superimposed at
station disturbances are severely attenuated. This is corroborated each blade-element location as a function of vehicle velocity,
by in-flight observation of the rotor tip path plane as compared with distribution according to the element's instantaneous
to vehicle responses. geometry.

THE SIMULATION Recent advances in the computational capacity of the simulation

A piloted simulation was performed on NASA Ames Research facility were significant in the decision to develop a model that
Center's Vertical Motion Simulator (VMS) in 1994 using the distributes turbulence over the rotor disc. This was anticipated in
GENHEL UH60A Black Hawk mathematical model of Reference 5:
Reference 1 as a platform to investigate the SORBET model.
This was a simulation technology experiment designed for the Progressive gust penetration by the rotor is an important
purpose of evaluating the influence of stochastic turbulence feature which should be included, where possible, as it is
created at the blade-element stations of a rotor system. likely to strongly influence the simulation results. Conse-
quently, the indi vidual blade mode with the gust field
This project was conducted primarily to determine if a repre- model should be used when high computation time is
sentative workload could be accommodated by our computer permissible.
system in real time, and to ascertain whether the cyclostationary
(periodically stationary) effects of Reference 2 were significant The entire stochastic rotor-element turbulence model is
in the motion environment of the VMS. Favorable pilot opinion described here, although for the initial proof-of-concept simu-
was somewhat surprising because independent longitudinal and lation only the verti cal component of turbulence was used.
lateral inputs were not used during the piloted runs. Further- Examples from this model are presented for the helicopter in
more, angular components of turbulence were not explicitly low-speed flight (10 knots).

*SYRE, SYSCON Corporation, 8110 Gatehouse Rd., Falls Church,

VA 22042.

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?",
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
11-2

In this study the Dryden spectral form was selected due to its \<n<N (blades)
computational efficiency, although a discrete model of a curve-
\<m<M (segments)
fit approximation to the von Karman form** does not offer any
particular technical challenge. As will be discussed, the rotor-to-
body attenuation is so severe that the original gust spectra appear Using the equal-annuli algorithm of Reference 1, the radii to the
to be of secondary consideration. From Reference 4, "When no blade stations may be computed from
comparable structural analysis is performed or when it is not
feasible to use the von Karman form, use of the Dryden form
will be permissible." The difference between the Dryden and \e+ef +
M
Y-{e+ef\
von Karman form is basically a small variation in the high
frequency content (Ref. 6).
where the rotor radius is R = 26.83 ft, the hinge offset is e =
1.25 ft, and the spar length is e' =2.25 ft.
SORBET Model Assumptions
Conventional MIL SPEC angular turbulence filters are a
substitute for the finite-element distribution of translational The azimuth angle of the nth blade with respect to the vehi cle's
turbulence in the air mass over the airframe. In the current study aft centerline is given by
only the rotor disc is considered as an airframe, and the transla-
tional turbulence in the air mass is distributed over the rotor disc
Wn - \Qdt + 2J—
to each blade element. This distribution produces rotor moments
as well as forces. Hence, only the MIL SPEC translational turbu-
lence filters were used in the SORBET model. where £2 = 21 rad/sec is the nomi nal rotor RPM.

From Taylor's hypothesis, the turbulent flow is stationary and In the SORBET model, rather than defining the translational
homogeneous in the longitudinal direction. Turbulence filter disturbance veloci ties at the vehicle center of gravity, they are
outputs are statistically valid at either fixed spatial locations or defined at two onset points, located along an onset line that is
at points translating with the vehicle (as is the case in the perpendicular to the horizontal mean wind vector v# shown in
conventional body-fixed formulation). Two points of applica- Figure 1. Using Gaussian inputs, three independent turbulence
bility are selected along an onset line that is perpendicular to velocity histories are computed at each of these two points, in
the vehicle flightpath, and tangential to the leading edge of the real time. These histories are updated each cycle time, and the
rotor disc. pertinent values from these histories are distributed to the indi-
vidual segment locations by temporal and geometrical
The outputs of turbulence filters are applicable at points on the algorithms.
onset line at the discrete time points. Interpolation perpendicular
to the onset line requires a transport delay that is proportional to
the distance along the flightpath to the element location, and
inversely proportional to the vehicle's aerodynamic velocity.

Simulating transport delays requires that the turbulence filter

outputs be tabled, and updated each computer cycle. A practical
limit on the table size produces a corresponding minimum
velocity of propagation through the rotor disc in simulation.

The MIL SPEC translational filters contain a pole proportional

to the aerodynamic velocity. This velocity is limited to the value
determined from the selected table size.

The turbulence is assumed uncorrelated at each side of the rotor

disc, and homogeneous in the lateral dimension. The statistical
properties of the turbulence are then invariant over the entire
rotor disc by using what we call Gaussian interpolation between
values applicable on each side of the rotor disc.
*„= Vn+ß V"
Geometrical Considerations
The Black Hawk rotor system has four blades (JV = 4). For each Figure 1. Rotor geometry.
of the blades the number of segments used in the simulation was
five (M = 5). The blade and segment indices are defined
Where Uf, is the vehicle's longitudinal aerodynamic velocity and
vb is the lateral aerodynamic velocity, the in-plane aerodynamic
velocity may be defined (approximately) by

V
**The earliest edition of such a model that we have discovered is in an
H =A/«F^
unpublished Boeing document (YC-14-FC-111R) by James H. Vincent
in December 1973.
 11-3

Using this velocity and the sideslip angle ß, the perpendicular velocities than the minimum, such that stationary outputs are
onset line is as shown in Figure 1. The aerodynamic azimuth available for table size increases caused by decreases in velocity.
angle (wind axis) for the nth blade is then The hypothetical case where the entire rotor disc receives the
same table value could only occur for velocities in the thousands
Vn=Vn+ß of knots.

where y/n is the «th blade's geometrical azimuth angle from aft. Lateral Distribution
Having determined the longitudinal point of application (or
This geometry is shown below to be sufficient to establish both index km „) for a given rotor element, the turbulence velocities
the longitudinal and lateral distributions. from the tables are identified as applicable at both the left and
right sides of the rotor disc. Lateral interpolation must be used to
Longitudinal Distribution determine the element's final turbulence velocities. Although the
The left and right Dryden filters receive Gaussian inputs, so the lateral proportion (from the left-hand side) to an element may
outputs of these linear filters are also Gaussian. Furthermore, a then be determined from the geometry of Figure 1
transport delay in these outputs does not alter the statistical
characteristics. The outputs of the filters are stored in tables, _1 rm>nSinyn
Pm,n - 2 + 2R
which are updated each cycle time. The value applicable for a
given element is determined by selecting a value from the tables.
Given that the table size is established (KM = 500) for storing this proportion cannot be applied in a linear fashion. Using the
time histories of the outputs of the filters, the minimum aero- vertical dimension, for example, although both AwL(kmn) and
dynamic velocity v^ that may be accommodated is determined AwflC^m,«) are Gaussian random variables with zero mean and
by distributing the table over the rotor diameter. RMS value given by <JW, in order to preserve the statistical
properties over the rotor disc we use Gaussian interpolation, so
2R that the combination retains the same variance (and all other
KuAt
^M
moments) as the original Gaussian variables.

If a smaller minimum velocity is required, a larger table size A variable x is normally distributed with zero mean and variance
must be selected. For the SORBET simulation cycle time Ar = o2 when its density function is
0.012 sec, the minimum velocity for KM = 500 is 8.944 ft/sec
(5.3 knots). The rotor-plane aerodynamic velocity used in the
turbulence filters is restricted to this minimum. /(*) = CTA/27T e-m
"min v
H<vmh The variance is defined by
V V v
H H £ min

= x2f(x)dx = a2
Independent of vehicle velocity, exactly KM At = 6 sec of data are j-
stored in the tables. Time histories of the outputs of six tur-
bulence filters are contained in six separate tables, each of for a single random variable. When we combine two indepen-
length KM- At the minimum velocity the individual cells corre- dent random variables at some interior point, we must consider
spond to a length of 0.10732 ft (such that 500 of them span the
their proportion p.
rotor diameter), whereas at a velocity of 100 ft/sec (59.2 knots),
for example, the cells are each 1.2 ft long such that only 45 of When a proportion/? has the range (0 < p < 1), as it does in pmn
them are needed to span the rotor diameter. above, a linear combination of two normally distributed vari-
ables may be forced to have the same statistical properties as the
In order to select the velocities to be used from the tables, for originals by normalizing the density functions as follows:
any blade element an integer index is computed to determine the
element's perpendicular distance from the onset line. Defining
/(*l) */(*)
^P2+(l-/

the index is given by

/te)= , }~p »/(*)
2
A/P +(I-P)

«At As applied to variables with a zero mean value, this is a

consequence of the arithmetic mean theorem given in
where the upper brackets denote the integer ceiling operation. Reference 7, where
The tables extend beyond the rotor disc for higher vehicle
11-4

/
nLv Ss + V»
PH (!-P)*2
Av(s) = -
%

%
pxj+(!-/>) x2 +2p(\-p)x\X2
v
p2+(\-pf 'JtU V3, + " Av
Aw(s) = -
Z2 „2
/%f+(l-/>) 4
E = CT
For conventional aircraft (and our baseline configuration of the
UH60A) the angular transfer functions are developed from
because Eyx\X2~[ = 0 for independent variables, and partial differential equations relating these translational distur-
bances to the immersion rate of the vehicle into the turbulent
E\XT = £\x\ \ = a . The variance of the combined dis tribu t
field. These equations use a parameter b, which is the wingspan
then remains a2. Gaussian interpolation is therefore given by length of the aircraft.

Pm,nAwR {km,n) + (l ~ Pm,n )AwL {km,n )

Aw„ oJfh.pus;
Pm,n +\}~Pm,n) Ap(s) =
Li}\s +
Ab
This combination is applicable for all three translational velocity
components at an element location.
iAw(s)
The Dryden Model Aq(s)-- Ab
From Reference 4 the low-altitude verti cal turbulence scale s+-
Ab
length is given as a piecewise continuous function of altitude h

[10 A < 10 ft —sAv(s)

y
Ar(s): 3b '
Lw=\h 10<h< 1000 ft
[lOOO h > 1000 ft 3b

and the in-plane components are given by the functions These angular disturbances are not used in the SORBET model
because the geometry is considered in the points of application
of the translational turbulence excitations. The asymmetric
75.64 h < 10 ft
12
turbulence velocities produce moments, which in turn produce
'-m ~ *-v — s hfuv 10<A<1000ft angular activity.
1000 ä> 1000 ft
All of the baseline case transfer functions have been presented
where because their spectra are displayed on certain graphs herein for
reference purposes. Also, this baseline model is actually avail-
/uv= 0.177+0.000823A able as an option in our UH60A simulation model, and was used
by pilots for comparison purposes.
The horizontal turbulence RMS intensi ties are a function of the
selected vertical turbulence intensity aw Discrete Implementation
Continuous transfer functions involving random inputs are typi -
-0.4 cally converted to discrete form using the zero-order hold
:
0"v = °wfu formulation, where the input is assumed constant over each
computer cycle At. The z-transforms of Laplace functions fj(s)
The results shown in this paper are for a„ = 5 ft/sec, h = 200 ft, then become
and vH = 16.9 ft/sec (10 knots).

Using Taylor's frozen field hypothesis from Reference 4, the Ft(z) = Z

ve-^A m
Dryden form of the linear turbulence trans fer functions are
given by
which may then be converted to difference equations for discrete
a, l2v«v. xL„ implementation. This process is nicely explained in Reference 8.
Au(s) = -
Using this technique on the Dryden transfer functions produces
the coefficients
 11-5

Yu = vuvAt/Lu created at the two onset points and then distributed and super-
imposed at the various blade-element locations. They produce
forces and moments because they modify the angle of attack of
each element.

/2=<r»(l-/l)V2/^ Element Velocities

Forces developed at individual blade-element locations are
Yv = VuvAt/Lv invari ably summed to create total rotor forces and moments. In
this section the spectral filtering consequences of the summation
operation are illustrated by using the turbulence velocities them-
gl=2e~yv selves. Then, in the next few sections, this is further illustrated
using the blade flapping spectra and the total vehicle responses.
Although a blade variable may exhibit much harmonic content,
the summation over the blade index effectively removes many of
the harmonics (Ref. 9).

Figure 2 displays typical time histories of vertical turbulence

gA=- [ove-7v /Vr7][l-^rv +(V3-l)yv] produced at the two onset points (AwL A.wR). These histories are
the uncorrelated outputs of the Dryden vertical filters, excited by
Gaussian noise with 0^, = 5 ft/sec.
Yw = VHV^/^W

15 —r
h\ = 2e_r>v

A2 = - e"2rw

h =[cw/Jy^][l-e-Y» +(43-\)e-Y»yw}

h4 =-[<ywe~7w/^][l-e~rw +(V3-l)yvv]

These coefficients are used in the following six difference equa- i 1 i

tions, each driven by an independent Gaussian noise source 77,-, 0 10 20 30 40
with a zero mean value and a unity standard deviation. Time - sec
AuL{k) = f\AuL{k-\) + f2ri\(k) Figure 2. Onset point velocities.
A
uR{k) = fxAuR(k-\) + f2ri2(k)
The velocities of Figure 2 are the ori gins of the turbulence
velocities for each blade element. In response to these inputs,
AvL{k) = gxAvL{k-l) + g2AvL{k-2)
Figure 3 is an example of the velocity of just one element.
+ 83n-s(k) + gAT)il(k-\) (outboard, m = 5) of one blade.

AvR{k) = gxAvR{k-\) + g2AvR{k-2) 15-

+ g3rl4(k) + g4rl4(k-l) 10-

5-
AwL(k) = hlAwL(k-l) + h2AwL(k-2)
< 0-
+ h3ri5(k) + h4ri5{k-\)

AwR(k) = hlAwR(k-l) + h2AwR(k-2) £ -10-

LU
+ h3ii6{k) + h4ri6{k-\) 15-^ 1 1 1 r
10 20 30 40
These computationally efficient equations produce stable
outputs for all stable inputs. Time - sec

For the body-fixed formulation these velocities are super- Figure 3. Element velocity.
imposed at the center of gravity. In SORBET these velocities are
11-6

Because the velocities at an element are rotationally sampled, The power spectral density (PSD) or autospectrum of this signal
Figure 3 displays considerable one-per-rev frequency content, is shown in Figure 6.
bounded approximately by the input curves shown in Figure 2.

Each and every element experiences vertical velocities similar to N o .

that shown in Figure 3. However, the summation of all of the
I 10 —"
—.
20 elements for each time point selectively eliminates most of O) "
Ü2 -5 _
the high frequency content. Figure 4 displays this phenomenon. ? 10 -
What remains after summation is essentially the average of the Q -
original onset velocities. 9? -io —_
Q-10 "
u>
c ■

9- -15 _
Sio —
LL -

10 10 10" 10
5 Frequency - Hz
<
Figure 6. Flapping PSD without turbulence.

LU The significant frequencies at the blade level of examination are

the blade harmonics, beginning with the fundamental, at 4.297
Hz (27 rad/sec). These harmonics theoretically extend to infinity.
Frequencies on this graph that are observed below the funda-
mental are, for the most part, aliased harmonics (originating be-
yond the Nyquist frequency).
Figure 4. Summed velocities.
When the vertical turbulence (aw = 5 ft/sec) is introduced into
the rotor system under the same conditions, the time history of a
This phenomenon occurs independent of turbulence. In the next blade displays more frequency content, as indicated in Figure 7.
three sections this is further illustrated using both rotor and
vehicle variables.

Rotating Frame Blade Spectra

The flight condition documented here is for a trimmed vehicle in
10-knot hori zontal flight at an altitude of 200 ft. Under these
conditions, when turbulence is not present, any given blade
produces the flapping (ßn of Ref. 1) time history as shown in
Figure 5.

Figure 7. Flapping with turbulence.

The background spectrum of Figure 6 is then elevated by a few

orders of magnitude, as shown in Figure 8.
I ■ " ■ I " ■ ■ I ■ ■ ■ ' I
12 3 4
Time - sec

Figure 5. Flapping without turbulence.

 11-7

The PSDs of these Fourier coeffi cients are given in

N o : • Figures 9-11. The lower lines show the behavior without
110- 11
—■ •""^Sk L turbulence, and the upper lines show the behavior with vertical
' I* Ji rotor turbulence. In comparing these figures with Figures 6 and
m
-5 - ■■!■ ■^vrvi. ■ ■Ji■t.
"? 10 - 8 it is seen that only blade harmonics that are a multiple of the
Q
CO in -
::::::::r:::ff:^k:: : *Wm
number of blades survive the summation operation, whereas
the stochastic contributions are retained in the spectra of the
0-10" - rotor disc.
c
; : i ■■ • ■■
&io - : : i
LL
1 1 1 1 1
10 10 10 10 10
Frequency - Hz

Figure 8. Flapping PSD with turbulence.

The significant periodic components in these spectra are created

by the blades essentially tracking each other, independent of
turbulence. Turbulence elevates the background spectrum, but
since the inputs are not uniform over the rotor disc it does not
have a significant influence on the rotor harmonics. Notice that
in Figure 8, except for the first harmonic, the blade harmonics Frequency - Hz
almost disappear into the turbulent background, and they are not
amplified by the introduction of turbulence. This leads to the Figure 9. AQFspectra.
conclusion that at the blade level of consideration the intro-
duction of turbulence does not produce significant cyclo-
stationary phenomena.

The lagging spectrum is not shown here. At this flight condition

it is about an order of magnitude smaller than the flapping
spectrum.

Fourier Coefficients
Forces and moments generated by blade activity are transmitted
to the airframe through summation functions. Given that each
simulated blade has identical physical properties, such as mass,
length, and relationship to adjacent blades, the consequence of
summation is to eliminate all integer harmonic multiples of the
RPM that are not a multiple of the number of blades (Ref. 9). 0 1
This is illustrated by considering the principal flapping Fourier 10 10
coefficients. Frequency - Hz

The principal flapping Fourier coefficients (main distortions of Figure 10. A, F spectra.
the rotor plane, in nonrotating frame) are given by the following
summations of the individual flapping angles (ßn of Ref. 1).

%F:
£> n=\

n=\

N
BlF = -—2_jßn^nVn
n=\ 10 10 10
Frequency - Hz
where A 0F is the coning (steady flap ping) angle, A j F is the longi-
tudinal first harmonic, and B\ F is the lateral first harmonic. Figure 11. B-\ F spectra.
11-8

Harmonics of N/rev seen in these figures are not caused by

turbulence (see Ref. 9). They are a normal consequence of blade
summation of periodic phenomena. Of course, interior har-
monics will be observed if the blades are not tracking each other
in their rotational environment. This occurs during transients,
and is also a consequence of mismatched blades:

The one-per-revolution (27 rad/sec) frequency content that

can be seen in the flight-test data is caused by a mis-
matched rotor blade that was not tracking with the other
three on the test aircraft; therefore, this frequency is not
observed in the simulation results. (Ref. 10)

Baseline Traces 10 10 10' 10 10

The baseline spectra are the responses for the conventional im- Frequency - Hz
plementation of turbulence, typically used for fixed-wing aircraft
but also available in our UH60A model. In the baseline configu- Figure 13. Aub, vertical input.
tation the inputs are at the vehicle center of gravity. The transfer
functions for this formulation were presented in the section
named The Dryden Model. The baseline spectra also describe
the outputs of SORBET's right and left filters.

The baseline configuration has been available for the UH60A

model (and other helicopters) for a number of years, although
pilots have invariably been critical of its performance during
low-speed flight. This baseline configuration is nonetheless
interesting for comparison purposes. In Figure 12 the vertical
velocity baseline spectra are shown, with parameter a = v/Lw.

For the flight condition of 10 knots at an altitude of 200 ft, the

parameter is about 0.04.
10 10 10 10
Frequency - Hz
N
X Figure 14. Avb, vertical input.

ü
0)
w

Q
C/D
O-
n
<

-2 -1
10 10 10 10 10 <D -10

Frequency - Hz
>io S
10 10 10 10 10
Figure 12. Baseline spectra. Frequency - Hz

Figure 15. Awb, vertical input.

At the 10-knot flight condition the baseline PSDs are presented
in Figures 13-18 as dotted lines. In the configuration studied
here only the vertical axis of SORBET had an input. However, It is useful to think of the body-fixed spectra (dotted lines) as
all axes (solid lines) display outputs. representative of the velocity of the air mass, as experienced by a
point traversing through it. If this air mass velocity is rigidly
attached to the vehicle, as it is in the conventional formulation,
 11-9

then the vehicle itself experiences this velocity. If, however, it is

attached to the velocity of a blade element, then it produces an
alteration in that element's angle of attack, which in turn con-
tributes to the production of rotor-system forces and moments.

The body-fixed baseline spectra could only be the same as the

helicopter spectra if unity transfer functions existed from the
rotor to the body. Indeed, the existence of nonunity transfer
functions in this regard is obvious because rotor craft responses
differ from those of conventional aircraft. Nonetheless, these
spectra are useful in determining just what the rotor-to-body
effects are. Only "for high speed flight, where helicopters
behave more like conventional aircraft" t could these dotted
baseline curves possibly represent desirable responses. 10 10 10 10
Frequency - Hz
For completeness the angular PSDs are also shown in
Figures 16-18. All angular disturbances observed in SORBET Figure 18. Arb, vertical input.
were induced responses from the application of vertical
turbulence. The baseline curves on these plots were obtained
from the conventional transfer functions, using an arbitrary In Figures 13-18 only the vertical component of turbulence into
value of wingspan length b of 20 ft. the rotor system was used, as was the case during pilot evalua-
tions. Induced effects are seen to occur on all other axes. Pilots
generally concurred that this system felt quite realistic, with the
exception that some airspeed and lateral variations were lacking.
This is illustrated in Figures 13 and 14.

Rotor-Airframe Transfer Function

In Figure 15 the vertical vehicle response to vertical rotor turbu-
lence was shown. Since the dotted line represents the spectral
input throughout the rotor disc, clearly much energy is lost in
the rotor-to-body transfer function. This transfer function was
determined by emulating a single-input/single-output (SISO)
system using the same statistical value for each element at each
time point.

10 10" 10 The baseline vertical Dryden PSD as shown in Figure 15 is

Frequency - Hz plotted in Figure 19 as a dashed line. Using identical random
inputs at each element, the SISO PSD curve of Figure 19 was
Figure 16. Apö, vertical input. created. Comparing the SISO curve to the input spectrum of the
air mass (dashed line), there is considerable attenuation. Also
shown in Figure 19 is the multiple-input/single-output (MISO)
PSD curve, as replicated from Figure 15. More attenuation
occurs on this curve, when the inputs are not uniform (MISO)
throughout the rotor disc.

SISO Transfer Function

10 10" 10
Frequency - Hz

Figure 17. Aqb, vertical input.

10 10" 10
Frequency - Hz
'''Remark by Norman D. Ham of Massachusetts Institute of Technology
during his class on Aerodynamics, Stability, and Control of Rotorcraft
and Other VTOL Aircraft, Ames Research Center, 1978. Figure 19. Awb, same vertical inputs.
 11-10

The rotor-to-body transfer function (nondimensional) is also Pilots generally agreed that the turbulence magnitude should be
shown in Figure 19, as computed from the SISO curve with a function of velocity. Above 40 knots, the subjective values for
respect to the baseline curve. The attenuation becomes more light and moderate turbulence were selected as <\, = 5 ft/sec and
pronounced beyond about 4.3 Hz, which happens to be the rotor 8 ft/sec, respectively. Near hover, pilots selected standard devia-
RPM. This phenomenon is a consequence of the blade summa- tions that were half these values. The reason for this is unknown,
tion process operating on signals that are identi cal for each and should be investigated further. Although the frequency
element. content is acceptable, the magnitude must be decreased in low-
speed flight. For the body-fixed formulation in low-speed flight,
When turbulence was permitted to vary over the rotor disc pilots invariably criticized the baseline configuration. A typical
according to the algorithms, the slightly different results of comment was "It was like a washboard."
Figure 15 were produced, shown in Figure 19 as the MISO
curve. The transfer function for the MISO system has not been Pilots are clearly sensitive to the more nonlinear aspects of
identified, although it must be similar to the SISO system given atmospheric turbulence. From Reference 12, "turbulent flows are
in Figure 19 because of the vehicle response similarities shown diffusive and intermittent" and, moreover, "the element of
by the SISO and MISO curves. surprise that is characteristic of turbulence must be present."
Pilots were quick to point out that although the stochastic turbu-
PILOT OPINION lence was realistic, it was too continuous and lacked occasional
From Figure 19, the rotor system clearly attenuates high- sharp gusts. Their comments reflected Reference 13 quite
frequency nonuniform inflow (as in blade-element turbulence) to closely:
a greater extent than it attenuates uniform inflow. Nonetheless,
lower frequency distortion of the rotor disc caused by turbulence One of the most common grievances helicopter pilots have
may be an important piloting cue. Distortions are an indication with simulated atmospheric turbulence is that it is too
of activity in the stability augmentation system (SAS), and may regular, and does not contain 'patches' of turbulence
influence speed selection. Some of the simulation pilots stated followed by periods of calm air.
that in real helicopter flight they sometimes "observed the rotor
plane undergoing large responses to turbulence, while the air- Due to the large number of pilot remarks in this area, we ac-
frame's motion was relatively smooth." In response to pilot commodated them on certain points. During the simulation, a
requests for this vi sual cue, the tip-path plane will be included in few additions to the SORBET model were made, as described
the visual display for the next rotorcraft simulation on the VMS. below. The pi lots agreed that these additions provided even
more realism.
To assure that simulation pilots had similar conditions in which
to assess the model, they were given identical tasks of following Occasional vertical gusts were simulated by changing the
recorded flightpaths. Our chief test pilot for the simulation magnitude of the mean vertical wind at random times. Gusts
established these basic paths (without turbulence) by flying a set were programmed to arrive more frequently and more abruptly
of runs over the terrain, and the spatial positions and orientations at high speeds (about every 3 sec) than at low speeds (every
of the helicopter were recorded. Using a sophisticated playback 12 sec). The resulting random vertical gusts were praised by the
feature of our heads-up display, other pilots were shown these pilots as extremely characteristic of real turbu lence.
paths by using both a graphical lead aircraft and a visually
superimposed stick-figure canyon (Ref. 11). This is depicted in Moments of calm air were introduced by scaling the magnitude
Figure 20, where both the lead aircraft and the velocity vector of the stochastic turbulence to zero for random time periods
(circle) are indicated. (between 2 and 6 sec). This was followed by smoothly applying
random percentages of the full magnitude, also for random time
periods. This technique created the effect of patches of different
levels of turbulence.

With the addition of vertical gusts and patches of turbulence, the

vertical turbulence was accepted by all the pilots as very realistic.
One deficiency they found was in the reduced level of lateral,
longitudinal, and yaw turbulence cues. Since the model did not
directly influence these axes, these comments were expected. As
a partial solution, a stochastic input, perpendicular to the tail
rotor, was added. While the pilots noticed the extra yaw varia-
tions, they generally agreed it was not a great improvement, and
Figure 20. Playback canyon. the lack of lateral inputs detracted from realism.

One major pilot concern involved the way the turbulence was
Pilots had the task of following the lead aircraft (at speed) and not correlated to the terrain. This was not within the scope of
staying within the canyon. Without turbulence, this task was SORBET, but work is being done in this area (Ref. 14).
quite simple for all pilots. The task became more difficult when
turbulence was added. In summary, pilot comments were quite favorable for SORBET
during the 2-week motion-based simulation study. Representa-
All pilots found the SORBET model realistic, and their com- tive comments are listed below, where the italics are the
ments were used to improve the model during the simulation. authors'.
 11-11

1. The chief NASA test pilot for this simulation, Munro In-plane turbulence applied at the rotor elements produces
Dearing, commented during low-speed flight that "this minimal vehicle responses. This is therefore an unsatisfactory
would be typical of light-to-moderate gusts in hover." technique for modeling stochastic turbulence in the horizontal
However, at 100 knots he noticed "there is a lack of heave." plane, and would probably produce erroneous results using gust
These comments were made prior to the introduction of the models. Hence, in-plane disturbances should probably be
embellishments described above. handled more or less as in the conventional formulation, with
the inclusion of the correlated influence on the tail rotor.
2 NASA test pilot Bill Hindson commented on SORBET
once the occasional gusts were included that "the low The vertical component of turbulence as input at the blade radial
frequency pitch and roll felt is also characteristic of turbu- stations produces significant differences in vehicle performance
lence," and "it is a fairly realistic simulation." from that of conventional body-fixed techniques. The differ-
ences include considerable attenuation of higher frequencies in
3. NASA test pilot Tom Reynolds commented on SORBET the vertical and all rotational axes. These phenomena elicit
when the occasional quiet periods were introduced that favorable pilot opinion, and probably identify the major differ-
"this is much like what I expect is real," and "overall, this ence between helicopter and conventional aircraft responses.
feels pretty good, fairly realistic light-to-moderate
turbulence." In establishing a uniform level of turbulence over a flight regime
our test pilots preferred a standard deviation that varied linearly
4. NASA test pilot Gordon Hardy commented on the with velocity, roughly doubling from hover to 40 knots, where it
SORBET simulation that "these are pretty typical became constant. This is an unknown phenomenon that deserves
excursions compared to real flight," but also said "I'd further investigation. Also, criteria for implementing patches of
expect more lateral motion." different levels of turbulence should be established, because
they correspond to pilot experience.
5. Retired NASA test pilot Ron Gerdes commented "there is
nothing that I would change in the model" REFERENCES
1. Howlett, J. J., "UH-60A Black Hawk Engineering
6. Navy pilot Tim Sestak commented "this is what turbulence Simulation Program: Vol. I—Mathematical Model,"
feels like." NASA CR-166309, Dec. 1981.

IN-PLANE TURBULENCE 2. Prasad, J. V. R., Riaz, J., Gaonkar, G H., and Yingyi, D.,
Although piloted runs were not made with the in-plane "Real Time Implementation Aspects of a Rotorcraft
longitudinal and lateral turbulence options enabled, it has been Turbulence Simulation Method," 49th Annual Forum of
established that their inclusion would have added very little to the American Helicopter Society, St. Louis, Mo., May
the simulation. From Reference 15, "Current articulated and 1993, p. 459.
semi-rigid rotors are insensitive to in-plane gusts." Our own data
support this statement. 3. Robinson, John E., Weber, Timothy L., and Miller,
David G, "Real-Time Simulation of Full-Field
In fact, adding these in-plane components to SORBET produced Atmospheric Turbulence for a Piloted Rotorcraft
changes that were too trivial to show here. For this reason we Simulator," 50th Annual Forum of the American
postulate that in-plane translational turbulence components Helicopter Society, Washington, D.C., May 1994.
should probably act at the vehicle center of gravity, just as in
the conventional formulation. Of course, using these in-plane 4. United States Department of Defense, "Flying Qualities
components also means that consideration should be given to of Piloted Aircraft," Military Specification MIL-F-
correlated tail rotor responses. The tail rotor becomes much 8785C, Nov. 1980.
more significant when gusts are considered, rather than just
stationary turbulence alone. 5. Dahl, H. J. and Faulkner, A. J., "Helicopter Simulation
in Atmospheric Turbulence," Vertica, Vol. 3, 1979,
In our next investigation we plan to include in-plane components pp. 65-78.
for turbulence that will probably be applied at the vehicle center
of gravity. This proposed system has also been investigated; it 6. Gerlach, O. H., van de Moesdijk, G., and van der Vaart,
produces almost identical spectral responses as those shown in J. C, "Progress in the Mathematical Modeling of Flight
Figures 13-18, except that the in-plane vehicle responses of in Turbulence," AGARD Conference Proceedings
Figures 13 and 14 then become superimposed with the dashed No. 140 on Flight in Turbulence, Paper No. 5, 1973,
lines (as in the conventional aircraft case). This model should pp. 1-38.
then include the in-plane disturbances that pilots expect.
7. Sokolnikoff, I. S. and Redheffer, R. M., "Mathematics of
CONCLUSIONS Physics and Modern Engineering," McGraw-Hill Book
Adding turbulence to the Black Hawk simulation added less than Co., Inc., New York, 1958, pp. 667-668.
10 percent to the required cycle time. Hence, blade-element
implementations of turbulence are now feasible using modern 8. Smith, J. M., "Recent Developments in Numerical
computers. Integration," Journal of Dynamic Systems,
Measurement, and Control, Transactions of the ASME,
No. 73-WA/Aut-6, Detroit, Michigan, March 1974.
11-12

9. McFarland, R. E., "The N/Rev Phenomenon in 13. Costello, M. F., "A Theory for the Analysis of Rotorcraft
Simulating a Blade-Element Rotor System," NASA Operating in Atmospheric Turbulence," Proceedings of
TM-84344, March 1983. the 46th Annual National Forum of the American
Helicopter Society, Washington, D.C., May 1990,
10. Ballin, Mark G. and Dalang-Secretan, Marie-Alix, pp. 1003-1015.
"Validation of the Dynamic Response of a Blade-
Element UH-60 Simulation Model in Hovering Flight," 14. Clement, Warren F. and Gorder, Peter J., 'Terrain-
Journal of the American Helicopter Society, Washington Correlated Atmospheric Turbulence Modeling for Real-
D.C.,Mayl990,p. 84. Time Simulation of Nap-of-the-Earth (NOE) Flight
Operations, USAAVSCOM TR 92-A-011, May 1992.
11. Zelenka, Richard E., Clark, Raymond F., and Branigan,
Robert G., "Flight Test of a Low-Altitude Helicopter 15. Judd, M. and Newman, S. J., "An Analysis of Helicopter
Guidance System with Obstacle Avoidance Capability," Rotor Response Due to Gust and Turbulence," Vertica,
AGARD Avionics Panel Symposium on "Low-Level Vol. 1,1977, pp. 179-188.
and Nap-of-the-Earth Night Operations," Rome, Italy,
Oct. 24-27,1994.

12. Panofsky, Hans A. and Dutton, John A., "Atmospheric

Turbulence," John Wiley & Sons, New York, 1984,
pp. 5, 278.
 12-1

C-160 GROUND HANDLING MODEL UPDATE

USING TAXI TEST DATA

D. Fischenberg
W. Mönnich

Institut für Flugmechanik

Deutsche Forschungsanstalt für Luft- und Raumfahrt (DLR)
Lilienthalplatz 7, D-38108 Braunschweig
Germany

SUMMARY Abbreviations:

The need for representative ground handling mod- A/C aircraft

elling for flight simulators according to modern ap- CG center of gravity
proval standards is generally strong. A main prob- FAA Federal Aviation Administration
lem is to find an adequate level of model accuracy U/C undercarriage
for the complex physics of ground related maneu- 6-DOF six degree-of-freedom
vers. In this paper some examples of the develop-
ment process of a C-160 ground handling model are
presented. The original analytically derived model 1. INTRODUCTION
is updated and validated with taxi test data using
system identification methods. Not only the iden-
Under contract, DLR Institute of Flight Mechan-
tification of model parameters is discussed, but it
ics participated in the development of a new flight
is also demonstrated how to derive equivalent sub-
simulator for the military transport aircraft C-160.
models in order to reduce the complexity of the data
This simulator is being built by the Simulator Divi-
bases without hardly any loss of simulation fidelity.
sion of Thomson-CSF according to the FAA simu-
lator approval standard Level D [1]. DLR's main
tasks were to instrument a test aircraft, to con-
NOMENCLATURE
duct flight tests, to prepare validation and proof-of-
ax body-fixed x-acceleration match data, and, last not least, to develop a suitable
V nosewheel steering angle math model for the aircraft dynamics. An extensive
Fs oleo damper spring force flight and ground test program was carried out in
Fd total oleo damper force order to derive a mathematical model from flight
Fb brake pressure scale factor test data. Using system identification methods, a
■fx,acc equivalent tire acceleration force completely new aerodynamic derivative model was
Fx tire longitudinal force estimated [2]. This is nowadays a common tech-
Fy tire side force nique for determining a high quality simulator data
Fz tire normal force base [3-5].
Joleo oleo compression Besides the aerodynamic model, simulator model fi-
I impulse delity essentially depends on a second math model
A'i, K2 damping coefficients part, namely the ground handling model with
m aircraft mass undercarriage kinematics and tire characteristics.
P friction coefficient This model is always necessary for operations with
fir roll friction coefficient ground contact: takeoff, landing, and taxiing, which
maximum brake friction coefficient
/^max are extremely important maneuvers for pilot train-
Pb brake pressure ing. Nowadays, the quality of ground handling mod-
r aircraft's yaw rate els is considered as not sufficient, as the validation
T skid angle of simulator models for ground contact within the
t time stringent FAA simulator qualification tolerances al-
UK,VK body-fixed components of flight path
ways causes problems [6-7]. The need for more rep-
velocity resentative, but computer implementable undercar-
X state vector
riage (U/C) models is strong [8].
%nw 1 ^mw i body-fixed x-, y-distances of a tire
Vmw contact point from CG A first problem of ground handling modelling is the

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?'
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
12-2

decision on how detailed the model structure has to A/C states

be. The aircrew is only aware of the loads induced

by the undercarriage, not e.g. of the oilflow within
one shock absorber. This crew awareness should be
Ground
the main criterion within the model development control inputs: forces
process [9]. Fidelity problems are mainly based on Handling moments
nosewheel
the availability of data bases: (a) spring and damper steering angle, Model
brake pressure
characteristics of suspension and tires and (b) char- left and right
acteristics of the tire-runway contact (tire models).
Generally, tire models are based on measurements, I U/C data base: "1
but these are rarely available for aircraft tires [10]. | - suspension geometry
Often data bases and models are taken from the au- | - runway condition
tomobile industry, cf. [11], which can be adapted to | - tire characteristic
| - friction coefficients
aircraft characteristics [12]. | - spring and damper
I characteristics
This paper presents some results of an investigation
aimed at improving the simulation quality of a C-
160 ground handling model. Using the measured
Fig. 1. Ground handling model black box
taxi test data of maneuvers gathered for simulator
structure
qualification, it is tried to update essential parts
of the purely analytically derived model with sys-
tem identification methods in order to meet aircraft
specific characteristics. Generally, it is sufficient are mounted with a considerable inclination angle
for a model update to adjust important parameters to the fuselage reference line, and the oleo dampers
within its structure, however, sometimes the model are compressed by a rotational motion of the lower
structure has to be changed or expanded, too. In struts.
this paper, some examples for the improvement of
different parts of the C-160 ground handling model
will be presented and discussed.

2. BASIC GROUND HANDLING PHYS-

ICS

The ground handling model is one of the three main

submodels in the structure of a complete simula-
tor model (aerodynamics, ground handling, engine).
A simplified structure of a ground handling model
with inputs, outputs, and information used within
the black box 'ground handling model' can be illus-
trated as in Fig. 1. The inputs for a conventional
transport aircraft are the nosewheel steering angle
and the right and left brake pressures. The out- Fig. 2. C-160 aircraft and undercarriage design
puts are the U/C forces and moments acting at and
about the aircraft's center of gravity.
A lot of information has to be provided for that The ground handling model of the C-160 was
black box: (a) the aircraft state vector X_ (rota- formulated analytically and describes the dynam-
tional and translational speeds, Euler angles, and ics of the suspensions for the nose gear and the four
CG altitude) and (b) the U/C data base (suspen- main gears separately. The model includes two sus-
sion geometry and location relative to CG, runway pension devices for each single landing gear caus-
condition, tire characteristics, friction coefficients, ing dynamics between the airframe and the runway
stiffness and damping characteristics of the suspen- (Fig. 3).
sion devices). The black box contains the differen- The oleo-pneumatic damper is mounted with an in-
tial equations of the dynamic systems (e.g. tires), clination angle of about 40° to the fuselage reference
the kinematics of each suspension device, and the line. The kinematic model between the tire axes
application rules for the data base. and the airframe uses the geometric lengths, an-
Fig. 2 illustrates the U/C design of the C-160. It gles of inclination, and damper strokes of the struts.
consists of a two-wheeled, steerable nose gear and The oleo-pneumatic damper force Fd is the sum of
four two-wheeled main gears. The main gear struts damper stiffness and damping force. The damper
 12-3

Hbmax is the maximal brake friction coefficient just

before the tire begins to slip on the runway. This
value depends heavily on the surface consistency
(dry, wet, icy ...). The factor Fb scales the brake
pressure pb, which is set by the pilot, to values be-
TCG
tween 0 and 1.
altitude

A suitable implementation of the roll friction co-

oleopneumatio
efficient is necessary for simulation of maneuvers
damper
with ground contact. This roll friction coefficient
is set according to the runway surface consistency.
The C-160 rolling on dry concrete is modelled with
fj,r = 0.03, whereas for rolling on soft soil \iT = 0.11.
During model validation, the roll friction coefficient
can be adapted to simulate correct ground speed
or ground distance, but it should deviate not more
Fig. 3. Model structure of one C-160 landing gear than 10 % from the handbook values.

For modelling a correct braking force, the brake

pressure scale factor Fb can be adapted using suit-
stiffness F, of the struts is a nonlinear function of
able taxi test data. A collection of test data at
the damper compression and is known from man-
different ground speeds, combined with some dif-
ufacturer's data. The damping is modeled velocity
ferential left and right braking can be evaluated
squared (/ojeo) with different constants for compres-
with system identification methods using the full 6-
sion (Ki) and decompression (K2):
DOF model for the aircraft including engine model
compression and U/C model. Important sensor signals for the
Fd (1) parameter determination are longitudinal acceler-
1 F.-K2ftUo decompression
ation, ground speed, and ground distance. With
these test data it is possible to estimate the brake
The tire is the second suspension device. Two tires scale factors Fj, and Fbr for left and right brake
are mounted on each strut. The nonlinear tire stiff- separately. In theory, both should have the same
ness is modeled using manufacturer's data tables. value, but it has been found that they often differ
A tire damping force is neglected as well as the tire up to 5-6%. The consideration of the difference be-
mass. The geodetic z-motion of the tire axis for tween left and right is important for lateral stability
each strut is thus described as a first order differen- during taxi maneuvers with braking.
tial equation.
Fig. 4 shows a typical identification result for (a)
differential brake operations and (b) rolling with a
3. ROLLING AND BRAKING total stop at the end. Longitudinal acceleration and
ground speed show a proper fit and each brake oper-
Both, the tire side force, responsible for the lateral ation can be seen clearly in both the measured and
directional behavior of the aircraft, and the longi- the simulated data. Additionally, the aircraft pitch
tudinal tire force are dependent on the tire normal angle and the nose wheel oleo compression is plot-
force at each contact point to the runway. This nor- ted. Each brake operation results in a small pitch
mal force is mainly a function of the oleo-pneumatic down rotation, which is verified by a corresponding
damper force, which can be computed as a function nose wheel oleo compression (cf. Fig. 4). Even the
of A/C attitude, velocity, altitude, rotation, and pitch up after the total stop at the end of the second
U/C kinematics and characteristics. time segment can be seen in both the measured and
the simulated data.
Having determined this normal force Fz for each
tire contact point, the longitudinal force Fx can be To verify correct rolling and braking, Fig. 5 shows
computed by multiplying by the friction coefficient two FAA validation tests: (a) ground acceleration
V~ and (b) deceleration time and distance using only
Fx=Fz-n (2) the wheel brakes (no reverse thrust). For (b) the
The friction coefficient fi is the sum of the roll fric- identified brake pressure scale factors are used. The
tion coefficient \iT and the brake friction coefficient, roll friction coefficient is set for rolling on dry con-
that depends on the magnitude of brake operation crete (0.03). Both validation tests show a proper
by the pilot. fit, satisfying the Level D requirements. In (b), the
anti-skid system activation for the right brake can
H = Pr+Fb-pf Hb„ (3) be seen clearly.
12-4

60 80
bright brake
brake pressure 4Q
,left left
pressure left and right,
left and right, right bar
bar 0
60-

longitudinal
acceleration
m/sec2
t -imp ground
speed,
m/sec
30
main gear
touch down
_ ,
nose gear
touch down
1
0
-3'
600
15-
ground
ground distance, 300
speed,
m/sec
(a) time.sec

pitch
angle, ys^p—^Y^>^
deg brake
pressure 40
-1.5 left and right,
.12' bar 0
nose gear
oleo
compress,
JwuJl^
m
.06'
0 time.sec 25 0 time.sec 20

(a) (b)
ground
distance,
Fig. 4. Identification of brake pressure scale factors
with differential braking, simulation model out- i r
20 time.sec
put ( ) compared to taxi test data ( )

Fig. 5. Ground acceleration (a) and deceleration

4. TIRE ACCELERATION AT TOUCH with brakes (b) validation, simulation model
DOWN output (—) compared to taxi test data ( )

Another example of C-160 ground handling model

update is the aircraft derotation after main gear
touch down. The measured longitudinal accelera-
tion shows a considerable peak at main gear touch impulse time t using several measured time histo-
down, which originates from the force needed to ac- ries of the longitudinal acceleration ax. For C-160,
celerate the eight wheels of the main undercarriage the equivalent impulse time was identified to app.
(Fig. 6). In Fig. 6a this tire acceleration force is ne- 0.2 sec, the corresponding force for each of the ten
glected in the simulation model, and a deficiency in wheels to Fx>acc « 6,000iV. A ground speed de-
pitch angle can be seen, which is violating the FA A pendency was neglected. Fig. 6b shows the simu-
Level D requirement for the pitch angle accuracy lation fidelity using this touch down impulse mod-
(±1.5°). eling. The fit of longitudinal acceleration and pitch
The computation of this force modifying the roll angle is considerably improved and is now within
the FAA Level D tolerances. The correct touch
friction coefficient of eq. (3) is hard to realize, as
the runway contact condition changes in a strongly down point simulation of the main undercarriage
can be seen, and even the nosewheel touch down af-
nonlinear way from total slip to full adhesion. A
ter about 6 sec can be detected in both the measured
simplified approach is adopted: the change of the
and the computed acceleration. Furthermore, the
aircraft's impulse at main gear touch down
touch down point is confirmed by the oleo compress
A7 = m ■ ax ■ t = Fx ■t (4) plot of the main front gear. This oleo compress sim-
acc
ulation, which is an important variable within the
can be computed, knowing the aircraft's mass and undercarriage force computation, shows an obvious
estimating the x-force FXi<iCC and the equivalent improvement, too.
 12-5

ing gear precisely. This is important for the correct

longitudinal computation of the skid angle at each tire contact
acceleration, point to the runway. The skid angle is an essential
m/sec2 main input for the tire side force computation. As illus-
gear touch down
-3
10

pitch steering angle T\

angle,
deg oH first i ^
ground contact skid angle T
-5
.15 nosewheel
drift angle
main front gear s

oleo compress,
m 1r\'~~~''/
/ v~-/ nose-
v +rx
K nw
\ 1/
wheel A
.oo-j —A- f^i
5 time, sec
(a) nw

longitudinal
acceleration, Fig. 7. Definition of the tire skid angle
m/sec2

trated in Fig. 7 for the nosewheel, the skid angle

T is the difference between the steering angle TJ and
pitch the drift angle, that depends on the aircraft's yaw
angle, rate r and the location of the gear strut XCG-
deg
VK + r- xnv
T] — atan (5)
uK
main front gear
oleo compress, The consideration of the yaw rate is important for
m the nose gear drift angle (considerable lever arm
from the center of gravity to the gear contact point
to the runway), but also for the main gear struts,
(b) 5 time, sec
although there are only small lever arms in the air-
craft's x- and y-directions. Considering these lever
arms, the skid angle at the main gear struts can be
Fig. 6. Influence of tire acceleration on touchdown formulated:
simulations: (a) neglecting tire acceleration,
(b) modelling tire acceleration; model output vK + r • x„
= —atan (6)
( ), ßight measured ( ) UK ~r-y„

This exact formulation is important for simulation

quality at low speed taxi [12].
5. LATERAL DIRECTIONAL BEHAVIOR After having described the kinematics precisely, the
ground handling model can be updated with the
identification results for the tire side force charac-
At the beginning of the development process for a teristics. Analytically derived 'deformation models'
realistic lateral directional submodel, it is desirable [11] are suitable, computer adapted formulations to
to analyse the complex physics. The division into describe the complicated nonlinear tire behavior in
indispensable submodels (e.g. damper character- contact with the ground. These models consider
istics) and submodels, which are nice to have but not only the skid angle and the tire load, but also
not important for pilot's awareness, can reduce the ground speed and runway and tire parameters. The
complexity of the ground handling model structure tire parameters are often available for automobile
noticeably. and truck tires, but rarely for aircraft tires.
Before trying to update the lateral tire characteris- The taxi test data provide a lot of information about
tics with measured taxi test data, it is essential to the aircraft's behavior at a certain runway condi-
model the geometry and kinematics of each land- tion, usually for dry concrete. Since the kinematic
12-6

variables can be computed from the onboard meas- '\ / "\

urements, the tire parameters can be identified as a nosewheel -A \ ' \
steering o
/
function of the tire skid angle using taxi tests with > \ /
angle,
nosewheel steering angle inputs. It is essential that '
deg
-20
N.
v'
there are strong steering inputs at different ground
12
speeds. Fig. 8 shows the identified side force as a /p- j> '/^
function of the tire skid angle on dry concrete. For yaw •/ ^ V
^i zr
'/ /
rate,
deg/sec
o
-- / >\
v*v y,
'/, /
\ -"/
W
10
-- ■- —- •- -- -. . ^
ground
speed,
1, _, . .- -- — ■
•- -- — -_ _~"^
nVsec
5
-■
*"* "**■*

Fig. 9. Rate of turn versus nosewheel steer-

ing proof-of-match, simulation model output
( ) compared to taxi test data ± FAA Level
D tolerances ( )

skid angle t, deg

6. BACKWARD TAXI

During C-160 taxi operations, it is possible to move

Fig. 8. Identified C-160 tire side force (Fy) charac- the aircraft in backward direction by selecting a neg-
teristics on dry concrete; measured truck tire
ative propeller blade angle. This backward taxi is
characteristics on dry and wet surface (con-
stant normal force Fz)

rudder;
comparison, the side force from the analytically de- nosewheel
rived C-160 model (using truck tire data) is plotted, steering
angle,
too. On dry runway, there is a linear dependency deg
up to 6° skid angle. For greater skid angles, there
is no full adhesion anymore but partly slip and the
side force is depending on the ground speed. Fig.
9 shows a typical rate of turn versus nosewheel an- lateral
acceleration, o
gle proof-of-match plot on dry concrete using the m/secZ
identified side force characteristics. For wet run-
ways, truck tires show similar linear characteristics
for full adhesion (up to 4°) as on dry runways, but
a considerable reduction of the maximal side force.
Since the identified C-160 tire characteristics on dry yaw
concrete are very similar to the measured truck tire
characteristics and since no C-160 test data were
sy? o-inAV^VyV^
gathered for wet runways, the truck tire parame-
ters are taken for taxi modeling on runways under
adverse conditions (wet, icy, snowy). azimuth
angle,
deg \f\
Fig. 10 presents the lateral directional simulation lift off rotation
-116
quality with the complete C-160 simulator model 20 time.sec
for a typical normal takeoff from brake release up
to 200 ft altitude. Directional control is maintained
by the pilot by using the nosewheel steering and
the rudder. These pilot control inputs are used di- Fig. 10 Lateral directional control validation dur-
rectly as inputs for the simulation model, which was ing normal takeoff, simulation model output
updated with the identified tire side force character- ( ) compared to flight test data ( )
istics.
 12-7

not a FAA Level D required validation test, but for 60

nosewheel
C-160 it was found to be desirable to have a valid steering
simulator model for this operation. angle,
deg
For backward taxi, only minor changes in the -60
ground handling model formalism were necessary. .5
One point is the formulation of the friction coeffi- lateral

cients (eq. (3)) depending on the aircraft's orien-

tation and movement, as the friction always acts
against the direction of movement. This can be
acceleration,
m/sec2
-A/^/ytyAyfr
modelled taking into account the sign of the x-
component of the aircraft ground speed UK ■ Eq.
(3) is now:

H= (Pr + Fb-pb ■ fibmax) ■ sign(uK). (7)

As a second point, the nose wheel and main wheel

skid angle computations of eqs. (5) and (6) have to
be expanded:

Tnw = rj ■ sign(uK) - atan [ ,u , "*" )(8)

l x|
vK + r ■ xmw
= —atan (9)
\UK - r ■ ymw

In order to verify these modifications, Fig. 11

presents a backward taxi validation plot with three
repetitions. For this validation simulation, the full lateral
ground
6-DOF model including engine and ground handling distance,
model was implemented. The lateral and longitudi- m
nal ground distance, which were provided by a laser time.sec
tracker, are simulated very properly. Some addi-
tional signals, i.e. lateral acceleration, yaw rate,
azimuth angle, and ground speed, show the simula-
Fig. 11. Backward taxi validation (3 repetitions),
tion quality of the model. simulation model output ( ) compared to
taxi test data ( )
7. LESSONS LEARNED

The examples of updating the C-160 ground han-

2nd order to 1st order. The validation of the model
dling model have clearly brought up the need for
with flight test data and taxi test data proved that
model validation. Nowadays, it is not the computer
this causes hardly any reduction in model accuracy.
power, but it is a clearly arranged model structure,
The next step could be to neglect tire dynamics at
that determines the scope of the simulator math all, probably the pilot would hardly notice it.
model. So a main task during the model develop-
ment process is to find that model structure com- Another example is the consideration of the tire ac-
plexity, which obtains a predefined accuracy [1,13] celeration impulse (Fig. 6), which is very impor-
in the most efficient way. tant for pilot acceptance. In this case, an equivalent
model is used to describe the effect sufficiently, as
For ground handling models, it is a tough decision
the correct computation of the actual friction coef-
on how detailed the model has to be. It is clear
ficient from total slip to full adhesion would require
that not all physical effects are noticeable for the
a huge data base. It is desirable to use those equiv-
pilot. The task is what amount of detail is needed
alent models, because they reduce the complexity
in order to reproduce essential effects. The main
and help to get a clear structure within the whole
criterion for establishing the necessity of modelling
math model.
an effect should be, besides the level of accuracy, the
pilot's awareness. In order to be independent of the Especially for ground handling models, there is a
pilot's opinion, flight (taxi) test data can support strong need for the use and the determination of
the decision on what is essential. Fig. 3 for example, such equivalent models. As shown in this paper,
presents the dynamic model of each gear suspension system identification methods are capable of sup-
without the mass of the tire. This simplifies the porting a procedure, which can be split up into four
differential equation for the tire axis altitude from steps:
12-

• detection of model reduction possibilities, Concluding, some aspects of effective ground han-
• determination of equivalent submodels, dling modelling are discussed. It is essential to find
• identification or adjustment of model parame- an adequate level of model detailism, for which the
ters using measured data, and criterion is - besides the required quality for ap-
• model validation. proval - the pilot's awareness. A clearly structured
model, which is easy to adapt to specific character-
This procedure can be a powerful tool, not only for istics, can be achieved using equivalent models for
model development, but also for validation. complex physics. For determination of equivalent
A simulator approval according to modern stan- model parameters, system identification applied to
dards requires a lot of flight tests in order to show taxi test data can be a proper tool.
the model quality compared to measured data. The
most convenient way to collect the additional data 9. ACKNOWLEDGEMENTS
necessary for the development of a new model struc-
The authors gratefully acknowledge the support
ture is to expand the taxi and flight validation tests.
provided by the German Air Force, the German
Procurement Organisation (BWB), and the Ger-
man Air Force Flight Test Center WTD-61.
8. CONCLUSIONS

This paper discusses some aspects of the develop- 10. REFERENCES

ment process of a ground handling model in order
[1] FAA; Airplane Simulator Qualification. FAA Advisory
to achieve a predefined model accuracy in an ef- Circular, AC 120-40B, Interim Version, Feb. 1991.
ficient way. Taxi test data, which were gathered
[2] Krag, B., Jategaonkar, R. V., Mönnich, W., and Fis-
during the validation tests for a simulator approval, chenberg, D.; Estimation of an Aerodynamic Data Base
are evaluated with system identification methods in for a New C-160 "Transall" Flight Simulator from
order to achieve a high quality model. Several ex- Flight Data. RAeS Symposium on "Data Issues for
Flight Simulators - An On-going Problem", Paper No.
amples for a C-160 ground handling model update
7, London, UK, Nov. 1993.
are discussed. [3] Maine, R. E. and Hiff, K. W.; Identification of Dynamic
Rolling and braking depend on the normal force Systems - Application to Aircraft. Part 1: The Output
Error Approach. AGARD-AG-300, Vol. 3, Dec. 1986.
and the corresponding friction coefficient. Evaluat-
ing differential braking taxi tests, two scale factors [4] Allen, L. D.; Evolution in Flight Simulation. AIAA
Atmospheric Flight Mechanics Conference, Paper No.
for the left and right brake pressures can be identi- 93-3545, Monterey, CA, Aug. 1993.
fied. The correct friction coefficients are validated [5] Baillie, S. W., Hui, K., and DeLeeuw, J.; The Flight
with two typical validation tests: ground acceler- Test and Data Analysis Program for the Development
ation and deceleration after touch down using the of a Boeing/deHavilland Dash 8 Simulator Model. 80th
Symposium of the AGARD Flight Mechanics Panel on
brakes.
"Flight Testing", Paper No. 30, Crete, Greece, May
A correct modelling of the tire acceleration at touch 1992.
down from total slip to full adhesion is difficult. As [6] Neville, K. W.; Simulaior-to-Flight Validation - Meth-
this effect is essential for pilot's awareness, an equiv- ods, Problems and Dilemmas. RAeS Symposium on
"Data Issues for Flight Simulators - An On-going Prob-
alent model is formulated, which uses the total ac- lem" , Paper No. 14, London, UK, Nov. 1993.
celeration impulse at each landing gear. Evaluating [7] Kullar, I.; Simulation Data Issues for Large Civil
flight test data, the parameters of this model are Transport Airplanes. RAeS Symposium on "Data Is-
identified and validated with landing impact flight sues for Flight Simulators - An On-going Problem",
test data. Paper No. 2, London, UK, Nov. 1993.
[8] Barnes, A. G.; Modelling Requirements in Flight Sim-
Lateral directional behavior with ground contact ulation. Aeronautical Journal, Dezember 1994 (1994)
depends on the tire side force model. An important [9] Doyle, G. R.; A Review of Computer Simulations for
part is the side force as a function of the tire skid Aircraft-Surface Dynamics. Journal of Aircraft Vol. 23
angle, which can be determined using taxi test data No. 4 (1986)
with strong nosewheel steering inputs at different [10] Barnes, A. G., Yager, T. J.; Simulation of Aircraft Be-
ground speeds. These identified characteristics are haviour on and Close to the Ground. AGARDograph
No. 285 (1985)
validated with lateral directional control test data
[11] Kortüm, W., Lügner, P.; Systemdynamik und Regelung
during takeoff. von Fahrzeugen. Springer Verlag, Berlin Heidelberg
As a last example, it is discussed how to achieve New York (1994)
a proper lateral control simulation quality during [12] Fischenberg, D., Mönnich, W., Krag, B., und Jate-
gaonkar, R. V.; Aspects of C-160 Simulator Model
backward taxi. Changing the sign of the nosewheel Determination and Validation on and Close to the
steering angle and of the friction coefficients, and a Ground. AIAA Flight Simulation Technologies Con-
modification of the tire skid angle computation are ference, Paper No. 94-3404-CP, Scottsdale, AZ (1994)
necessary to achieve a simulation fidelity according [13] Royal Aeronautical Society; International Standards
to FAA quality criterions for ground contact ma- for the Approval of Airplane Simulators, draft (1992)
neuvers.
 13-1

TACTICAL ENVIRONMENT SERVERS

Francois D. Heran
SOGITEC Division Electronique
4, rue Marcel Monge
92158 Suresnes Cedex
FRANCE

SUMMARY computer generated forces (CGF) will be used to provide the

required level of tactical realism to training or research and
Mission training or evaluation on a manned military simulator development simulators, whether they are new designs or
requires the presence of friend and foe forces. These forces, legacy systems adapted for DIS connectivity.
either computer generated or man-controlled, must react in an
intelligent manner, in real time. Due to restrictions in software This idea is the basis of the work that we have been
tools and computing power (or manpower in the case of conducting since 1991 on tactical environment servers, which
dedicated manned stations), as well as emphasis put on has been funded in part by the Delegation Generale ä
procedural and skill training more than on mission training, l'Armement (French MoD).
many existing simulators feature little or no tactical
environment. WHAT IS A TACTICAL ENVIRONMENT SERVER ?
This paper presents a concept, the tactical environment server, We will introduce this section by stating what it is not: it is
which has been developed to populate the battle space of real neither a very accurate model for operational evaluation
time combat simulators with various computer generated purposes, nor a constructive simulation. Potential problems of
friends and foes. using constructive simulations in conjunction with manned
virtual simulators are linked to their larger granularity (cycle
In the first part, tactical environment requirements for air
time, terrain cells, size of units); they also require operators.
combat and land battle simulators are discussed. Coherence
between the server and its clients, one of the most problematic The tactical environment server is designed to be a plug-in
technical issues, is analysed. unit requiring little or no human intervention during the
exercise, easy to set-up, preferably from the IOS of the
A second part addresses the algorithmic and technological
simulator to which it is connected. Implemented on a standard
trade-offs one has to consider to implement a real-time tactical
Unix workstation, its real time simulation models of friend
environment server. The connection of the server to the client
and opposing forces are adapted to individual or crew training.
simulator(s), via a network using the DIS protocol or via other
solutions, is also discussed.
TACTICAL ENVIRONMENT
Implementation of a tactical environment servers for a Mirage
2000 and a combat helicopter simulator are presented and The natural and man-made environment is comprised of 4
conclusions are drawn. domains:

This project is sponsored by the Delegation Generale pour - terrain,

l'Armement (French MoD). - weather,

INTRODUCTION - electromagnetic,

Modern unpredictable and remote conflicts, advanced weapon - infrared.

systems and C3I, decreasing military budgets and In this synthetic battle space, computer generated entities
environmental concerns are factors in favour of simulation- evolve dynamically in interaction with the manned
based mission training, which we define as training in the simulator(s). In addition, tactical events such as detonations
fulfilment of an operational mission in the presence of active occur.
forces. This training requires the presence of friend and foe
forces, either computer generated or man-controlled, which Although our aim is to develop a common foundation for the
act and react in a reasonably intelligent manner, in real time. simulation of air operations and land battle, tactical
These forces enhance the realism of the tactical environment environment requirements are quite different for these two
and create the desired workload and stress which are classes of simulation.
associated with combat. Engineering simulators, used to
design weapon system interface, also have to simulate the Simulation of Air Operations
crew's workload in the most demanding phases of the battle. Air operations which would most benefit of a tactical
Distributed Interactive Simulation promotes the creation of environment are :
synthetic tactical environments through the connection of - interception, where a friendly patrol seeks and fights enemy
multiple simulations over remote, geographically distributed intruding aircraft,
sites. Although the connection of manned simulators through
wide area networks in Europe is technically possible, local - air police,

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?",
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
13-2

- A/G strike, with multirole fighters or mixed aircraft, C3I entities are virtual entities that are necessary to model the
tactical C3I system (for instance, the link between air defence
- standoff A/G strike, units).
- reconnaissance.
Entity behaviour
Close air combat is not one of our design requirements, and is
better practiced with networked manned simulators. Entities exhibit at most two sets of behaviour rules (or
"brains"):
The trend in military training to include mission rehearsal
capabilities in the requirements of mission simulators should - one set is dedicated to piloting the platform (for instance, the
be addressed. The link between mission preparation and the "pilot" is able to follow flight plans, avoid obstacles or follow
simulated tactical environment is tight in this context and will a leading platform),
be taken into account for the design of future air operations - the other set controls the weapons system including ECM
tactical environment servers. and ECCM, gives orders to the "pilot", sends and receive
In the study we have conducted on tactical environment for tactical information (C3I interface).
combat aircraft simulators, we have focused on the Both "brains" have access to the environment through
challenging electromagnetic environment encountered by specialized server functions, such as height above terrain
modern aircraft and created by radars and jammers of ground processing or radar processing.
and airborne entities.
High-order logical rules may be edited to define particular
Adoption by the French Air Force and Navy of the AWACS aspects of the behaviour.
and E2C may also increase the amount of C3I required in the
tactical environment simulation.
TERRAIN AND TACTICAL DATABASES
Although modeling the electromagnetic environment does
Computer generated entities of the tactical server interpret
take into account the terrain, coherence requirements are not
natural and man-made environment at two levels of
extremely stringent in this domain. It should also be stressed
knowledge: the geometrical level and the tactical level.
that the aesthetical correctness of tank motion (or whether it is
Interpretation at the geometrical level, which is the level
moving at all) is quite irrelevant to a fighter aircraft.
usually fulfilled by the CIG in a manned simulator, uses files
describing the geometry of the environment. This is used for
Land Battle Simulation the assessment of intervisibility between entities, CGF and
Manned simulated platforms which are potential users of that remote, for computing altitude, height above terrain and for
class of tactical environment are tanks, armoured infantry modeling physical interactions such as driving on the ground,
vehicles, helicopters. landing, colliding,...
Simulation of the land battle usually involves: At the tactical level, a description of the world, which is in our
case an object-oriented database, feeds the decision processes
- scores of opposing forces, such as route planning, hiding and combat.
- coordinated friendlies, The following objects are extracted from the terrain database:
- area effect weapons, such as artillery fire and mines. - RIDGES and VALLEYS,
This tactical environment is closely linked to the natural - WOODS,
environment (terrain, objects, weather) and therefore
coherence requirements with the manned simulator - RIVERS,
environment are strong. - BRIDGES,
CGF ENTITIES - BUILDINGS,
- TREES,
Entity type
- ROADS.
Mobile entities are characterized by the following sets of
parameters: Coherence between the manned simulator and its externally-
generated tactical environment is a cruel necessity that has
- dynamic performance (performance of the platform been particularly emphasized by the DIS community.
expressed with 10+ parameters which include , for aircraft, Obviously, the geometrical and tactical databases must match
minimum and maximum speed at sea level, mass, max as closely as possible the polygons that form the visual
positive and negative g load, max turn rate and so on), database of the manned simulator although this may not be
- weapons system characteristics (radars, ECM, missiles and practically achievable due to the levels of detail implemented
so on), for the CIG.

- behaviour rules. Terrain related functions have demanding requirements on

coherence control: intervisibility, attitude and elevation of the
Fixed weapon systems may be seen as mobile entities with the computer generated entities relative to the terrain, collision
paradoxical characteristic of being immobile. A truck- avoidance (placement and existence of objects), road
mounted SAM site or a ZSU 23/4 exemplify that a distinction following.
with the previous type is vain.
 13-3

Intervisibility no major problems have been encountered when

implementing this protocol, there are a few domains not yet
The process of assessing whether a synthetic actor is able to
covered or unsatisfactory:
visually "see" another entity is very complex and does not
yield a binary result. This decision is reached after successive CGF initialization: Since there is at the moment no standard
stages: messages to initialize CGF with the DIS protocol,
initialization of the entities is described by means of a text
- at the filtering stage, the relief of the terrain and the
file.
maximum detection range are used to compute the coarse
geometrical visibility. This allows to conduct finer processing Electromagnetic PDUs: the DIS protocol still needs
for a reduced number of entities. improvement in the domain of electromagnetic interactions.
PDUs do not cover the full range of interactions (Eg IFF) and
- at the visibility stage, the overall visibility factor is
must be trimmed.
computed based on background and foreground, and
environmental conditions. The depth of scene analysis that is C3IPDUs: the DIS protocol defines a Signal PDU but tactical
necessary to derive a fine intervisibility information could messages have not been standardized. Some standardization
ultimately be close to the processing applied in a CIG. for this type of messages has been established by NATO but
not to the point of being easily interpreted by simulations and
- at the human modeling stage, physiological and
interoperability of simulations remains a long-term goal.
psychological conditions of the simulated observer have to be
taken into account with a set of deterministic (sensor
APPLICATIONS
characteristics, entity motion, visibility factor) and heuristic
rules (random time to detection).
MIRAGE 2000-5 Weapons system trainer
Terrain following This trainer, under development, allows training in air combat
and incorporates its own tactical environment server which
Since it cannot be expected that the terrain database used by
provides for friendly aircraft, airborne targets and ground
the tactical server is exactly correlated with that of the manned
threats. It will be connectable through DIS with other trainers
simulator, the attitude and elevation of a computer generated
and simulators.
land vehicle will generally look somewhat unrealistic to the
crew of this simulator. Terrain following for ground vehicles
can be achieved on the visual displays by requiring the CIG to Land Battle Tactical Environment Server
clamp the vehicles to the ground although this introduces a A tactical environment server dedicated to a manned attack
discrepancy between their computed attitude and elevation helicopter simulator is under development. The tactical
(ground truth) and what is visually perceived by the crew. environment features anti-tank helicopters, hostile attack
helicopters, tanks and armoured SAM carriers.
SOFTWARE TECHNOLOGY: FROM AI to C++
This server will be connected to the TIGER simulator in
Preliminary studies have been conducted on a SUN 3/160 operation at the Flight Test Centre in Istres, France.
workstation using LISP and the KEE environment from
Intellicorp. Although these tools allowed to test and validate Combat between helicopters will be simulated although
basic behaviour models and tactical databases, they would not restricted in a first stage to engagement with A/A missiles.
allow for real-time performance. In the search for this goal,
object oriented design and languages have been identified as CONCLUSIONS
more critical for these applications than AI techniques. We Our experiments with Land Battle applications have shown
therefore switched to the C++ language and a CLASS that our original goal of severely restricting human
RELATION based CASE tool. CLASS RELATION (Desfray intervention during the course of the exercise is very difficult
1994) is a full-fledged object model, based on the tight to achieve when, in some situations like hiding from an enemy
integration of three complementary models: in combat, computer generated forces will not escape from an
- the structure model describes application classes, their unrealistically static position. In real life, orders would be
properties (methods, attributes, integrity constraints), the expected from the higher echelon or actions would be inferred
relation and inheritance diagram and the schema and domain- from subjective reasons rather than from doctrine. A
based structure, complaint of the users we did not expected was that the
computer generated enemies were sometimes too smart and a
- the operating model defines how classes are used, the object degradation in combat performance was subsequently
lifecycle, and the methods usage modes, introduced.
- the dynamic model describes the reactive behaviour of The key word to user acceptance of computer generated forces
objects and the methods invocation dynamics. The CLASS is then flexibility, in the preparation and the execution of the
RELATION method is industrialized by SOFTEAM through exercise.
its CASE tool, OBJECTEERING
In the preparation of the exercise, not only the type, number
COUPLING THE TACTICAL SERVER: DISTRIBUTED and mission of the entities but also their behaviour must be
INTERACTIVE SIMULATION (DIS) PROTOCOL definable. It must be possible to assign subjective behaviour
attributes to units such as morale and proficiency, allowing the
The DIS protocol was selected for this programme as the simulated crew to choose between conflicting rules (protect
connection protocol with the manned simulator(s). Although your life and equipment vs destroy opposing forces).
13-4

During the exercise, we plan to give to the instructor the

possibility to enter new mission orders and to take control at
any time of a computer generated entity (Eg, fly an enemy
aircraft).
Further developments we plan to conduct for aircraft
simulation applications are related to the C3I environment.
Modern combat aircraft such as the Rafale feature
sophisticated tactical displays fed by data links and are tightly
integrated in a tactical network (in that case, the SCCOA)
which plays an essential part in the preparation of the mission
and in its execution. The tactical environment server must
generate the interface between the aircraft simulator and the
C3I entities and it must have the capability to use the output of
the mission preparation systems or to reproduce their
functionality.
We plan to study the simulation of tactical data links, airborne
radar systems and to extend the capabilities of the tactical
environment server in the mission preparation and rehearsal
domain.
 14-1

UK Attack Helicopter Flying Qualities: The Role of Piloted Simulation Evaluation

in Supporting the Procurement Decision Making Process

M T Charlton f
G D Padfield f
J T Green $

t Flight Dynamics & Simulation Department

Defence Research Agency
Bedford, MK41 6AE
United Kingdom

$ Attack Helicopter Project Office

MoDUK
London
United Kingdom

Summary: 1. Introduction

Contenders for the UK Army's Attack Helicopter In 1991 a Staff Target (ST) was raised for an
(AH) competition, were subjected to a Technical Attack Helicopter (AH) for the British Army to be
Assessment during the period November 1993 to procured 'off the shelf on a competitive basis. In
December 1994. The Defence Research Agency's this context, off the shelf meant that the UK MoD
Flight Dynamics and Simulation Department were would not fund a separate development contract
Lead Assessors for the flight control system and and did not intend to be the first user of the
handling qualities aspects of the assessments. FDS aircraft system. Hence, it was important to
carried out a programme of off-line and piloted determine exactly what was being offered by
simulation activities in support of the handling prospective bidders, to assess potential shortfalls
qualities assessments, using the DRA's HELISIM against the targets and to identify areas of risk. To
simulation model. A piloted simulation evaluation this end, the MoD AH Project Office initiated a
was completed using the DRA's Advanced Flight technical assessment of candidate aircraft against
Simulator, where the objective was to evaluate the the Staff Target. While much of the assessment
contenders handling qualities and agility in the was expected to be based on expert opinion,
context of the AH mission. The paper describes whenever possible assessors were tasked to
the test techniques and procedures used in the tests provide independent estimates for compliance
and discusses the background details of the checking and validation purposes, either by
handling qualities assessment methodology. Some analytical means or simulation modelling. One
conclusions are also drawn regarding the important area was the relative agility of the
effectiveness of simulation in supporting defence contending aircraft and the flying qualities
procurement objectives, and potential future necessary to underpin the aircraft's mission
developments. effectiveness.

Flight Dynamics and Simulation Department

(FDS) of the Defence Research Agency were lead
Presented at AGARD FVP Symposium 'Flight assessors for the flight control system and flying
simulation: Where are the challenges?', qualities aspects of the UKAH. In consultation
Braunschweig, Germany, 22-25May 1995. with the Project Office, FDS planned a
® British Crown Copyright 19951 DBA - Published coordinated programme of off-line and real time,
with the permission of the Controller of Her piloted simulation activities to support the
Britannic Majesty's Stationery Office. assessments. Piloted simulation evaluations of the

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?",
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
14-2

AH contenders were subsequently carried out at subsequently allowed five months in which to
the DRA's Advanced Fight Simulation (AFS) complete an initial appraisal of the TDRL data.
facility at Bedford, using the Large Motion During this time, they raised any necessary points
Simulator (LMS). This was the first time in the of clarification and requested any vital information
UK that such an exercise had been carried out to which appeared to be missing.
support a major procurement activity. The test
techniques and evaluation procedures were drawn The ITT was issued to five prospective bidders
from previous FDS experience of research and one additional unsolicited bid was received.
associated with new handling qualities criteria, One of the major difficulties faced by the Project
such as those related to the US Army's Office was the different developmental status of
Aeronautical Design Standard for the RAH-66, the contenders. At the time of receipt of the
ADS-33 (Ref 1). Tenders, none of the aircraft existed in the form in
which they were offered; the scale of difference
This paper discusses the background to the varied from a design concept, through an aircraft
assessments, the FDS approach to handling which had recently completed its first flight to
qualities evaluations, and the aims, test procedures aircraft being substantially updated from the
and results achieved in the piloted simulation trial. models currently in service. In many areas this
The paper also addresses simulation fidelity and was not necessarily a problem because the bidders
validation issues and the lessons learned. In the could provide sufficient data on which to base a
interests of commercial confidentiality, only risk assessment. Regarding the flying qualities
general results are given and there are no aspect, in view of the aircraft's need for an all-
references or inferences to any particular weather capability (supported by sophisticated
aircraft. visual aids), it was considered to be high risk to
rely solely on a 'paper' assessment and hence, that
2. Background and overview to UKAH the bids be subjected to a more thorough
Assessment evaluation. As the lead assessor for the flight
control system and handling qualities, FDS was
The UKAH Staff Target described the attributes of tasked with this undertaking.
the weapon system that the Army required
together with the primary mission details and As noted above, the flying qualities of the
desired performance in the form of a Cardinal contenders were evaluated through various
Points Specification (CPS). The UK MoD's simulation activities, of which the focal point was
Procurement Executive was tasked with identifying a piloted simulation evaluation using the AFS.
suitable equipment which might be available on the Other elements included an off-line assessment of
world market and assessing it in terms of cost each contender against the flying qualities criteria
effectiveness against the CPS. The technical specified in Aeronautical Design Standard ADS33
aspects of the assessment were carried out with the (Ref 1), the latest US specification for handling
assistance of almost 100 specialists drawn mainly qualities of military rotorcraft, using a specifically
from the Defence Research Agency. developed software 'Toolbox' (Ref 2). The
handling qualities Toolbox derives ADS-33 criteria
In terms of timescale, it took eighteen months based on inputs from test data or from the
from the time of the formation of the Project responses of the imbedded HELISIM simulation
Office to the issue of an Invitation To Tender model. In another activity, inverse simulation
(ITT) to the bidders. During this time, assessment techniques using the Glasgow University/DRA
teams were formed and briefed, who then assisted HELINV model (Ref 3), were used to predict the
in compiling a Technical Data Requirements list performance capabilities and control workload of
(TDRL) as part of the ITT. The TDRL comprised each contender in AH related mission tasks.
nearly 700 target parameters required by the ST
together with over 2000 back up items of data. The viability of the evaluations was critically
Bidders were required to return their proposals dependent on the quality of the data provided by
within nine months and assessors were the bidders through the TRDL. FDS had specified
 14-3

a number of TDRL requirements that were design and evaluation methodology is illustrated in
intended to elicit key data sets for building Fig 1, developed from Key (Ref 4). The detailed
HELINV and HELISIM configurations, and response type requirements follow from the
appropriate flight or model data for calibrating the user-defined missions and operational
DRA model responses. In the event, the data environments, and hence the usable cue
provided were adequate on both counts and environment (UCE). Resultant handling qualities
enabled satisfactory models to be constructed to levels are judged on a combination of results from
meet the aims of the evaluation plan. clinical open-loop and demonstration closed-loop
test manoeuvres.
3. Flying Qualities assessment methodology
vision aids
and displays

Good flying qualities underpin mission tables of response

user defines
effectiveness and flight safety. Establishing flying operational types for each £, response type

missions and - mission task

quality requires a combination of quantitative environment - visual cue

criteria, that define the customer's best

operational response type characteristics
understanding at a given time, and subjective flight
envelope hover and low forward
opinion of how well the aircraft is fit for purpose. speed flight
<45kn >45kn
At the time of writing, the available quality helicopter helicopter -equilibrium
standards for helicopter flying qualities are design and
fabrication
~& characteristics - response to controls
- response to disturbances
reasonably comprehensive. However, existing A A - controller characteristics

criteria relate to single axis response characteristics flight tests

failures
- clinical
and pilots rarely fly single axis tasks when - task manoeuvres
required levels
conducting a nap-of-the-Earth mission. A of handling
qualities
thorough test of quality therefore requires assigned levels
of handling
evaluation in task; the interplay between qualities

quantitative criteria and pilot subjective opinion of

task-worth characterises the DRA approach to
flying qualities assessment, and its application to Fig 1 ADS-33 requirements capture, design &
the UKAH, as outlined below. evaluation Methodology

3.1 Flying Qualities Synergy and the ADS-33 Any helicopter designed to the ADS- 33 standard
Standard - A Resume should exhibit very good qualities; formally the
ADS-33 standard states that a helicopter should
The DRA approach to flying qualities evaluation exhibit level 1 qualities (desired performance
is based on the concept that flying qualities consistently achievable at low pilot workload)
represent the synergy between the internal throughout the operational flight envelope. In this
attributes of the vehicle - its stability and control sense the standard has to be seen in the context of
characteristics, cockpit ergonomics etc., and the high levels of flight control augmentation, that
external factors that influence the piloting task - tame the natural tendencies typified by the lack of
threat level, atmospheric disturbances, quality of carefree handling, strong cross couplings and
visual cues etc. Implicit in this approach is the poor stability. The question then arises as to what
assumption that flying qualities are task-oriented as value is ADS-33 in judging the capabilities of
reflected in the new standard, ADS - 33D (Ref 1), existing aircraft or, more generally, aircraft not
anchored in a unique test database derived from designed to this Standard? This question is
advanced simulation studies and in-flight validation particularly relevant to the UKAH evaluation.
studies over the last 15 years. ADS-33 is formally Research experience to date suggests that most
a US Army standard for the RAH-66 helicopter, current operational helicopters exhibit a wide
but has been developed out of International range of Level 2 characteristics combined with
Collaboration and, in its structure and form, is some Level 1 and even Level 3 characteristics. A
applicable to all roles and types. The framework Level 2 helicopter can still perform missions with
for using ADS - 33 as a requirements capture, adequate performance but the pilot is likely to
14-4

1 ;— i " i i

—Q- low aggression

have to work much harder to compensate for -t^y- mod. aggression
Level 3
deficiencies. The ADS-33 standard has been -O- High aggression
-A- ADS33C !l,gnr.
developed to discern flying quality across all three tesi results
(ref 7.11)
—
A^ A
Levels, and hence is properly applicable to
Level 2 T ^"*^\\
existing aircraft as well as super-augmented
aircraft of the future, in both normal and failed
*LT^*^ T X :

conditions (where some degradation into Level 2 Level 1

and 3 is allowed). Indeed, much of the database
used in the development of ADS-33 from current 2.5 A 3.0 3.5
types was used to substantiate the new criteria in 1
no3
':
I
T
T306
ci)bw~rad/sec

the Level 2 region. One of the outstanding issues

ADS33C
in all flying qualities work relates to the effects of Level 1/2 u>m
(general MTEs)
a combination of Level 2 characteristics on pilot
workload and task performance and the uncertainty Fig 2 AFS Sidestep handling qualities results
surrounding these effects is perhaps the single versus ADS-33 flight test data
most important reason for the continued strong
emphasis on the need for pilot subjective the degree of conformity with the ADS-33 flight
evaluation in mission tasks. test data. Fig 2 presents roll axis handling
qualities results for rate command response type
ADS-33 states that, "Compliance with the aircraft flown in a sidestep mission task element
requirements will be demonstrated using analysis, (Ref 7). The key handling qualities parameter
simulation and flight test..." This places initial relating to closed-loop piloting is the attitude
emphasis on capability demonstration during bandwidth, defined conceptually as the highest
design, through analysis and simulation. frequency that the pilot can close a task loop
Confidence in the results of compliance without threatening stability (Ref 1). The
demonstration in design is critically dependent on ground-based simulation data are shown compared
the simulation fidelity level, including modelling with the ADS-33 flight test data, indicating very
and cueing environment. For the UKAH good correlation with the Level 1/2 boundary
evaluations, it was important that any limitations predicted at a roll axis bandwidth of about 2.5
caused by simulation infidelity were well rad/sec - about 25% higher than the, more
understood. Modelling issues are discussed briefly conservative, ADS boundary itself. The AFS data
in Section 4.2, but even with perfect aircraft also confirmed the importance of motion cueing to
modelling there is still a question over whether pilot control strategy, acting as a realistic filter to
tests conducted in ground-based simulation can 'high gain' activity on the one hand and
accurately predict flying qualities levels. suppressing the over-controlling typical without
motion, especially in the vertical axis (Ref 7).
A review paper by Condon (Ref 5) presents data
showing the extent to which ground-based The substantiating evidence of good fidelity,
simulation has improved during the 1980s. coupled with the engineering experience with flight
During the formative years on ADS-33, in the and simulation trials over many years, made the
early 1980s, there was a clear disparity between AFS an ideal tool for evaluating the UKAH
ground-based simulation and flight test data (Ref contenders flying qualities, relative to the ADS-33
6). Pilots were not able to achieve Level 1 standard. Two guidelines were established. First,
handling qualities with rate command response in general, it would be expected that the simulated
types in simulation, while the flight data predicted aircraft would be marginally more difficult to fly
a genuine Level 1/2 boundary. Problems were than the real aircraft. Second, that the quality of
attributed to poor visual and motion cueing in the the phototextured visual scenes and motion cueing
simulation and the ground-based data were were expected to be sufficiently realistic to expose
discounted. When the DRA's AFS became any potential pilot - induced - oscillations, which
available with a large motion capability in the can threaten flight safety at the higher levels of
early 1990s, one of the first tasks was to establish task aggression.
 14-5

3.2 The 3 - Stage Evaluation Methodology (iii) pilot in-the-loop simulation using the AFS;
This element of the methodology forms the main
The approach taken by DRA can be described topic of this paper of course and will be discussed
under three headings as follows: in more detail in later Sections. Underpinning the
piloted evaluation is a well defined series of
(i) clinical tests and the HQ toolbox analysis; The mission task elements or flight test manoeuvres
ADS-33 standard contains a set of criteria for with well defined desired and adequate
different response types and different control axes performance levels. These need to be (clinically)
(Fig 1). These response characteristics are further representative of operational situations, reflecting
subdivided into criteria for different ranges of in the present case the UKAH role. Pilots need to
frequency and amplitude. For example, agility be able to perceive accurately their achieved task
characteristics are represented by large amplitude performance, dictating careful design of the MTE
(control power) and moderate amplitude visual cueing. Pilots also need to be familiar with
(quickness) criteria, while stability characteristics the roles being considered and properly trained in
are represented by long term open-loop (e.g. the use of the Cooper-Harper Handling Qualities
phugoid, Dutch roll damping) and short term Rating (HQR) scale (Ref 9). The latter is
closed-loop (bandwidth) criteria. Quality criteria particularly important for achieving consistency
for the different forms of cross-coupling are also between pilots regarding the interpretation of low,
defined. These criteria are typically formed into moderate and considerable levels of compensation,
2 - parameter diagrams with defined boundaries for example. Finally, HQRs need to be arrived at
between Level 1, 2 and 3 quality standards. The following structured dialogue between the trials
DRA handling qualities toolbox (Ref 2) has been engineer and test pilot, that serves to document the
developed to derive these parameters from flight reasons for the HQR.
or simulation test data and to present results
automatically on the HQ charts. The DRA The three - stage approach contains a number of
Helisim model is an integral part of the Toolbox, synergistic features. The Toolbox analysis can
and pre-defined or custom test control inputs can focus attention on areas of apparent serious
be applied to the simulation model to produce deficiency while the HELINV results can identify
responses from which the HQ parameters can be limiting conditions to support the design of the
derived. The HQ Toolbox is developed within MTEs. In the next Section, the approach to and
the MATLAB/SIMULINK environment. Areas of results from the AFS trials are described in more
particular interest in the UKAH evaluation were detail.
agility, stability and cross coupling. Evaluations
were made with and without stability and control 4. UKAH AFS simulations - the approach
augmentation.
4.1 AFS trial objectives
(ii) inverse simulation; The Glasgow
University/DRA inverse simulation approach The overall objective of the FDS assessment
integrated into the software package HELINV (Ref strategy was to evaluate the contenders flying
3) was used to predict the limits to agility in qualities and to check that these would not unduly
MTEs. HELINV takes as input the MTE, defined constrain the levels of 'useable' agility, in the
in terms of flight path kinematics, along with context of the AH's primary mission. The AH will
aircraft limits, e.g. control, power. The HELINV be required to operate in the battlefield
algorithm effectively inverts the simulation model environment, primarily flying anti-armour, ground
to compute the rotor loads and hence controls suppression and anti-helicopter missions. For
required to fly the manoeuvre. Some validation of mission effectiveness, it was stipulated in the
this approach has been conducted with Lynx flying target operating characteristics that "...The AH
slalom MTEs (Ref 8), where comparison between should have handling and engine response
flight test and HELINV results indicated that qualities appropriate to accurate flight path control
control limits were reached at very similar levels with low pilot workload in the NOE, battlefield
of agility, at a slalom aspect ratio of about 11%. environment. Suitable means should be provided to
14-6

allow for exploitation of the full flight envelope manoeuvring. Comparisons with the test data
when flying 'eyes out' without the risk of provided by the AH bidders confirmed this. The
inadvertent and unacceptable excursions beyond linearity assumptions also become increasingly
it....". In addition, a number of key point fragile a high speeds, but, since only low - mid
performance characteristics for given flight states speed MTEs were flown, this weakness was not
were also defined, which specified the desired considered to have a significant impact on the
acceleration and speed capabilities for the aircraft. results.
Taking the two issues together, it is implicit that
the aircraft should embody good agility and One of the known failings of a flap-only model
manoeuvrability coupled with handling qualities with simple 3-component inflow modelling is the
that allow the pilot to exploit the available poor fidelity of cross coupled pitch/roll responses
performance, with confidence and safety. and the HELISIM versions of the UKAH
contenders were no exception. Comparisons
The specific aim of the AFS trial was to check the varied from poor to fair and the general approach
pilot in-the-loop flying qualities, levels of taken during the off-line analysis and piloted
workload, task performance and agility for each of simulations was to reduce the emphasis on the
the contenders, and to provide important data for cross coupling quality criteria. During the piloted
comparison with inputs from the off-line trials, particular care was taken to identify any
simulation predictions. Specifically, the objective adverse comments relating to a characteristic that
of the trial was to conduct piloted evaluations of was known to be poorly modelled. In the event,
the contenders' handling qualities in mission- none of these areas appeared to be critical to the
orientated tasks extracted from key flight phases of test pilots, who were encouraged to give emphasis
the AH's primary anti-armour role. to the primary control response (agility) and
stability characteristics.

4.2 Simulation Models Creation and Rotorspeed was assumed contant for all
Validation configurations. HELISIM does feature a generic
powerplant/rotorspeed governor/fuel flow model,
Simulation models of the UKAH contenders were but insufficient data were provided to model the
created based on the data provided in the TDRL different configurations. Constant rotorspeed will
responses from bidders. Currently there are three on the one hand obscure any handling features
versions of the generic DRA HELISIM, relating to delayed engine response or torque
distinguished largely by the complexity of rotor overshoots; also, the instantaneous engine response
modelling. The hi-fidelity version employs an is likely to result in less representative yaw
aeroelastic rotor model with non-linear unsteady coupling. Any pilot comments relating to these
aerodynamics and is currently undergoing issues were noted, as with pitch to roll couplings,
integration into the real-time AFS environment. although, once again, they did not appear as a
The Helisim version adopted for the HQ Toolbox major driver to the HQRs.
and HELINV analysis employs a rigid-blade disc
approximation, with dynamic multi-blade Configurations were modelled together with the
coordinate representations of blade coning and stabilisation components of the stability and
cyclic flapping motion; blade aerodynamics are control and autopilot augmentation systems, again
linear (Ref 10). For the real-time simulation, the using data supplied by the manufacturers.
rotor blade degrees of freedom were further Autopilot modes were excluded since all the MTEs
approximated by quasi-steady representations of evaluated were essentially full attention active
flapping and coning. This level of approximation flying tasks.
is known to give moderate levels of fidelity across
a frequency range between zero and about 10 Overall, correlation with the test data provided by
rad/sec, in terms of primary axis control response, the bidders showed adequate correlation for
in the absence of aerodynamic nonlinearities e.g. primary response characteristics in terms of
caused by interactional effects or rapid control power and damping. This conclusion is
 14-7

supported by previous validation work conducted motion cueing was provided by the LMS. Key
using HELISIM with Lynx, Bol05 and Puma features are summarised below.
flight data. As noted above, cross coupling was, in
general, poorly modelled, although the levels were General
such that, in a broad sense, similar handling
qualities would be expected between model and the Specific features include:
real aircraft, e.g. Level 2 handling qualities for
pitch/roll/pitch coupling described in terms of the • Electric feel-system with Lynx mechanical
ratio of off-axis to on-axis response span the wide controls (centre-stick, rudder pedals and
range from 25 to 60 % in ADS-33. collective)

4.3 Test and evaluation methods • Lynx seat featuring vibration cueing scheduled
with airspeed and normal 'g'
The test and evaluation procedures used for the
AFS assessments were based on well tried and • Link Miles Image 600-PT visuals featuring 3
robust techniques, developed during previous FDS windows with maximum field of view:
handling qualities research (Refs 7,11,12,13)
through the complementary use of the ground - Total azimuth: +/- 63deg,
based AFS and Lynx/Puma airborne test facilities. - Centre window vertical: +/- 18deg
From previous experience, notwithstanding the - Side windows vertical: +/- 24deg
limits of simulation capability, it was considered
that the results would provide a valuable insight • LMS motion cueing: Maximum accelerations
into the contenders primary axis handling including -
characteristics to complement the HELINV and
Toolbox analyses. Moreover, given the importance - Vertical: +/- lg
of motion cueing for piloted handling qualities - Lateral/Longitudinal: +/- 0.5g
evaluations, the AFS with its Large Motion - Angular: +1-2 rad/s2 pitch,
System (LMS) was considered to be well suited 3 rad/s2 roll, 1.5 rad/s2 yaw
for the AH assessments.
• Head down display of primary flight
The AFS configuration, test procedures and test instruments, e.g artificial horizon, airspeed
manoeuvres are discussed in more detail below. indicator, normal 'g' meter, main rotor speed and
engine torque indicator

4.3.1 Simulator Configuration • Data logging facilities:

The available time and resources precluded a - automatic recording of pilots control activity,
detailed representation of each aircraft's cockpit, aircraft responses and flight path coordinates via
controls, cockpit systems and displays. Each computer disk and pen chart recorders
contender was evaluated in a 'standard'
configuration, with the assumption that they would - video records of pilot's eye-view from the
be equally affected by any deficiencies. centre window

The simulator configuration used in each case Controls configuration

featured a single-seat helicopter cockpit with a
Lynx seat and controls and a 'standard' set of Where possible, the controls were configured as
head-down flight instruments. Visual scene content friction devices or with static force gradients and
was displayed via a Link-Miles Computer breakout characteristics using information supplied
Generated Imagery (CGI) visual system and three by the bidders. Alternatively, the controls were
cockpit mounted monitors, arranged to provide configured with Level 1 characteristics in
one centre and two side windows. 'Platform' accordance with Def Stan 00-970 (Ref 14) and
14-:

AIRCRAFT DEMANDS ON THE PILOT PILOT

ADEQUACY FOR SELECTED TASK OR
CHARACTERISTICS 1 SELECTED TASK OR REQUIRED OPERATION* RATING
REQUIRED OPERATION*

Excellent Pilot compensation not a lactor lor

Highly desirable desired performance

Good Pilol compensation not a lactor lor

Negligible deficiencies desired perlormance

Fair — Some mildly Minimal pilol compensalion required loi

unpleasant deficiencies desired perlormance

N
Minor but annoying Desired perlormance requires moderate
pilot compensation

Pi
deficiencies
Deficiencies Moderately objectionable Adequate perlormance requires
warrant considerable pilot compensation
improvement
deficiencies
-&a
Very objectionable but
tolerable deficiencies
Adequate perlormance requires extensive
pilot compensation fcl
Adequate performance not altainable w
Major deficiencies maximum tolerable pilol compensation.
Controllability not in question
Deficiencies
requite Major deliciencies
improvement
Intense pilot compensation is required to
Maior deficiencies (etain control

H Maior deficiencies
Control will be lost during some portic
required operaiion

Fig 3 Cooper-Harper rating scale for handling qualities

ADS33C criteria. Regarding dynamic contenders, the AFS latency would not have a
characteristics (frequency, damping and inertia), significant impact on the validity of the
for the centre-stick and rudder pedals data evaluations.
representing measurements from a Lynx were
used. The rudder pedals and collective controls 4.3.2 Test procedures
had trim force release buttons, positioned on the
collective control grip, while the cyclic control had To meet the assessment objectives, test pilots
trim follow-up and trim release buttons, mounted evaluated the contenders' handling qualities in a
on the hand grip. These functions were not set of mission-related flight tasks, or 'mission task
considered to be part of the assessment and were elements' (MTE). MTEs form the basis for
generally only used when setting up trim 'stylised' tasks specifically designed to enable
conditions. formal handling qualities evaluations using the
Cooper-Harper rating scale (Ref 9), shown in Fig
Simulation transport delays 3. To help confirm pilot impression and to enable
them to review the handling qualities in a total
With the CGI visual system, the total system mission context, pilots were also required to fly a
latency, ie. time between initiation of a control simulated AH mini-mission. In this task, pilots had
input and visual system response, has a mean of to complete an NOE flying sequence, interspersed
115ms +/- 10ms. The latency is an important with discrete mission task elements, in order to
factor in handling qualities evaluations because it accomplish a number of mission objectives, eg.
directly influences the minimum achievable phase target acquisition, general reconnaissance. For the
delay and maximum bandwidth that can be formal evaluations, the pilots were required to
modelled. However, checks on altitude bandwidth achieve the tasks within given accuracy constraints
using the handling qualities Tool-box indicate that, and special visual cue arrangements were used to
when compared to the predicted data for the assist the pilot in judging the level of task
performance achieved. Pilots were also required to
 14-9

evaluate the tasks at different levels of aggression, its primary role, the AH will typically be expected
where aggression refers to pilot control strategy to spend a high proportion of time in NOE flight,
and may be taken as an indication of how hard the at speeds below 80kn, and in manoeuvring at low
pilot is 'driving' the aircraft, or the level of speed close to the hover. As an agile combat
inherent aircraft performance that is exploited in helicopter, it must be capable of delivering rapid
the execution of the task. Experience has shown and accurate control of flight path. To this end,
(see Refs 11-13) that testing over an increasing the speed and precision with which the pilot can
range of aggression can expose potential handling redirect the rotor thrust, through control of
qualities 'cliff-edges', which signify a rapid rise in attitude, will be a major factor. Hence, the roll
workload as the pilot strives to maintain adequate axis response characteristics, and to lesser extent
task performance under increasing time pressures. those of the pitch axis, play a key role in
Hence, the intention was that pilots should explore determining the suitability of the aircraft's
the effect of task aggression on task performance, handling qualities for the role.
workload and agility in their handling assessments.
The MTEs chosen for the evaluations included two
In order to achieve a reasonable consensus, hover and low speed tasks, the lateral sidestep and
evaluations were completed by three different the quickhop, and one forward flight task, referred
pilots, who were each allowed three sorties in to as 'lateral jinking'. These MTEs had been
which to assess each contender. The first sortie developed and tested in previous FDS research
was allocated to training and familiarisation and (Refs 11,12) and through participation in the
the following two for formal evaluations. Before European collaborative programme on ACT for
assigning a rating, the evaluation pilot was allowed rotorcraft (Ref 13). They represent handling
to practise the task until a consistent level of task qualities evaluation tasks with well defined control
performance had been attained; the on-line data strategies and manoeuvre objectives, and with
logging was used to provide feed-back information clear performance goals and levels of task
to the pilot on applied levels of task aggression aggression. Handling qualities in the tasks are
and task performance achievement. Handling dominated by the primary roll or pitch axis
qualities ratings were recorded using the Cooper- response characteristics, where key parameters will
Harper 'decision tree' and scale and, in addition, be roll/pitch controllability (control power and
a special questionnaire was used to record sensitivity), roll/pitch attitude quickness and
supporting comment and opinion. Pilot's control closed-loop stability (attitude bandwidth and phase
demands, flight path coordinates and vehicle delay)
responses were also logged for all designated
evaluation runs and subsequently used to confirm The sidestep and quickhop are essentially hover
achieved levels of task aggression and task re-positioning manoeuvres (see Figs 5 & 6), which
performance. At the end of each sortie, pilots might entail moving from one point of cover to
completed a further questionnaire as a means of another with minimum exposure time, or perhaps
providing a more detailed debriefing on their a move from cover to complete an observation
ratings and assessments. They were also asked to task. The sidestep task is dominated by the. roll
complete a summary report on their overall axis response, but at the same time a multi-axis
impression of each configuration, based on their control strategy is needed to coordinate the
experience in flying the simulated mission. heading (rudder pedals), height (collective) and
track over the ground (longitudinal and lateral
4.3.3 Test Manoeuvres cyclic). In handling terms, the sidestep is
characterised by three distinct phases; an initial
From ADS33C, a mission task element or MTE, roll to accelerate the aircraft, followed by a roll
is defined as "...An element of a mission that can reversal to decelerate and a final roll out to stop
be treated as a handling qualities task...". For the and re-establish a hover. In the same way, the
AH evaluations, MTEs were selected on the basis quickhop features a pitch down, pitch reversal and
of the primary mission profile and the piloting pitch down phases associated with accelerating and
tasks associated with key phases of the mission. In stopping the aircraft.
14-10

Table 1: MTE task performance requirements

MTE MTE Phase Heading/ Height Ground track Over/ Plan position Speed
balance -ft- - m - under - m - -kn -
- deg- shoot
- m -

Sidestep Transition + /- 15 D > 5 D + /- 3 D + /- 3 D na na

+ /- 25 A > 10 A + /- 6 A + /- 6 A

Terminal +/- 5D + /- 5 D na na + /- 1.5 D na

hover +/- 10A + /- 10 A + /- 3 A

Quickhop Transition + /- 15 D > 58 D + /- 3 D + /- 10 D na na

+ /- 20 A > 65 A +/- 15 A

Terminal +/- 3 D + /- 5 D na na + /- 3 D na
hover +1-6 A + /- 10 A + /- 6 A

Lateral Overall +/ 5D + /- 2.5 D na na na + /-5D

jinking +/-10 A + /- 5 A +/- 7.5A

Tracking + /-5D +/- 2.5 D +/- 3 D + /- 3 D na +1-5 D

+ /-10 A +/- 5 A + /• 6 A + /- 6 A + /- 7.5A

Table 2: MTE task aggression requirements

MTE Aggression Level of Aggression

parameter
Low Moderate High

Sidestep Net roll attitude change during 8-12 deg 18-22 deg 28-32 deg
accel/decel

Quickhop Net pitch attitude change 8-12 deg 18-22 deg 28-32 deg
during accel/decel

Lateral jinking Maximum roll attitude during 10-20 deg 25-30 deg 40-50 deg
the 'jink'

Table 3: AH Simulated mission sequence

Ele- Location Objective

ment

1 Centre of farmyard Hover at 25ft AGL, spot turn & transition to river bed

2 Northwards along river bed Follow river bed maintaining height 25-30ft & speed 30-50kn

3 First village Enter village, decelerate to hover and spot turn to reconnoitre
buildings

4 First thicket on right bank of river Return to NOE flight following river
- Rapidly decelerate to hover within cover of trees

5 - ditto - Execute bob-up to reconnoitre church, followed by bob-down to cover

6 - ditto - Execute sidestep to right to reconnoitre surrounding terrain and return

to cover

7 Northwards along river bed Execute rapid sidestep to left from cover, turn rapidly & resume NOE
flight following river

8 Second village, right bank of river Rapidly decelerate to hover within cover of building

9 - ditto - Execute bob-up to reconnoitre church & and bob-down to cover;

bob-up to acquire & track building

10 - ditto - Turn & make rapid withdrawal along the river bed (70-80kn).
 14-11

Lateral jinking (see Fig 7) is essentially a roll axis review their impression of the handling qualities in
slalom manoeuvre, combined with sequences of a broader range of mission tasks. The general
tracking elements, and is designed to test the mission plan and sequence of events are
capability for accurate control of flight path in low summarised in Table 3. A representative view of
level NOE flight, in the presence of obstacles in the CGI scene detail for the NOE database is
the ground plane. The test objective is to check shown in Fig 4.
lateral/directional handling qualities in transient
turning manoeuvres in the mid-speed range and in
acquiring and maintaining a designated track.
Again, in handling terms, the task is characterised
by an initial roll in followed by a roll reversal to
achieve the flight path re-alignment, and then a
roll out to acquire and hold the tracking line. The
manoeuvre is dominated by the primary axis roll
responses, but again a multi-axis control strategy
is required for maintenance of height, speed and
balance.

Task performance and aggression requirements are

defined in Tables 1 & 2. Regarding task
performance, note that where appropriate,
different requirements are given for 'desired' and Fig 4 NOE scene detail
'adequate' levels. The target levels of task
aggression were specified through an appropriate
parameter associated with the primary control axis;
for a sidestep for example, aggression is specified
in terms of the maximum roll attitudes to be
achieved during the acceleration and deceleration
phases of the manoeuvre. The three levels are also
indicative of the maximum angular rates, attitudes
and translational rates to be achieved during the
manoeuvre. In relation to the AH mission, the
intention was that 'low' aggression represents
unhurried or cautious manoeuvring in the presence
of threats, or perhaps manoeuvring in poor
visibility or confined places etc. Similarly,
'moderate' represents manoeuvring with 'normal'
levels of mission urgency where there may be no
direct or imminent threats, while 'high' represents Fig 5 Sidestep MTE
rapid weapon deployment, direct threat
avoidance/evasion, or rapid withdrawal from
danger zone. 4.3.4 CGI Task cue arrangements

Referring to Fig 5, for the sidestep the principal

As noted above, pilots were also required to fly a cues are provided by a building-like structure and
simulated AH mini-mission as part of the sighting arrangement. The near and far sights are
assessment. The mission sequence was flown after designed to give height and plan position error
the MTE tests and although not formally feed-back relative to the 5/10ft (desired/adequate)
evaluated, it gave the pilots a final opportunity to requirement. The road in the foreground is
14-12

intended to give additional longitudinal position 5. UKAH simulations - the results

cueing during the lateral translation. The posts
provide additional height and longitudinal position The principal test objectives as discussed in the
cueing for precision hovering. previous section were met. Sufficient information
was provided by the bidders for building
simulation model configurations and the calibration
exercise gave confidence that the model responses
were acceptable for the piloted simulation test
objectives. Results achieved in the subsequent
simulation evaluations were consistent with the
findings of previous FDS flight control related
research and were also in good agreement with the
off-line assessments using the handling qualities
Tool-box. In particular, pilots comments and
ratings were well matched with predicted handling
qualities levels for roll and pitch axis bandwidth,
attitude quickness and control power criteria.

Handling qualities evaluations were carried out by

Fig 6 Quickhop MTE three pilots at moderate levels of task aggression,
where in all cases either desired or adequate task
performance requirements were achieved. The
Task cues for the quickhop are shown in Fig 6. tasks were also attempted at higher aggression
Height and position cueing are given by a road levels and in nearly all cases, the pilots were able
running between a building and a gantry- like to achieve at least adequate performance levels.
structure. The building window line shows the Most important, pilot comment showed that,
target task height and its upper extremity generally speaking, simulation constraints did not
delineates the desired task performance range. The unduly influence their perception of handling
initial and final hover positions are given by qualities. One formal set of evaluations at high
adjacent vertical black lines on the wall and the aggression was carried out and the results are
gantry, which for correct positioning should be presented below, together with those for moderate
aligned in the centre of the side windows. The aggression. Some general comments on the overall
road edges provide lateral positioning information. results are also given.
The vertical posts in the forward field-of-view are
placed to give height and track cues during the
final 'flare'.

For lateral jinking, see Fig 7, the task layout

consists of a sequence of turning gates located on
either side of a roadway. The gates are represented
by sets of four posts, comprising a small inner
pair, which represent adequate tracking
performance (12m wide), and a larger outer pair
which serve as a height reference. The direction of
the roadway defines the tracking lines, while its
width (6m wide) defines the desired tracking
performance.
Fig 7 Lateral jinking MTE
 14-13

5.1 Pilot ratings reduce the scope for 'backing-off on control

demand. As noted above, testing across a range of
A summary of pilots ratings for handling qualities aggression can give evidence of a handling
is shown in Fig 8. The plot shows the mean and qualities 'cliff-edge'.
spread of ratings for each configuration, for the
three evaluation tasks.
5.2 Pilot comments and opinion
0 Moderate agg ^Highagg |
From pilot comment, for the tasks evaluated
.£ 6 simulation related limitations/deficiencies were not
T considered to be unduly intrusive, and were
generally judged not to have had a significant
., T impact on the results. A representative comment
a from one pilot was that:

•o
c
ra "... The AFS provided a good method of
X comparison of the three aircraft. Some simulation
Max/min/mean
ratings
limitations detracted from the overall realism but
probably had little effect on the results achieved.
SIDESTEP LATERAL JINKING QUICKHOPS The visual system was limited at high pitch up
Fig 8 Pilots ratings for handling qualities attitudes which precluded conducting truly
representative tactical manoeuvres, however the
For moderate aggression, the spread of ratings was required level of aggression was achieved in the
generally within one scale point, indicating a good assessments. The motion system gave good cues
consensus of opinion between the pilots. In three for most tasks, but when manoeuvring
instances scatter increased to 1.5 or 2 scale points, aggressively the lateral and yaw cues became
and in one of these cases, ratings also crossed the unrepresentative..."
Level 1-2 rating boundary. The variation may be
explained by differences in control strategy and As commented, field-of-view limitations in the
task aggression. Within the targeted task quickhop task were probably the most noteworthy
aggression, at low to moderate levels, the pilot can limitation, and may have given rise to an estimated
use significantly lower roll rates to achieve the degradation in ratings of 1 scale point at the
change in roll attitude, as indicated by lower moderate level of aggression. At higher
attitude quickness (Ref 7). This may reflect choice aggression, motion cueing was noticeably more
of a more relaxed control strategy, or may result intrusive, although pilots still preferred to fly with
from poor handling qualities, where the pilot is the motion system engaged. From past experience,
using a strategy that reduces the need for excessive tasks flown at high aggression in fixed-base
compensation. Also, from observation, application simulations tend to give rise to pilot disorientation
of aggression tended to be a function of pilot and even nausea. This is particularly true for NOE
background, ie. experienced 'attack' helicopter tasks which involve dynamic roll axis manoeuvring
pilots tended to achieve higher attitude quickness such as lateral jinking.
values when compared with transport helicopter
pilots. The head-down instrumentation was reported to be
generally too far from the pilots normal scan to
Ratings for the high aggression case show a clear provide more than minimal assistance, particularly
increase in mean rating compared to the moderate in high aggression tasks. Regarding torque
aggression results. Such results reflect the need for margins, at the target levels of task aggression,
a sample of different subject pilots and to test pilots were mostly able to achieve the task within
across a range of task aggression. As aggression the desired margin, particularly for the lateral
increases, the combination of task performance jinking task which was evaluated at a speed close
demand and increasing time pressures on the pilot to the minimum power case. However, for the low
14-14

speed tasks at higher levels of aggression, when potential modifications and upgrades which might
manoeuvre attitudes exceeded 25deg, there was a be needed, either to address shortfalls in current
noted tendency to exceed the defined limit. operational capability, or to ensure continued
effectiveness throughout its service life. Such
There were no adverse comments regarding the activities will usually entail an extensive trials and
cockpit inceptors. Where used, the Lynx controls evaluation programme to provide the necessary
breakout forces and gradients were considered to substantiation data. For an 'off-the-shelf buy,
be satisfactory, and there were no reported while existing test data (supplied by the contractor)
handling qualities deficiencies due to control may already be available, further testing will
characteristics. almost certainly be required.

From a Project Office perspective, accomplishing

6. Future challenges test plans in a timely and cost effective manner is
a key issue. Flight testing is an expensive and
6.1 Future applications sometimes 'risky' business, and simulation can
potentially help to improve the cost effectiveness
The direct use of handling qualities simulations and safety of the evaluation process. For example,
during an aircraft procurement has been a novel simulation can be used to increase the productivity
and successful experience. For a competitive of flight trials by identifying and directing
selection process such as the UKAH, the AFS was attention to safety critical conditions within the
demonstrated to be a valuable tool for assessment flight envelope. Simulation allows flight up to and
of handling qualities, whereby it allowed piloted beyond flight envelope limits to be explored,
evaluation of representative aircraft models in an without the attendant flight safety risks. Again,
even environment against predetermined simulation can provide a flexible tool for
objectives. The ability to compare results against preliminary assessment of airframe or flight
other areas of the technical assessments increased control system modifications, eg. airframe mass
confidence in the overall level of credibility. re-distribution (ie. addition to or variation of
Although actual flight evaluation of current models weapon load), or an AFCS modification, for
of the contending aircraft was undertaken, these impact on handling qualities. The capability for
activities were regarded very much as assessing flight in degraded visual conditions is
complimentary to the simulation activity, with another important application of simulation, and
different primary objectives for each. However, it facilities like the AFS are regarded as essential
is important to note that in common areas, the tools for the development and testing of helmet
findings from the piloted simulation were mounted displays (HMD). From a research
consistent with those for the aircraft evaluations. standpoint, simulation provides an important
resource for investigation of technologies and
Following the successful completion of the AH concepts that might be exploited in a mid-life
assessments, it is of interest to review the potential update, the application of Active Control
role of simulation in future procurement activities, Technology (ACT) being a prime example.
both from a general standpoint and in association
with the chosen AH. Reviewing a typical For the potential to be fulfilled, it is important that
procurement process, there are three key follow-on the issue of simulation fidelity is addressed. From
phases where support might be considered. Firstly, the 'customers' standpoint, there is a need to
following contract award, for 'acceptance off define adequate levels of fidelity in relation to
contract' there is a need to check the 'deliverable' specific simulation objectives, and to specify the
against the agreed performance specification. substantiation criteria that will be needed. In view
Secondly, for an aircraft it is then necessary to of the ongoing requirement, it is important that
establish flight envelope limits for mandatory there is a development strategy for facilities like
airworthiness requirements and for in-service use the AFS that will provide the desired capability.
of the aircraft. Finally, in parallel with the other
activities, it is essential to assess the aircraft for
 14-15

6.2 Simulation requirements Visual system:

For future applications, to meet the customers For evaluation of handling qualities in NOE flight,
needs, it would be clearly advantageous to be able where flying is essentially 'eyes out' of the
conduct evaluations in specific mission cockpit, it is important to achieve a good
environments and operational scenarios. Regarding representation of the outside visual scene. The
battlefield helicopter roles, operation in the DVE outside visual cues provide the pilot with the
is a high risk area for handling qualities information needed for accurate flight path
requirements and hence it is considered that there control, and will have a critical impact on the
is a particular need to achieve a high fidelity ability to adopt a realistic control strategy; poor or
simulation of the helicopter and its visual missing cues will almost certainly demand some
aids/displays in the DVE. Such a simulation would degree of pilot compensation which will serve to
serve as a cost effective tool to support generic spoil the impression of handling qualities. Hence,
research into visual system development, factors such as the available Field-of-view (FOV)
assessment and compliance testing of future and the representation of fine detail and textural
projects. This set of research is currently receiving cues can have a critical influence in piloted
attention in a pan-DRA activity funded by the handling qualities evaluations (Ref 15). Similarly,
MOD's Applied Research Programme. where there is a requirement to simulate DVE
conditions, the fidelity of the visual environment
Key elements of a high fidelity simulation include model will be of primary concern. For features
models of the aircraft and its mission systems, such as fog and ambient light modelling, there is
including the ability to represent two crew the problem of defining the cueing requirements
operations, the natural environment, including and levels of fidelity needed to achieve an
visual scene and effects of wind and turbulence, adequate simulation.
and the operational scenario, including modelling
of threats. The AFS has demonstrated its In addition to FOV, the issue of scene detail also
capability for supporting fundamental rotary wing needs to be addressed. The resolution provided by
handling qualities research over the past four years typical visual systems is best suited for forward
and more recently, it has demonstrated the flight and improvements are needed for helicopter
capability for supporting work on visually coupled low speed flight in the NOE. The problem is most
systems (VCS) for DVE operations. At other sites accute for very low speed tasks close to the hover,
within the DRA, there simulation facilities which particularly when high precision is needed, eg.
already provide sophisticated mission simulation positioning or weapons aiming tasks. To put this
capabilities. Hence, for the future, the AFS may into perspective, pilots have reported that in real
best be adapted to meet the customers needs NOE flight, they use details such as the movement
through 'networking' to run parallel simulations and aspect of blades of grass for information
with those facilities. regarding relative wind direction, speed over the
ground and flight path direction. To some extent,
especially for transient manoeuvring cases, the
6.3 Facility configuration situation can be improved by adopting task designs
such as those used in the AH assessments. Here,
Generally speaking, the AFS is well suited for the the objective is to enrich the visual scene with
type of assessment programme that was carried out patterns of objects, which, as the vehicle moves
for the AH. The facility will be subject to even slowly, serve to create a discrete 'flow field'
upgrades programmed throughout its life that will effect, from which relative translational rate and
enhance its capability to meet the requirements positioning cues may be perceived.
discussed in the previous section. The main
elements of the simulation and the potential Cockpit, pilots controls and instruments
influence on handling qualities are discussed in the
following paragraphs. For mission simulations for a two seat aircraft like
the AH, as noted above, it should be possible to
14-16

simulate 2 crew operations through networking. manoeuvre situations, onset of 'g' is an important
For the pilot station, from a handling qualities indicator of aircraft flight state and has a primary
standpoint, it will be important to provide suitable influence on pilot control strategy. For a pilot in-
inceptors and flight instruments. The inceptors the-loop control task, normal 'g' provides a feed-
dynamic and static force displacement and back loop by which the pilot judges and moderates
trimming characteristics can have a significant the strength and progress of the flight path
impact on pilot impression of handling qualities. response. From Ref 18, missing cues can give rise
Hence, it is important that they are adequately to over-controlling and increased workload and use
represented in the simulation. of a 'G' seat can help to recover the situation.

As a source of primary cueing information for From Ref 19, fixed-wing research has shown the
piloting control, cockpit instrumentation is another importance of motion cueing in achieving the
important factor in handling qualities evaluations. necessary fidelity to simulate high gain, pilot
For helicopters, the need to observe critical flight control tasks, showing that motion was needed to
envelope limits such as normal 'g', rotor torque in order to capture pilot-induced-oscillation
and RPM serves as a primary source of workload; characteristics (PRO). Similar results have also
audio and tactile warning devices are sometimes been reported for helicopter simulations, see Refs
available to provide some degree of 'carefree 7. In this case, it was concluded that pilots have a
handling'. Hence, missing or poorly positioned tendency to fly more aggressively with no motion
instruments can also serve to spoil otherwise cueing due to the missing acceleration cues. There
satisfactory handling qualities. was also evidence that the increased aggression
gave rise to significant over-controlling. For the
'G' Seat. Sound & vibration cueing heave axis, it was found that lack of motion cues
gave rise to a PIO tendency in tasks that demanded
Sound and vibration cues not only serve to create accurate control of height.
a more realistic impression in a simulation, but
they are an important source of secondary cues for System latency
vehicle flight state. The amplitude and frequency
of sound and vibration in helicopters are generally As noted above, system latency is an important
strongly modulated by flight condition. When the factor in simulating handling qualities because it
pilot is flying eyes-out, such cues are used by the can limit the minimum phase delays and maximum
pilot as indicators of potential rotor bandwidth that can be achieved. Latency is
under/overspeed or over-torque situations, and associated with the computing time delays in
large normal 'g' loads. Hence, they can be a converting pilot's control demands into simulator
contributory factor in reducing pilot workload. cue demands and responses at the cockpit. In a
real aircraft, the achievable attitude bandwidth is
Motion cueing influenced by time lags associated with the
actuation, flight control and rotor system
The principal function of motion cueing is to characteristics. For aircraft with high speed rotors,
provide cues of the onset of acceleration through basic system time delays in the roll axis can be as
stimulation of the body's motion sensory low as 50ms, while helicopters with slower rotors,
mechanisms (Ref 16). From Ref 17, motion cues lags can be greater than 100ms. For an aircraft
are important for handling qualities simulations with full authority ACT, additional time delays can
concerning aircraft with marginal stability, high be expected, depending on the computational
gain tracking tasks and flight in disturbed efficiency of the system. The ADS-33 bandwidth
conditions. The AFS' provides both platform criteria reflect the predicted deterioration in
motion and pseudo-motion through a G-seat. handling qualities due to increasing time delays.
As noted in Section 3 of the paper, the current
The so-called 'G' seat provides the onset of AFS configuration is adequate for predicting the
normal 'g' cueing that the pilot would normally Level 1/2 handling qualities boundary for rate
feel through 'the seat of the pants'. In transient command helicopters; in these experiments, as
 14-17

with the UKAH trials, the time delays from the pilots conducted the evaluations in three mission
physical elements of the aircraft were effectively related tasks - the sidestep, lateral jinking and
simulated by the AFS system delays. quickhop - followed by a UKAH mini-mission,
assembled as a series of contiguous tasks. Flying
at moderate levels of aggressiveness, the pilots
6.3 Simulation model were able to complete all the tasks within adequate
performance standards for all configurations
Regarding the simulation model, there are several flown. Confidence in the results was increased by
known areas of flying qualities deficiency that are the low spread of pilot handling qualities ratings
currently not modelled adequately by the Helisim and the overall pilot impressions that the
model, and hence will not have been addressed in simulation was representative for the tasks flown.
the assessments. In the low speed MTEs, the Research also correlated well with off-line
influence of the main rotor wake on the tail rotor Toolbox and HELINV analyses.
and empennage can have a significant effect on the
pilot's ability to hold a steady flight condition and Simulator limitations encountered during the trial
perform accurate flight path control. The extent of did not seriously detract from the objectives nor
this effect varies considerably from type to type, compromise the conclusions drawn, but were
depending largely upon the development work noted and have provided the focus for technical
carried out by the manufacturer. A second discussion on fidelity requirements for handling
example concerns the effects of blade stall on investigations in this paper.
manoeuvrability at mid-high speeds. Performance
in tasks requiring accurate flight path control and
attitude tracking could be degraded depending on Specific points are as follows;
the extent of this problem. No conclusions can be
drawn about potential problems in these areas as a (i) the combination of visual, motion, audio and
result of the FDS flying qualities evaluations. Hi- tactile cues work synergistically to give the pilot a
fidelity modelling developments planned for the realistic impression of flying.
AFS over the next few years will provide the
capability to examine such areas. (ii) visual cues provide the most compelling inputs
to pilots controlling a helicopter's flight path and
attitude in the NoE. The field of view of the CGI
7 Conclusions display in the AFS helicopter cockpit was adequate
for most of the MTE flying, giving strong periphal
This paper has described the approach taken to the cueing laterally, but the range in the vertical FoV
assessment of the flying qualities of the UKAH limited the level of aggressiveness possible in pitch
contenders, with particular emphasis on the piloted manoeuvres. Textural detail, provided by the CGI
simulation element. The approach has photo-texturing, proved vital for good low speed
demonstrated the complementary capabilities of velocity and attitude cues and was supplemented
off-line evaluations, using the ADS-33 Handling by a range of artificial objects to assist pilots judge
Qualities toolbox and inverse simulation, and desired and adequate levels of task performance.
piloted tests to establish flying quality. The
background to the UKAH project and the overall (iii) motion cues provide the pilot with important
evaluation programme have been summarised, lead information, in general working to contain
followed by a resume of the DRA approach to control aggressiveness to realistic levels and also
flying qualities assessment. Simulation models of proving vital to the correct prediction of PIO
the contenders were assembled, within the DRA boundaries. In the UKAH trials, motion cues
Helisim framework, from data provided by the proved effective. At high levels of aggressiveness,
bidders. The simulation facility used for the control strategy became affected by spurious
piloted evaluations, the Advanced Flight cueing, particularly in the roll/sway axes. Work
Simulator, and the associated experimental design is in hand to improve the motion drive laws in this
have been described in some detail. Three test area.
14-18

(iv) the design of a piloted simulation trial can be G W Condon, "Simulation of

guided to great effect by off-line analysis as Nap-of-the-Earth Flight in Helicopters",
demonstrated in the DRA's approach to the UKAH AGARD CP 473, Computer Aided System
assessment. Conducting analysis of a helicopter's Design and Simulation, 1990
response characteristics ä la ADS-33 gives a clear
indication of a helicopter's strengths and J M Morgan, S W Baillie," ADS-33C
weaknesses. Effective use of the HQ Toolbox and Related Handling Qualities Research
HELINV inverse simulation has been made in this Performed Using the NRC Bell 205
context to focus attention during the piloted Airborne Simulator", Proceedings of the
evaluations. AHS/NASA Conference on Piloting
Vertical Flight Aircraft, San Fransisco,
In this, first of a kind, exercise in the UK, much Jan 1993
valuable information has been captured to assist in
the overall assessment of the competing weapon G D Padfield, M T Charlton, A M
systems and the DRA approach to flying qualities Kimberley, "Helicopter Flying Qualities in
evaluation has proved successful in providing Critical Mission Task Elements: Initial
insight into the mission capability of the UKAH Experience with the DRA(bedford) Large
contenders. Piloted simulation has demonstrated Motion Simulator", 18th European
its worth in highlighting handling characteristics Rotorcraft Forum, Paper No F2, Avignon,
relevant to the UKAH role. This assessment has 1992
also provided the opportunity to identify areas
where simulation fidelity needs improvement to G D Padfield et al, "Where does the
realise the full potential in supporting design, Workload go when Pilots Attack
evaluation and, ultimately, airworthiness Manoeuvres? An Analysis of Results from
certification. Flying Qualities Theory & Experiment",
20th European Rotorcraft Forum, Paper
No 83, Amsterdam, October 1994

References: G E Cooper, R P Harper, "The use of

Pilot Ratings in the Evaluation of Aircraft
Handling Qualities", NASA-TN-D5153,
Anon, "Handling Qualities Requirements 1969
for Military Rotorcraft", Aeronautical
Design Standard ADS33C, August 1989 10 G D Padfield, "Theoretical model of
Helicopter Flight Mechanics for
J Howitt, "A Toolbox for Evaluation of Application to Piloted Simulation", RAE
Flying Qualities based on ADS-33 TR 81048,1981
criteria", DRA Working Paper (in
preparation) 11 M T Charlton, G D Padfield, R I Horton,
"Helicopter Agility in Low Speed
D G Thomson, R Bradley, "The use of Manoeuvres", RAE TM FM 22, April
Inverse Simulation for Conceptual 1989
Design", 16th European Rotorcraft
Forum, Glasgow, Sept 1990 12 M T Charlton, S S Houston, "Flight Test
and Analysis Procedures for New
D L Key," A New Handling Qualities Handling Criteria", RAE TM FM 26,
Specification for US Military Rotorcraft", May 1989
Proceedings of International Conference,
'Helicopter Handling Qualities and Flight
Control', RAeSoc, London, 1988
 14-19

13 D Schimke et al, "The European ACT

Programme: Complementary use of
Ground based Simulation facilities and
Experimental Fly-by-Wire/Light
Helicopters", 18th European Rotorcraft
Forum, Avignon, France, 1992

14 Anon, "Design and Airworthiness

Standards for Service Aircraft", Defence
Standard 00-970

15 R Hoh, "Investigation of Outside Visual

Cues required for Low Speed and Hover",
AIAA-85-1808, 1985

16 B N Tomlinson, "Simulator Motion

Characteristics and Perceptual Fidelity",
AGARD CP 408 "Flight Simulation",
1985

17 AM Kimberley, "A Survey of Validation

Techniques for the assessment of
Helicopter Simulator Handling Fidelity",
DRA Working Paper FS(B) WP(92) 076,
February 1993

18 AD White, "G-Seat Heave Motion

Cueing for improved Handling in
Helicopter Simulators", RAE TM FM 33,
October 1989

19 AD White, J R Hall, B N Tomlinson,

"Initial Validation of an R & D Simulator
with Large Amplitude Motion", AGARD
FMP Symposium "Piloted Simulator
Effectiveness", October 1991
 16-1

NASA Ames Research Center

Mail Stop 211-2
Moffett Field, CA 94035-1000
USA

THE USE OF PILOTED SIMULATION FOR

CIVIL TILTROTOR
INTEGRATED COCKPIT DESIGN

William A. Decker, Rickey C. Simmons, George E. Tucker

NASA Ames Research Center

Moffett Field, California, USA

Ground-based flight simulators are receiving increased use in the

design of civil aircraft. In addition to traditional simulation roles in
support of cockpit control and display design, simulators are now used
to develop new flight procedures and to assist in airport design. This is
particularly true for the concept of a civil tiltrotor transport. This
presentation summarizes recent simulation activity at NASA's Ames
Research Center focused on the design requirements for the
introduction of tiltrotor aircraft as economic vertical flight transports.

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?'
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
16-2

Presentation Outline

• Introduction
• Experiment
-Facility
-Design/Conduct
• Results
-Controls
-Displays
- Procedures
-Infrastructure
• Conclusions

This presentation will begin by discussing the motivation and

background for the civil tiltrotor simulation investigations. This will be
followed by a description of the simulation facility and the experimental
design for a series of experiments. The results discussion will range
from traditional simulation roles of control and display development
through the new application for civil aircraft operation of terminal area
procedure development and airport design.
 16-3

CTR on Approach to SFV

The successful flight demonstration of the tiltrotor aircraft concept by
the XV-15 and its subsequent development for the U.S. military in the
V-22, Osprey, program suggest the potential for a civil tiltrotor transport
operating from "vertiports" located near transportation market centers.
16-4

Motivation for Civil Tiltrotor

Investigations

NASA-FAA-lndustry studies show potential civil

tiltrotor market.
Certification and operations criteria required for
this new class of powered-lift aircraft.
Define infrastructure requirements:
• Vertiport
• Air Traffic Control

Market studies have identified such a market as well as technical,

economic and political issues to be resolved.

Building upon an interim powered-lift aircraft certification criteria, both

the U.S. FAA and European certification authorities have begun a
review of those criteria as their application to an aircraft becomes more
real.

In the course of the aircraft design and operations criteria review,

design guidelines for the ground facilities (vertiports) and operating
procedures are also reviewed and commented upon.
 16-5

Tiltrotor Operating Challenge

£SfS?,». ~ POWER CONTROLS

CONTROLS AIRSPEED
ALTITUDE AlHbPEED

o
o
? 8-
X
: ^
a
i
1

\ i % /
R REQUIRED

LTM0DE
\|/m,7' /
1

HELICOPTER j ^ AIRPLANE MODE

MODE «*°
LU 2-
!
o
a.
" i i i i i
0 50 100 150 200 250
AIRSPEED- knots

The tiltrotor operating challenge is exemplified by the power-required

curve, shown here as a function of airspeed and flight mode. A tiltrotor
aircraft has a series of power curves based on nacelle angle (rotor
orientation). These extend from airplane mode through the tiltrotor's
unique "tilt" mode back to slow-speed operation in helicopter mode.
These curves show both the performance potential of an airplane that
can operate as a helicopter and an airplane and everything in between.
They also help illustrate the operating challenge presented to the pilot.
At higher speed forward flight, the pilot principally controls altitude or
flight path angle with pitch attitude. As the aircraft slows down, the pilot
shifts control strategy to use power as the principal flight path or altitude
control. Through a series of simulation investigations, we have seen
this "backside" of the power curve operation requires cockpit aids for
precise operations such as steep glide slope tracking.
16-6

Civil Tiltrotor Experiments

CTR-1: Conversion location

CTR-2: Raw data guidance on steep
approaches
CTR-3: Flight Director and Flight Path Vector
Displays on steep approaches
CTR-4: Terminal operating procedure
development
CTR-5: One Engine Inoperative Operations

To date, a series of five major piloted simulation experiments at NASA's

Ames Research Center have investigated various topics important to
cockpit design and terminal area operation of a civil tiltrotor transport.
The first experiment investigated where during the approach to convert
from airplane mode to helicopter mode-in level flight or on glide slope.
The second and third experiments investigated cockpit guidance
requirements for operation on steep instrument approaches. The most
recent experiments expanded the terminal area operations
investigations with consideration of missed approaches and one engine
inoperative concerns.
 16-7

Experiment Facility
Vertical Motion Aircraft
Simulator • Generic Tiltrotor
VIM NOMINAL OPERATIONAL MOTTO ■JLIMTTS
Altfl

vsimcAi.
OMWL
,*0
VELOCITY

,.
ACCEL
a*
Simulation Model
LATKRAL 130 • I«
LONGITUDINAL
ROLL
i-4
Ml
4
«0
10
• 40,000 lbs.
PTTCM
YAW
11»
1J4
•O
«4
ALL NUMBERS, UHfTS n, 4ag, • -c
Transport
• Attitude Command
SCAS
• Torque Command
and Limiting
System
• Automatic Flaps
• Nacelle Angle
Control Systems

All of the experiments were conducted in the Vertical Motion Simulator.

This facility provides an extremely large motion base platform, ideal for
terminal area investigations. In particular, this six degree-of-freedom
motion base features a large vertical travel and a choice of one large
horizontal axis to provide high bandwidth acceleration cuing. For
tiltrotor investigations, the longitudinal axis is emphasized for
accelerating and decelerating operations .

The aircraft was simulated with the Generic Tiltrotor Simulation model,
configured as a 40,000 pounds transport. Attitude command control
augmentation and a torque command and limiting system were
provided. Handling qualities evaluations from the earlier experiments
led to the development of automatic flap and nacelle angle control
systems that will be discussed later in this presentation.
16-8

CTR-2/3 cockpit
The experiments concentrated on handling qualities evaluations by the
pilot on controls with only the right cockpit seat provided in the
simulator. Shown here is the cockpit configuration used in the early
experiments as viewed by the pilot. The visual scene shows the aircraft
on short final approach to a vertiport pad located within an "urban
canyon." Those are buildings on either side. Cockpit controls include a
center stick, pedals, and a thrust control lever with geometry similar to
the original V-22 design. Located on the thrust control lever is a thumb-
wheel proportional rate control for nacelle movement. A c.r.t. display is
located in the center of the panel. Either an electronic analog of
conventional round-dial instruments or an advanced primary flight
display concept could be shown on this c.r.t.
 16-9

Flight Director Diagram

Nacelle Roll command

Pitch
movement attitude
command^A command
Raw
Power -—xn' glide
lever slope
command error

Raw localizer error

For conventional round-dial instruments, the pilot concentrates on the

attitude indicator during instrument operations. As shown here, the
raw-data glide slope and localizer error are displayed on the periphery
of the attitude indicator. Flight director functions could be displayed
with pitch and roll command bars, a power lever cue and a nacelle
angle movement prompt, making a four-cue flight director.
16-10

CTR-5 cockpit
The most recent CTR experiment used this cockpit arrangement. The
aircraft is shown on approach to a notional vertiport located south of
downtown San Francisco. We are at 1000 feet on a nine degree final
approach glide slope. Prompted by pilot commentary, the thrust control
lever has been changed to a near-vertical movement, similar to the XV-
15 control. On top of the thrust control lever head are several switches
including a central rocker switch used for a semi-automated nacelle
angle control that will be described more later. The panel now includes
two c.r.t. displays, inspired by modern fixed wing transport "glass
cockpit" layouts. The central attitude indicator on the primary flight
display could portray flight director command bars or flight path vector
guidance and control symbology under development at Ames.
 16-11

Flight Path Vector Display

FLP 60 UME 00.G ALT 13
ID 52 HDG LOC
MLS ON
Deceleration command

Ü05<g 50.
2550

Nacelle
\
Longitudinal Nacelle
corridor acceleration angle

The flight path vector display concept portrays the instantaneous flight
path-vertical and horizontal velocity vectors. The pilot controls the
aircraft in such a fashion as to overlay the aircraft flight path vector
symbol over a leader aircraft symbol. The leader symbol is driven
along a "perfect" flight path at a position three seconds ahead of the
own-ship. Both the the flight path vector and flight director displays
were described in a 1993 AHS paper.
16-12

Approach Task

Capture localizer prior to

converting
CCapture
Convert to approach confiauration glide slope
** "'-V >

Decelerate on glide slope " Land

S^ visually
Experiment Variables
* Glide slopes: 6, 9,15,25 degrees '////
* Approach speed adjusted for glide slope angle
* Calm or crosswinds and turbulence
* Clear or limited visibility

The principal evaluation task for the experiments involved fundamental

building blocks of terminal operations: a localizer capture, a level flight
conversion from airplane to helicopter mode plus a deceleration to the
final approach configuration, glide slope capture and tracking to a
landing decision point, and a vertical landing relying on visual
references. Glide slopes investigated ranged from 6 to 25 degrees,
with the approach speed adjusted for the glide slope angle. Task
meteorological conditions included clear-calm and degraded wind and
visual conditions.
 16-13

Steep Approach Operations

Approach Rule of Thumb:
"Descend at no more than 1000 fpm below
1000 feet above ground level."
Glide slope Airspeed Nacelle Angle
degrees knots degrees

6 80 80

9 55 80

15 35 90

25 20 90

Using an approach operations planning rule of thumb requiring a

descent rate lower than 1000 feet per minute on final approach, it
becomes a geometric exercise to calculate nominal approach
airspeeds. For steeper glide slopes-speculated as necessary for noise
minimization and obstacle clearance-very low approach airspeeds are
required. Approach nacelle angles were selected to provide nearly
deck-level operation. A tiltrotor aircraft has independent control of body
attitude at low airspeeds through use of the nacelle angle control.
16-14

Task Standards
Maintain flight within:
Desired Adequate
Level fliaht
Heading 5 deg 10 deg
Altitude 50 feet 100 feet
Airspeed 5 knots 10 knots

Glide slope tracking

Localizer 1/2 to 1 dot 1 to 2 dots
Elevation 1/2 to 1 dot 1 to 2 dots
Airspeed 5 knots 10 knots
1 dot = 1.25 deg. elevation error or 2.5 deg. azimuth error

Task performance standards must be defined for proper use of the

Cooper-Harper Handling Qualities Evaluation System. Adequate
performance was based on the FAA's Airline Transport Pilot test
standards. Desired performance was set at just half of the adequate.
Particularly important for tiltrotor terminal operations are altitude control
during a level-flight conversion and glide slope tracking on approach.
 16-15

Conversion Systems

Nacelle angle control

-manual beep
-detents
-semi-automated
Automated Flaps
-preliminary military schedules
-NASA developed schedule

Pilot comments led to the development of cockpit controls to aid the

flight mode conversion. For nacelle angle control, the earliest
experiments used the "manual beep" control where the pilot had to
keep his thumb on the nacelle control movement switch to continuously
move the nacelles. This is the system used in the XV-15 and the V-22.
Pilot comments suggested the use of a nacelle angle detent system
with preselected stops. The pilot could hold down the nacelle
movement switch, concentrating on. other flight tasks, knowing the the
nacelles would stop at the preselected angle. Pilot comments
suggested further development, resulting in our current pilot-initiated
semi-automated nacelle movement system. The pilot depresses a
switch once to move nacelles between the airplane mode at zero
degrees and tilt mode at 60 degrees at a slow movement rate. A
second switch depression moves nacelles aft to 80 degrees. A third
depression moves the nacelles aft to the hover stop, typically between
85 and 90 degrees. The operation is similar to flap selections in a fixed
wing transport.

As with the V-22, our CTR transport has automated flaps. Beginning
with simple flap schedules used for the original military program, we
eventually developed a flap schedule based on both airspeed and
nacelle position for the civil transport task. This NASA-developed
schedule intends to minimize trim pitch attitude changes.

The handling qualities results of these control developments is shown

next.
16-16

Conversion Handling Qualities

10
Average Range of
9H Rating X Ratings
8
Cooper-Harper INADEQUATE
Handling 7
Qualities
6-
Rating
5 ADEQUATE

4-

3-

2- SATISFACTORY

1 -

1 1 1 — 1
Flaps #1 Flaps #1 Flaps #2 Flaps #3
Beep Nacelle Nacelle Detenl Semi-Auto Nacelle Semi-Auto Nacelle
Raw Data Raw Data Flight Director Flight Director

Shown here are Cooper-Harper Handling Qualities ratings for four

arrangements of cockpit controls and displays. Beginning with the
simple systems of manual beep nacelle control and an abrupt flap
schedule, only adequate handling qualities were rated on average with
a wide dispersion of ratings depending on pilot tiltrotor experience and
weather conditions. The addition of the nacelle detent system provided
negligible improvement in quantitative ratings, but associated pilot
comments indicated the potential for further improvement. Provision of
the semi-automated nacelle control system, a gentler flap schedule and
a flight director improved the handling qualities ratings to marginally
satisfactory with a greatly reduced dispersion. Finally, our current
system with nacelle movement system refinements and the NASA flap
schedule provides solidly satisfactory ratings. The simple system
evaluations were provided by 6 to 8 pilots while the more recent
systems were evaluated by as many as 14 pilots.

Turning now to glide slope tracking.

 16-17

Gli(ie slope t raclring handling qualities

tn T
• 'Rawdatz * Mean X 1 Range of
9 - A Flignt dire ctor raling _ / ratings S
3-
«
3 T ~°
a
T —
f 7 ♦ -
O
6 \ i
1
Q.
«
\\ \ i S
S \
101 < % V
\
\\ I 1
T T
?3> 4
O.
<
O
° 3J
3 L \ i
Ü -(
. 1 . o
2
a
a
1 a
tn
0
Calm Turb Calm Turb ' Calm Turb Calm Turb
6" glide slope 9° glide slope 15° glide slope 25° glide slope

Shown here are Cooper-Harper Handling Qualities ratings as a function

of glide slope angle, weather condition (calm or crosswinds and
turbulence) and primary cockpit display. Using only raw data glide
slope error guidance (shown by the dots on the left of each group),
pilots rated this task as marginally satisfactory for only the shallowest
glide slope tested, 6 degrees. Steeper glide slopes which require slow
approach speeds on the "back-side" of the power curve drove the
ratings into the adequate range or worse. Provision of appropriate
cockpit guidance in the form of a flight director or flight path vector
display led to more uniform ratings with marginally satisfactory ratings
provided on up to a 15 degree glide slope. With their slower approach
speeds, crosswinds became more of a problem on the steeper glide
slopes. Beginning at 15 degrees, pilot commentary indicated cockpit
field of view had become an issue affecting ratings.
16-18

CTR-5 cockpit
on approach to SFV
Shown here is the pilot's view from a CTR on a 9 degree final
approach. Note that the glide slope aim point at the center of the
vertiport is low in the field of view. On a 15 degree glide slope, even
with deck level, we are perilously close to the 17 degree look-down
angle, typical of many cockpits. Minor variations in flight path caused
by atmospheric conditions easily lead to obscuring the aim point under
the nose of the aircraft. Certification evaluator-pilots commented
negatively on this and degraded their handling qualities ratings as a
result.

The field of view issue becomes even more important close to the
landing decision point.
 16-19

CTR-5 LDP @ OAV

Shown here is the view at a 200 feet high landing decision point for our
notional Oakland vertiport. The landing surface is fully in view, but note
that any approach lights are obscured below us-precisely at the point
in the approach where they are most needed--the transition from
instrument to visual flight references. Conversely, "approach" lights on
the departure side ARE an aid to landing-both for height control in
limited visibility (the horizon obscured) and for lateral line-up. Also
shown are a large vertiport symbol, clearly identifiable from the air, and
an alternate departure "runway" to better clear obstacles upon
departure.

In the low and slow flight condition at the end of the approach, one
engine inoperative concerns become essential to operation planning.
16-20

One Engine Inoperative Power

Power required
% HOGE

20- -O- Nacelle = 90°

-D" Nacelle . 80°
~$- Nacelle - 70°
-Ä- Nacelle = 60°
I I I
20 40 80 100 120

Airspeed, knots

Shown here is the power required versus airspeed performance for a

variety of nacelle angles from the most recent experiment. The power
is normalized at 100% by the power required for hover out of ground
effect. A base power was established for each engine by a
conservative requirement to provide for OEI service ceiling in the fixed
hover configuration. Above this base continuous engine rating, 15%
and 30% short-time contingency ratings (30 second ratings) were
established for the experiment. 15% contingency ratings are typical of
current design practice. A 30% rating is a considerable technological
stretch, but provides for hover-in-ground effect up to a 10 feet wheel
height for this aircraft.

Using these engine power ratings as experimental variables, a variety

of terminal operating strategies were investigated.
 16-21

Approach Procedures
Without OEI Hover capability
Landing Decision Point
50 knots
80 degrees nacelle angle
40 degrees flap angle
Landing completed with -provides OEI missed approach
visual reference using
"nacelle flare"
■ 600 feet -
With OEI Hover capability

_ Landing Decision Point

Continuous deceleration \ /"based on visibility only
on glide slope

«*- 200 f ee^

For an aircraft without OEI hover capability, an approach procedure

was developed to maintain safe OEI missed approach capability until a
landing commitment-in visual conditions-could be made. This
necessitated keeping 50 knots airspeed and an appropriate aircraft
configuration down to 200 feet altitude on a 9 degree approach. Upon
landing commitment, the pilot performs a "nacelle flare" (maintaining
pitch attitude while moving nacelles aft to to the hover position). This
slows the aircraft vertically and horizontally to make a vertical landing.
The operation is manageable by the pilot in a repeatable fashion
through use of the semi-automated nacelle movement system. This
procedure required guiding to the center of at least a 600 feet long
vertiport.

Provision of OEI hover capability permitted an alternate approach

strategy. The pilot could be guided to a continuous slow deceleration
on glide slope to achieve a vertical landing. This strategy received
better handling qualities ratings and would support a lower landing
decision point, but requires higher power in the event of an engine
failure.
16-22

Departure operations
Non-hover OEI capability
-Vertical departures
» In H-V avoid region: difficult to control rejected takeoff
-Airspeed over altitude departures
» Poor repeatable rejected takeoff landing performance
-Rolling takeoff (ESTO)
» Takeoff decision point (similar to fixed-wing)
» Low airspeed system required

Hover in ground effect OEI capability

-Add hovering back-up departure from confined spaces

Hover out of ground effect OEI capability

-Pure vertical operations possible

Turning to departure operations, a variety of strategies were

investigated. A vertical technique was quickly discarded since it clearly
encroached on a height-velocity avoid region of the flight envelope.
Using an "airspeed over altitude" technique-maintaining a low altitude
(under 20 feet wheel height) until a safe climb out speed is achieved,
pilots still had difficulty judging their touchdown for a rejected takeoff. A
better technique with limited OEI performance is an extremely short
take off technique. Using airspeed-based take-off decision points (30
knots was recommended), a take-off roll of only a few hundred feet
provided for safe, repeatable takeoffs and rejected takeoffs. Indeed,
the evaluating certification pilots found they could safely takeoff at a
lower airspeed, 20 knots, than the minimum required for out of ground
effect level flight. Note that both of these airspeeds are below the
speed for reliable use of typical pitot-static airspeed systems. Thus, the
need for a good low airspeed system was identified as critical to gaining
the most performance (payload) from the aircraft. It should also be
noted that the short-time contingency rating was needed for no more
than 30 seconds. In all instances, pilots could safely land or accelerate
to a safe flight condition within 30 seconds. The tiltrotor's thrust-
vectoring is a great help in this operation.

With in-ground-effect hover capability, a hovering back-up technique

could be used, permitting operation from confined spaces.
 16-23

Conclusion
Civil tiltrotor aircraft and infrastructure designs
are being developed with extensive use of ground-
based simulation.

Traditional Handling Qualities Design Use

• Controls
• Displays and Guidance

Developing Simulation Roles

• Procedure Development
• Vertiport Design

In conclusion, the civil tiltrotor simulation program has been used for
traditional handling qualities design purposes-control and display
design. Developing roles, new to civil aircraft design include procedure
development for new aircraft types and development of airport facility
designs for these aircraft.
 17-1

SIMULATION APPLICATIONS IN V-22 AND RAH-66 DESIGN AND DEVELOPMENT

R.E. Sheffield
L.E. Lalicker
Helicopters Division
Boeing Defense & Space Group
P.O.Box 16858 M/S P30-27
Philadelphia, Pennsylvania 19142-0858 USA

In addressing these technical challenges, one must not lose

1. SUMMARY sight of fundamental simulation objectives. As an airframe
The Boeing Company Defense & Space Group, Helicopters manufacturer and system integrator, organic simulation
Division, has increasingly applied constructive and virtual capabilities must be oriented toward influencing, in a positive
simulation in support of its aircraft developmental programs way, the development process. This process includes design,
which include the V-22 Osprey and the RAH-66 Comanche. systems integration, testing, fielding and training. For most
From their conceptual phases to present, methodologies and major military development programs, simulation is also used
mission simulation capabilities have evolved rapidly due to to (1) support the decision process at major milestones, and (2)
tremendous advances in computing power. By interacting
to help convince political leaders on the importance of a
directly with the systems engineering process, simulation has particular development program. To limit the scope of this
proven to be extremely valuable in optimizing rotorcraft paper, we will focus on the application of simulation as it
survivability and mission effectiveness. At the same time, pertains to design, systems integration and test support to the
simulation has proven to be an effective means of reducing risk RAH-66 and V-22 programs.
and costly flight testing. Recent focus has been on evolving
our simulation capabilities for participation in Distributive Simulation within the U.S. Department of Defense normally
Interactive Simulation (DIS). These capabilities and their has three major categories: constructive, virtual and live (see
acquired techniques provide potential opportunities to (1) better Figure 1). Constructive simulation includes wargames and
apply simulation to increase operational effectiveness of future math models used as analytical tools to predict system or
rotorcraft, and (2) facilitate customer involvement in the subsystem performance. Performance includes mission
developmental process. effectiveness, detectability and survivability. Virtual
simulation is a virtual reality environment for human
2. LIST OF SYMBOLS
interaction. Flight simulators fall into this category. Live
DIS distributed interactive simulation
simulation is exactly what the word implies. It includes
V-22 Osprey tiltrotor aircraft evaluation of hardware on instrumented ranges and larger
RAH-66 Comanche armed reconnaissance helicopter
exercises such as REFORGER (REdeployment of FORces to
TACAN tactical navigation device
GERmany).
FLIR forward looking infrared
Lab laboratory
AAA anti-aircraft artillery
SAM surface to air missile
ATCOM advanced tactical combat model

3. INTRODUCTION
The demanding environment and missions for which future
Equipment and Humans'
rotorcraft must be developed and operated are presenting Conducting Operations in
helicopter manufacturers with a truly formidable challenge. In Representative
and of itself, rotorcraft flight simulation is a difficult challenge Pilot-in-the-Loop Environments or
which is made more complex by the application of new Synthetic Environments Instrumented Ranges
technologies. The RAH-66 integrates advanced technologies in
the areas of weapons, sensors, signatures, cockpit displays and
other mission equipment that require major enhancements to
simulation tools previously used. The need to properly
represent the influences of terrain, weather and technical
sophistication of threat systems are a continuing requirement.
The simulation challenges for the V-22 run from the relatively
simple to the complicated. For example, proper visual
representation of the V-22's nacelles as either vertical or
horizontal depending on the flight mode is easy to implement.
A more complex challenge is the mathematical calculations
required to properly simulate aerodynamics associated with the FIGURE 1. Simulations, as Defined by the U.S.
helicopter-to-airplane transition state. Department of Defense, are Divided Into Three
Categories

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?",
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
17-2

4. DESCRIPTION OF RAH-66 AND V-22 The Marine Corps will use the aircraft to satisfy their medium
The RAH-66 Comanche is the U.S. Army's only new helicopter lift requirements. Because of some unique requirements for
development program. It is presently in the special operations, a variant of the V-22 is under development
demonstration/validation phase. The RAH-66 Comanche is a for the Special Operations Command. This variant is
twin-turbine, two-seat (tandem) helicopter that is designed to designated as "CV-22." Significant systems and key design
perform armed reconnaissance for cavalry units and attack features of the Osprey include hover flight (helicopter mode),
missions for the light divisions. For the heavy divisions, forward cruise flight at 275 knots (airplane mode), an advanced
Comanche will serve as the aeroscout for the AH-64 Apache. digital cockpit, composite wing and fuselage construction, and

• Centerpiece of U.S. Army's Aviation

Restructure Initiative
• High Survivability
• Improved Operational Effectiveness
• High Reliability

FIGURE 2. The RAH-66 Comanche Program Presents Several Opportunities to Apply Simulation to Meet its Key
System Design Requirements
Significant systems and key design features of the Comanche a cabin that accommodates 24 seats (see Figure 3). The
include low signatures (radar, visual, infrared, acoustic); Osprey is being developed under the auspices of the U.S. Naval
improved target acquisition sensors; increased maneuverability, Air Systems Command by a team of Boeing Defense & Space
agility and speed; increased survivability; significantly reduced Group, Helicopters Division and Bell Helicopter Textron.
operation and support costs; and reduced supportability
requirements that includes a simple remove-and-replace 5. VIRTUAL SIMULATION
maintenance concept (see Figure 2). The Comanche will Virtual simulation allows pilot interaction with emulated or
replace U.S. Army AH-1, OH-6 and OH-58 attack and actual flight controls and displays. Historically, flight
observation helicopters. First flight is scheduled for late 1995. simulation has lead the way in successful application of virtual
Team Comanche is led by Boeing Defense & Space Group's simulation. Flight simulators have been used extensively for
Helicopters Division and Sikorsky Aircraft Corporation, United pilot training, and they have been used to develop and optimize
Technologies. cockpits for efficient human interaction. Since these
applications are well known and understood by the aviation
The V-22 Osprey is a twin-turbine, vertical-lift, tiltrotor community, this paper will focus on a relatively new
transport. Originally conceived as a joint program for the U.S. application of flight simulation. This application deals with the
Marine Corps, Navy, Air Force, and Army with high dual use demonstration, validation and test phases of the development
potential as a civil transport. Initial production is now targeted process.
for the Marine Corps and U.S Special Operations Command.

• Satisfies U.S. Marine Corps'

Medium Lift Requirement
• Increased Survivability
• Improved Operational Effectiveness
• Highly Versatile Tiltrotor

FIGURE 3. The V-22 Osprey Program Introduces Several Unique Simulation Challenges
 17-3

The V-22 program implements a concept for digitally linking a be quickly evaluated. Moreover, feedback to the designers and
flight simulator to the laboratories that perform "bench testing" programmers can also include pilot qualitative evaluations. A
of hardware and software. Mission equipment hardware, flight massive amount of software is required by the V-22 and RAH-
displays, flight controls, mission computers, avionics, and data 66, so this concept provides a safe means of evaluating system
busses provided by various subcontractors are mounted in the functionality prior to actual flight.
simulator or connected to the digital network such that they are
V-22 benefits derived from this concept will be applied to the
stimulated by the flight simulator as it performs simulated
RAH-66 program. The V-22 program has incorporated a "triple
missions. A fiber optic reflective memory network called
lab tie-in" approach to its integration efforts. The triple lab tie-
SCRAMnet (Shared Common Random Access Memory
in digitally links three laboratories (the flight control
network) is used to connect the various functions being
laboratory, flight simulation laboratory, and mission
evaluated. This type of network allows different types of

ScramNET Ring FLIGHT CONTROL

MODELING
COMPUTER

AIRCRAFT
ACTUATOR
HARDWARE
FLIGHT
CONTROL
COMPUTERS

COCKPIT CONTROLS, SWITCHES, & BEZELS

1553 AVIONICS BUS

REMOTE REMOTE DISPLAY

TERMINAL TERMINAL HARDWARE VIDEO
SIMULATOR TACAN GENERATOR
FLIR
CHAFF DISPENSER
etc.

ScramNET Ring

FIGURE 4. The Triple Laboratory Tie-in Approach Allows Development Aircraft to be Virtually "Flown" in a Laboratory
Environment
computers to operate on the same network. Of course basic computer/avionics software laboratory) together. In this way,
functionality of a piece of hardware is checked prior to testing the V-22 can be virtually "flown" within a laboratory
it as part of the aircraft system. environment. Figure 4 illustrates the triple lab tie-in concept
used for the V-22 program. The RAH-66 program plans to use
The real benefit comes from evaluating functionality of these
the same concept with the possibility of expanding it for
sub-systems as they are being controlled by software that
additional laboratories, thereby accommodating target
resides in the mission computer and software interfaces
acquisition sensors and weapons systems.
between the mission computer and a particular subsystem. This
allows nonconfigured software or software patches to be tested Multiple lab tie-in evaluations have been used for cockpit
in a realistic environment. Such tests (using non-certified management system development relative to a dynamic flight
software) would not be acceptable for actual flight testing. scenario environment. Figure 5 illustrates the flow of a typical
Another benefit is the speed at which integration concepts can evaluation.

Test
Plan
Test Plan

t
Memorandum
Assignments Test Cards 8mm Data Tape

Test Test Test Test

Brief Setup Conduction Debrief

Laboratories
Engineering
Analysis *
Test
Report

FIGURE 5. A Typical V-22 Triple Laboratory Tie-in Evaluation Mimics the Procedure for a Flight Test Profile
17-4

The multiple lab tie-in reduces risk and, through reductions in program underwent a very thorough Verification, Validation
required flight test hours, there is potential for substantial cost and Accreditation (VV&A) process which was performed by U.
saving. Based on rough estimates from the V-22 program, 43 S. Army agencies. The greatest discriminator and challenge for
hours of flight-test reduction would completely pay for the selecting, developing and validating the models was in their
costs associated with the triple lab tie-in (flight testing is ability to simulate the effects of terrain on detection, lock-on
approximately 50 times more expensive that simulation). range and end-game engagements.
Multiple lab tie-in is also a versatile tool for troubleshooting
The primary application of these models has been to support
flight anomalies once flight testing begins. trade studies and parametric or sensitivity evaluations of
alternative design/integration concepts. Other uses include
6. CONSTRUCTIVE SIMULATION evaluation of candidate technologies for application to the
Both V-22 and RAH-66 programs have made rather extensive
helicopter, periodic mission effectiveness/survivability
use of constructive simulation. Because these developmental
assessments to measure change as the program progresses and
aircraft have different capabilities and missions, the
to develop, relative to the current fleet of Army helicopters,
methodologies and models used differ somewhat. However,
comparative data for use in convincing Government officials on
due to contractor teaming arrangements and the need to limit the merits of the program. Of course the latter is of little
the scope of this paper, only the RAH-66 constructive
significance to the design process. However, it is an important
simulation activities will be presented.
aspect of a development program during defense cut-backs.

MISSION PROFILE SCENARIOS THREAT DIGITAL DMA TERRAIN

AIRCRAFT SUSCEPTIBILITY AIRCRAFT SURVIVABILITY BATTLEFIELD EFFECTIVENESS

DETECT Radar Detection *RADGUNS Threat AAA ATCOM

"CASTFOREM Bn - Bde
RCM Radar Clutter
SEKE RF Propagation
*ESAMS Threat SAMs
SPAR RF Signal Processing *HELIPACATRACES Air-to-Air
*IMARS
LOCK-ON IRSeeker
"LOADLINE -f DIS MODELS
'ACQUIRE \- E-O Detection
"TARGAC -1 ATCOM 100 Combat
ITEMS Entities Per
GLINT Canopy Glint "ModSAF Model
VULNERABILITY
ilLITY I
Data
SUPPORTING MODELS
APART Atmospheric
*LOWTRAN7 Atmospheric
*EOSAEL Weather Data

*U.S. Government Sponsored Models

FIGURE 6. A Family of Constructive Simulation Models Are Being Used on the RAH-66 Program to Support Trade
Studies and Sensitivity Analyses

Past use of constructive simulation to support design and A general methodology presented by Figure 7 provides a top-
development of aircraft programs has been predominantly with down flow of major constructive simulation activities. A
one-on-one and one-on-few models. However, during the player-interactive battlefield simulation model was used, with
competitive proposal phase of the RAH-66 program, the value help from experienced Army pilots, to evaluate several Army
of simulating team tactics, techniques and procedures and the scenarios. These simulations served to develop the flight paths,
synergism of combined arms operations for mission observation positions, and firing positions used by the
effectiveness and survivability became very apparent. This helicopters relative to the position of threat systems defined in
prompted our use of battlefield simulation models to augment the Army scenarios. Using this data, the detectability and
the more traditional survivability models. survivability models were run to provide a more detailed
evaluation of detection or engagement outcomes. The player-
Figure 6 depicts the family of constructive simulation models interactive battlefield simulations also served to provide force
being used on the RAH-66 program and their purpose. Note movements and other data necessary to run the Army's
that many of these models were developed by U.S Government CASTFOREM (Combined Arms and Support Task Force
agencies. The other models were developed or purchased to Evaluation Model) simulation which is not player-interactive.
accommodate specific program needs. The models used on the
 17-5

Digital Terrain Scenarios Threats System Effectiveness Data

Mission-Threat Analysis I
User Assistance Battlefield Simulation I Exposure statistics
ATCOM I

¥ Flight Paths, Threat Locations

ah/sis I
Susceptibility Analysis
►- Probability of Being Detected (Pd)
DETECT * LOADLINE
LOCK-ON

y Point on path where detection occurs

Survivability Analysis
*- Probability of Survival, Given Detection
•RADGUNS *ESAMS

H Probability of kill data

Battlefield Effectiveness
Battlefield Survivability & Effectiveness
* CASTFOREM

* Government Developed/Sponsored

FIGURE 7. Methodology Depicting Flow of Major Constructive Simulation Activities

most cost-effective when several small changes or variations to
The RAH-66 program is using a Computer-Aided Design a system or subsystem need to be evaluated.
(CAD) tool called CATIA (Computer-Aided, Three-
dimensional Interactive Application) that is having a positive Most constructive simulation models do not run in a real-time
impact on both constructive and virtual simulation activities. mode which is necessary for virtual simulation. To meet the
On the constructive side, this design tool provides, in digital real-time requirement imposed by virtual simulation, faster
form, the data necessary to define the physical characteristics of computers and multiple processors may be required for
the aircraft for visual, infrared and radar signature predictions. constructive models.
However, the process of using this information is not a simple
A relatively new concept being exploited within the U.S.
as it may seem. Information from the computer-aided design
Defense Department is Distributed Interactive Simulation
tools must be processed into a form useable by other models.
(DIS). This concept allows dissimilar models distributed over a
The other models have their own unique input format, so
large geographical area to interact with each other to form a
numerous conversion programs must be developed to make use
common simulation. It includes connection of live,
of data developed by this computer-aided design tool.
constructive, virtual and training simulation under a common,
synthetic environment, (see figure 8). These simulations
7. VISION FOR THE FUTURE
communicate over local and wide area networks. Key elements
A natural progression of virtual and constructive simulation is
of this concept are:
to interface the two capabilities. For constructive simulation,
this would allow human factors (pilot-in-the-loop) issues to be
• No central computer for scheduling or conflict resolution.
analyzed and quantified when desired. In other words,
variations in cockpit design could be evaluated and quantified
• Each node is responsible for maintaining the state of one
in terms of their potential contributions for mission
or more simulation entities.
effectiveness and survivability. For virtual simulation,
comprehensive battlefield scenarios could be exercised to
• Common protocols for communicating entity
provide a more realistic simulation environment. Also, more
status/activity.
complex interactions between the simulator and other
• Receiving nodes determine what is perceived.
battlefield entities or units could be simulated and evaluated.
Presently, there are at least three ways to interface constructive
• Only entity state changes are communicated by each node.
and virtual simulation. They are shared memory, reflective
memory and common protocol networks.
• Each node uses dead reckoning to limit communications.
Although interfacing virtual and constructive simulation may
appear costly, there are potential cost savings. Virtual- Considerable progress has been made with the DIS concept.
Constructive simulations performed with the flight simulator The U.S. Department of Defense is exploiting it for three major
fully manned could be recorded for multiple unmanned reruns domains: Research, Development, and Acquisition (RD&A),
with a constructive model. This record-playback capability is which includes testing; Military Operations, which includes
useful for parametric and sensitivity analyses which can be planning and rehearsals; and Training, which includes
executed much quicker without the manned simulator. This is integrating live exercises with simulated activities. The DIS
17-6

concept has already influenced RAH-66 Comanche program 8. CONCLUSION

activities, and we are preparing to use it to support V-22 Osprey Simulation is playing a significant role in the development
program activities. phases of the RAH-66 and V-22 programs. We, in concert with
our team members, are applying both virtual and constructive
simulation to optimize the combat effectiveness and
survivability of the aircraft. Concurrently, we are recognizing
opportunities for better application of simulation, and we are
rapidly improving our capabilities in this area. As computer
processing speeds increase and networking capabilities
improve, simulation will play an even greater role in
development of rotorcraft. With future application of virtual
prototyping, the necessary synergism between design,
manufacturing simulation, and testing will provide new
challenges with many potential benefits. Time reduction,
efficiency, manpower reduction and risk reduction/abatement
are some of the more obvious benefits to the contractor. To the
military customer, it will be a better product at a more
affordable price.

FIGURE 8. Distributed Interactive Simulation Allows

Dissimilar Models Distributed Over a Large Geographical
Area to Interact with Each Other

Because of the computing power becoming available and

rapidly improving quality of simulation, a new concept called
"virtual prototyping" is emerging. The U.S. Department of
Defense defines the process as:

"The process of using a virtual prototype, in

lieu of a physical prototype, for test and
evaluation of specific characteristics of a
candidate design."
A virtual prototype is a computer-based, digital representation
of a system or subsystem. This concept involves utilization of a
collection of simulation tools to guide product design from
concept formulation to prototype with emphasis on integration
and optimization rather than hardware. It is envisioned that the
simulation tools used will bring warfighters, scientists,
engineers, testers and manufacturers together to develop and
evaluate, both qualitatively and quantitatively, concepts in
simulated environments. This would be accomplished by a
confederation of computers connected by local and wide area
networks. Once a system is optimized and it enters production,
the associated simulation tools would also be used for training,
doctrine and other purposes.
 18-1

Piloted Simulation in Rotorcraft Design & Development

EUROCOPTER Experience and Future Irends

La simulation pilotee pour la conception et le developpement

des appareils ä voilure tournantes

L'expenence d'EUROCOPTER et les tendances pour le futur

Philippe ROLLET
EUROCOPTER-FRANCE
13725 Marignane Cedex - France

STJMMARY RESUME

Piloted simulation has largely been used in Depuis 10 ans, la simulation piloted est largement
EUROCOPTER's developments for the last 10 years. utilis6e dans les developpements d'EUROCOPTER.
For the TIGER attack helicopter, piloted simulation Pour l'helicoptere d'attaque TIGRE, la simulation
was extensively used from the very beginning of the pilot6e fut employee de facpn intensive du d6but du
development to the prototypes' flights. Today, the d6veloppement jusqu'aux vols des prototypes.
development of the NH-90 helicopter requires a lot Actuellement, le developpement de l'helicoptere
of experimentations on the SPHERE simulator. NH-90 nScessite de nombreuses experimentations
Piloted simulation is also used for research related to sur le simulateur SPHERE. La simulation pilotee est
handling qualities, such as the ACT European 6galement utilis6e pour des recherches relatives aux
program, or to Man-Machine-Interfaces (MMIs). qualites de vol, tel que le programme European ACT,
The implementation of demonstrators, such as the ou aux interfaces homme-machine. La mise au point
FBW DAUPHIN, as well as the study of the de demonstrateurs, tel que le DAUPHIN CDVE,
EUROFAR Tilt-Rotor have also largely involved ainsi que l'Stude du convertible EUROFAR ont aussi
piloted simulations. A significant experience in largement fait appel ä la simulation pilotee. Une
testing procedures has been acquired. Environmental experience significative en matiere de procedures
and human factors appear to be of the greatest d'essais a €t€ acquise. Eenvironnement et les facteurs
importance. In the future, evolutions are planned to humains apparaissent fetre de la plus grande
further enhance the representativity of SPHERE importance. Dans le futur, des evolutions sont
simulator. The emergence of new design standards, privues pour ameliorer encore la representative du
such as ADS-33C, leads to envisage new applications simulateur SPHERE. Lemergence de nouvelles
of piloted simulation. normes de conception, telles que PADS-33C, permet
d'envisager de nouvelles applications de la simulation
piloted.

1.ABREVIATIONS IFR Instrument Flight Rules

LLLTV Low Light Level TV
ACT Active Control Technology MFD Multi Function Display
ADS Aircraft Design Standard MMI Man-Machine-Interface
AFS Advanced Flight Simulator MTE Mission Task Element
CDVE Commandes De Vol Electriques NFH NATO Frigate Helicopter
CELAR ment
Centre d'ELectronique de l'ARmement NLR
Nationaal Lucht—en—
CEV Centre d'Essais en Vol Ruimtevaartlaboratorium
CPU Central Processor Unit NOE Nap-Of-the-Earth
DRA Defence Research Agency PAH Panzer Abwehr Hubschrauber
FBW Fly-by-Wire PIO Pilot Induced Oscillations
FCC Flight Control Computer PFCS Primary Flight Control System
FLIR Forward Looking Infra-Red RFI Radio Frequency Indicator
GE General Electric RISC Reduced Instruction Set Computer
GFF Government Furnished Facility SIMCO SIMulation COckpit
HAC Helicoptere Anti-Char TTH Tactical Transport Helicopter
HAP Helicoptere Appui-Protection UCE Usable Cue Environment
HMS/D Helmet Mounted Sight/Display VMS Vertical Motion Simulator

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?",
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
18-2

2. INTRODUCTION cabine est transfer6e dans le dome de la DASA, d'un

diametre de 10 m et situ6 sur le mSme site, oü un
Bien qu'utilis6e depuis longtemps dans le domaine
g6nerateur d'image GE COMPUSCENE IV ä 6
des avions, ce n'est qu'au debut des ann6es 80 que la
canaux de projection permet d'obtenir une image du
simulation pilotee s'est imposee en tant qu'outil
paysage avec un champ de vision d'environ 140°(H) x
d'aide ä la conception et au developpement des
110°(V)(Fig.2&3).
appareils ä voilures tournantes. Cet avenement a
result^ de la conjuncture de plusieurs facteurs Le comportement de l'helicoptere et sa reponse aux
determinants. D'une part, l'accroissement commandes sont simulSs par le modele gen6rique de
considerable de la puissance des calculateurs m6canique du vol GENSIM developp6 par
num£riques a rendu possible la simulation en temps EUROCOPTER-DEUTSCHLAND. Le calcul des
reel avec des modeles rotor gen£riques non—lineaires efforts aerodynamiques sur la cellule utilise ä la fois
permettant obtenir une bonne representative dans des donn6es de soufflerie et des formules anah/tiques.
l'ensemble du domaine de vol. Dans le meme temps, Les efforts rotor sont calculus par un modele de type
les nouveaux programmes militaires se sont Orientes element de pale avec degres de liberty en battement et
vers le developpement de machines tres sophistiqu£es en rotation. Les moteurs MTR-390 sont 6galement
congues en tant que Systeme d'Armes et pour mod61is6s, ce qui permet de simuler les variations
lesquelles il devenait indispensable de recourir ä la transitoires de regime rotor lors des manoeuvres.
simulation des le stade de la conception. Enfin, Pour augmenter encore le realisme de la simulation,
l'emergence des Commandes De Vol Electriques on peut adjoindre au modele de mecanique du vol un
(CDVE) a fait apparaitre l'interet d'Studier en modele de train d'atterrissage, ainsi qu'un modele
simulateur, puis en vol, de nouvelles lois de simplifie de bruit interne. Un calculateur HARRIS
commande visant ä require au maximum la charge de NIGHTHAWK est utilise' pour le calcul et la gestion
travail de pilotage. des modeles de simulation. Le temps de cycle varie de
30 ä 40 millisecondes, suivant le degrd de complexity
Dans ce contexte, EUROCOPTER a et6 amene ä choisi pour le modele de mecanique du vol (ref. 2).
concevoir et ä mettre en place des moyens de
simulation piloted, mat6riels et logiciels, qui sont Jusqu'ä aujourd'hui, Pactivit6 SIMCO reprgsente
actuellement utilises avec profit pour le plus de 650 heures de simulation effectuSes par les
developpement des nouvelles machines, telles que le pilotes de l'Industrie et des Services Officiels Clients.
TIGRE et le NH-90.
Les simulateurs CELAR et CEV
Cette conf6rence präsente les principaux aspects de
l'experience acquise par EUROCOPTER en matiere Les simulateurs du CELAR (Centre d'Electronique
de simulation piloted, ainsi que les evolutions de l'Armement) et du CEV (Centre d'Essais en Vol)
envisages pour le futur. sont tous deux des installations GFF ("Government
Furnished Facility") appartenant ä l'Etat Francais et
3. LE PROGRAMME "TIGRE" situ£es respectivement ä Bruz, pres de Rennes, et ä
Lhelicoptere d'attaque TIGRE est le premier Istres. Ces installations sont, d'un point de vue
programme d'EUROCOPTER pour lequel la materiel, tres semblables et sont utilis£es
simulation pilotee a 6t6 utilisee de facon intensive des essentiellement pour le developpement de la version
le debut du developpement. Trois installations Appui-Protection du TIGRE (HAP) (r6f. 3).
d'essais ont 6te necessaires pour mener ä bien les
activites de simulation: SIM CO et les simulateurs du Le simulateur du CEV est utilise pour les etudes
CELAR et du CEV (centres d'essais etatiques) . d'interface Equipage—Systeme specifiques ä cette
version telles que la coordination entre le pilote et le
Le simulateur SIMCO tireur et Pintegration du Systeme d'arme HAP au
Systeme de base. De plus, de par sa proximity de
Le simulateur SIMCO ("SIMulation COckpit") est Marignane, ce simulateur permet 6galement
install ä Ottobrunn sur le site d'EUROCOPTER d'assurer un soutien aux essais en vol des prototypes.
DEUTSCHLAND. II est utilis6 pour le
developpement et la validation des fonctions Le simulateur du CELAR est utilise pour le
d'interface homme-machine (MMI) communes aux developpement des systemes de tir canon, missiles
trois versions du TIGRE: HAP, HAC and PAH-2 Air—Air, et roquettes de l'HAP.
(ref.l).
Les simulateurs du CELAR et du CEV sont dquipes
La cabine de simulation est une replique fidele du de 2 cabines representatives d'un point de vue
cockpit en tandem du TIGRE (Fig. 1). Lavionique et ergonomique des fonctions MMI ä evaluer et en
les amenagements internes sont tres repr6sentatifs et accord avec les evolutions successives de definition.
certains 6quipements sont ceux de l'appareil r£el Les cabines "Pilote" et "Tireur" sont separges et
(commandes de vol completes, HMS/Ds, panneaux de chacune comprend son propre Systeme d'image de
commande, RFIs, et Systeme d'alarmes). synthese pour la vision du monde exterieur. La cabine
La cabine peut fonctionner de facon autonome pour du pilote est 6quip6e d'un systeme hydraulique
les essais oü la vision du monde exterieur n'est pas restituant les efforts de commandes, d'un g6n£rateur
indispensable. Lorsque les essais le n6cessitent, la d'environnement sonore et d'un siege vibrant.
 18-3

La vision exterieure et les images senseurs sont dont les dimensions, la position des commandes et le
obtenues par des gene>ateurs d'image SOGITEC champ de vision peuvent etre considered comme
GI-10KSOGITEC. Le Systeme de projection ä trois repr6sentatifs d'un appareil de type NH-90, bien que
canaux dans un dome d'environ 10 m de diametre pas tout ä fait conformes ä la definition actuelle de cet
permettent au pilote une vision du monde exteneur helicoptere.
sur un champ de 200 °(H) x 55 "(V). Les images vues
Seule la place droite est utilised pour les simulations
par les differents senseurs de tir sont envoyees au PFCS. Un Systeme hydraulique de restitution
tireur dans Poculaire du viseur et (ou) sur les ecrans
d'efforts MAC FADDEN permet d'obtenir des
multi—functions (MFDs). La base de donnees du
caract6ristiques d'effort et de deplacement des
Systeme de vision a 6t6 spScialement developpSe pour
commandes conformes ä celles visees pour l'appareil
le vol helicoptere ä proximite du sol et des obstacles r£el. Sur la planche de bord, des ecrans graphiques
(vol NOE). Eimage d'un autre helicoptere en vol
permettent d'afficher les parametres de vol
(escorte ou cible) peut egalement Stre pr£sent6e.
necessaires pour le pilotage de base et pour realiser
4. LE PROGRAMME NH-90 les elements de mission. La symbologie utilisee est de
type expeiimentale et n'est pas completement
Le developpement de cet helicoptere, qui a d6but6 en representative de celle pr6vue pour l'appareil reel.
91 et dont le premier vol est prevu fin 95, engendre un
volume tres important de simulations. D'une part, le Le cockpit experimental est plac6 sur une base fixe
developpement actuellement en cours des dans le dome SPHERE de 8 m de diametre. Un
Commandes De Vol Electriques (CDVE) impose une Systeme de projection d'image de synthese ä 3 canaux
utilisation intensive de la simulation piloted pour SOGITEC GI 10K permet d'obtenir une vision du
l'etude et la definition des lois de commandes, ainsi monde exterieur sur un champ de 180°(H) x 80°(V)
que pour la validation avant vol du fonctionnement de (Fig. 5 & 6). La base de donnees principale, de
ce Systeme. D'autre part, les performances de mission dimensions 14 x 10 Km et hautement d6taill6e, a 6t6
elevees requises pour cet appareil, que ce soit en specialement concue par SOGITEC pour le vol
version terrestre (TTH) ou marine (NFH), tactique helicoptere et comporte de nombreux
conduisent ä des Etudes de definition obstacles naturels et artificiels. Une autre base de
particulierement poussees des interfaces donn£es, plus vaste mais moins d6taill6e, comporte
Equipage—Systeme (MMIs) et qui devront Stre une 6tendue maritime avec un navire, de type frigate,
validees par de nombreuses evaluations en pouvant etre anim£ de mouvements en relation avec
simulateur. Le volume des evaluations necessaires en l'etat de la mer. Ceci permettra, pour la version navale
simulateur est par ailleurs accru du fait de la diversite NFH, d'analyser en simulateur certains aspects
des missions terrestres et navales, resultant eile meme relatifs aux manoeuvres d'atterrissage sur pont de
du nombre important d'Armies—Clientes (4 navire.
Marines, 3 Arm6es de Terre). Depuis le debut du Le comportement de l'helicoptere et sa reponse aux
programme NH—90, plus de 1000 heures, dont commandes sont simul6s par le modele gSnSrique de
environ 1/3 avec des pilotes, ont €t€ effectu6es en mecanique du vol S89 developpS par
simulateur pour la definition et le developpement du EUROCOPTER-FRANCE. Des essais effectues au
PFCS (Primary Flight Control System). Les NLR (Nationaal Lucht en Ruimtevaartlaboratorium,
simulations pilot6es relatives aux MMIs doivent Netherlands) avec des maquettes de soufflerie
dSbuter en 96. motorisSes et non—motorisees ont permis de recaler
Les simulations pilotees du programme NH-90 le modele de calcul des efforts a6rodynamiques sur la
utüisent essentiellement le moyen d'essais SPHERE cellule. Les efforts rotor sont calcul6s par un modele
implants sur le site d'EUROCOPTER-FRANCE ä de type analytique avec degr6s de liberty en
Marignane (Fig. 4). battement et en rotation. Une version avec modele
type "616ment de pale" est Sgalement disponible. Les
Les simulations PFCS moteurs ne sont pas pour le moment modelisSs, mais
un modele dynamique des moteurs RTM 322 sera
Ces simulations permettent de developper les lois de integrS ä la simulation courant 95.
commande (ou lois de pilotage) du PFCS de fagon ä
satisfaire les objectifs retenus en matiere de qualit6s Le PFCS est simulS par un modele utilisant la meme
de vol. Ces objectifs sont denvös des criteres de specification de code que celle prevue pour les
1'ADS—33C (rei. 4) en tenant compte de la nature des calculateurs FCC r6els (Flight Control Computers).
missions des versions TTH et NFH. La simulation Ceci permet de garantir Pidentite' des lois de
piIot6e est utilis6e d'une part pour valider en temps commande entre l'appareil simulg et l'appareil r6el.
r6el le fonctionnement des lois de commandes, et Les reponses des servo-commandes et des capteurs
d'autre part pour s'assurer que la structure et le sont simul6es par des modeles en accord avec les
r6glage de ces lois permettent effectivement d'obtenir specifications de ces 6quipements (bandes passantes,
une charge de travail de pilotage faible, done des retards) et seront recalls ultgrieurement, si
qualites de vol de Niveau 1, pour chaque element de n6cessaire, en fonction des performances reelles
mission (MTE). constatees.
Le cockpit utilis6 est une cabine de pilotage Eensemble des calculs relatifs aux modeles de
cöte-ä—cote multi—usages (cockpit experimental) simulation sont effectuSs par des CPUs de
18-4

technologie RISC implantees sur des cartes VME 32 directement interpr6tables, permettant aux clients de
bits. Le temps de cycle global est de 40 milli-secondes suivre au plus pres et de valider les differentes Stapes
(25 Hz). Le retard de l'image est entierement des developpements. EUROCOPTER a done 6t€
compense' par un algorithme d'extrapolation (r£f. 5), amen6 ä definir et ä organiser des sessions de
ce qui reduit au minimum le delai entre une action simulations sp6cifiques pour presenter aux clients les
pilote et la reponse visuelle du modele. choix techniques en matiere d'interfaces
homme-machine (MMI) et de systemes de
5. RECHERCHE APPLIQUEE commandes de vol.
En parallele aux activites liees aux developpement Les clients sont en general fortement impliqu6s dans
d'helicopteres nouveaux, la simulation pilot6e est les processus de definition des interfaces
egalement largement utilised par EUROCOPTER homme-machine. Dans le cadre du programme
dans le cadre de programmes de recherche appliqu6e TIGRE, les pilotes-clients ont participe' ä de
ou du developpement de d6monstrateurs nombreuses sessions de simulation pilotee en vue
technologiques. d'optimiser et de valider la definition des interfaces
Le programme ACT (Active Control Technology), de pilotage, notamment en ce qui concerne les
realis6 en cooperation entre la France, l'Allemagne et symbologies de planche de bord. Un processus
la Grande Bretagne, a pour objectif d'etudier les similaire devaluation en simulateur est prevu pour le
nouveaux concepts rendus possibles par NH-90äpartirde96.
l'introduction des commandes de vol 61ectriques (ou En matiere de d6veloppement de systemes de
optiques) telles que les lois de commandes evoluees et commandes de vol, la participation des clients aux
les mini-manches (r6f. 6). Les simulations pilot6es simulations pilotees doit etre organised de iaqon ä
relatives ä ce programme ont 6t6 effectu6es sur obtenir des resultats coh6rents et non biais6s malgr6
differents moyens d'essais mis ä disposition par Pinfluence deiavorable de Penvironnement
PIndustrie (simulateur SPHERE ä Marignane, simulateur sur les evaluations de qualitds de vol.
simulateur DASA d'Ottobrunn) et par les organismes
de recherches (simulateur AFS du DRA Bedford). Dans ce but, les pilotes repr6sentant les Services
Officiels clients sont invit6s ä effectuer, comme pour
EUROCOPTER effectue egalement des recherches des essais de recherche, les evaluations sur des
en matiere d'interface homme—machine. La manoeuvres types pr6alablement d6finies et ä
simulation piloted est par exemple utilisee pour quantifier la charge de travail suivant PSchelle de
evaluer Pimpact de nouvelles symbologies de planche Cooper-Harper (r£f. 10). Les notes sont ensuite
de bord sur la charge de travail de l'Squipage. moyenn6es afin obtenir une estimation du niveau de
La mise au point des lois de pilotage de Ph61icoptere qualit6 de vol pour chacune des manoeuvres.
DAUPHIN 6001, appareil demonstrates de Le recours aux manoeuvres types est essentiel pour la
Commandes de Vol Electriques (r6f. 7 & 8), a fait coherence des r6sultats car e'est la seule facpn de
largement appel ä la simulation pilot6e, dans un garantir que les tous pilotes effectuent les memes
premier temps sur les installations 6tatiques (CEV, täches de pilotage. L'expenence passee
CELAR) puis sur SPHERE. Le DAUPHIN 6001 est d'EUROCOPTER a montrS que des evaluations
actuellement utilise dans le cadre du programme ACT libres sans programme pr6cis conduisaient ä des
pour revaluation en vol de nouvelles lois de pilotage. resultats tres disparates et difficilement exploitables,
Des lois de pilotage adapters ä un appareil de type chaque pilote ayant tendance ä se focaliser sur un
convertible ä rotors basculants (Tut-Rotor) ont aspect particulier de la simulation.
egalement ete etudiees en simulateur dans le cadre du De meme, Putilisation de Pechelle de
programme europ6en EUROFAR (r6f. 9). En attente Cooper-Harper oblige les pilotes ä rationaliser leurs
du lancement de la phase de definition d'un appareil jugements, ivitant ainsi d'introduire trop de
demonstrateur, les simulations pilot£es sont subjectivit6 dans les appreciations. Pour que les
actuellement orient6es vers Petude des trajectoires resultats soient valides, Ü est egalement important
d'approches ä forte pente. que les pilotes se referent au questionnaire de
Pechelle avant d'6mettre une note. Beaucoup de
6. PROCEDURES D'ESSAIS ET FACTEURS pilotes ont malheureusement, tendance ä vouloir
HUMAINS court-circuiter cette 6tape. Pour 6viter de telles
Lorsque la simulation pilotee est utilisee dans le cadre erreurs de proc6dure, une table de Cooper-Harper
de programmes Statiques majeurs, tels que le TIGRE est systematiquement pr6sent£e au pilote dans le
ou le NH-90, diff6rents types d'essais doivent etre cockpit ä la fin de chaque manoeuvre-type.
considers. Les effets induits par Penvironnement simulateur
En effet, bien que la simulation soit essentiellement doivent egalement £tre consid6r6s lors de Panalyse
un outil d'6tude ä la disposition des ing6nieurs et des r6sultats, surtout compte tenu du fait que certains
pilotes de PIndustrie, les Services Officiels clients pilotes des Services Officiels clients peuvent n'avoir
demandent toujours ä Stre impliqu6s dans ces que peu d'expenence en simulateur, ou bien une
activites. Ceci est principalement H6 au fait que la experience limitee aux simulateurs d'entrainement au
simulation pilot6e donne des resultats concrets et vol IFR.
 18-5

Lorsque des difficulty de pilotage sont rencontrdes simulateur ä base fixe mais equipe d'un Systeme de
pendant la realisation d'une manoeuvre, la premiere vision performant. Ce choix a ete egalement en partie
tendance des pilotes est souvent de conclure qu'il y a justifie par les bons resultats obtenus avec le
un probleme de contrölabilite du modele alors que simulateur ä base fixe de la DASA, notamment lors
ceci est parfois la consequence d'un manque des simulations SIMCO associees au programme
d'adaptation ä Penvironnement de simulation. En TIGRE.
particulier, l'experience a montre que ceci pouvait
conduire certains pilotes ä se gdneier artificiellement, Visualisation du monde exterieur
par leurs propres actions aux commandes, une charge La qualite de l'environnement visuel depend de
de travail excessivement 61ev6e m6me dans des cas de plusieurs facteurs. Letendue du champ de vision est
vol oü le modele est repute tres stable et facile ä determinante pour la perception des assiettes et des
contröler. Lexp6rimentateur doit alors trouver les vitesses et ceci d'autant plus que la vitesse
justes arguments pour amener le pilote eValuateur ä d'avancement est faible, ce qui est le cas typique d'un
reconnaitre que ses difficult6s sont, en partie ou en heiicoptere en evolutions autour du stationnaire ou
totality, dues ä un manque d'adaptation au en vol tactique (NOE). Lors du developpement du
simulateur. simulateur SPHERE, le champ de vision horizontal a
D'un point de vue general, l'experience a montre que ete, comme prevu, etendu d'abord de 60° ä 120°, puis
les pilotes ayant l'habitude de voler sur simulateur, de 120° ä 180°, chaeune des etapes correspondant ä
meme sur une autre installation, r6alisent les l'adjonction d'un canal de projection supplementaire.
meilleures performances sur les manoeuvres-types U fut alors interessant de constater, pour un meme
et sont egalement plus efficaces pour faire la modele d'heiicoptere, l'ameiioration de la pilotabilite
distinction entre les problemes lies ä l'environnement en stationnaire au fur et ä mesure de l'augmentation
et les vSritables defauts de qualitis de vol du modele du champ de vision horizontal. Lameiioration, bien
de Simulation. Le fait de possSder ou non une qu'ayant ete laplus significative entre 60° et 120°, est
formation de pilote d'essais, et en particulier aussi apparue nettement perceptible lors du passage
l'habitude d'utiliser l'echelle de Cooper-Harper, de 120° ä 180°, ce qui demontre bien l'importance de
apparait Egalement determinante mais sans doute ä la vision peripherique pour le vol ä basse vitesse.
un degr6 moindre que l'experience en simulateur. Les retards induits par le temps de calcul de l'image
7. ENVIRONNEMENT ont aussi beaueoup d'importance et doivent Stre
autant que possible minimises. Sur simulateur ä base
Lenvironnement du simulateur doit fournir les fixe, les retards peuvent pratiquement etre
sensations n6cessaires au pilotage, en complement compenses en totalite au moyen d'un algorithme
des informations affichles sur laplanche de bord. Son extrapolateur tel que celui decrit dans la reference 5.
niveau de representativite doit etre adapts ä la nature Une teile approche n'est pas toujours possible avec
des essais ä effectuer, le but etant d'eviter les exces de une base mobile oü, pour assurer la coherence entre
details inutiles et couteux. l'image et le mouvement, le retard image doit rester
comparable ä celui de la plate-forme (r6f. 11).
Base fixe ou base mobile
Le niveau de detail de la base de donnees a egalement
Un des premiers problemes auquel a ete confronte une influence importante sur les sensations visuelles
EUROCOPTER lors du developpement du lors de la realisation de manoeuvres de precision pres
simulateur SPHERE etäit de determiner s'il etait du sol. II est cependant apparu que la nature et la
necessaire ou non de restituer les accelerations ä disposition des obstacles representes etaient bien plus
court terme au moyen d'une base mobile. Dans ce but, determinants que la densite totale de details figurant
une equipe composee d'un ingenieur et d'un pilote a dans l'image. Ainsi une zone tres riche en details n'est
eu pour mission de visiter et d'evaluer differentes pas obligatoirement celle qui procure les meilleures
installations existantes de simulation heiicoptere, sensations visuelles pour le pilotage. Cet aspect
avec et sans base mobile. La conclusion de cette pourrait etre consider par les concepteurs de bases
mission a ete que, avec les debattements de de donnees afin de limiter la puissance de calcul
plate-forme relativement faibles habituellement necessaire pour la generation des images.
utilises, les accelerations restituees par le mouvement
sont insuffisantes, ou trop rapidement evanouies, La base de donnee VOLTAC utilisee dans le
pour ameiiorer de fagon significative la precision de simulateur SPHERE a ete specialement concue pour
pilotage lors des evolutions ä basse vitesse pres du sol le vol heiicoptere pres du sol (NOE). En complement
et qu'il est plus important de priviiegier la qualite du des details integres au paysage (arbres, ponts,
systäme de vision. Les plate—formes ä grand maisons, ...etc.), des obstacles artificiels ont ete places
debattement, telles que le VMS (NASA) et le AFS en differents endroits pour servir de reperes visuels
(DRA), semblent actuellement etre les seules lors de la realisation de certaines manoeuvres-types
capables de restituer de facpn satisfaisante les (Fig. 7). Eimportance de la disposition des obstacles a
accelerations mais restent de part leur coüt et leur ete clairement mise en evidence lors d'essais avec le
complexite de mise en oeuvre reserves aux grands modele NH-90 depourvu de tout dispositif de
centres de recherche etatiques. EUROCOPTER s'est stabilisation (appareU nu), une nette tendance au
done logiquement Oriente vers le choix d'un pompage pilote (PIO) apparaissant dans certains
18-6

endroits de la base de donnees, ou lorsque Oriente de la puissance de calcul, il serait possible d'ameiiorer
d'une certaine facpn par rapport ä un obstacle. D'une la fluidity de l'image lors des manoeuvres de rotation
maniere generate, les defauts d'environnement visuel rapides, notamment en roulis et en lacet, en
ont d'autant plus d'influence sur la pilotabilite que le multipliant par 2 la frequence de generation d'image
modele hilicoptere est instable (appareil nu ou avec (50 Hz au lieu de 25 Hz). Cependant, le beneTice
stabilisation coup6e). Ceci est tout ä fait en accord apporte par une teile modification serait faible
avec la philosophie de PADS-33C qui etablit une compare ä son coüt. EUROCOPTER prefere done
correlation entre le niveau d'environnement visuel attendre la disponibilite des gdnerateurs d'images de
(UCE) et le niveau de stabilisation n£cessaire pour nouvelle generation pour faire 6voluer le Systeme de
obtenir des qualites de vol de niveau 1. vision de SPHERE. Ces nouveaux g6n6rateurs
d'image devraient permettre, grace aux textures plus
Sons et vibrations fines, d'ameiiorer de facpn significative la perception
des distances par rapport aux obstacles et au sol, cette
La restitution de Penvironnement sonore et des
perception etant actuellement jugee insuffisante par
vibrations, peut egalement apporter des sensations
utiles au pilote lors de l'execution de certaines les pilotes.
manoeuvres. Ceci a €t€ utilise avec profit dans le Des enrichissements de la base de donnees sont aussi
cadre des simulations TIGRE. prevus en fonction de revolution des besoins des
utilisateurs. Actuellement, EUROCOPTER
Le cockpit du simulateur SIMCO est restitue developpe un modele d'animation de mouvement de
l'ambiance sonore dans les ecouteurs du casque du navire en fonction de l'etat de la mer, ceci en vue de
pilote et les vibrations par l'interm6diaire d'un permettre des simulations d'appontage de la version
coussin de siege gonflable. Les sensations ainsi marine du NH-90 (NFH).
obtenues sont apparues utiles pour evaluer certaines
La representativite du modele temps reel de
conditions de vol, en particulier les manoeuvres ä fort
mecanique du vol heiicoptere sera egalement
facteur de charge, les manoeuvres de flare, les
ameiioree dans le futur par l'introduction de
descentes rapides et les autorotations.
modeiisations plus completes des rotors
Les cockpit des simulateurs CEV et CELAR sont (mouvements de trainee et de torsion des pales) et des
interactions aerodynamiques entre les rotors et la
egalement 6quip6s d'un tel Systeme, avec en plus la
capacite de restituer le bruit produit par le tir des cellule.
armements. Dans le cadre du programme NH-90, un cockpit de
simulation totalement representatif de la definition
8. EVOLUTIONS FUTURES reelle de l'appareil seraprochainement mis en oeuvre
A partir de cette experience acquise, sur SPHERE pour realiser les simulations MMI
EUROCOPTER a identify les Evolutions relatives au Systeme de base et ä la version transport
souhaitables en vue d'ameiiorer les performances et tactique (TTH).
la repr6sentativit6 de ses moyens de simulation, et en En ce qui concerne les procedures d'essais, il est
particulier celles du simulateur SPHERE. vraisemblable que dans le futur, les aspects MMI et
H n'est pour le moment pas envisage d'evoluer vers un Qualites de Vol tendront de plus en plus ä etre traites
Systeme ä base mobile. Les nombreuses simultanement au cours des developpements.
experimentations effectuees sur SPHERE depuis sa Comme le precise l'ADS-33C, la charge de travail de
mise en service en 92 semblent confirmer que, pour ce pilotage depend non seulement de la reponse de
qui est des simulations relatives ä la conception et aux l'appareil aux commandes, mais egalement de la
developpement d'appareils nouveaux, une qualite des informations visuelles dont dispose le
installation ä base fixe est süffisante dans la mesure oü pilote. Les developpements en simulateur de
celle—ci est principalement utilisee par des pilotes nouvelles lois de commandes optimisees pour le vol
d'EUROCOPTER accoutumes ä ce moyen d'essais. en conditions d'environnement visuel degrade
Seul l'absence d'acceierations verticales apparatt devront done s'effectuer en parallele avec ceux des
quelque peu penalisant, en particulier lors des images et symbologies associees aux systemes d'aide ä
manoeuvres ä tres basse vitesse pres du sol et lors des la vision (FLIR, LLLTV). Par ailleurs, l'experience a
atterrissages. Pour compenser en partie et au moindre montre que lors des evaluations de lois de commande
coüt ce manque de sensation, il est pr6vu ä partir de 96 avec mini-manches, une part significative de la
d'6quiper les cockpits de simulation de sieges ä charge de travail de pilotage pouvait Stre induite par
coussins gonflables permettant de restituer ä tres des problemes ergonomiques. Les developpements
court terme le facteur de charge vertical (g-seat). futurs utilisant des nouveaux concepts de commande
devront done integrer, des la phase de conception
Le Systeme de visualisation actuelle SOGITEC initiale, les aspects Qualites de Vol et MMI.
GI- 10K ä 3 canaux r6pond egalement bien au besoin
des essais effectu6s. Le champ de vision actuel (80 "x 9.REFERENCES
180°) apparait tout ä fait süffisant pour effectuer
1. Hellmuth, Kampa, Karl, "The TIGER Cockpit and
toutes les manoeuvres types, y compris celles
comportant des phases de translation laterales its Simulator", 18* ERF, Sep. 92.
(sidestep, pirouette). Par le biais d'un accroissement
 18-7

2. Braun, Kampa, Schimke, "Mission Oriented 7. Damotte, Mezan, "Evaluation of Advanced

Investigation of Handling Qualities through Control Laws using a Sidestick on the Experimental
Simulation", 17th ERF, Sep. 91. FBW Dauphin Helicopter", 18th ERF, Sep. 92.

3. Brunand, De Villemagne, "MMI simulation for the 8. Heng, "FBW Dauphin System Demonstrator",
TIGER-HAP", 20th ERF, Sep. 94. Symposium: In-Flight Simulation for the 90's,
Braunschweig, July 91.
4. "Handling Qualities Requirements for Military
Rotorcraft", Aeronautical Design Standard - 33 9. Rollet, Thibaudat, "EUROFAR simulation trials
(ADS-33C),Aug.89. on EPOPEE simulator", 18th ERF, Sep. 92.

5. Mc Farland, "Transport Delay Compensation for 10. Cooper, GE. and Harper, R.P., "The Use of Pilot
Computer-Generated Imagery Systems", NASA rating in the Evaluation of Aircraft Handling
TM100084. Qualities", NASA TN D-5153, Apr. 69.

6. Schimke, Guyomard, Lane, Bellazi, Charlton, 11. Mitchell, Hoh, Atencio JR, " Ground Based
Pausder, "The European ACT Programme - Simulation Evaluation of the Effects of Time Delays
Complementary use of Ground Based Simulation and Motion on Rotorcraft Handling Qualities,
Facilities and Experimental Fly-by-Wire/Light USAAVSCOM TR 91-A-010, Oct. 92.
Helicopters, 18th ERF, Sep. 92.

SIMULATOR COCKPIT DOME

VIRTUAL COCKPIT

GRAHIC DEVELOPMENT STATION

SILICON GRAPHICS '.0/25

SW-DEVELQPMENT STATION
SIMULATION CONSOLE
SUN WORKSTATION

Figurel. Simulateur SIMCO

18-8

*5§
70
I/ "
Zf7*0
4L2W_ gy " SOÄ-^
f/ 1 r/:•= 4030 LVAlJ
77 > ®\\J3:
/
L \ \
/ / / / / 1 ' I \ \
140 160 180
-180 -160 Li40 1-120 i-lOO -80 -60 -401 -20 D 20 40 60 80 lOOJ 120/
-r—lOn
-20. J ' /
-1LTT
1 *A * V\ -30 (2)// /(B)/
\ -50
\\-60 / /
$$i5-80.
^^-90"

Figure 2. Champ de vision du dome DASA

Figure 3. Vue de la simulation SIMCO (dome DASA)

 18-9

EXPERIMENTAL COCKPIT
/ SPHERE

GRAPHIC DEVELOPMENT STATION

SRICON GRAPHICS VAPS M9GNER

SIMULATION TARGET
Gl 10 000 SOGITEC REAUIME OS
MimiPROCESTJNG
SW-DEVELOPMENT STATION
1/oconmoL SIMULATION CONSOLE
FIIGIKMANAGEMENI
FIKSHIMECANHCS SUN WORKSTATION

Figure 4. Simulateur SPHERE

140 160 180

Figure 5. Champ de vision du simulateur SPHERE

18-10

Figure 6. Vue de la simulation PFCS NH-90 (SPHERE)

Figure 7. Obstacle artificiel pour manoeuvre "Lateral Jinking"

 19-1

Technological and Economical Limits

Experienced for R&D Mission Simulation

P.Castoldi M.Allocca S.Lo Presti M.Trifoglietti

Alenia Aeronautica
corso Marche 41
Torino - Italia

SUMMARY programs and the different roles of the related

This paper illustrates some significant lessons aircraft.
learned through R&D activities performed at the
Flight Simulation Center of Alenia Aeronautica AMX and EF2000 FLIGHT SIMULATORS
in Turin, highlighting experiences in Data Base Two independent flight simulators are working
Generation, Graphics and Graphics Prototyping, in the Flight Simulation Center in Turin.
Particular Simulated Tasks and Integration of They are mainly structured following the same
Dedicated Devices such as Target Projector and philosophy, but with some substantial
g-Seat g-Suit System. differences, due to the different roles of the
For each specific mission task, the problems, aircraft.
solutions, alternatives and resulting system Both systems have:
limitations involved in identifying the required
system characteristics are discussed. . Cockpit on fixed base
. 30-feet inflatable dome housing the cockpit
INTRODUCTION . Projection system inside the dome
In the R&D Mission Simulation, it is easy to run . Host computer and Special Purpose
into technological and economical roadblocks. Computer (CIG, Graphics Work Stations,
Consequently it is the best area to investigate and etc.)
define the characteristics and the specifications . Control room
of particular devices for specific tasks.
AMX Simulator
Design, prototyping, production and new It is the oldest one and its configuration allows
versions development are the phases of the R&D and pilots training activities.
aircraft life and consequently of the associated The cockpit is fully equipped with the real
simulator. hardware (h/w) including the MIL-STD-1553
For each phase, the tasks to be performed are instrumentation.
different, generating additional requirements for The real HUD has an additional lens to collimate
the existing simulation tools, which have to be the simbology to the image on the internal
tailored, upgraded or modified. surface of the dome.
Simulation has achieved its highest engineering The projection system is composed of three
expression in the modern aircraft flight TALARIA Light Valve Projectors PJ5155 with
simulators. standard lenses, located behind and above the
Ground-based flight simulation increased its pilot's eye, projecting a three windows image of
importance and its legitimate domain of 160* horizontal x 50° vertical: one high
investigation and application both as research resolution channel (1 million pixels), two medium
and development tool and as a training aid. resolution side channels (0.5 million pixels each).
Computer flight simulation is a development tool A Control Loading System is integrated with the
with unique capacity to develop and to evaluate stick for the simulation of powered and manual
design information during the whole life of the flight controls including hydraulic failures.
aircraft program. A pneumatic/electromechanical g-Seat/g-Suit
system is integrated in the cockpit for tactile
Two simulators are considered in this paper: sensations.
A very simple sound system is implemented.
- A/G mission AMX
- A/A mission EF2000 EF2000 Simulator
The cockpit is full equipped with commercial/
The first aircraft is well into production while homemade and representative displays and real
the second one is in the prototype phase, equipment such as:
consequently the two simulators are in different - primary flight controls (stick, throttles,
states of integration. pedals) interfaced via MIL-STD-1553 buses;
Different requirements emerge for these - seat mock-up;
simulators, due to the different status of the - console panels.

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?"
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
19-2

The HUD symbologies are projected on the coordinate sets;

scenario by a CRT projector in the correct FOV
to the pilot's eye. . special effects - e.g. explosion,
The projection system is composed by three lightning;
TAL ARIA PJ5155 equipped with wide-angle
optics and located below the cockpit projecting . universal features - features often
three windows for a continous image of 190* instanced in the data base, such as trees
horizontal and 50* vertical: high resolution in the and houses;
central channel and medium in the side channels.
- EDB related data, such as collision detection
A dual slewable and zoom optics for independent and target data;
targets is integrated with a LV9000 SIM
projector. A Silicon Graphics workstation - light and color tables;
provides two independent target images.
The dual optics is able to project in any azimuth - texture data, needed to project raster images,
and elevation of the dome two targets controlled acquired through a scanner, upon terrain or
by either the second dome or through the control culture faces.
room console or by an "intelligent" software (s/w)
program. The previous data types, with the exception of
moving models and special effects, are actually
organized in two separate mathematical models:
a coarse data base, active in the background, and
Data Base Generation a fine data base, active in the foreground. The
two data bases are organized according to the
a - Generic data base limitations. following hierarchy:
The image generators (IG) employed in the
Alenia Aeronautica Simulation Centre ( CIV and - the most elementary unit is a square degree;
CIV-A) are able to display only data bases
developed by means of Data Base Generation - each square degree is subdivided into coarse
System (DBGS), an integrated combination of regions, rectangular in shape and uniform in
dedicated software and hardware tools (essential size all over the data base;
h/w: Sun workstation - Eikonix scanner -
Summagraphics digitizing table). - each coarse region is split into fine regions,
whose size must be some even subdivision of
A DBGS data base is a mathematical model of the coarse region size, while their boundaries
the entire visual scene available to the image can be irregular, reflecting the actual terrain
generator; it usually represents some specific outline;
geographic area in the world, but it may also
depict a generic area, provided it contains - eventually, each fine region is comprised of
suitable visual cues for training on specific tasks. clusters, consisting of one or more terrain faces
The base components of such data base are plus all surface culture insisting on them.
terrain and culture, that are, respectively:
b - Italian D.B. generation: advantages,
- a network of triangular or quadrangular limitations.
faces, matching the actual ground course, in At the beginning, data bases provided by the IG
as much detail as wished by the user; manufacturer were used, mainly the broadest
one, representing the region of Lake Mead
- the representation of the most relevant (USA) surrounded by a wide generic area,
three dimensional features, either natural or notwithstanding that the pilots who were trained
man made, present in the area of interest. in Alenia Simulation Centre found them not
satisfactorily fit for realistic simulation.
Besides, a data base is comprised of several data
types loaded into the IG memories either at The demand for a new data base, including the
mission initialization or during a real time greatest possible area of Italy, sprang from the
mission: task of integrating a Mission Planner, developed
by the Alenia Artificial Intelligence Laboratory,
- Environment Data Blocks (EDB), of fixed into the AMX simulation program.
size, containing the basic geometric
description of: Even Lake Mead data base was unfit: source
data were not available, which prevented us from
. static features - the majority of the data working out the simplified terrain and culture,
base, including terrain,surface culture, strictly coherent with the data base itself,
static 3-D objects; essential to the mission planner; in addition, it
covered too limited an area, where the eligible
. moving models - separately controlled way-points were too few and scattered for the
 19-3

operative requirements of the planner itself. base should be designed to largely avoid overload
conditions.
Therefore, we chose to generate a brand new
data base, covering a global area of sixteen Indeed, non generic data bases are usually built
square degrees (from 38N-14E to 41N-18E). starting from Digital Land Mass System (DLMS)
data, generated from satellite and other aerial
First of all, we had to face the limits common to reconnaissance photos; these data include both
all DBGS data bases due to IG memory and terrain elevation grid data and overlaid cultural
processing constraints, starting with the data, describing the location, size and main
maximum on-line storage provided by both CIV attributes of all significant features. Such data
and CIV-A (only 512 EDB's) and the maximum were available only for four square degrees at
number of faces contemporaneously displayed, Level 1 standard (not including linear features),
just 4000, comprised of terrain, surface culture therefore all the remaining area had to be built
and models. from terrain contour lines and culture features
digitized on the spot from 1:250000 pilotage
Such limits involved the need for a careful, charts.
overall data base design.
This caused a lot of additional problems, which
For instance, the DBGS offers a choice of fixed turned evident only eventually, while flying the
or variable density terrain structure: while fixed data base, such as:
density terrain yields explicit user control over
terrain density, therefore a more regular form - mismatching of neighbouring degrees at the
and a predictable load, variable density terrain boundaries, due to mistakes in digitizing
provides a better fit of terrain triangles to the overlapping contour lines;
actual area modelled, through the generation of
a less uniform structure. - generation of uneven terrain surface
underlaying hydrographic features;
Another critical choice is that of the coordinate
system: as to the earth shape, one can select - inaccuracy in the resultant altitude and
either flat earth with convergence or geocentric neighbourhood of particular way-points.
earth and also the units, normally feet, can be
designated as centimetres to provide greater Such problems could be solved mainly by turning
resolution. the terrain structure of the wrong areas into
variable density, inside the overall fixed density
However, no matter how careful the global terrain, correcting some of the contour lines
design is, forcing the data base to stay within the and/or digitizing some new ones, inserting/
IG processing constraints is a very complex task deleting some terrain triangles directly.
with many variables, such as the number of
viewpoints, the number of channels and the field After all, a final check of the data base quality
of view per channel, the activation range by had to be performed in the dome to take into
feature type, the number of terrain clusters per account the lower contrast and brightness and the
coarse region, the predominant height and speed different field of view generated by the
the data base will have to be flown at. Indeed, projector system on a white screen; some peculiar
the IG updates the display 60 times per second: colour matching suitable for the monitors often
during each cycle, a lot of parallel computations did not prove the best harmonization inside the
must occur and be completed; in addition, most dome.
internal processing lists have limited length.
c - Texturing requirements.
For instance, in dual viewpoint systems both Cell texture generation is the process of
viewpoints must share the same processing capturing images, mainly drawings or photos, as
capacity so that the data density of each must be cell maps and then associating the maps
less. The data density can increase as the field of themselves with specific data base faces; the
view is narrowed, since a smaller portion of the process can be split into three steps: capturing a
data would be visible; likewise, as activation source image, generating the related cell map,
ranges are reduced for coarse and fine terrain placing the map over one or more faces.
and for data at each level of detail, the data
density can increase again. The number of fine In order to get a suitable cell map, the modeler
detail terrain clusters per coarse region is critical usually needs to modify the initial result,
to the priority processing time; a conservative whatever its origin, increasing or sweetening its
number is 80 and an upper limit of 160 clusters/ contrast, keeping only a part of it, combining it
coarse region should be observed. with a different picture, inserting or getting rid
of some details, eventually smoothing the
The IG contains real time overload monitoring matching of two opposite edges (right/left or
and automatic adjustments to correct overloads upper/lower), if the image has to be repeated
with degraded performance, however the data where projected.
19-4

Afterwards, to position a cell texture map on a objects to make them recognizable and true-life:
face, several items of information are required: for a city, streets (highway, if any), railroads,
crossroads, bridges; for an airport, the tower,
- rotation, offset and scaling : orientation, large buildings, bushes, trees, groups of houses;
location and size of the map; for a lake, hotels, trees, gardens and so on.
- translucency: relative opaqueness, transluce,
or transparency of the cells; e - Time & Cost
- modulation: contrast range of the cells; The generation on site of the data base described
- edge control: bits that determine the above required the efforts of a dedicated person
wrap-around conditions and the edge blending for two years and a half, six months of which
options for the cell textured face; the option of thoroughly devoted to learning how to exploit the
wrap around is provided so that a very large subtlest features of DBGS.
face may have a pattern continuous up to the
limits of its edge. Digitizing and testing largely proved the two
most time-consuming operations; while the
The overall process is quite long and demanding, former could have been saved, provided the
moreover the texture can be actually tailored DLMS data had been available, there was no way
only while flying: for instance, broad translucent to avoid the latter, since only the pilots'
areas may cause overload only when inserted comments could allow to realize which details
inside regions already too rich in details, which were actually essential to the simulation goals.For
cannot be realized a priori. instance, the choice of colouring the terrain
faces, the insertion of objects in front of each
Despite a good deal of efforts, black-and-white airport end, the effort to cover large culture
texture may fail its aim of making the landscape areas with several different texture patterns so to
depicted true-life, especially as to large surface make the data base lively even where objects and
culture; however in our data base the worst further details were lacking, were all answers to
effect, highly annoying, was attained by the needs expressed by the pilots under training.
terrain itself, which was satisfactorily solved by
changing the colour of terrain faces, choosing The testing activity was also the most expensive
automatically one out of ten close tonalities of item: each significant modification required the
brownish. use of both the image generator stand-alone and
of the simulator with a pilot in the loop, so
d - 3D Objects requirements. summing up to a whole bunch of hours.
A model is a group of faces and possibly point
lights stored in a model library and subsequently We would not be far from truth stating that
placed in the environment as a group, as far as digitizing and processing contour lines and
channel assignment, priority and level of detail culture took about one year, while the second
decisions are concerned; any combination is year had to be devoted to correcting and
valid, provided it does not exceed a single EDB improving the data base.
limit.
In addition, the generation of moving and fixed
A maximum of 32 models into some meaningful models took only about six months, since some
arrangement, named model complex, can also be general model libraries had been supplied by the
stored on the model library and therefore placed manufacturers.
in the data base by simply specifying its location.
For sure, a new-generation device, able to
Eventually, a supermodel is an assembly of up to actually take care of the image generator real-
8 individual models defining a single moving time limitations, would have helped in cutting
model, such as an aircraft carrier or a fighter. costs, saving at least the stand-alone tests, while
Each part of a supermodel is called a partition colour texture would have reduced the need of
and is stored under its own name on the model mixing different patterns.
library; all partitions are modelled relative to a
single origin, usually the resulting supermodel's Operational Issues
center of gravity. Simulation tasks often require trade-off between
It is quite intuitive that supermodels are the most different possible technical solutions in order to
demanding 3-D features, but some static models achieve the best available realism without
in our data base proved challenging, too: for increasing costs beyond acceptable levels;
instance, 3-D cities, not supplied by DBGS, were therefore, the most careful attention must be
implemented as properly textured raised forests, payed to all those features of the visual system,
while small clouds were accomplished as trees the cockpit (controls and displays), the motion
with huge, invisible trunks and white, system (if required), the scenario, needed to
translucent textured leaves. present a true-life reproduction of the
environment experienced by the pilot during an
Much care was given to target areas, selected actual flight.
way-points surrounded by a whole bunch of
 19-5

Such goal becomes vital in case of peculiar Simulation of the typical IR steerable head of a
environmental conditions, such as low level night laser designator pod requires the use of the IG
flying, or high demanding tasks, e.g. formation fourth channel, displaying the same
flight or air-to-air refuelling. monochromatic standard of the remaining IG
channels.
Low Level Night Flying
The field of view (FOV) of the overall The line of sight (LOS) of the fourth IG visual
background image stays 160*h x 50*v, but also channel, instead of being fixed as for the
the middle channel is allowed just medium remaining three channels, is steerable; steering
resolution (1024x512); a fourth channel modalities are the same of those envisaged for
(1024x512), associated to a second viewpoint, is the real targetting pod (either in automatic or in
projected on the cockpit multifunction display. manual mode), thus allowing an accurate
simulation of the targetting pod seekerhead
Such solution, alongside the level of detail and aiming procedure.
the image resolution, proved effectual to simulate
the tasks related to the air-to-ground role typical The fourth channel viewpoint position is the
of the AMX aircraft, providing also good same of the seekerhead pod installed on the
response to the secondary air-to-air tasks aircraft (fuselage centerline station); moreover,it
simulation. is possible to increase/decrease the channel FOV
in order to simulate even the real pod zoom
One of the most demanding tasks in air-to- capability.
ground attack role, low level flight at night or in
marginal weather conditions has become more Night Vision Goggles
and more a frequent task in modern flight Simulation of NVG operational use requires the
operations, during which sophisticated surface- evaluation of several aspects related to the image
to-air weapon systems are to be faced. presentation and also, not a less critical factor, to
Training a pilot in performing the procedures the equipment bulk and weight.
required by such a mission involves the
simulation of the different electro-optical (E/O) Several solutions can be worked out, taking as a
sensors used by the pilot as flight support: standpoint the use of a typical GEN III NVG
equipment and the specific simulation tasks.
- Navigation FLIR
At first, the use of a steerable projector was
- Targetting Pod evaluated, so to visualize an appropriate image of
adequate FOV (45* circular for each eye,
- Night Vision Goggles (NVG). typically) in the pilot's area of interest, therefore
simulating the "tunnel" effect due to the reduced
Navigation FLIR NVG FOV.
In order to simulate the operational use of a
raster/stroke HUD with navigation FLIR, the IG In order to lower costs, the pilot FOV was
representation of the external world has been afterwards reduced, so allowing only the required
elaborated, providing a monochromatic (32 levels portion of the whole 160*x50° IG image to be
of gray) image similar to the one supplied by an seen at each moment; this was achieved opening
IR sensor. two port-holes corresponding to the pilot's eyes
in the transparent visor of a standard HGU-2P
The typical raster/stroke HUD FOV limits the flight helmet, by means of an appropriate
image presented to the pilot to 30ex20<> in template. For the same reason, we chose also to
boresight direction; this restricted FOV is not disregard the pilot's physical stress, focusing only
obtained via software, instead an appropriate on the visual aspect of NVG operational use, no
template has been elaborated so to hide the IG matter how bulky the used device turned out to
image but for the FOV corresponding to the be.
HUD.
Using the monochromatic table described above
This solution allows us to quickly restore the for the FLIR simulation proved the least
original 160"x50* FOV when required (e.g. demanding, though satisfactory, way to show the
when simulating the use of NVG) avoiding NVG vision on the IG.
modifications of the IG parameters during the
simulation session and the related time The final result has proved effectual in
consuming procedures. simulating the obstruction and the image quality
The monochromatic image (Sensor Color Table) of the NVG; visibility inside the cockpit is
of the IG is sufficient to satisfy the task involved similar to the one allowed by the use of a typical
without implementing a very sophisticated and in-line optical scheme.
expensive Sensor Data Base.
Both introduction of NVG compatible cockpit
Targetting Pod lighting and masking of cockpit light sources
19-6

proved unnecessary and also simulation of the accuracy whenever passive target detection is not
NVG image blooming due to intense light was a critical factor.
not deemed a major factor and therefore
discharged. Cockpit
The realistic cockpit reproduction in terms of
Besides, simulation of a night-time environment controls and displays available to the pilot is
and of the related NVG/FLIR sensors imagery mandatory for simulation effectiveness.
relaxes the requirements for depth, definition Response of each aircraft system should be
and speed cues, otherwise mandatory in case of replicated in the most precise way.
low level flight simulation task. When a new system architecture has to be
evaluated, modifications to the cockpit layout are
Visibility was adjusted to allow recognition of often necessary, expecially if the Man Machine
ground targets at the distances typical of the Interface (MMI) is heavily involved.
relevant E/O sensors. In such case the introduction of an additional
display is needed.
Alongside the E/O sensor, simulation of a A rapid prototyping front panel of the cockpit
multimode radar is an important issue when would be more beneficial in terms of minor h/w
assessing AMX night attack version capabilities. modification and quickest representation of the
new modification.
Multimode Radar
Simulation of a multimode radar is aimed to Wide angle HUD symbology is generated by a
providing the pilot with the workload he has to Silicon Graphics workstation linked in real time
face during actual radar operations. to the simulation computer and projected inside
the dome superimposed to the IG image, while
Thanks to the specific simulation tasks, we could the real conventional HUD available on the flight
spare developing complex radar models, focusing simulator is inactivated.
all our efforts in generating a realistic simulation
model, furnished with several different Formation flight
modalities: This task is performed in the AMX simulator as
well in the EF2000 one, exploiting different
- A/A Ranging conditions.

- A/S Ranging In the AMX simulator the h/w key elements are
the visual system, the g-seat/suit system and the
- Ground Mapping sound system.

- Terrain Avoidance The visual system is limited in horizontal to 80°

left and right, which compels the second aircraft
- Contour Mapping. to be displayed inside a strict FOV.

Since A/A attack capability was not to be The g-seat/suit system helps the pilot to control
assessed, no further A/A mode has been the aircraft in minimum movements at high
developed. frequency, providing the tactile sensation so
significant for such task.
The major effort in this simulation is the
landmass definition, to be performed according The sound system is an additional feed-back
to the IG database; in this way it was possible to giving several aural cues related to the dynamic
include correlated returns of ground tracks. of:
The landmass simulation is possible thanks to the - aerodynamic noise;
use of a VGXT 420 workstation. - the engine RPM variation;
The radar returns are presented inside the - APU noise / ECS noise;
cockpit on a proper multifunction display upon - landing gear touch-down;
pilot request; radar fixing procedures are - wheels rolling on the runway.
available to the pilot in Ground Mapping Mode.
No motion system is present, but the control of
A dedicated control panel is included inside the the aircraft in all possible formation flight
cockpit for radar modes management, antenna configurations proves satisfactory and feasible:
tilt and gain control; appropriate HOTAS controls actually, the worst drawback is just the limitation
are allocated as well. of the visual scene FOV, which will be partially
solved installing a slewable optic behind the
The radar simulation provides effectual aid to pilot's eye, capable of projecting a second
the flight in night-time or marginal weather aircraft and the background inside a window of
condition, allowing the pilot to recognize 50°hx35*v in azimuth and elevation.
waypoints and target locations out of the E/O
sensors range, exploiting navigation and attack As highlighted before, in the EF2000 simulator
 19-7

the characteristics of the hardware are different: manoeuvring. However some degradation was
neither a g-seat/suit system nor a sound system experienced with more rapid manoeuvres,
are available, while a wider FOV of the visual especially lateral ones.
scene and a new generation target projector are
integrated. In the third case, the lack of both systems caused
a very difficult control of the aircraft in fine
The target projector system consists of a dual adjustments, but again, after a certain amount of
indipendent high resolution, light-valve target attempts the pilots were able to reach a
optics, complete of zooms and shutters, satisfactory level of aircraft control.
projecting two circular windows ranging from 20
to 7 degrees; the two target images, produced by Moving into the EF2000 dome, the pilot found
a Silicon Graphics workstation and therefore himself in the same situation as the third case of
thoroughly independent from the CIV, can be the AMX (neither sound nor g-seat/suit system),
projected both instead of the IG models and but for the implementation of a slewable
wherever in the dome such models cannot be projector of two indipendent targets. The results
displayed. were comparable to the third case experienced
with the AMX simulation.
A correct evaluation of the conditions where
either the IG or the Silicon Graphics targets have A formation flight with two independent targets
the better effectiveness has not been yet has been tested in different configurations and in
investigated in deptht: it depends on the distance, any value of azimuth and elevation of the targets.
the relative attitude of each target, the weather The limitations of this configuration are
conditions and the time of the day. highlighted in the 20*maximum FOV of the
target image outside the background scene
Due to the engine and flight controls of the (190*hx50°v). This means that an aircraft with
EF2000, performing close formation flight is the dimensions of a fighter like the EF2000 can
more difficult than in the AMX simulator; be completely displaied not closer than 40m.
according to the pilots' analysis, to overcome the No major differences from the results obtained
problem the availability of the sound system in the AMX simulation have been highlighted,
would be much more essential than the one of apart from the different behaviour of the aircraft
the g-seat/g-suit system. In order to verify such in terms of flight control system.
hypothesys, different assessments in both
simulators have been performed varying the In-flight refuelling
cueing environment: Another task experienced in both simulators is
the in-flight refuellig.
a) AMX simulator without sound system; The configuration of both simulators has been
described above.
b) AMX simulator without g-seat/suit system; In order to give the visual cue of the probe
attached to the manned simulated aircraft, a
c) AMX simulator without either sound or g- model of it has been generated in the IG and
seat/suit system; displayed in the scene as a fixed part of the "box"
of the own aircraft (the own aircraft is designed
d) EF2000 simulator with one, or with two syntethically as a box).
targets.
The probe has been depicted as a model
In the first three cases, a target controlled by a composed by 6 EDBs using about 60 faces getting
dedicated station in the control room was the data of the AMX real one with the farthest
displayed in the scene as a wingman, once on the point fixed at 2m forward and lm to the right in
left, once on the right and once in front as a front of the pilot.
leader. The tanker is a Boing 707 TT used by the Italian
The target's manoeuvres were selected starting Airforce equipped with three baskets generated
with a simple linear velocity, then smooth left by the DBGS as well.
and right turns with pre-calculated roll angle The dynamics of the baskets' length and the
(30*, 45<\60'>), eventually the complete dynamic turbulence behind the tanker have not been
of an aircraft with the same performances of the simulated.
manned simulated one. To simplify the task to be performed, the tanker
In the first case, the pilots experienced flies at a constant altitude and velocity,
difficulties during the very close formation positioned at half mile in front of the pilot.
flight, but after several attempts they proved The image quality and the visual cue of the
able to perform the task, within adequate probe are realistic even if the 3-D is absent and
criteria. the tanker model is very close to the real one,
complete of shading, texture and light points.
In the second case, after a period of
familiarization with the alterated cues the pilots Flying towards the tanker is a task that does not
were able to fly in formation with gentle involve difficult controls, but as soon as the
19-8

fighter gets close to the tanker and the pilot starts performance graphics running at 30 - 60 Hz with
engaging the basket with the probe a very a time delay of about 50 ms.
uncomfortable feeling appears, causing the task
to abort after several attempts. The workstations have the possibility to send
Such failure is due to two major reasons: video signals in different standards compatible
- lack of vestibular cue; with the display system in the cockpit, in the
- lack of 3-D visual cue. dome and in the control room: the most used
standards are 30 Hz interlaced, PAL, NTSC and
The vestibular cue helps controlling the aircraft 60 Hz not interlaced.
close to the tanker in the fine adjustments Particular attention is devoted to the monitoring
especially around the longitudinal axis, when the in control room; the aircraft behaviour is
tanker is above and the basket is less than 10 represented in real time at 30 Hz in terms of
meters far from the probe. attitude, surface movements, stick position,
engine values, landing gear movements and other
The 3-D visual cue plays an important role, dedicated visualizations.
when the objects in the scene are closer than 20 Related to the system modelling and relative
meters. The sensation of the in-flight refuelling failures, a monitor is devoted to the graphic
in 2-D is comparable to a flat picture in which representation of the functionality of general
understanding the object order in depth is systems such as electrical, power and secondary
impossible. power system, completed of animation of these
The final result is a very difficult task, risking to systems components during the normal and
engage a pilot coupling mode in longitudinal failured behaviour.
axes.
In order to partially mitigate such unwanted The limits of this graphic network system are
effect, an artificial translucent disc has been determined by the number of the workstations
drawn and added on the top of the probe. working at the same time. For different tasks and
This disc, the same size as the basket, helps the different configurations, different allocations of
pilot in better understanding the basket position the workstations have to be defined in order to
relative to the probe. split these between the AMX and EF2000 flight
Some improvment has been obtained, but the simulators, both using video switches controlled
task still remains very difficult, successful only by the pilot in the cockpit and/or by the control
after many attempts. room.
Some Airforce, for example, use special facilities
to train pilots to this specific task: Conclusions
their simulators reproduce an h/w scaled mock- Several technological limitations of the existing
up tanker and the probe is a 1:1 scale mock-up. simulators are evident, depending on the tasks to
With a special optic system it is possible to be performed.
combine the view of the h\w (probe and tanker) Using an R&D simulator means to be able to
with the background scene, generated by an IG. touch many aspects of the simulation spectrum:
The 3-D image so got succeeds in training - trade-off evaluation between operative
properly without the motion platform stimulating configurations;
the vestibular system. - ground and general handling simulation;
- handling simulation with manual flight;
Graphics - failures simulation and analysis of the pilot's
The graphic symbologies generated by the reactions;
workstations connected with the real-time - emergency procedures;
simulation are dedicated to the onboard Multi - data bases and interactive scenario generation
Function Display System and to the and integration;
operator/instructor console Color Monitors. - operational mission simulation, including
These graphic representations are mainly 2-D, phases like A/A combat, formation flight, in-
therefore not very demanding for the flight refuelling;
workstations CPU; however, the graphic - human engineering assessment;
processor is heavily loaded due to the high image - development of Crew Assistant System;
refresh rate. - simulation engineering development.
In the Alenia Flight Simulation Center, Silicon During all these tasks is easy to impact into the
Graphics workstations connected in LAN on difference between the reality of the simulation
Ethernet with TCP/IP protocol are being used. and the reality of the real world.
Three of these are high performance, capable to Many trade-off analyses and compromises have
generate complex graphics such as A/G RADAR to be done task by task and different solutions
representation, two independent targets, colour have to be chosen case by case depending on the
maps and synthetic interactive scenario. architecture of the simulator in use and eventual
C language under Unix system is used to improvements that can be implemented.
generate the graphics running at different rates,
depending on the task.
HUD, AOI, dual targets and others are high
 19-9

REFERENCES

1 "Piloted Simulation in Low Altitude, High

Speed Mission Training and Rehearsal",
AGARD FMP WG20 report (draft)

2 "Fixed Wing Night Attack Mission:

Assessment in a Flight Simulation
Environment" AGARD FVP symposium on
"Low-level and Nap of the Earth (N.O.E.)
Night Operations", Pratica di Mare (I), 1994.

LIST OF ABBREVIATIONS

A/A Air-to-air
A/G Air-to-ground
APU Auxiliary Power Unit
DBGS Data Base Generation System
DLMS Digital Land Mass System
DMA Defence Mapping Agency
ECS Environment Control System
EDB Environment Data Block
FLIR Forward Looking Infrared
FOV Field Of View
HOTAS Hand On Throttle and Stick
HUD Head Up Display
H/W Hardware
IG Image Generator
IR InfraRed
LOS Line Of Sight
MMI Man Machine Interface
NVG Night Vision Goggles
S/W Software
 20-1

Utilisation de la simulation
pour la conception du SNA du Rafale
R. Goussault
Chef du Departement Simulation Validation
DGT/DSA
DASSAULT AVIATION
78, Quai Marcel Dassault
92214 Saint-Cloud
France

M. Leclere
Adjoint au Chef du Centre de Simulation
CENTRE D'ESSAIS EN VOL
Base d'Essais d'Istres
13128 Istres Air
France

1. SOMMAIRE 3. LES INNOVATIONS DANS LA DEMARCHE DE

Dans cette presentation elaboree conjointement par Dassault CONCEPTION
Aviation et le Centre d'Essais en Vol (CEV), nous indiquerons Pour faire face ä de tels enjeux et surmonter ces difficultes, une
d'abord Ies innovations importantes adoptees dans la demarche demarche totalement nouvelle a ete adoptee pour permettre de
de la conception du Systeme de Navigation et d'Armement maitriser la conception du Systeme d'armes.
(SNA) de l'avion Rafale et en particulier le role fondamental
de la simulation. Nous montrerons ensuite la synergie existant 3.1 La vision avancee Systeme
entre Dassault Aviation et le CEV dans la definition des La conception du SNA Rafale (!'architecture materielle et
besoins de cette simulation, dans le choix des modeles et des logicielle, les regies generates, la definition de la cabine et de
moyens qui la composent et dans son utilisation par l'interface pilote Systeme) n'etait plus compatible d'une
Yarchitecte industriel et l'Etat. demarche consistant ä additionner les functions operationnelles
Quelques precisions sont necessaires avant d'entrer dans le vif apres les avoir definies separement. II a fallu au contraire se
du sujet: projeter dans l'avenir et bätir, le plus tot possible, une vision
avancee du Systeme prenant en compte l'ensemble des
• dans le programme Rafale, Dassault Aviation est
functions operationnelles et techniques, ce qui a permis de
Yarchitecte industriel du SNA
definir leurs priorites respectives et les principes de leur mise
• le CEV est, au sein de la Delegation Generale pour
en oeuvre. II est par exemple fondamental de considerer tous
l'Armement, l'organisme etatique expert pour les
les differents capteurs destines ä equiper l'avion ä Tissue de
essais en vol ainsi que la mise au point des materiels
leur developpement, ceci de fagon ä etablir les regies de leur
aeriens et systemes embarques militaires
partage par les fonctions utilisatrices, les logiques de gestion
• la simulation designe ici la simulation pilotee
ou de ralliement ä leur appliquer et les lois d'association ou de
fusion des informations qu'ils fournissent en vue de leur
2. LA SITUATION AU DEBUT DU PROGRAMME
presentation ä l'equipage. II faut egalement que cette vision ä
RAFALE terminaison comporte une premiere definition de la preparation
Par rapport ä celle des systemes des differentes versions du
de mission, qu'elle permette une cooperation entre plusieurs
Mirage 2000, la conception du Systeme du Rafale est sans acteurs interactifs et qu'elle integre l'utilisation du Systeme
commune mesure. En particulier, la maitrise de la definition de Rafale au sein de reseaux de contröle et de commandement.
son SNA, pour prendre en compte une expression de besoin
Ainsi, la connaissance des differentes conduites de tir
tres ambitieuse, a necessite un changement et des innovations envisagees permet de definir ['architecture la mieux adaptee
dans la demarche de conception. En effet, plusieurs difficultes pour optimiser les ressources materielles et logicielles du
fondamentales se sont nettement imposees des le lancement du Systeme, tout en privilegiant la modularite et l'evolutivite
programme : necessaires aux besoins operationnels futurs. De cette vision
• la complexite de 1'analyse technique du besoin avancee etablie tres tot dans le developpement, sont ensuite
operationnel et notamment la polyvalence demandee au deduites les versions intermediaries du Systeme qui, pour des
Systeme, raisons techniques et economiques, est realise progressivement.
• le niveau d'integration du Systeme qui en decoule, pour
ne pas augmenter la charge de travail de 1'equipage que 3.2 Le groupe de conception Systeme
ce soit dans le cadre d'une version monoplace ou Pour concevoir cette Vision Avancee Systeme et la faire vivre,
biplace, il faut savoir « regarder loin » et avoir les « bonnes idees » tant
• la maitrise des coüts et des delais sur le plan operationnel que sur le plan technique. Ce travail

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?"
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
20-2

est realise par un Groupe de Conception Systeme, compose de et aux decideurs de faire ces choix ä bon escient, ce qui
pilotes d'essais et d'ingenieurs de facon ä prendre en compte ä renforce le besoin de la simulation. Pour cela, la simulation est
la fois les aspects ergonomiques et les aspects techniques. Le developpee avec le souci de representativite sur le double
röle de ce groupe est de concevoir ä partir des specifications aspect d'un environnement realiste pour placer le pilote dans
techniques de besoin, un Systeme utilisant les techniques et les les conditions süffisantes d'appreciation et d'une architecture
technologies emergentes pour satisfaire des performances (en particulier le decoupage fonctionnel et le respect des
exigeantes mais qui doivent rester compatibles de la charge de interfaces aux bornes des equipements) deduite de celle du
travail des pilotes et de la fiabilite indispensable ä leur securite. Systeme reel.
Ainsi, ä partir du dossier de construction Systeme definissant Pour concretiser ce besoin, le Centre d'integration
les grandes orientations de la solution retenue (architecture et Hybridable (CIH) a ete cree et installe au sein du bureau
premier dimensionnement des ressources necessaires), ce d'etudes de Varchitecte industriel pour accueillir le maquettage
groupe de conception conduit et Oriente les travaux des equipes du SNA du Rafale. II s'agit d'un outil qui permet de montrer
specialises en charge du developpement et de la realisation du une image informatique du comportement du Systeme dans un
Systeme. environnement representatif simule. Lance en 1989, il est
operational depuis la fin 1992, date oü a ete presentee la
3.3 Des outils : la simulation et le centre d'integration premiere image informatique complete, appelee Vision ä
hybridable Terminaison du SNA du Rafale. Le CIH est organise de
La demarche de conception ainsi adoptee pour le Rafale maniere ä favoriser le travail en Synergie des principaux
permet de maitriser le developpement du SNA, ä condition de acteurs de la definition, e'est ä dire :
s'assurer que la Vision Avancee Systeme, d'une part repond • les pilotes d'essais car la cabine et son environnement
aux besoins operationnels et d'autre part est techniquement visuel respectent les regies ergonomiques principales
realisable. II apparait ainsi necessaire que le travail de de la cabine du Rafale. Les pilotes peuvent done se
definition realise puisse etre represente et utilise par les pilotes, prononcer sur l'utilisation du Systeme ä partir de son
d'oü le besoin d'une simulation temps-reel. Ceci d'autant plus image informatique animee en temps reel et juger des
que la documentation traduit mal pour les pilotes les aspects de retouches (simplification, optimisation) de la definition
superposition des fonctions operationnelles, mises en oeuvre initiale qui evitent de coüteuses remises en cause
par des fonctions techniques. ulterieures,
L'approche sur le dimensionnement et la faisabilite technique • les ingenieurs concepteurs : la modelisation temps reel
est initialisee par une premiere estimation des ressources du Systeme respecte le decoupage en equipements et en
techniques et budgetaires necessaires ä sa realisation. Les grands modules logiciels (d'oü la denomination du
allocations qui en resultent sur les ressources les plus critiques Centre d'integration Hybridable), ce qui permet aux
(charges de calcul, volume memoire et debit d'informations sur ingenieurs de s'assurer des ordres de grandeur des
les bus) sont ensuite affinees et recalees par un rebouclage principaux dimensionnements,
permanent. Un accord entre Dassault Aviation et les • les responsables etatiques du programme qui, grace ä la
realisateurs du logiciel embarque permet d'exercer un controle disponibilite d'une simulation similaire au CEV (cf.
de ces ressources. Le produit final qui sera utilise par les paragraphe 4), disposent avec une anticipation
pilotes operationnels est ainsi le resultat de choix et de süffisante des elements de choix des compromis
compromis; il faut done permettre aux pilotes, aux ingenieurs techniques et budgetaires.

VISION AVANCEE SYSTEME

VISION A TERMINAISON
( DOSSIER DE CONSTRUCTION SYSTEME )
E
Q
U
I \
P STANDARDS UTILISATEURS
E REMARQUES
IMAGE DE REFERENCE
( DEFINITION GLOBALE ) ::::;:
::
::v:;:v.I:.:; :::::::
I
N
T
E
G
R
E
E
DEFINITION DETAILLEE
REALISATION
ESSAIS

Figure 1. Contribution des simulations au processus de definition du SNA

 20-3

3.4 Le travail en equipe integree 4.2 Röle de Dassault Aviation

L'etablissement, dans les phases amont du programme Architecte industriel du programme Rafale, la societe Dassault
(environ dix ans avant le terme de son developpement), de la Aviation a la responsabilite de la coherence d'ensemble du
Reference du Systeme ä terminaison necessite de s'assurer de Systeme d'armes. Elle est responsable de la definition de la
son adequation aux besoins des pilotes operationnels. Ainsi des cabine et de l'interface homme-systeme. Elle doit maitriser
le travail de conception ceux-ci sont associes aux choix de cette definition (adequation au besoin, respects des coüts et des
base; au fur et ä mesure que la definition se precise et en delais) et notamment s'assurer en permanence de la faisabilite
particulier lorsque des compromis s'imposent pour tenir de ce qui est propose. Dans cette optique, Dassault Aviation
compte de la faisabilite technique, leur avis est indispensable. s'est dote de moyens et a mis en place des methodes
La simulation est l'outil privilegie de ces differents travaux et particulieres dont les enjeux ont ete presented en premiere
comme nous le verrons par la suite, le travail en equipe partie de ce document.
integree, tantöt sous responsabilite industrielle, tantot sous
responsabilite etatique, est systematiquement mis en oeuvre. 4.3 Role du CEV
De facon generale, le CEV est au sein de la Delegation
3.5 Approbation de la definition avant realisation des Generale pour l'Armement (DGA) l'organisme expert pour les
materiels et des logiciels essais en vol ainsi que la mise au point des materiels aeriens et
La vision ä terminaison est plus ou moins precise suivant systemes embarques militaires. Le CEV dispose d'un centre de
qu'elle correspond ä des fonctions dejä maitrisees sur d'autres simulation, situe sur la base d'Istres, au coeur des essais en vol
avions. Elle va cependant permettre d'etablir l'organisation du industriels et etatiques, ce qui permet de profiter d'une
Systeme : utilisation nominale des ressources par chacune des synergie apportee par la confrontation entre travaux de
fonctions et regies de priorites entre elles. Une fois ces grandes simulation et vols d'essais.
lignes tracees, une definition globale est etablie pour les Dans le cadre du programme Rafale, le CEV est responsable
fonctions dont la realisation est demandee en priorite dans les d'une partie des travaux relatifs ä l'interface homme-systeme
differentes versions successives du Systeme: les standards relevant de täches etatiques. En particulier, il assure la synthese
utilisateurs. Pour chacun d'eux, une image de reference de la des avis formules par les representants de l'Etat et est
definition est realisee. Celle-ci est systematiquement soumise ä responsable des evaluations en vol et en simulation pilotee. A
l'approbation de l'Etat avant realisation des materiels et des cette fin, des simulations Rafale ont ete mises en place au
logiciels par les industriels cooperants du programme (voir CEV, et ce des le debut du programme. Profitant de moyens de
Figure 1). simulation importants et de sa situation au coeur des essais en
vol, cette simulation etatique permet un travail approfondi en
contexte operationnel et une utilisation intensive en soutien
4. CONSTRUCTION ET EXPLOITATION DE LA aux essais en vol. Elle est un des outils ä la disposition de la
SIMULATION - SYNERGIE DASSAULT direction de programme pour la maitrise du developpement.
AVIATION - CEV L'utilisation complementaire des differents moyens de
simulation du programme dans le processus de developpement
4.1 Introduction Systeme peut-etre resumee sur le schema presente en Figure 2.
Nous l'avons dejä precise plus haut, dans le cas d'un Systeme
complexe comme le SNA Rafale, I'analyse du besoin 4.4 Methode de travail et organisation des travaux
operationnel et l'adequation du produit final ä celui-ci sont des Pour rendre plus efficace le dialogue entre l'industrie et l'Etat,
defis majeurs. L'Etat et l'industrie doivent y repondre, chacun le travail en "equipe integree", tant au niveau des ingenieurs
dans son domaine de competence, en privilegiant le dialogue et que des pilotes, est privilegie pour l'ensemble du programme
la collaboration constructive. En particulier, l'industriel Rafale, sans pour autant diluer ou confondre les responsabilites
responsable de la definition du SNA ne pourrait pas se passer de chacun. La simulation pilotee, image dynamique du Systeme
de la plus-value apportee par les futurs utilisateurs dans les d'armes et outil de dialogue privilegie pour les principaux
phases de conception, notamment au niveau de la redaction des acteurs de la definition n'echappe pas ä cette regie.
specifications globales servant de base ä la realisation, et Pour memoire, les essais en vol du programme Rafale, font
chaque fois que des choix importants sont ä faire, surtout dans l'objet d'un protocole signe entre l'industrie et l'Etat, qui
les domaines touchant ä l'emploi du Systeme. prevoit sa participation active des la phase de mise au point
Pour mettre en forme cette plus-value, une organisation des chez le constructeur.
travaux de definition a progressivement ete mise en place, avec De la meme facon, une organisation des travaux de simulation
le concours des partenaires industriels et etatiques. En pilotee prevoit un fonctionnement en "equipes integrees" ä
particulier, le CEV et Yarchitecte industriel travaillent avec une deux niveaux:
synergie que nous allons preciser dans cette seconde partie, en • coherence technique simulateurs (Equipes
s'interessant plus particulierement aux travaux relatifs aux CEV/Dassault Aviation/cooperants)
simulations pilotees. • travaux de definition Systeme (Equipes Etat/Dassault
Tout d'abord, nous rappellerons les roles respectifs de Dassault Aviation)
Aviation et du CEV, puis nous decrirons la methode de travail. Un ensemble de documents definit les responsabilites
Le developpement et l'exploitation des simulations pilotees (notamment pour la conduite des essais, la gestion courante et
seront ensuite exposes de facon plus precise. les servitudes) et les regies applicables.
20-4

Place des simulations pilotees

Etudes de concept
dans le developpement d'un SNA

Evaluation
Besoins operationnels operationnelle

\ /
Specifications Essais en Vol
fonctionnelles
Simulateurs CEV
Simulateurs d'etudes
industriels et CEV /
(Roles complementaires)
\ Integration fonctionnelle
Conception architecture Mise au point/developpement
materielle et logicielle

\
Conception des materiels Integration partielle
et des logiciels embarques des sous-systemes

Figure 2. Simulations pilotees et developpement des systemes

La coherence technique et calendaire des travaux de simulation de l'emploi du Systeme dans les phases de reflexion et de
est assuree par la mise en place d'une equipe integree definition preliminaire. Le CEV est mandate pour coordonner
"simulateur" reunissant des membres des differents organismes ces activit.es et en faire la synthese. En particulier il anime le
(CEV, architecte industriel, cooperants). Elle s'assure de "groupe pilotes" compose de representants du CEV et des
l'adequation de I'etat de simulation aux besoins des differents forces, designes pour participer au developpement du SNA
utilisateurs industriels et etatiques, et permet d'eviter les Rafale. Si une partie de ces travaux se realise sur les
surcoüts que pourraient entrainer des doubles developpements simulateurs "d'aide ä la specification" de Varchitecte industriel
et de garantir une optimisation de l'emploi des moyens de tels que le CIH, l'exploitation formelle des simulations des
simulation du programme en respectant les calendriers prevus. etats du Systeme est realisee sur les simulateurs du CEV
Un travail important a notamment ete fait pour reduire les (presentations et evaluations).
delais de transfert des logiciels des sites industriels vers les Le processus de construction puis d'exploitation des
sites etatiques. simulations pilotees est schematise en Figure 3.
Les travaux de definition Systeme necessitent la consultation Nous allons le decrire plus precisement dans les paragraphes
par Varchitecte industriel d'experts etatiques dans le domaine qui suivent.

Sites d'integration
Sites de developpement et simulateurs Simulateurs CEV
Sites
industriels de I'Architecte industriel

Validation Avis de
industrielle I'Etat

Developpement
modeles SNA
Integration
> et mise au point
Avis
Pilotes
Portages
Integration Mise au
1, PRESENTATIONS
et
EVALUATIONS
1, Utilisations
utterieures
point
Enrichissement: Pilotee
- Moyens d'essais
Travaux
^O^ - Contexts Ops
effectues >^ /\
Developpement V^D
modales öquipements
et armements

dont Env tactique

Responsabilite Industriels CEV/Industriels CEV

Industriels
technique

Industriels puis Equipe integree

Participation Equipe integree Operationnels
Equipages Equipe integree

Figure 3. Developpement et utilisation des simulations Rafale

 20-5

4.5 Definition des moyens de simulation Le souci de maitrise de la faisabilite technique du Systeme
Les choix en matiere d'architecture et de moyens de (charge de calcul, volume memoire, debit sur les bus) a
simulations sont guides par plusieurs considerations : egalement conduit ä choisir des moyens sur lesquels la
• compte tenu des roles respectifs de Dassault Aviation repartition des differents modeles soit representative des
et du CEV et de leur complementarite, un travail de equipements du Systeme reel.
definition du besoin en simulation est mene en L'hybridation (remplacement de modeles par des equipements
commun. II est initialise par Dassault Aviation apres reels) etant envisagee, le respect des echanges entre modeles et
identification du contenu des specifications techniques la conformite aux interfaces des equipements sont recherches
de besoin et des niveaux de mise au point envisages; il pour faciliter cette evolution ulterieure (voir Figure 4).
est discute et valide par le CEV et complete par
l'examen de sa compatibilite avec les moyens 4.5.2 Choix des moyens
d'environnement du CEV (environnement englobant Dans le respect de l'architecture definie, les moyens de
les materiels choisis ainsi que les logiciels de structure Simulation sont choisis en fonction essentiellement de deux
d'accueil et certains logiciels applicatifs). criteres : performances (charge de calcul, volume memoire et
• les moyens de simulation du programme font gestion des entrees/sorties) pour les calculateurs et
egalement l'objet d'une «Vision ä Terminaison» representativite par rapport aux equipements reels pour la
technique et budgetaire. Les choix sont guides par la cabine.
volonte de respect des principaux standards Les travaux menes etant de nature differente, les materiels
industriels: Unix, VME, Ethernet,... de facon ä retenus peuvent etre differents chez Dassault Aviation et au
beneficier des compatibilites ascendantes des materiels CEV (Cabine et environnement). En effet, les travaux menes
et des logiciels necessaires ä la realisation et ä la mise sur les simulateurs CEV necessitent une tres bonne
en oeuvre de la simulation pilotee. Ils sont egalement representativite de la cabine, pour obtenir un jugement
coherents des moyens utilises par les cooperants pertinent sur 1'interface homme Systeme. Dans la mesure du
industriels pour developper et exploiter les modeles possible, des equipements reels sont utilises (manche et
qu'ils developpent pour cette simulation. manette, visualisations, interfaces physiques...) et
l'encombrement relatif prevu est strictement respecte. D'autre
4.5.1 Architecture part, pour restituer des conditions d'emploi operationnel lors
Le besoin de representativite de la simulation en termes des evaluations, le CEV met en oeuvre des moyens de
d'interface homme Systeme et de temps de reponse du Systeme Simulation tres importants (visuels et bases de donnees, spheres
entraine le besoin de representativite de l'architecture de la de combat, serveur de simulations...). A titre d'exemple les
Simulation par rapport ä celle du Systeme reel. En particulier, le moyens mis en place pour l'etude des conduites de tir
respect du decoupage fonctionnel et de l'architecture video est multicibles sont decrits dans le texte d'une conference
important pour permettre de juger la configuration de la cabine presentee ä l'AGARD en meme temps que celle-ci. Ces
selon la selection des fonctions et les choix effectues en differences entre les moyens Dassault Aviation et CEV sont
preparation de mission par le pilote. identifiees de facon ä maitriser les adaptations necessaires au

CIH = PROJECTION MATERIELLE DU SNA

CALCULATEUR DE SIMULATION
GESTION
POSTE
OENERATION SIMULATION RESEAU MIDS
ENVmONNEMEMT D'ENVIRONNEMENT
CAPTEURS MODELE
BOTTIER AVION
STOCKAGE RADAR
DONNEES

GESTION CONTREMESURES
ECHANGES
BUS opmowouE
POSTE CALCULATEUR
SURVEILLANCE CALCULATEUR
ELABORATEUR
PARAMETOES
DE
BOTHERS TRAJECTOIRE
MTERFACE
CABINE

VISUALISATION VISUALISATION VISUALISATION VISUALISATION

DESUVI LATERALE TETE TETE LATERALE
DE LA MISSION BAUCHE MOYENNE HAUTE DROHTE

Figure 4. Representativite de l'architecture de simulation vis-ä-vis de celle du SNA

20-6

moment des transferts de logiciels (« portages »). Compte tenu equipements reels. L'outil correspondant permet d'une part la
de cette demarche et de l'etablissement de regies de portabilite, production des specifications detaillees et d'autre part celle
les durees des portages de simulation complete entre Dassault d'un code executable sur des generateurs de symboles
Aviation et le CEV sont faibles. industriels, ce qui permet de l'utiliser en simulation, que ce soit
avec des equipements de visualisation simules, ou reels dans
4.6 Realisation des modeles certains cas (au CEV notamment).
L'architecture de la simulation est definie en tenant compte des
fonctions operationnelles ä modeliser et de la representativite 4.6.3 La modelisation de la preparation/restitution de
demandee. Celle-ci correspond ä la projection de la definition mission
et ä son decoupage fonctionnel. Deux types de modeles Dans le prolongement des differents programmes precedents,
peuvent etre distingues: les modeles « capteurs », auxquels il mais avec une ampleur renforcee par la polyvalence du
faut associer la modelisation de I'environnement tactique et les systeme et la superposition des fonctions qui en resulte, la
modeles du « coeur systeme » correspondant ä la partie du preparation/restitution de mission effectuees au sol joue un role
SNA relative aux calculateurs de mission et ä l'interface pilote- tres important dans le SNA du Rafale.
systeme. Ceci revient ä dire que celui-ci possede un segment bord mais
aussi un segment sol constitue par la preparation et la
4.6.1 Les modeles de capteurs et environnement tactique restitution de mission, et dont la conception ne peut etre menee
Pour concevoir le systeme d'armes, la connaissance des independamment l'un de l'autre. Dans l'attente de disposer du
capteurs et des armements (classe de performances, moyen qui sera utilise dans les forces armees, un systeme de
caracteristiques, mise en oeuvre) est primordiale car eile preparation/restitution des donnees necessaires aux essais en
conditionne fortement leur utilisation ainsi que la presentation simulation est developpe comme les autres modeles. Durant la
de leurs informations et de la synthese qui en est presentee ä mise au point puis 1'evaluation de la definition, il est en effet
l'equipage. D'autre part, aucun biais ne doit apparaitre dans le necessaire de pouvoir modifier les scenarios d'essais mais
jugement porte sur l'utilisation du systeme ä cause d'une aussi la preparation/restitution de mission, pour la rendre
modelisation erronee ou trop simpliste de ces equipements. coherente avec le scenario choisi. II est ainsi possible de
Une attention particuliere est done portee sur leur montrer aux pilotes les parametres modifiables soit au sol, soit
modelisation. Les modeles sont developpes par les industriels en vol; ces choix fondamentaux ont en effet des repercussions
cooperants responsables de la realisation des capteurs ou importantes sur le dimensionnement des ressources du SNA.
armements reels, ä partir des specifications de besoins etablies
par Dassault Aviation et le CEV, en fonction des objectifs des 4.7 Integration - mise au point
essais en simulations pilotees. Les modeles sont developpes par les cooperants industriels et
II convient de noter qu'il existe une forte imbrication entre la par Dassault Aviation, dans le respect des regies de portabilite
modelisation de ces elements et celle de I'environnement qui sont un des points cle de la reussite du developpement de
tactique (SNA amis ou ennemis, modeles de terrain, la simulation. Comme le prevoit le planning general de
environnement atmospherique, ...). Pour celle-ci le CEV se realisation de la simulation globale, ces modeles sont ensuite
dote de moyens adaptes aux essais qu'il conduit, qui sont mis en commun au bureau d'etudes de Dassault Aviation sur
utilises par les industriels, et servent de reference pour les l'outil SAMOS (Structure Adaptee ä la Modelisation
specifications de modeles. Les responsables du programme Operationnelle des Systemes). Celui-ci est constitue d'un
Rafale ont confie au CEV le role particulier devaluation de la ensemble de stations de travail cablees en reseau et dont la
representativite des modeles par rapport aux specifications structure d'accueil assure l'integration des modeles et la mise
techniques, sachant qu'il a par ailleurs la responsabilite au point informatique. Cet outil dispose de moyens d'analyse
d'approbation du fonctionnement des capteurs de l'avion avant permettant de resoudre les problemes de coherence et de
leur integration au systeme d'armes de l'avion reel. Des que completude de la definition dont la simulation est un bon
cela est possible, le CEV confronte, grace aux outils et revelateur. II peut aussi mettre en oeuvre simultanement
mefhodes dont il s'est dote, les resultats des vols d'essais plusieurs simulations, ce qui permet de mener les differents
capteurs ou armements sur avions bancs d'essais ou travaux en parallele. Ses limitations sont d'une part un
prototypes, afin de recaler les modeles ou d'analyser les fonctionnement generalement non temps-reel (tous les
dysfonctionnements eventuels. Cela permet d'ameliorer en processus sont appeles de maniere sequentielle, mais ils sont
permanence la pertinence des resultats d'essais en simulation. executes plus lentement) et d'autre part l'absence d'ergonomie
puisque les commandes simplifiees sont tres differentes de
4.6.2 Les modeles du coeur systeme celles utilisees dans l'avion.
Ils sont developpes par Dassault Aviation qui a la Pour s'adapter ä l'eloignement geographique entre le bureau
responsabilite du decoupage fonctionnel et du d'etudes de Dassault Aviation et la base d'essais en vol et
dimensionnement du SNA. Ces modeles doivent etre permettre notamment le developpement et la mise au point de
representatifs et en particulier ceux destines ä simuler les l'interface Homme Systeme ä l'aide de la simulation, une
visualisations presentees au pilote qui a besoin de cette partie de ces travaux est menee ä Istres ä proximite immediate
representativite pour mettre au point la definition. Ainsi, des la des pilotes d'essais, par une equipe specialisee sur une cabine
conception, un decoupage precis et detaille entre les images et representative; il s'agit d'OASIS (Outil d'Aide ä la
les traitements est realise conformement ä celui de l'avion. En Specification d'Interfaces Systeme).
partenariat avec l'industriel charge de la realisation du logiciel Lorsque la simulation est conforme ä la definition, eile est
embarque, la definition de ces informations est saisie dans un alors transferee au CIH dont 1'architecture et les moyens
formalisme informatique adapte ä sa realisation dans les assurent un fonctionnement temps reel de la simulation et sa
 20-7

mise en oeuvre avec une interface dont l'ergonomie est 4.9 Autres utilisations
representative. Cet outil, comme nous I'avons vu, facilite le Les simulations du CEV utilisees dans les phases de definition
travail en commun des differents specialistes du SNA avec les du Systeme peuvent evoluer et repondre ä d'autres besoins
pilotes (industriels dans un premier temps, puis equipe integree relatifs au programme:
Industrie/Etat), pour analyser les difficultes rencontrees et
mettre au point la definition par iterations successives, ä l'aide • les etudes de concept ponctuelles liees ä des
de la simulation dans le souci constant de la faisabilite complements de developpement, ou amendements de
technique. la definition
Enfin, une fois la simulation representative de la proposition • les activites de soutien aux essais en vol et d'aide ä la
de Varchitecte industriell eile est transferee sur les simulateurs qualification
du CEV, ou eile est enrichie des moyens de simulation
etatiques, et preparee pour une utilisation en contexte Dans le cas des etudes de concept (confiees ä des partenaires
operationnel. L'etat de simulation final fait l'objet d'une industriels), le processus de developpement de l'etat de
« validation industrielle », realisee par Dassault Aviation et les simulation, puis d'exploitation des simulations est tout ä fait
cooperants, ce qui garantit la reconnaissance des resultats similaire ä celui que Ton vient de decrire, la responsabilite des
obtenus pendant les phases de presentation puis d'evaluation. evaluations finales restant du ressort du CEV.
Le besoin d'utilisation des simulations pilotees en complement
4.8 Presentation - evaluation des vols d'essais a ete mis en evidence et analyse par le CEV.
Deux grandes phases d'utilisation de la simulation pilotee pour Cette analyse decoule notamment de l'experience acquise dans
chaque etat de definition du Systeme devant effectivement etre ce domaine sur les programmes precedents (en particulier dans
realise (etapes intermediaires et standards utilisateurs) sont le cadre du developpement du Mirage 2000 ou des etudes
prevues : generates sur le multicible). Le soutien aux essais en vol
repose sur la capacite de disposer d'une simulation du SNA
• les presentations officielles par l'lndustrie, conforme ä l'etat des logiciels embarques, completee de
moyens d'analyse harmonises avec les outils d'exploitation des
• les evaluations etatiques.
vols d'essais, ce qui permet:
La presentation est le processus au cours duquel l'etat de
• de preparer efficacement les vols d'essais, pour en
definition du Systeme est presente ä l'Etat par Varchitecte
diminuer le nombre (entrainement des pilotes,
industriel pour approbation avant realisation des logiciels
definition des profils ou points d'essais, repetitions,...),
embarques. La simulation pilotee joue le role d'une image • d'analyser plus completement les vols d'essais,
dynamique des specifications qui permet aux pilotes etatiques
notamment par la capacite de "stimulation" (injection
de porter un jugement sur la fourniture, ce qui ne serait pas
dans une simulation de donnees issues de vols reels), et
possible par une simple lecture critique de documents. Cette eventuellement d'etudier des propositions de
presentation a normalement lieu sur les simulateurs du CEV modifications,
(sans delai par rapport aux travaux industriels, grace au travail • de proposer une extension du contexte effectivement
technique des equipes integrees "simulateurs"). Le contröle de
accessible aux cours des essais en vol en fonction des
la representativite de la simulation et la Synthese des avis des moyens disponibles ou de l'avancement reel du
representants de l'Etat sont de la responsabilite du CEV. A la
developpement des differents equipements ou
suite de ces travaux, un etat accepte des specifications est
armements (aide ä la mise au point puis ä la
elabore, en concertation avec les differents intervenants. qualification).
L'evaluation est un processus purement etatique, grace auquel
l'Etat peut porter un jugement approfondi sur la fourniture
5. CONCLUSION
industrielle (expertise amont pendant la definition et la
A Tissue de cette presentation qui a permis d'exposer la
realisation du Systeme). Au cours de cette phase, le CEV etudie
demarche utilisee pour concevoir le SNA du Rafale, notre
l'emploi du Systeme dans des conditions les plus proches
conclusion est destinee d'une part ä etablir un bilan global des
possibles des missions operationnelles prevues, parfois avec
resultats de cette demarche, et d'autre part ä degager des
des donnees confidentielles vis ä vis de certains des industriels.
perspectives d'avenir sur le role de la simulation.
D'autre part, ä l'aide des specialistes du domaine, il analyse les
En premier lieu, il est important de souligner que le
problemes lies ä la charge de travail de l'equipage sur des
programme Rafale se deroule conformement aux previsions, ce
scenarios d'emploi operationnels realistes. Les resultats de ces
qui valide les efforts consentis et les innovations adoptees : la
evaluations peuvent se traduire par:
definition generate, construite et mise au point grace ä la
• une prise en compte des remarques dans des etats Vision Avancee Systeme, repond aux besoins exprimes et le
ulterieurs du Systeme, premier standard utilisateur va entrer en service, dans le respect
des coüts et des delais. Les travaux en equipe integree des
• des actions de correction de la definition en cours de equipes d'essais ont permis d'aboutir ä des resultats tangibles.
realisation (differents cas contractuels). C'est ä la fin de 1992 qu'a ete presentee la Vision ä
Par l'intermediaire de ces deux phases d'utilisation des Terminaison, constitute ä partir d'une premiere definition de
simulations, le CEV formalise ainsi vis ä vis de l'lndustrie une l'ensemble des fonctions du SNA Rafale, avec toutes les
prise en compte des besoins operationnels dans le domaine de capacites Air/Air et Air/Surface. Au debut de 1994, la
l'emploi du Systeme, le plus tot possible avant realisation simulation du premier standard utilisateur a ete presentee puis
definitive, mais en fondant les avis sur une image dejä bien evaluee et la definition correspondante finalisee et approuvee
avancee du produit final. par l'Etat, permettant ainsi le lancement de sa realisation.
20-8

II faut egalement signaler la qualite des resultats techniques differents standards du SNA Rafale devront etre essayes en vol
obtenus dans les travaux en equipe integree pour definir et et qualifies. L'experience acquise lors des developpements des
construire la simulation pilotee. Ceux ci se traduisent avions de combat precedents a montre tout l'interet d'une
notamment par des delais de portage totalement maitrises utilisation intensive des simulations pilotees lors des phases
puisqu'ä l'heure actuelle, des simulations dont les modeles d'essais en vol et de qualification des systemes d'armes
representent environ 300000 lignes de code et sont interfaces aeroportes. C'est ce qui a conduit le CEV ä mettre en place et
par un dictionnaire de 10000 variables, sont transferees en une faire evoluer ses moyens de simulation, ainsi qu'ä definir les
duree inferieure ä une semaine. De meme, des developpements outils et methodes de travail pour completer les travaux
ou des specifications sont menes en commun pour d'une part d'essais en vol et en ameliorer l'efficacite.
beneficier de 1'experience et des competences reciproques et Les enjeux techniques du programme Rafale (complexite et
d'autre part faciliter les transitions entre les travaux de mise au haut niveau d'integration du Systeme, polyvalence, innovations
point et ceux d'evaluation. Enfin, la coherence et techniques pour les equipements...) et la complexite de la täche
l'homogeneite des moyens de simulations industriels et de qualification necessiteront probablement une evolution des
etatiques sont une realite. Les differences, tenant compte des techniques employees pour la simulation. En particulier, le
roles respectifs et de la nature des travaux effectues, ont ete recours ä une simulation pilotee hybride, c'est ä dire integrant
clairement identifiees et sont prises en compte pour faciliter dans des proportions variables des equipements embarques
l'integration des modeles. avec leurs logiciels est envisage pour repondre aux besoins de
Ces resultats engagent les responsables du programme ä la «branche montante» du V methodologique de
poursuivre les travaux avec la meme methodologie pour la developpement du Systeme.
suite de la definition du Systeme Rafale. En definitive, simulations pilotees et organisation originale des
Mais nous l'avons vu, l'utilisation des simulations pilotees ne travaux entre l'Industrie et l'Etat concourent de facon
s'arrete pas une fois la definition obtenue et approuvee. Les essentielle ä la reussite du programme Rafale.
 21-1

ARC SEGMENT ATTITUDE REFERENCE - ASAR

Simulator Application during the Development Process of a

New Attitude Reference Symbology

W. Fuchs and G. Fischer

Daimler-Benz Aerospace
Department LME 51
D - 88039 Friedrichshafen
Germany

1. SUMMARY The first ASAR Version presented 1987

Pilots of modern fighter aircraft with high pitch and extremely A circular arc segment was used as pitch reference and the
high roll rates complain about fast moving and twisting pitch angle between aircraft symbol and center of the arc as roll
ladders on the head-up display (HUD) and the necessity to reference (Fig. l).The dimension of the arc is defined by either
virtually concentrate their entire attention on the attitude 180 degrees minus pitch angle (9) times two (180° - 2 6) or
references in order to maintain orientation in space when flight path angle (y) (180° - 2 y).
performing aggressive three-dimensional maneuvers.
The first ASAR version, presented to pilots, consisted of two
For this reason we were considering a more stationary flight concentric arc segments with center and end markings. Fig. 2
attitude symbology to allow for attitude awareness, which shows horizontal flight where a 180° arc is displayed below the
requires less attention and concentration. The result was the aircraft reference symbol. The dimension of the arc segment
introduction of the Arc Segment Attitude Reference (ASAR) therefore will decrease with increasing positive angles for 0 or
Symbology. Y or increase with negative angles for 9 or y.

After a number of alternating simulator trials and flight tests for The final ASAR Version in 1994
further development and refinement of the symbology, a final
version was agreed and already implemented in experimental After 3 simulator trials and 2 flight test evaluations a final
programmes. version was agreed and specified in 1994. As shown in Fig. 3
additional dots and gaps have been introduced to mark pitch
angles of 9 (y) = + 30 degrees and ± 60 degrees. The dots of the
2. PREFACE lower segment portion remain displayed for angles above the
horizon in order to improve the identification of the semi-
The Key Role of the Development Simulator circular shape.

In 1987, the ASAR symbology, was demonstrated for the first A dash mark was introduced at the outside of the segment for
time to flight test and operational pilots during simulator trials better roll reference and a triangular mark at the segment center
aimed at improving the attitude reference symbology for the highlights the direction of the line perpendicular to the earth
Alpha Jet aircraft. surface.

The simulator demonstration triggered German MoD funding

and was the kick-off for further development and flight tests. 3. THE KEY ROLE OF THE SIMULATOR
The change requests and recommendations derived from the
airborne evaluations were programmed in the HUD computer Close interaction between simulator trials and flight test
symbol generator of the simulator, evaluated and prepared for evaluations were required during the development process of
the next flight test portion. the new attitude symbology.

The simulator equipment was continuously improved to meet

EF 2000 standards in hardware and software. Therefore rapid 3.1 The Initial Push
implementation of advanced mission phase symbology,
including sensor and weapon indications in combination with In 1987 simulator trials for improving the Alpha Jet HUD
high maneuvering potential, became available and allowed the symbology were performed in the Dornier Simulator. As an
extension of the symbology evaluation beyond the capabilities add-on demonstration we offered the segment attitude
of the Alpha Jet flight test aircraft. reference, and the pilots who were very skeptical and reserved

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?"
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
21-2

during the static demonstration ...."leave us alone with new • A/A weapon envelopes
gimmicks"... changed their opinion totally after a few minutes • Approach and landing guidance
of dynamic maneuvering in the simulator. • A/G LOFT maneuvers

"This is a promising symbology; we want to see it in flight and combined the EF 2000 mission phase symbology with the
test". ASAR. The pilots were able to switch without interruptions or
delay between pitch ladder (PL) and ASAR symbology
including the mission phase symbols during maneuvers in any
3.2 First Flight Test Evaluation in 1989 mission phase.

The ASAR version as shown in Fig. 2 was programmed for the For mission phases which do not require dynamic maneuvering
HUD of a prototype Alpha Jet assigned to the German Armed but precise pitch references a simplified PL was added to the
Forces Flight Test Center. ASAR as primary mode.

60 flights with 16 flight test and GAF pilots involved were 12 pilots were involved in this simulator trial (8 test pilots + 4
performed with this symbology. The outcome of this operational pilots).
assessment was positive, with a number of recommendations
and change requests to be incorporated, tested and agreed in the The test pilots refused to apply rating scales, like Cooper-
simulator prior to the next flight tests. Harper or modifications thereof, stating that the tasks were too
complex with too many parameters involved, so that standard
ratings scales could not be applied. We ended up with around
3.3 Second Simulator Trials in 1990 40 pages of written comments.

All change requests and recommendations which arose during Task set-up for this simulator trial
flight test were integrated as far as possible. The simulator trial
was conducted with 3 flight test and 6 GAF pilots involved. • Evaluation of the latest changes to ASAR
• Evaluation of Unusual Attitude Recovery (UAR)
There was unanimous agreement that the symbology was • Evaluation of A/A-symbology in combination with PL or
superior for air-to-air (A/A) combat, coarse maneuvering and ASAR
unusual attitude recovery (UAR) but needed further flight
• Evaluation of A/G-symbology (LOFT maneuvers) in
testing and possibly refinement for low level navigation, air-to-
combination with PL or ASAR
ground (A/G) and instrument flight applications. There were
also some reservations concerning HUD clutter with ASAR and • Evaluation of basic flying phase symbology in combination
mission symbology combinations. with PL or ASAR + PL for Approach, Landing, Take-off
and IFR-Navigation.

3.4 Second Flight Test Evaluation in 1991/1992 Simulation Results

The basic symbology which was developed during the second • Basic ASAR symbology as demonstrated was accepted and
simulator trials was also flight tested with an Alpha Jet aircraft frozen.
with 40 flights and 13 pilots involved (5 flight test and 8 GAF).
The final report was available in early 1993. Again, the • Unusual Attitude Recoveries were accomplished faster with
common opinion was: The symbology is better than the pitch ASAR, up to 30 % for pilots with previous ASAR
ladder for A/A, UAR and attitude awareness with some experience
reservations concerning low dynamic flights and HUD clutter
as soon as complete mission phase symbology is added to the • A/A symbology with ASAR was rated superior to the PL,
ASAR. The latter was already commented during the preceding even pilots who were skeptical before changed their mind
simulator trials.
• A/G symbology with ASAR better or equivalent to PL
At that point we reached the limits of the Alpha Jet as test
aircraft. Because of the HUD with limited field-of-view (FOV) • Basic instrument flying symbology better or equivalent to
and the lack of airborne radar, advanced sensor and weapon ASAR + PL compared to PL only.
symbology could not be generated.
Overall opinion of the pilots was that the ASAR symbology
was better than PL for all mission phases for HUD and helmet
3.5 Third Simulator Trials in 1994 mounted display (HMD) applications.

At this time we had our EF 2000 research cockpit and aircraft

software available to simulate the symbology in a highly 4. PRESENT ASAR APPLICATIONS IN R&D
dynamic environment. PROGRAMMES

We prepared computer models for: 4.1 X-31 Research Aircraft

• maneuvering targets The ASAR and the THETA-symbology is used on the X-31
• Radar A/A combat modes HUD for attitude reference.
 21-3

The THETA symbology is a three-dimensional globe-type

display with the aircraft symbol in the center.

The test pilots of this programme favour the THETA display,

but this aircraft has no mission phase symbology which, most
probably, would totally clutter the THETA display.

4.2 Tornado Helmet Mounted FLIR Display

In this German MoD funded research programme the ASAR

symbology is used for attitude reference in the HMD in
combination with the FLIR picture.

According to the draft of the final report the test pilots are
happy and feel comfortable with this flight attitude reference.

5. FURTHER PLANNING

The next simulator trial is planned for June/July 1995 after the
latest EF symbology changes have been programmed. Presen-
tation to EF 4 nations Cockpit Assessment Working Groups
pilots is planned by end of the year with a HMD integrated into
the simulator cockpit. On this occasion we also intend to
integrate angle-of-attack reference with the ASAR using a
variable aircraft symbol.
21-4

Fig. 1 Arc Segment Attitude Reference - ASAR Explanation

180° -29 or
180°-2y. /'
Pitch Reference

Roll Angle

Fig. 2 First ASAR Version -"The Orange Peel"

-A-
m
Horizontal Straight Straight
Flight up down
 21-5

Fig. 3 Final ASAR Version

.-,/

Level Flight y =0° with roll

0 = 0° marks I 0 = 30°
\ /

\ /
/ \

y =-45° y =-75°
0=0° 0=0°

y =80°
0=0" 0=0°

y =45° y =-45°
0 = 45° 0 = -60°
 22-1

Issues in the Flight Clearance of Vehicle Management Systems

DJ Diston & BRC Weiler
Systems Engineering Research and Development,
British Aerospace Defence Limited
Warton Aerodrome,
Preston, Lanes. PR4 1AX
United Kingdom

The main challenge concerns the safety case and, therefore,

Summary the flight clearance of a software-controlled system which is
complex and highly integrated.
The adoption of highly-integrated and complex aircraft systems
will facilitate the development of optimum strategies for vehicle This paper will consider the principal issues in the flight
management. This will provide benefits in functional perform- clearance of a VMS. It will be argued that clearance depends
ance and operational effectiveness. The drive towards stan- upon properties of determinism, robustness and testability.
dardisation implies the adoption of some form of modular This is influencing the research agenda in a number of key
avionic architecture, thereby adding the benefits of improved areas, especially A3 operating systems, analytical control
logistics. In this paper, it is argued that these developments theory and 'virtual rig' simulation.
force a fundamental review of existing approaches to flight
clearance. A number of challenges are proposed in the areas
of Operating Systems, Analytical Control Theory and 'Virtual 2 Concept of Vehicle Management
Rig' Simulation. In conclusion, the paper anticipates the evol-
ution of a "modular" procedure which might provide a legal
An essential feature of VMS philosophy is to provide an
basis for the certification of aircraft systems.
integrated control system which flies an Aircraft and provides
management of Stores (i.e. weapons, external fuel tanks and
so on) and Energy (i.e. fuel, power, cooling and so on). Overall
1 Introduction objectives include:
Vehicle Management System (VMS) concepts are accepted • 'carefree handling' (including high-g/, high-a,
as the basis for improving the performance of future aircraft. severe turbulence and engine surge)
Traditional airborne systems satisfy objectives which have • reduced pilot workload via vehicle automation
been specified within narrow functional boundaries. The new • rapid response and precision flying tasks
opportunity is to apply software control in order to coordinate (especially weapon delivery)
and optimise dynamic interaction and transient response • efficient airframe resource management
across all aircraft systems. In this way, new system objectives • environmental protection of installed equipment
can be pursued. For instance, energy efficiency can become This demands a unified approach to functional specification
a design driver whereas, hitherto, this particular property was and design which can be applied to all airborne functions. It will
unmeasured and uncontrolled. require a comprehensive method for analysing system
requirements and for assigning functions to appropriate
The scope of VMS functionality creates the dual problem of implementation methods. A high level of functional integration
complexity and diversity. Increased connectivity enables is assumed and, accordingly, the behaviour of each part of the
functions to exploit shared data but the open nature of the system will be dependent upon correct behaviour of other parts
architecture removes many of the traditional defences against of the system.
comprehensive fault propagation. Attempting to integrate
diverse functions introduces a range of interfacing problems.
Also, it becomes very difficult to establish the most appropriate 2.1 System Attributes
optimisation policy in the face of large-scale system models.
In the most general context, an airborne System provides the
An emerging factor for VMS implementation is the drive towards host aircraft and its users with an operational capability within
standardisation. This will lead inevitably to the adoption of some a prescribed environment. The limits of that capability are
form of Advanced Avionic Architecture (A3) although safety defined by an Operational Envelope which is appropriate to the
considerations will rule out the use of fully reconfigurable, fault current system configuration. This is a parametric description
tolerant computing. This aims to achieve cost reduction through of all reachable states of a system (e.g. position, velocity,
the provision of standardised modules, housed in modular temperature) and all constraints imposed by its environment
racks and connected by a communication network. Feasibility (e.g. structural strength, control power) or its user (e.g. g-onset
studies have been conducted under the Allied Standard leading to sudden loss of consciousness). Note that the concept
Avionics Architecture Council (ASAAC) and elsewhere, but a of reachability has meaning if and only if the system is
large amount of new work is required to how generic resources predictable and controllable.
can be managed by a distributed operating system and be used
for safety-critical applications. The organisation and interconnection of system elements is
called an Architecture. The type of element depends upon the
The principal benefits of VMS integration lie in four main type of design activity being conducted at a particular stage of
objectives: development and the type of partitioning which is being applied
• rational grouping of system functions to the system. Early in the development cycle, the type of
• unified approach to system implementation element is a function; during implementation, it will be a item
• flexible allocation of system resources
• scalable architecture which supports expansion

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?",
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
22-2

of hardware or software. The hardware elements which make units of execution for system functions and these will
up the system infrastructure, together with their management implemented as items of hardware and software. The eventual
software, are called Resources. implementation will comprise the Software Architecture and will
produce so-called applications from software modules (stan-
The behaviour of a system element has five basic attributes, dard services and specialised code). The structure of system
as follows: resources is defined by the Hardware Architecture. Note that
Operation The way in which the system element is used the user interfaces for the system are defined at the level of
to support the system operation. functional architecture and then mapped on to lower levels.
Task What the system element does when it is activated, The traditional model of system implementation is based upon
as described by a constitutive relationship and an a federated architecture of Line Replaceable Items (LRIs).
interface specification. Each has its own dedicated resources and can operate
Performance Speed and accuracy of response of the independently; integration is performed at the functional level
system element when it is activated, expressed as and is facilitated by a communication network. A simple
combinations of value and time. structure can be identified, as follows:
Resource The dependency of operation, task and per-
formance on system resources. DEVICES APPLICATION
Failure The effect of loss of system resources on the Line Replaceable Item
system element and the way in which such a loss can
NETWORK
occur.
This provides a complete operational definition, which can be A partition is applied down to the network level, according to
applied to system elements (either individually or collectively) whether functions are to be implemented using devices (i.e.
at any level of architectural decomposition. sensors, actuators, controls and displays) or applications (i.e.
software items). In this model the application software controls
Applying these attributes at system level, the following char- virtually all aspects of hardware (LRI) utilisation. The process
acterisations can be defined: architecture furnishes an LRI Processing Specification which
dictates the software construction for particular target hard-
System Type Attributes Requirement ware.
Real-time System Performance Deadline Timing The implementation model for an A3-based system is inter-
Control System Operation Continuous preted as a more elaborate structure, as follows:
Task Modify Environment
Performance Periodic Deadline Timing APPLICATION
plus Frequency Response APOS Interface
A Real-time System has timing constraints imposed such that DEVICES OPERATING SYSTEM
the system functions deliver a service within a known time
interval after they have been activated. Note that 'real-time' MOS Interface
does not necessarily mean 'high performance'. A Control MODULE
System is any real-time system which modifies the dynamic
behaviour of its environment. This is achieved by means of NETWORK
applied power which is continuously modulated as a result of
The rationale for this type of layered abstraction is the need for
operating commands and physical measurements.
hardware transparency atthe application level. This will support
(in principle) the use of different modules, perhaps incorpo-
2.2 System Abstraction rating new technology, without the need for software re-en-
gineering. The same partition is applied between devices and
applications, but the application mapping is structured by two
System design can be interpreted as a hierarchical abstraction, interface layers, which enforce standardisation.
as follows:
The purpose of each layer of the implementation model is
Functional Architecture
described, as follows:
Process Architecture • DEVICE layer contains functional mechanisms which
I Hardware Architecture Software Architecture embody mainly the analogue process definition of
the system and the interface between analogue and
The intention is to impose a structure on the information needed digital processes.
in order to develop a system concept into a system product. • APPLICATION layer contains executable software
The hierarchy contains a mapping between different archi- which embodies most of the digital process defi-
tectural definitions. Operational requirements are applied at the nition of the system.
top: implementation constraints are applied at the bottom.
• APOS Interface layer provides a virtual machine
Within the hierarchy, system requirements are passed top- interface for "application to operating system" (or
down and system constraints are passed bottom-up. APOS) mapping.
The behavioural model of a system is defined by the Functional • OPERATING SYSTEM layer contains a set of ser-
Architecture. The overall structure of resource utilisation, both vices which manage the process execution defined
to satisfy functional requirements and to provide operational in the application layer. The operating system
integrity under failure conditions, is represented by the Process 'owns' all system resources and all system time.
Architecture. This is the computational model which will dictate
how the system is to be built. Conceptually, processes are the
 22-3

• MOS Interface layer provides a transparent interface development methodology which is modular and can be
for "module to operating system" (or MOS) mapp- applied at different levels of product definition. This would
ing. support re-use of system components and a phased clearance
• MODULE layer contains processor and memory of new system configurations, starting with simulation and
resources and provides a physical interface to the progressing towards full implementation.
network.
At every stage of development, a modular approach is required
• NETWORK layer contains physical resources for data which can handle all issues of functional complexity at a
communication. detailed level of design and concentrate on integration issues
The particular value of this model is to convey a generalised at an aggregate level. This leads to a dual interpretation,
concept of system resources - partly hardware and partly namely [i] a set of interacting system elements or [ii] a network
software. Thus, the MOS will specify standard hardware of flowing information. A modular approach to system design
resources and the APOS will specify standard software is possible by decomposing the overall specification into
resources. In the process architecture, both types of resource individual specifications for elements of the system architec-
are available and the aim is to maximise their usage in order ture. A modular approach to system qualification is then
to minimise the amount of application-specific software to be possible such that each system element is shown to satisfy its
developed for any given system. For instance, where hardware own specification and the network connections provide a
redundancy management is required, it is possible that the traceable link between systems elements. The definition of
voter-monitor algorithms (which are currently treated as part of system architecture depends upon the level of abstraction that
the Application) will be incorporated as standard services is appropriate to the stage of system development. What is
within the Operating System. important is that traceable links can be demonstrated between
different levels of abstraction and that, within each level,
compatibility can be demonstrated.
2.3 VMS Implementation Requirements
Equipment standardisation and cost reduction are well-es-
The perceived benefits of systems integration are two-fold, tablished as system design drivers. The logistic gains which
namely improved functionality (performance, fuel economy, may be achieved from A3 are large, provided that safety issues
etc) and improved logistics (reliability, availability and main- can be resolved. It is clear that benefits can be realised by
tainability). This can be exploited in order to reduce design adopting a common architecture for VMS functions which can
margins (either to increase performance or to reduce cost) but, support the availability requirements for safety-critical appli-
as a result, the system safety assessment will be complicated. cations. The argument rests on pragmatism such that a
The critical trade-off for the customer will be cost-effectiveness, common module type could be specified with a level of
that is 'cost of ownership' versus 'operational effectiveness' at hardware isolation suitable for all critical control systems (with
acceptable risk. a compatible segregation policy for the backplane). This is less
ambitious than applying fault tolerant computing techniques
Achieving these benefits is the responsibility of the system and it will relax the operating system requirements. In this way,
designer. Typically, the design is conceived middle-out rather functional isolation can be supported by physical isolation
than top-down. This allows due weight to be given to existing where necessary, without stipulating logical or temporal iso-
engineering knowledge without being forced to use a "clean lation. The real requirement is for safe and secure isolation
sheet of paper" every time. It also provides flexibility in without excessive development cost or timescales.
reconciling technology selection, customer requirements and
commercial constraints. The most pressing priorities for A3 management concern the
scheduling and allocation of system resources. For VMS
At early stages of development, a system implementation has implementation, the principal aims are to achieve hard dead-
not yet been built and experience with real hardware and lines for software execution, distributed resource management
software is generally lacking. This is especially true for A3- and deterministic operation. The deadline policy must be hard
based proposals because even the hardware, software and real-time in order to satisfy the performance and safety
system management concepts have not been fully defined. requirements associated with the external environment of the
Also, as system complexity increases, the implementation system, especially the avoidance of hazardous events. Dis-
issues become much more difficult to grasp. For VMS inte- tribution is a requirement which might be derived from system
gration, the building blocks are quite large, complex subsys- architecture, software size or functional isolation. It is likely that
temsin their own right, e.g. propulsion, flight control, secondary the allocation of resources will be predominantly static and
power and environmental conditioning. Functional integration redundant, with some level of dynamic switching (e.g. to isolate
serves to bind these subsystems together in a manner which resource failures). Note that the need to share data between
hitherto was not possible. This in turn opens up new opportu- resources creates an 'internal' requirement for hard deadlines
nities for equipment design because a suite of equipment can on processing and communication because, otherwise, itwould
be configured ab/n/'f/bwith integration in mindand this becomes be impossible to satisfy 'external' hard deadlines. In all of these
an objective in the design optimisation. This is particularly true respects, operating system research will be critical.
in the area of engine thrust/power offtake, where new pack-
aging concepts are being developed for power generation. In summary, the benefits of increased efficiency and decreased
logistic support are central to VMS concepts. The challenge is
All of these factors point to considerable uncertainty in design to actually achieve these benefits within cost and safety limits.
and implementation of a future VMS. There is a dual problem
of complexity and diversity; the existing problem of handling
complex designs is likely to be more severe and be com- 2.4 MMI Requirements
pounded by new integration concepts which encompass a
much greater range of functionality than hitherto. This demands There are important questions which must be raised in con-
tools which are capable of simulation of large-scale diverse nection with the Man-Machine Interface (MMI) requirements
systems, at any level of abstraction. It also demands a for VMS. These concern the control/display modes which
22-4

support, firstly, an appropriate level of automation for particular These represent a large number of system design options and
functions and, secondly, the actual level of influence which can depend heavily on the concept of optimality being employed.
be exercised over particular functions. All of these modes are different and, more importantly, all of
these modes will be present in future aircraft (plus a few more,
Piloted simulation is an essential aspect of system clearance, no doubt). Perhaps what is really needed is a new breed of
particularly in the areas of flight control and cockpit instruments. super-human test pilot. Figure 1 suggests a possible long-term
The rationale for building expensive facilities and for supporting option although there is some concern regarding high-g
research into Virtual Reality is that a human operator will manoeuvres!
exercise a system simulation in a representative and
exhaustive manner if the psychological cues are believable. Serious questions must be asked when assessing current
Human factors are of importance, not only because the pilot is practice against future VMS requirements. Integration will yield
the ultimate end-user of an aircraft but also because subjective more combinations of functions than can be tested and much
assessment by an independent expert (i.e. a test pilot) gives of the functionality will be transparent to the user and, therefore,
invaluable insight into the adequacy of system design. Rating also transparent to a test pilot who is flying a simulator. Thus,
schemes exist for aspects such as handling qualities (cf. a new approach will be required for automated test procedures,
Cooper-Harper ratings) and these serve as the basis for which can identify what test coverage has been achieved and
comparing and contrasting different human reactions, thereby furnish objective evidence that the coverage is sufficient to
accommodating the type of variability that would be address safety requirements.
encountered in service. In addition, an independent expert can
play "devil's advocate" and, as such, prompt further develop-
ment of a system. 3 Concept of VMS Simulation
The new challenge which arises in a complex and highly-in- Vehicle management involves a wide range of airborne func-
tegrated system is the method of exercising system functions tions, integrated in an optimum manner. What is readily
in a way which has proved so successful for, say, flight control apparent is that extensive simulation will be essential in order
systems. Much of the VMS functionality will be automated and to furnish a proof of concept which will justify the large
the behavioural characteristics willdepend upon interfaces with investment required for development. Effectively, simulation
other functions. How then can effective simulation experiments must support a preliminary "design clearance" and, accord-
be designed? The scientific imperatives are not clear in this ingly, the scope and depth of simulation must be such that
area and research is urgently required. detailed analysis of VMS behaviour is possible.

By analogy with existing practices in flight simulation, the This section will introduce and describe briefly seven com-
obvious first step is to attemptto emulate human characteristics plementary but different viewpoints on VMS structure. The aim
in determining simulation objectives, some of which are highly is to demonstrate not only the complexity of the system but also
structured, some are random and some are context-dependent the number of overlapping interpretations which exist. Each
(e.g. workload). In other words, a pilot is partly ruled-based, relates to a different type of behaviour and, thus, a different
partly fuzzy and partly chaotic! Training will encourage struc- type of simulation model.
tured behaviour but chaos might still occur under conditions of
high stress or uncertain information. As far as is known, this Aircraft Viewpoint
mix of imperfect human qualities gives high coverage of A simple block diagram is a good starting point for introducing
potential problem areas and is a proven technique for finding more detailed descriptions of a VMS. In figure 2 an aircraft
snags. platform is shown schematically, carrying weapons and con-
taining a system (the VMS) and a cockpit. The cockpit supports
The challenge to simulation specialists will be to recreate this the aircrew and provides them with controls and displays
level of confidence for complex systems which do not rely upon (designated 'C and 'D', respectively) which allow them to
human action or intervention. An overview of possible con- interact with the system. The system interacts with the aircraft
trol/display modes is instructive: (because it is controlling the aircraft!) and the aircraft interacts
with the outside world environment (airmass, terrain and so
Control Characteristics on). Finally, the outside world provides visual cues for the
Modes :^Y: ■:■■:■■■ aircrew; motion cues are acquired directly through the aircraft.
Manual Full manual operation
Protected Carefree manual operation via monitoring Physical Viewpoint
Augmented Assisted manual operation For illustration of the physical integration of the aircraft and its
(high performance) embedded equipment, a schematic is presented in figure 3.
Selected Pre-programmed automation The Aircraft is described in terms of its longitudinal aerody-
Pre-selected Adaptable automation namics and its major airframe 'components', namely Engine,
Automatic Full automation (autonomous operation) Fuel, Cockpit and a set of other enclosures which, for
convenience, will be called Zones. The intention is to indicate
Display Characteristics the principal sources of interaction between the aircraft and its
Modes embedded systems and, to this end, the schematic should be
Essential data self-explanatory. Attention is drawn to the so-called Integrated
Continuous
Informational data to assist manual operation Power Offtake System (IPOS), which is a merely a packaging
Cooperative
concept covering a range of airbleed and power offtake
Interactive Generated data in response to
manual operation requirements.
Selectable Non-essential data
Advisory Event-driven data
Diagnostic Fault-finding data
 22-5

Equipment Viewpoint 4 Concept of Flight Clearance

A more uniform structure is shown in figure 4, which con-
centrates on the functional partitioning of equipment within the
Flight Clearance is the process of validation and verification
airframe. It also shows the interactions between system,
which demonstrates that an aircraft is safe to fly. It must ensure
cockpit, aircraft and outside world, as contained in the Aircraft
an appropriate level of dependability for the aircraft and its
Viewpoint. Interfacing takes the form of Controls (C), Displays
embedded systems. It must be based upon a set of design
(D), Actuators (A), Sensors (S), Inputs (I) and Outputs (O). It
concepts, implementation standards and operating procedures
is assumed that the so-called Zones will contain the system
which minimises or removes the risk of accidents.
processing resources.

Functionality has been added for primary actuators (which are

essential for aircraft flight), secondary actuators (for ground
4.1 Dependability
handling and optimised flight) and aircraft sensors (as distinct
from system sensors). There are various interpretations of the term Dependability
although all convey the general consensus that a system is fit
Communication Viewpoint for purpose. The particular interpretation adopted here is based
The passage of information around the system is handled by upon those concepts which are critical to flight clearance. A
a data network, as shown in figure 5. Note that a digital process summary is given below.
is characterised by information transfer whereas an analogue • Certification
process is characterised by energy transfer. This provides all Legal recognition that a system complies with statutory
the interfacing that was identified in the Equipment Viewpoint, requirements, e.g. airworthiness regulations and health
with the exception of actuators and sensors which are local to and safety legislation.
the equipment zones. • Qualification
Documentary evidence that a system will operate safely in
Functional Viewpoint accordance with its specification and that adequate pro-
Figure 6 shows the arrangement of 'control' and 'manager' cedures were imposed during development in order to
functions which provide the inner-loop connections around the provide proof of design.
airframe equipment. The distinction is based on stereotypes.
Control functions are considered to be high-performance and • Safety
non-optimising: manager functions are low-performance and Confidence that a system will behave in accordance with
optimising. its specification without causing harmful effects.
• Survivability
The underlying philosophy should be self-evident although it is Ability of a system to continue operating within a specified
worth noting that Flight Control and Airframe Control are not hostile environment, as defined by the notions of sus-
shown as independent functions. This reflects the current ceptibility (i.e. the probability that a fault may be induced
debate regarding the layout of future aircraft platform sensors. by a defined threat) and vulnerability (i.e. inability to tolerate
For convenience, it is assumed that a]| sensor information will damage resulting from an encounter with a threat).
be coordinated through a Sensor Manager. • Robustness
Ability of a system to accommodate failure effects or
Processing Viewpoint changes in the system environment, thereby providing a
The 'control' and 'manager' functions, identified in the Func- margin of protection against unplanned events which would
tional Viewpoint will be physically incorporated within an otherwise have unpredictable and possibly catastrophic
equipment zone, e.g. an A3 modular rack. Figure 7 attempts consequences.
to show the general form of partitioning that is envisaged
although no attempt is made to allocate particular functions to • Predictability
particular modules. The intention is to indicate that, somewhere Stipulation that a system will behave in a known manner
in the Zone, there resides a Flight Control function (say) which within its prescribed operational envelope.
is served by a local data network and a local power network. • Testability
Ability to perform operational checks on a system.
Zone-specific features are highlighted, namely the Interface •Availability
Module, Power Supply Module and Zone Manager. Essentially Time/resource constraints within which a system can
the Zone Manager is responsible for all aspects of energy and provide its intended operational capability.
resource management within its zone, such that it handles
cooling, power distribution and computing hardware at local
level. The role of sensors and actuators is seen here as an 4.2 Fault Propagation Model
environmental task, without loss of generalisation. The
important point to note is that measurement and control of the For current purposes, a Fault in a system element is defined
zone environment are handled locally w'athe Interface Module
to be any unintended change in its behaviour (i.e. outside
and the Zone Manager. normal operating tolerances). A Failure is an extreme change
in behaviour such that, for all practical purposes, the system
Integration Viewpoint element is no longer operational. The relationship between the
Figure 8 offers a broad overview of system functionality and, two is shown in figure 9, together with a perspective on possible
most importantly, an interpretation of functional integration. It
fault mechanisms.
shows a composite of Aircraft Viewpoint and Functional
Viewpoint, with an exploded view of the VMS plus additional
functionality which supports a mission capability.
22-6

System properties can be affected by component defects (e.g. • Fault Tree Analysis is performed on a system, in the
a manufacturing flaw oradegradation in material performance) context of its environment and operation, in order
or by environmental threats (e.g. electromagnetic interfer- to find all credible ways in which a predefined
ence). Both phenomena can result from inadequate design, hazardous event can occur. Note that a fault tree
arising either from poor component quality or insufficient is not in itself a quantitative model, nor is it a model
protection. However, assuming competent design, the main of all possible causes of system failure.
problems concern components failing with age or an excessive • Zonal Analysis is performed a system in order to
threat impinging on the system. A component defect might establish [i] the effects of adjacent failures within a
render the system vulnerable to a threat or, alternatively, a bounded area of the aircraft, [ii] design installation
threat might result in component damage. If defects become and regulations and [iii] maintenance and operating
significant, either in severity or quantity, they will eventually aspects (including human factors).
constitute a fault; if threats cause large disturbances in the
• Energy Trace Analysis identifies circumstances in
system environment, they too will eventually constitute a fault.
which an energy source transfers unwanted energy
to a vulnerable target in the absence of adequate
Particular difficulties arise when a system element can proceed
barriers.
from a fault condition to a failure condition with no evidence
that the fault has occurred. This is called a Dormant Failure Hazard tracking is performed so that all risks to the system are
because it is hidden until the system element becomes critical. reported and the status of risk reduction activities is clearly
The issue of dormancy is not easy to solve and ultimately must visible.
be addressed in the system design, taking due care in analysing
failure probabilities.
4.4 System Integrity
The concept of a system Error relates to the capacity of the
system to associate behavioural anomalies with known fault A High Integrity System is one for which incorrect behaviour
mechanisms and to raise an alarm. Assuming that the fault can would be unacceptable. The requires a high level of confidence
lead to a failure, the system must initiate protective action in that the system implementation is correctand, thus, that system
order to reduce or remove the failure effects. If the fault behaviour is predictable. An appropriate methodology will
corresponds directly to a failure, then corrective action is minimise the likelihood of design deficiencies arising from
required in order to recover from the failure effects. There human errors and computer-aided tools by eliminating the
appears to be no universal agreement on terminology in this circumstances which permit such errors to occur and go
area but it is clear that 'protection' involves Fault Masking and undetected. These requirements are expressed by a set of
that 'correction' involves Fault Tolerance. A more basic form criteria for Design Integrity, as follows:
of intervention in a system is performed at module level and
comprises a set of operational checks which are designed to Design Integrity Attributes Requirement
detect and locate component defects. This is generally called
High' Operation Predictable
Built-in Test and is intended to prevent defects from becoming
Task Predictable
faults.
Performance Predictable
Resource Predictable
4.3 Hazard Identification Medium' Operation Short time at risk3
Task Predictable
A Hazard is a condition which is identified as having the Performance Predictable
potential to cause harm to the aircraft, its occupants and its Resource Dissimilar4
environment. Obviously, this does not include planned weapon Low2
delivery! Typical hazards include
- No specific requirement
1
• Loss of structural integrity Incorrect behaviour is unacceptable
2
• High energy uncontained explosion Incorrect behaviour is acceptable provided it is
• Uncontained debris contained
3
• Fires Mitigates the overall risk if and only if the time at risk
• Escape of gases or liquids is short
4
(especially if hot and/or pressurised) Substantially reduces the probability of incorrect
• Inadequate protection against environmental threats behaviour
• Controlled flight into terrain In addition to design deficiencies, it is recognised that incorrect
The hazard risk assessment is based on [i] the worst possible behaviour can arise during system operation due to failures.
(and credible) consequence of a hazardous event and [ii] the Thus, a high integrity system must embody a set of design
likelihood of a failure which would cause a hazardous event to concepts, implementation standards and operating procedures
occur. which minimises or removes the risk of unacceptable failure
effects. This requires that, firstly, the system failures are
Analytical methods are used to identify the hazards which are predictable and, secondly, stability and performance are
linked with run-time faults and failures. The principal categories robust to any change in system characteristics. This require-
of analysis are, as follows: ment is expressed by a set of system-level criteria for Oper-
• Failure Modes, Effects and Criticality Analysis is ational Integrity, as follows:
used to assess the cause and effect (including
limits) of all possible failure modes, together with
estimates of the probability and criticality of
occurrence.
 22-7

Operational Attributes Requirement Design Feature Requirements

Integrity Design Integrity High
Fail Operational' Operation Continuous/Essential or
Performance Robust Medium
Failure Predictable
Resource Static
Fail Stop2 Operation Selectable/Non-essential Allocation or
Failure Predictable Static with hard-deadline switching
or
1
Failure effects are unacceptable Dynamic with hard-periodic resets'
2
Continued operation leads to unacceptable effects
Resource Static
Note that a Safety Critical System is a special type of high Scheduling or
integrity system for which 'unacceptable' means 'hazardous', Pre-emptive with hard deadlines
either resulting directly in an accident or substantially or
increasing the probability of an accident. In fact, it is common Dynamic within hard-periodic frames'
to use the safety analogy when discussing the risks associated
with any type of critical system. For this reason, the term "fail ' Only applicable to microprocessor operation
safe" is usually preferred to "fail stop".
The default requirements would be for [i] the highest level of
design integrity, [ii] static resource allocation and [iii] static
Using the above criteria, the formal options for safety-critical
resource scheduling. In order words, everything about the
system implementation are [i] a high or medium level of design
run-time behaviour of the system is completely pre-determined.
integrity and [ii] fail operational or fail safe response (as
appropriate). The particular choice of option depends upon the
The other design options respect the onward development of
nature of external hazards and upon the requirement for
dependable computing techniques. It is reasonable to expect
continued operation of the relevant part of the system.
that some relaxation of the default requirements will be possible
although it must be stressed that this is not an end in itself. For
the foreseeable future, system-level resource management will
4.5 Flight Clearance be essentially static. However, if support becomes available
for hard real-time operation, then pre-emption becomes
For any system intended to be safety critical it is essential that possible. At processor level, it is possible that optimised VLSI
such a requirement be identified at an early stage of devel- design features (such as pipelining and multiple-level cache)
opment, preferably at the outset. "Safety" cannot be applied will be used. However, for these non-static features, it will
retrospectively without incurring significant penalties, because always be necessary to constrain theirusage in order to render
the level of design integrity has to be increased. As has been them open to analysis and test. Without constraints a safety
demonstrated on numerous occasions, any attempt to do so is case cannot be constructed.
generally very expensive, even assuming that it is possible!
Contrary to popular belief, if a safety critical methodology is
applied from the start, the resulting overhead may not be 4.5.2 Robustness
particularly large. Indeed, as a result of the greater rigour
applied to the development process, some costs might actually
Robustness is concerned with being able to accommodate any
be reduced because the system is more likely to work first time, unintended behaviour of a system. In practice, this means being
thereby removing the need for re-work and re-testing.
able to maintain acceptable stability and performance margins,
firstly, under internal failure conditions and, secondly, under
Given the scope of functionality involved, it is clear that a VMS external variations (arising either from environmental changes
is clearly going to be safety critical. Its operation will be essential
or from an inadequate model of the intended behaviour!).
to aircraft flight and, therefore, it is therefore possible to proceed
directly to define both the properties which must be exhibited
In the functional architecture, the concept of robust design is
by any VMS concept and the proof required in support of such
concerned with control law development. For a VMS or any
a concept. The properties needed are:
other complex, highly-integrated system, the control strategy
Determinism will be multivariable. This demands a new engineering
Robustness framework for analysis and design, and one which retains the
Testability high visibility and common intuition that is associated with
These must be incorporated at all stages of design (i.e. in each classical Nyquist-Bode techniques. This is, in fact, possible to
definition of system architecture) in order to provide high construct if it is recognised that the most important properties
integrity. Effectively, what is being stipulated is that system of any control system are of signal transmission and no! those
behaviour is pre-determined as far as is practicable (under both of arcane and recondite mathematics.
nominal and failure conditions), stability and performance is
robust (to both failure effects and dynamic variations) and that The generalisation of classical control requires that, for initial
these properties can be demonstrated with sufficient rigour to exposition only, certain constraints be imposed on the use of
satisfy airworthiness authorities. multivariable feedback. Thus, by considering diagonal control
and square systems, a concept of multivariable structure can
be formulated which incorporates phase and gain margins as
4.5.1 Determinism primary indicators of robustness. This is justified by a
straightforward piece of algebra which shows that the structure
The concept of determinism is used here to stipulate design of a multi-input multi-output (MIMO) system is described by an
requirements, as follows: equivalent set of single-input single-output (SISO) channels
which preserves the role of loop interactions. The phase and
gain margins are applied to these SISO (or individual) channels
22-8

and demonstrate immunity from the effects of dynamic at very low frequencies. Note that, if the y,-locus were to pass
uncertainty and, by extension, structural transitions such as through the point (+1,0), the channel transmittance would
loop failures. Once having established the initial exposition, the become identically zero at some frequency!
constraints can be removed and the analysis can be applied
to non-diagonal and non-square systems. It is convenient to construct a model of uncertain multivariable
structure by drawing a disc around any given point on the
A conventional model of feedback control is seen in figure 10, 7,-locus [cf. Figure 12]. The centre of the disc gives the nominal
comprising a 'system' G(s) and a 'controller' K(s). The open- value of y,(s) and the radius gives the lowest upper bound for
loop and closed-loop dynamics are described by the following the uncertainty at that point; a directed line segment drawn to
relationships: the point (+1,0) defines the gain and phase contributions to the
Open-loop System y = G(s)K(s)r + nP channel transmittance due to the presence of other feedback
loops in the multivariable system. The actual value of the
Closed-loop System y = T(s)r + S(s)nP - T(s)nM multivariable structure function corresponds to a point located
where the system output y is derived from the system input r, somewhere inside the disc. If the disc is large' and 'close to
process noise nP and measurement noise nM. The transfer the point (+1,0)' then the gain and (especially) phase at that
function matrices T(s) and S(s) are defined as follows: frequency become more sensitive to uncertainty, i.e. they
Tracking Function T(s) = [l+GK] 'GK = GK[I+GK]' exhibit a larger range of possible values. In extremis, if the disc
Matrix encloses (+1,0) then the problem of phase sensitivity
becomes acute (up to ±180°). This introduces the possibility of
Sensitivity Function S(s) = [l+GK]"1 non-minimum phase behaviour which, if it were to occur near
Matrix the system bandwidth, would jeopardise stability.
For the closed-loop system T(s), an individual channel can be
defined between any selected input-output pair. The feedback Stemming from these insights, three sets of gain and phase
loops can then be separated, one belonging to the closed-loop margins can be defined for the i* individual channel of a system
channel and the remainder belonging to the rest of the G(s):
closed-loop system. By removing its feedback connection, the Channel Robustness Gain margin of c, when ZC|=-180°;
channel becomes open-loop; in fact, it constitutes a special Measures Phase margin of c, when |c,|=1.
type of open-loop subsystem because it incorporates a com-
plete description of the closed-loop subsystem with which it Structural Sensitivity Gain margin of yr when Zy,=0o;
interacts. The concept of individual channels is absolutely Measures Phase margin of y, when |y,|=1.
fundamental to multivariable feedbackas it provides an intuitive Phase Sensitivity Gain and phase margins of y, when
basis for system decomposition, a testable basis for system Measures zc,=-180°;
analysis and a practical basis for system design. Gain and phase margins of y, when
|c,|=1.
Structural equivalence exists between the tracking function
matrix T(s) and the closed-loop dynamics of individual chan- These are shown schematically in figure 13. The channel
nels, as shown in figure 11. The open-loop dynamics of the i" robustness measures are akin to the classical phase and gain
individual channel can be written in the form margins forstability robustness; structural sensitivity and phase
c,(s)=k|(s)gM(s)(1-yl(s)), where the term y,(s) is called the sensitivity are indicators of performance robustness. If both
multivariable structure function. The corresponding structural sensitivity and phase sensitivity are benign, then the
closed-loop dynamics are calculated as t|,(s)=c,(s)/(1 +c,(s)). classical phase and gain margins associated with the individual
channel are valid measures of channel robustness (provided,
The relationship between uncertainty and robustness is of course, that the usual 'gradient rules' are observed at the
assessed by design margins. The expected frequency gain and phase cross-over frequencies).
response characteristic for an individual channel corresponds
to a nominal transfer function model [cf. Figure 12]. However,
there is a range of gain and phase values at any given frequency
4.5.3 Testability
and a range of possible gain and phase cross-over frequencies.
The actual frequency response lies somewhere inside an The system design must be capable of being tested at a number
envelope of possible characteristics. With reference to the of levels and, therefore, appropriate access mechanisms must
familiar Nyquist stability criterion, if this envelope is large be provided by design. Specific requirements exist for tests to
enough to enclose the point (-1,0) then the number of be performed in the following circumstances:
encirclements may change, thereby rendering the closed-loop • design and development
system unstable. This indicates the problem of channel • flight clearance
robustness. ForSISO systems, this would furnish a complete • maintenance
interpretation of robust control: for MIMO systems, the situation • pre-flight
is more complicated. • in-flight
With increasing complexity of system design, the role of
The term 1 -y, is critical in dictating the achievable performance
simulation will become more and more significant, both in
of the individual channel c,(s). Its variation over frequency is establishing the expected behaviour of the system and in
implied by the polar plot of the multivariable structure function,
determining the most appropriate test procedures. This
7j(s) [cf. Figure 12]. Uncertainty in G(s) generates an envelope involves making assessments of failure effects and worst-case
which bounds the locus of y,(s). If any part of the envelope interactions in order to ensure adequate test coverage.
encircles the point (+1,0) then there is a problem of structural
sensitivity \n that a change in structure can generate additional
zeros in the right-half s-plane. Incidentally, this gives rise to a
new'Nyquistminimum phase criterion". Theeffect is pernicious
because the resulting loss of phase drives the system unstable
and it is particularly awkward because the instability can occur
 22-9

In the context of the present discussion, it is appropriate to

concentrate on the relationship between testability and simu-
4.5.4 Flight Clearance Activities
lation. The use of simulation tools for VMS design and devel-
opment enables a much greater consideration of different Having formulated the working definitions of determinism,
system options and a more rapid assessment of different robustness and testability, then flight Clearance can be shown
optimisation strategies. The principal benefit lies in the ability to depend upon:
to perform extensive analysis and to generate benchmark Design
results in support of flight clearance activities. Note that Analysis
simulation will not in itself reduce development timescales! Testing
Clearance or "proof of operation" is not a single activity,
It is possible to construct simulation models at all levels of performed once, but is made up of many activities, performed
architectural abstraction but, in practice, it is probably throughout a programme and pervading every facet thereof.
cost-effective to stop at the level of process architecture. A Some further brief comments will suffice in order to highlight
staged approach is necessary, establishing the viability of an some of the key aspects.
overall system concept model first of all and, then, proceeding
in a modular fashion to substitute more detailed models where Design requires a well-defined methodology and good con-
required for the particular development activity being under- figuration control. Although it is accepted that practical design
taken. Concept models are especially useful because they proceeds middle-out, it must conform to the well-established
convey basic information (e.g. parametric trends, dominant notions of structured decomposition and modular partitioning.
dynamics) without undue elaboration. Subsequent tests must To this end, the widespread use of structured semantics (for
verify that model substitution, while giving more accuracy high-order language programming) and graphical editors (e.g.
results, does not fundamentally alter the conclusions regarding for simulation block diagrams) is of great help. Potentially, the
the overall system concept. use of object-oriented languages (such as C++) will have a
profound effect on engineering simulation and design.
It should be recognised that, from the various viewpoints which
overlook VMS [cf. section 3], the structure of the system is Analysis must be performed in order to show that a system is
parallel (or concurrent). Considerable progress has been made acceptable both in terms of the control design (adequate gain
in understanding the issues involved in constructing large-scale and phase margins, efc.) and in respect of the hardware design
diverse simulation models using parallel machines. This (failure modes and effects, efc). Software items must be
technology can offer for the first time at an affordable price a subject to appropriate analysis to verify that they are structured
real-time simulation capability which has wide applicability in a manner which is satisfactory for safety-critical applications.
across aircraft systems. More work is needed, however, in Typically this would include examination of program structure,
order to satisfy demands for generic simulation platforms which data and control flow, and ideally would be performed by same
can access graphical notations (such as those generated using form of computer-based tool, such as SPADE®. It is essential
MATRIXx" or EASY5®) and map a simulation model on to a that analysis proceed in parallel with design; it is not an
network of processors, e.g. transputers. This will have a "incremental" activity done later.
profound impact on productivity of so-called "simulated
engineering" and, at last, enable the development of a proper Testing is usually understood to be the "verification" that a
Virtual Rig environment. design is correct and the "validation" that it fulfils its specifi-
cation. Unlike analysis, which generally takes place on a
Because simulation will play a very important in VMS design general-purpose workstation, testing in support of flight
decisions, new concerns are being raised about Dependable clearance has to make use of actual flight-standard hardware
Simulation. A system architecture (to any level of detail) can or, at least, its functional equivalent. Simulation can assist test
be encapsulated as the functional architecture of a simulation. activities by providing an independent model against which the
The subsequent design of that simulation can be interpreted operation of the target system can be compared. It can also be
as a hierarchical abstraction, in exactly the same way as for used during the design stage to gain an insight into system
system design [cf. section 2.2]. The resource utilisation con- operation, particularly in the presence of systematic faults.
tained in the process architecture will include facilities such as
aGraphical User Interface (GUI), Computer-Aided Engineering Due allowance must be made in system design as to the
(CAE) tools and general-purpose interfacing for "Hardware In intended purpose of a system in order to make testing practi-
The Loop" testing. cable, rigorous and cost-effective. Some past designs,
although nominally satisfactory in functional terms, were almost
The concept of system integrity which is relevant here is impossible to test because testability was not considered at the
perhaps best described as Design Critical. This implies a design stage. For the complex and highly-integrated systems
particular type of high integrity system for which 'unacceptable' envisaged for the future, such omissions might mean that
means 'incorrect calculation', resulting in incorrect design systems would be completely impossible to test to any degree
decisions or invalid clearance (with potential safety implica- of confidence. Thus, VMS implementations must be provided
tions). Using the existing criteria, the formal options for with features which enhance testability and enable systems to
design-critical system implementation are usually considered be constructed in a staged manner as hardware/software
to be [i] a medium level of design integrity and [ii] fail safe becomes available during development.
response. These are justified, respectively, because simulation
must always be reinforced by independent calculation and
because continued operation (due to an undetected failure) 5 Where are the Challenges?
leads to an unacceptable outcome (i.e. wrong results). There
is no requirement for a fail operational response: on the
Traditionally, for relatively simple systems, development has
contrary, there is a positive requirement to stop the simulation
been split into two disciplines, namely Control Law Design
when a failure occurs in order to generate diagnostic infor-
(servo loop design) and System Design (including software).
mation.
Future systems will not be so convenient. Proper operation,
particularly in the presence of one or more failures, requires
22-10

the design to be properly understood and coordinated if the • Safety

criteria of Determinism, Robustness and Testability are to be • Legislative
met. This in turn raises challenges in a number of key research • Quality
areas. These are summarised below. The top priority is to sponsor work which will generate a draft
standard on Certification Considerations for Modular Aircraft
A3 Operating Systems Systems, taking into account the decomposition of system
Research into operating system concepts is required urgently. development activities, the role of modelling and simulation,
As currently conceived, A3-based systems cannot support independent qualification testing, compatibility of system fail-
safety-critical applications as they do not provide [i] hard ure models, integration of system design data, flexible allo-
real-time scheduling or [ii] allocation services for segregation cation of system resources and so on. The recent contribution
of system resources. "Hard real-time" is an essential require- by the SAE Systems Integration Requirements Task Group,
ment for determinism; this is an issue for both system design dealing with 'complex and highly-integrated aircraft systems',
and system simulation. The "segregation" issue could be is a most welcome first step. However, the clearance philos-
resolved by modifying the A3 hardware concept. ophy must evolve very quickly if it is to keep pace with
technology and if it is provide a legal basis for the certification
Analytical Control Theory of aircraft systems.
An exhaustive assessment of robustness can be performed on
multivariable control systems, using phase and gain margins.
This provides immunity from the effects of dynamic uncertainty. 7 References
Further work has provided the extensions necessary to handle
loop failures and, also, systematic changes due to moding and Carter HN, 1993, A Subsystem Integration Technology
logic. Current priorities for VMS flight clearance are concen- Concept, SAE Paper 93-1382.
trated on non-robust failure modes and "worst case" effects
arising from multivariable structure, nonlinear dynamics and Edwards RA, 1994, ASAAC Phase 1 Harmonised Concept
time variance. Summary, ERA Technology 1994 Avionics Conference and
Exhibition: "Systems integration - is the sky the limit?", Ramada
'Virtual Ria' Simulation Hotel, Heathrow, UK, 30 November - 1 December 1994.
The reliance on simulation technology is likely to increase
significantly as a result of VMS development, because of Johnson BW, 1989, Design and Analysis of Fault-Tolerant
factors like "size", "connectivity" and "optimisation". It is certain Digital Systems, Addison-Wesley: Reading, MA.
that much of the evidence in support of flight clearance will be
computer-generated and will involve diverse models and Leithead WE, O'Reilly J, 1992, Performance issues in the
sophisticated analytical tools. The issues of Virtual Rig and individual channel design of 2-input 2-output systems:
Dependable Simulation are closely related and there is a clear Part 2. Robustness issues, International Journal of Control,
need to establish industry standards for development and 55(1), 3-47.
qualification of commercial simulation and support tools. Note
that the objective is not to enforce the use of Commercial O'Reilly J, Leithead WE, 1991, Multivariable control by
Off-The-Shelf (COTS) products which are inadequate, but to 'individual channel design', International Journal of Control,
encourage the creation and adoption of appropriate standards 54(1), 1-46.
against which future COTS products can be qualified for
safety-critical system development. SAE Systems Integration Requirements Task Group, AS-1C,
ASD, 1994, Certification Considerations for Highly-Inte-
grated and Complex Aircraft Systems, SAE Document ARP
6 Conclusion 4754.

VMS concepts contain complexity and diversity. They are

subject to uncertainty, both because the integration issues are
not particularly well understood and because the applicable
technologies have not yet been fully defined. It is certain that
partitioning must be adopted for development in order that the
specification, design, implementation and test activities for the
various parts of the system will be tractable.

The hope is that a push will be made towards the creation of

a "modular" flight clearance procedure. Recent work on
architectural definition (especially A3) and object orientation
indicates that this should be possible. What is needed is a
concerted effort to formalise qualification requirements at
system-element level, within an overall framework that allows
flexibility in connecting system elements and traceability
between internal failure modes and external hazard identifi-
cation. This would allow system configurations that are
non-permanent, possibly through the useof dynamic switching,
as discussed already.

Simulation will be vital in this context although gaining

acceptance of the underlying philosophy will pose perhaps the
greatest challenge of all. The implications are far-reaching and
influence many requirements, such as
 22-11

Aircraft
Wing Aerodynamics

Tall Aerodynamics
Downwash Flowtteld

Jet Interference

Engine

IPOS

Airframe

. COCKPIT
EfM^f

Figure 3 Physical Viewpoint

Figure 1 A New Breed of Test Pilot?
X ±

[~i
-Q

>
z
c

r
0 @
Cockpit

[Ö] 0
[D]^^^^^P[C ., rf ZONES

111 Aircraft | C^Pit Outside s t

■ ■■ World ÖS
Q
1 | Weapons |
3
= g,
C i
PRIMARY
ACTUATORS

SECONDARY
ACTUATORS

L Aircraft
fr
L
Ql SENSORS
Figure 2 Aircraft Viewpoint

3
Outside
World
n
Figure 4 Equipment Viewpoint
22-12

D4M NETWORK

L-CEK

a a Eh
ä B-
Interface Module ST
ZONE MANAGER
Cockpit

es FLIGHT CONTROL

PROPULSION CONTROL

AIRFRAME CONTROL

H—• DISTRIBUTION CONTROL

POWER MANAGER
H3
FUEL MANAGER

C Aircraft
SENSOR MANAGER

.__f
V—Q
c Power Supply Module

-Q Outside
World

Figure 5 Communication Viewpoint Figure 7 Processing Viewpoint

MSSION MANAGER
MISSION PLANNING VEHICLE MMI
PRIORITISATION MISSION MMI
VEHICLE ASSESSMENT
DEFENSIVE ACTION
OFFENSIVE ACTION

VEHICLE GUIDANCE
AUTOPILOT
FLIGHT DIRECTOR
OBSTACLE AVOIDANCE

AIRFRAME CONTROL
TRAJECTORY MANAGER
LANDING GEAR
TRAJECTORY PLANNING GROUND HANDLING
WAYPOINT NAVIGATION WEAPON RELEASE
TERRAIN AVOIDANCE
COLLISION AVOIDANCE

ENERGY MANAGER
WEAPON MANAGER
DISTRIBUTION MANAGER
STORES MANAGEMENT
POWER MANAGER
WEAPON SELECTION
WEAPON PREDICTION ZONE MANAGER
WEAPON FIRE CONTROL FUEL MANAGER

DATA MANAGER SENSOR MANAGER

DATA FUSION INERTIAL
TERRAIN MATCHING ELECTRO-OPTICS
TARGET RECOGNITION RADAR
TARGET PREDICTION COMMUNICATIONS

Figure 6 Functional Viewpoint Figure 8 Integration Viewpoint

 22-13

Exceedance
Susceptability Unreliability

'
m
Z
< _ X CO
o§ o O
23
3! Z
m 2» m
03 ^3
~TH|

ss «s
S m
H>
O
JL 3
1*

H m m
>
i-

D 3 D
ID (u ü>
Vulnerability ^ DIT H (Q <D
C
s
■
a
Malfunction
3
to
FAULT
rv * K »A

Fault Masking
r m
Ms) |— 1 \
o
o m
3 ymmW I ' - M*>> J 1
ERROR W

B>.
Fault Tolerance s
■

FAILL RE - '

Figure 9 Fault Propagation Model Figure 11 Individual Channel Decomposition

Individual Channel

uncertainty Bound

K(s) G(s)

Control Law System Dynamics

Phas« Margin
Uncertainty

Multivariable Structure
Structural Sensitivity
Phasa Margin

Figure 10 Automatic Feedback Control

Skuctural Sensitivity
Gain Margin

Figure 12 System Uncertainty

22-14

Individual Channel

gm = Gain Margin
pm = Phase Margin

Figure 13 Phase and Gain Margins

 23-1

THE DUTCH NATIONAL SIMULATION FACILITY:

ADVANCEMENTS IN SIMULATOR TECHNOLOGY AND APPLICATION
by
H.A.J.M. Offerman
Deputy Head of the Flight Simulation Department
National Aerospace Laboratory NLR
Anthony Fokkerweg 2, 1059 CM Amsterdam
The Netherlands

simulation and simulator requirements,

1. SUMMARY pilot workload in various tactical
The Dutch National Aerospace Laboratory situations has become more and more
NLR has developed during the last three important, which led to requirements
years an advanced Full Mission to extend NLR's simulation
Simulation capability for fast jet capabilities.
fixed wing and rotary wing aircraft.
This Full Mission simulator, named the Not only should this extension follow
National Simulation Facility NSF, is the Mid-Life Update of the F-16, but
created for research in training it should also be possible to simulate
methodologies, investigation of rotary wing aircraft. Even, when the
simulation and simulator requirements need was felt, it should be possible
for specific training tasks and to to simulate fast ships or land
support industrial development of vehicles.
cockpit and aircraft systems. The
current capability of the simulator Financed by the ministries of Economic
focuses on the F-16 Mid-Life Update Affairs and Defence, the Royal
configuration, but will be extended to Netherlands Air Force and NLR's own
accommodate manned helicopter research budget, the NSF was developed
simulation later this year. The latter by NLR researchers in a relatively
will be directed towards the recently short time span of three years.
(by the Dutch Air Mobile Brigade)
procured Boeing CH-47D Chinook medium-
heavy transport helicopter, and the
McDonnell Douglas AH-64D Apache attack
helicopter.
This paper will address the various
technical aspects of the National
Simulation Facility.

2. INTRODUCTION
The National Simulation Facility (NSF)
which has been developed at the Dutch
National Aerospace Laboratory NLR
situated in Amsterdam, is an extension
to the already existing NLR Research
Flight Simulator. This simulator is
operational since 1976 and is mainly
active in the field of civil oriented
research simulation. However, military
simulation could also be accommodated,
illustrated by the fact that the
mechanization of the digital flight
control system of the Israel Aircraft
Industries LAVI fighter aircraft was
designed and refined on this same
facility.

During the last years the need for

research into training methodologies, Fig. 1: The National Simulation Facility NSF

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?"
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
23-2

3. PURPOSE OF THE NSF checks before actual flight testing

The National Simulation Facility is of the F-16 MLU aircraft;
designed as the primary facility in the Simulations must support air-to-
Netherlands for advanced simulation and air, air-to-surface, low-altitude
training research and development, and high-speed missions with a maximum
for the (inter)national aerospace of realism. This implied that not
industry to have an instrument to only the visual system should have
evaluate and test aircraft a very high fidelity, but also that
(sub)systems, hence its name. The NSF F-16 weapons, avionics systems and
focuses in its applications primarily the mission environment should be
on Defence related research: modelled with great precision and
the concern on growing pilot realism;
workload due to the increasing Simulator investigations
complexity of avionic systems needs concentrating on human factors is
to be addressed in the early stages considered to be one of the prime
of system design. To understand and issues of the facility's
assess pilot workload during capabilities: research into
tactical missions and combat, training and training
controlled conditions must be effectiveness, assessment of pilot
created which can replicate the workload during simulated combat
situations encountered in current missions, etc. This implied that
fast jet and helicopter aircraft. the NSF must be capable of
Aircraft and avionic manufacturers integrating physiological
must integrate their concepts in the measurement equipment.
earliest possible moment during Assessment of simulator training
aircraft design. Early pilot in the effectiveness and (training)
loop evaluation of these concepts simulator component requirements
can only be achieved in a simulator should be possible by selective
capable of integrating real and 'degrading' of its components;
emulated avionic systems. The NSF is Simulation of avionics should
designed to integrate real avionics support integrating both aircraft
or to simulate it through its equipment through its standard
software counterpart. interface ("hot bench") and its
Training systems should not functional software equivalent;
represent a mere replication of - The design and development should
aircraft systems and their functions include the future growth to enable
but should be precisely designed helicopter simulation;
instruments to increase the pilot's All hardware and software
efficiency when flying his aircraft. components should be designed in
Research into the "unknowns" of such a way that maximum
simulator training is currently a interchangeability with the other
"hot topic". Various (inter)national NLR research flight simulator would
programs like the EUCLID CEPA 11 be ensured.
address simulator and their benefit
in training. The NSF has a A great many more requirements could
significant role in CEPA 11. be presented here, but this would
distract the reader from the more
From these and other purposes the NSF essential information in this paper.
was put forward by the Dutch Government
as a prime requirement. Making maximum
use of existing facilities, 4. FULL MISSION SIMULATION IN A
infrastructure and expertise, NLR was RESEARCH ENVIRONMENT
put forward as the preferred base for
the NSF. 4.1 Introduction
A flexible and modular setup of
The aim for NSF is to create a full hardware and software components can
mission, both manned F-16 Mid-Life be regarded as the conceptual heart of
Update and helicopter simulation every research simulator. The
capability for research and components by itself show in some
development. The top level cases already unique features. One can
specifications could be defined as: see this at the various Research
The facility should be capable to Flight Simulation establishments in
perform, on behalf of Lockheed's F- Europe and beyond. The British Defence
16 Mid-Life Update codevelopment Research Agency created for advanced
programme, one of the final helicopter research their Large Motion
functional cockpit mechanisation System, a facility identifying the
very large vertical excursion
 23-3

requirements for helicopter simulation.

4.2 NLR's Research Flight Simulator

NLR operates a research flight
simulator since 1976. At that time
simulations were carried out with a
generic fighter (modified F-104
"Starfighter") cockpit which was placed
on one of the first platform motion
systems with hydrostatic bearings. Some
years later a (DC-7) transport cockpit
was refurbished to accommodate multi-
crew experiments. This cockpit could be
exchanged with the generic fighter
cockpit and placed on the motion
platform. Both the transport and the
generic fighter cockpit have been
modified during the years following to
the ever changing cockpit technology.
Currently (fig.2) the transport cockpit
features a lay-out to functionally
simulate a Boeing 747-400 or similar
"glass" cockpit aircraft, including Fig. 2 : Research Flight Simulator RFS
Control & Display Units (CDUs) for
operating the Research or Experimental 4.3 Research applications for the NSF
Flight Management Systems, touchpads The NLR National Simulation Facility
for Air Traffic Management Datalink NSF was designed for fast jet and
communication etc. This simulator has helicopter simulation research. Based
recently been used in a number of on the specifications mentioned
experiments for conceptual evaluation earlier, several concepts of building
of using ATC-aircraft datalink, the NSF were possible. One of the most
evaluations of Take-Off Performance important requirements is the ability
Management System and investigations to change the facility for different
into the establishment of handling research objectives. This includes
qualities for fly-by-wire large exchanging cockpits on the motion
transport aircraft. Especially in those system, change the flying
projects where operationally realistic characteristics of the aircraft,
flights are essential to the outcome, a change the interior of the cockpits
civil equivalent of a full mission and/or change the way the information
scenario must be created. This includes is presented to the pilot. In most
radio traffic, genuine ATC messages and cases this prohibits the use of the
directives, visual converging aircraft actual aircraft instruments. The
(to react on TCAS-messages), proper capability to adapt them to the
radio and beacon frequency selections specific simulation requirements is
etc. Apart from simulating emergency essential. For instance, the Multi
and system malfunction conditions, this Function Displays in the F-16 cockpit
Research Flight Simulator could also be are driven by graphics workstations
used for training: the big difference instead of the Modular Mission
to a training simulator, however, being Computers HUD Electronics Unit. The
the very modular and flexible software latter provides fixed, non-changeable
and hardware setup. The simulator can (currently) monochrome display
be changed from a Boeing 747-400 to a symbology. A requirement for future
Fokker 100 in a couple of hours, or research, however, is the ability to
from a traditional B-747 equipped with colour symbols and investigate the
Primary Flight and Navigation Displays human factors implications of it.
to a fly-by-wire aircraft operated by
sidesticks and equipped with CRTs The interaction of the pilot with his
displaying information in three environment is a major research topic.
dimensions. Environment in this context must be
seen in a wide sense: from situational
awareness issues during specific
combat missions to definition of skill
acquisition in a training task
analysis.
23-4

In the ongoing EUCLID (European Simulator. The growth foreseen in

Cooperation for the Long term In software code and its processing due
Defence) programme, the NSF is to the requirements of needing a full
scheduled in two Research and avionics suite and a full mission
Technology Projects: RTP 11.1 on environment, had an enormous impact on
"human factors" and RTP 11.2 on the existing infrastructure. For
"simulation technology". Both RTPs instance the computer systems
consider the lack of knowledge in resources should quadruple (memory,
the current (pilot) training throughput, input/output, disk
processes an area worth looking into capacity etc.), which could not be
in great detail. Basis for both RTPs handled within the same host computer
is the definition of "skill" and operated at that time. The TV-
skill acquisition and the way it can modelboard visual system could hardly
be influenced by training, both in be used for the fast flying and fast
the real aircraft and in the manoeuvring F-16 fighter: it not only
simulator. The definition of lacked the necessary response speeds,
objective measures and subsequently but more important, the visual field
the definition of the training task, of view of 40 x 30 degrees was deemed
its context and the technology to be insufficient.
used is the prime outcome of these However, not all required components
RTPs. The NSF will be used as the for the NSF would have to be newly
validation tool, as it is considered developed or bought: the six degrees
to be one of the most advanced of freedom platform motion system,
(research) simulators available at designed for simulation of fast jets
present. Influencing the training and helicopters was already
context, being for instance the size operational. Large parts of the F-16
of the field of regard or the size simulation software, the aerodynamics,
of the field of view of the visual flight control system, engine and
scene, is of utmost importance. undercarriage models had already been
Although many studies in the past validated. And last but not least, the
have already been performed with expertise and knowledge of our
reconfigurable simulators, the NSF simulation group played a major
combines for the first time high contributing factor in the subsequent
performance cueing in motion (both development period.
platform and direct body cueing), At the moment the required funds for
full field of regard high resolution the NSF development were granted, a
visual and advanced audio cueing in large scale system definition task
a full mission environment. took place which final outcome can be
- Already mentioned is the observed in the components, in detail
participation of NLR in the Lockheed described below.
F-16 MLU codevelopment program. The
NSF will be used to perform a 5.2 Computing Technology
cockpit mechanization checkout to One of the first (new) systems to be
look into the various specific decided upon was the simulation host
European MLU-unique items. In the computer. Having operated a Concurrent
checkout simulations, pilots from MicroFive computer with its dedicated,
five nations (USA, Belgium, Norway, proprietary operating system for many
Denmark and The Netherlands) will years, the need was felt to change to
evaluate whether all the changes UNIX-based computers (the Concurrent
made in the avionics functionality is still being used for the other
are implemented correctly and civil-based Research Flight Simulator
whether the avionics meet all the -RFS-). After a careful selection
specifications. process, the Silicon Graphics
Challenge L was chosen of which two
5. INNOVATIONS IN RESEARCH FLIGHT systems were procured. For many years,
SIMULATION software development and real-time
operations were carried out on one
5.1 Introduction computer system with one very big
Conception of the NSF took place some drawback: during busy simulator
five years ago. Ideas were formed and experiment periods software
brain storm sessions with the development would almost come to a
operational air force community took standstill. For the NSF two identical
place to establish the NSF's prime Challenges are used, one of which is
requirements. From these discussions it dedicated for software development,
became clear that development of the the other is dedicated for real-time
NSF could not be regarded as a "normal" operations. The four MIPS R-4400
extension of NLR's Research Flight processors in each Challenge operating
 23-5

at 100 MHz have sufficient computing For real-time interfacing with the
power to run the entire F-16 simulation graphics computers and for non-
programme at nominal 50 Hz update rate realtime interfacing with all other
(on only one processor). For software computers, a dedicated Ethernet is
development we decided to change from used. The graphics computers are used
the standard ASCII terminals to an "X- to generate the various displays for
windows" environment. the F-16 cockpit and control desk,
including the Multi-Function Displays
and the Head-Up Display. Instead of
routing dedicated RGB-cables to a
specific location, and thus creating a
non-modular use of. expensive graphics
Sound Visual
system system
computers, a Video Distribution System
I PC network server
was created. This system connects,
through a cabinet, each graphics
CHALLENGE CHALLENGE 6-DoF
motion system
computer with a display location. In
Development Real Time
this cabinet, however, connections can
F-16 be rerouted. A multitude of (Silicon
cockpit Graphics) graphics computers are used:
from the IRIS 4D/85, -210 and 310 VGX
:10X-terminals Mock-up Tactical Env. G-cueing

(F-16) ITEMS system

(equipped with video-splitters) to the
more current Indigo W8C and Extreme.
- SCRAMNet

Instrument Graphics Computers - Ethernet 5.3 Visual Cueing

Probably the single most important
Fig. 3 : NSF Flight Simulator Interface System attribute in a simulator is the way
the pilot (or driver) receives his
Changing to another computer system led visual information. Taking into
to other implications. All RFS account the required research
simulator hardware systems are applications, it proved to be
connected to the Concurrent via difficult to define the requirements
dedicated I/O channels. The digital- for the NSF visual system. The
analog and other converters are part of technical solutions available for
this host computer and can not be presenting a 360° field-of-regard Out-
easily reconfigured. A more user- of-the-Window (OTW) image to the pilot
friendly I/O-system was, at the did not have sufficient overlap. The
beginning of the NSF-project, already prime requirement could be narrowed
in the process of being designed. Its down to the following: does the use of
concept: each major hardware system operational equipment, normally worn
(visual, motion etc) receives its data by the pilot, have to be considered of
through a node connected to a networked more importance than a high-
fibre-optic, digital interface system. resolution, high-brightness image.
Each node has a processor which would Ergo, the choice between dome or
pack and unpack the data necessary, and helmet mounted display. For the NSF
performs calibration type of the decision was made for a dome with
calculations. Essentially this a head-slaved area-of-interest
interface concept has been used in the projection display system, the best
design for the NSF (fig.3). A SCRAMNet compromise between using operational
based, fibre-optic network connects all equipment versus brightness and
major subsystems, next to a dedicated resolution.
Ethernet. The SCRAMNet-system operates
as a "reflective memory", thus sharing The NSF visual system contract was
almost instantaneously the relevant awarded to Evans & Sutherland, which
data between the various (front-end) could offer a total, integrated image
computer systems. Sharing of data generator and display system. The
occurs at a speed of 6.4 Mbytes per head-slaved projection system, called
second for the time being and at the VistaView®, offered an "affordable"
current update rate fast enough to compromise between field of regard,
satisfy our requirements. An important resolution and brightness. Eye-
lesson learned using this concept is tracking was not really considered an
the fact that a computer's internal option. Although the Link-Miles ESPRIT
memory-to-memory transfer speed, simulator was already equipped with
especially with large amounts of data eye-tracking equipment, we didn't feel
to be transferred, can become a sufficiently confident to incorporate
bottleneck. eye-tracking at that time.
23-6

The image generator, a three channel onfiguration file. In this way the FOV
ESIG-3000® AT/GT combines the latest can be changed from the nominal 120
technology available within one system. degrees in azimuth to 180 degrees. The
Two channels are configured to handle elevation FOV changes accordingly.
1000 polygons in the inset and 3000
polygons in the background, thus giving The dome has a high-gain coating (gain
a balanced polygon density over inset 5) which was felt as a good compromise
and background. The third channel is between brightness fall-off and the
used for sensor simulation. Either an available brightness (measured
optical sensor (including zooming 6.8 ftL) with a contrast ratio of
features) or an infra-red sensor can be 54:1, at the pilot's eye point.
simulated. Although the dome consists of 24
Quite unique is the capability to sphere-parts, no seam can be observed
modify (through datafiles) the IG through the use of special coatings
throughput delay (which can be lowered and integration techniques.
to 40 milliseconds delay) and the
distribution of polygons over inset and During the design and testing phase,
background. special attention was paid to the
required stiffness of the projection
The projection system displays a head- system and dome. No relevant resonance
slaved background (142 x 110 degrees) frequencies due to the imposed
and a head-slaved inset (51 x 36 frequencies of the platform motion
degrees) on a 16.6 feet diameter system, should be noticeable in the
graphite composite dome. Figure 4 shows region between 0 and 25 Hz. This was a
the area-of-interest concept. very tough requirement for the
projection system and could only be
Both the images for background and satisfied after some mechanical
inset are pre-warped in the IG, where modifications had been made to its
also a dynamic non-linear image mapping pedestal and the attachments to the
function corrects for not having the platform. The dome, however, has an
exit pupil of the optical system in the even higher stiffness: a measured
centre of the dome. After de-rotation first resonance mode of 80 Hz.
the single image is projected by means
of a servo-controlled lens. Capability 5.4 Motion cueing
of the servo system is better than 4800 Much can be said on the relevance of
deg/sec2 and 1500 deg/sec, which for motion cueing in fast jet flight
future growth allows the servos to simulation, a difference can be made,
track the eye line-of-sight for however, for training simulators and
variable acuity IG techniques. those to be used for research and
development. Based, amongst others,
For various research studies it should upon our experience from the IAI-LAVI
be possible to change the field of view digital flight control law development
of the system, a feature which can be in which the necessity of platform
implemented by changing an optical motion was proved for evaluating
element in the VistaView® and handling qualities, a clear need for a
reconfiguring the IG through its platform motion system was present.
AREA OF INTEREST AREA OF INTEREST
PROJECTOR PROJECTED IMAGE AREA OF INTEREST
OCCULTATION PROJECTOR
OCCULTATION
COCKPIT
OCCULTATION COCKPIT
OCCULTATION

Fig. 4: Area-of-Interest Concept

 23-7

Fortunately the NSF could be built on 5.5 Cockpit fidelity

the synergistic six degrees of freedom For the large amount of human factors
motion system already made on specific studies to be performed with the NSF,
NLR requirements by the Dutch firm a choice was made for a high fidelity
Hydraudyne. The NSF 6-DoF system can be cockpit. Instead of generic controls
regarded as a high-performance system, and displays, a dedicated cockpit is
its main specifications listed in table needed when training studies involve
1. parameters like pilot workload.
An F-16 MLU cockpit was procured from
Table 1: NSF 6-DOF motion system specifications Lockheed Fort Worth Company by the
Royal Netherlands Air Force (RNLAF).
The NSF F-16 cockpit consists of a
real aircraft section, adapted by
displacement velocity acceleration Lockheed for simulation purposes and
for placement on a motion platform.
Most of the instruments, displays and
pos neg (pos/neg) (pos/neg) controls are exactly as in the
aircraft, however some specific
instruments (like altimeter) had to be
longitudinal 1.72 1.34 m 0.8 m/s 8 m/s2 exchanged by special simulator
lateral 1.39 1.39 m 0.8 m/s 8 m/s2 instruments.
vertical 1.01 1.14 m 0.8 m/s 10 m/s2

roll 30 30 deg 30 deg/s 200 deg/s2 Another important aspect of creating a

high fidelity environment is the
pitch 29 29 deg 30 deg/s 200 deg/s2
ability to reproduce correct
yaw 40 40 deg 30 deg/s 1 50 deg/s2 communication and aircraft related
sounds. The sound generation system is
a digital system based on the Paradigm
fully hydrostatic actuators Emax II sound synthesizer and state of
the art digital sound processors
band width: 45 degrees phase lag at 4 Hz
integrated through a SGI-Indigo with
AudioWorks and built by the Dutch firm
CHeSS. The sound system not only
Nevertheless, for fast jet (and performs all sound generation for the
helicopter) simulation some additional F-16 cockpit such as aerodynamic hiss,
motion cueing is necessary. The high-g engine and Voice Message Unit sounds,
environment in which a fast jet pilot but handles also all communication
operates during attack missions can be routing between cockpit, control desk,
regarded as a "disturbance" to his mockup and experiment controller.
normal behaviour. Apart from a human The sound system is prepared to handle
centrifuge, some sort of equivalent three dimensional sounds, which will
flight perception can only be generated possibly can be a next step in
by a G-cueing device. Because of lowering pilot workload and improving
latencies normally found in "older" g- situational awareness in the aircraft.
seats due to the pneumatic systems
used, specifications for the NSF g-seat 5.6 Simulation Software
stipulated very low (less than 70 Incorporating the necessary innovation
milliseconds) delay. in simulation software for the NSF led
to one of the project's critical
The contract for the NSF G-cueing development items. Having made the
system was awarded to Sogitec, which decision to change the host computer
develops and integrates the system and the type of operating system, the
based on NLR-specific requirements. It NSF projectteam faced a major redesign
will consist of a G-seat, an anti-G of the existing real time scheduling
suit system, helmet loader and lap-belt software and the existing (simulation
restrictor. A Partial Pressure Positive model) software development tools.
Breathing system had also been From the end of 1992 until the
considered, but the medical beginning of 1994, NLR researchers
implications to use it on "real" pilots together with subcontractors Fokker
outweighed its usefulness. Space & Systems and BSO/Origin
Integration of the G-cueing system will developed, a completely new software
take place in the beginning of 1996. tool, called PROSIM (Programme and
Real time Operations Simulation
support tool) (Ref. 3). PROSIM
incorporates the latest information
23-;

technology on real time scheduling and troubles are now slowly being solved
software development. and done with.

The predicament "full mission" for a

simulator can only be given when not
only the aircraft's internal systems
Mission are correctly simulated to respond on
Planning System external threats, but also when the
external environment reacts in the
same correct fashion. An aircraft
Global Data definition
Global Data Description operating below a certain altitude and
Scheduler
Error Definition making use of surrounding high terrain
Global Data Initialisation
Mission Definition can shield itself from being detected
by missile sites. The NSF makes use of
terrain masking models, having been
converted from the visual database, to
make ownship and threat line-of-sight
calculations. This data is also used
by ITEMS, the "Interactive Tactical
Environment Management Simulation"
package of CAE Electronics. ITEMS was
Data Processing procured to handle all threat, target
Software
and networking (including DIS)
computations. Operating on a stand-
Fig. 5 : Top-level functionality of PROSIM alone SGI Indigo W8C, it hooks in the
simulation software through a SCRAMNet
A completely new aspect, compared to reflective memory interface.
operating the RFS, was the necessity of
having a full avionics simulation suite 6. NSF RESEARCH PROJECTS
for the NSF. During the entire Current research experiments scheduled
existence of the NLR Research Flight for this year focus both on pure
Simulator, no detailed functionality of research and system development. It is
all of the aircraft avionics systems paramount to our facility that all new
was needed. To be able to develop this systems incorporated in the NSF
avionics simulation software for the F- function in a fully synchronized
16 MLU, engineers from LFWC's Flight manner. Correct functioning of the
Simulation Laboratory were contracted various motion control and wash-out
to assist the NSF team. Large portions schemes have to be established for
of the avionics software used at the various flight conditions. Also all
Lockheed simulation facility were subsystem computational delays and the
transferred to the NSF, thereby total system time delay will be
creating a nearly 100% concurrency of researched.
the NSF with the MLU aircraft. In the second part of this year a,
Before converting and integrating this government funded, technology
software, a special team had laid out a demonstration project will show the
design to be able to incorporate an benefits and drawbacks of using speech
avionics "hot bench" functionality. recognition and commands in the F-16
When available, the NSF can operate MLU aircraft operating in different
real aircraft avionics or switch to its threat conditions.
software counterpart; whatever is
needed for a specific experiment. During 1996, the NSF will be heavily
involved in the EUCLID CEPA-11
To have all decentralised computer programmes RTP-11.1 and -11.2, amongst
systems operate in a coherent fashion, others.
proved to be no sinecure. Although most
interface computers operate with 7. CONCLUSION
identical processors and software, each Development of the National Simulation
computer runs, in a sense, autonomously Facility phase 1 has been finalized
from the host: the reflective memory end 1994. The research facility is
data being the only part shared at the unigue in offering capabilities such
same moment. Specially during the first as high-performance motion and visual
days of operating in real time mode, cueing together with a full- mission
the various interface computers had a environment. The NSF is the world's
life of their own. The "Power ON/OFF" first motion-based F-16 MLU simulator.
button had to be used on many The facility will not only offer
occasions. However, these teething extensive research and development
possibilities for both Air Forces and
 23-9

industry, but from an operational point 8. ABBREVIATIONS

of view, it will also be an asset in
defining training (simulator) ATC Air Traffic Control
requirements. CEPA Common European Priority Area
DIS Distributed Interactive
During the 1995 time frame, the NSF Simulation
will be extended to allow for advanced EUCLID European Cooperation for the
helicopter simulations. Based on high- Long term In Defence
fidelity simulation models, the FOV Field Of View
intrinsic value of the simulator will HUD Head-Up Display
increase dramatically. By virtue of I/O Input/Output
designed and incorporated flexibility IRIS Integrated Raster Imaging
and modularity, the National Simulation System
Facility presents itself as an ITEMS Interactive Tactical
international testbed for advanced Environment Management
simulation possibilities. Simulation
LFWC Lockheed Fort Worth Company
MFD Multi Function Display
MLU Mid Life Update
NLR National Aerospace Laboratory
NSF National Simulation Facility
OTW Out-of-the-Window
RFS Research Flight Simulator (of
NLR)
RNLAF Royal Netherlands Air Force
RTP Research and Technology
Project (EUCLID)
TCAS Traffic-alert and Collision
Avoidance System

9. REFERENCES

1. Offerman, H.A.J.M. 1994

"Development of the Dutch National
Simulation Facility NSF"
Proceedings of the AIAA 1994, Flight
Simulation Technologies Conference,
Scottsdale Arizona, USA

2. Offerman, H.A.J.M. 1995

"Full Mission Simulator for Research
and Development
Proceedings of the 1995 International
Training Equipment Conference (ITEC),
The Hague, The Netherlands

3. Brouwer W. et al., 1994

Fig. 6 : NSF cutaway drawing "The National Simulation Facility NSF:
The application of the Simulation
The NSF will be extended towards Support Tool PROSIM"
simulation of two helicopters, recently Proceedings of the ESA 3rd Workshop,
procured by the Dutch Government, the Simulators for European Space
Boeing CH-47D medium-heavy transport Programmes, ESTEC, Noordwijk, The
helicopter and the McDonnell Douglas Netherlands
AH-64D Apache attack helicopter.
The NSF wil support the Air Force in
operating these helicopters, a new
tactical and operational element to be
used by the recently established Air
Mobile Brigade, the Dutch rapid
deployment force.
 24-1

Simulateur multicible pilote pour le developpement des conduites de tir air/air

J.-E. Chevillot
Chef de la section Etudes et Simulation
Centre d'Essais en Vol
Base d'essais d'Istres
F-13128IstresAir
France

menaces sol/air. Le cockpit des a6ronefs pilotes est soit une

1. RESUME cabine simplifi6e avec toute la planche de bord sur un m6me
La section Etudes et Simulation du Centre d'Essais en Vol grand ecran tactile et un retroprojecteur presentant l'image du
(Ministere de la Defense) participe etroitement au developpe- monde exterieur, soit une mini-sphere COSMOS developpee en
ment des systemes de navigation et d'armements (SNA) de tous cooperation entre Thomson TTS et le CEV. Dans les deux cas,
les programmes d'aeronefs (Rafale, Tigre, Mirage 2000, ...). manche et manette proposent les vraies commandes temps reel.
De plus, tous les SNA proviennent du portage des logiciels
Les essais en simulation de ces SNA necessitent la mise en SNA de simulations d'etudes precddentes d'avions de combat,
oeuvre de cibles ou d'equipiers reconstituant ainsi l'environne- permettant ainsi un choix de plusieurs aeronefs realistes et
ment tactique du pilote. Mais en fonction de l'etat d'avance- valides sans developpement supplemental.
ment des travaux, le besoin varie depuis un environnement
tactique automatique jusqu'ä la mise en oeuvre d'equipier ou de Le document presente le description d£taill6e du moyen, les
cibles pilotes pour un plus grand niveau de realisme. premiers resultats de son utilisation en evaluation opera-
tionnelle ainsi que les axes de developpement ulterieur.
Ce premier besoin a entraine le developpement d'un simulateur
complet capable non seulement de restituer l'environnement 2. CONTEXTE
tactique en mode preprogramme, mais aussi de simuler jusqu'ä La Direction des Constructions Aeronautiques (DCAe) a
trois ou quatre aeronefs pilotes en assurant un parfait concentre au sein du Centre d'Essais en Vol (CEV) l'ensemble
determinisme et le controle de l'echange des donnees des SNA. des activites de simulation pilotee au profit du developpement
II etait de plus necessaire de pouvoir coupler ce moyen ä des grands programmes aeronautiques (Rafale, Tigre, Mirage
chacun des simulateurs d'etudes de la section. 2000,...).

Le Systeme appele HEP ou Hostile/Equipier Pilot6 consiste en Situee sur la base d'essais d'Istres du CEV, au coeur des essais
un calculateur central temps reel simulant deux ou trois en vol de prototypes et d'integration SNA aussi bien des avions
aeronefs pilotes avec un SNA complet, jusqu'ä 20 cibles air/air d'armes que des helicopteres, les missions de la section Etudes
automatiques avec leur SNA lui aussi automatique et 100 et Simulation sont:

Evaluation
Besoins operationnels
operationnelle

\ /
Specifications
Essais en Vol
fonctionnelles

Simulateurs CEV
/
Simulateurs d'etudes \ Integration fonctionnelle
industriels et etatiques Conception architecture
Mise au point/developpement
=> materielle et logicielle
dont CEV

\
Conception des materiels Integration partielle
et des logiciels embarques des sous-systemes

Figure 1 - la simulation dans les processus de developpement

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?"
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
24-2

a. etudier ä moindre coüt les concepts nouveaux, aussi bien aussi faible que possible. II est done indispensable de faire
dans les domaines de l'aviation civile que de appel pour ces simulations ä des SNA bien identifies et au
l'aeronautique militaire, comportement valide.
b. fournir les elements d'appreciation technique ddterminant
l'approbation ou les evolutions de specifications 3.2. Besoins materiels
necessaires ä l'adequation des systemes evalues aux Cependant, si le realisme du SNA doit etre recherch6, que ce
besoins operationnels, soit pour un essai avec une patrouille homogene ou pour une
c. analyser les situations ä risque ou ä charge de travail patrouille heterogene de deux aeronefs amenes ä cooperer, il est
elevee, souhaitable de pouvoir faire appel pour la simulation de cet
d. rechercher les causes d'anomalies decelees en vol par equipier ä une simulation legere au niveau de l'environnement
rejeu ä partir des donnees enregistrees en vol, materiel du pilote pour des raisons evidentes de coüt d'une
e. de former ou d'entrainer les equipages d'essais. simulation d'etudes complete. Aussi bien pour l'environnement
visuel que pour le cockpit lui-meme, la mise en oeuvre de
Ainsi, cette intervention de la section Etudes et Simulation du materiel economique etait recherchee.
Centre d'essais en vol dans le processus de developpement des
programmes militaires et civils depuis les etudes de definition II reste une contrainte ä respecter au niveau des actions reflexes
jusqu'aux evaluations operationnelles des systemes embarques du pilote, le plus souvent effectuees ä travers les commandes
peut etre resumee par le schema de la figure 1. temps reel des manche et manette, en application du concept
3M (Mains sur Manche et Manette en francais) ou HOTAS
3. BESOINS (Hand On Throttle And Stick en anglais). Cet environnement
simplifie doit done utiliser les manche et manette reels.
3.1. Besoins fonctionnels
Les essais en simulation des systemes d'armes multicibles 3.3. Autres besoins
necessitent de mettre en oeuvre des scenarios composes d'un D'autre part, des simulations pilotees simplifies sont
grand nombre de cibles ayant des comportements de plus en necessaires pour:
plus realistes. Pour cela, l'animation des cibles peut etre realisee
simultanement de deux facons distinctes : assurer des fonctions d'enseignement pour l'Ecole des
Personnels Navigants d'Essais et de Reception (EPNER),
automatiquement par l'intermediaire d'un logiciel de ou de formation propres ä la section Etudes et Simulation,
gestion des trajectoires, avec possibility de contröle par un la aussi avec un certain niveau de realisme dans la
Operateur unique. Ce logiciel est soit implante sur le simulation du SNA avec des concessions acceptables dans
calculateur de simulation en etude, soit integre ä un le domaine de l'environnement materiel, l'objectif etant
gestionnaire d'environnement tactique sur station de une demonstration pedagogique et non une mesure ou une
travail dediee ä cette fonction, reelle evaluation de l'interface homme/machine,
permettre de desactiver certaines chaines d'etudes
par pilotage des leaders respectifs des groupes de bom- devenues trop onereuses ou attribuees ä d'autres essais
bardiers ou d'escorteurs au moyen d'un simulateur tout en conservant la memoire des essais effectues et des
simplifie pilote. Dans ce dernier cas, les trajectoires et le SNA valides.
SNA des equipiers de ce leader pilote doivent etre
asservis automatiquement ä celui-ci par le gestionnaire 4. ETUDES PRELEVIINAIRES
d'environnement tactique du paragraphe precedent. Dans ce but, le Centre de simulation a developpe ces dernieres
annees un logiciel de gestion automatique des patrouilles et de
Si le choix de scenarios automatiques est indispensable pour les l'environnement tactique, ainsi que plusieurs concepts de
phases de mise au point initiale d'une conduite de tir ou de simulation simplified repondant ä des objectifs bien definis.
familiarisation de l'equipage d'essais avec le nouveau Systeme ä
evaluer, il est en revanche necessaire de prendre en compte le 4.1. Logiciels DALAI et ENVTAC
comportement du pilote dans les cibles lors des evaluations Le premier besoin presente au paragraphe 3, la gestion
operationnelles et des phases de qualification du Systeme automatique des trajectoires des cibles, a entraine le
d'armes. developpement du logiciel DALAI (Dispositif d'Animation des
Logiques Aeriennes d'Interception) par la section Etudes et
Un deuxieme grand besoin a couvrir est la capacite ä restituer le Simulation. Une des caracteristiques les plus originales de ce
contexte operationnel du travail en patrouille. En effet, avions logiciel est la gestion de la coherence de la patrouille suivant
d'armes ou helicopteres, rares sont les cas oil l'aeronef travaille des principes mis au point avec des pilotes operationnels, en
seul. Le plus souvent, le dispositif basique est la patrouille adaptant les evolutions au sein de celle-ci au type de formation
legere de deux avions, et il est essentiel de restituer la charge de retenue (formation de manoeuvre offensive FMO ou defensive
travail liee ä : FMD, patrouille serree) et au facteur de charge de Involution.
Ainsi suivant les cas, il peut y avoir croisement ou non au sein
la conduite de la patrouille, et la gestion de la coherence de la formation. Ce logiciel est capable de gerer plusieurs raids
de celle-ci, de bombardiers avec leurs escortes air/air ainsi que d'eventuels
la repartition des objectifs ä traiter au sein de celle-ci. brouilleurs stand-off.

Ceci ne peut etre assure de maniere realiste que si deux equi- Tres rapidement, il s'est avere necessaire d'ajouter ä cette
pages "humains" (et non preprogrammed) sont dans la boucle. gestion des trajectoires des cibles air/air une simulation
automatique de leur systemes d'armes de facon d'une part
Enfin, il faut pouvoir mesurer les resultats obtenus et il faut que d'exciter les contre-mesures de la simulation en essais mais
l'accoutumance au pilotage de ces cibles ou de cet equipier soit aussi d'obliger le pilote de cette derniere ä tenir compte des
 24-3

capacites de tir adverse. Le logiciel DALAI s'est done vu d'une image television alors que ce type de figuration est
integre au sein du logiciel ENVTAC (ENVironnement habituellement genere par des tubes ä balayage cavalier.
TACtique). Celui-ci est capable de simuler 20 menaces air/air et
leur Systeme, jusqu'ä 100 menaces sol/air et 10 bateaux. Enfin, des tests pour l'ajout d'une fenetre frontale d'une image
du monde exterieur ont ete realises et semblent prometteurs.
Au logiciel ENVTAC est associee une console de contröle sur
laquelle est presentee une situation 2D du scenario en cours 4.4. Enseignements et limitations
ainsi qu'un poste de commande. Cette console est destinee au Ces premiers developpements ont ete effectues sur des
suivi du scenario, et ä l'intervention sur les composantes non structures informatiques independantes, stations de travail pour
pilotees pour modifier le deroulement des scenarios le PPS et les logiciels DALAI et ENVTAC, chassis VME
preprogrammes, declencher des evasives ou des manoeuvres de specifique pour le COSMOS. Si les concepts ont donnes
combat, activer/desactiver des menaces sol/air,... De plus, satisfaction, d'importantes limitations sont vite apparues. Les
l'ingenieur dispose d'une image pseudo 3D vue depuis le premiers enseignements sont done les suivants :
cockpit d'un quelconque aeronef du dispositif voire d'un
missile, presentant les parametres essentiels et la position le logiciel ENVTAC repond aux besoins d'essais et ses
angulaire relative de tous les mobiles. Cela permet en capacites en nombre de menaces sont süffisantes pour la
particulier un suivi temps reel des tirs simule et l'interpretation complexite des scenarios ä traiter. Les outils de
des comportements de selection de cibles par l'autodirecteur. preparation de scenarios en temps diffdre et le poste de
contröle en temps reel necessitent certes un apprentissage
4.2. Postes de pilotages simplifies (PPS) mais sont d'une convivialite süffisante. Enfin, la
Le premier simulateur simplifie developpe a ete le concept PPS presentation 2D du dispositif associee ä la vue cockpit
(Poste de Pilotage Simplifie). II s'agit d'une cabine en bois lorsque cela est necessaire permet une interpretation aisee
reprenant les formes principales exterieures et interieures du de la situation. II n'y a done pas de besoin de recherche
cockpit reel, dans lequel le pilote dispose des manches et complementaire pour une presentation 3D plus complexe.
manette reels ou au moins fonctionnellement identiques. le premier PPS disposait d'un logiciel SNA simplifie.
L'ensemble de la planche de bord est regroupe au sein d'un Cela posait des problemes de mise en oeuvre par les
grand 6cran tactile qui permet la selection des differentes pilotes qui ne retrouvaient pas un avion aux fonctions
fonetions au SNA ne necessitant pas une action reflexe du connues. Cela a confirms la necessite de simuler des SNA
pilote. Le monde exterieur, dans lequel les reticules du reels et validds.
collimateur t6te haute sont directement incrustes, est presente la multiplication des structures temps reel posait des
sur un retroprojeeteur place devant la cabine. problemes de mise en oeuvre (lancement non centralise)
et de synchronisation. Le non determinisme des echanges
ä travers des liaisons Ethernet et de la structure temps reel
meme des PPS entrainaient la multiplication de modules
Retrop 'ojeeteur
de datation et d'interpolation dont la validation
fonctionnelle posait quelques problemes en ce qui
concerne les conduites de tir missiles et la guerre
Ecran tactile,
electronique.
la multiplication des structures posait des problemes de
Cabine maintenance, aussi bien des systemes informatiques que
des logiciels applicatifs modelisant les SNA et les sym-
bologies associees, et entrainait des surcoüts materiels.

Figure 2 - Poste de pilotage simplifie 5. SPECIFICATIONS SYSTEME

Un groupe de travail interne ä la Section a ete cree avec pour
Ce concept de cockpit est particulierement bien adapte pour un objectif de definir une architecture informatique capable de
avion en mission air/sol en basse altitude oü la prise en compte federer l'ensemble de ces simulations simplifiees, dösormais
du relief est essentielle. En revanche, il ne permet pas les baptisees HEP soit Hostile/Equipier Pilote, et l'environnement
engagements air/air car le champ du retroprojeeteur de l'ordre tactique automatique complementaire.
de 40°*303 est trop limite pour les evolutions de combat.
Tout d'abord, la realisation d'essais multi-cibles et/ou multi-
Aujourd'hui, le CEV a developpe deux types de cabines suivant aeionefs impose, en raison de la necessaire capacite ä
ce concept, une monoplace et une biplace. interpreter les resultats obtenus :

4.3. Mini spheres COSMOS le determinisme de toutes les composantes en jeu et des
Pour les essais ä dominante combat air/air, le CEV et Thomson couplages entre ces differentes composantes, ä savoir qu'ä
TTS ont developpe le concept de minisphere COSMOS de iso conditions initiales, les phenomenes observes sont
seulement 2 metres de diametre, qui permet de disposer d'une identiques, en particulier non pollues par des processus
presentation omni directionnelle de cibles generees par deux propres au calculateur et externes ä l'objet simule (delais,
projecteurs laser. Le pilote dispose d'un horizon de 180° de retards,...),
champ obtenu par un projecteur fish-eye et d'une symbologie la realisation de rejeux et d'enregistrements.
tete haute directement presentee sur la surface de la sphere par
un troisieme projecteur. Le fait de disposer d'un projecteur En consequence, il faut mettre en oeuvre une liaison
specifique pour cette derniere est essentiel car il permet de deterministe ä haut debit entre la simulation d'6tudes et la
garantir une bonne resolution ä cette projection sous forme structure HEP, ou entre les differents elements du HEP, ä base
de memoire reflechie ou partagee.
24-4
ecran tactile (2 dans le cas d'un biplace) et soit une station de
Deuxieme necessite, la mise en oeuvre de SNA bien connus, travail g6nerant le monde exterieur et les reticules tete haute
reels et valides impose, afin d'eviter des surcoüts de incrustes dans le cas d'un PPS, soit la mini sphere COSMOS
developpement, de reutiliser des logiciels issus des simulations avec sa baie VME qui assure les asservissements de projecteurs,
d'etudes precedemment utilisees, ce qui permet de limiter les les generations de cibles et de la symbologie tete haute.
operations de creation de nouveaux hostiles ou equipiers pilotes
au seul portage de logiciels des simulations d'etudes vers la 6. ARCHITECTURE RETENUE
simulation HEP. II est done indispensable de respecter des
normes strides aussi bien au niveau de la structure d'accueil que 6.1. Generalites
des döveloppements des logiciels de facon ä minimiser les Le but de ce chapitre est de decrire les grands principes de
temps de portage. l'organisation fonctionnelle du Systeme HEP, qui ont presid6 ä
la realisation de la structure d'accueil temps reel.
La necessite d'une maintenance simple impose une meme
structure temps reel aussi bien pour les PPS, les COSMOS que L'ensemble est base sur la repartition des modeles en deux
l'ENVTAC. L'analyse des coüts conduit alors ä un compromis : classes : les modeles critiques et les modeles non critiques. Par
une grosse machine est plus economique que plusieurs petites modele, on entend un module logiciel qui simule le
liees par memoire reflechie, mais est moins souple d'emploi comportement d'une plate-forme ou d'un de ses equipements,
principalement en terme de modularite. En fonction de la ou une fonction (ou partie de fonction) de son SNA. Par
complexite des essais vises opposee ä la necessite de pouvoir exemple, ce peut etre le modele radar, le modele pilote
les maitriser et les exploiter, le choix s'est porte vers une automatique, le modele symbologie tete haute, le modele sons
structure capable de mettre en oeuvre 2 aeronefs pilotes au sein et alarmes...
de l'environnement tactique automatique plus une simulation
simplifiee autonome (instructions ou etude legere). Un modele critique produit des donnees qui doivent Stre
fournies avec des contraintes temps reel ä d'autres modeles qui
Dans le meme esprit de simplification des operations de main- les consommeront: par exemple, les donnees issues d'un des
tenance et d'evolutivite, il a ete decide d'uniformiser le principe radars sont des entrees des contre-mesures des autres SNA, etc.
d'un grand ecran tactile simulant la planche de bord identique Pour un modele non critique, un retard induira peut Stre une
pour les PPS et les COSMOS, en utilisant pour la generation de legere perturbation, mais ne remettra pas en cause la coherence
cette planche de bord graphique le meme progiciel. de l'ensemble. Dans l'exemple precedent, un retard de quelques
3 ms pourra suffire pour que les contre-mesures des autres
Ainsi, le schema de principe de HEP est le suivant: systemes ne travaillent pas au cycle suivant sur les dernieres

Host/Eq

Host/Eq

Host/Eq

Host/Eq
Reseau
type Scramnet
Figure 3 - Principe general HEP

Le coeur temps reel HEP peut ä etre couple ä chacune des informations radar. Ceci pourra aller jusqu'ä retravailler sur les
simulations d'etudes (rafale, 2000-5, TIGRE, ...) par memoire donnees identiques ä celles du cycle precedent. Le
reflechie. Chaque objet au sein du Systeme HEP, teile une cible fonctionnement des SNA en sera done perturbe. En revanche,
ou un equipier pilotee occupe un processeur auquel est couple un meme retard de 3 ms sur la production des reticules de
par liaison Ethernet le materiel necessaire ä la simulation du pilotage ne sera pas percu par le pilote, qui est pourtant en
cockpit: la station de travail generant la planche de bord sous boucle fermee sur la simulation. La generation des figurations
 24-5

tete haute, et plus generalement de l'ensemble de 1'IHS

(Interface Homme Systeme, ou MMI, Man Machine Interface Dans le cas d'un HEP couple ä une simulation d'etude et
en anglais) est done paradoxalement non critique au sens HEP. eventuellement d'autres HEP, trois mecanismes sont mis en
jeu : un mecanisme de contröle et de synchronisation en
6.2. Description fonctionnelle d'un HEP provenance de la simulation maitre, un mecanisme d'echange de
Decrivons maintenant un Hostile ou Equipier Pilote1 donn6, tel donnees avec les autres constituants de l'essai, et un m6canisme
qu'il est resume par la figure 4. de configuration de l'ensemble du Systeme HEP.

Contröle et Donnees exterieures Configuration

Synchronisation (BDD, intervenants multicible externes,...)

JÜ :

Donnees propres
H
HEP E n
P

3mndea

-
Modales critiques H
(producteurs de donnees ä synchroniser) E n
P

Mod.S

Modeles de simuls
valides et
disponibles -
H
E n
P
Interface Interface Intsrf.
ft

Mi. -
Modeles non Modeles environnement
critiques: cabine: H
Commandes E n
- Visus - Horizon et cibles COSMOS Hardware HEP et P
physiques -Sons - Visuel exteneur modeles correspondant:
- Acquisitions - Poste instructeur - Poste simplifie Mod. A
-COSMOS
-CabineSEM,...

HEP

Repräsentation d£taillee Representation schematique

Figure 4 - Description fonctionnelle d'un HEP

II est constitue: 6.3. Cas particulier de l'ENVTAC

Du point de vue fonctionnel, l'environnement tactique est un
de modeles critiques, simulant le comportement d'un HEP comme un autre. La figure 5 ci-apres resume la
aeronef pilote (plate-forme, capteurs, conduites de tir,...); modelisation. La difference reside dans l'environnement propre
ces modeles sont actives cycliquement, de ce HEP qui n'a pas de cabine, et dans le fait qu'une console
d'interfaces de couplages entre les modeles critiques et les specifique est mise ä disposition de l'ingenieur d'essais en salle
moyens d'environnement faisant partie des HEP; ces d'ecoute et de conduite d'essais.
moyens sont consumes, pour un HEP donne :
. d'une cabine ou d'une mini sphere COSMOS constituant 6.4. Description fonctionnelle du Systeme HEP complet
le poste de pilotage duquel sortent les commandes La figure 6 presente le Systeme HEP complet dans son
physiques, notamment celle utilis6e pour le pilotage fonctionnement multicible et autonome. On distingue :
proprement dit,
. de modeles d'environnement cabine permettant de le simulateur d'etudes utilisant le mecanisme d'echange de
presenter au pilote un visuel exterieur, un horizon, des donnees avec les HEP (production et consommation), le
cibles , et ä l'instructeur son poste de travail, mecanisme de contröle et de synchronisation, le
. de modeles non critiques presentant au pilote (et ä mecanisme de configuration des HEP,
l'instructeur le cas ech6ant) des figurations, des signaux la salle d'essais comprenant le poste ingenieur d'essais et
sonores, etc. ou faisant l'acquisition de donnees cabine le poste environnement tactique,
(par exemple un poste de commandes de Systeme la gestion du Systeme HEP: administration,
d'armes). developpement, gestion de la configuration du Systeme,
de structures de donnees propres au HEP considere la plate-forme HEP support des HEP individuels dans leur
manipulees par les divers modeles. ensemble, fournissant les services de mise en oeuvre
24-6

Contröle et Donnees extörieures Configuration

Synchron isatbn (BDD, intervenanls muriicible externes,.

Donnees propres
ENVTACT

Modeies critiques
(producteurs de donnees ä synchroniser)

Modeies
ENVTACT

Interface

Modeies non
critiques:
• Visus Station de travail:
- Commandes -Visu DALAI
• Poste de commands

ENVIRONNEMENT TACTIQUE

Representation detaillee Representation schematiqu

Figure 5 - Description fonctionnelle de l'Environnement Tactique

operationnelle (couple ou autonome) et de 7. PREMIERE CONFIGURATION D'ESSAIS

fonctionnement global en multicible couple,
l'ensemble des environnements propres ä chaque HEP. 7.1. HEP prototype
Deux raisons principales ont conduit ä la realisation d'un
Lorsque le Systeme HEP est couple ä une simulation d'etudes, prototype. D'une part, il fallait valider l'ensemble des principes
pour des essais multicibles ou simplement pour la simulation retenus, materiels, informatiques et applicatifs sur un objectif de
d'un equipier pilote, c'est le calculateur de la simulation d'etude taille raisonnable. D'autre part, il est dangereux dans tout projet
qui est le calculateur maitre de la simulation globale. informatique de fixer des objectifs trop lointains.

Coeur de la plate-forme HEP, la structure d'accueil gere les L'objectif de ce prototype est done le suivant: un Systeme HEP
couches donnees, modeles synchrones (critiques et cycliques) et comprenant un environnement tactique automatique, au sein
interfaces des HEP. Elle ne gere pas les environnements duquel deux Mirage 2000 DA peuvent etre pilotes, l'un en
propres ä chaque HEP. Elle fournit les services suivant: cabine PPS et l'autre en COSMOS. Le Systeme HEP doit etre
couple ä la simulation RAFALE.
initialisation et activation des modeles correspondant ä la
configuration des HEP choisie, en multicible ou en De plus, un HEP autonome doit pouvoir 6tre active"
autonome, independamment de l'utilisation ou non du reste en mode
Synchronisation entre HEP participant ä un meme essai en couple, ä des fins destructions ou de demonstration. II doit
ce qui concerne les donnees et l'activation des täches s'agir soit d'un Mirage 2000 DA, soit d'un helicoptere de
synchrones critiques (modeles et interfaces), transport.
synchronisation des HEP participant ä un essai
multicibles avec le simulateur d'etude implique; ceci En ce qui concerne les services implementes dans la structure
concerne les echanges de donnees, et le contröle de d'accueil, l'essentiel des grandes fonctions demandees est
l'activation des täches synchrones, disponible. Le detail est decrit en annexe. En revanche, l'etude
dialogue avec les machines supportant les modeles d'un point important, ä savoir la capacite de disposer de deux
asynchrones (symbologies par exemple); ce dialogue HEP couples ä deux essais independants, fera partie d'une phase
pourra etre de type synchrone ou asynchrone, critique ou ulterieure.
non critique suivant les cas.
 24-7

$1 £-
!£ I | SALLE D'ESSAI
CONSOLE INGENIEUR
SIMULATEUR D'ETUDE • Lancement
• DevetoppemBnt Essaia Mutu-cibtesi
Gestioo de configuration
Structure Modeles
•^L d'accueil
simulations
d'etudes CONSOLE ENVTACT
Donnöes - Visualisations
• Commandos

Ji A~

» JpLATE-FORMEHEP
IGESTIONNAIRE SYSTEME HEP
Ort on

B
da**»
GERANT HEP:
- Administration
K"°""">I -OevftloppemsntHEP
' Gestion da oonfifluration

HEP .
±MB&\

HARDWARE HEP

Figure 6 - Description fonctionnelle du Systeme HEP

Le choix de deux simulations simplifies Mirage 2000 DA, Enfin, il s'agit d'un objeetif de minimisation des travaux
l'une en cabine PPS et l'autre sur COSMOS, repond ä plusieurs necessaires ä la realisation et la validation de ce prototype. En
raisons. Tout d'abord, le Mirage 2000 DA est un avion bien particulier, l'arrivee du second COSMOS n'etait prevu qu'au
connu des ingenieurs de la section et le SNA est, pour les debut 95, ce qui n'autorisait pas un delai süffisant pour une
criteres d'aujourd'hui, relativement simple. Cela presente done utilisation en essais.
l'avantage de limiter les aleas de portage (meme s'il en existera
toujours). D'autre part, les premiers resultats concernant l'environnement
tactique (paragraphe 3) ayant montrö que le produit etait satis-
En meme temps, il s'agit d'un avion bien connu des pilotes du faisant, il a suffi de redecouper le logiciel ENVTAC suivant la
CEV et de 1'Armee de l'Air, ce qui facilite la prise en main du decomposition fonctionnelle presentöe au paragraphe 6.3.
simulateur et permet de pouvoir porter l'accent sur la validation
des concepts de cette simulation simplifiee, sans rencontrer de 7.2. Configuration HEP prototype
problemes sur l'applicatif. II sera ainsi possible de conclure sur La configuration materiel pour le coeur temps reel retenue est
les capacites et les limitations respectives du PPS et du un calculate«- Harris Night Hawk 5836.
COSMOS, ce qui n'aurait pas ete le cas si les aeronefs simules
avaient et6 differents. En particulier, une grande partie des Pour la memoire refl6chie entre ce calculateur et la chaine de
interceptions se passe aujourd'hui ä proximite du sol, et, en simulation d'etudes, le choix s'est porte sur un SCRAMNET™
l'absence (provisoire) d'image du monde exterieur dans le LX de la societe Systran. II permet en particulier une certaine
COSMOS, le moyen le plus adapte en fonetion des essais souplesse sur la distance entre les calculateurs, ce qui est
envisages sur la simulation d'etudes ne pouvait pas etre tranche. appreciable dans un bätiment oü les salles de calcul sont assez
eloignees l'une de l'autre, lorsque le Systeme pourra Stre couple
Limiter le nombre de HEP pilotes ä 2 correspond aussi ä un ä tous les simulateurs du centre.
souci de souplesse de mise en oeuvre. La composante pilotee
n'a d'interet que si eile est mise en oeuvre par des pilotes formes Pour la generation des planches de bord sous ecran tactile, le
ä l'emploi des conduites de tir simulees ou similaires. Sinon, le progiciel graphique VAPS™ de la socidte VPI a 6te retenu, en
caractere realiste et operationnel disparait et il vaut mieux raison de l'expenence dejä acquise par la section sur ce produit.
revenir ä des scenarios automatiques valides avec des II präsente certes l'inconvenient de demander des puissances de
operationnels. calcul relativement importances si l'on veut conserver une
frequence de rafraichissement des images süffisante mais
24-8

permet une reconfiguration aisee des planches de bord en decrite au § 7.3. Ces essais fitant aujourd'hui acheves, il est
fonction des besoins. possible de dresser un premier bilan de cette utilisation.

7.3. Configuration de l'essai RAFALE/HEP L'apport de la plate-forme HEP r6side principalement dans le
Un des points majeurs de l'essai sur le simulateur RAFALE realisme accru qu'elle a confere ä la simulation, et, par voie de
6tait 1'evaluation des conduites de tir air/air et du combat ä consequence, ä la qualite et ä la force des conclusions de l'essai.
proximite du sol. La cabine de simulation est done pour cela
placee dans une sphere de projection qui presente deux cibles L'amelioration constatee, par rapport aux moyens
generees par des stations de travail sur toute la surface de la d'environnement multicibles de la generation precedente
sphere et une image frontale du monde exterieur sur un champ (DALAI), se situe ä deux niveaux :
de 130°x35° generee par une machine 3 canaux VIS A4 de
Thomson TTS. Naturellement, un traitement logiciel permet de d'une part les manoeuvres effectuees par les "mobiles
selectionner au sein de l'environnement tactique les deux cibles HEP" sont plus representatives (niveau elementaire),
prioritaires qui seront prises en compte par les g6nerateurs de d'autre part les scenarios sont globalement plus realistes
cibles et les projecteurs specifiques correspondants, en (niveau macroscopique).
privilegiant les a6ronefs pilotes par les deux HEP, les autres
mobiles etant presented par la machine VISA4, mais de ce fait Le niveau elementaire s'explique par la quality du modele avion
limites ä la fenStre frontale. simule sur HEP, qui est sans commune mesure avec un modele
simplifie de cible, associee au fait qu'il y a reellement un pilote
En raison des objectifs de l'essai, deux points ont demande une aux commandes. Grace ä cette amelioration des qualites
validation particulierement soignee: manoeuvrieres des cibles, les modes et commandes specifiques
au combat de l'avion en etude ont pu etre evalues avec une
les interactions radar de l'un avec les contre-mesures des dynamique tres proche de celle d'un combat aerien reel, et, de
autres et ce dans les deux sens de facon ä ne pas plus, multicible.
desequilibrer les rapports de force. Sur ce point, la
maitrise des echanges de donnees et la synchronisation Le niveau macroscopique decoule inevitablement du premier,
des structures temps reel du RAFALE et des HEP sont mais y participent egalement:
des elements essentiels ä la credibilite et au realisme des
scenarios. l'existence de systemes d'armes hostiles ou amis tres
l'intervisibilite entre les differents mobiles, et done la complets interagissant avec celui de l'avion en essai,
prise en compte d'une base de donnees de terrain (BDD) la possibility de simuler une mission avec un 6quipier reel
commune au Systeme dans son ensemble. et les aleas que cela comporte,
l'impact psychologique non n^gligeable qu'engendre sur
Ce dernier point merite quelques precisions car la coherence des les pilotes evaluateurs, la presence d'un ou plusieurs
BDD de terrains est un element critique ä toute simulation autres intervenants humains dans la mission simulöe, et
interactive distribute. II faut non seulement disposer en au qui se traduit par une motivation et une agressivite
moins un endroit de l'ensemble d'un module calculant l'inter- accrues. Ce point est tres important : en effet, cette
visibilite de tous les mobiles actifs, pilotes ou non, et des interactivite entre plusieurs pilotes contribuent ä faire
menaces sol/air au sein de l'environnement tactique, mais il faut oublier le simulateur, done ä renforcer la credibilite des
aussi distribuer cette information sur tous les systemes d'armes. resultats et leur transposition vers un cadre r6el.
La encore, la synchronisation des structures temps reel et la
maitrise des echanges de donnees par memoire reilechie a 6te Tous ces points ont concouru ä augmenter la charge de travail
un atout considerable. Dans le cas present, e'est l'environnement du pilote jusqu'ä un degre comparable ä celui rencontre lors
tactique qui dispose d'une copie de la BBD et calcule ces d'une mission de combat.
intervisibilites.
II convient de revenir sur la coexistence de simulations de
II faut aussi assurer un visuel du monde exterieur coherent pour systemes d'armes complets, rendue possible par le concept
tous les aeronefs qui en disposent, c'est-ä-dire dans le cas HEP. Elle a permis durant 1'evaluation pilotee du RAFALE,
present le simulateur RAFALE et le HEP en cabine PPS (le d'obtenir un contexte d'interaction entre avions du point de vue
COSMOS devant en disposer dans un prochain developpe- de leurs radars, armements et contre-mesures respectifs d'une
ment, voir paragraphe 3.3). Pour des raisons de coüt, le PPS ne richesse inegalde. C'est dans ce domaine, on le verra plus loin,
dispose que d'un visuel bas de gamme sur stations de travail, que s'ouvrent les perspectives essentielles de developpement
mais la coherence altimetrique est assuree par la reprise d'une des HEP.
BDD ayant une altimetrie identique ä celles de la VISA4 du
RAFALE et de l'environnement tactique. Ce visuel dispose Les deux types d'interface differents (cabine PPS et mini-sphere
pour plus de realisme de motifs de phototexture. Enfin, des COSMOS) se sont revelds etre complementaires. La cabine PPS
objectifs reconnaissables similaires ont ete places aux memes a ete plutot utilisee dans le cas d'avions cens6s ne pas engager le
coordonnees sur cette BDD et la BDD VISA4. combat ä vue, bombardiers par exemple, alors que la mini-
sphere COSMOS a ete attribuee ä des aeronefs dont la mission
8. ENSEIGNEMENT - DEVELOPPEMENT ULTEREEUR etait, a contrario, l'escorte ou l'interception, que ce soit en tant
qu'hostile ou en tant qu'equipier du RAFALE. Dans ce premier
8.1. Enseignements essai, la capacity ä travailler sur une image du meme terrain que
L'evaluation en simulation pilotee du premier standard l'avion en etude s'est revelee etre un atout significatif de la
utilisateur du RAFALE a constitue la premiere utilisation cabine PPS. Ceci etait en particulier le cas lorsque le profil de la
operationnelle de la plate-forme HEP, dans la configuration mission simul6e donnait une grande importance relative aux
phases en basse altitude : il a alors ete parfois prgfere de limiter
 24-9

le role du HEP au travail au delä du visuel des cibles (BVR en 8.2.4. Moyens d'environnement
anglais) et de faire l'impasse sur la capacite ä engager le combat Les cockpits "bas coüts" utilis6s ont donn6 entiere satisfaction.
En particulier, la mise en oeuvre de commandes temps reel rea-
listes permet une accoutumance tres rapide du pilote ä cet envi-
8.2. Developpements ulterieurs ronnement simplifie, l'activation des fonctions ä travers l'ecran
tactile n'interferant pas avec les phases ä forte charge de travail.
8.2.1. Generalites
Ce premier essai a d6montre tout le potentiel du HEP dans le En revanche, des developpements sont des maintenant en cours
domaine de l'interaction entre les systemes d'armes puisqu'il pour equiper les cabines COSMOS d'une capacite de visuel
permet d'integrer n'importe quelle simulation existante et, par lä frontal du monde exterieur. Cela permettra ainsi de couvrir tout
meme, d'adapter le contexte en fonction des themes d'etude le domaine d'utilisation depuis le vol basse altitude jusqu'ä
(air/air, air/sol, air/surface etc.). On peut egalement envisager, l'engagement du combat ä vue.
au-delä des etudes de developpement, des simulations multi-
HEP destinees ä parfaire les tactiques d'emploi des systemes Ulterieurement, il pourrait etre interessant d'augmenter le
d'armes embarqufis. La seule limitation ä ces developpements nombre de mobiles aeriens projetes dans les COSMOS.
reside dans le fait qu'il faut assurer une coherence süffisante
entre le niveau de representativite de la modelisation des 8.2.5. Systeme HEP
differentes simulations mises en oeuvre, or celles-ci pouvant ne Plusieurs evolutions du Systeme HEP sont envisagees ä plus
pas avoir ete concues pour interagir avec un environnement long terme. Parmi celles-ci citons le developpement d'une
aussi riche, il conviendra de les adapter voire de les enrichir capacite de couplage DIS, qui permettrait un interessant
afin d'utiliser pleinement les possibilites offertes par le concept elargissement du champ d'action et d'exp6rience compares de
HEP. ces deux concepts d'interactivite aux objectifs tres differents.

8.2.2. Environnement tactique 9. CONCLUSION

Le principe lui-meme de l'ENVTAC ne devrait que peu evoluer. En resume, revenons ä la question fundamentale: pourquoi
En revanche, les capacites devraient augmenter en terme de HEP a-t-il ete d6velopp6 ? II s'agissait, pour evaluer finement le
fonctionnalites pour l'utilisateur, de fonctions simulees et de comportement d'un Systeme d'armes dans un environnement
nombre d'entites. L'6volution la plus notable au niveau des realiste, de creer un reseau local faisant interagir un nombre
fonctions va etre l'introduction d'un simulateur de r6seau MIDS limite d'aeronefs pilotes, typiquement inferieur ä 8, au sein d'un
dont les specifications sont en cours d'elaboration. environnement tactique complexe et varie.

En ce qui concerne le nombre de menaces, la situation actuelle II fallait done creer, sans contrainte de recup6ration
permet dejä des scenarios complexes. Cependant, pour plus de d'architecture existante, une structure permettant d'etablir et de
souplesse dans la definition des scenarios, le nombre de mobiles maitriser le dialogue et done les interactions entre plusieurs
air/air va augmenter de facon ä disposer d'un contexte aerien modeles de systemes d'armes. Le probleme est done inverse de
important pendant toute la duree de missions longues. celui pos6 par DIS, ou plusieurs simulateurs existants, pouvant
etre tres eloignes les uns des autres et de caractere parfois tres
Un autre axe de developpement ä court terme est l'introduction het6rogene, doivent interagir.
d'un module de combat automatique du type de ceux dejä
utilises dans certains simulateurs d'entrainement operationnels. Aujourd'hui, HEP repond, des sa version prototype,
Le principal interet est de gerer une patrouille d'escorte de parfaitement au probleme pos6 et permet d'assurer un haut
maniere entierement automatique avec une qualite acceptable niveau de r6alisme des engagements a6riens, avec un souci de
par rapport ä un aeronef pilote, en permettant de concentrer les minimisation des coüts de l'environnement materiel en fonction
deux COSMOS maintenant disponibles ä la simulation d'une des besoins des pilotes grace ä trois types de cockpit : cabine
autre escorte homogene. d'etude complete pour l'essai proprement dit, mini-sphere
COSMOS ou poste de pilotage simplified
8.2.3. Aeronefs simules
Des maintenant, le developpement d'un aeronef capable de Le champ d'application de HEP apparait tres vaste, pour toute
mission air/sol tout temps est en cours. II s'agit de la application locale demandant le realisme de la modelisation
reconfiguration de l'actuel PPS 2000D au standard HEP. globale et la capacit6 d'analyse des phenomenes rencontres:
Lorsqu'il sera operationnel, cela permettra de disposer d'un
dispositif de raid adverse constitue' d'une patrouille de 61aboration des tactiques d'emplois des nouveaux
bombardiers en tres basse altitude, capable de suivi de terrain systemes embarques,
automatique, accompagn6 de deux escortes air/air dont tous les etudes de concept amont dans le domaine n contre p afin
leaders de patrouille sont piloted, leurs equipiers ayant des de valider les orientations de developpement des systemes
trajectoires automatiquement asservis ä celui du leader. Le futurs.
leader de la patrouille de bombardier 2000D utilisera la cabine
PPS biplace, les deux leaders des escortes pilotant les deux Les outils sont lä pour garantir le realisme des comportements
COSMOS. relatifs des differents intervenants, ä condition que le niveau de
repr6sentativite des modeles soit süffisant. Paradoxalement e'est
La transposition des autres simulations d'etudes actuellement la structure informatique, frein classique ä l'affinage des
disponibles ä la section Etudes et Simulation vers la plate-forme modeles informatiques, qui, dans ce cas precis, motivera une
HEP suivra ensuite en fonction des besoins des essais en cours. mise ä niveau de la mod61isation.

ANNEXEA
Planches de description du concept COSMOS.
24-10

ANNEXE A

COSMOS VISUAL SYSTEM

DESCRIPTION

Up to 4 tar£

Micro-dome

HUD

Forward view
terrain Horizon
line

Pilot's cockpi

Figure 1 : Description de la cabine COSMOS

 24-11

O
s
o

3
ft

11
s
M
 27-1

Issues in the development of training analysis methodologies

Terry Simpson
Thomson Training & Simulation Ltd
Gatwick Rd
Crawley
West Sussex, RH10 2RL, UK.

first solution proposed may not be the best one. This

SUMMARY is because the problems expressed above are multi-
dimensional and there are many different ways of
Training analysis is a topic that is often talked about addressing each of them. The methods of 'training
and less often carried out. Various methods have needs analysis' or 'training analysis' provide a
been proposed in the past, but many of these have systematic way of examining problems such that
been too complex and too prescriptive for practical training solutions can be based on a credible process.
purposes. Major system procurements on cost-plus
contracts in the past may have justified detailed The scope of training analysis is extremely wide and
methodologies, but the majority of modern training this has led to the term sometimes being used
analysis is required on small fixed price contracts. inappropriately. This paper describes some of the
The training analyst is often solely responsible for the issues that are involved in training analysis so that no
work, or working in a small group of 2 or 3. matter what particular methodology is used, the
Consequently, the analyst has to use techniques that sponsor of the analysis should be able to recognise
can be easily applied and be flexible enough to carry the key elements.
over from one contract to another.
This paper describes a range of issues that affect
training analysis and presents a simple methodology 2. MYTHS OF TRAINING
that can be easily adopted by analysts and applied to
commercial situations. (After references 1 and 2)

2.1 Myth: practice is training.

1. INTRODUCTION There are several forms to this myth, for example an
over-reliance on experience rather than training 'I
Consider the following situations: don't need courses, I have done this job for years'.
• An air force is introducing a new aircraft type. Even those designing training programmes can fall
• An army faces an increasing number of low- into the trap: 'sit by Nellie' or 'job shadow', are both
intensity, multinational operations. popular techniques that rely on the theory that
• Naval personnel with experience using observing practice and taking part constitutes
mechanical equipment find that their jobs have training. A more successful approach is to
changed and they now have to operate electronic decompose the job into tasks that require specific
equipment. skills and to train the individual skills separately
• Office workers find themselves surrounded by before combining them. Consequently, learning to fly
computer technology that they are unable to use an aircraft is separated into psychomotor skills such
effectively because they don't understand it. as manouvering, landing etc, procedural skills such
• The boss of an engineering firm says that many as checks, and cognitive skills such as navigation.
people seeking work do not have the skills for the
jobs available. Practice is important for a range of reasons including:
confidence building, providing a transition from
Each of these situations will be improved by training. training to working, embedding skills and
Therefore we can say that they all have 'training knowledge, providing a period where errors can be
needs'. It is not always easy to define the best tolerated, etc. However, simply carrying out mock
solution to any set of training needs. It is often the tasks is not necessarily useful and may serve to
case that the person who identifies the problem has demotivate trainees since they are unlikely to perform
also instinctively identified a solution. However, the well.

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?",
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
27-2

2.3 Complex tasks require complex training

It is important for the development of staff skills that solutions
the correct balance is made between practice and Since aircraft are very complex pieces of machinery
training. In the military context, practice (exercises) with complex tasks, there is often the assumption that
involves a large number of people. Some people may the training also needs to be complex. The budgets
be required to carry out undemanding tasks to for training are high enough to buy complex training
support the exercise. It is important to try and devices. However, some complex training can easily
maximise the training benefit for as many people as be carried out in a classroom. For example, team
possible. The larger the scale of the exercise, the tasks tend to be very complicated social interactions,
more difficult the problem becomes. yet some useful skills can be learnt with simple
training aids. It is not necessary to equate the task
complexity or the system complexity with the
2.2 Myth: Training requires realism complexity of the training solution.
The reasons that realism is regarded so highly
include:
• training programme designers sometimes do not 2.4 Expert users know how to train
know what are the important cues for a task so Training requires specialist skills. These skills are not
they want to have them all just in case they miss necessarily the same as those in other jobs. Whilst
one; expert users are sometimes very good trainers, it is a
• realistic training is so hard to achieve that it mistake to believe that this is always the case or that
acquires a high status associated with expensive the experts need little additional training or support
solutions; for developing the specialist skills they need to pass
• subtask training is sometimes difficult and their expertise onto others.
trainees would often prefer to progress to the real
task.
2.5 Myth: stressful training is good training
The whole of the simulation industry is based on Novices sometimes find that the process of learning
providing analogues of the real world. The demand new skills is stressful. This is often because several
for high quality representation of the real world skills have to be adopted at once. For example, flying
(fidelity) has provided a driver for sophisticated aircraft circuits requires many skills e.g. manual
devices. In some areas, fidelity is simple to provide tracking tasks, procedural tasks such as conducting
(e.g. cockpit switches), in other areas it is extremely checks, perceptual tasks such as searching the
difficult (e.g. high resolution daytime images). The ground, airspace and instruments, communication
never ending demand for better fidelity makes it easy tasks such as making radio calls and handling radio
to confuse simulator fidelity with simulator calls as they come in, etc. All of these activities have
effectiveness. In any case, fidelity is much easier to to be carried out in real-time and the novice has to
measure than training effect. devote conscious attention to many of them. This is
why the novice feels overloaded when the expert
The important issue for training programme design is does not. Most of these lower level skills have
that the correct cues are available to support the task behavioural components e.g. speaking on the radio,
being trained. An air combat simulator may have a where performance is easily assessed, and stress is
requirement for high resolution visual detail on target apparent e.g. in voice pitch.
aircraft but very little on the home aerodrome.
Conversely a civil aircraft simulator will need very With strategic tasks activities such as time planning
little detail on other aircraft and very high resolution are more important determinants of performance than
on the home aerodrome. behavioural skills. This makes performance less easy
to assess and stress becomes less apparent. The
It is of course, very difficult to assess what is the trainee working moderately hard may appear to be
minimum cue required for training effect on any task. calm. This calm exterior can lead an instructor to
Indeed, most simulators are required for a variety of believe that no training is taking place, the temptation
tasks so it is tempting to specify the highest possible is therefore to add extra tasks into the training session
cue fidelity. However, it is essential to remember that until trainee stress becomes visible.
demanding realism will not necessarily provide good
training.
2.6 Myth: training that was good for me will be
good for others
This myth arises from resistance to change if an
alternative training strategy is proposed. It also
 27-3

comes about from the following logic: I can do the

job. A trainee wants to do the job. Therefore the
trainee wants to be like me. Therefore the trainee
must repeat the training that I had.

The initial training programme encountered by the

skilled operator may well have been designed 25
years ago. This may have been supplemented by
additional training or experience over many years.
The new trainee will have an entirely different
background and will be familiar with interacting with
people and technology in a different way. Therefore
it is important that training programmes should be
updated to take these factors into account.

2.7 Training analysis is great, but I don't need

methods. I can use my own judgement and
common sense.
Since everybody is human, everybody has some
expertise in human science. We have all experienced
training, and it is likely that we have all experienced
both good and bad training. Therefore instinctive and
common sense opinions on training are worth
listening to. However, common sense can be wrong,
e.g. it gave us the flat earth theory.

3. TRAINING ANALYSIS METHOD

There are a variety of methods that are proposed for

the analysis of training. Ultimately, the objective of
training analysis is usually to design new training
programmes. For this reason, many training analysis
methods are called Instructional Systems
Development (ISD) models. According to sources in
Reference 3, there are over 60 ISD models. One of
these models is the Interservices Procedures for
Instructional Systems Development (IPISD) which
has led to a military standard (Mil Std 1379D).

All such models are tools to assist the practitioner

and it is the emphasis on achieving the objective of
good training systems that is important. Each training
analysis has different requirements and therefore a
method has to be specific to the application, or
sufficiently generic to be applied across a range of
domains. In the case of generic methods, Mil Std
1379 has a problem for practitioners because it is too
prescriptive to be used in acquisition programmes
that are greatly divided up into individual fixed price
subcontracts.

What has been found to be useful in Thomson

Training & Simulation is a model that is generic and
less prescriptive so that it can be adapted to the
commercial environment. This is under continuing
development, but the diagram below illustrates the
model.
27-4

System analysis

' r

Task analysis

Define training i
objectives

Define training
Plan curriculum
'' > device
functionality

i r

* Develop training
Plan lessons t-
device

■ '

Conduct training

ir

^
bvaiuat e training

Figure 1. A simple training analysis method

 27-5

4.5 Equipment manufacturers

Simulators have to replicate the appearance and the
4. DATA SOURCES functionality of equipment in the target system. This
means that very specific details are required by a
simulator manufacturer. In a commercial
4.1 Contract environment where the equipment manufacturer is
Contractual obligations have to be met regardless of competing with the simulator manufacturer, data
whether there is training value in the activity. transfer becomes a very important issue.
Sometimes the analysis may highlight a mismatch
between contractual requirements and training needs. Equipment manufacturers should have information
If it is a serious problem, the analyst may have to about equipment design philosophy which is
draw the attention of the customer to the mismatch. extremely important to understanding training
This will allow the customer to make a modification requirements.
to the contract.

In most cases, contractual requirements impose little 4.6 Predecessor systems

extra workload for the analysis. Some training systems have to be designed even
before the target equipment is in service. This is quite
understandable since users need to be trained before
4.2 Customer input they use the equipment. However, this presents real
Customer satisfaction is important to commercial problems for training analysis since the level of user
organisations. Therefore if a customer has a knowledge about the system is likely to be extremely
preference for a system configuration or feature, then low, possibly only available from test users operating
that is what they will get. In training system design it prototypes.
is important to acknowledge that the customer will
have an influence. A common way of dealing with this problem is to
extract data from users of existing systems
It is also important to note that the customer is not performing the same or similar functions.
usually the end user. The person or people
responsible for contract specification, award and
acceptance can be quite different types of people 4.7 Courseware
from the instructors and trainees, indeed some of Most training system design occurs as part of a
them may be in a different domain e.g. development of existing systems. Therefore there are
administration rather than aviation. usually documents such as curriculums and training
manuals. In some cases, all that is needed is an
There may also be people within the customer update of the existing courseware, in other cases,
organisation who have expertise about training there will need to be a reconfiguration of the whole
system development that can be valuable to the system. As an early step in the analysis, it is essential
training analysis. Good management of the to ensure that the courseware is obtained and
customer/producer relationship can produce a examined.
higher quality analysis.

4.8 Subject Matter Experts

4.3 Standards Training analysis depends on having information
There are various military and civil standards that about the task. Some of this information may be
may be complied with. For example, Mil Std known by designers and some of it may be available
1379D. Standards are often difficult to work with, in other documentation. However, most training
may contradict each other and may have been analysis work requires direct access to users for task
superseded by current knowledge. However, if a knowledge elicitation. Since the system is designed
contract requires compliance with standards then for human use, it is clear that the users should be
they will affect the training system. involved in the design process.

It often surprises those attempting to conduct training

4.4 Authorities analyses how difficult it is to obtain user involvement
Authorities can influence the training system either in system design. One way to make user involvement
by the production of standards or by other means easier is to have Subject Matter Experts (SMEs) as
such as certification of the equipment. user representatives. One definition of SMEs is that
they will satisfy the following criteria:
27-6

5.1 Sophistication.
• Appropriate qualification. If the task to be trained Tasks can be relatively simple, such as learning how
is flying a Tornado, then the person should be to use a word processor or they can be complex such
qualified to fly that aircraft. as learning how to deal with aircraft malfunctions. In
the former case, the task is accessible to the analyst
• Task knowledge. They should have task but the latter is so sophisticated, it is almost
knowledge. If the task is multi-aircraft tactics, impossible for a training analyst to have complete
then the person should know how to carry out that awareness. This affects how the analysis is
task. conducted.
• Expertise. They should be good at the task.
• Currency. Their expertise should be current. In
complex socio-technological systems such as an 5.2 Safety.
air force, tasks change very rapidly. Sometimes Some systems have safety as a priority. Examples are
information that is only one year old can be nuclear power stations and civil aircraft. This biases
incorrect. training towards error reduction.

These requirements are quite difficult to achieve, but

it is important to strive for the ideal. If access to 5.3 Performance.
SMEs is not possible, then non-SMEs may be used as Military fast jet training prioritises performance and
long as their limitations are acknowledged. It is the time domain becomes more critical.
sometimes a problem with training analysis that
somebody is made available to the study and
declared to be an SME when they do not match the
Ease of use.
above requirements. This is a particular problem in
aviation where there are lots of enthusiasts who have Civil systems such as cars, ticket machines, and
some limited knowledge of the domain. office equipment tend to focus on ease of use.
Modern computerised equipment is now expected to
be operated without any specialised or dedicated
training at all. The training therefore needs to be
4.11 Research
reduced to a minimum or embedded within the
A lot of work is being done in research centres on operation of the equipment.
training. This research changes the thinking about
how training needs can best be met. It also changes
thinking about methodologies.
Scope.
Close links with research organisations can help It is important that the scope of the task is defined.
provide an effective analysis. The training analysis can only examine a range of
tasks and the scope of the tasks has to be defined.
Scope definition prevents the analysis from being
drawn outwards. For example, scope can define a
4.12 Observation
start point as being qualified to operate the aircraft
Documents can be quite specific about what happens and an end point as being qualified to carry out a
in any job, and SMEs can also give details about given range of missions successfully.
what happens. However, it has been demonstrated on
many occasions that what people say happens and
what actually happens are two separate things. 6. SYSTEM FOCUS
Therefore it is important to observe tasks whenever
possible. The observer will obtain vital information
about the task that would be unavailable otherwise. It
is often the most important information that cannot 6.1 Technology.
be reported verbally or in writing since it sometimes Complex training devices are produced by
can involve users doing what they may think is not engineering organisations where the system focus is
allowed. quite natural on technology. Cost is a technology
factor which affects training analysis, it is now
possible for the analyst to consider computer based
5. TASK FOCUS training since the costs have reduced radically in
recent years. Feasibility is another technology factor
which affects training. For example, high resolution
visual displays representing daylight terrain and
airports can now be suggested on simulators.
 27-7

Something that meant training on real aircraft is now Relatively slight modifications to the training devices
feasible in a simulator. The training need has not procured as a result of the analysis can help training
changed, only the technology. for the other jobs.

6.2 People. 7.4 Range of experience.

Systems involve people as well as technology. This is There will be a range of experience in system users
often forgotten, probably because so much effort is from trainee to test pilot. There may be the person
expended getting the technology to work. The focus who is trying to obtain remedial training to retain a
of the analysis is on the whole system and that means job. This person will want simple clear training
that a lot of effort needs to be spent looking at the systems with tactful handling of mistakes. There may
people issues. This is why user involvement is so be the expert instructor with combat experience who
important. relishes challenging situations and needs to explore
the limits of an aircraft envelope.

7. USER FOCUS
7.6 Team stability.
Teams may be stable because they often work
7.1 Stakeholders. together and may even live together, as in the
In any training development programme there is military. They may be unstable so that they may have
usually a range of people who have a legitimate less awareness of what to expect from each other, as
interest in the system. These can be called in civil airlines. Stability will affect team interactions
stakeholders. There will be what are usually called and this affects how people should be trained. Crew
end users. So a word processor or an aircraft may Resource Management is a recent initiative in the
have end-user i.e. those people who have their hand airline industry which addresses the training needs of
on the controls. However, if we consider that most this issue.
jobs serve other people we can consider the end users
as primary e.g. the secretary or the pilot, and we can
also consider secondary users such as a manager or 7.7 Trainee or instructor.
the passengers who are directly affected by the Training systems have to meet the needs of trainees,
performance of the primary user. There may also be but they also have to meet the needs of instructors.
various tertiary users such as the company or airline As training devices get more complex and options
staff. increase, instructors face an increasing workload.
The system should enable instructors to produce
In addition to users there are other stakeholders. interesting and challenging training with minimum
Maintenance staff have an interest in training effort which allows them to concentrate on trainee
programme development. Often their needs are not activity.
considered until late in the analysis process. System
designers have an interest since they have to work
with limited resources and may be recalled to modify 7.8 Future proofing.
a system. Complex and unique solutions may meet Training system development is a time consuming
the training needs, but they may be difficult for and expensive business. Solutions should ensure that
device designers to produce. there is some future proofing built into the system.
Thus if a modification to the real world system is
A Target Audience Description is a useful document made, then the training programme can be updated
to produce as part of the training analysis. This relatively simply.
contains a list of the types of people will be trained
and their characteristics. This can be agreed with the
customer, and the very process of producing the 7.9 Credibility
document helps extract useful user information. There may be some features of training that are done
more for user credibility than for training effect.
Examples include: having a particular lesson with the
7.3 Range of jobs. real equipment rather than in a simulator, enhancing
Many training systems are used for a range of jobs. A simulator fit to include extra controls and displays,
simulator may be used by a pilot, a copilot, an making the training devices look attractive. These
engineer, cabin crew etc. Although the explicit features may not affect training directly, but they may
demand for training may be for one job, the facilities have an indirect effect. The trainee that believes that
may usefully provide training for several jobs.
27-8

training system works well is likely to be more

receptive to the training.

8. TRAINING EFFECT

• Changing people. Courses of training or sessions

using a simulator must be used to change people.
Many people experience the compulsary course
where attendance is all that is required and no
real difference is made to the trainee. The training
programme should be designed around the
objective of producing a change in people rather
than just spending course time on a range of
topics.
• Directed. Training must also be directed. We can
change people, but it must be in ways that we
want. Programmes need to be developed to
ensure that the training objectives are achieved.
• Effective. Training has to be effective. Thus the
benefit must be in terms of task performance,
operational (and training) costs, safety, or
operator satisfaction. It is often difficult to
measure training effect but we should still try to
make sure that training systems are designed for
effect.

9. CONCLUSIONS

Training analysis has been shown to have many

influencing factors. These have been described with
suggested way of dealing with them. Various popular
myths about training have been described with
explanations as to why they are incorrect. A simple
method has been shown which is simple and flexible
enough to cope with the limited scope, resources and
timescale of the modern contract.

10. REFERENCES

1. Klein, G.A., Zsambok, C.E., and Thordsen, M.L.,

"Team decision training: five myths and a model."
Military Review. April 1993, pp 36-42.

2. Pheasant, S., "Bodyspace: Anthropometry,

ergonomics and design." Taylor and Francis, 1986.

3. Patrick, J., "Training: research and practice."

Academic Press, 1992.

4. Mil Std 1379D Department of Defense, USA,

1990.
 28-1

Delayed Pilot Response in Windshear

G. Schänzer & J. Krüger1
Institute of Flight Guidance and Control
Technical University of Braunschweig
Hans-Sommer-Str. 66
D-38106 Braunschweig
Germany

Abstract of the aircraft approximately one nautical mile

prior to the runway threshold (fig. 1). The aircraft
Windshear can cause fatal accidents. To pass attitude is sufficient, the rate of descent is
windshear situations safely, the automatic flight moderate, but the ground is not prepared for an
control should be employed. This advice is in aircraft touchdown. This can be observed
contrast to an FAA recommendation to switch off especially when the aircraft is flown manually. A
the autopilot in a windshear and to perform a go- comparison with fully automatic approaches
around procedure. under the same conditions show a delayed throttle
With conventional instrumentation and manual response caused by the pilot. The delay time of
approach extremely large time delays up to 40 pilots is more or less correlated with the cycle
seconds could be identified. It is still unclear what time of the aircraft phugoid motion and is in the
the psychological reasons for this delay are. There order of 30 seconds. There is still no sufficient
are some indications that the human being reacts physiological explanation for this enormous time
like a band pass filter. With proper display delay. In this paper some simulation results are
information a thrust command director can be presented to evaluate this time-delay effect. The
realized that may force the pilot to react correctly selected windshear type is a low level jet, best
and safely in windshear situations. suitable to produce clear results.

List of symbols 2. Low Level Jet

F thrust The term "low level jet" is used to describe wind

H height phenomena of the lower part of the boundary
t time layer, characterizing jet like wind profiles with a
Uwg horizontal wind component low altitude wind maximum. This kind of wind
wWg vertical wind component profiles has been observed in connection with
V airspeed specific local terrains, thermal effects in mountain
VK flight path velocity valley regions, frontal activities, and the nocturnal
Vw wind speed boundary layer. Usually the nocturnal low level
w weight jet appears in the time between late afternoon and
X, Xg distance/flight path dawn under clear nocturnal sky when a strong
Y flight path angle radiation temperature inversion develops. Because
of the strong stability in the inversion layer
friction disappears and the unbalanced Coriolis
1. Introduction and pressure gradient forces produce an
acceleration of wind speed. Reference [1]
Several fatal accidents, especially those caused by describes the evolution of low level jets as a
windshear finish with an undesired ground contact nonstationary process where the vector difference

1
Since February 1995 Flight Operations Inspector / Flight Simulators at the Federal Civil Aviation
Administration (LBA)
Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?"
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
28-2

between the actual wind and the geostrophic wind one or two minutes. In this case the temporal
is rotating nearly circularly around the evolution is not relevant and modeling can
geostrophic wind (inertial oscillation), see fig. 2. concentrate on quasi-stationary engineering
models. Velocity profiles like low level jet wind
profiles have been observed in fluid dynamic
•ijlUllllUinn mi'
'//Illinium " research of free jet and wall jet. As a first low
^hkU I mill'" level jet approximation a superposition of a
'XM II11""' boundary layer profile (for instance the power
100 law) and a plane free jet velocity profile is used
f/Z/Z/Z/h^J"""
-'//////ZZ/JII11 %^»;,'' (see fig. 4). Measured data and equations for the
so / / / ru ^ free jet have been published among others by [6]
and [7].
-3500 -3000 -2500 -ZOOO -1500 -1000 -500
DISTANCE, x , m
* METAWäremen Airport
++ HEW-MAST.Morne I5:S1 GMTi
Figure 1: Landing approach through a downburst « NKW-MAST. Stade
-.- fixed controls
.... autopilot (no autothrottle)
— reconstructed flight path of an
accident occured in New York on
June 24, 1975

—(r~
/i
10m/j 120° 160» 200°
Wind Speed | Wind Direct.| |Temperature|

Figure 3: Low level jet measured in-flight

"I
Figure 2: Inertial oscillation of wind vector

In the northern part of Germany the low level jet

can be found approximately in 10 percent of all
nights. Hence, a lot of data has been collected in
the last years by means of tower and aircraft
measurement [3, 4, 5]. Fig. 3 shows a typical low w
W.m V„
level jet sample recorded during a windshear
measuring project by means of an Airbus A 300
Figure 4: Composition of the low level jet model
operated by Lufthansa. A wind maximum of the
1.8-fold value of geostrophic wind speed and the
For a large number of data records a comparison
4.5-fold value of reference speed (Href = 10 m)
was observed as well as a change of wind of measured data with modeled low level jet has
been carried out. The examples of tower data (fig.
direction of about 90 degrees in a vertical layer of
300 m height. Investigations in the plains of 5) and aircraft data (fig. 6) are in good agreement
northern Germany carried out on two with the model.
meteorological towers of 300 m height [5] show a
cycle period of 14,5 hours for the inertial
oscillation. During take-off and landing approach
the critical zone of windshear is passed in only
 28-3

Locality: Gartow
Date : 16.05.1979
Time : 05:45 MEZ
__!_.
i
i
\
I ....
f
.\
y
,.-■*
,'
r
}
i 1
15 17 19 21 23

Locality: Gartow
Date : 21.06.1979
Time : 06:30 MEZ
__ .__*_
>
Hi 280 "
1 l(

240 - hi<
M iAfeUFlEf* ENTf ^-MODEL
200 - I
\1 V
*
160 - x I
»--. }
120 - -- > - -
1
iv
•>
80

40
•ir*
>
<
s
JS . -'■'

0 19 20 21 22 23 24
0 12 3 4 5 6 140 150 160 170 180
Locality: Gartow
Date : 17.04.1980
Time : 06:10 MEZ

I
\
%
-4-

/
^.< '*'

r .
«*■""

1
! 2 3 4 5 6 7 8 1« 150 155160 165 170 175180 185 8 9 10 11 12 13 14 15

WIND ANGLE. DEGREES TEMPERATURE/ C

WIND SPEED/ M/S

Figure 5: Approximation of tower data

28-4

Worst case profiles for the low level jet have been -•-Khabarovsk, 23.1.1976 Model
m
approximated from Soviet measurements (fig. 7) Hz«J
/
i
no
[8]. WO
i
m
ao \
A typical feature of low level jets is the extremely
low degree of turbulence as a consequence of the
in
i» -
no.
sN s
^
180- \ i

stable atmospheric conditions. This fact is J \

100 -
/'
important for the simulation results, as pilots /
cannot identify the windshear simulation through
a correlated turbulence. This is in contrast to the y
s .'' f\i
» j « < i n n ii ii ii ID a -16 -15 -M
typical downburst, where high windshear
gradients, high downdrafts and extreme
turbulence are correlated. Figure 7: Approximation of a worst case profile

In all cases the maximum windshear gradient UWH

A»»«rt
0«t<
:»r«.tii
.»1.10. mo
(«) WTM
is greater than 0.2 s"1 according to the ICAO's
Co« :ol:ll W

+J S<Hl««««ii«(*):ta/< 1J0-1TI»
definition (table 1).
JC 500
soo

t, /
•r»
HI
m (00 ; \ intensity of influence on the variations of wind UWH
(00 windshear flight path speed within
300 AH = 30 m
\
S 3
/ 1
300
low neglectable 0-2m/s 0.0 -0.07
200
200 ; \ moderate significant 2-4m/s 0.07-0.13

y . Jt 100 strong heavy 4-6m/s 0.13-0.20

100 very strong dangerous >6m/s >0.20
/
0
5 io <n« is 150 wo no» 10 "C12

Table 1: Classification of vertical windshear

Airport : frtirtrvrt (•) KTM
Oal« : 11.11. IM» gradients (ICAO) [9].
1>n< :«f:W Off
RHUWOT : If
+J Sartot« wn< (»I: I.5Q.W11
.c
en Such a windshear may be defined as dangerous.
In contrast to this definition, in most flight
measurements no pilot comments were reported
and no significant deviation of flight path and
airspeed were obtained. Only the extremely low
level jet of Khabarovsk (fig. 7) caused a fatal
S mA 10 200 ■rr* accident.
I«) «1»
Based on these reported low level jet
0«1«
Tim«
: u.a.mo
: ■•:14 CUT
measurements a set of four windshear profiles
:li
Swl«»Wfi< («l:5./t t00-llt*
were generalized for simulator campaigns to
evaluate the pilot response in such windshears.

3. Simulators and Pilots

Two different types of simulators have been used

22Öb 2 t °C 6 for the tests:
Wind Speed Wind Direction Temperature
• Airbus A 340 full flight simulator,
designed and produced by CAE and operated by
Figure 6: Approximation of flight data the Berlin Center for Flight and Simulation. This
simulator can be operated in two modes:

a) as a certified training simulator

b) as a research simulator

• Fixed base simulator of a turboprop

powered aircraft. This simulator is based on an
 28-5

original Metro II cockpit including The driving force is the wind acceleration as
instrumentation. It was designed and is operated function of the windshear gradient dVw/x
by the Braunschweig Technical University,
Institute of Flight Guidance and Control. The
copilot's control panel was modified to take up a dVw dx _ dVw
(2) Vw = V,
programmable Electronic Flight Instrument dx dt dx
System (EFIS).
With an ideal flight control system, the airspeed
In total 8 airline pilots and 6 test pilots, all of
them very experienced, performed 86 evaluable has to be constant and therefore V = 0. This
approaches. results in

dV,w
4. Simulator Results (3) VK = Vw = V„
dx
A simulated approach through the Khabarovsk
low level windshear is shown in fig. 8. After In the ideal situation (V = const.) the aircraft has
passing the wind maximum in a height of 120 to be accelerated with the wind variation.
meters, the decreasing headwind moves the Equation (3) is a differential equation with its
aircraft below the glide path. The pilot initiated a inherent dynamical characters.
go around manoeuvre and could survive, but the
height above ground was of only a few meters. A The required thrust of an aircraft normalized with
small building or a tree would have caused a fatal its weight W is
accident (approximately 1200 meters before the
nominal touch down point). The small airspeed , wg . U

variation is remarkable. The approaches of 14 (4) A— = VK + A/+^l cos/ +—-sin?'

W V
pilots were very similar.
If the flight path is maintained constant the flight
path angle deviation Ay has to be zero. With a
neglecting vertical wind component and small
flight path angles wWg we get

(5) A-^ = VK + ^L
Y
W V
4000 3000 2000 101
distance to threshold [ m ]

If the airspeed is constant (V = 0) we obtain the

thrust control law by introducing eq. (3) into eq.
Figure 8: Approach through a low level jet (5):

(6) A—= -^VKK

The ideal thrust control law of eq. (6) has been W dx
used as a reference system in the simulations and
it will be implemented in the Institute's research If we multiply eq. (6) by the speed VK we get the
aircraft DO 128. The principle of throttle control specific excess power:
is based on fundamental flight mechanical
equations [12]. As the flight path velocity (inertial
velocity) VK is the sum of airspeed V and wind (7) SEP = HE = ^Z-VlK
speed Vw, the respective accelerations (time dx
derivatives) are
The energy variation expressed as energy height
(1) VK =V+VV (or total energy) is
28-6

v^ down since the landing jerk is a typical indication

(8) HE={HEdt = -^ + H for motion. A surprising result was that pilots
2g could not correctly identify if the motion system
was switched on or not (fig. 9) [13]. The time
If we apply the thrust control law (eq. 6) then we delay was not influenced by motion. This minor
get the thrust required when passing through the effect of motion gives an indication that the
low level jet of fig. 8. With this ideal thrust results obtained with the fixed base simulator
control the aircraft passes the windshear without have the same validity as those obtained with the
any deviation of speed and flight path. The more sophisticated motion simulator.
required thrust is within the limits of the engines.
The commanded thrust rate is fairly low and no 37 5%
modarat«
problem to be realized. Even in a downburst, inlluanca

thrust and thrust rate lie comfortably within the

engine limits [12].

A comparison between the ideal thrust and the

pilots thrust settings demonstrate that both 12.5%
responses are similar in principle (fig. 8). The was nacassary for a
realistic approach

pilots thrust setting is typically digitized. This is

quite normal since experienced pilots do not set
the throttle continually but in steps. A surprising
result is the delay of the pilot thrust response Figure 9: Pilot judgement on the influence of the
compared to the ideal thrust response by a "dead simulator's motion system (motion off)
time" of about 30 seconds depending on the kind
of low level jet. This includes a response delay of 5. Thrust Command Director
approximately 2 seconds of the engine. The
extreme time delay was nearly the same for all 14 The enormous time delay of the involved pilots
pilots involved in this simulator campaign for the could be gained in both flight simulators for wide
fixed base simulator with conventional body aircraft as well as for turboprop aircraft.
instrumentation. Variation of different parameters The scatter of the time delays between the
demonstrated primarily the effect of the height of different pilots was small. The effect could be
wind maximum in the low level jet (fig. 8). A quantified but we are still far away to give a
wind maximum in H = 340 m resulted in a delay proven explanation for the pilot's behaviour. It
time of 40 sec. A less significant parameter is the may be assumed, that the human being has not
magnitude of the wind maximum itself. only limitations in the higher frequency range, but
also in the lower frequency domain. The human
Similar results were gained in the Airbus A 340 being has more or less the frequency response of a
flight simulator in Berlin. When discussing the band pass. The higher frequency domain is well
results one should keep in mind that the Airbus A examined as handling qualities. Do we have to
340 simulator had an airspeed trend indicator expand the handling quality domain to lower
included. The basic equation for this speed trend frequencies? The difficulty to identify
bar is accelerations of longer duration e.g. in a lift is
well known as "adaption" in the medical
dVmeasured community. Nevertheless, the question can not be
(9) Vindicated = Vmeasured +• answered in this paper.
dt

The advance time constant T being equal to 10 To keep the pilot in the control loop the flight
director is a well proven and successful device.
seconds.
The function of the pilot is more or less that of an
As the response in a low level jet is rather slow actuator. The pilot has to keep the needles of the
and turbulence is not present the question arose flight director close to zero. The needles itself are
whether the motion cue of the simulator would controlled by the flight control system. With such
affect the results. Approaches were realized with a flight director an aeroplane can be controlled
and without motion. During these trials the precisely and safely. The pilot can supervise the
motion system was always switched off at touch control system easily.
 28-7

We have pointed out that the throttle control will

follow the thrust command formula (eq. 6) to red
keep airspeed and flight path constant even in commanded thrust
yello
severe windshear simulation. Can we persuade the
pilot to response in the same manner?
L / Ü green

Pilots have learned to set the throttle only if n"n green

necessary. High throttle activity is typical for a
yellow
flying student. What we need is the indication of
actual thrust
the actual and the commanded thrust. The actual red
thrust is represented in the vertical speed
indicator. If we add a second pointer to represent
the commanded thrust, all necessary information a) b)
for the pilot is available. A previous simulator
Figure 10:
investigation [16] with the DLR motion based
a) Thrust and thrust command indicator
simulator (out of service since several years) was
b) Energy height indicator with alert level
only partly successful. Only ten of fourteen pilots
dependent colouring
involved in the simulator programm "survived"
the well known downburst that caused a crash in
New York on June 24, 1975 (fig.l). The main
problem was the typical position of the vertical
speed indicator. Measurements of the eye
movement indicated that this instrument was
scanned only every 7 seconds. This scanning rate
was probably not high enough. The investigations
were stopped with the result that an indicator
located in the center is necessary to present the
relevant information for the pilot. A solution for
this is the electronic flight instrument system
(EFIS) that can be easily reprogrammed. An
easily programmable symbol generator has been
especially developed at the Insitute of Flight
Guidance and Control to represent digital maps
on a commercial EFIS display [14]. With such a
Figure 11: Modified primary flight display (PFD)
display thrust and thrust command can be located
in the central panel area (fig. 10, 11) [13].
Additionally, the pilot needs an indication of the
With the thrust command director human pilots
energy situation to distinguish if throttle response
are able to fly the aircraft with similar results as
is required or not. A column indicator (fig. 10) for
those obtained by automatic flight control
the energy height deviation from the trimmed
systems. The total delay time is reduced down to
prosition changes its colour depending on the
engine delay time (fig. 12). Additionally, the
level of alert (green, yellow or red). An energy
results with conventional instrumentation as well
height deviation of more than 10 meters is
as with speed trend indication in the Primary
dangerous [10, 15]. With such a situation
Flight Display (PFD) are demonstrated. The
indicator the pilot provides a non linear control.
relevant flight and thrust variations are plotted in
In the green range no input is needed and the
fig. 13. The relative risk as well as the energy
yellow range already indicates that throttle
height deviation are plotted in fig. 14 for the
activities are required. The integration of these
different type of indicators. Table 2 demonstrates
symbols is shown in fig. 11.
the percentage of accidents and go around
decisions obtained with different indicator
concepts.
28-8

25
Wind speed
,
20
(Headwind)
[m/s] i [%]
[Isll integral of energy
10
height error
5
^gj integral of
0 relative danger
-6000 -5000 -4000 -3000 -2000 -1000

F/G 0.25
(Pilot) [ | Conv. Instr.
0.2 "1 Speed Trend Ind.\
A F/G [.Thrust Cmd. Dir. \ \
(Ideal Thrust
0.15
^ \
Cmd.) [-] 0.1 ^ \
(Ideal Thrust^
>*- \
0.05
\Command J ^V-A »1

V.'.'.'.'.'.'.i
0

-0.05
-6000 -5000 -4000 -3000 -2000 -1000 conv. PFD with PFD with
instr. Speed Trend Thrust Cmd. Dir.
Height ^400 J 1 1 1
[m] ' Thrust Command Director.
350 Conventional Instrumentation
300 Speed Trend Indicator \ \ Figure 14: Energy height error and relative risk
^s. IdealThrust \ \ \
250 Command \ \ \ during the approaches related to
200 conventional instrumentation
150

100 \v
\ \\
A
50
6. References
0
-6000 -2000 -1000

Distance to Threshold [m]

1. Swolinsky, M.: Wind Shear Models for
Aircraft Hazard Investigations, AGARD-CP
Figure 12: Approaches through 'worst case* low 470, 1989
2. Blackadar, A.K.: Boundary layer wind
level jets with different indicator
maxima and their significance for the growth
concepts
of nocturnal inversions, Bull. Am. Meteorol.
H (Vw max) [ m ] Soc. 83, pp. 283 - 290, 1957
3. Swolinsky, M.; Krauspe, P.: Windbestimmung
aus Flugmeßdaten eines Linienflugzeugs,
Meteorologische Rundschau 37, pp. 72 - 81,
June 1984
4. Swolinsky, M.; Krauspe, P.; Vörsmann, P.:
Wind Determination and Wind Shear
Detection from Flight Test and Airline Flight
-0- PFD with Thrust Cmd. Director Data, International Conference on Aviation
-B- PFD with Speed Trend Indicator
-Ö- Conventional Instrumentation
Weather Systems, American Meteorological
10 20
Society, Montreal, May 1981
total delay time [ s ] 5. Kottmeier, C: Die Vertikalstruktur Nächt-
licher Grenzschichstrahlströme, Dissertation at
the Institute for Meteorology and Climatology
Figure 13: Comparison of the total delay time
obtained with different indicator of the University of Hannover, 1982
6. Reichardt, K.: Gesetzesmäßigkeiten der freien
concepts
Turbulenz, VDI-Forschungsheft 414 (1942),
2nd edition 1951
7. Schlichting, H: Grenzschicht-Theorie, Verlag
conv. instr. speed trend EMSIS automatic G Braun, Karlsruhe, 1982
accidents 32% 3% 0% 0% 8. Shelkovnikow, M.S.: Strujnye techeniya na
go-arounds 28% 21% 7% 0% malykh vysotakh (Low level jets),
Meteorologiya i gidrologiya, 1983, No. 11, pp.
Table 2: Accidents and go-around decisions for 44-46
the different indicator concepts (EMSIS: 9. Badner, J.: Low-Level Wind Shear: A Critical
Energy Management, Surveillance and Review, NOAA Tech. Memorandum NWS
Information System) FCST - 23, National Oceanic and
 28-9

Atmospheric Administration, Silver Spring,

Md., 1979
lO.Schänzer, G.: Influence of Windshear on
Flight Safety, AGARD-CP 347, 1983
ll.Redeker, A.: Beiträge zur Verbesserung der
Führungsgenauigkeit von Flugreglern,
Dissertation Technical University of
Braunschweig, 1986
12.Schänzer, G: Influence of Windshear,
Downdraft and Turbulence on Flight Safety,
AGARD-CP 470, 1989
13.Krüger, J.: Untersuchungen zum Piloten-
verhalten beim manuell gestuerten Einflug in
Scherwindsituationen während des Lande-
anfluges, DGLR-Jahrbuch, 1994
14.Bavendiek, K.: Ein Beitrag zur Realisierung
von Echtzeit-Graphik in Fahrzeugen mit
hochdynamischer Bewegung, Dissertation
Technical University of Braunschweig, 1991
15.Swolinsky, M.: Beiträge zur Modellierung von
Scherwind für Gefahrdungsuntersuchungen,
Dissertation Technical University of
Braunschweig, 1986
16.Alles, W.; Krauspe, P. et. al.: Untersuchungen
technischer Möglichkeiten zur Erhöhung der
Sicherheit bei Scherwindanflügen, report no.
TB Nr. 1257/82 of the Federal Ministry of
Transport (BMV), March 1982
 29-1

ASRA - A New Tool for In-flight Simulation

Current Status and Actuation Studies
J.Murray Morgan and Stewart W Baillie
Flight Research Laboratory
Institute for Aerospace Research
National Research Council of Canada
Ottawa, Ontario, Kl A 0R6
Canada

1. ABSTRACT decision to acquire this model came only after a short, but
intensive, flight test study conducted at the Bell facility at
The Flight Research Laboratory (FRL) of the Institute for Mirabel in Canada [2]. With the cooperation of Bell
Aerospace Research (IAR) is in the process of developing an Helicopters/TEXTRON, the company demonstration aircraft
Advanced Systems Research Aircraft (ASRA) based on a Bell was instrumented by the FRL . The instrumentation package
412HP helicopter. This paper describes the current status of was the FRL Micropack [3] , a high quality instrumentation
this facility, and discusses the approach being taken towards and data recording system developed at the laboratory and
fly-by-wire actuation in this high control power host vehicle. used extensively for parameter identification studies,
The primary problem faced in designing a research fly-by-wire augmented by specific pilot control input sensors. In this
(FBW) actuation system for this aircraft is that of maintaining configuration the aircraft was flown through a complete set of
an adequate level of flight safety throughout the entire flight ADS-33C Part 4 manoeuvres and subjected to specific
envelope, while limiting its simulation capabilities as little as parameter identification control sequences (modified 3,2,1,1
possible, this is discussed. A scheme for compound actuation inputs and manually flown frequency sweeps).
of the 'critical' axes is introduced. Initial off-line simulations
of various failure modes are described and the results This exercise confirmed the suitability of the Bell 412 for its
presented. The purpose of this paper is to present to the intended research role and the aircraft was acquired in April
community a proposed approach to this problem and to 1993.
describe some initial studies in support of the design concept
used. 3 . THE DEVELOPMENT PROCESS

2. BACKGROUND 3.1 Overview

In common with many similar organizations internationally,
The FRL had operated variable stability helicopters in the the National Research Council has been subject of severe
airborne simulation and systems development modes for budget restraint for the past several years and there is no
nearly three decades. The current working aircraft, the indication that this will change in the foreseeable future.
Airborne Simulator, based on a Bell 205 A-l is, in fact the Therefore, the approach to development of ASRA is perforce
third generation of such machines developed at the laboratory. one of slow stages towards the final goal of having a fully fly-
The FRL approach to fly-by-wire for research has always been by-wire general purpose flight mechanics research vehicle. At
a generalized one, with the aircraft not married to a specific the same time, it was felt desirable to have the machine
control system architecture. The Airborne Simulator produce useful research work before that final goal was met.
comprises the host aircraft, a set of dual mode, full authority, This has led to a staged development process, in which the
high bandwidth actuators, a general purpose computing aircraft is to be instrumented and provided with in interim
system a set of state and pilot input sensors and a variety of data processing and recording system with the ability to
pilot displays. There is no embedded software and no present the pilots with project specific instrument displays
dedicated fly-by-wire control system as such. This has left the before the application of fly-by-wire actuation and the final
laboratory free to adopt control system structures appropriate flight control computing system. It was also determined that,
to the experiment in hand: state feedback, control feed to exploit its potential agility to the fullest, it would be
forward or model following techniques are all used as the necessary to install a real-time loads measuring system on the
demands of specific projects dictate. aircraft and that this installation had to precede any further
parameter identification studies.
Some years ago, particularly when developing a data base for
use in drafting ADS-33C [1], it was realized that the Bell 205, 3.2 Current Status
with its two bladed teetering rotor system, had neither the At this time, sufficient instrumentation has been added to the
control power nor the agility to investigate those areas of aircraft to permit useful research, particularly in the area of
increasing importance to the helicopter community. It was civil helicopter IFR flight. The present fit comprises:
therefore decided to replace the Airborne Simulator with a • a project power distribution system
new research aircraft having a rotor technology more • an engineering workstation
representative of today's production models. After a study of • a Litton LTN 92 Inertial Reference System with Integrated
several available helicopters, with either fully articulated or GPS
hingeless, bearingless rotor systems, it was decided to • a Trimble GPS receiver with a UHF data link for real time
purchase a Bell 412HP to be the new host vehicle. The final differential GPS positioning

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?'
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
29-2

• static pressure, dynamic pressure and total temperature already in house. Consideration of a conventional control
sensors loading system to simulate conventional controls will be
• pilot control input sensors addressed at a later time. The flight control computer will be a
• provisions to drive the standard electromechanical 68040 processor based VME system operating under OS -9.
instruments from the workstation. This will provide compatibility with the on board engineering
• a DAT based data recording system, controlled by the workstation and with identical hardware being installed in
workstation other laboratory aircraft. Actuators have been selected and
ordered, but the actuation scheme has yet to be finalized and
As a research system, the aircraft installations are is, in fact the subject of current study. The remainder of this
supplemented by a compact, self contained GPS ground paper will be devoted to discussion of the actuation system.
station which produces the real time corrections sent to the Although the actuation scheme has yet to be finalized and is,
aircraft via the UHF data link and a VAX workstation for data in fact the subject of current study, actuators have been
transcription and analysis. selected and ordered.. The remainder of this paper will be
devoted to discussion of the proposed actuation system.
3.3 Planned Installations - Near Term
Two significant installations are planned for the near term. 4.0 FLY-BY-WIRE ACTUATION IN ASRA

Speech Recognition System. As part of the private sector 4.1 General Considerations
support for ASRA, the Canadian Marconi Company (CMC) "Hardover" Incidents in the Bell 205. The Bell 205 Airborne
has donated a CMA 2082 CDU for installation in ASRA. The Simulator has always been driven by full authority high
specific unit designated for this will contain a speech bandwidth actuators which, in a single channel system have
recognition capability, which represents the product of always left the aircraft potentially vulnerable to a full
research conducted at this laboratory by a cooperative group deflection hardover in the FBW mode. The FRL has relied on
comprising CMC, the NRC and the Neil Squire Foundation hardware health monitoring and software safety related
with the support of DND (CRAD). In addition, a complete modules to assist the safety pilot in preventing catastrophic
suite of bus controllable avionics has been ordered to replace results from such an occurrence. In practice, very few/«//
the standard items presently in the aircraft. This will enable hardovers have ever been noted and none were permitted by
the conduct of specific and realistic research into the use of the safety pilot to progress to a point at which the aircraft was
Direct Voice Input (DVI) in the cockpit. at serious risk. Only one hardover has ever been due to failure
of an actuator or its associated electronics, one was
Stereoscopic Flat Panel Display. In cooperation with Litton attributable to a sensor failure and one due to a switching
Systems Inc, the laboratory is shortly to install a unique logic error in the early days of the aircraft's operational life,
stereoscopic head-down display for generic display research. when the flight control computations were predominantly
The unit is an active Liquid Crystal Display, capable of mono analogue. This implies that the majority of hardover or
or stereo graphics without encumbering the pilot with any unplanned step displacements of the controls experienced in
additional equipment. The unit is currently running in a the Airborne Simulator were commanded by the flight control
ground based display development facility and has been used computing system. Such commanded occurrences have
to develop a range of three dimensional "highway in the sky" become extremely rare since flight control computer
symbology sets for steep decelerating IMC approaches to a programming was switched, first to floating point assembler
helipad. In the aircraft these will be used to assess pilot and then to a high level language from assembler with integer
performance while using and acceptance of displays of this arithmetic.
nature relative to more conventional displays and flight
directors in the same task. Safety Factors in the Bell 205. Although it would be difficult
to prove academically, the FRL is convinced beyond doubt
3.4 Fly-by-Wire System that the ability of the safety pilot to ensure the aircraft's
Converting the ASRA to a fly-by-wire aircraft involves several integrity under failure conditions is due to three main factors.
distinct tasks: Firstly, all actuator commands generated by the FBW system
• Selecting, constructing or purchasing suitable pilot control are reflected back to the safety pilot's controls. This implies
inceptors and mounting these at one of the pilot stations. that all actuation must be in parallel with the safety pilot's
• Selecting, mounting and integrating a suitable flight controls. Secondly, the safety pilot remains always hands on
and is provided with an instinctive FBW disable/disengage
control computing system.
control in the form of a cyclic paddle switch, mounted so that
• Determining a suitable scheme for actuation.
it can be activated quickly and instinctively by one of the two
• Acquiring suitable electrically signaled actuators
fingers which are always kept resting upon it. The safety pilot
• Designing the physical installation of the selected
quickly learns what is the norm for any given control system
actuators. and tends to react very quickly when the perceived pattern
• Performing the physical installation changes for no clear reason. Finally, the 205 has a
comprehensive health monitoring system that continually
The constraints placed on these activities by current fiscal examines the states of electrical power supplies and hydraulic
restraint in Canadian Government funding have had a major pressure at the actuators and disengages the FBW system the
influence on this entire process. Not only are funds limited, moment a fault is detected. Based on these observations made
but the scheduling of available moneys is also a factor. over the 22 years of the Airborne Simulator's service life, the
Because of this some decisions have been heavily influenced FRL has decided that ASRA should share the same essential
by the availability of in-house equipment. Pilot inceptors will,
at first, be one or other of the several side-stick controllers
 29-3

characteristics, even if they are achieved by different technology, led us towards a dual actuated approach to
techniques. motivating the pitch and roll channels. In this approach, one
Considerations for ASRA. Despite the excellent record of actuator is permitted full authority, but at a limited rate of
safety under abnormal conditions experienced in the 205, it is extension, while a second is permitted high bandwidth, but
considered that an identical system would provide inadequate limited authority. The output from the rate limited unit is
safety margins were it to be fitted to the Bell 412. The Bell differenced with the command signal and this "error" is used
Soft In Plane rotor system of the 412 displays appreciably to drive the high bandwidth device. This scheme is shown in
higher control power and considerably less response delay Figure 1
Flow Limrter
than is seen in the 205 (Approximately 72 ms for the Bell 412 LF Actuator

compared to 180 ms in the Bell 205). It has been felt from the Command
G(s)
first that a method must be found to limit the aircraft's
response to hardover in the actuation hardware itself (in +^ Swashplate
addition to an essentially similar system to that in the 205).
Several options were examined, including: HF Actuator Attenuator

Vw\ G(s) K
• shaped pilot valve porting to tailor the
frequency/amplitude response of a single actuator
• the use of multiple actuators per channel and Figure 1 Compound Actuator Schematic
• the use of software monitoring of the actuators to inhibit This scheme provides a system in which the high bandwidth
fast, large amplitude responses. actuator moves only when the full authority actuator reaches
its rate limit and there will be a linear relationship between
The first option, the more elegant and simpler to install, had command and aircraft control run displacement until such
to be abandoned because funding limitations would not permit time as the rate limited actuator and the high bandwidth
our paying for the non recurrent engineering which would be actuator are simultaneously limited, hi practice, the rate limit
involved in the design and development of such actuators. The and the contribution to final output of the fast actuator will be
third option implicitly rate limits the actuators. This was felt set physically by use of a hydraulic flow limiter and a
to be undesirable because of the high probability of distorting mechanical summing linkage between the actuators.
the simulation responses of the aircraft. It was decided to
adopt dual actuators (or rather, two actuators operating as a 4.4 Basic Actuator Performance
single 'compound actuator') for the critical axes (pitch and Given an actuator configuration of this nature, it is necessary
roll) and to use a certified flightworthy design and to employ to accept that it will impose some limitations on the FBW
single actuators of the same type in the yaw and heave axes. envelope of the host vehicle due to its inability to respond in a
linear manner to all driving signals. The actual device chosen
4.2 The Actuation Scheme. for use in ASRA is the HR Textron Part No 41000360, a three
Again, the FRL experience with the Bell 205 has helped to inch throw linear electrically signaled actuator with built in
ease the decision making process for ASRA actuation. The fail-safe (free output rod) mode. This unit is designed to an
close similarity of the two helicopters enabled us to decide airworthiness critical specification and has been employed on
that the collective and yaw channels could be safely serviced other fly-by-wire vehicles, both fixed wing and helicopter
by a single actuator each and to limit compound actuation to (notably the Sikorski Shadow). The manufacturer gives a
pitch and roll, these being deemed the critical axes for failure small signal model for this actuator in servo form as shown in
at very low altitude. Figure 2.
Yaw channel hardover failures in the Bell 205 have proved to + 10 ma + 1.5 in
be startling but never to endanger the aircraft significantly and
the tail rotor of the Bell 412 is identical to that of the 205.

A full collective DOWN failure at very low level could cause -10 vote

the aircraft to strike the ground heavily, but probably not in

such a way as to put the crew at risk. Such a failure is also, to
an extent self limiting. If the aircraft is VERY close to the LVDT
ground, then there will be insufficient time for it to reach a
high vertical velocity, while as the height becomes greater,
there is more time for the safety pilot to react to ameliorate Figure 2 Servo Block Diagram Model
the situation.
This block diagram was implemented in a non real time
Pitch and roll disturbances, in contrast to those in yaw and simulation and subjected to frequency domain analysis, the
results being shown in Figure 3. In this figure, the amplitude
heave, can rapidly become a danger to the aircraft when very
close to the ground, especially during manoeuvring flight. ratio plotted has the units of in/volt and the driving signal was
This led us to search for a method which would enable us to a frequency sweep of 0.01 to 40 rad/sec in 120 seconds, with
achieve limited aircraft upset in the event of an actuator an amplitude of 3.33 volts (33% of maximum).
malfunction or, the more real case, an inadvertent very large
actuator step demand. The success of some models used in 4.5 Compound Actuator Performance
The structure in Figure 1 was modeled, with the actuator
recent experiments in the area of limited authority attitude
blocks G(s) being replaced with the structure shown in Figure
SCAS systems [4], themselves borrowed from old autopilot
29-4

2. The flow limiter values and the gain K were varied to 5.2 Aircraft Modeling
examine their effects on the frequency response of the For the purpose of this investigation, a simple second order
compound actuator. A sample of the results is shown in Figure transfer function model of the primary lateral control response
3 (See section 5.2 for Case definitions). This figure clearly characteristics of the unaugmented Bell 412 at the hover was
shows that the lower the authority of the high bandwidth constructed. This model was built empirically to match data
actuator, the greater the performance loss in the compound taken by the laboratory on the actual host airframe (Reference
actuator. However, .the figure also shows that it is quite 2). To obtain a good match between the phase roll off of the
possible to build a compound actuator which approaches aircraft and model required a time delay of some 72 ms,
closely to the performance available from a single however, since the form of the phase curve may be related to
other high order effects, the time delay was subsequently
Amp Hallo (db) In/nil removed to ensure a 'worst case' low frequency response from
the model when subjected to step inputs. The model was
^ ; r : : subjected to excitation from four separate sources; these were
the outputs from actuator models which themselves were
V-f j-ftiTil subjected to 100% step input demands, the actuator models
were
1 1 ! i ! Psi? !
• A basic single actuator as described in Figure 1
• The output of the compound actuator of Figure 2
-*- Bwld AccluMor -*-"CM» i*xi -& C»M iss» ! ! ! ::: • The individual outputs from the component values of the
i ; ; i ;;:; ; :::,;;, i i _J i—■—■—i i compound actuator, that is, the rate limited output of the
1 10 LF actuator and K times the output of the HF actuator.
FfMUMoy (rad/Mo)

Figure 3 Actuator Matching Performance A series of test cases was run using this set-up, making the
rate limit and gain K experimental variables. For convenience,
unit, while maintaining a significant measure of protection to
the configurations used have been assigned case identifiers for
the crew. It should be noted, too that these simulations were
future reference: the case identifier is a four figure group, the
conducted using a relatively large amplitude driving signal,
first two being the rate limit on the LF actuator in percent per
covering a range in which the vast majority of routine, even
second, the second two the extent limit on the HF actuator in
aggressive pilot inputs fall. However, it is not sufficient to
percent, eg. case 1520 would define a compound actuator with
show that the actuation scheme can be built to provide
a rate limit of 15%/sec and a limit on the HF actuator of 20%
adequate performance for the simulation task without
examining the response of the aircraft to hardover failures of total throw.
with the various combinations of rate limit and authority
For each case, results were obtained for three major
proposed for this system.
conditions. Firstly, the aircraft was allowed to respond open
loop to the actuator signals; this was to obtain a feel for the
5. AIRCRAFT RESPONSES
kinds of disturbance possible in the first few seconds after a
failure or abnormal command. Secondly, two forms of pilot
5.1 Swashplate Geometry and Worst Case Failure.
response were simulated, the first based on pilot intervention
It is intuitive to accept that a worst case failure scenario for
after a certain recognition time, the second triggered by the
the low level helicopter is that which would produce the
aircraft reaching a given attitude change from trim. Both these
largest attitude disturbance in the least time. The geometry of
may be valid trigger mechanisms to induce safety pilot
the swashplate/aircraft actuators in the Bell 412 is such that
intervention. The first implies the case of a step input being
the simultaneous and controlled operation of two actuators is
fed back to the safety pilot's cyclic stick; rapid intervention is
necessary for a clean pitch or roll response from the aircraft.
assured (this remark is based on the author's personal
Both cyclic actuators contribute to both the pitch and roll
experience of having flown as safety pilot in the Airborne
commands to the main rotor that is, if both actuators move in
Simulator for over 15 years). The second case is similar to the
unison, the swashplate rotates in pitch, while if they move in
kind of response engendered in the presence of a more subtle
opposition, the swashplate rolls. It follows that if a single
failure, such as the runaway of a rate limited actuator, hi both
actuator runs away due to a fault downstream of the drive
cases the pilot model was a simple low pass filter at 5 rad/sec
signal, the aircraft will suffer a two axis disturbance, but that
plus a 300 ms delay which returned the control signal to zero.
the swashplate will deflect, in each axis, only about one half
This was intended to represent a rather abrupt resumption of
of its total capability in that axis. At the same time, if the
control and initial recovery by a safety pilot pre conditioned to
FBW system is currently active, any disturbance caused by
anticipate a failure. The 300 ms is intended to represent the
failure of a single actuator, unknown to the active control
rninimum neuromuscular delay expected of an alert safety
system, then the error condition will be automatically opposed
pilot and the return to zero a rather conservative response,
by the remaining actuator, softening the total effect of the
since experience shows that a safety pilot usually applies a
failure. This , together with the observations made in
positive control spike against the initial disturbance before
paragraph 4.1 lead us to believe that the 'commanded'
entering a closed loop stabilization mode of control.
hardover, where both actuators receive a maximum drive
command from the FBW computer due to programming or
sensor error, represents the worst case failure mode.
Consequently, investigations of potential aircraft disturbances
have been concentrated on this concept.
 29-5

6. SIMULATION RESULTS close to the ground. To assist in making this decision, the two
'pilot intervention' cases were examined.
6.1 Open Loop Responses.
Figure 4 shows typical examples of the open loop response of Table 2 Disturbance Characteristics for Failure of LF
the aircraft to various hardover failures. On this are identified Actuator Only Without Pilot Intervention
the traces of how the aircraft would respond to each of the Time to 45 Max Attitude Time Max
fault conditions specified in 5.2 above. The advantage to the (sec) (deg) (sec)
safety pilot of compound actuation is clear from this diagram, 10 3.54 75.11 6.25
even a commanded hard-over gives the pilot a significant 15 2.68 81.11 4.85
increase in available reaction time before excessive attitude 20 2.25 84.07 4.26
disturbances occur. 25 1.98 85.61 3.94
Bank Angb <d»g>

Table 3 Disturbance Characteristics for Failure of

Both Sections of the Compound Actuator Without
Pilot Intervention
Case ID Time to 45 Max Attitude Time Max
(sec) (deg) (sec)
1010 3.03 75.28 5.73
Tim» From Fair- 1015 2.35 81.39 4.51
Case Figure 4a 1010 Response, No pilot Intervention 1020 2.01 84.30 4.01
1025 1.79 85.79 3.74
Bank Anak* (dag)
1510 2.54 75.96 5.17
1515 2.05 82.15 4.17
1520 1.79 84.86 3.77
1525 1.61 86.20 3.55
2010 2.12 77.49 4.61
2015 1.79 83.36 3.86
2020 1.60 85.67 3.55
2025 1.46 86.77 3.39
L* t 2510 1.80 79.91 4.10
Tim* From F«ikr 2515 1.58 84.83 3.59
2520 1.44 86.58 3.37
Figure 4b Case 2525 Response, No Pilot Intervention 2525 1.33 87.39 3.25

The disturbances for the 16 cases studied are summarized in

Tables 1 to 3. Table 1 shows the maximum attitude reached at 6.2 Pilot Intervention Cases
first peak when only the HF actuator fails, Table 2 contains Pilot Intervention After Time Delay. The pilot recovery action
three items per LF actuator rate limit, the time to reach 45 described in 5.2 was modeled with time delays to failure
degrees of attitude, the peak attitude reached and the time to recognition of 0.25,0.5,0.75,1.0,1.25 and 1.5 seconds. Only
reach peak attitude. Finally, Table 3 summarized in the same the full compound actuator hardover was considered, and the
way the results obtained for runaways of the complete results have been characterized as the peak attitude reached
compound actuator, using the case identifiers described earlier after failure and subsequent pilot intervention in Table 4. M
to characterize the compound actuator characteristics. this table, cases which resulted in an attitude excursion of
over 20 degrees have been highlighted. A set of typical time
Table 1 Maximum Attitude Excursion for Step
histories for this model are shown in Figure 5
Failure of HF Actuator Without Pilot Intervention
HF Limit (K) Max Attitude Time to Max Buk Angla (dag)

(deg) Attitude
0.10 8.86 2.83
0.15 17.73 2.86
0.20 26.59 2.88
0.25 35.44 2.90

These data give considerable confidence that the basic

compound actuator strategy is a viable one for maintaining an
acceptable balance between fly-by-wire performance and crew o.» a« 07« 1 IM

safety, especially as a simple actuator of the same type Tim» From Failure (sec)

indicated 1.15 seconds to 45 degrees of bank with a maximum

attitude of 88.45 degrees in 3.05 seconds. Selection of a set of
compound actuator characteristics now depends on a judgment Figure 5 Case 2525 with Pilot Intervention After 0.5
of how great a disturbance can be tolerated when operating seconds Delay
29-6

Pilot Intervention Triggered by Attitude. These cases were 7. DISCUSSION OF RESULTS

examined because experiences with the Bell 205 have shown
that subtle, or insidious failures are possible which lead to a 7.1 Permissible Attitude Excursions. The results presented
slow attitude divergence without sufficient feedback to the in the previous section offer a rough guide to what might be
safety pilot's cyclic controller to trigger intervention. Such expected in the way of attitude excursions with a nominal
conditions have on occasions even been long and slow enough safety pilot reaction to abnormal control activity or
to permit verbal discussion between the pilots, typically the uncommanded aircraft motion. What remains is to select a
safety pilot saving "Are you doing this ?" to the evaluation maximum attitude excursion which may be considered
pilot before a failure condition was confirmed. Depending on acceptable and from there determine the physical
the rate limit finally chosen for the ASRA LF Actuators, characteristics to build into the compound actuators. In
considering acceptable excursions, thought must be given to
Table 4 Peak Attitudes Following Compound the types of manoeuvres expected of the aircraft in the FBW
Actuator Failure and Pilot Intervention After a mode and when close to the surface and the role of the safety
Recognition Delay pilot in preventing accidents due to system malfunctions. In
Delay Delay particular, it is necessary to consider the normal manoeuvring
Case Delay Delay Delay Delay
attitudes achieved and what would be the final attitude of the
ID 0.25 0.50 0.75 1.00 1.25 1.50
aircraft if the most adverse failure were to occur when at that
1010 5.92 9.02 12.51 16.33 20.41 26.67
attitude. For example, if the aircraft was at 45 degrees of bank
1015 7.57 11.20 15.19 21.33 27.09 33.18
in a lateral acceleration at, say 10 feet AGL, would a potential
1020 9.16 13.34 17.83 26.34 33.77 41.69
excursion of 20 degrees (final attitude therefore 65 degrees of
1025 10.70 15.43 20.43 \ 31.34 40.45 50.21
bank) be acceptable ?
1510 7.17 11.27 16.03 19.45 23.91 28.47
1515 8.81 13.46 18.71 24.45 30.57 36.97
7.2 The Role of the Safety Pilot
1520 10.40 15.59 21.34 29.45 37.25 45.48
. Note: The following remarks are based on the personal
1525 11.94 17.68 23.94 34.45 43.93 53.56
experiences of the author, who has flown as safety pilot in the
2010 8.41 13.53 19.55 22.54 27.38 32.26
205 for the past 15 years and his observations of three other
2015 10.06 15.71 22.22 27.53 34.04 30.75
40.71 49.25 pilots who have shared this duty in the same period
2020 11.65 17.85 : 24:68 32.53
2025 12.89 20.11 28.73 37.53 47.38 56.63
15.79 23.07 25.60 30.82 36.03 Table 4 indicates that,, for commanded hardover failures, the
2510 9.66
11.30 17.97 25.74 30.58 37.48 44.51 attitude excursions is critically dependent on the safety pilot's
2515
13.18 19.94 27.45 35.57 44.14 53.00 delay in recognizing the failure condition. Experience at the
2520
22.19 30.97 40.47 50.63 59.45 FRL, unfortunately supportable only anecdotally, is that when
2525 14.43
in a potentially high risk situation, the safety pilot is so
"spring loaded" for failures that recognition time is very short
failure cases similar to these may be anticipated. Again, the indeed. The few failures documented in the past show a total
same conditions were applied to the simulated pilot response, reaction time under these circumstances in the order of 350 to
that is, after the response trigger, a neuromuscular delay of 500 milliseconds, this including both recognition and reaction.
300 ms was followed by a very simple low pass return of the There have been several occasions recorded wherein the
control to neutral. Since this type of failure is much more safety pilot beat the hardware safety system in disconnecting
likely to be related to a hardware (particularly analogue the fly-by-wire system. There are also occasions when the
electronics) failure than a software or computer hardware
safety pilot causes inadvertent or nuisance disconnects
failure, a runaway of the LF actuator was considered to be the
because of his readiness to react very quickly - these are most
most representative case and possibly the worst case, since prevalent when new control systems are in use and he is,
only a subtle ramp signal would be felt at the safety pilot's
therefore, unfamiliar with the normal spectrum of control
cyclic, not the step plus ramp which would be the case if the
activity they produce or when he has an unknown 'high gain'
full compound actuator failed. The results for 10, 15 and 20 pilot in the evaluation position. It is therefore with confidence
degree attitude triggers are shown in Table 5 that we can claim that a human safety pilot, experienced and
well trained, has proved a successful final link in the safety of
Table 5 - Peak Attitudes Attained and Elapsed Time flight during fly-by-wire operations of the Airborne Simulator.
when Pilot Response was Triggered by Attitude If, therefore, the reaction of ASRA to commanded hardovers
Divergence can be made no more severe than that of the 205, then we can
confidently continue this mode of operation without limiting
Rate Trigger Trigger Trigger the FBW flight envelope of ASRA.
Limit lODeg 15deg 20deg
8.0 FUTURE DEVELOPMENT
Attitude Time Attitude Time Attitude Time
10% 16.12 2.04 21.58 2.34 26.75 2.62 8.1 Hardware
15% 18.31 1.82 24.18 2.04 29.70 2.25 It is planned to mount a set of actuators on a ground
20% 20.24 1.69 26.45 1.88 32.23 2.04 development rig and examine the efficacy with which the
25% 21.97 1.60 28.48 1.77 34.54 1.91 modeled structure can be implemented in practice. The
actuators will also be measured for large and small signal
performance and for match to the manufacturers model.
 29-7

8.2 Off-Line Simulations

Off line simulations will be continued to investigate various
forms of actuator monitoring to provide protection against post
digital hardware failures (amplifiers, demodulators etc).
Specific attention will be given to using a software model of
the actuators as comparators and, in conjunction with the
ground test rig mentioned above, to determine possible error
level triggers for automatic disconnect. Such a scheme will
have to account for varying actuator loads and system noise.

8.3 In-Flight Manned Simulations

It is proposed to conduct manned simulations using the NRC
Airborne Simulator to confirm safety pilot reaction times to
the kinds of failures detailed in this paper before committing
to a specific set of compound actuator characteristics. The
results of these simulations will also be used to determine
what kind of monitoring/comparing system might be required.

9. CONCLUSION
On the present time table, it is expected that ASRA will be
available for fly-by-wire experimentation in the fall of 1996.
Once the conversion has been completed, the vehicle will
provide a platform for advanced systems and flight mechanics
research which will be significantly more powerful than the
current Airborne Simulator

10. REFERENCES

1. Anon "Aeronautical Design Standard - Handling Qualities

Requirements for Military Rotorcraft" US Army AVSCOM,
St Louis, Mo. ADS 33-D, July 1994

2. Baillie, S.W., Kereliuk, S., Morgan, J.M. and Hui, K. "An

Evaluation of the Dynamics and Handling Qualities of the
Bell 412 HP Helicopter" National Research Council of
Canada, Institute for Aerospace Research LTR-FR-121, June
1993

3. Beauchamp, G.M. and Lum, K.K., "Design and

Development of a Portable, DSP Micro-processor Bases, High
Accuracy Data Acquisition System" National Research
Council of Canada Report NRC-32146, March 1989.

4. Baillie, S.W., Morgan, J.M., Mitchell, D and Höh, R, "The

Use of Limited Authority Response Types to Improve
Helicopter Handling Qualities during Flight in Degraded
Visual Environments", AHS 51st Annual Forum, Fort Worth,
Tx., May 1995
 30-1

RESEARCH APPLICATIONS AND CAPABILITIES OF THE NASA/ARMY

ROTORCRAFT AIRCREW SYSTEMS CONCEPTS AIRBORNE LABORATORY
(RASCAL)

Edwin W. Aiken
Robert A. Jacobsen
William S.Hindson

NASA Ames Research Center

Mail Stop 211-2
Moffett Field, CA 94035-1000
USA

comparatively benign ground simulation environment

ABSTRACT can result in additional development costs when the
The Rotorcraft Aircrew Systems Concepts technology is transferred to flight vehicles.
Airborne Laboratory (RASCAL) is a UH-60 Black Consequently, an in-flight simulator in which to
Hawk helicopter that is being modified by NASA and evaluate and validate system integration designs is
the US Army for flight systems research. The principal becoming an increasingly important requirement.
systems that are being installed in the aircraft are a
Helmet-Mounted Display (HMD) and associated In-flight simulators have been an integral part of
imaging systems, and a programmable full-authority the development of aircraft technology since the first
Research Flight Control System (RFCS). In addition, one was developed in 1947 at NASA's Ames Research
comprehensive instrumentation of both the rigid body Center. In the past, these R&D facilities have
of the helicopter and the rotor system is provided. This typically been limited to the investigation of handling
paper describes the design features of this modern qualities requirements as manifested primarily in the
rotorcraft in-flight simulation facility and their current flight control system. Now, the strong impetus to
state of development. A brief description of initial integrate complex avionics and mission equipment in
research applications is included. both military and certain civil helicopters is placing
critical new demands on the pilot. The influence of
INTRODUCTION these demands on safety and mission effectiveness that
can only be fully understood in the real flight
The application of modern digital technology is environment. The RASCAL is the latest in-flight
accelerating with the development of sub-systems, such simulator to be developed at Ames, and it is the most
as flight and fire control systems integrated with helmet versatile. The combination of its advanced fly-by-wire
mounted displays, that tax the capability of the pilot to flight control system and displays capability make it
accomplish new complex mission tasks. Fortunately, uniquely suited for the simulation of today's complex
new technologies, such as digital flight control systems, piloting tasks.
are also available to offset these new demands on the
pilot. It was determined (Ref. 1) that the Army/Sikorsky
UH-60A Black Hawk helicopter was the most
Much of the work needed to develop practical appropriate baseline vehicle for RASCAL development.
pilot-vehicle systems that do not exceed the capability In October of 1989, a JUH-60A, originally used as the
of the pilot can be done in ground-based simulation. Army's Advanced Digital-Optical Control System
However, as pilot workload and the operating (ADOCS) demonstrator aircraft (Ref. 2), was loaned to
environment approach limits, such as in nap-of-the- NASA-Ames Research Center by the U.S. Army to
Earth (NOE) operations, conclusions drawn from the develope the RASCAL research facility.
Presented at the AGARD Flight Vehicle Integration
Panel Symposium on "Flight Simulation - Where are PROJECT OBJECTIVE AND RESEARCH
the Challenges?," May 22-25, 1995, Braunschweig, APPLICATIONS
Germany. The objective of the RASCAL facility
development project is to design and develop an

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?'
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
30-2

airborne laboratory capable of supporting the flight problem

research requirements of several different programs.
• Integration of vision sensors; including
These programs involve the close cooperation of NASA
television, infrared, and millimeter wave
with the Army, FAA, and the rotorcraft industry.
radar with an on-board terrain data base
The joint NASA/Army programs have a common and appropriate guidance algorithms to
objective: to develop integrated control, sensor, and allow manual and pilot-directed,
display technology to increase the capability and safety automated low-level flight
for all-weather, low-level rotorcraft operations. The
individual elements of these programs are: • Development of the pilot-vehicle
interface, including an advanced helmet-
• Development of flight-validated modeling mounted display and integrated, active
methods employing system identification controllers to enhance the pilot's
and a rotor state measurement system to situational awareness and increase the
ensure the development of realizable operational maneuver envelope of the
control laws vehicle
• Development and validation of multi- NASA programs that are in support of FAA and
input/multi-output control synthesis the civil rotorcraft industry include the Civil Rotorcraft
techniques suitable for rotorcraft which IFR Terminal-Area Technology Enhancement Research
allow physical insight into the design (CRITTER) program that is conducting research to

System Control Monitor and Keyboard

Computer

Video Cameras Scramnet

I"
.ViSSVWi
Image Processor

Display Electronics
Unit
Programmable Display
Generator

Head Tracking System

MM'm
Differential GPS E<AHAkm>A«A Panel Mounted Display V*. .a?.** a/
Computer

LN-93 INU

Helicopter Side Arm Controller

Air Data System

Analog Sensors
- Rigid Body Research Flight Control
- Rotor States Computer Assembly
- Propulsion
- Air Data

MIL-STD-1553B

ArT 33
33
Research Servos

Figure 1. RASCAL Research System Architecture

 30-3

improve the operating safety and community propulsion system sensors, suitable for the
acceptance of rotorcraft, including Tilt Rotors. acquisition of experimental data and for
Preliminary research describing the development of implementation in flight control
one-engine-inoperative flight techniques to improve applications.
safety and pay load is described in Refs. 3 and 4. In
addition to this NASA-initiated research, the RASCAL • A programmable, fly-by-wire research
helicopter can be used by the helicopter manufacturing flight control system including high-
industry as a common national facility to conduct flight performance servos, a flight control
computer with a hardware/software
tests.
architecture necessary for the
RESEARCH SYSTEM DESCRIPTION computational throughput requirements of
the various research control concepts, and
The RASCAL is being developed in an incremental a high-speed data bus with sufficient
fashion which allows the conduct of research programs capacity for the anticipated bus traffic.
during the development period. As additional
capability is installed, it remains part of the research • An in-flight researcher interface for
system for future use in research programs. Flight test monitoring the experiments and for
programs which have been conducted on the RASCAL effecting configuration changes to allow
include: productive use of flight time.
• Acquisition of video data which, when • An on-board precision navigation system
coupled with the recorded aircraft rigid- suitable for low-altitude flight, including a
body state data, allow the development of ring-laser gyro Inertial Navigation Unit
obstacle ranging information for (INU) and a Differential Global
presentation to the pilot (Ref. 5.) Positioning System (DGPS).
• Development of a laser-based • Appropriate passive (e.g., TV or FLIR)
measurement system to determine the and active (e.g., radar or laser) sensors for
flapping, lead-lag, and pitch angles of image-based guidance and navigation
each of the rotor blades. Initial data have including obstacle detection/avoidance.
been acquired and are being evaluated.
These data will allow the development of • An extensive on-board computational
refined models of rotor dynamics for capability for real-time image processing,
simulation use. vehicle motion estimation, guidance
algorithm generation, and pilot's display
• Development of a helicopter approach generation.
guidance methodology using differential
GPS to minimize noise for helicopter • Terrain data base storage for low-altitude
operations in urban environments. (Ref. 6) navigation.
The RASCAL helicopter is configured as an in- • A programmable cockpit display system
flight simulator with the evaluation pilot in the right consisting of two panel-mounted color
cockpit where the controls are mechanically displays suitable for the presentation of
disconnected from the aircraft and electrically guidance information and digital maps.
connected through a full-authority fly-by-wire flight
• A color, wide field-of-view, Helmet-
control system developed specifically for the R&D role.
Mounted Display (HMD) combining
The left cockpit controls remain mechanically
flight status and command information
connected to the JUH-60A to serve as a safety pilot
with sensor-based imagery and computer
station for monitoring the system and for assumption of
based imagery.
control in the event of a system fault. The RASCAL
research system architecture, shown in Figure 1, Two key elements of the RASCAL research system
provides a flexible and highly capable research facility which support operation of the JUH-60A as an in-flight
while ensuring flight safety. This system will include: simulator are the Helmet-Mounted Display System and
the Research Flight Control System. The remainder of
• A high quality data acquisition system, this paper describes the requirements and the designs of
including rigid body, rotor state, and those two systems.
30-4

Control Panel
t .
Audio In
Control ^ Boresight
Audio Out Reticle
r r r Unit
_ HTSData ON/OFF CONT BORE-
BRT SIGHT
System
Control l/F
Programmable
^ Boresight
Display
<l>
Generator Display Helmet Audio Out
ü Electronics
co Mounted
t Unit Power ^ ^ Audio In
o
Display
J Video
Sweeps ^ »
% Video In ^ Imagery
i_
Sirius Video ^
.E Video Out Expansion
<

Sensor
^ Data

Head Tracking
System (HTS) Excitation HTS
Source

Figure 2. Helmet -Mounted Display System

relay optics necessary to project the CRT image to the

Helmet-Mounted Display System pilot, and a head tracker sensor. The color display is
The RASCAL Helmet-Mounted Display system is generated by two high-brightness monochrome CRTs,
being developed for NASA by Kaiser Electronics and one for each eye, in front of which are placed rotating
consists of the helmet with its color display generating color filters, divided into two or three color sections.
capability, a head tracking system to allow presentation The data on the CRTs are synchronized with the
of airframe or inertially stabilized information to the rotating color filters, so that the appropriate datum
pilot, a display control panel allowing the pilot to appears when the corresponding color section of the
control the system, a programmable display generator wheel is in front of the CRT viewing surface. This
to process aircraft state and guidance information and mechanically shuttered field-sequential color generation
produce the symbology for presentation to the pilot and technique provides the RASCAL HMD with two-color
a boresight reticle unit for registration of the display or full three-color capability depending on the filter set
symbology and imagery to the aircraft axes. A that is installed. To permit operation in locations of
functional block diagram of the system is illustrated in high ambient lighting, the color filters can be removed
Figure 2, and the principal system features are listed in resulting in a monochromatic presentation. The
Table 1. Additional elements include the video sensors performance provided by this technology is listed in
which provide a forward-looking image to the HMD Table 1. The relay optics are a 10 element color-
system. corrected system which provide a 40 degree monocular
Helmet-Mounted Display (HMD): The HMD field-of-view. The optics use a combiner element
consists of three parts: the Aircraft Retained Unit which provides 30% see-through to the outside world.
(ARU), the Pilot Retained Unit (PRU), and the Display The ARU is connected to the DEU via a cable assembly
Electronics Unit (DEU). The ARU includes two which includes a quick disconnect allowing rapid
cathode ray tubes (CRTs), mechanical color filters, emergency egress.
 30-5

Table 1. RASCAL Helmet-Mounted Display Principal Features

PARAMETERS PERFORMANCE

40 Degrees Circular (monocular)

Field-of-View Overlap 50%
Total FOV - 40 deg (Ver) X 60 Deg (Hor)
1 Color Yellow 1136 fL u' =0.146 v' = .565
2 Color Green 297 fL u' =0.1066 v'=.5819
Modes of Operation, Red 83.55 fL u' = 0.4701 v' = .529
Brightness & Chromaticity 3 Color Green 297 fL u' =0.1066 v' = .5819
Red 41.7 fL u' =0.4701 v' = .529
Blue 8.81 fL u' = 0.0696 v' = .3177
Exit Pupil 15mm on axis; 10mm Off-axis
Eye Relief 17mm, Eye Glass Compatible
1 Color Yellow CR 4.34 10,000 fL Ambient, Class II visor
2 Color Green CR 3.91 3,000 fL Ambient, Class II visor
Contrast Ratio, Viewing Condition Red CR 1.81 3,000 fL Ambient, Class II visor
3 Color Green CR 52.4 30 fL Ambient, Class I visor
Red CR 8.23 30 fL Ambient, Class I visor
Blue CR 2.52 30 fL Ambient, Class I visor
Line of Sight Positional Accuracy Horizontal - 5.5 mrad Vertical - 3.5 mrad
1 Color Yellow 4.48 SOG 10,000 fL Ambient, Class II visor
2 Color Green 4.18 SOG 3,000 fL Ambient, Class n visor
Shades of Gray, Viewing Condition Red 1.97 SOG 3,000 fL Ambient, Class II visor
3 Color Green 11.67 SOG 30 fL Ambient, Class I visor
Red 6.32 SOG 30 fL Ambient, Class I visor
Blue 2.92 SOG 30 fL Ambient, Class I visor
Luminance Variation ±20% within 10 degree diameter
±30% within total FOV
Combiner Transmissivity 30%
Brightness (left/right)
Display Controls Contrast (left/right)
Boresight On/Off
Input Power 115 VAC 3 Phase 400 Hz
Registration (in overlap) Horizontal ± 5mrad Vertical ± lmrad
Ocular Alignment +1/2 diopter, -1/8 diopter

The PRU consists of the helmet shell, liners, CRT drive signal generation, color filter control, and
microphone, and earphones. The helmet is custom fit power for the ARU. The total head-borne weight of the
to an individual pilot's head. No alignment is necessary HMD is less than 5.5 pounds, and the center-of-gravity
other than preflight boresighting. The liner, when is +0.7 in. ±0.1 in. forward and +0.6±0.1 in. above the
inserted into the PRU and fitted, provides a close fit center-of-gravity of the head. The helmet-mounted
with air circulation for comfort. With the exception of display and ARU are shown in Figure 3 and Figure 4,
the comfort fit liner, no part of the head-borne HMD respectively.
system components make facial or eye contact. The Display Control Panel (DCP): The DCP provides
ARU and PRU are firmly held together by a bayonet- the pilot with the capability to control power to the
type mechanical latch. The DEU provides the HMD system, adjust brightness and contrast of each
electronic interface between the Programmable Display display and to conduct the initialization boresight
Generator (PDG), the ARU, and the Display Control operation.
Panel (DCP). It provides video signal conditioning,
30-6

PRUCG

AD/Neck CG ,— ARUCG

CRT/Optics CG

Figure 3. Helmet-Mounted Display

Latch

.- ARU Frame

CRT Assy.

Optics Assy.

Shutter Assy.

Figure 4. Aircraft Retained Unit (ARU)

 30-7

Head Tracking System (HTS): The HTS is a capability to provide the required field-sequential
ruggedized version of the Polhemus FASTRACK™ imagery to the HMD. The PDG is programmable in the
tracker. It consist of an aircraft-mounted source, a C programming language and provides access to low-
helmet-mounted sensor, and a rack-mounted electronics level graphics primitives from the high level language.
chassis. The HTS provides the following performance: The hardware and software are capable of generating
complex, real-time, dynamic displays and computer
• Angular coverage: ±180 degrees azimuth, ±90 generated imagery, utilizing double-buffered hardware
degrees elevation & roll that produces flicker-free images. The PDG is also
• Translational coverage: 16 inches on each capable of communicating with the in-flight researcher
interface system. The PDG accepts inputs from the
axis
head tracking system and provides input/output (I/O)
• Angular resolution: 0.44 mrad via Ethernet, MIL-STD-1553B and RS-232 data busses.
• Translational resolution: 0.1 inch HMD Symbology Software: The initial flight
display software to be used during HMD acceptance
• Static line-of-sight accuracy: 2.6 mrad RMS testing and early flight evaluations was derived from a
display format developed by NASA and implemented
• Update rate: 60 Hz on the Army's STAR UH-60 (Ref. 7). The symbology
• Transport delay: < 16.67 msec associated with the display is shown in Figure 5. It
features inertially referenced flightpath symbology as
• Noise: no perceptible dither to the displayed well as screen-fixed symbology which can be viewed
image against the background real world or overlaid on video
Programmable Display Generator (PDG): The imagery from the forward-pointing cameras. Initial
PDG is a Silicon Graphics Onyx Reality Engine 2 research to be conducted with this format will extend
deskside work station which has been ruggedized for the effort reported in Ref. 7 (which used a first-
the flight environment It is packaged with a VME generation narrow field-of-view monochromatic
compatible chassis which provides system expansion monocle display) to take advantage of the RASCAL'S
capability. The PDG and associated software and wide field-of-view color capability.
firmware constitute a programmable
graphic/alphanumeric/symbology generator with high- Nose Mounted Camera System: The camera
resolution color and monochrome display generation apparatus consists of two monochrome interlaced video

N 01 02 03 04
I I I I I I I I I
Airspeed Phantom
flight director aircraft

IhPathway
Aircraft nose "symbols

Horizon line Flight path

I-vector
predictor

Pitch attitude
reference lines'

Figure 5. HMD Symbology

30-8

cameras mounted 1 meter apart on a horizontal bar Boeing Defense and Space Group together with a team
attached to the nose of the RASCAL helicopter. The which includes Lear Astronics and Moog are under
cameras have a focal length of 6 mm, a field of view of contract to design and build the RFCS. Boeing and its
58X45 degrees, and they are electronically shuttered sub-contractors, Lear Astronics and Moog, have
with a 1/1000 sec exposure time to reduce image smear presented a detail design of the RFCS at a recent
due to camera motion. The video imagery from the Critical Design Review. Information provided at that
cameras is time-tagged using video time inserter units review is the source for many of the details presented in
and recorded using two video recorders onboard the the following sections.
RASCAL. The images are acquired at the rate of 30 RFCS Architecture. The RFCS architecture,
frames/sec per camera. shown in Figure 6, implements a full-authority fly-by-
wire capability while retaining the unmodified JUH-
Research Flight Control System 60A mechanical flight controls as backup. This
The requirements and design philosophy for the architecture is fail-safe; that is, critical RFCS faults
development of the RASCAL Research Flight Control cause immediate disengagement of the RFCS and
System (RFCS) are described in Ref. 8. Key aspects of reversion to the JUH-60A mechanical flight controls.
the RFCS architecture and components are presented
below. Currently, the Helicopter Division of the

Pilot assist actuators Swashplate

Safety pilot

Tail rotor research servo

Vw.iu!*w - Mixing
|c^0?§g::i:g:::
\:^::-:;:tin^g;:::i::xi:j:i::::>:::

Displacement iöwttaps
Collective
Controller :'S^:<rftSSee

Figure 6. RFCS Architecture

 30-9

The main components of the RFCS are: provides inputs to the tail rotor servo while backdriving
the safety pilot's pedals and collective control. The
• Research Flight Control Computer Assembly transfer system provides the means for transferring
(RFCCA) control form the JUH-60A mechanical flight controls to
• 3 Main Rotor and 1 Tail Rotor Research the RFCS and vice-versa. It is designed with sufficient
Servos redundancy to ensure that either of the pilots or the
safety monitors resident in the RFCCA can reliably
• Transfer System disconnect the RFCS. The RFCCA has a multitude of
I/O resources to support many pilot controller and
• Pilot Controllers aircraft state sensor types. Initially, the RASCAL will
• Aircraft State Sensors be equipped with a 3-axis side-arm controller and a
displacement collective control. In addition, two
• Pilot and System Operator Cockpit Display Cockpit Display Units (CDUs) will provide the
Units (CDUs) capability for the pilot or the system operator to set
control modes, adjust control parameters, and get RFCS
The RFCCA is a high performance computer health information.
assembly which will process sensor input, compute the
experimental control laws, and monitor RFCS RFCS Components. Salient features of the major
performance. It controls the three main rotor research RFCS components are described below.
servos and the tail rotor research servo. The main rotor RFCCA: The RFCCA architecture, shown in
research servos are mounted on the upper crown area of Figure 7, includes the two functional components, the
the JUH-60A and provide inputs to the primary boost
Flight Control Computer (FCC) and the Servo Control
servos while backdriving the safety pilot's cyclic and Unit (SCU), whose requirements were described in Ref.
collective controls. The tail rotor research servo is 8. This architecture provides in one line-replaceable
mounted in the vicinity of the tail rotor gear box and unit an FCC with high-performance I/O and

FCC

1 ,.-*•.. 1
.♦
SCU SCU
256 Hz IOP1 IOP2

■
Legend
_ Parallel Bus
1024 Hz AIM1 AIM2 Serial Link
a

■ Channel Fail Discretes

Synchronization Discrete
DUAL CHANNEL SCU

Figure 7. RFCCA Architecture

30-10

computation throughput for aircraft state acquisition frame. The IOPs provide the FCC with a substantial
and control law computation and an SCU for fail-safe capacity to perform sensor I/O and filtering. A
RFCS control and monitoring. Inside the RFCCA, the summary of FCC I/O resources, most of which are
FCC and the SCU are physically segregated from each provided by the IOPs, is shown in Table 4. Operating
at a 60% duty cycle, each IOP can sample, scale, and
float 60 analog inputs, process eight 20-word MIL-STD
1553B messages, and filter (with 4th order filters) 85
Table 2. IOP Attributes variables every 3.9ms frame. I/O data is passed
between an IOP and the CLP through the IOP's 2Kx32
Microprocessor: TI320C30, 32-bit bytes of dual-port RAM. The CLP accesses the dual-
Digital Signal Processor port RAM through a 32-bit system bus.
Clock Speed: 21 MHz
The software structure in the IOPs and the CLP
Throughput: 10.5 MIPS
allows the configuration of the FCC I/O to be managed
Program Memory 64Kx32 (EEPROM) through configuration tables in the CLP software.
Data Memory 8Kx32 (SRAM) These tables are downloaded to the dual-port RAM on
2Kx32 (Dual Port RAM) power-up to provide each IOP with information
defining analog input scaling and filtering, analog
Table 3. CLP Attributes output scaling, discrete input debouncing and logic, and
MIL-STD 1553B message definitions. Thus, only the
Microprocessor: IDT R3081W with on- CLP software needs to be changed in order to
chip floating point unit accommodate changes in the sensor types and
Clock Speed: 33 MHz attributes.
Throughput: 24MIPS/4MFLOPS
The SCU is a dual-channel design with each
Program Memory: 1MX32 (FLASH)
channel consisting an IOP module (identical to those
1MX32 (SRAM)
used in the FCC) and an Actuator Interface Module
(AIM). It is responsible for controlling the four research
servos, the evaluation pilot's collective actuator and
Table 4. FCC I/O Resources clutch, and the JUH-60A stabilator in response to
DC Analog Inputs 160 (40 per IOP) commands from the FCC. It also provides inputs to and
DC Analog 16 (4 per IOP) monitors the operation of the transfer system. Most
Outputs
important, though, is the function it performs as the
36 (9 per IOP)
RFCS safety monitor. An extensive set of monitors in
AC Analog Inputs
the SCU are designed to detect failures in the RFCS
Discrete Inputs 52 (12 each for IOPs 1,2,
components and anomalies in the experimental control
and 4; 16 for IOP 3)
laws that would result in unacceptable servo motions or
Discrete Outputs 20 degrade the fail-safe features of the RFCS. In the case
MIL-STD-1553B 4 dual-redundant buses of such faults, the SCU will disengage the RFCS
Serial Data Buses (1 per IOP) without unacceptable transients. A functional
RS422 Serial Data 2 (CLP) schematic of some these monitors is shown in Figure 8.
Buses 4(1 per IOP) Many of the other RFCS component monitors are
Ethernet 1 (CLP) omitted from this figure for clarity. Servo commands
are monitored for excessive step size and for commands
that exceed the control limits. In addition, the servo
other. They communicate through dedicated high- rates are monitored to ensure that rates above 1/3 of the
speed serial links, synchronization discretes, and maximum servo rate are not allowed to persist. This
monitor uses a time threshold that is inversely related to
channel fail discretes.
the servo rate. The difficulty in developing the
The FCC is comprised of 4 digital signal appropriate monitors and in selecting the threshold
processor-based Input/Output Processors (IOPs) and a levels in order to detect sensor and control law failures
RISC-based Control Law Processor (CLP). IOP and without experiencing excessive nuisance disconnects
CLP attributes are summarized in Table 2 and Table 3. was described in Ref. 8. In order to provide some
The IOPs and the CLP execute at a 256 Hz frame rate flexibility in this area, the SCU will contain four sets of
and are required to synchronize with each other every
 30-11

scu Servo Command Research

Monitoring Servo

Servo Commands
FCC Servo Command
Mixing & Limiting
Servo Loop
Closure -H:

Servo Position
Demixing

Servo Rate Monitoring

&
Command - Position
Synchronization
Monitoring

Evaluation
Pilot Collective
Collective Motor Command
Collective
Position
Limiting &
Monitoring

Stabilator

Stabilator Command
Stabiiator
Command Stabilator
Limiting & Amplifiers
Airspeed Monitoring

Figure 8. SCU Monitoring Schematic

threshold levels from which the pilot can select prior to assembly uses a semi-balanced piston-cylinder design.
engaging the RFCS. Each assembly contains a conventional flapper-type,
four-way electro-hydraulic servovalve (EHV) with a
Research Servos: The three main rotor research
dual-coil torque motor. Each coil in the torque motor is
servos and the tail rotor research servo provide high-
driven by an SCU channel. The second stage EHV
bandwidth, high-rate, and full-authority inputs to the
slots are unequal in area to provide a symmetric
JUH-60A primary servos. The four servos are identical
velocity response for the servos. EHV spool position
except for the tail rotor research servo which has a
feedback is used for loop damping as well as for servo
longer stroke and faster maximum rate. A functional
fault detection. In addition, each servo includes
schematic of the servos is shown in Figure 9. The servo
30-12

Engage Arm Return

Pressure Pressure Pressure

SCU1 Dual-coil 4-way LVDT

SCU2 flapper EHV -SCU 1/2

r~[
Uo zvv
Arm
Bypass Valve,
Pressure Operated
SCU1/2-
LVDT
t
SCU 1/2 "
n /vs/
Engage
Bypass Valve,
Pressure Operated

LVDT { y

Relief Relief
Valve Valve

SCU 1/2 SCU 1/2-

Pressure Transducer
Pressure Transducer

LVDT

G D Semibalanced
Actuator

SCU 1/2

Figure 9. Research Servo Functional Schematic

pressure sensors to measure differential pressure across small-signal bandwidth of 8 Hz (defined by phase < 30
the piston for resonant mode stabilization and load degrees). The tail rotor research servo has a maximum
limiting. Pressure relief valves provide a redundant rate of 7 inches per second and has a bandwidth similar
means of load limiting. Each servoactuator package to that of the main rotor servos.
contains arm and engage bypass valves which provide Transfer System: The transfer system is the set of
a dual-redundant means to put the servo into bypass cockpit switches, relays, and hydraulic shutoff valves
mode in the event of RFCS disconnect. These bypass that control transfer between the RFCS and the
valves are pressure operated and feed back spool mechanical JUH-60A flight control system. A
position to the SCU for monitoring of the research schematic of the transfer system is shown in Figure 10.
servo state. The main rotor research servos have a The state of the RFCS is determined by the state of a set
maximum rate of 5 inches per second (equivalent to the of dual-redundant relays. These relays directly control
JUH-60A primary servo maximum rate) and have a the hydraulic shutoff valves that control hydraulic
 30-13

28Vdc

To
Shutoff
Valves RFCS
Hydraulic
System
To Trim and
Boost Release Pilot
Engage *• Assist
Servos
Disengage
Relay
Coil Stabilator
Grounds Stabilator
RFCCA A- Amplifiers
V V
Stabilator
Commands
4\
and Wraparounds

Hardware Acknowledge Signals

Legend:
—►- 28 Vdc Power
^> Signals to/from RFCCA

Figure 10. Transfer System Schematic

pressure to the arm and engage bypass valves of the the FCC. A medium-displacement left-hand collective
research servos. In addition, these relays are used to controller provides the capability for conventional
turn off the JUH-60A trim and pilot-assist servos when collective inputs to the FCC. The collective controller
the RFCS is engaged. Another set of relays controls the includes a low-rate elctromechanical actuator to allow
switching of the stabilator commands between RFCS the FCC to backdrive the controller when the RFCS is
control and the JUH-60A stabilator system AUTO disengaged. This will ensure that the evaluation pilot's
mode control. Through the RFCS control panel and collective is aligned with the safety pilot's collective
controller grip switches, the evaluation and safety pilots prior to RFCS engagement, necessary to prevent engage
control the 28Vdc going to the arm and engage relays. transients. This controller lever/actuator mechanism is
Each channel of the SCU controls the ground path for based on that used on ADOCS while its grip is from the
one of the dual-redundant relays. With this design, Comanche program.
each pilot and each SCU channel has an independent
Sensors: The wealth of FCC I/O resources provides
means to disengage the RFCS.
the RASCAL facility the capability to accommodate a
Pilot Controllers: The RASCAL helicopter has the large number of sensors of various types. The initial
capability to evaluate various pilot controller sensor suite that is directly interfaced with the RFCS
configurations. Initially the RASCAL will be includes the following:
configured with a small-displacement 3-axis force-
• LN-93 Inertial Measurement Unit
gradient right-hand controller for longitudinal, lateral,
and directional inputs to the FCC. This controller is • Blade Rap, Lead-lag, and Pitch Laser
identical to that developed for the RAH-66 Comanche Position Sensors
helicopter. It also has a limited vertical motion,
• Blade Hap and Lead-lag Accelerometers
intended for aircraft trim commands. The grip used
with this controller is also that from the Comanche. It • Rotor Azimuth Encoder
includes 8 switches of various types that will be read by
30-14

• Flight Control Position Sensors REFERENCES

• Nose Boom and Body Mounted Airspeed

and Pressure Altitude Sensors 1. Aiken, E.W., Hindson, W.S., Lebacqz, J.V.,
Denery, D.G., and Eshow, M.M., "Rotorcraft In-Flight
• Helicopter Air Data System with fully
Simulation at NASA Ames Research Center: A Review
gimbaled pitot/static probe
of the 80's and Plans for the 90's," International
• Radar Altimeter Symposium on In-Flight Simulation for the 90's,
Braunschweig, Germany, July 1-3, 1991 (NASA TM
In addition, other sensor information (e.g. DGPS)
is available from the PDG computer via the MIL-STD- 103873, August 1991).
1553B data buses. 2. Dones, F.D., Glusman, S.I., et al., "Advanced
Digital/Optical Control System Final Technical Report,
Cockpit Display Units: Two CDUs will be installed
Volume 2 - Systems Engineering," AATD Report No.
on the RASCAL for control, setup, and monitoring of
D358-10055-2, Applied Aviation Technology
the RFCS. The CDUs will be used for selecting control
Directorate/Boeing Helicopters, Dec. 1989.
modes, adjusting control law parameters, displaying
fault messages, executing pre-flight and maintenance 3. Sharma, V., Zhao, Y., Chen, R.T. N., and
checks, and performing INU alignment. A pilot CDU Hindson, W.S., "Optimal OEI Clear Helicopter
will be mounted in the cockpit center console. A Operation of a Multiengine Helicopter," Presented at
second CDU will be installed at the system operator the 51st Annual Forum of the American Helicopter
station in the aft cabin to allow the system operator to Society, Ft. Worth, TX, May 1995.
participate in test setup and system monitoring. The
4. Iseler. L., "Piloted Simulator Investigation Of
CDUs will interface with the RFCCA over one of the
Category A Civil Rotorcraft Terminal Area Cockpit
MIL-STD-1553B data buses.
Displays," Presented at the 51st Annual Forum of the
Control Laws: A baseline set of control laws will American Helicopter Society, Ft. Worth, TX, May
be delivered with the RFCS. This set of control laws 1995.
will be based on those developed for the ADOCS
5. Smith, P. N., Sridhar, B., and Hussein, B.,
program, but will only implement the core functions of
"Vision Based Range Estimation Using Helicopter
those control laws that are needed to provide good
Flight Data," NASA TM 103930, June 1992.
handling qualities. The core functions will include
command shaping, automatic trim, rate and attitude 6. Hindson, W.S., and Chen. R.T.N., "Flight Tests
stabilization, heading hold, and turn coordination. of Noise Abatement Approach Profiles for Rotorcraft
These control laws will facilitate the rapid integration Using Local Differential GPS Guidance," Presented at
and checkout of the RFCS before proceeding with the the 51st Annual Forum of the American Helicopter
implementation and testing of experimental control Society, Ft. Worth, TX, May 1995.
laws.
7. Zelenka, R.E., Clark, R.F., and Branigan, R.G.,
"Flight Test of a Low-Altitude Helicopter Guidance
CONCLUDING REMARKS
System with Obstacle Avoidance Capability,"
Presented at the AGARD Mission Systems Panel
The unique and highly capable pilot display Symposium on "Low-Level and Nap-of-the-Earth
capability that the RASCAL will provide, combined (NOE) Night Operations," October 24-27, 1994, Pratica
with its advanced flight control capability, create a di Mare (Rome), Italy.
research tool which will far surpass the traditional in- 8. Aiken, E., Jacobsen, R., Eshow, M., Hindson,
flight simulators of the past. The combination of these W., and Doane, D., "Preliminary Design Features of the
capabilities creates, in the RASCAL, a full-environment RASCAL - A NASA/Army Rotorcraft In-Flight
simulator which extends the simulation technology to Simulator," AIAA/AHS Flight Simulation
which we have become accustomed in the ground- Technologies Conference, Hilton Head Is., S.C., August
based facilities into the unlimited motion-base 24-26, 1992.
environment of the in-flight simulator. The RASCAL
is precisely positioned to meet the most demanding
research and development needs for the rotorcraft
industry in the next decade.
 31-1

ATTAS & ATTHeS In-Flight Simulators

Recent Application Experiences and Future Programs

J. J. Buchholz
J.-M. Bauschat
K.-U. Hahn
H. J. Pausder
Institut für Flugmechanik
Deutsche Forschungsanstalt für Luft- und Raumfahrt e.V.
Lilienthalplatz 7
38108 Braunschweig
Germany

• Establishment of credible data for the definition of generic

1 SUMMARY evaluation criteria
This paper gives an overview of DLR's latest flight experi- • Investigation of interference effects between required
ments, both with the fixed-wing in-flight simulator ATTAS overall system response, display format, and inceptor type
and the helicopter in-flight simulator ATTHeS. After a de- • Requirements for system mode blending characteristics
tailed description of both testbeds and the corresponding • Investigations of allowable response degradations which
ground based simulators, flight test results are presented. occur after system failures
These include experiments on in-flight simulation, rate satura-
tion, reconfiguration, flight control laws, handling qualities, All investigations can be performed without endangering men
and vision based hovering. The paper concludes with a dis- and/or vehicle. In addition, the development and operation of
cussion on certain aspects of simulation fidelity. in-flight simulators inherently include the necessity to solve
many problems of a highly augmented system. Consequently,
2 INTRODUCTION many of the lessons learned and the technical solutions can be
Flying qualities play an essential part related to pilot/vehicle transferred directly to the design of operational vehicles.
performance (task effectiveness) and flight safety. Due to the
3 ATTAS AND ATTHES SYSTEM OVERVIEW
large amount of parameters influencing the flying qualities
(control system, information/display system, vehicle behavior, For more than 20 years DLR has been successfully operating
environment, flight task, etc.) pilot-in-the-loop simulation is a in-flight simulators for flying qualities research. From 1972 to
fundamental element in flight control law design. In-flight 1984 a HFB 320 Hansa Jet was extensively used as an in-
simulation (IFS) is recommended as one of the most important flight simulator by DLR, the German industry, and foreign
tools to be used complementary to ground based simulation. institutions. Handling quality assessment technology and data
IFS can provide substantial benefits in evaluating control ef- evaluation procedures have been developed and applied. In
fectiveness of highly augmented flight control systems, new the early 80's the modern flying simulator and demonstrator
controller types, and pilot information systems. It gives the aircraft ATTAS (Advanced Technologies Testing Aircraft
designers more flexibility to alter the response characteristics System) was developed by DLR and Daimler-Benz Aerospace
of the overall system and to tailor the desired flying qualities. Airbus (formerly MBB) supported by the German Ministry of
Research and Technology. Furthermore the Institute of Flight
The great advantages of an in-flight simulator are the realistic Mechanics of the DLR has developed a research helicopter
visual and motion cues which are the key feedback parameters based on a BO 105. DLR has been operating the ATTHeS
for the pilot to adapt his control strategy to the vehicle and the (Advanced Technologies Testing Helicopter System) as a
flight task. The results obtained from ground simulation stud- testbed and in-flight simulator for more than a decade.
ies have to be verified in the "real world", which means, that
they have to be tested in flight. To avoid the appearance of 3.1 ATTAS In-FIight Simulator
problems later in the development phase, which causes ex- The complex structure of the ATTAS testbed is outlined in
pensive delays and increasing costs, requires the utilization of several papers. A brief summary is given in [3]. ATTAS is
a flying testbed. The demands on high fidelity, high flexibility, based on a VFW614, a twin-turbofan, short-haul, 44-
and high quality of pilot cues formulate the need to use flying passengers aircraft (Fig. 1).
simulators [1].
The great advantage of an in-flight simulator, compared to a
ground simulator, is to fly the system in the real world with
adequate visual and motion cues for the pilot, and, compared
to a technology demonstrator, the capability to vary the system <>*"%.
response characteristics. Hardware components like new type '■'.«Sfflwiiw -äSsl; -
of inceptors or information displays can be installed. Besides
the complementary incorporation of an in-flight simulator in
the development approach of a new flight vehicle and the in-
tegrated technologies of pilot vehicle interfaces, additional Fig. 1 DLR In-Flight Simulator VFW 614 ATTAS
areas of use are more general and basic research topics [2]:
The VFW 614 is ideally suited as a general purpose testbed
• Control law design due to its size, cabin space, loading capacity, and flight per-
• Investigation of the response system required by a flight formance. With full fuel, about 3.5 tons of test equipment can
task be loaded.

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?"
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
31-2

The flight performance, with cruising altitude of 30 000 ft, • Stabilizer

maximum cruising speed of 288 kts calibrated airspeed (CAS) • Rudder
and a rather low landing speed of about 100 kts, is adequate • Both ailerons
for a large transport a/c flight regime representation. Fig. 2 • Both engines
illustrates the ATTAS flight envelope. • Landing flaps
• Six direct lift flaps
Flight • Onboard operator consoles with programmable quick look
Level
displays
300 • Nose boom with angle of attack, angle of sideslip, and true
airspeed probe (flight log)
200 - An automatic elevator trim system operates the stabilizer in
such a way that the aircraft is always in trim condition. In case
the safety pilot reconverts to the basic control system the re-
100 sulting column and aircraft transients will be negligible.
The ATTAS installed Rate Command - Attitude Hold system
was designed especially for the Fly-By-Wire Mode to stabilize
0 kts 500 the aircraft before switching to the Simulation Mode to run the
True Airspeed experiment (e.g. in-flight simulation). There must be a feature
to fly the ATTAS in the easiest way. For this the pilot controls
Fig. 2 ATTAS-Flight Envelope the a/c with all control elements in the direct mode. But in
The heart of ATTAS is the fly-by-wire/light system. It is based case he puts the side stick into the zero position, the last
on a five computer system. [4] and [5] describe the system achieved attitude in pitch or roll is kept independently in the
performance requirements as follows: pitch and roll axis. For level flight also a wing leveler function
• Maximum computational cycle time of 20 ms for all func- is active if the roll angle is less then 2 degrees [3].
tions
3.2 ATTHeS In-FIight Simulator
• Redundancy for flight in critical conditions
• Airborne equipment and interfacing of aircraft systems A description of the ATTHeS in-flight simulator and its ca-
(ARINC 429, MIL-Bus 1553B, etc.) pabilities is given in [2]. The research vehicle is based on a
• Availability of computer capacities for user applications in BO 105 helicopter (Fig. 3). The next generation of helicopters
a high level programming language are required to fly high pilot gain, high bandwidth, and high
precision flight tasks, which have to be performed in a heli-
To meet these requirements the system has been designed as a copter airborne simulator, too. The capability to cover the re-
two channel computer network consisting of four processors quired range of dynamic response behavior is essentially lim-
in each channel with one common central processor for com- ited by the characteristics of the helicopter being the host for
munications and data recording. All onboard computers are of the in-flight simulation. The high control power and the quick
MIL-specified LORAL/ROLM types (MSE14 and initial response on pilot's inputs of helicopters with a hingeless
HAWK/32). 1.4 MFLOPS are available for experimental rotor are excellent prerequisites. Correspondingly, the BO 105
functions in the HAWK/32 32-bit computer with up to 8 Mb helicopter is well suited to be a host for an in-flight simulator.
of memory. The network in each channel is based on a ring The high level of interaxis coupling and the high order re-
structured serial fibre optical bus system. sponse induced by the rotor system complicate the design of a
To give ATTAS a 5-DOF simulation capability five independ- full authority digital flight control system which is needed for
ent control surfaces must be available. Therefore, ATTAS was a capable in-flight simulator.
equipped with a specifically developed "Direct Lift Control"
System (DLC) for pitch/heave motion decoupling and
gust/load control. For low frequency DLC operation the basic
VFW 614 landing flap system can be deflected electrically
between 1 and 14 degrees. The rear parts of the landing flaps
have been divided into six (three on each wing) fast moving
flaps having about 85 deg/sec flap rate and ±35 degrees flap
deflection capability for high frequency direct lift modulation.
The main modifications, test equipment, and features of the
ATTAS aircraft are summarized as follows:
• Right hand seat safety pilot with conventional control
system
• Left hand seat evaluation pilot with fly-by-wire controls Fig. 3 DLR In-FIight Simulator BO 105 ATTHeS
• Freely programmable flight instruments/displays, CRTs The testbed is equipped with a full authority non-redundant
• Digital onboard computer system with fibre optical data fly-by-wire control system for the main rotor and a fly-by-light
bus providing freely programmable control laws and fly- (FBL) control system for the tail rotor. The test helicopter re-
ing qualities quires a two-men crew, consisting of a simulation pilot and a
• Duplex inertial reference and digital air data systems safety pilot. The safety pilot is provided with the standard me-
chanical link to the rotor controls whereas the simulation pi-
• Comprehensive onboard data acquisition system, record- lot's controllers are linked electrically/optically to the rotor
ing, and PCM-telemetry controls. The FBW/L actuator inputs, which are commanded
• 15 electro-hydraulic self-monitored actuators, partly du- by the simulation pilot and/or the flight control system, are
plex, linked to the FBW system by MIL-BUS 1553B mechanically fed back to the safety pilot's controllers. With
• Dual redundant electric and hydraulic systems this function, the safety pilot is enabled to monitor the rotor
• Fly-by-wire motivators for control inputs. This is an important safety aspect, because the
• Elevator
 31-3

safety pilot can evaluate, whether the inputs are adequate to control system is held in the trim position. The control system
the flight task. The safety pilot can disengage the FBW/L starts, when the simulation pilot switches on the control status
control system by switching off the FBW/L system or by and the computer generates a subcycle of 1/5 of the frame time
overriding the control actuators. In addition, an automatic (8 msec). The subcycle allows refreshing the FBW/L actuator's
safety system is installed, monitoring the hub and lag bending inputs in a lower time frame than the sampling frame. For the
moments of the main rotor. The testbed can be flown in three implemented explicit model following control system, which
modes: is described in the next paragraph, this refreshing time is as-
1. The FBW/L disengaged mode, where the safety pilot has sessed with 16 msec. The overall computation time for the
the exclusive control commanded model and the control laws is 13 msec. All data
2. The 1:1 FBW/L mode, where the simulation pilot has the regarding cycle times and the measured test data included in
full authority to fly the basic helicopter this paper are related to a PDP 11/73 control computer. This
3. The simulation mode, where the simulation pilot is flying control computer is now replaced by a PDP 11/93, which re-
a simulated helicopter system with full authority duces the sampling and computational time to half the values.
In the 1:1 FBW/L and the simulation mode the flight envelope The data recording computer is equipped with a 64 channel
is restricted to 50 ft above ground in hover and 100 ft above A/D converter. All sensor signals are sampled with a fre-
ground in forward flight. quency of 100 Hz. A sampling frequency, significantly higher
than the control computer sampling frequency, has been
To incorporate the digital control system for in-flight simula- specified to achieve a more precise assessment of the overall
tion purposes, an onboard computer and a data acquisition system performance. Both computers are linked by a dual port
system have been installed. In the specifications for the design memory. The measured signals, which are used in the control
the following system conditions and requirements have been computer, and the signals, which are calculated in the control
considered: computer, are transmitted via the dual port memory for record-
• Limited space is available in the helicopter. ing. The data are recorded onboard on a floppy disk. In addi-
• Software modifications in the control system must be ac- tion, the data are transmitted to the telemetry. The telemetry
complished in a host computer on the ground. data are only used for quicklook purposes in the ground sta-
• A system simulation facility which is compatible to the tion. The ground station also contains a host computer, which
onboard system is needed to check any software modifica- is compatible to the control system computer. Any modifica-
tions before going into flight. tions of the software code including changes of the control
• The onboard system tasks, the control system, and the laws and the command model parameters are first performed
evaluation of the control system performance have to be in this host computer on the ground and then transferred to the
clearly separated. onboard computers via floppy disk.
• The flight tests have to be observed and managed from a 3.3 Ground-Based System Simulators
ground station.
The real-time simulation of the in-flight simulators on ground
is an important tool. The structures of both the ATTAS and
Sensors
the ATTHeS ground-based simulators are shown in Fig. 5.
... ^ . The ATTAS Simulator [3] simulates the a/c as good as possi-
i ble without motion cue and a visual system. The onboard data-
Signal Signal
Conditioning * Conditioning ; processing system consists, as in the real a/c, of MIL-specified
computers. An original ATTAS-cockpit is part of the simula-
i •■ i tor.
A/D A/D
Converter Converter
The VFW 614 ATTAS is simulated in real-time on an Applied
Telemetry
' ""■ T Dynamics AD10. The programming language on this multi-
:
Data Rec.
Computer
processor system is MPS10. The necessary interface between
digital and analog world is a VME-Bus system [6]. The main
t Floppy Disk
task of the VME-computer is to transport the simulation data
Control Computer between the a/c real-time simulation on the AD 10 and the on-
board data-processing system via hardware interfaces. It is
" 1 1 designed to cover the special needs of the onboard MIL-
D/A Status
Floppy Disk specified computer network. The MIL-Bus is a bi-directional
Converter Keyboard
interface-bus transporting input- and output-data.
The VME-Bus computer contains two CPU-boards with
Actuators 68020 processors equipped with 1 Mbyte local memory each,
communicating via 64 kbytes shared memory. One processor
performs the analog in- and output, the other computes the
Fig. 4 Structure of ATTHeS Onboard System rest of the interfacing including the user interface.
Fig. 4 shows in a block diagram the onboard system. Two The operating system PDOS is in use for developing and run-
PDP 11 computers, ruggedized for operation in the airborne ning the software on the interface computers. The program-
environment, are installed. The data recording task and the ming language is C.
control system task are assigned to the computers, which allow
a largely autonomous treatment of the data streams needed for Corresponding to the ATTAS simulator described above an
the control laws and needed for the data recording for the ATTHeS simulator has been developed. The ATTHeS system
control system performance evaluation. simulation tool permits the examination of the real time soft-
ware. The basic helicopter and the actuating system are repre-
The simulation pilot's inputs and the state variables, which are sented by a non-linear model. The onboard computer system
used in the control laws, are obtained directly from the pre- has been duplicated in the ground system simulation. This tool
conditioned sensor signals with a 16 channel A/D converter. A
sample rate of 25 Hz is realized. After the initialization the
31-4

FLIGHT COCKPITS FLIGHT - OR EQUIVALENT FLIGHT VEHICLE AND SYSTEM

EHICLES GROUND COMPUTER REAL - TIME SIMULATION
SYSTEMS

SIMULATION
MULTIPROCESSOR
AIRCRAFT
SYSTEM
ADIO(RTSIO)

7>
TO CENTRAL
MAINFRAME

l£ 3£
HOST AND SW- PERIPHERY
DEVELÖPMENT - DISKS
COMPUTER ■TAPE
- PRINTER
VAX 3200 -TERMINALS
(SUN - SPARC) • PLOTTER

7>

\y

SIMULATION
MULTIPROCESSOR
ROTORCRAFT
SYSTEM
AD 100
MOTOROLA 68030

Fig. 5 Structure of ATTAS and ATTHeS Ground-based System Simulators

allows a compatible hardware and software in the loop testing is intentionally limited, either by a software rate limiter some-
before an implementation in the flying ATTHeS helicopter. where in the control system, or simply by the pilot, if the stick
is moved with a limited rate.
Every experiment flown on both in-flight simulators can be
checked on the ground. The scientist is able to validate his
software and obtain first results. Another important fact is to
train the pilots and find out their opinions before going air-
borne. The ground simulation reduces the costs and also a lot
of development risks. The standard of the ground-based real-
time simulators allows the realization of typical experiment
concerning simulation technique.

4 ATTAS FLIGHT TEST RESULTS

4.1 SCARLET
Within the SCARLET project (Saturated Command And Rate
Limited Elevator Time delay) the influence of actuator rate F^T
t, t,
saturation on the handling qualities of modern fly-by-wire air-
craft has been examined. The aim was to develop an alternate Fig. 7 Time Delay
control scheme to eliminate the time delay between actuator Consider now a situation in which the rate limited command is
command and reaction. faster than the actuator, as shown in Fig. 7. In the first time
period (0 < t < t,), the actuator strives to reach the command
c > act_out
with its maximum rate; however, a discrepancy will develop
acUnp Rate
Controller limiter
Motor between the magnitudes of the command and the output due to
the difference in rate limits. When the command then changes
direction at th its magnitude is greater than that of the actuator
output, and therefore the command begins to move into the
direction of the output while the actuator continues in its
Fig. 6 Rate Limited Actuator original direction as it tries to meet the command. Only when
Actuator rate saturation can lead to a significant time delay the magnitudes of the two signals meet at t2 will the actuator
between the actuator input and the actuator output. This time finally change direction to follow the command. The time pe-
delay arises due to the interaction of two rate limited elements riod between the reversal of the command and the reversal of
in the control system: the actuator itself, and the command to the actuator output is the time delay TD.
the actuator (actuator input). The actuator is rate limited due
to 'real-world' effects. The time delay could be eliminated if the actuator reversal oc-
curred at the same point as the command reversal, as shown in
Fig. 6 depicts a typical actuator consisting of a motor or hy- Fig. 8.
draulic booster and a controller. The error between the actua-
tor command and the output is calculated by the controller and Therefore, a solution is suggested through the following Al-
used to drive the motor. The motor, however, can only re- ternate Control Scheme (ACS):
spond with a limited rate due to constraints in electric current
or hydraulic flow. The actuator command, on the other hand,
 31-5

the actuator. Thus the time delay is eliminated.

If the actuator is rate saturated,
then coordinate the reversal of the actuator rate 1b
Tracking Task ^-^
with the reversal of the command. (Pitch Angle) f \
1
^ /i \ 7 ' '. \
0 i \ i / ^^s '_• i \ f~
1
\ ' / / ' \ i

Actual \
Pitch Angle \ '

Without ACS \
?s
0 10 20 30 sec 50
Fig. 10 SCARLET Flight Test Results without ACS
Fig. 10 shows an ATTAS flight experiment in which the ma-
log_out ximum elevator rate has been artificially limited to 0.467sec.
Without the alternate control scheme the above mentioned
Fig. 8 Time Delay Elimination time delays lead to an uncontrollable, PIO-like situation. The
The ACS can be implemented through the use of a logic block safety pilot had to take the wheel.
placed directly before the actuator in the control loop, as de-
picted in Fig. 9.
f \ Tracking Task !
act_inp log_out act_out
(Pitch Angle) \
Logic Actuator \

t
\ /
Fig. 9 Smart Actuator
The actuator command from the control system is fed into the
logic block. The logic then determines whether ACS is re-
quired based on information about the command and the ac-
tual actuator output. When ACS is to be activated, the logic Fig. 11 SCARLET Flight Test Results with ACS
block provides an output signal which serves as a new actuator
input. This modified input is calculated to produce the desired With the alternate control scheme no time delays and therefore
actuator output. If ACS is not required, then the logic block no pilot induced oscillations occur (Fig. 11). Even though the
simply passes on the normal actuator command as input to the pilot has some problems to fulfill his task because of the ex-
actuator. treme rate limit, he is permanently in control of the aircraft.

This process is clearly illustrated in Fig. 8. In the first time 4.2 Reconfiguration
period the command demands more than the actuator can Control system reconfiguration becomes necessary with the
achieve, such that as before a discrepancy develops between occurrence of failures or damages in the aircraft actuator sys-
the magnitudes of the signals. However, when the command tem (elevator, aileron, rudder, ...). The control system has to
changes direction, ACS switches on and produces the output be adapted in order to have intact actuators take over the tasks
shown, which serves as the new actuator input. The actuator of the impaired ones (Fig. 12). The inherent redundancy in
now strives to follow the logic output and thus changes direc- most modern actuator systems is the main prerequisite for
tion immediately, and therefore the time delay disappears. The such a substitution. Damage of the elevator for example can be
basic design philosophy is that when conditions are right for compensated at least partly by the stabilizer; rudder and
the occurrence of the time delay (i.e. the actuator is saturated), asymmetrical thrust both produce a yawing moment. If there is
then the normal control system structure is bypassed, and the no possibility to fulfill the pilot's commands because of the
information about the command reversal is passed directly to

Fig. 12 FAST (Fault Accommodation by Structure Tuning)

31-6

Side-
Elevator f stick Elevator
// * f Contr ^ // *
ATTAS Pilot recon- ATTAS
) ^ figures j ^. .-
Trim
button Stabilizer Stabilizer

Altitude Profile Altitude Profile

without with
Reconfiguration Reconfiguration
600 sec 1000 0 200 400 600 sec 1000

Fig. 13 FAST Flight Test Results

degradation, the pilot's demands are translated into suitable command the aircraft. The reconfigured control system trans-
commands that can be realized. forms his command, that had formerly been sent to the intact
elevator, into equivalent commands for the stabilizer. By this,
4.2.1 FAST the reconfiguration enables the pilot to fulfill his task much
more precisely.(Fig. 13 right).
FAST (Fault Accommodation by Structure Tuning) is a new
control system reconfiguration concept for a contemporary fly- 4.2.2 GRACE
by-wire aircraft. DLR's in-flight simulator ATTAS is used as
test bed for the validation of the concept. Negative effects of actuator faults on the operation of a flight
control system unable to be taken into account by fault identi-
The controller consists of a feedforward and a feedback part fication and knowledge based reconfiguration can be mini-
(Fig. 12). The reconfiguration of the feedforward block repre- mized during an adaptation period. The analyzed flight control
sents the main task. Besides knowledge based reconfiguration system contains an explicit model following control structure
for that part of the fault that can be quantitatively gauged, with model, feedforward, feedback, and the basic aircraft
automatic control system parameter adaptation by means of process (Fig. 14).
robust genetic algorithms (see chapter 4.2.2) can take care of
those fault impacts that have not been numerically determined.
The fault is supposed to be known. It has to be determined by
a fault detection routine. FAST itself consists of two succes-
sive steps. At first the qualitative and quantitative information
GRACE
i i
J
about the fault is used to restructure the control system. This 1
'
first part of the reconfiguration has to be done immediately
Feed-
after the fault information is available, in order to ensure con- Model
forward i
tinuous stability and maneuverability of the aircraft. The sec-
ond reconfiguration task can then be performed in a much
broader schedule: An automatic parameter adaptation routine
can be applied to take care of those impacts of the fault that ( Feeäd-
could not be quantitatively gauged. Such an adaptation basi- * ba ck
cally involves an optimization to alter the parameters of the
control system via a certain strategy within predefined limits Fig. 14 GRACE (Genetic Reconfiguration Algorithm for
in order to maximize control performance. Since in most cases Control System Enhancement)
this optimization is only a "nice to have"-feature, the conver- In order not to end up with unnecessary stability problems,
gence speed of this procedure is not as important as in the case only the feedforward parameters are adjusted by the adaptation
of the preceding immediate knowledge based reconfiguration. algorithm. The optimization technique used for the parameter
Robustness, which in this context is the ability of the optimi- adaptation is a genetic algorithm applying the principles of
zation algorithm to find and keep the global optimum, is of natural evolution, i.e.
much more concern.
• Selection,
Fig. 13 shows the results of two ATTAS flight experiments, • Recombination, and
during which the elevator was disengaged. In the first experi- • Mutation,
ment the pilot was asked to use the stabilizer trim button in-
stead of the longitudinal stick to cover a predefined altitude to technical problems. Such an algorithm produces near-
profile. He encounters severe problems stabilizing the phugoid optimal solutions very reliably and was successfully tested
mode with the slow stabilizer (Fig. 13 left). both in simulation and flight experiment. The result of the
conducted flight experiment demonstrates a significant reduc-
After a control system reconfiguration in the second experi- tion of the control error after adjusting the feedforward pa-
ment the pilot can use the stick and the old control law to
 31-7

Model Feedforward Uff u

Pilot —»0— Aircraft
Aircraft Controller

Feedback
Controller
Ax
-+Ö*-
Fitj. 15 IFS Model Following Control Concept

rameters caused by an unpredicted elevator effectiveness fail- tem by an individual interface program for the control of the
ure of fifty percent. different experimental functions [9]. Numerical codes and
supporting software elements were developed to generate
4.3 A3XX In-Flight Simulation those specific control law input signals required by the flight
The in-flight simulation available on ATTAS is based on an control laws and not provided by the comprehensive ATTAS
explicit model following control system (Fig. 15). The first of standard sensor and measurement system (Fig. 17). The final
two essential parts of this system is a feedforward controller, adjustment of the flight control laws, the adaptation to the
which represents the inverse dynamics of the host a/c. The ATTAS real world system, the experimental check out, and
second part is a feedback controller with proportional and in- the pilot's flight test training were performed using the ground
tegral elements. based ATTAS system simulator [5].
The most recent application of this system has been the air- | Supporting Software
borne simulation of the Airbus A3XX transport a/c, which is (Experiment ATTAS Experimental System
Dependent)
presently under development. The model is based on prelimi-
nary data, is unaugmented and has no elastical degrees of
freedom. A typical flight-test result is illustrated in Fig. 16. Pilot
SUB1

Experiment
^S^BÜsBiää» Adaptation
. Roll Command

^fflöM W1^
I s
Flight Control Real
Laws Environ-
(FCL) ment
Contractor
Data Recording
Software

Fig. 17 SAFIR Flight Test Concept on ATTAS

The flight control laws include normal laws (full controlled
200
a/c), direct laws (direct links between cockpit control devices
300 sec
and control surfaces), and flight envelope protections which
Fig. 16 A3XX In-Flight Simulation were investigated during the flight tests:
The pilot had the task to perform a turn reversal. With his ai- • Pitch axis
leron inputs he excited the Dutch roll motion. The sidestick • nz-law
commands illustrate, that at first he had problems to perform • Turn compensation
the task because of the oscillation. The match of the flight • Roll axis
states (bank angle and angle of sideslip) of model and ATTAS • Roll rate command bank angle hold law
indicate a sufficient performance of the model following sys- • Yaw axis
tem. • Rudder command
• Turn coordination
4.4 SAFIR • Yaw damper
Within the SAFIR project (Small Airliner Flight-Control- • Compensation of non-symmetrical thrust
Laws Investigation and Refinement) flight control laws for a • Protections
future 100 seater transport type aircraft have been investigated • Pitch attitude protection
which were designed by the Daimler-Benz Aerospace Airbus. • Bank angle protection
The aim of this contract work was the validation, demonstra- • High angle of attack and high speed protection
tion and valuation of the flight control laws by a flight test • nz protection
program performed on the DLR testbed ATTAS. Especially
flying qualities and flight envelope protections were analyzed. A typical flight test result is given in Fig. 18. It shows the time
histories of a left hand turn flight maneuver. Roll normal law
During the first flight test campaign (SAFIR 1) the flight con- is a roll rate command/bank angle hold law. The roll rate
trol law package of the Daimler-Benz Aerospace Airbus was command is proportional to side stick deflection and limited to
implemented on the Experimental and Control Computer of Ipl < 157s. Bank angle hold is provided up to l<|>l = 33° with
ATTAS. It was coupled with the ATTAS data processing sys- automatic turn compensation. This allows turns to be flown
31-8

without manual pitch inputs. For bank angles l<|)l > 33° spiral Cl<pl < 33°) was generally rated as Level 1. Leaving the flight
stability is introduced. The maximum bank angle in ATTAS envelope of normal operation (l<])l > 33°) the design of the
flight tests was limited to k|>l < 45° (k|>l = 30° in AOA Protecti- FCL's demand more pilot activity to increase his situation
on and High Speed Protection). Turn compensation is reduced awareness. This intended effect causes the degradation of pilot
in accordance with the bank angle so that pilot action for pitch ratings.
control is required [10]. The effect of the bank angle protecti-
on is illustrated in Fig. 18. The defined maximum bank angle
Altitude
is not exceeded although full side stick deflection is comman-
ded. After stick release to neutral position the bank angle of
l(|)l = 33° is established.

Sidestick 1-5
Maximum Distance x
Roll Input
Deflection
-.5
6 Distance y
90 deg turn © Descent with
with bank angle 90 deg turn and
Aileron deg <|> = 45 deg bank angle of
-4 <)> = 45 deg
©■
20 Precise
Bank --© Precise
level flight
Attitude deg \V ^y
level flight
-60 ©--
0 10 20 sec 30 Climb with
90 deg turn ® Level flight
Fig. 18 Bank Angle Protection (Flight Test Results) x
90 deg turn
The pitch normal law is a load factor law with automatic trim
function. At low speed load factor is blended with pitch rate.
r
A load factor limitation allows adequate maneuvering without ®
risk of structural overload. If a predefined angle of attack is
Distance x
exceeded (Oprot), an angle of attack command law becomes
active, protecting the aircraft against stall [10]. Fig. 19 gives Fig. 20 SAFIR Maneuver
the flight test results of the angle of attack protection which
Effort 10 ...Q... Heading
prevents the aircraft from pitching to excessive high angles-of- Rating
attack (a < a^) even for full aft stick deflection. After stick . . -X- • - Airspeed (<> < 33')

release to neutral position the angle of attack is lowered and A Airspeed (rj) > 33°)

an,,, is established. 6+

4
Sidestick -5 ..q, • 'P----D,
Pitch Input v
—^FuH Pitch upl . * X / X- - '
/
2 \ t>-
-1.5!
-+- -t-
2.0 8 10
Elevator deg Maneuver No.

-.5 Fig. 21 Effort Ratings for Heading and Airspeed Control

Angle of . Effort 10 .--0- --Climb Rate ($< 33°)
Rating
Attack de9 ■ -A. . - Sink Rate ($ > 33°)
6 X Total Task

Fig. 19 Angle of Attack Protection (Flight Test Results)

Pilots judged the in-flight results to be consistent with the
simulator results. The transition from a-command to nz-
command mode was found to be very sensitive to pitch rate
and load factor cues and can therefore only be optimized by 2 4 6 8 10
means of flight testing [10]. Maneuver No.
For the handling qualities investigations a simple but powerful Fig. 22 Effort Ratings for Vertical Speed Control
flight maneuver was designed containing seven different seg- Right now the SAFIR 2 project is successfully running. This
ments (Fig. 20). From this SAFIR maneuver the pilots derived flight test program is directed to the investigation of system
COOPER-HARPER ratings and general comments in a ques- aspects such as
tionnaire.
• Dissimilar software,
The outcome of results for the handling qualities is illustrated • Data consolidation, and
by Fig. 21 and Fig. 22. Both figures show the effect of the roll • FCL degradation (direct laws).
rate command/bank angle hold control law with full turn com-
To carry out these investigations a flight-worthy HARRIS
pensation (for turns with bank angles of l<|>l < 33°) compared Night Hawk 4401 computer has been connected to the ATTAS
with turns where pilot action for bank and pitch control is re- EDP-System (Fig. 23). The computer calculates the flight
quired (for bank angles of k|>l > 33°). Although for this princi- control laws quasi-parallel using two different codes
ple study of SAFIR the control laws were adapted but not op- (FORTRAN, ADA).
timized the quality of the FCL's for normal operation
 31-9

ARINC-429-lnterface
- 6 Highspeed Channels
mand were defined for the vertical and pedal axes. The re-
'mm : 3 ZKR -> FCLC
3 FCLC -> ZKR
FCL-Software
• FORTRAN
sponses to the control inputs were fully decoupled, except for
-ADA an altitude hold term. The response characteristics were varied
ZKR FCLC
in bandwidth and time delay by a variation of the primary ex-
Chanal 1: 20 Data (Pilot Inputs, Sensor Signals)
Label 000 to 077 perimental variables roll damping and time delay for the RC
EXPO-Data Chanal 2: 13 Data (Sensor Signals)
fK — i-.C. Chanal 3: 16 Data (Sensor Signals, calc. Parameters) <
Label 100 to 177
Label 200 to 277
and natural frequency and time delay for the AC (damping
ratio was held constant at 0.7). With these variables, band-
EXPI-Data Chanal 4: 24 Data

FCLC-» ERR (Control Commands, Display, Status Signals)

Label 000 to 077 width could be varied between 1.45 rad/sec and 4.3 rad/sec.
Starting with the ATTHeS inherent equivalent time delay in
Chanal 5: 40 Data for PCM-Recordinq Label 100 to 177
PCM roll axis of 110 msec, time delays were added in steps of 40
Chanal 6: 40 Data for PCM-Recordinq
Recording Label 200 to 277
msec up to 160 msec resulting in overall system phase delays
Cycle Time 20 ms between 80 msec and 205 msec.
Fig. 23 Integration of the Flight Control Law Computer Fig, 24 shows the data points from the flight tests in the ADS
format. The data point of the basic BO 105 is added. The test
5 ATTHES FLIGHT TEST RESULTS
results can be summarized:
5.1 Bandwidth Criteria Evaluation
1. A high degree of consistency between the rate and the atti-
The introduction of high authority flight control systems in tude command configurations exists. This confirms not
rotorcraft created the need for new helicopter handling quali- only the independence of the bandwidth and phase delay
ties requirements. In response to this, a new helicopter han- criterion with regard to the response type, but also illus-
dling qualities specification was developed under the leader- trates the appropriateness of the slalom tracking task used
ship of the US Army and published as Aeronautical Design for the investigation and demonstrates the suitability of
Standard 33 (ADS 33) [11]. In a new bandwidth and phase ATTHeS to cover also high bandwidth configurations with
delay criteria format, requirements are specified relating to the good fidelity.
rotorcraft's ability to perform small amplitude and high pilot
gain tasks such as ground and air tracking, slope landing, etc. 2. The shape of the boundaries recommended by the test data
The bandwidth criterion is an application of the pilot cross- are significantly different from the ADS 33 requirements.
over model. It provides a measure of the maximum frequency The phase delays are limited for Level 1 at about 0.1 msec
a pure gain pilot can achieve without closed loop instability and for Level 2 at about 0.17 msec.
and without any tendency of pilot induced oscillations. Low
Phase
bandwidth of a system indicates the need of pilot equalization delay
during high frequency tasks. The helicopter phase delay de-
scribes how quickly the phase lag increases beyond the neutral 0.3
/' I evel 2 /' Level 1
stability frequency. When the pilot has to control above the ' ADS 33 / ADS 33 O Rate Command
neutral stability frequency, a significant amount of pilot lead A Attitude Command
compensation is required which increases the pilot's workload J
@ A Level 1
and deteriorates the handling qualities evaluation. For the de- 0.2 ©A Level 2
sign of actively controlled helicopters, bandwidth and phase i
• ▲ Level 3
delay are two important design parameters. Bandwidth is a
basic parameter for control system design. Phase delay can be
directly related to the equivalent time delay of a system and is ©; ©©
composed of the contributions of the basic helicopter and of 0.1 obAo£0A
portions produced by a control system. < €J @C (~@^&-^& M.

Since its introduction, research has been conducted to extend Level 3 Level 2 Le v e 1
L, ! Basic Bo 105
the data base on which the ADS 33 criteria are based on. A DLR DLR DLR
1 1 i i i i
review of the data base used for the definition of the ADS 33
rad/s 6 Bandwidth
bandwidth requirements showed that these were primarily ob-
tained from ground-based simulations and flight tests with low Fig. 24 Flight Test Data of the Bandwidth Study
bandwidth variable stability rotorcraft. The ADS requirements In the updated ADS 33D version, these modified requirements
allow increasing phase delays with increasing bandwidth. Sin- are accepted as roll axis requirements for target acquisition
gular data points achieved in previous tests with an early con- and tracking tasks in hover and air combat tasks in forward
trol system design for BO 105 ATTHeS with equivalent time flight.
delays of more than 200 msec showed significant discrepan-
cies with the ADS requirements. Although the data points 5.2 Rotor Decoupling Study
were clearly within the ADS Level 1 limits the pilots rated
Typical high agility, high bandwidth helicopters, like the
poor handling qualities in high gain piloting tasks. Recogniz-
BO 105, exhibit severe pitch-roll coupling. This coupling is
ing these discrepancies an activity was started to extend the
produced by the stiff rotor system and the large hinge offsets
data base and to verify the ADS level boundaries for the roll
of the blades which are required to generate the large rotor
axis in forward flight.
moments especially needed for the agility and the responsive-
The variable stability BO 105 ATTHeS was used to investi- ness in high gain piloting tasks. High authority control sys-
gate the effects of varied bandwidth and phase delay for a high tems can provide an active decoupling of the helicopter to re-
gain slalom tracking task [12]. The capabilities of ATTHeS duce the pilot workload to an adequate level. Whether a con-
achieved by the high bandwidth characteristics of the basic trol system should be designed to eliminate all coupling is
BO 105 helicopter and the specifically designed explicit questionable and will be balanced by the operational needs,
model following control system were well suited for these the technological effort for the control system, and the costs.
tests. Two response characteristics were implemented: a first In addition, an augmented helicopter must be controllable af-
order rate command (RC) and a second order attitude com- ter a control system failure, too.
mand (AC) system. These command modes were defined for
roll and pitch axes. Rate of climb response and sideslip com-
31-10

In ADS 33 the maximum allowable levels of pitch-roll cou- tive coupling amplitudes at the bandwidth frequencies of the
pling are specified in a time domain format. The criteria de- compensatory axis are crossplotted.
fines ratios of peak coupling responses to peak on-axis re-
sponses to given cyclic inputs which are 9pk/(|) for pitch-due-
to-roll and <t>pk/0 for roll-due-to-pitch. By defining a 4 sec
B Control
segment for the measurement of the ratios, the criterion pri-
marily considers the mid to long term effects. However, in a 0 ORate
A Combined
high gain pilot task the mid to long term coupling responses
can be more easily compensated by the pilots and therefore © Washed out
dB
may secondarily influence the pilot workload. Whereas the
suppression of the short term coupling response will drasti-
cally increase the pilot workload. Furthermore, the short term -10
effects are particularly important for the control system design #>
and will lead the effort in the design. Long term coupling can
be easily eliminated by a simple feedback control system.
In cooperation with the US Army a research program was -20 Level 2
conducted to establish a more comprehensive crosscoupling DLR
data base. The in-flight simulator ATTHeS was used to inves- Level 1
tigate the effects of different types of coupling with a strong DLR
emphasis on the inclusion of coupling types which are repre-
-30 O ##*
sentative for highly augmented helicopters [13]. The imple-
mented model consisted of two parts: An uncoupled baseline -30 -20 dB
model which was evaluated as Level 1 and a pitch-roll cross-
coupling model. The baseline model was a first order rate
command with a roll axis bandwidth and phase delay of Fig. 25 Flight Test Data of the Coupling Study
3.44 rad/sec and 77 msec, respectively and with a pitch axis
bandwidth and phase delay of 2 rad/sec and 114 msec, respec- A two sided format allows to consider the effects of the indi-
tively. Three different types of crosscoupling were realized vidual and combined roll-to-pitch and pitch-to-roll couplings
and varied in magnitude and dynamics: in one diagram. An additional benefit of the frequency domain
criteria is that the required data can be obtained from the
1. Control coupling bandwidth tests and analysis, thus eliminating the need for
2. Rate coupling additional compliance test data.
3. Washed-out coupling
Control and rate coupling represent the coupling behavior of 5.3 Lynx In-FIight Simulation
an unaugmented helicopter. In an actively controlled helicop- For the purpose of a rotorcraft simulation in flight a special
ter, any off-axis responses from control or rate coupling will simulation model to be followed has been developed for AT-
be reduced to zero by a feedback system which can be charac- THeS. This real-time command model consists of a linear 9
terized by a washed-out type response. DOF model which is extended with nonlinear terms for an
An intensive analysis of the pilot control strategy to compen- improved simulation fidelity [14]. Either from flight test data
sate the coupling responses gained the fundamental under- or from runs of a generic simulation program the model dy-
standing of how coupling affects the pilot workload. In the namic matrix A and the model control matrix B are calculated
tracking phases of the roll axis slalom task the pilots used a using system identification procedures [15].
feedback control strategy. While acting as a feedback system, The necessary nonlinear terms for coordinated turns, gravity,
the pilots primarily controlled the on-axis roll task and used changes of flight path, and the Euler equations are pro-
spare capacity to remove the unwanted coupling. As coupling grammed explicitly. In addition, a four axes stability and con-
increased more attention was needed for the off-axis control. trol augmentation system (SCAS) for the simulated helicopter
This resulted in poorer task performance, lower lateral input can be incorporated. The SCAS can be engaged or disengaged
power, and poorer handling qualities ratings. With increasing during flight with software switches. For the investigation of
coupling response in the pitch axis, the pilots increased the SCAS failures, several failure situations with changeable deg-
longitudinal input power and used higher frequencies in the radation and reconfiguration blending can be simulated. When
longitudinal control. For severe coupling cases, longitudinal adequate models of a helicopter are available, the helicopter
input activity shifted to the bandwidth and neutral stability simulation with ATTHeS can be used in the approach to de-
frequency of pitch axis. For some of the most severe coupling velop and adapt a control system to the helicopter dynamics.
cases, input activity above the neutral stability frequency was The evaluation of software and hardware components in the
observed resulting in mild pilot induced pitch oscillations. integrated system, with the pilot-in-the-loop, and in a real
This frequency dependent feedback strategy accounts for the flight environment will support the design approach and will
inconsistencies between the control/rate and the washed-out drastically reduce the risk of a system development.
coupling when using a criterion format in the time domain.
The ability of the pilot to suppress coupling will depend The in-flight simulation of the Lynx helicopter serves as an
largely on the system capabilities of the axis he has to use to example to describe the simulation capability. The Lynx heli-
compensate coupling. Parameters which are a function of the copter has some differences in the response characteristics
frequency characteristics of the compensatory axis yield an compared to the basic BO 105, for example some couplings
adequate correlation with the pilot ratings. In Fig. 25 the rela- are opposite to the corresponding couplings of the BO 105.
 31-11

Longitudina bU
Control %
30
Collective 50
Control %
10
Lateral 45
Control %
25
Forward 50
Speed m/s
20
Pitch .2
Attitude rad
-.2
Roll .2
Attitude rad
-.2
160 sec 180

Fig. 26 Simulation of Lynx Helicopter Dynamics

Fig. 26 demonstrates the simulation fidelity of the in-flight loops were preoptimized in a non-real-time simulation. Before
simulator ATTHeS. The left diagram provides flight test data performing any flight tests, the position hold hardware and
of the simulated basic Lynx without a SCAS in a speed varia- software had to be integrated and evaluated in the real-time
tion maneuver. The time histories in the right diagram show ATTHeS system-simulation, which is used as the standard
test data with a SCAS specifically designed for the simulated pre-flight test procedure.
helicopter. Experienced Lynx pilots who have flown the tests In the flight tests the position target was presented by a black
commented the compatibility with the operational Lynx heli- square which was mounted on the top of a car. The task of the
copter as very satisfactory. test pilot was to fly the helicopter, which had implemented a
decoupled rate command attitude hold control mode, above
5.4 Vision Based Hovering Experiment the target. When the target was within the range of the video
The demands to extend the helicopter mission profiles, i.e. camera the position hold mode was engaged by the pilot
that helicopters can be flown at night and in adverse weather, pressing a button on the MFCS keyboard. When the pilot
drives the need to design advanced control systems. Especially moved the cyclic control the position hold was automatically
in hover and low speed where the helicopter is flown close to disengaged. The position hold control system had to fly the
the ground, augmentation for the pilot is required to reduce helicopter above the target in constant altitude and with con-
the pilot's workload to an acceptable level. A special task in stant heading while the pilot had the hands off the controls.
various mission phases is the position hold under gusty condi- When the target was ground fixed the achieved position accu-
tions over a ground fixed or moving target which could be a racy was about 1 m. To simulate a moving target, the car then
shipboard or lifeboat reference. drove a circle (40 m radius) with 15 km/h. With the moving
target, the position errors deteriorated up to 2.5 m.
DLR has equipped ATTHeS with an innovative system for the
measurement of a reference hover position. A video camera
together with a sophisticated computer system for processing
the optical information was used as an integrated target scan-
ning sensor. Based on the existing model following control
system of ATTHeS, the control system was modified by add-
ing a feedback control structure for minimizing position errors
[16]. For a helicopter in hover, the longitudinal and lateral
accelerations can either be controlled by changing the pitch Heading ^Au^AAjVA~^yA^y^\yv-^V\y/'\s^^y^^
and roll attitudes or for short term corrections directly by the
sideforce capability of the rotor system. The helicopter posi- Altitude
tion in relation to a ground target is a second order differential
equation with a nonlinear function in the attitudes. The model
Status MFCS active : :
j Position Hold active
following control system in hover highly emphasized the
stabilization of the attitude loops. For the position hold task, ■■ ■ ■ ■ ' ■"

30 sec 60
additional terms in the attitude loops were incorporated. These
terms concern the relation between the commanded pitch and Fig. 27 Helicopter States during Autonomous Hover
roll attitudes for the relative position errors and the corre- Fig. 27 shows time histories of a test which was performed in
sponding velocities. The coefficients of the additional control a 15 kts wind condition and with gusts up to 30 kts.
31-12

1300 Cumulative
Flight Hours HQ-Flight Test
1200 • Criteria (Moll)
• ACT-Continuation
Test Pilot Training
1100 ETPS
EPNER

1000 European ACT Side Stick

Programme Integration
900 H • Handling Qualities
• Control Laws
HQ-Flight Test • Inceptors
800 Bandwidth (MoU)
Cross Coupling

700
In-Flight Simulation
• Control Law Optimization
600 • Response

500 4 Axes Maneuver

Demand System
400 Installation of
Variable Controllability Fly-by-Light
• Slalom Yaw Axis System I
300
(Low Airspeed SensorJ:
FBWBO105-S3
200 into Service
at DLR Experimental Equipment
100 Dev/DemA/al

1982 83 84 85 86 87 88 89 90 91 92 93 94 1995
Fig. 28 ATTHeS Flight Test Statistics

5.5 Flight Test Statistics

Fig. 28 overviews the flight test activities with ATTHeS.
The fly-by-wire helicopter BO 105-S3 came into service at
DLR in 1982. Since that time several flight test programs have
been performed demonstrating the multipurpose capability of
the airborne simulator. Beside a continuous improvement of
the simulation hardware, the utilization of ATTHeS can be
summarized:
• Development and improvement of the explicit model fol- Model Verification
lowing control system
• Development of advanced control systems
• Generation of credible handling qualities data
• Test pilot training
On average, about 150 flight test hours a year demonstrate the '-~*Jx&^

effectiveness and applicability of the in-flight simulator AT- Fig. 29 Modeling of a Dynamic System
THeS. Fig. 15 illustrated two essential sources, which can be respon-
sible for deficiencies concerning this particular simulation
6 ASPECTS OF SIMULATION FIDELITY
technique:
During the last decades the ground-based a/c simulation has
become a well-known and accepted technique. It is not only in • A/C model to be simulated: Fig. 29 shows the elements,
use for training purposes but also for scientific applications. which have to be taken into account developing an a/c
The already mentioned in-flight simulation is as well known, model:
but because of its costs it can only be found in a few centers in • The real system: The requirement is to reproduce par-
Germany, Russia and the USA. However, both techniques ticular characteristics of this system. These can be the
have to deal with the problem that the computer based simula- dynamic characteristics as well as functional aspects or
tion of dynamic systems has inherent deficiencies. First the in- whatever the aim of the simulation is.
flight simulation: • The numerical model: It contains the mathematical de-
scription of the system to be simulated. The parameters
and the order of the equations are the result of an inves-
tigation with the real system as central element.
 31-13

• The system model in the computer: The basis of this error curve, which will be shown in this section (see Fig. 33).
system is the algorithm of the numerical model. It has to State of the art software systems allow a quantitative analysis
be transferred to a higher programming language. The of the model following results. It can be based for instance on
combination with a database describing the real system statistical algorithms which are dependent on the time or the
allows it to imitate the dynamic of the real system in a frequency. Also very particular evaluations are possible, based
given quality. on quality criteria related to the given technical problem. The
• Model following system: The feedforward controller of the following two subsections deal with two examples for analyti-
model following system is based on a flightmechanical cal evaluations.
database of the host a/c. Adequate model following can The first approach is defined in the frequency domain. It gives
only be achieved if the database represents all dynamic information about aspects of the dynamic performance and the
characteristics of the respective flying testbed. short term model following quality. The second method is one
The second example is the motion-based flight simulator. [21] example how model following fidelity can be handled in the
touches some typical problems these systems have: time domain. The quality in the long-term range and for in-
• In the case of a fixed-based simulator there are no pro- stance drift effects can be evaluated.
prioceptive cues. The experience which has been gained working with both ap-
• A motion-system has physical limits and therefore some proaches is, that frequency domain and time domain are both
cues are more or less suppressed (e.g. only 10-15% of the necessary to get information about overall fidelity aspects.
real roll acceleration p are available).
• Some cues, such as the load factor n2, are missing. 6.1.2 Frequency Domain
• Because of washout filtering, some cues are generated It is essential for every in-flight simulation, that the host a/c
which never appear in a real airplane. The design of wash- covers the spectrum of the model system dynamics in a given
out filters is still a kind of black art. range. Requirements in the frequency domain provide the op-
• The harmonization between a/c motion-system and visual- timal background to investigate this aspect.
system dynamics is a problem area.
The existing experience in the field of handling qualities in-
• The workload of the pilot in a simulator and in real flight
vestigations can be used. [22] found out that especially for
is generally different. Investigations concerning PIO-
pilot-in-the-loop situations the total system bandwidth and the
effects, for instance, have shown this.
phase delay are essential. These are the critical parameters for
6.1 Model Following Fidelity Evaluation Methods system stability and handling qualities.

6.1.1 General Remarks Other investigations, e.g. [23], have shown, that added dynam-
ics of about ±2 dB in magnitude and about ±10° in phase at
The evaluation of model following fidelity is usually based on the total system bandwidth frequency are below pilots percep-
qualitative and quantitative aspects. The more qualitative tibility. Based on this result [20] defined a simulation fidelity
method is to plot the flight states of model a/c and host a/c in criterion, which is illustrated in Fig. 30 in principle. The fig-
one diagram. It is then possible to see how well the two curves ure shows the above mentioned boundaries applied to the
match. This optical check can be supported by an additional

1r Aircraft to be
Simulated = Model A/C
Model A/C Model A/C
Model A/C
pilot
Control —► Actuator —► Dynamics
System System

Pilot Input

ATTAS
o ATTAS IFS ATTAS
ATTAS
Simulation Actuator
u - Dynamics *-► x
pilot & Control System
System

20
o 180

M
dB
^> >» 1 ■«_
- + 2cIB-
Phase c%w = Frequency at
-135 deg phase angle
of model open loop
frequency response
0
\
-10
*— «■"

1 Openloop = Loop not

-20
0. 1
II
1 0 racJ/s 100 100
closed by
the pilot

Frequency

Fig. 30 Simulation Fidelity Requirements in the Frequency Domain

31-14

transfer function of simulator response x versus the response a/c) and the output of the system to be controlled (host a/c)
xm of the simulated a/c in the sense of simulation fidelity. match. This way is supported by the following approach
which is called "delta rating" (AR). It is defined by the ratio:
The simulation fidelity transfer function x/xm comprises all
elements of the model following system given in Fig. 15. The Model Following Error
function x/xm is nearly identical to the ATTAS actuation sys- Delta Rating
tem dynamics and the computational time delay caused by the Desired Signal
onboard computer system. The reason is based on the fact that The mathematical description is based on the following as-
the ATTAS-inverse and the ATTAS-dynamics cancel each sumptions and equations:
other up to relatively high bandwidth. Actuation system dy-
namics and computational time delay are the primary parame- • The states of model and host are given in a time interval
ters which are responsible for dynamic in-flight simulation [ti, t2].
fidelity and performance. • The reference states in this time interval can be determin-
ed.
Any a/c which has to be simulated contains high order dynam-
ics produced by actuator dynamics and control system delays. It is now essential for the idea of the delta rating, that the de-
The aircraft to be simulated can therefore be described by an viations from the given reference state of the model a/c are
ideal a/c dynamics and an equivalent time delay representing evaluated. These deviations are integrated using the following
high order dynamics. In order to avoid additional time delays equation:
on the in-flight simulator the model implemented in the 12

ATTAS onboard computer system contains only the ideal (2) X(t) = j|x(t)-xoffsc,|dt
model characteristics with zero time delay because ATTAS'
inherent time delays cannot be reduced. As a result the simu- In the case of a flight state, which has to be integrated with
lation fidelity defined by the allowed phase lag as a function equation (2), xoffse( is the constant reference flight state of the
of frequency is given by the difference of high order dynamics current flight segment. Its value has to be found out in ad-
of ATTAS and the model. This conclusion can also be under- vance.
stood in a simplified way as the difference of the equivalent
time delays (phase delays, respectively). As long as the model The model following error (Eq. (1)) is integrated in the follow-
phase delay is smaller than the phase delay of ATTAS a phase ing way:
deviation exists. In the case that both phase delays are equal
(zero phase deviation) an optimal simulation is given. For (3) AX(t) = ]|Ax(t)|dt
model time delays greater than that of ATTAS, the additional
delay has to be taken into account in the model description. With the integrals (2) and (3) a mathematical interpretation of
the delta rating can be given:
6.1.3 Time Domain
- AX(t)
Discussing the fidelity of in-flight simulation in the time do- K(4) ARR = —±t
main usually confines to one central aspect. It is the evaluation X(t)
how well the curves describing model state and host a/c re-
The simple nature of AR provides a direct impression of the
sponse match. This more quantitative method based on an op-
tical check can be supported by an additional error curve model following quality. Two typical values can be distingu-
given in Eq. (1): ished in an exemplary way:
(1) Ax(t) = x„(t)-xM(t) XR = 0 : The two curves, which have to be compared, are
identical.
The result of Eq. (1), the error Ax(t), can be used for statistical
XR = 1: The plane under the error curve is equal to the
data evaluation. The standard deviation s4X of the error can be
determined. plane under the curve of the desired signal. This
would be an example of a bad ratio.
At the beginning of this section, we described how the engi- Fig. 31 illustrates an example with the help of three sinusoidal
neer usually evaluates the result of a model following control: functions. The constants are selected as follows:
He tries to find out how well the signal generator input (model
Sinusoidal Function Sinusoidal Function
— a + sin(x)

f^\ a + b sin(x)

\J
>

■ A-Rating: 0.2

Error

i . i . -i 1 1—
x in radian 12 9 x in radian 12

Fig. 31 Examples of Delta Rating Applications

 31-15

a = 1-5 = x„lrscl rating can also be applied to the flight-test result given in Fig.
33. The rating for the model following quality for the pitch
b = 1.2
angle is: ARe =0.37 .
c=2
[18] describes this method in greater detail. Pitch Angle in deg ( Model, - Host)

6.2 Applications of the Evaluation Methods

,.v, .^rV^v-'
6.2.1 In-Flight Simulation of a Spaceplane
Error in Pitch Angle in deg

Amplitude
-~~~~*-^V^fy^ ■-WAä/«*>™A|^

Fig. 33 Pitch Angles during an ILS-approach

Phase
deg It can be seen that the standard deviation as well as the delta
0 rating indicate a good model following for the pitch angle.
~~~~^_^J~~ . ^
The Euler angles play a dominating part in the in-flight simu-
-45 Rollrate Ratio lation, because they are controlled directly by the pilot. They
P ATTAS
^^^
P HERMES
are modeled adequately in this case and not only in the longi-
i 1 t 1 1 .1'

10 rad/s 20
tudinal motion, which is presented here, but also in the lateral
Frequency to motion (see e.g. [18]). During the last five years, numerous
flight-test hours with about fifteen experienced test pilots have
Fig. 32 HERMES Dynamic Simulation Fidelity been performed. They all gave this in-flight simulation only
Fig. 32 shows the application of the criterion described in good or excellent ratings.
section 6.1.2 to the in-flight simulation of the European
spaceplane HERMES (the HERMES project was canceled 6.2.3 Effects of Unmodeled Dynamics
meanwhile). It can be seen that the simulation fidelity require-
A criterion for the acceptance of the simulation by pilots has
ment is completely fulfilled by the ATTAS in-flight simulation
been derived by [17] for fixed wing aircraft. Here, the fre-
system.
quency responses of measured simulator rates are compared to
6.2.2 In-Flight Simulation of a Widebody Transport A/C the commanded model rates. Ideally, the frequency response
should have a magnitude relationship of one and a phase angle
Fig. 33 shows the pitch angles of model a/c and host, which of zero for a broad frequency range. For practical use, bound-
were recorded during a standard ILS-approach simulated in aries of so called unnoticeable dynamics were defined, where
flight. The simulated a/c was a nonlinear model of a wide- pilots will not notice simulation deficiencies. Using this crite-
body transport a/c (two engines, 115 tons). rion also as a quality assessment for the ATTHeS simulation,
Fig. 34 compares the frequency responses for measured roll
The testpilot performed the landing approach as follows:
rate to model roll rate for two different cases:
Flying straight at a level of about 820 m (2690 ft) the pilot • An identified conventional 6 DOF rigid body model was
captured the localizer of the particular airport. At 100 sec the used to define the flight control system
pilot reached the glidepath of the ILS. He changed the confi- • An identified extended model with rotor DOF was applied
guration of the model and the host a/c to landing configura-
for the flight control system design.
tion. The landing gear was extended and the simulated a/c as
well as ATTAS began the descent. The standard deviation for It is clearly seen, that only the ATTHeS frequency response
the error curve in this example is: S4Q = 0.344°. based on the extended model lies within the boundaries. As
also good pilot ratings for the simulation where given, it con-
This error is not perceptible by the evaluation pilot. The delta firms the definition of the boundaries and seems to indicate

Frequency
Modeled Rotor Dynamics
Unnoticable Dynamic Effects
Unmodeled Rotor Dynamics

Fig. 34 Effect of Unmodeled Dynamics in the Frequency Domain (ATTHeS)

31-16

that they can be used as a more general quality criterion for Aerodynamics and Aeroacustics of Rotorcraft, Berlin,
high bandwidth flight control systems. Germany, October 1994.
[20] Hanke, D., et al., "In-Flight Simulator ATTAS - System
7 REFERENCES Design and Capabilities", CP DGLR-91-02, 1991
[21] Harper, R. H., "The Evolution of In-Flight Simulation at
[I] Gmelin, B., et al., "Mission Oriented Flying Qualities
Calspan", CP DGLR-91-05, Paper 91-05-01, 1991
Criteria for Helicopter Design via In-Flight Simulation",
[22] Hoh, R. H., et al., "Bandwidth - A Criterion for Highly
AGARD-CP-423, Paper No. 4, 1986.
Augmented Airplanes", AIAA Paper 81-1890, 1981
[2] Von Grünhagen, W., et al., "A High Bandwidth Control
[23] Wood, J. R. and Hodgkinson, J., "Definition of Accept-
System for a Helicopter In-Flight Simulator", Journal of
able Levels of Mismatch for Equivalent Systems of
Control, London - Washington, DC, ISSN 0020-7179,
Augmented Aircraft", MDC Rept. A 6792, December
1994.
[3] Bauschat, M. and Lange, H.-H., "ATTAS and its Con- 1980
tributions to System Design and In-Flight Simulation", 8 ACKNOWLEDGEMENT
International Conference Aircraft Flight Safety, Confer-
This paper is dedicated to the late BO 105 ATTHeS flight test
ence Proceedings, pp. 403-413, 1993.
pilot Klaus Sanders and his companion mechanic Hans-Jürgen
[4] Lange, H.-H., et al., "ATTAS Flight Testing Experi-
Zimmer, who decisively contributed to all success of this
ences", AGARD-CP-452, Paper No. 2, 1988.
unique flight test vehicle.
[5] Hanke, D., et al., "The Role of System Simulation for
the Development and Qualification of ATTAS",
AGARD-CP-513, Paper No. 26, 1991.
[6] Bensberg, S. and Saager, P., "Interfacing AD10 and AD
100 Systems via VME-Bus with Special Flight Vehicle
Components", ADIUS90-Europe, 16.-18. October, Celle
(Germany), 1990.
[7] McRuer, D., "Human Dynamics and Pilot-Induced Os-
cillations", Minta Martin Lecture, Massachusetts Insti-
tute of Technology, December 1992.
[8] Buchholz, J. J., "Time delay induced by Control Surface
Rate Saturation", Zeitschrift für Flugwissenschaften und
Weltraumforschung, 17, 1993, pp. 287-293.
[9] Anon, "Statusbericht 1989-1993", Institut für Flug-
mechanik, Deutsche Forschungsanstalt für Luft- und
Raumfahrt, pp. 403-413, 1993.
[10] Luckner, R. and Heintsch, T., "Development of Manual
Flight Control Functions for a Small Transport Air-
craft", ICAS Congress, Congress Proceedings ICAS-94-
7.6.1, Anaheim, 1994.
[II] Anon, "Handling Qualities Requirements for Military
Rotorcraft", Aeronautical Design Standard ADS-33C,
Aug. 1989.
[12] Pausder, H.-J. and Blanken, C. L, "Investigation of the
Effects of Bandwidth and Time Delay on Helicopter
Roll-Axis Handling Qualities", Presented at the 18th
European Rotorcraft Forum, Avignon, France, Sept.
1992.
[13] Blanken, C. L., et al., "Rotorcraft Pitch-Roll Decoupling
Requirements from a Roll Tracking Maneuver", Pre-
sented at the American Helicopter Society 50th Annual
Forum, Washington D. C, May 1994.
[14] Bouwer, G. and von Grünhagen, W., "Lynx helicopter
In-Flight Simulation with ATTHeS", DLR Institute Re-
port, IB 111-92/47, 1992.
[15] Kaletka, J. and von Grünhagen, W., "Identification of
Mathematical Derivative Models for the Design of a
Model Following Control System", Presented at the
American Helicopter Society 45th Annual Forum, Bos-
ton MA, June 1989.
[16] Bouwer, G, et al., "Autonomous Helicopter Hover Po-
sitioning by Optical Tracking", To be published in ZFW
(Zeitschrift für Flugwissenschaften und Welt-
raumforschung.
[17] Anderson, M. R., "Robustness Evaluation of a Flexible
Aircraft Control System", AIAA Journal of Guidance,
Control, and Dynamics, May-June 1993.
[18] Bauschat, J.-M., "The Role of In-Flight Simulation for
the Definition of Simulation Fidelity Criteria",
ICAS/AIAA CP, Volume 1, Paper ICAS-94-8.1.2, pp.
261-271, 1994.
[19] Hamel, P. G. and Kaletka, J., "An Overview of AGARD
FVP Working Group 18", AGARD Conference on
 32-1

Variable Stability In-Flight Simulator Test Aircraft (VISTA)

by
Kurt Buehler
AFFTC/XPX
1 S. Rosamond Blvd.
Edwards AFB, CA 93524-1036

Phil Reynolds
Calspan Corporation
Flight Research Department
150 N. Airport Dr.
Buffalo NY 14225-1473

Steve Markman & Gary Hellmann

WL/FIIA '
Building 450
2645 5th Street, Suite 16
Wright Patterson AFB, OH 45433-7922

ABSTRACT: DEVELOPMENT OF VISTA

The VISTA is the United States (US) of The high-performance in-flight simulator used in the
America's next generation in-flight simulator, replacing past by the USAF was the NT-33A. The NT-33A
the aging NT-33. The VISTA is an F-16D with the delivered in 1951, is the oldest flying aircraft in the
Israeli Peace Marble II configuration. The flight control USAF inventory. It was developed into an in-flight
system has been extensively modified to create a state of simulator in the late 1950's. The NT-33A has simulated
the art in-flight simulator. The VISTA Phase II nearly every new fighter aircraft that has entered the
Development, Test and Evaluation (DT&E) at the Air USAF inventory, plus several aircraft for the Navy.
Force Flight Test Center (AFFTC) started in July 1994 NASA, and allied nations. The NT-33A has also
and was completed in January 1995. The flight test performed research to help develop a Military
team consisted of personnel from the United States Air Specification for flight control and handling qualities.
Force (USAF), Lockheed Fort Worth Company, and As the in-flight simulation demands of new aircraft have
Calspan Corporation. The VISTA flew 62 test sorties become more complex, the NT-33 A no longer
for a total of 138 flight test hours at the AFFTC meeting represented modern high-performance aircraft. The NT-
or exceeding all design specifications. The VISTA is 33A is also no longer logistically supportable based on
owned by Wright Laboratory's (WL) Flight Dynamics cost. Seeing the end of the useful life of the NT-33 A a
Directorate. The Calspan Corporation maintains and contract was awarded to the Calspan Corporation in
operates the VISTA for WL. Using the aerodynamic, 1982 to define the requirements for a replacement.
mass, inertia, and controller characteristics, and the
flight control laws of other aircraft, the VISTA can The F-16D was selected as the best replacement for the
simulate the aircraft "in-flight" determining the flying NT-33A meeting the following requirements: 1) Two-
qualities and handling characteristics. The USAF and seat fighter-type aircraft with the front cockpit to be the
Navy Test Pilot Schools utilize the VISTA in their evaluation cockpit with variable feel controls with
curriculum. The VISTA is available to support the programmable displays and the rear cockpit to be the
USAF, other branches of the US Department of Defense, pilot-in -command (safety pilot); 2) The Variable
commercial aircraft development, National Aeronautics Stability System (VSS) must be capable of all-attitude
and Space Administration (NASA), and foreign military model-following and response feedback simulations; 3)
aircraft development programs. This paper presents a The in-flight simulator must be able to control the six
brief summary of the development of the VISTA, the degree-of -freedom forces and moments to satisfy
hardware and software modifications that were made to mission and simulation fidelity requirements,and; 4) It
create the VISTA, the VISTA Phase II flight test plan must be relatively inexpensive to operate.
and results, and discusses the future direction of in-flight
simulation. The F-16D was modified during production at Lockheed
Fort Worth, TX to produce the VISTA aircraft. The
VISTA modification contract was awarded to Lockheed

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?'
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
32-2

in August 1988. All the VISTA hardware and software

modifications were incorporated under this contract.
The VISTA development approach was dictated by
funding, with the baseline effort consisting of a five
degree-of-freedom capability, hydraulic systems
upgrades, a response feedback variable stability system.
and extensive cockpit modifications. Planned upgrades
will add a model-following capability, extend the
simulation envelope to supersonic, increase the
bandwidth of the integrated servoactuators, add tactile
cueing to the aft seat controllers, provide a variable feel
sidestick and rudder pedals, and add thrust vectoring
capability.

The baseline VISTA development approach included the

DT&E flight test effort which was completed in two The VISTA cockpits are extensively modified from the
phases. Phase I of the DT&E flight test effort was production F-16 cockpit baseline configuration. Major
completed at Lockheed Fort Worth in April 1992. The changes include the addition of several controls and
Phase I DT&E demonstrated that VISTA could displays to both cockpits and the relocations of primary
fly/operate like a F-16 and consisted of five sorties and 5 controls and displays to the aft cockpit. The front
flight test hours. The VSS and safety trips were not cockpit or evaluation cockpit has a movable variable feel
tested. Funding constraints prevented the USAF from centerstick controller in addition to the F-16 sidestick
starting the Phase II DT&E effort immediately after the controller. Both sticks, the rudder pedals and the throttle
Phase I effort was completed. The VISTA was put in interface with the VSS during a simulation. The throttle
flyable storage until it was modified and used as the test in the front cockpit is not mechanically connected to the
bed for the Multi-Axis Thrust-Vectoring (MATV) throttle in the rear cockpit. The aft or safety cockpit has
engine test program at EAFB during the July 93 through all primary controls required for single pilot operations
March 94 time period. The airplane was modified back as well as secondary controls. A HUD monitor, capable
to the original VISTA configuration after the MATV of repeating the HUD information, has been installed on
flight test program was completed. The VISTA Phase II the safety pilot's glareshield.
DT&E flight test program flew 62 sorties, 138 flight test
hours, started July 94, and was completed January 95. VARIABLE STABILITY SYSTEM (VSS)

TEST ITEM DESCRIPTION: The key to the modification of the VISTA is the
installation of the VSS that interfaces with the Digital
The production VISTA F-16 (Type Version IV) includes Flight Control System (DFLCS) allowing the aircraft to
the large dorsal fairing (extending aft from the canopy to fly with a variety of control laws. The VSS utilizes the
the vertical tail) and heavy weight landing gear. This response feedback method for simulation. The VSS
Block 30 configuration is based on the Israeli Air Force provides limited five degree of freedom simulation
Peace Marble II (PMII), Type Version 4K, F-16D capability. The system commands symmetric and
airframe. Unnecessary systems and hardware such as asymmetric horizontal tail movement, symmetric and
the 20mm gun system, ammo drum, radar warning asymmetric flaperon movement rudder, and throttle
system, chaff/flare dispenser, nuclear weapon capability, control. The VSS does not control the leading edge flaps
AMRAAM missile capability, and HUD expanded and speedbrakes. Future expansion for explicit model
envelope gun sight capability were deleted from the following capability is an option. Major airborne
VISTA design. The aircraft is powered by a General components of the VSS include three Rolm Hawk 32
Electric F110-GE-100 engine, and has an APG-68 radar, digital computers with an input/output expansion
and a digital flight control system. The hydraulic system chassis, a microcomputer, sensors and sensor signal
flowrate was increased, the flaperon and horizontal tail conditioners, circuitry interface chassis, and cockpit
actuators modified for increased rate, and modifications controls and displays.
to the electrical and avionic systems were made to the
VISTA. The VISTA was completed by adding the VSS The three Rolm Hawk-32 digital computers control the
equipment and making the cockpit changes required to all-attitude response feedback simulation. The
support the VSS operation. The general arrangement of computers have the capacity to accommodate and
the airframe and the VSS equipment is shown in Figure execute software for feel system characteristics,
1. command feedforward/response feedback mechanization,
simulation/configuration management, graphic
functions, VSS operational envelope limiting and
warning and safety monitoring. A Titan computer
performs the function of controlling the variable feel
 32-3

system on the center stick. The VSS is a single-strand, cleared by analysis to 80% of the allowable design limit
fail safe system, and commands the quad redundant F-16 load resulting in reduced g envelopes. Special
control activation system. Figure 2 shows how the VSS monitoring functions within the DFLCS have been
interfaces with the DFLCS. designed to disengage the VSS when commands from
the VSS to the DFLCS result in loads greater than 80%
of allowable design limit load.

VISTA/F-16 FLIGHT ENVELOPE

ii^ 16° W = 22756 (Wf =1000)
INITIAL VSS ENGAGE/DISENGAGE TRANSIENT
SUPPRESSION APPROACH
F-1GCD

ALTITUDE
(1000 FT.)

'CONTROL LAWS INITIALIZED TO STEADY STATE CONDITIONS

UPON MODE TRANSITION
0.2 0.A 0.6 2.0 2.2
"SUMP CHANGED UPON DISENGAGEMENT
FIGURE 2

VISTA/F-16 LOADINGS
CARRIAGE

The rear cockpit of VISTA is always a F-16 using a "B ; .0.^,0.

normal F-16 side stick controller and production flight \ fWTT^f SPEED

MACH)
a".';
lose) ""'"
1 2 3 4 5 6 7 8 9 "" T
control laws. The front cockpit functions in one of three '■ j?. j8 ,,., 25 1

basic modes of flight control; VSS mode, F-16 ...... s.

,,. » S, ■;.;' 25 1

convenience mode, and F-16 emergency mode. VSS can " B. .,;,. ® s. '?.!' 25 1

only be run from the front cockpit. Only one cockpit is ,. J8 ® ® *n ja ','.'.' IM 16 1

», „Si JF.L '.",' 16 2,3

active at a time. The VSS mode is used during actual ,„
„ A '.'•,'
® JKL 16 2,4
simulation work. The F-16 emergency mode is designed
2. AIM-9S MUST BE CARRIED AND THEY MUST BE THE "L" VERSION
to allow the front cockpit pilot to fly the aircraft using A. 7.M SYMMETRIC g AND 5.86 ASYMMETRtC g ALLOWED WHEN IHTEHNAL FUEL '. 4000 LBS. (300 G. TK: EMPTY)

the side stick and normal F-16 control laws. Its purpose
is to allow the front seat pilot to recover the aircraft
should the rear cockpit safety pilot become incapacitated. TEST OBJECTIVES/SCOPE
The F-16 convenience mode is very similar to the
emergency mode in that it also allows the front seat pilot The general test objectives for the VISTA Phase II flight
to fly the aircraft while the rear cockpit pilot attends to test program were:
other duties such as modifying the simulation gains.
With either the emergency or F-16 convenience modes 1. To verify the safety and integrity of the aircraft with
selected, the sidestick and rudder pedal controller will the VSS engaged.
interface with the DFLCS and the aircraft will operate
with normal F-16 control laws. The difference between 2. To expand the VSS envelope by determining gain and
the two modes is that in the convenience mode all maneuver limits.
manual safety trips which revert control to the rear
cockpit are still available, and the safety pilot can 3. To demonstrate the aircraft's simulation capabilities
maintain throttle control if desired. for research and student training missions.

The flight envelope cleared for the VISTA is shown in The specific test objectives for the VISTA Phase II flight
Figure 3. The standard F-16 flight envelope was reduced test program were:
for the F-16 host aircraft for initial operations to reduce
the required amount of analysis and testing. The 1. To evaluate the transients during VSS engagements,
variable stability flight envelope is further reduced to disengagements, and mode changes in 1 g and
subsonic operation with the option of expanding the maneuvering flight.
envelope in the future. The external loadings cleared for
VISTA are shown in Figure 4. For the F-16 host 2. To verify and develop the functions of the VISTA
aircraft operations, structural loads have been cleared to Integrity Management (VTM) and VSS safety trip
100% of the allowable design limit load for external svstems.
loadings. For VSS operations, structural loads have been
32-4

3. To develop and determine the effectiveness and utility ENGAGEMENT AND PILOT TRANSFER
of using aeroservoelastic (ASE) detection software to TRANSIENTS
augment the other VSS safety trips.
The test program was initiated by performing functional
4. To determine what limits exist on the VSS feedback check flights. The VSS engagement transients and pilot
gains. transfer transients were tested next to determine the
response characteristics of the aircraft after VSS
5. To qualitatively evaluate the simulation capabilities engagements and pilot transfers at various initial
over a broad range of maneuvers and flight conditions. conditions. Testing was conducted at medium altitude,
beginning at medium airspeeds and progressing to low
6. To determine the effectiveness of the direct lift flaps then high airspeeds and dynamic flight maneuvers. The
to support direct lift and pitch pointing maneuvers. pilot transfer transients were evaluated by engaging the
VSS, getting to the desired initial condition, and having
7. To determine the thrust dynamics of the F110-GE- one pilot manually transfer control of the VISTA to the
100 engine. other pilot . These tests also evaluated the safety pilot
control force safety trips. There were no significant
8. To determine basic airframe stability and control transients experienced during these tests.
parameters to provide a baseline for future simulation
developments. VOM AND ASE TESTING

9. To develop and evaluate the safety pilot's ability to The VISTA Integrity Manager (VM) rate limit safety
safely assume control and recover the aircraft during trip testing was performed by sending automatic test
VSS approach and landing tasks. inputs to drive individual flight control surfaces below
rate trip levels to verify that rate limits do not occur and
10. To evaluate compliance with the Statement of Work to drive surfaces at their respective rate limits to cause a
requirements that the aircraft be able to simulate selected safety trip. After the VTM was verified some work was
rigid-body, modal parameters within given ranges. performed on the Aeroservoelasticity (ASE) detectors.
These detectors were considered a backup safety trip and
11. To develop and evaluate the aircraft's ability to were not required to perform flight test. The ASE
simulate specific airplane configurations for research detector development was ongoing during the remainder
type programs using the X-29 as a model. of the flight test program.

12. To obtain user evaluations through demonstration The VTM structural monitor tests and the remaining
flights to potential VISTA users. VTM monitor tests developed and verified the automatic
safety features of the DFLCS. These tests completed the
FLIGHT TEST RESULTS safety trip evaluations. These tests were performed in a
buildup manner, beginning with medium/low airspeeds
The VISTA Phase II flight test plan shown in Figure 5 and benign maneuvers, then progressing to higher
delineates the order in which the test blocks were airspeeds and more aggressive maneuvers. The VTM
completed. The safety/safety trip related flight test structural monitor safety trips generally tripped as
points were flown first. The gain build up testing designed, but there were some cases where the safety trip
followed determining the gain limits. Once the safety design was more conservative than the prediction. Both
trips and gain limits were determined the in-flight manual and automatic inputs were used to verity' the
simulation capabilities test points for VISTA were flown VTM safety trips. Table 1 list some of the VSS VTM
followed by the demonstration test points. trips . An example of the structural limit boundary is
given in Figure 6 for the asymmetric horizontal tail. The
VISTA/F-16 FLIGHT TEST PLAN structural limits affect simulation if negative flap
Jul-94 AUG SEP OCT NOV DEC Jan-95
deflection is required such as when simulating an n/a
SAFETY OF FLIGHT
Functional Check Flights
lower than the F-16's.
VSS Engagement Trans.
Pilot Transfer Transients
VIM Rate.Umitsafety;Trips
ASE Detector Tests
_ " ..;■
VIM TRIPS A FFECTLNG VSS
OPERATE G ENVELOPE

■-■
Initial VSS Gain Buildups m^^^^^^H
VIM Structural Monitor Tesls ^^^^^^^^^^^^H 7.33,-2.4 g
Urn Petsislenc
none
VIM Tests
±.5g non=
Final VSS Gain Buildups
16., -10. deg none

DETERMINATION OF CAPABILITY 440 kis 2nd M = .9 see FiSu »one

1)
Large Amplitude Maneuvers
Flap Effectiveness Tesls ±300. deg/stc none
Thrusl Dynamics Tests

■
±30.deg none
Parameter Indentification ±70. icg/sec 115 ms
VSS Landing Buildups 115ms
±70. deg/sec
Modal Parameter Range
±70. deg/see 500 ms
SIMULATION DEVELOPMENT
±70. deg/sec 500 ms
X-29 Sim Development
Test Pilot School Demos ±100. degree 250 ms
no trip, deg. _
N42.N43 dtg 7 VSS fram
N44.N45 dcg. 7 VSS fram
N46.N47 deg 7 VSS frame

N4B.N49 dtg. 7 VSS frjme

 32-5

The large amplitude maneuvers were designed to allow

the pilots to fly VISTA aggressively at all attitudes and
at a variety of conditions to determine if the safety trips
°HA<d<>9>
were too conservative. The design of the safety trips was
! considered acceptable by the pilots. This subjective
evaluation was made by all pilots. The flap
ASYMMETRIC 1 effectiveness testing recorded data at trimmed flight
HORIZONTAL 10 \ conditions with various flap deflections. The thrust
TAIL VIM A*',
BOUNDARY
■ : ;
/
dynamic testing evaluated throttle servo response and
.i.- • ;
aircraft response to throttle commands. Parameter
q =400 psf
' 8
fA<d
eg)
identification tests provided the data necessary to derive
n = 1g
\
-10
VIM TRIPS the basic airframe longitudinal and lateral directional
„, ALLOWABLE REGION
"•DUE TO TRIP TRANSIENT equation parameters.
I qc>267psl
(Vc, > 275 Ms)
MODAL PARAMETER TESTING

The VISTA modal parameter design goals follow:

Short Period:Frequency 0.0 to 15 radians/sec

GAIN BUILD-UP TESTING Damping Ratio -0.1 to 1.1

The initial VSS gain build up testing was performed to Dutch Roll: Frequency 0.0 to 8.0 radians/sec
clear VSS configurations that were used in the structural Damping Ratio -0.1 to 1.0
monitor and VIM tests. The final gain buildups cleared
the remainder of the VISTA envelope and these test Roll/Spiral: Frequency 0.0 to 5.0 radians/sec
points were completed after the VTM tests were Damping Ratio -0.1 to 1.0
completed.
ANALYSIS APPROACH
The gain build up testing was performed using a single
gain procedure, and double gain procedure. The gains Time response analysis was used to determine the range
were increased until the modal parameter goals in the of dynamic characteristics that the VISTA can simulate.
statement of work were reached or when the aircraft For each dynamic mode, a specific VISTA response was
response no longer fit the desired lower order equivalent matched to an equivalent lower order model response to
system model response due to ASE interactions. The determine the modal parameters. The VSS control law
single gain procedure evaluated primary gains, which configuration used for modal parameter flight tests is
are defined as most often used in in-flight simulations. shown in Figure 7.
An example of a primary gain is elevator per angle of
attack rate (E/aDOT). The double gain procedure tested
a new primary gain in conjunction with a previously
tested gain. The new primary gain was increased to its Command
Control
Command
Feedforward ►-{ )-*-] Lead/Lag |-#£ Aclualo.s L-H V,
STA (- ,*-
Syslem Sjsljm
gain limit while keeping the previously test gain at a
Feedback
constant value. This would be done three times by Syslem

increasing the previously tested gain until the gain limit WSTA VSS CONTROL CONFIGURATION

was reached (example: 40%, 80%, and 100%). In the FIGURE 7

gain build-up tests the gain was run open loop and 2-25
Hz white noise was injected into the flight control system
exciting the desired response. Using the real time
frequency analysis system MASSCOMP (Nyquist plot & For modal parameter tests, the Command Control system
frequency spectrum plot), a gain margin and value for was represented by a pure gain. The command
instability was predicted. The starting point for the close feedforward system distributes the command control to
loop testing was then taken as 50% of the predicted the various VISTA control surfaces. The Response
instability values. Accelerometers on the wing tips, Feedback System provides feedback from VISTA sensors
horizontal tail, and vertical tail allowed the structural to the control surfaces. The actuator lead/lag
mode damping to be monitored. compensation provided the capability to use high sensor
feedback gains necessary to achieve the goal for the
LARGE AMPLITUDE MANEUVERS, FLAP modal parameter range with the existing low bandwidth
EFFECTIVENESS, THRUST DYNAMIC TESTING, actuators. Specific test inputs (step and doublet) were
AND PARAMETER IDENTIFICATION TESTING used to excite the VISTA for various feedback
configurations. To determine the modal parameters, a
32-6

specific VISTA response was matched to the response of .Output Ke~"

an Equivalent Lower Order Model (LOM). The (4)
equivalent LOM is shown in Figure 8. Input (5 + X\)(s + AT)

Lower The model response was matched to the flight response

Command by adjusting the parameters K, Q, and co to satisfy the
Input ->- e"TtS Order
Model error criterion. The equivalent time delay, xe, was held
fixed at 100 msec during the matching process. This
time delay value was based on the best available estimate
EQUIVALENT LOWER ORDER MODEL of equivalent time delay due to the VISTA actuator,
computations and signal conditioning and confirmed
through flight data analysis.
FIGURE 8
The results for the Short Period modal parameter
The LOM parameters are determined by matching the optimization runs cruise condition are shown in Figure
LOM response computed for the same command input 9 and the runs for power approach condition are shown
used during flight tests, to the VISTA response. The in Figure 10.
LOM is either first order or second order depending on
the VISTA response being matched. The LOM
l Short Period

parameters are adjusted to minimize the error between

the model and VISTA responses. The criterion used for 1 ]; /:/ //
optimizing the equivalent model parameters is defined as
ill/7' 7
1 ■ / /':/'//
Ti
1 «?| fpv/ji./ /; / f
(1) Error = IX,■model -XVISTAl dt \ 1 i I h/ ]/A
Tt-T\ Ti

1lulMä^-
\ / / /■/)(///■

where Xmodei is the LOM response, XVISTA is the VISTA

flight response, and Ti and T2 are the start and the final
4
_
\4lW^>^- ::
\ 'jn/y^i^
7/^SSS?»*- :jf
i
j?

times of the flight test, respectively. A particular VSS

gain set or configuration is accepted if the lower order
model parameter set satisfies the following requirement:
T2
1 Short Period, Power Approach
(2) Minimum Error < 0.15 IX,-model Idt
Ti-T^

The match of the LOM time response to the VISTA

response was accomplished through nonlinear
optimization. The error criterion of Eq (1) was
minimized using the technique of Fletcher and Powell.

MODAL PARAMETER RESULTS

The Short Period mode was extracted by fitting a second

order model to the VISTA angle of attack response to a
horizontal tail test input. Two types of second order
model were used. For the complex short period mode, X, + ij or 2to„ (rad/sec)

the second order model form used was: SHORT PERIOD MODAL ANALYSIS RESULTS

FIGURE 10

Output Ke The dark lines in the figure represent the goal for the
(3)
Input s2 + 2£ü)S + co2 Short Period modal boundary. The points that are
circled represent the modal parameter optimization runs
For the case when two Short Period was represented by that slightly exceed the cost criterion but are acceptable
two real roots, the second order model form used was: based on the observed fit. The goal for the modal
parameter limits are reached with the specified feedback
gains and any point within the boundary can be obtained
The response feedback gains used were EAx and
 32-7

E/ccDOT, representing the feedback from the to the short term VISTA roll rate response, where ^ is
complementary filter angle of attack and angle of attack the estimated roll mode and X2 is the spiral mode to be
rate, respectively, to the horizontal tail. estimated. For the case where the roll and spiral modes
were not separated widely, the second order model of
The Dutch Roll mode was extracted by fitting a second equation (4) with two real roots was used to fit the
order model to the VISTA sideslip response to a rudder VISTA bank angel response. For the case where the roll
test input. As with the Short Period case, two types of and spiral modes combined to form a complex pair, a
second order model were used: the first type to extract second order model with damping and frequency was
the damping and frequency and the second type to used to fit the VISTA bank angle response. Figure 13
extract two real roots. Figure 11 presents the results for presents the results of the roll spiral modal parameter
the Dutch Roll analysis for cruise condition and Figure optimization runs for cruise condition and Figure 14
12 present the power approach condition results. presents the results for the power approach condition.

s
1
1 III'/
itftr
T#77
1/7
/ / / '/
/////■: >*

&^x

,<1.
Rdl-Spi«] Mcfc!

Dutch Roll, Power Approach Roll-Spiral, Power Approach

Vf ./!///
/ Tv / /
\
\ , / / //
\l i/^^
rx i

3Vf
The dark lines in the figure representing the goal for
the Dutch Roll. The response feedback gains used were The response feedback gains used were A/P and A/O
R/ß and R/ßDOT, the feedback from the complementary representing the feedback from the VISTA roll rate and
filter sideslip and sideslip rate, respectively to the bank angle, respectively, to the flaperon. The time delay
rudder. The actuator time delay was set at 100msec for for each run was 100 msec.
each run.

The Roll Spiral modes were extracted using three LANDING BUILD-UP RESULTS
different types of lower order models. For the case
where the roll and spiral time constants were far apart, The purpose of the landing build-up tests was to
the roll mode time constant was extracted by fitting a demonstrate that the safety pilot could recover VISTA
first order model,
from an upset during a landing approach. The testing
started at 10,000 feet AGL and worked down to lower
Output Ke altitudes and finally touchdown. All four pilots
(5)
Input (s + fa) practiced recoveries at 10,000 feet AGL. The high
altitude tests did not provide evidence preventing
32-8

repetition of the tests at lower altitudes. Simulated research-type programs. Specific points in the sky were
approaches were then flown in the landing pattern with selected and verified that VISTA could simulate the X-
known and unknown upsets input at 500, 250, and 200 29 at those test points.
feet AGL. The six upsets used were pitch up, pitch
down, roll (left and right), yaw (left and right), slowing FUTURE DIRECTION OF USAF IN-FLIGHT
to high AOA, and throttle back. The upsets were input SIMULATION
by the evaluation pilot when the safety pilot knew what
the upset was going to be. The evaluation pilot then The end of the Cold War generated a new
input upsets when the safety pilot did not know which political environment which reshaped the USAF of the
upset he would have to recover from. The altitude lost 1990's. These new realities are forcing the USAF to
after an upset varied with the different pilots. The pitch increase efficiency with fewer people and lower funding.
down upset caused the largest altitude loss as expected, The remaining resources must be reorganized and
about 80 feet. The pilots were not pressed by a high prioritized to allow greatest benefit for the dollar to be
sense of urgency to recover at these altitudes. realized. The Air Force can no longer afford to develop,
own, and operate facilities that duplicate those already in
The next step was to perform approaches with known existence, regardless of whether they are government or
and unknown upsets to 50 feet AGL. This allowed the civilian owned. Any duplication must be justifiable
safety pilot to sense visual cues such as sink rate, runway based on military need. If a service can be obtained
alignment, and aircraft attitude. The safety pilots had an elsewhere in a timely manner and satisfies the need, then
increased sense of urgency to recover the aircraft it must be obtained from that source. USAF test
quickly. The standard recovery technique was to level facilities are being reviewed to eliminate duplication and
the wings, raise the nose, apply full throttle, and close in some cases eliminate a test capability. In-flight
the speed brakes. This recovery technique worked well simulators were identified as a unique military facility
for all six upsets. with no comparable civilian capability; however, their
survival is not assured. In today's environment, the
Approaches were then made below 50 feet AGL with a survival of in-flight simulation depends on keeping
75 msec time delay added in the roll and pitch these aircraft flying while minimizing the fixed costs.
commands. None of the six hard over upsets were input The realities of the 1990s defense budgets forced the in-
by the evaluation pilot below 50 feet AGL. Small flight simulation to branch out from the traditional
nondivergent Pilot Induced Oscillations (PIO) were missions and pursue non-traditional missions and non-
induced when flying the configuration with added time traditional customers, presenting exciting new
delay. Several safety trips were encountered at 10 feet challenges and possibilities.
AGL and the safety pilots recovered easily without any
problems.
TRADITIONAL IN-FLIGHT SIMULATOR MISSIONS
The actual landings were performed with the VSS
programmed for Cooper-Harper Level 1 flying qualities The traditional mission of the in-flight
and Level 2 flying qualities. The Levels were simulator can be categorized into the following four
determined up and away. Inserting a time lag in the areas; pre-first flight evaluation of new/modified aircraft,
pitch and roll axis allowed Level 2 flying qualities to be research and development, specialized training, and
obtained. All landings were touch and goes with the flying laboratory.
Level 1 VSS configurations and were uneventful. The
Level 2 handling qualities landings were essentially The pre-first flight evaluation of a new or modified
uneventful so a turbulence generator was added during aircraft is what in-flight simulators are most noted for,
the approach, again recoveries were made with no but this type of simulation is actually a small portion of
significant degradation. No Level 3 handling qualities the total in-flight simulator work because the U.S.
approaches were flown. military does not develop new aircraft often. USAF
programs that utilized in-flight simulations include
bombers such as the B-l and B-2, fighters such as the F-
DEMONSTRATION AND X-29 SIMULATION 15, F-16, F-18, JAS-39 Gripen, Lavi, YF-22, and YF-23.
TESTING research aircraft such as the X-15, AFTI/F-16, and X-29,
the Space Shuttle and commercial developments.
The Test Pilot School (TPS) test flights demonstrated
that VISTA could satisfy the curriculum of both the Air Research and development is the mission for which in-
Force and Navy TPS's. Pilots from both TPS's flew flight simulators first were developed and operated.
VISTA and were satisfied that their respective in-flight They are ideally suited for flying qualities research and
simulation curriculums can be completed with VISTA. many other R&D efforts because the flight
The X-29 simulation development evaluated the ability characteristics can be easily changed and the effects of
of VISTA to accomplish specific airplane simulations for new controls, controllers, or control mechanizations can
 32-9

be demonstrated to test pilots and flight test engineers. verifying, and flight testing the new code is a lengthy
This is especially true for evaluating new flight control process. Proposed software changes normally must
concepts, developing flying/handling qualities specifica- undergo an extensive verification and validation (V&V)
tions, and testing displays and human factors concerns. process before being tested in an aircraft to verify that
the original problem has been corrected. Using the
Specialized training missions primarily support training VISTA, proposed software changes that have not been
at the Air Force and the Navy test pilot schools. The V&Vd can be installed in the variable stability system
specialized training missions account for about half of and flight tested to verify proper operation. If the new
the in-flight simulator flying hours, which helps to software change fixes the original problem, it can then
decrease the cost to all in-flight simulation customers. be V&Vd for fleet-wide use. It would also be possible to
Examples of the specialized training mission include the install an actual F-16 digital flight control computer in
NT-33 which was used for nearly twenty years to VISTA in addition to the variable stability computers for
demonstrate a wide variety of flying qualities and HUD testing new code. This use can significantly decrease the
formats. In a one and a half hour flight, about thirty cost of developing new software for operational F-16s.
different combinations of aircraft dynamics, stick
dynamics, control system feedbacks, time delays, and This concept can be extended to other existing aircraft
HUD formats were demonstrated. The VISTA replaced and aircraft under development by using the VSS to
the NT-33A in this role, with a new flight syllabus replicate the bare airframe characteristics of those
oriented toward the VISTA's capabilities. The NC-131 aircraft.
is used to train pilots and engineers to test avionics. Test
techniques are taught in a comfortable but real airborne The technology demonstrator and test bed capability of
environment. In this role, the NC-131 is configured the VISTA will be augmented by the installation of a
with radar, FLIR, and E/O sensors, and INS, GPS, and permanent thrust vectoring capability. This upgrade will
LORAN navigation systems. Many internal parameters make VISTA the Air Force's only thrust vectoring/high
of these avionics systems can be accessed through a angle of attack research aircraft, in addition to giving it a
special computer and data bus. greatly increased simulation capability. This effort is
currently funded and the work breakdown is being
In-flight simulators make excellent flying laboratories. adjusted to match the funds available over the next
The MATV program is an example of how VISTA has several years. With a shrinking number of test assets and
been successfully used as a flying laboratory. The customers, the in-flight simulators will be used as
aircraft are already highly instrumented, and the technology demonstrators and test beds provided other
procedures for designing and approving the installation critical programs can still be supported.
of new or special test equipment are well established.
NON-TRADITIONAL IN-FLIGHT SIMULATOR
NON-TRADITIONAL IN-FLIGHT SIMULATOR CUSTOMERS
MISSIONS
With fewer projects being sponsored by traditional
The non-traditional mission for in-flight simulators is to customers, new customers are being sought. In addition,
serve as technology demonstrators and test beds. some regulations/instructions that are no longer
Technology demonstration programs are not new for in- consistent with the new economic environment are being
flight simulators, but they were not used significantly in rewritten making in-flight simulation available to almost
the past. In the past, special test aircraft were dedicated anyone that can pay the program cost.
to demonstrating new technologies. Examples included
the AFTI-16 and AFTI-111 programs. In-flight In the past, virtually all in-flight simulations were
simulators were seldom used for these applications performed by either the Air Force, Navy, or NASA.
because it required the aircraft to be dedicated to that The new, more stringent economic situation is forcing
effort for extended periods denying other programs program management to look beyond the traditional
access to their aircraft. Also they seldom needed customer base for new project sponsors. Newer
simulation capability to accomplish their mission. The customers include U.S. industry, foreign government
Air Force's in-flight simulators will certainly compete in programs performed through Foreign Military Sales, and
the future to perform this function. It is anticipated that foreign industry. American commercial programs have
such programs, if structured carefully, can be performed been performed since the mid-1980s, but new regulations
without causing undue loss of ability to support other now make it relatively easy to allow industry to use
needs. unique military facilities. Foreign military programs
were supported since the early 1980s. Supporting
Another non-traditional mission for the in-flight foreign commercial programs is relatively new. A recent
simulator is to serve as a test bed for developing F-16 TIFS simulation program was the first foreign
flight control system upgrades. With production F-16s commercial development ever supported by the Air Force
having a digital flight control system, new software is (with no regulations in place to provide guidance for this
simple to install in the field; however, developing. situation, lengthy high-level coordination was required
32-10

until permission to perform the program finally was concept has been documented elsewhere . Prior to the
obtained). Efforts to find new sponsors will be expanded VISTA becoming operational, this operating concept
in the future, as well as continuing to make the aircraft was reviewed and determined to still be the most viable
available to traditional customers. means for operating in-flight simulators.
OPERATION CONCEPT FOR IN-FLIGHT
SIMULATORS FUTURE IN-FLIGHT SIMULATION PROGRAMS

The new economic environment is impacting the Despite the decrease in R&D budgets, the in-
operation of in-flight simulators, but not as badly as flight simulation program remains viable. These aircraft
many other R&D facilities. Up until the mid-1970s, the proved they provide a unique capability, and serve an
TIFS and NT-33A were completely funded by Wright essential need not duplicated anywhere within the
Laboratory. Most projects flown were to support in- government or civilian sectors.
house needs, but based on priority, projects for other
government agencies also were flown. This changed The TIFS/NC-131H is committed to three
about 1975 when the laboratory considered terminating programs in the near future. These are to support
the program. The program was saved, but made self- NASA's effort in developing technologies for the High
supporting. The lab would provide some seed funds for Speed Civil Transport, to continue to perform research
a very basic level of maintenance, proficiency, and in the area of large aircraft flying qualities for Wright
system upgrades. The program management would have Laboratory and Douglas Aircraft, and to simulate the
to identify all costs for performing specific projects and Indonesian Nusantara N250 commuter aircraft.
the sponsor would have to pay those costs in order to see
their project flown. By the late 1980s, this support grew The VISTA also has several commitments.
to about $200,000 per year from the laboratory with an Following installation of a permanent thrust vectoring
additional $500,000 per year or more from headquarters capability, VISTA will be the Air Force's high angle of
for depot-level maintenance. attack/post stall maneuvering research aircraft. It will
also support development of the F-22 and Indian Light
Over the last three years the above mentioned support Combat Aircraft. The Joint Advanced Strike
funds were reduced and ultimately terminated. The new Technology (JAST) program plans to use VISTA to aid
environment's impact in the in-flight simulators is not in developing technologies for future strike aircraft, but
procedural, but monetary. With support funds gone, all exact utilization has not been determined. The Air Force
costs now must be paid by sponsors as part of using the and Navy test pilot schools also plan to continue using
aircraft. With the TIFS and NT-33A each flying about VISTA. The aircraft made its first deployment to the
200 hours per year, the impact is a cost increase of over schools in March and April 1995.
$1500 per flight hour.
CONCLUSION
The new environment extends also to spare parts
obtained from Air Force depots. In the past, the depots The Phase II VISTA flight testing at the AFFTC showed
were funded to provide spares and overhauled that VISTA can support the high performance in-flight
components for whoever needed them. About three simulation of the future for the US AF. Once the VISTA
quarters of the cost was absorbed by the depots, making upgrade program is completed VISTA will have a thrust
many of the needed components "free" to the users. This vectoring capability. Thrust vectoring will allow VISTA
changed as well; all parts now must be paid for by the to support high AOA technology demonstrations and in-
user and the cost passed on to the project sponsor. flight simulations. The VISTA is available for new
missions as a technology demonstrator and as a test bed.
The operating concept for Air Force in-flight simulators New VISTA customers are also being pursued including
remains unchanged. The four key elements remain foreign commercial customers.
government ownership, contractor operation, base at
contractor facility, and use of a task ordering contract.
This concept evolved over thirty-five years of operating
1. Markman, Steven R., USAF In-Flight Simulation: A
in-flight simulators. It allows the Air Force to take
Cost-effective Operating Approach, AIAA Flight
advantage of the positive elements of government and
Simulation Technologies Conference, Monterey CA,
industry working together, while eliminating the
negative aspects of each. A detailed discussion of this
 33-1

EXPERIENCES ACQUISES LORS DE LA MISE EN OEUVRE DES PROTOCOLES SID

AVEC DES SIMULATIONS DEJA EXISTANTES.

Claude Crassous de Medeuil

THOMSON TRAINING & SIMULATION SA.
1, rue du General de Gaulle BP 226
Cergy Pontoise CEDEX FRANCE

SOMMAIRE LISTE DE SYMBOLES

Le besoin de coupler des simulateurs ou des simulations croit CATS Combined Arms Training Systems
chaque jour et la diversite de ces simulations est tres grande. CCTT Close Combat Tactical Trainer
Des experiences passees ont montre qu'il est presque impos- GPS Global Positioning System
sible d'interconnecter des simulations dejä existantes s'il faut I/ITSEC Interservice / Industry Training Systems
pour cela utiliser des methodes et des structures communes and Education Conference
imposees ä chaques simulations. L'approche des Simulations IP Internet Protocol
Interactives Distribuees (SID) permet de coupler des simula- ITEC Industry Training Equipement Conference
tions heterogenes sans avoir ä les modifier de maniere impor- MBONE Multicast Backbone
tante. De plus les simulateurs dejä existants ont certaines PDU Protocol Data Unit (Le message elemen-
particularites quand on les compare aux simulateurs recents. taire defini par la norme IEEE 1278-1)
Ils sont souvent limites en memoire, puissance de calcul et RNIS Reseau Numerique ä Integration de
ont rarement des moyens de couplage ä des reseaux de com- Services (ISDN)
munications. SID Simulations Interactives Distribuees
SIMAN Simulation Management
Ce papier decrit les methodes et les techniques utilisees pour
STOW-E Synthetic Theater of War - Europe
developper des bibliotheques logicielles mettant en oeuvre les
TCP Transmission Control Protocol
fonctions d'interoperabilite des SID conformement ä la norme
UDP User's Datagram Protocol
IEEE 1278. Les exigences principales pour ces bibliotheques
etaient: portabilite, masquage pour l'utilisateur des details du
protocole, et flexibility pour pouvoir s'adapter au large even- INTRODUCTION
tail des simulations utilisees dans le groupe THOMSON- Au debut des annees 90 THOMSON-CSF a commence ä
CSF. ressentir le besoin de coupler entre elles des simulations. En
ce qui concerne les simulateurs d'entrainement, le couplage
Les conclusions qui peuvent etre tirees ä la suite de trois
permet 1'entrainement collectif de l'equipe de base teile que le
experiences sont les suivantes :
peloton de char ou la patrouille aerienne, avec une totale
- l'utilisation des methodes SID pour coupler des simulations interaction entre les individus qui composent cette equipe.
existantes resout les problemes d'echange de donnees, de Avec plus de simulateurs on pourra mettre sur pied des
format et des differences de Systeme Sexploitation, mais exercices plus importants (par exemple CATS en Grande
pas les problemes de coherence entre bases de donnees et Bretagne ou CCTT aux Etats Unis). Pour les simulations
de modeles, technico-operationnelles le couplage permet d'augmenter le
realisme de l'environnement utilise pour valider un systeme.
- les täches d'adaptation sont specifiques ä chaque couplage,
Comme aucune simulation isolee ne couvre la totalite des
- comme les traitement SID sont coüteux en calculs, il est entites presentes sur un theatre d'operation, le seul moyen de
necessaire de disposer d'une bonne marge de puissance, construire des simulations realiste ä grande couverture est de
sinon des messages seront perdus dans des tampons satures coupler des simulations. Apres une phase d'investigation,
dans les couches reseaux, THOMSON-CSF a reconnu l'interet d'utiliser les methodes
SID et en 1992 la decision de developper un outil Thomson ä
- il est indispensable de disposer d'outils d'analyse des
ete prise comme methode d'acquisition d'experience et de
echanges sur le reseau en temps reel,
savoir faire, pour pouvoir developper les fonctionnalites
- il faut utiliser des bibliotheques d'interoperabilite dejä specifiques dont le groupe pourrait avoir besoin et aussi parce
validees ou bien disposer d'une reference validee pour se qu'ä l'epoque aucun produit commercial n'etait disponible en
contröler, France ou en Europe.

- une bonne competence en reseau et Systeme est necessaire

en plus de la connaissance de la simulation ä coupler. RECENSEMENT DES BESOINS
Le recensement des besoins a fait apparaitre les caracteristi-
ques suivantes :

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?",
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
33-2

ADDITION EMBEDDED EXTENSION

HOST CPU

HOST
SIMULATION

SHARED DATA

SIMULATION
SHARED DATA SIMULATION PROC2
SIMULATION SHARED DATA

♦
ACCESS TO DIS

♦
DIS Libraries DIS Libraries DIS Libraries

♦
NETWORK NETWORK
♦ NETWORK

D SCPU 4 HC)ST CPUT PROC

7
Figure 1 : Les 3 methodes d'implementation de bibliotheques SID

Simulations ä coupler: d'Exploitation et il etait done impossible de developper une

version de l'outil pour chaeun de ces cas.
- Les simulateurs d'entrainement avec homme dans la boucle
et un fonctionnement temps reel,
CHOIX D'ARCHITECTURE
- Les simulations d'etude dont l'eventail va de la simulation Les contraintes ci-dessus menent ä des premiers choix d'ar-
tres detaillee, comme par exemple la restitution d'une
chitecture :
forme d'onde jusqu'ä des simulations tres globales de thea-
tre. En general ces simulations ne sont pas temps reel, cer- - les logiciels d'interoperabilite doivent pouvoir etre inde-
taines sont ä pas de temps, d'autres sont evenementielles, pendants de la machine et du Systeme d'exploitation de la
simulation ä coupler. Ce qui veut dire que dans certains cas
- le besoin principal est d'abord de recuperer l'existant, il ces logiciels ne pourront s'executer sur la machine de simu-
n'est pas envisageable de reecrire des simulations pour les lation mais qu'ils devront etre implementes sur des machi-
coupler et d'attendre que de nouvelles soient crees pour nes separees. Cette decision a ete prise d'autant plus faci-
beneficier des avantages du couplage. lement qu'il existe sur le marche plusieurs offres de syste-
Un certain nombre de contraintes sont aussi apparues a cette mes de memoires reflechies permettant ä plusieurs calcula-
phase de l'etude: teurs de type differents de partager une meme zone me-
moire. Tous ces dispositifs sont uniquement materiels et ne
- particulierement dans le cas des simulateurs d'entraine- necessitent pas l'emploi de logiciels de communications,
ment, beaucoup de configurations de calculs auraient ete les performances d'echange sont telles que en general on
tres coüteuses ä faire evoluer pour satisfaire les besoins en peut negliger les temps de transfert et assurer un fonction-
puissance de calcul requises par le couplage et les connec- nement synchrone entre machine de simulation et serveur
tions ä des reseaux de communications, d'interoperabilite. Certains de ces dispositifs sont meme
- les simulations existantes ont ete developpees sur une tres capables d'assurer les conversions d'ordre des octets dans
large variete de machines, de langages et de Systemes un mot (Poids forts en tete, Poids faibles en tete). Ce qui
mene aux trois possibilites suivantes :
 33-3

- Addition, operations d'entretien et de datation des PDUs (voir figure

- Extension 3),
- Integration (Voir figure 1)
- la plupart des simulations d'etude ne sont pas cadencees de
THOMSON-CSF a done pris la decision de developper les maniere reguliere, les calculs recommencent des que pos-
logiciels d'interoperabilite en C et sous UNIX "temps reel". sible. Le serveur d'interoperabilite doit offrir des services
Le choix du langage C a ete fait parce que e'est celui qui de cadencement analogues ä ceux des moniteurs temps reel
permet le plus facilement de s'interfacer ä des langages an- des simulations temps reel (voir figure 4).
ciens comme le Fortran ou ä des langages de haut niveau
comme ADA ou le C++.
FILTRAGE
UNIX s'est impose parce que e'est le Systeme d'exploitation le Pour diminuer les echanges inutiles entre la simulation et le
plus largement repandu. Les fonctionnalites "temps reel" serveur d'interoperabilite les principes suivants ont ete mis en
definies par la norme POSIX 1003.13 avaient ete jugees oeuvre:
necessaires pour maitriser correctement les ressources de la
- l'acces ä une entite en lecture ou sa mise ä jour en ecriture
configuration. Malheureusement ä cette date aucune version
doit pouvoir se faire seulement sur une partie de la struc-
temps reel n'etait disponible pour les stations de travail les
ture du PDU correspondant. En effet une partie importante
plus couramment employees dans les laboratoires. Ce qui
des champs du PDU vehicule des informations normale-
nous a oblige de commencer nos developpements avec une
ment invariables au cours du temps. Ce sont les champs
version standard dUNK.
Entity Type, Marking, Guise, etc. Cette approche a ete
reprise dans la proposition d'un PDU changement d'etat qui
CHOIX ^IMPLEMENTATION ne transmettrait que les changements du PDU (Ref 1)
Pour satisfaire l'imperatif de recuperation maximum des
- l'acces doit pouvoir se faire selon des criteres de nature afin
simulations existantes et pour obtenir un outil aussi universel
de permettre ä la simulation de n'acceder qu'aux entites qui
que possible, les principes suivants ont ete retenus :
l'interessent. Ainsi une simulateur de combat Air-Air ne
- comme l'entretien des positions presuppose un deroulement s'interessera pas aux entites n'appartenant pas au domaine
continu du temps, les simulations evenementielles devront air ou au domaine munitions. L'emploi de criteres de nature
etre adaptees pour fonctionner selon le mode "ä pas de permet aussi de faire des tris.
temps". Cela impose la gestion d'un evenement supplemen-
- le serveur d'interoperabilite doit pouvoir filtrer les PDUs
taire, le deroulement d'un pas de temps,
par nature. Ceci permet par exemple, ä une simulation qui
- toutes les operations d'entretien de position, de datation, represente un Systeme sans detecteur radar de ne pas rece-
d'extrapolation, de seuillage, de changements d'axe et/ou voir les PDUs Transmitter.
d'unites pour passer des axes et unites de la simulation aux
Nous avons aussi identifie le besoin d'un Systeme de filtrage
axes et unites specifiers par la norme doivent etre assurees
en distance pour ne recevoir que les entites situees dans le
par les logiciels d'interoperabilite,
domaine de perception de la simulation hote. En effet le
- les logiciels d'interoperabilite entretiennent une base de couplage de plusieurs simulations a en general comme conse-
donnee de la situation globale ä partir des PDUs (Protocol quence d'augmenter le nombre d'entites dans l'environnement.
Data Unit) recus de l'exterieur et des mises ä jour fournies Le traitement des entites hors du domaine de perception
par la simulation hote. La simulation hote accede en lec- represente une charge de calcul inutile. Ce filtrage pourrait
ture et en ecriture selon ses besoins propres. On obtient etre fait par le logiciel d'interoperabilite ou par la couche
ainsi une structure de serveur (voir figure 2), d'adaptation. Le filtrage au niveau du serveur d'interoperabi-
lite est plus efficace en terme de charge de calcul, puisque il
- ä cause de la tres grande diversite des organisations de
est fait au moment de la reception du PDU et qu'ensuite il rüy
donnees des simulations utilisees, il a ete decide d'acceder
a plus d'entretien.
ä cette base de donnee entite par entite, evenement ou
commande par commande, au moyen de bibliotheques de Neanmoins ce filtrage n'est pas toujours süffisant, en effet les
fonctions de lecture ou d'ecriture, simulations ne peuvent traiter qu'un nombre limite d'entites.
Ces simulations en autonome ont un generateur d'environne-
- comme les PDUs de type Etat sont emis systematiquement,
ment qui respecte les limitations du Systeme. Par contre une
ils sont moins importants que les PDUs de type gestion de
fois couplees, meme apres un premier filtrage en distance,
la simulation ou du type evenement comme le tir ou la de-
l'environnement peut depasser les capacites du Systeme. H
tonation. En consequence la priorite de traitement doit etre
faut done prevoir des filtrages complementaires specifiques
assuree pour ces categories. Pour cela nous les separons et
de l'application, par exemple en ne retenant que les entites les
traitons les PDUs d'etat apres avoir termine le traitement
plus proches ou bien celles situees dans un secteur privilegie.
des PDUs de gestion ou evenement,
Ceci ne peut etre fait qu'au niveau de la couche d'adaptation.
- le serveur d'interoperabilite doit se synchroniser avec le
cycle de la simulation hote pour assurer la coherence des
33-4

Architecture Generale InterSIM

Network Environment Memory Management

\ Etats

Evenements

/ \ Commandes

ne

Debut Simulation ne

Legendes
Abreviations
E.I Entites Internes
E.E Entites Externes
S.L Supervision Locale
MTR Moniteur Temps Reel

-C_.S.C,'

Activation
Acces

Figure 2 : Architecture generale d'lnterSIM

 33-5

debut de cycle debut de cycle

Simulation

j depart InterSIM ..tempsr

FrT^-.:.TO:::--''-i

%
^^~™~~y,~™~™''^^^^^^^^^
Adapi Ö4-sti(^^Mesti Kiitmm
»chec»
e
9.int. I e-int ||§§§§§| Hllliillllll :" ©i;:

|seau
eception sion
Figure 3 : InterSIM synchronise par la simulation note

Entites Internes : Entites gerees par la simulation höte

Entites Externes : Entites recues par la simulation höte
Entites externes 1 PDUs de type Evenement ou Commande. Ces PDUs doivent etre trait.es des la redeption, sinon ils
seront perdus.
Entites externes 2 PDUs de type Etat. Leur traitement peut etre retarde.
La täche d'adaptation des Entites Externes est activee systematiquement par une minuterie et peut interrompre la täche de
gestion des Entites Externes, de facon ä disposer de donnees ä jour pour le prochain cycle de simulation.

^^^^^^^^^^^^H Simulation
NMHMMHNHNH
Li hf*fsohfid,:: aL! 4"
/
&x:...-.-.-.i '!::■:■ ':':':*tf ■"■'■>$&*; . .■■■.:

1 Gestion Entites Internes

2 Gestion Entites Externes 1 (Messages)
3 Gestion Entites Externes 2 (Entretien des positions)
4 Adaptations Entites Externes
5Adaptations Entites Internes

Figure 4 : InterSIM synchronisant la simulation

33-6

Une autre methode interessante de filtrage a ete suggeree par - gerer la simulation en creant ou supprimant des entites,
plusieurs concepteurs. Cette methode consiste ä employer demarrer, arreter la simulation. Ce sont les PDUs de la
l'adressage Diffusion de Groupe (Multicast) prevu par les famille SMAN,
protocoles IP et ä affecter un groupe ä chaque zone geogra- - de plus il a ete reconnu que les besoins des simulations
phique. Chaque simulation inscrit le PDU qu'elle envoie dans d'etude rendaient necessaires l'acces ä des valeurs internes,
le groupe correspondant ä sa zone geographique et souscrit la possibility de modifier des parametres. La norme a done
aux groupes correspondants aux zones qu'elle est capable de prevu les PDUs Set Data, Send Message....
percevoir. cette methode ne pouvait pas etre implementee
jusqu'ä recemment parce que le mode Diffusion de groupe Comme le PDU de collision n'est utile que dans les simula-
etait peu disponible dans les stations de travail et pas du tout tions d'environnement terrestre avec visuel et que la logisti-
dans les routeurs. Le developpement des videoconferences que n'etait pas utilisee dans les simulations THOMSON-CSF,
sur le reseau Internet utilisant le Systeme MBONE et l'intro- nous avons decide dans une premiere etape de ne pas les
duction de 1'Assynchroneous Transfer Mode (ATM) sont en implemented Depuis le besoin est apparu et il a fallu les
train de declencher une offre de produits en forte croissance. rajouter.
L'experience Synthetic Theater of War Europe (STOW-E) de En plus de ces PDUs THOMSON-CSF a ajoute des PDUs
novembre 94 a utilise le mode Diffusion de Groupe. L'emploi supplementaires non prevus par la norme pour gerer la
de ce mode est prevu dans la version suivante de la norme synchronisation des horloges et le temps non reel.
IEEE en cours de ratification. THOMSON-CSF utilise ces PDUs pour :
- synchroniser les horloges de calculateurs dont les OS ne
CHOK DES PDUS. prevoient pas de fonction de synchronisation sur une hor-
La norme IEEE 1278 prevoit des PDUs (Protocol Data Units) loge de reference,
pour:
- synchroniser des simulations non temps reel presentant une
- decrire l'aspect exterieur des entites, tel qu'il pourrait etre charge de calcul fortement variable.
percu par des capteurs. Dans cette categorie on trouve :
- le PDU Entity State qui decrit la position et l'aspect vi- SYNCHRONISATION DES HORLOGES
suel des entites, Un des principes de base de la norme SID, e'est l'entretien par
- le PDU Emitter qui decrit remission electromagnetique extrapolation des positions des mobiles. En theorie cette
rayonnee par une entite, extrapolation se fait en integrant la vitesse et eventuellement
l'acceleration sur l'intervalle de temps separant l'instant oü
- le PDU Laser qui decrit la zone illuminee par un desi- l'emetteur a calcule le vecteur d'etat et l'instant ou la simula-
gnateur laser, tion emettrice fait un nouvel acces ä la position extrapolee.
- le PDU Transmitter qui decrit un emetteur radio ainsi Cette extrapolation se fait selon la formule generate suivante :
que le PDU signal qui achemine le contenu de remis-
sion, Pu=Pto + Vt0(Tl-TO)+lfto(Tl-TO)2
- decrire le tir d'une arme (PDU Fire) et l'explosion de la
munition tiree (PDU detonation), Formule 1 : Extrapolation de la position ä partir de la
vitesse et de l'acceleration.
- signaler la collision d'une entite avec une autre avec un
element du terrain (PDU Collision). Le but de PDU est de
signaler aux simulations receptrices qu'elles doivent modi- - T0 = heure de validite du PDU
fier l'entretien des positions des entites en collision. La
simulation receptrice doit arreter d'extrapoler et figer l'en- - Tj = heure courante
tite en collision en attendant un nouveau PDU Entity State Et pour cela la norme SID a prevu que chaque PDU soit
pour cette entite, accompagne d'une estampille de date (time stamp), dormant
- permettre la simulation des operations logistiques en l'heure de validite du PDU. Mais pour que le calcul ci dessus
decrivant les echanges entre entites demandeuses de servi- soit juste, il faut que les horloges de la simulation emettrice
ces d'approvisionnement ou de reparations et les entites et receptrice soient synchronisees entre elles. La norme SID
capables d'offrir ces services. Ce sont les PDUs Repair ou permet ä la simulation emettrice d'estampiller soit ä partir de
Ressuply, offer, request acknowledge. Ces PDUs servent ä l'heure en Temps Universel, soit d'utiliser une reference
decrire les quantites echangees et ä simuler les delais locale.
impliques par ces operations. Pratiquement lors des demonstrations faites lors des confe-
rences ITEC ou MTSEC seule la reference locale etait utili-
 33-7

see par simplification. Dans ce cas deux alternatives sont ou:

possibles:
- tQ: heure locale pour B de remission du PDU Requesting
- faire lliypothese que les horloges locales sont suffisamment Time
proches pour que l'on puisse negliger leur difference et la
- ti: heure de reference de A d'emission du PDU Servicing
formule 1 continue ä s'appliquer.
Time PDU
- faire lliypothese que la difference entre les horloges est
- t2: heure locale pour B de reception du PDU Servicing
plus grande que la somme des temps de codage ä remis-
Time .
sion, transfert sur le reseau et codage ä la reception. Dans
ce cas, il faut ä la reception du PDU remplacer l'estampille
emise par llieure locale de reception. FONCTIONNEMENT EN TEMPS NON REEL
La norme SID prevoit explicitement le fonctionnement en
Dans un reseau local sans synchronisation particuliere des
temps reel, qui s'impose bien evidemment pour les simula-
horloges la deuxieme hypothese est la plus valable. En effet
tions avec homme et/ou du materiel reel dans la boucle.
les temps de transfert sur un reseau local non sature sont
Cependant un grand nombre de simulations ne sont pas temps
faibles, et il est aussi possible de compenser les temps de
reel et il serait interessant de pouvoir les coupler entre elles
codage et de transfert en soustrayant quelques dizaines de
en utilisant les methodes des Simulations Interactives
millisecondes ä llieure locale de reception. Mais cette me-
Distributes.
thode oblige ä prevoir un moyen de changer cette correction
parce qu'elle change d'un exercice ä l'autre en fonction du Ceci est possible en utilisant la methode un facteur d'echelle
reseau et de la charge. variable pour le temps (Ref 2). Dans ce mode de fonctionne-
ment on definit:
Dans le cas d'un couplage au moyen d'un reseau ä longue
distance les temps de transfert peuvent etre beaucoup plus - un facteur de dilatation du temps, rapport Temps
importants et variables, par exemple dans le cas de reseau ä Simule/Temps Universel,
commutation de paquets. De meme certaines simulations
- une duree de temps simule (Time Slice) pendant laquelle
d'etude reclament une tres bonne synchronisation des horlo-
ce facteur de dilatation sera utilise.
ges pour assurer une bonne precision. Une erreur de quelques
dizaines de millisecondes provoque une erreur de quelques Chaque simulation demarre et va au bout de ce temps simule,
dizaines de metres sur la position d'un mobile evoluant ä passe en mode gel et fournit ä la station de supervision un
Mach 3. De plus des exercices de grande amplitude comme Time_Slice_ACK PDU qui contient llieure universelle
STOW-E ont montre qu'il etait quasi impossible de fusionner d'emission de ce PDU et le ratio K = Temps simule / Temps
des enregistrements faits sur chaque site s'ils n'etaient pas CPU consomme.
estampilles en Temps Universel.
Quand toutes les simulations ont termine et renvoye leurs
Dans ce cas les simulations doivent utiliser une reference de comptes-rendus, la station de supervision calcule un nouveau
temps unique. Cette reference peut etre obtenue par l'utilisa- facteur de dilatation et une nouvelle tranche de temps simule
tion d'un service de distribution de rheure s'il existe sur le ä executer en fonction des comptes rendus et renvoie ces
reseau et s'il a la precision requise. Une autre methode est informations aux simulations au moyen du PDU Time_Slice.
l'emploi d'une reference locale suffisamment precise et stable
A l'interieur de cette tranche de temps simule, plusieurs pas
pour satisfaire la condition d'unicite. Une teile reference peut
de simulation s'executeront. Pour reduire au minimum la
etre fournie par un recepteur GPS. II existe des cartes au
desynchronisation entre les simulations "lentes" et les simu-
format VME incorporant un recepteur GPS facilement utili-
lations "rapides", le Moniteur Temps Reel des bibliotheques
sable dans un calculateur. La principale difficulte est l'instal-
d'interoperabilite relance la simulation de facon ä ce qu'elle
lation de l'antenne qui doit avoir une vue directe des satellites
recommence son prochain cycle au bout d'un temps egal ä au
GPS. Si l'on ne cherche pas une tres grande precision, il est
moins:
possible de synchroniser les horloges en employant la me-
thode de "Flaviu". DT = Pas de temps simule x K
Dans cette methode un site maitre delivre llieure aux escla-
ves ä leur demande. Le simulateur "A" est responsable de la SECURITE.
reference. Le simulateur "B" se synchronise sur "A" au moyen L'emploi de PDUs normalises est potentiellement une source
d'un echange de PDUs, extension aux protocoles SID d'insecurite. Meme si les donnees vehiculees ne contiennent
(Requesting Time PDU, Servicing Time PDU). pas directement les caracteristiques des entites simulees,
La formule de synchronisation est: toute personne qui a pu ecouter le reseau est capable de
reconstituer les trajectoires, les performances, les distances
d'engagement etc. Si les echanges se font sur un reseau non
XT XT protege il faut employer des techniques de cryptage. Deux
IT _ ■n-Bt2 " flBtO _|_ II
refAtl types de methodes sont possibles. On peut crypter par logiciel
les PDUs au moment de leur calcul, juste avant le passage
33-8

dans les couches reseau, ce qui laisse en clair les en-tetes IP attend qu'il le rejoigne. La methode classique du filtre du
utilisees par les functions de routage. On peut aussi utiliser premier ordre appliquee ä une extrapolation sur la position
des Chiffreurs places ä la sortie du calculateur, ce qui eco- introduit une teile erreur, puisque la position evolue cons-
nomise de la puissance de calcul. Par contre il faut utiliser tamment. Le lissage lineaire (l'erreur de position 8 lors de
des Chiffreurs qui comprennent les en-tetes reseau pour les la reception d'un nouveau PDU est annulee en n cycles en
laisser en clair. De meme il faut utiliser des Chiffreurs per- utilisant n increments egaux 8/n) ne cree pas d'erreur resi-
mettant le mode diffusion vers plusieurs destinataires utili- duelle, mais il y a une discontinuite de vitesse qui est
sant une meme cle. quelquefois visible. Un filtre du deuxieme ordre avec
avance de phase donne de bons resultats. Pour l'attitude,
Les techniques logicielles sont coüteuses en temps calcul, comme celle ci n'evolue pas constamment, un lissage du
mais elles sont faciles ä mettre en oeuvre. C'est une methode premier ordre suffit.
interessante pour "cacher" seulement quelques entires sensi-
bles dans un environnement plus large et sans sensibilite
particuliere.
Le chiffrage, s'il est systematique, protege aussi contre les
ACTUAL POSITION
intrusions en cours d'exercice. Plus generalement il faudrait
n'accepter ä la reception que les PDUs emis par des machines
identifiees comme participant ä l'exercice.
Finalement nous avons conclu que les besoins etaient trop
divers et les solutions tres dependantes de la configuration
materielle et reseau pour pouvoir proposer une reponse
standard. Notre approche est d'utiliser un reseau local protege
et dans le cas de liaisons ä longue distance de resoudre le
Probleme au niveau des passerelles. Ceci a l'avantage de ne
pas imposer de contraintes particulieres au niveau des simu-
lations.
Figure 5 : Retard de position

LECONS TIREES DE L'UTILISATION DU LOGICIEL.

Ce logiciel a dejä ete utilise dans trois occasions differentes l'extrapolation d'un parametre pour compenser le delai de
en dehors du demonstrateur de validation qui a ete utilise transmission peut creer des transitoires perturbants lorsque
pour son developpement. ce parametre peut varier de maniere important en un laps
de temps du meme ordre que le delai de transmission (voir
figure 6). C'est le cas par exemple de la vitesse de roulis
Premiere experience. qui peut atteindre, sur un avion d'arme, des valeurs tres
Une premiere fois, dans une version preliminaire et ä titre de elevees en tres peu de temps. L'extrapolation de cette vi-
demonstration, lors du couplage de deux simulateurs d'avions tesse sur une duree egale au delai mene ä des depasse-
d'arme Mirage 2000C. Ces simulateurs sont situes sur les ments car en realite la vitesse s'est annulee en un temps
bases aeriennes de Cambrai et Orange, soit ä plus de 800 Km plus court. Cet effet a ete evite en n'extrapolant plus le
(500 Miles). Les principaux enseignements ont ete : roulis. Cet effet illustre une des limitations de la methode
- l'utilisation de liaisons RNIS ä bände etroite est une solu- SID. La methode ne s'applique bien que pour transmettre
tion pratique pour des liaisons non permanentes. On peut des phenomenes dont les constantes de temps sont supe-
multiplexer plusieurs lignes entre elles pour atteindre le rieures aux delais de transmission.
debit souhaite. Une fois etablie la liaison est permanente
de bout en bout et le delai de transmission est tres faible, Deuxieme experience.
au moins pour ces distances. Le out de l'abonnement est Pour cette occasion le logiciel a ete utilise pour une expe-
modique, compare aux lignes permanentes louees, et on ne rience de couplage de simulations qui a eu lieu pendant la
paie que l'utilisation. conference ITEC de 1994. Cette fois ci le but etait de coupler
- pour des simulateurs de vol l'aspect visuel est primordial. un simulateur de radar aerien developpe par LINK-MILES ä
Le saut de position ou d'attitude qui se produit ä la recep- l'environnement cree par les autres simulations. C'etait aussi
tion d'un nouveau PDU est parfaitement visible pour les la premiere fois que ce logiciel etait confronte ä d'autres
simulateurs en dome. II faut done lisser les valeurs de po- logiciels d'interoperabilite. Les enseignements qui ont ete
sition et d'attitude en sortie des programmes d'entretien. II tires de cette experience sont:
faut employer des methodes de lissage qui n'introduisent - le portage du logiciel, originellement developpe sur SUN,
pas d'erreurs residuelles car celles ci sont detectees lors des sur machine Silicon Graphics a ete fait sans aucun proble-
vols en formation. Une erreur residuelle de quelques me- mes, justifiant ainsi le choix d'UNLX comme Systeme d'ex-
tres est visible, car chaque pilote voit l'autre en "arriere" et ploitation,
 33-9

2
UJ

HI

DC

t1 +delay t2 + delay

Figure 6 : Depassement pendant une extrapolation

- la mise au point complete d'un logiciel de ce type ne peut Une autre caracteristique est aussi que les simulations de C2
se faire qu'en utilisant un serveur d'interoperabilite de refe- n'avaient prevu par simplification aucun contröle de la qualite
rence. A cette occasion nous avons decouvert plusieurs de la transmission, parce que dans une execution centralisee
erreurs de formatage et de programmation des algorithmes aucun rate de transmission n'etait ä redouter. Les principaux
d'estime. Ces erreurs nous avaient echappe parce qu'elles enseignements ont ete:
etaient masquees par des erreurs symetriques ä la
- la perte de messages, si eile n'est pas trop frequente, est
reception,
acceptable pour les messages d'etat grace ä la robustesse de
- il faut prevoir les moyens de changer rapidement les acces- la methode. Par contre des pertes, meme faibles, perturbent
ses reseau, les ports IP/UDP, les classes d'adressage de la la simulation des systemes C2. Pour fiabiliser la transmis-
machine hote pour s'integrer dans le schema d'adressage sion des messages C2 deux methodes ont ete envisagees,
d'une experience de couplage, soit utiliser le protocole TCP/IP, soit faire contröler la
transmission par le logiciel d'interoperabilite. D'une part
- l'utilisation d'outils permettant la visualisation en temps
TCP/IP offre la garantie de resultat recherchee, mais est
reel des PDUs qui circulent sur le reseau facilite beaucoup
coüteux en ressources et n'offre que des transmissions en
la mise au point. Les outils d'enregistrement concus pour
point ä point. D'autre part UDP/IP est plus rapide, plus
l'analyse ä posteriori ne sont pas utilisable pour une mise
simple ä utiliser et permet l'emploi du mode Diffusion de
au point.
Groupe. Le contröle de la reception par acquittement des
messages de type SIMAN ou logistique est prevu dans la
Troisieme experience. norme. Cette methode a finalement ete retenue, aussi parce
Cette experience a consiste ä coupler des simulations de qu'elle ressemble ä celle utilisee dans de nombreux proto-
theatre mettant en oeuvre des simulations de systemes de coles C2. Mais les performances ne permettront pas tou-
defense antiaerienne avec leurs systemes C2, opposees ä des jours de respecter les tres courts delais utilises dans cer-
simulations de raids aeriens. Le but etait de repartir ces tains protocoles. Nous nous retrouvons encore une fois
simulations sur plusieurs machines pour augmenter la puis- devant la principale limitation de la methode, l'impossibi-
sance de calcul disponible par opposition ä l'utilisation d'une lite de transmettre des donnees evoluant rapidement.
machine centralisee. Contrairement aux experiences prece-
dentes ce couplage met en oeuvre un grand nombre d'entites.
33-10

le couplage a ete fait en utilisant 1UNTX standard parce de ces adaptations est specifique ä chaque exercice de
que le seul disponible sur les machines cibles, et au debut couplage.
nous avons releve de nombreuses pertes de messages. Les protocoles SID sont generateurs d'une forte charge de
Apres investigation nous avons decouvert que ces pertes calcul, les utilisateurs ont toutes les chances de devoir
n'etaient pas dues ä des collisions sur le reseau, mais ä des augmenter la configuration initiale d'une maniere ou d'une
tampons satures dans les couches reseau. La raison en etait autre. Une des conditions pur ne pas perdre de messages
que le partage equitable des ressources, principe de base est d'avoir une bonne reserve de puissance pour absorber
dUNTX, permettait ä la simulation de creer plus d'entites ä les pics de transmission. C'est ä dire que la configuration
chaque pas que les couches reseau ne pouvaient en traiter doit etre capable de traiter ä chaque pas nettement plus que
dans le laps de temps alloue par UNIX. Pour corriger ce le debit moyen attendu. II est ä remarquer que certains pa-
defaut nos avons module l'ordonnancement des taches de piers ecrits apres STOW-E recommandent de mettre en
maniere ä attendre la liberation des tampons avant de re- oeuvre des methodes pour empecher 1'emission simultanees
commencer un nouveau pas de simulation. L'emploi d'un
de blocs de PDUs.
UNK temps reel nous aurait permis de resoudre ce Pro-
bleme beaucoup plus simplement en utilisant les mecanis- sur un reseau local le facteur limitant c'est la puissance de
mes de priorite. Neanmoins nous devons rester conscients calcul et pas la bände passante du reseau,
qu'une simulation qui cree de nombreux PDUs induit for- l'emploi d'outils de visualisation temps reel est indispen-
cement une forte charge sur les couches reseau. sable ä la mise au point en temps reel,
le nombre de messages ä traiter par les couches reseau on doit pouvoir ajuster les allocations de ressources de la
pourrait aussi etre reduit en utilisant le mecanisme de con- configuration pour eviter les desequilibres entre täches.
catenation des PDUs. En general un PDU a une taille infe- Ceci ne peut se faire qu'avec l'aide d'un specialiste Systeme
rieure ä celle d'une trame IP sur Ethernet; on peut done
et reseau.
concatener plusieurs PDUs dans une meme trame et re-
duire ainsi le nombre de trames ä traiter.
REFERENCES.
ces simulations de theatre utilisent des pas de calculs Ref 1 .: A new PDU to reduce DIS network traffic Steve
relativement importants, typiquement 5 sec, mais lors du Seindsticker and Norman Lawler. 9
01
DIS
calcul de la distance de passage d'un missile de positions Workshop.
intermediaires sont necessaires. L'estime systematique d'un
pas sur l'autre ne permettait pas de calculer ces positions Ref 2 : Variable Scaling Time Implementation. Jean Pierre
intermediaires. II a done ete decide d'offrir une fonction de Faye and David Taylor. 10th DIS Workshop.
lecture de position estimee pour la date fournie par
l'utilisateur.

CONCLUSION.
Toutes ces experiences nous ont mene aux conclusion
suivantes:
- les protocoles SID permettent effectivement de coupler des
simulations dejä existantes sans obliger ä des modifica-
tions importantes de celles ci, pourvu que leur organisation
initiale ait dejä bien separe la generation d'environnement
de la simulation proprement dite. dans ce cas il est facile
de remplacer la generation originelle d'environnement par
l'environnement cree par les autres simulations. Neanmoins
les methodes SID ne resolvent que les problemes de
transmission de donnees, mais pas les problemes de cohe-
rence des bases de donnees terrain, des modeles de surface
equivalente radar ou des bases de donnees de signatures
restent ä resoudre.
- SID est une methode saine pour coupler des simulations de
grande taille et permet de remplacer un simulation par une
autre ce qui donne une grande souplesse d'emploi,
- il est necessaire de developper les adaptations necessaires
au filtrage de l'environnement pour l'adapter aux possibili-
tes de la simulation. Ceci ne peut etre fait qu'apres avoir
clairement defini les scenarios et evalue le trafic. Chacune
 34-1

HIGH FIDELITY, MOBILE, NET WORKABLE TRAINERS

THE TRAINERS OF THE FUTURE?
Chad C. Miller / David Perdue / Scott Davis
4.3.2.2/5.1.6.1 Mail Stop 3
Flight Vehicle Simulation Branch / Aircraft Simulation Branch
Naval Air Warfare Center
Aircraft Division
Patuxent River, Maryland 20670
U.S.A.

ABSTRACT MFS Manned Flight Simulator Facility

WST Weapons System Trainer
The U.S. Naval Air Warfare Center Aircraft OFT Operational Flight Trainer
Division Patuxent River personnel recently IOS Instructor-Operator Station
completed a high fidelity, low cost mobile APT Aircrew Procedures Trainer
prototype AH-1W Aircrew Procedures Trainer SMT Small Mobile Trainer
for the United States Marine Corps at Camp FPT Fleet Project Team
Pendleton, California. The use of recent TNS Tactical Navigation System
technological breakthroughs in both super-micro
computers and visual display technologies made PROBLEM STATEMENT
this program possible. These breakthroughs
pave the way for future OFT/APT developments. The current need of the military is a more realistic
The AH-1W APT project demonstrated the training at a much lower, more affordable cost.
application of evolving simulation technology in In face of decreasing budgets, and ever more
four areas. First, that new technologies reduce expensive weapon systems, the military cannot
the physical size of room required to support afford to continue to construct and deploy
computational requirements of WSTs. Second, traditional trainers except at the expense of
newer technologically advanced COTS removing funding from the actual usable weapon
computers can be used in a rugged environment system. The trainers of the future must support:
without any special expensive considerations. broad geographical areas, various realistic
Third, existing full-size training systems could be combat conditions, provide usable and realistic
supplemented or replaced by such small, mobile experiences with weapons, improve aircrew
devices such as the AH-1W high fidelity trainer safety, be mobile and relocate-able, and support
prototype constructed at NAWC AD. Fourth, mission rehearsal requirements.
fixed wing tactical aircraft training systems that
do not require motion cueing devices are Modern technology has produced the "Weapons
particularly attractive for this application. Systems Trainer" or the "Operational Flight
Trainer" in the U.S. Department of Defense as an
Future considerations should include the use of answer to the added requirements placed upon
the actual aircraft as a training system. This could simulation in training. OFTs are being used as a
be accomplished by aircraft that employ substitute for actual aircraft flight instruction for
computer buses for aircraft control. This paper basic flight training. This is an expansion of the
will survey the current possible mobile trainer role of simulation in military training beyond the
configurations, and examine the use of actual use primarily for emergency procedures training
tactically deployed aircraft as the basis for that was common only a few years ago. WSTs
training systems. are used for basic flight training, but are also
designed to do the additional functions required
LIST OF ABBREVIATIONS to train the student, or refresh the veteran, in
delivering accurately the large amounts of
ordnance carried by today's attack aircraft.
DIS Distributed Interactive Simulation WSTs have supplanted the actual live firing of
COTS Commercial Off-The-Shelf modern, expensive, weapons to the extent that
NAWC AD U.S. Navy's Naval Air Warfare many pilots who have not flown in combat may
Center Aircraft Division

This paper is declared a work of the U.S. Government and is not subject to copyright protection in the United States.

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?'
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
34-2

have never fired one of the more exotic weapons provide training on the new Tactical Navigation
that they are expected to be able to use in combat, System, and receive fleet comments. The goals
or at best are allowed to fire one or two such of the second prototype (currently under
weapons a year. development) will expand on the capabilities of
the first to include the new Night Targeting
Modern WSTs and OFTs do a good job of System and weapons delivery training. Fleet
training modern pilots for the threats and comments will also be generated on the second
scenarios that they may meet in a still-hostile unit and noted when the specification is written
world, but they share one flaw that has become for industry to generate more devices if required.
more apparent to air crews and training systems
personnel alike: a fixed site availability and huge The MFS approach to the problem began with the
cost. Current WSTs and OFTs are large, development of a Fleet Project Team. The FPT
building sized devices, often requiring dozens of consists of active and reserve Marine pilots and
support staff and technical crew members to MFS engineers. This program was initiated in
ensure daily availability. 1992. The system was designed to fit within a
mobile unit complex consisting of three standard
IDEAL DESIGN REQUIREMENTS mobile units with minor modifications. These
units house the deployed system. In addition,
The ideal design for a mobile, high fidelity trainer these three units serve as the shipping containers
would incorporate at least the following nine when the device is being transported. The device
features: design called for a three channel visual system
for the first device, and a four channel visual
1. Mobile, including air, land or sea deploy- system for the second device. The final system
ability design layout is show in Figure 1.
2. High fidelity cockpit and visual system
3. Reuse existing software models
4. Easily adaptable to configuration changes
5. Employ COTS equipment
6. Capable of operation by flight crew without
trainer system knowledge
7. Capable of operation using portable power
supplies
8. Capable of operation within a short time
period after a movement
9. Capable of training all procedures, both
normal and emergency, found in the
NATOPS, including night vision capability.

For the purposes of this paper, such a system

will be called a Small Mobile Trainer (SMT).
FIGURE 1
System Design Layout
CURRENT MOBILE SYSTEMS:
AN EXAMPLE Computer Architecture
The MFS facility is a high technology Navy asset The computer system chosen for the APT was
of NAWC AD Patuxent River. MFS was selected the Digital Equipment Corporation's VAX
for the prototype construction because the unique Station Model 4000-90, with the VMS operating
nature of the facility and the type of simulation system. The MFS has had extensive experience
work done there could translate directly into the with this type of system in past years, and the
AH- 1W APT project. modular software systems developed there
operate best with this environment. The DEC
The project requirements called for MFS to Model 90 has a Linpack Single Precision rating
design and build two prototype APTs. The goal of 12.91 million floating operations per second
of the first prototype was to prove the basic (MFLOPS). This rating compares to the current
concepts of a mobile high fidelity simulation,
 34-3

host machines of the AH-1W WST's 1.8

MFLOPS rating. To allow large amounts of
space for future expansion in input/output
requirements, it was decided that two Model 90s
with BIT3 Corporation's Turbochannel to VME
adapters would be used as the host machines for
each APT. FIGURE 2
Forward Display Field of View
A shared VME backplane serves as the common
Input/Output system for the APT and as the path This front display provides an 18.4° vertical field
to the shared memory. One of the BIT3 VME of view, with 9.4° of view up from the horizon
adapters contains a daughter card populated with and 9.0° field of view below the horizon, with a
2 megabytes of shared memory. The majority of 24.3° horizontal field of view. This provides
accesses are performed by the airframe VAX, so side fields of view (FOV) of 8.4° up FOV, 26°
its BIT3 VME card hosts the shared memory. down FOV, and 45.7° horizontal FOV. This
This minimizes VME bus utilization. A FORCE results in a total of 115.6° horizontal FOV.
68040 VME Single Board Computer (SBC) is
used to process DIS protocol data units and is The entire system operates on one 50 ampere
also located in the VME chassis. The SBC DIS 120/208 Volt, 3 phase 60 Hz external power
software utilizes one megabyte of the shared supply. The mobile unit complex cooling and
memory to communicate with the visual system heating equipment requires two additional lines.
software. In addition, the machines were This amount of power can be generated by
equipped with Military Standard 1553 Bus several U.S. Military standard mobile generators.
interfaces for communication with actual aircraft The APT mobile units contain power
hardware. conditioning and power interruption protection
equipment to allow operation and protection to
The host computers measure 40x20x60 the system from any quality of power supply.
Centimeters and weigh less that 75 kilograms
each and mount in standard computer racks The user interface consists of a US patented
located at the rear of the cockpit. This can be video switching and software system that allows
compared to the room full of computers required the IOS "point and click" control screens to be
to run the same software for the existing WSTs. displayed on the forward field of view at the
touch of the cockpit mounted IOS controller.
Visual System The training time to teach a novice aircrew how
to use this system for effective training has been
The requirements for the APT visual image found to be under one hour.
generation systems were:
Summary of the Prototype
• small physical size
• textured visual scenes Together with the above visual and computer
• ability to use existing WST databases systems, the prototype consists of an actual nose
• low cost per channel section of a stricken AH-IT that was cut down
• ability to produce night vision device imagery. and modified to fit within the mobile units. The
cockpit section was then populated with gauges,
The system selected was the Evans and switches and panels all of which were NVG
Sutherland ESIG-2000 image generation system, compatible. The aircraft hardware utilized
which fulfilled all the above requirements. In included the sticks, collective and control
addition, the system is capable of 60 Hz linkages, TNS, HUD, Full Functional Signal
operation for minimal transport delay, laser range Processor and the TOW/Hellfire Control. All the
finding, and up to 63 moving models. This hardware above combined made an exact copy of
system is capable of using areas of the visual the AH-1W TNS cockpit area.
databases developed for the WSTs.
When it came time to interface hardware and
software, MFS started with the WST software as
a proven baseline. The code was then rehosted
to run on DEC computers and to use the MFS lab
34-4

executive and simulation control software. Only

the models dealing with the new TNS had to be
designed and written from scratch.

Since its delivery in September of 1994, the AH-

1W APT has been scheduled 50 hours a week
and has seen less then four days down time.
This small amount of down time is due to a
modem that MFS incorporated into the prototype
to allow access to the simulation when support is
needed. FIGURE 3
Simplistic Databus Flow Path
Since the trainer has no motion base and proven
COTS equipment, the amount of knowledge A static aircraft can be converted into a training
required to maintain the device is rather small. system by the addition of a visual system and a
The device incorporates software driven daily host computer. The host computer would listen
readiness checks and hardware diagnostic tools. in on the bus traffic to detect pilot control
With these tools, the simulation is maintained movements and cockpit switch selections. The
with minor aircrew involvement. host computer would then, using high fidelity
models, perform the usual simulation processes
The Navy's prototype incorporates all the desired to calculate the new states of the aircraft and the
ideals of a portable trainer. Improvements to the required cockpit displays. This information
system could be made in the visual system by would then be put back on the bus for display to
relaxing the requirement to use the mobile units the pilot/trainee by the usual aircraft displays.
in an unmodified way, or by going to a helmet
mounted display system. The fleet has used the The information flow would then look like:
prototype for six months, and they have found it
to be a very effective and reliable system. States feed
onto bus
FUTURE THRUSTS/CONSIDERATIONS: Aircraft i L
j j States^
USE OF ACTUAL ASSETS AS TRAINING
DEVICES
.£ZAl displayed ? state change
4
Modern aircraft (and other) weapon systems are
■
becoming more complex and dependent on Aircraft Sim. state
Inputs
electronic data buses for displaying information. detected response change
When almost all interaction between the weapon
system and the operator uses a data bus it
\
becomes possible to design training systems that Simulation
use the actual aircraft themselves as the training of a/c
assets. The design concept and solution that MFS
has created is called "Lamprey," after the fish that FIGURE 4
attaches onto a host and remains there, using the Lamprey Databus Flow Path
host for movement and food.
The following paragraphs will further explain
The concept is best illustrated with an example, how a Lamprey type system would work and
in this case using an F/A-18C or new F/A-18E/F incorporate the basic design requirements for a
aircraft. Both of these aircraft use MILSTD 1553 mobile, high fidelity trainer.
buses for displaying the bulk of information to
the pilot. This bus is also utilized to read The Lamprey system would be highly mobile and
switches and aircraft control positions. In very be able to be deployed anywhere. The entire
simplified form, the flow path in the aircraft is system could be made capable of running off the
shown below: same portable "yellow gear" power carts that
power the aircraft. Note that the aircraft itself
would have to be electrically powered for this
method to work. The portable training system
 34-5

would then reduce to a compact wheel-mounted A Lamprey system would be able to reuse the
system that could be moved from aircraft to software models written for the OFTs/WSTs.
aircraft as the situation warranted. There would be little change to the aerodynamic
models. The avionics software models would be
If the aircraft was initially designed with loop incorporated so that the actual expensive
holes for Lamprey systems in the future then hardware on the aircraft would not have to be
most of the aircraft systems could be turned off used. The power to the aircraft would only be
during training. Most if not all the special used to drive cockpit displays, control positions
situations described above could be taken care in and the data-bus to the Lamprey system.
the mission computer software load. With such a
design, once the Lamprey system is attached to Configuration changes to the trainer would be a
the aircraft the mission load would recognize a very simple task. Since the trainer incorporates a
training session and override normal aircraft helmet mounted display and the trainer is bought
responses. Once the Lamprey is disconnected, to the cockpit, to change the configuration all that
the aircraft would respond in the normal matter. will have to be done is to run different
This design concept would remove the chance of aerodynamic, avionics and aircraft interface
mistakes that would produce a mishap. software models.

Since the aircrew will be using an actual flight The whole Lamprey system could be COTS
asset for training the cockpit will not have to be supplied. The power for the system could come
modified to increase the fidelity. The from the same source that is generating power
technologies for man/machine interfaces in for the aircraft systems.
aircraft are pushing for "glass" cockpits and
control of systems through a data-bus will keep a The Lamprey concept could be incorporated in
Lamprey system from having a large integration future designs of aircraft to handle the data flow
factor when the trainer is to be used. Until and integration more cleanly. If the basic designs
modern cockpits are fully integrated onto a data- of aircraft incorporate "loop holes" in mission
bus, Lamprey will need to interface to these computer loads and aircraft data-bus integration
instruments. If these instruments are critical for ports, the Lamprey system would have a cleaner
training, flat-panel LCD display devices could be interface design and not be considered as an after
clipped into place over the existing non- market add-on. Full testing of the aircraft and
functional displays, or the faces of the dials could trainer union could be accomplished to reduce the
be programmed into the helmet-mounted visual chance of mishap.
system to update as required. The system would
appear as illustrated:
Benefits of Lamprey System
There are five major potential benefits of a
Lamprey system. First, and most obvious, the
use of an actual tactical asset negates the need to
build a high fidelity and often expensive
simulation system. First, the system does not
require a unique piece of training gear, which
also removes the need for an entire infrastructure
system for support personnel, spares, training of
FIGURE 5 the support personnel, etc. To fulfill the need to
Lamprey System Concept
maintain the trainer the existing aviation support
The visual system for the illusion of movement personnel could be cross-trained on how to
could be presented to the pilot with a boom- remove defective components from the boxy
supported helmet mounted display system. This training device and return them to a depot for
could to be attached to the box that housed the repair.
host computers and 1553 bus interfaces. There
will most likely be display helmet sensor Second, In practical terms, there is usually one
positioning and interference issues with the aircraft per squadron or airwing that is "down"
aircraft's canopy, but these should be workable. for extended repairs, the infamous "Hangar
Queens" or even a series of aircraft that are in a
34-6

non-flying status for a variety of reasons. If the airframe's actuators we insert a modification to
malfunction or failure of the aircraft was not the aircraft that causes the control surfaces not to
serious enough to affect the cockpit control respond when the pilot moves the control stick in
systems, an aircraft laid up for repair could be the aircraft when the Lamprey is attached.
used for the Lamprey device, making lemonade Obviously, leaving this training modification in,
out of lemons. This would turn a downed asset or inadvertently turning it on in flight would have
into the trainer of the week. disastrous consequences.

Third, and perhaps as important as any other Another problem is that there are some systems
function, the Lamprey could be used to help that could simply not be allowed to operate in
repair crews isolate a malfunction, and reproduce their normal manner without mission and fight
malfunctions seen in flight on the ground. The control computer loads changes. Using
Lamprey could be set to run a series of canned examples again, it would be unfortunate if a pilot
maneuvers "through" the aircraft, or even replay training in take-off succeeded in raising his gear
a recorded series of actual flight evaluations, to or a comrade succeeded in dropping a bomb
help reproduce and/or isolate the problem with training for a strike package while sitting in the
the aircraft. As a last method for fault isolation, hangar space of a carrier. The host computer
the pilot who reported the problem could fly the would have to recognize the command on the bus
system to the point where the problem occurred. "raise gear" and over-ride the actual aircraft
This capability would, of course, be limited to systems response, not allow the gear to
debugging electrical systems and hardware only. transition, and then display to the pilot that in his
virtual world the gear had moved.
Fourth, a device such as the one being outlined
would be a very cost effective way to profile Finally, while such a system could be easily
missions in a squadron or group setting. The integrated into an aircraft that is data-bus driven
major push for the next generation training and fly-by-wire, the system becomes harder and
system will be to accomplish mission planning harder to integrate into an aircraft that does not
and rehearsal. If enough devices exist, groups of use such systems for display of information. A
aircraft could practice maneuvers using the net- good example would be a 1960's designed AH-
workable nature of the device and include the 1W attach helicopter, which uses data-bus
latest information on the target's position and information only for a Heads-Up Display, with
capabilities. The addition of a photo-textured all the other displays being driven electro-
mission rehearsal visual system (with the mechanically. For a system such as this, most of
incurring of some large additional cost) would the aircraft's interior would have to appear as
add the ability to perform actual strike package virtual reality systems in the image generator,
rehearsals using the actual aircrew and aircraft which would drive costs up sharply.
that will fly the mission.
Costs from both type systems should range from
Finally, a Lamprey type system would be sizable $5 - $20 million per copy. The operational cost
to the mission desired. A pure cockpit of the Lamprey system is estimated at
procedure's trainer need not even have a visual $400.00/hr. The costs break down to
system, and this represents the lower end of the $200.00/hr. Escrow for repairs to aircraft
systems possibilities. Visual image generators hardware, $50.00/hr. for the pilot, and
could then be provided as required, up to the maintenance personnel, and $100.00/hr for
mission rehearsal type devices mentioned above. electrical power.
Drawbacks of the Lamprey SYSTEM USAGE EXAMPLE: MISSION
REHEARSAL
There are drawbacks to this type of design, and
some are potential "show stoppers." In the field the concept could be used as follows.
A deployed force of mixed aircraft type has three
One of the basic laws of any Lamprey type Lamprey type systems for fixed wing aircraft,
system should be: The flight asset cannot be and three SMTs for rotary wing aircraft. This
modified in any way. This would preclude any scenario assumes the usage of a photo-visual,
unfortunate events that might occur. As an rapid turn-around visual system and a mission
example, perhaps, to save wear and tear on the requiring a mix of aircraft types.
 34-7

Both systems share the advantage of being able

Days, weeks or even hours before the mission, to make use of today's advanced visual systems
reconnaissance assets over fly the area leading to and networking capability for mission rehearsals
the objective. The results are incorporated into over the actual tactical area with the actual flight
the visual systems for the SMTs and the Lamprey crews required, including, if desired, command
type devices. Planning begins at once. Using and control groups.
three squadron aircraft, the Lamprey systems are
attached, and the flight leaders use the system, SUMMARY
along with all the usual materials available to
them, to plan and profile the mission. The flight The technology exists today to produce a new
leaders of the rotary wing aircraft are performing type of high fidelity trainer. All the components
similar functions. of the systems discussed in this paper exist and
are off-the-shelf available. These trainers can
Once the mission is planned, flight leaders begin either be independent, mobile systems that move
leading their teams through the mission, linked to the point of greatest need, or be "strap-on"
and interacting only with similar aircraft types. systems that attach to actual tactical assets.
Usable in the field, on networks, and with
Finally, the entire mission group trains together, mission rehearsal visual systems, such systems
will all participants planning and training in can supplant or replace existing systems in usage
rotation through the mission until all pilots for low cost and with short development times
involved have practiced in a linked, interactive required. Small mobile trainers and strap-on
environment. Simple workstations allow training systems can meet the needs of today's
command and control forces to interact with the military both in the field and in the area of fiscal
group in real-time. reality.

The result is a mission group that has flown the BIOGRAPHY

mission many times in a virtual world before the
actual mission is undertaken. Command and Mr. Miller is the team leader for the tactical
control procedures have been rehearsed, as have aircraft simulation section of the MFS facility, a
problem scenarios and tactics. Of course, no section that encompasses the Navy's F/A-18,
such scenario can anticipate totally the response F/A-18E/F, AH-1W and EA-6B simulation
of the other side, but such a training session efforts. He received his B.S. degree in
could even be linked with the best aggressor Aerospace Engineering from North Carolina
doctrine and tactics models to produce a training State University of Raleigh, North Carolina,
environment that may be capable of closely U.S.A. in 1986. Mr. Miller has been a member
anticipating reality. of the AI AA since 1984.

COSTS AND BENEFITS Mr. Perdue is the senior Technical Specialist for
the Manned Flight Simulation Facility. He has
On the positive side, SMTs offer mobility, served in this position as the principal designer of
excellent training value, low cost, on-site the Manned Flight Simulator since 1986. He
availability and a mission rehearsal capability. received his B.S. degree in Electrical Engineering
Drawbacks include a lack of motion cueing from Virginia Polytechnic Institute and State
systems, space requirements onboard ship, and University in 1975, and his M.S. degree, also
the burden of additional support requirements for from VPI, in Electrical Engineering in 1976.
squadrons.
Mr. Davis is a lead avionics engineer for the
Lamprey type systems have as benefits the Aircraft Simulation Division of the Manned
following: re-use of expensive tactical assets, on- Flight Simulation Facility. He has been involved
site availability, maintenance usage and superior in the development of the AH-1W APT and the
realism and fidelity. Drawbacks include the V-22 GTPT simulations. He received his B.S.
following: Use of an expensive tactical asset, degree in Computer Science from Frostburg State
lack of motion cueing, and the technical University in 1988, and his M.S. degree from
complexity required for the visual image Florida Institute of Technology in Management
generation system. in 1990.
 35-1

SIMULATION, DISTRIBUTED SIMULATION,

and
SYNTHETIC ENVIRONMENTS
Future Applications and Challenges

Charles E. Adolph
Dr. Jack Thorpe
Science Applications International Corporation
1710 Goodridge Drive
McLean, Virginia 22102

SUMMARY
This survey paper reviews current practices in One of the objectives of this survey paper is to
simulation technology with emphasis on aircraft help the Flight Vehicle Panel develop your
development and testing, followed by a Aerospace 2020 concept paper. In this paper, we
discussion of present and future applications of briefly review current practices in simulation
distributed simulation and synthetic environments. technology with emphasis on development and
The aviation community has used a wide variety testing, then address the use of distributed
of simulation tools for over 35 years. The simulation and synthetic environments to support
benefits of current generation standalone the requirements, development and training
simulation tools, including system integration processes. We address these issues in a broad
laboratories, are well understood. Distributed context including, but certainly not limited to,
simulation is beginning to be used to support aircraft development.
training and requirements definition; test
applications are also beginning to be explored. This is a dynamic time for both government and
Advanced distributed simulation provides an industry. In the past few years, we have seen the
unprecedented opportunity to explore new Warsaw Pact and the Soviet Union disintegrate.
concepts as well as improve the efficiency of test These events have properly triggered a significant
and training activities.With the rapid increases in decrease in defense spending, but most people
information processing technology, synthetic recognize that there is a need to continue to
environments will be used increasingly to support maintain an effective military capability against a
test and training, both in standalone and range of regional threats. In addition, military
interactive distributed modes. There are forces are being called upon to perform a
numerous challenges in developing and applying spectrum of non-traditional roles, including
these technologies; including simulation fidelity humanitarian support activities, evacuation
and scalability, verification and validation. There operations, disaster relief, peace keeping, counter-
is also a need to decrease simulation costs. The terrorism and counter proliferation support.
emergence of distributed simulation has been the These new roles, when combined with a wide
catalyst for increased interaction between the variety of potential regional threat scenarios,
simulation world and test and training systems. makes the tasks of requirements definition,
Next generation range and platform testing, training, tactics and doctrine development
instrumentation architectures will be designed to far more challenging than in the past. Simulation
facilitate interaction between live players and technology can be a powerful tool in meeting
simulated entities. these challenges. This paper surveys the status of
these tools and addresses future applications with
INTRODUCTION emphasis on advanced distributed simulation
Good afternoon ladies and gentlemen. It is a (ADS).
pleasure to be here with you today. Over the past
several days speakers have reviewed the state of USE OF SIMULATION FOR AIRCRAFT
the art in flight simulation and simulations used DEVELOPMENT, TEST
for engineering and training. The concluding The aviation community has used a variety of
session of this symposium properly addresses the simulation tools extensively in support of system
broader issue of Future Applications of development and test and evaluation for at least 35
Simulation Technology. years. Simulations and physics-based models are

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?'
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
35-2

used at the subsystem level, in standalone, accommodate a fighter aircraft, an electromagnetic

engineering-level simulations as well as for warfare integrated systems test laboratory, an
mission-level assessments. Fixed-based and electromagnetic environment generation system,
motion simulators are key instruments in flight and a manned flight simulation facility. The
control system development. This Flight Vehicle ACETEF has been linked to most of the other
Panel symposium reviewed the state-of-the-art in facilities listed above, as well as a variety of other
flight simulation, visual and motion cueing, as simulation facilities throughout the U.S. It is
well as advances in simulation for aircraft design rapidly becoming a node in a variety of distributed
and development. One topic not previously exercise and training simulations, as well as a
addressed during this symposium is the increasing system development and test facility.
use of system integration laboratories in support
of aircraft development. These facilities, which It should be noted that the above facilities were all
are defacto complex hybrid simulation systems, originally developed to operate solely in a stand-
include actual aircraft subsystem hardware, alone mode. With the advent of distributed
subsystem models, and system stimulators. A simulation, most of these facilities are now being
system integration laboratory, located at the netted to operate in a mode not envisioned earlier;
airframe manufacturer's facility, has become an in a linked mode as well as stand-alone. They are
integral part of the development of each new becoming nodes in a number of on-going and
aircraft model. Integration laboratories are planned distributed simulation activities, which
sometimes located at the avionics integration are discussed later.
contractor's facility as well. Earlier generations of
avionics consisted of a number of federated Based on years of experience, the benefits of
subsystems, some of which were pod-mounted. using system integration laboratories and
The F-22 design is the first production application hardware in-the-loop simulators for development
of an integrated avionics suite on a new aircraft. and testing are well understood and accepted. The
With this generation aircraft, integration majority of the subsystem integration and
laboratories are becoming increasingly important checkout tasks can be accomplished in a system
in the development of extensive and complex integration laboratory. When compared to flight
interfaces, distributed on-board as well as off- testing, the process is far more efficient, the
board, including aircraft-aircraft links, ground and experiment is controlled and test execution is
satellite links. rapid, relatively simple and inexpensive. The
inefficient fly-fix-fly approach is minimized.
There is another set of generic simulation- Simulation has also long been used in direct
intensive aircraft system development and test support of flight testing in the following ways:
facilities which have evolved to support the • Pre-test analysis, scenario development
development of aircraft electronic warfare
• Test planning
systems, weapons sensors and guidance systems.
Some of the major facilities are listed below. • Crew test procedure rehearsal
• Air Combat Environment T&E Facility • Post-test evaluation
(NAWC, Patuxent River, MD) • Anomaly investigation
• Electronic Combat Integrated Test Facility
(Edwards AFB, CA) SIMULATION CATEGORY
DEFINITIONS
• Guided Weapons Evaluation Facility (Eglin With this background, we will transition to
AFB, FL) emerging developments in distributed simulation.
• Theater C2 Simulation Facility (Kirtland AFB, Some definitions and concepts are useful aids to
NM) the discussion. First, a 1992 U.S. Defense
• AF Electronic Warfare Evaluation Simulator Science Board Study on "Simulation, Readiness
Activity (Ft Worth, TX) and Prototyping" defined categories of simulation
• Real-Time Digitally Controlled Analyzer as follows:
Processor Activity (Buffalo, NY) • Live: Operations with real equipment
in the field
The Air Combat Environment Test and Evaluation • Virtual: Systems and troops in
Facility (ACETEF) is the most versatile and simulators fighting on synthetic
advanced electronic combat integrated test facility. battlefields
It features an anechoic chamber large enough to
 35-3

• Constructive: Wargames, models, analytical ...that can be entered efficiently from remote
tools locations (no need to travel to a gaming
The Defense Science Board observed in the study center or exercise range)...
that everything is simulation except combat. This ...by large numbers of warfighters...
is a more global view of simulation than the more ...at a high degree of granularity and fidelity if
widely held construct which makes a distinction required...
between actual hardware and a system which is ...for a variety of applications and purposes
replicated in some fashion. This broader (training; mission planning; dress rehearsal
construct has implications for system development operations, command, control; after
and test. We will discuss these implications in mission review; concept development;
view of emerging simulation technologies; in requirements definition, simulation based
particular advanced distributed simulation and design; engineering development test and
improved synthetic environments. evaluation; force structuring; tactics &
doctrine development; etc.)...
ADVANCED DISTRIBUTED ...where commanders can use operational
SIMULATION OVERVIEW communications and operational systems
Advanced Distributed Simulation (ADS) may be (actual command, control, intelligence
defined as the set of technologies used for creating systems)...
realistic synthetic environments that can be entered ...outcomes can be visualized, inspected,
from worldwide locations in real time. Advanced analyzed, and understood better than before
Distributed Simulation (ADS) is not a single thing ...and results can be communicated to senior
or model that can be rigorously specified and military and political decision makers using
built, but rather a suite of constantly improving more effective and vivid visualization
tools and techniques for assembling dynamic and techniques which characterize performance
disparate federations of models and simulations as effectiveness and limitations.
well as real platforms and real systems of all types
and sizes that interoperate together. Interoperation ADS is based upon distributed computing
is the single most critical feature of ADS and structures, so the key enabling technologies
represents a dramatic departure from earlier include the rapidly increasing power of
generations of models and simulations that were workstations, microprocessors, data bases,
incapable of working together, indeed were never software tools, and emerging nationwide and
thought of as needing to work together. worldwide digital networks that tie all of these
The fundamental change in the last ten years is the together. With their new processing power it is
realization that modern models and simulations are possible to support the requisite realism in graphic
based on the same underlying information rendering and visualization of the simulated
technologies and can and should be designed to be environment to interact effectively with a human
connectable as a guarantee of interoperability and in real time. ADS has also been a catalyst for
compatibility. This change will result in a more good information technology practices such as
generic approach to modeling and clearer common data dictionaries and open architecture.
responsibility for ownership of system/weapon
characteristics and effects (the algorithms). This With that overview, we will review several charts
is a more cost effective approach for customers which illustrate the concept. Figure 1 illustrates a
since fewer redundant (and conflicting) diverse set of players interacting on a distributed
simulations should result. This will , in turn, network which provides the battlespace
alleviate current verification, validation and environment. The players broadcast and receive
accreditation problems. There has been state data from the network. Figure 2 (AWACS
increasingly widespread recognition of this example) illustrates the ability to mix and match
change, especially by the DoD Defense Science real and virtual players in an ADS environment.
Board and the science and advisory boards of the The virtual players could either be created on-
military Services. board the AW ACS, or uplinked from the ground
from a virtual AWACS. In a beyond visual range
ADS offers the potential for fundamentally new environment, virtual players can be injected
capabilities: seamlessly. To illustrate the flexibility the concept
...The creation of very large, complex provides, assume that a training mission is
battlefields... scheduled and the AWACS has to cancel or abort
for some reason. Segments of the mission could
still be performed from the ground-based virtual
35-4

AWACS, which could be patched into the field tests for a variety of reasons including cost,
exercise initially in a parallel passive monitoring safety, availability, logistics, and environmental
mode. The implementation of this concept conditions. Applications include:
requires that ground based systems be built to the
same architectural standards and have functional • Adding assets
compatibility with the airborne system and
associated simulators. Friend, threat, unidentified
Two-way interface between virtual and
Figure 3 illustrates the options associated with the live forces beyond visual range
use of sensitive or classified models. Assume the • Expanding scope, events, trials
task is to evaluate an aircraft with low observable
characteristics against a potential threat acquisition • Near real-time endgame analysis
or tracking radar. For a mission planning system, Leveraging human factor/live response testing
one would need access to detailed models of the • Improving environmental robustness
aircraft as well as the potential threat radars (top
illustration). However, if the aircraft were one of Natural
several types used in an exercise for which access Weather, diurnal effects,
to the specific aircraft characteristics was not obscuration
allowed, the simulation would only provide bi- Man made
modal detect/non-detect information based on the
aircraft's attitude and bearing from the radar. This EW, EMI, target signatures (IR, UV,
information could be altered further as appropriate noise, etc.), obscuration
to provide protection against revealing actual Terrain
detection ranges. Topography, cultural features
With that brief overview of the advanced Manmade features
distribution simulation concept and a look at some Terrain changes: military action,
sample applications, it is worthwhile to make weather
some assumptions. First, it is assumed that next
generation military leaders will be increasingly Within the next decade, advanced distributed
comfortable with and supportive of simulation simulation will have relatively few applications to
technologies. Second, military commanders will system development and testing when compared
be receptive to "productized" ADS that they can with training applications. However, many of the
use "on their watch" to solve current problems. same simulations will be used in a standalone
Third, next generation test and training mode for test support. The focus during system
instrumentation will be ADS compatible. On- development and test is properly in the actual
board interactive embedded training subsystems hardware/software being developed. The
will be incorporated during production of next requirement to connect geographically dispersed
generation aircraft, ground vehicles, and ships. elements in near real-time to support development
The initial focus of ADS will be on training and is limited. As has been previously mentioned, a
tactics involving large numbers of players. wide variety of simulation and system integration
Emphasis will be placed on: visual environment; laboratory tools are used to support development.
human in-the-loop activity (generating The majority of these tools are interactive, but are
requirement for near-real-time); and keeping not normally distributed.
bandwidth down to accommodate many players.
The following list contrasts some of the
1 differences in simulation requirements between
Synthetic environments will be used increasingly
to support test and training, both in standalone test and training.
and interactive distributed modes. Simulation • Testing may not require many players
offers the potential to provide representative test
environments that cannot be created physically for • Test accuracy may require larger bandwidths
• Testing requires high fidelity physics-based
1
models of individual entities
Some members of the simulation community restrict
the term synthetic environments to natural physical • Test data rates are generally higher
environments such as weather and topography. It is • Some test applications do not need human-in-
used in a broader sense in this paper to encompass loop interaction
any element that is modeled or simulated.
 35-5

• Near real-time may or may not be a test • Cultural: Acceptance by test community
requirement
• Architectures and data structures may have to There are some special challenges relating to the
be test/event specific use of ADS in support of operational testing. In
general, the operational test community is more
• Testing is "scripted" vs training "free play" suspicious of models and simulations than are
material developers, developmental testers, and
The emergence of advanced distributed simulation trainers. However, as resource, safety and other
is the catalyst for increased interaction between the constraints become more restrictive, the
simulation world, test and training. This has operational test community will be forced to make
major implications for next generation range more use of simulation tools. As an example, the
instrumentation and simulation systems as well. number of live missile shots required to achieve a
Range and platform instrumentation architectures high level of statistical confidence may not be
will be designed to facilitate live player and affordable. As a result, simulations will be
simulation interaction. There will be increased increasingly used in conjunction with actual test
commonality between mission control displays results. Operational test information will be one
(test and training) and simulation displays. Data data source to independently verify, validate and
from live players will be merged with simulated accredit (W&A) the simulation. The operational
friendlies and threats. This has command/control test community may also be reluctant to use ADS
implications, which will be addressed later. Post- tools, primarily because of the additional W&A
mission display /debrief systems will be used to burden associated with the elements of a
support simulation verification and accreditation distributed simulation not under their control.
activities, as well as their normal functions.
There are several command and control challenges
Figure 4 (Simulation Applications) summarizes that must be addressed if the objective is
the relative utility of stand-alone and distributed "seamless" integration of live and virtual players.
simulations to support system requirements There will be a range safety requirement for
definitions, developmental and operational testing, positive differentiation between real players and
training and tactics and doctrine development. virtual entities for range safety, air/ground traffic
This chart is constructed to compare the scope of control and search and rescue purposes. Rules of
applicability of standalone and distributed engagement will require attention. This is a
simulation tools. "Low" is meant to convey a potential constraint on system architecture, which
relatively limited number of applications rather will dictate at least three data paths; one for range
than a value judgment about the applications for safety command/control and one for individual
which it is used. "players," and one "player" command/control
which will be further subdivided into friendly and
ADVANCED DISTRIBUTED adversary command/control.
SIMULATION CHALLENGES
There are a number of challenges in developing Threat representation challenges include the need
and applying advanced distributed simulation. to simulate numerous potential regional threats in
Technical challenges associated with test a wide variety of threat scenarios. Because of the
applications include: potential for numerous combinations of red, gray,
• Scalability blue threat arrays, it is impractical/impossible to
• Fidelity test/train across the threat spectrum. Synthetic
environments provide an opportunity to alleviate
Higher resolution entities this constraint.
Integration of live, virtual and constructive
entities Finally, the variety and scale of joint task force
Threat emulation using distributed scenarios in a NATO environment make it
signature models impractical to consider staging large exercises as
the principal means of evaluating new systems or
Terrain/environment representation tactics. By using ADS, synthetic environments
• Data rates, latency, update rates and scenarios can be created to evaluate the
• RF data links for virtual threat injection effectiveness of a new system, doctrine, or
potential threat scenario.
• Verification, validation, accreditation
• Lower cost
35-6

KEY ACTIVITIES/DEVELOPMENTS technology and will include stimulation of on-

Next we will give a brief overview of several key board platform sensors.
activities and developments that are underway in
the U.S. which involve advanced distributed The objective of the Joint Advanced Distributed
simulation and synthetic environments. This Simulation Joint Test and Evaluation is to
overview will provide some insight into the investigate the utility of Advanced Distributed
diversity of applications of these emerging Simulation for both developmental and operational
technologies. testing. Two major test activities are currently
• Synthetic Theater of War planned: a systems integration test involving an
advanced missile system; and end-to-end theater
• Joint Tactical Combat Training System battle arena systems test. The objective of the test
• Joint Advanced Distributed Simulation Test activities is to evaluate ADS technology, rather
and Evaluation than test the systems involved in the experiments.
• Ship Self Defense Synthetic Test Range
The objective of the Ship Self Defense Synthetic
• Joint Precision Strike Demonstration Test Range is to develop a synthetic environment
where the ship self defense problem can be
Each one of these activities could be the subject of assessed in a manner that preserves tactical and
a detailed discussion. Our purpose here is only to technical details. The demonstration will include
briefly summarize the activity to provide some both real systems and models. The goal is to
insight into the breadth of developments relating examine system performance and operation in a
to advanced distributed simulation. variety of high fidelity scenarios.
The purpose of the Synthetic Theater of War The goal of the Joint Precision Strike
(STOW) is to provide the ability to construct a Demonstration program is to introduce and
variety of synthetic environments to perform implement new technologies into the defense
several functions including training and readiness, arena that address the precision strike mission. A
doctrine development, requirements definition, live-simulated virtual environment is being used to
and equipment design and prototyping. These evaluate technologies, train users, and perform
functions will be performed primarily with virtual experiments necessary to reduce sensor-to-shooter
simulations with the option of integrating live timelines and to attack high-value, time-sensitive
forces and constructive simulations. The STOW targets. Whenever possible, JPSD uses the actual
goal is to develop a suite of simulation hardware to conduct these demonstrations.
technologies including behaviorally accurate
computer generated forces, high resolution CONCLUSIONS
environmental representations, and distributed The next decade will be exciting and challenging
networking technologies. as the NATO defense community continues to
adjust to profound political and budget changes.
The objective of the Joint Tactical Combat There are potentially new adversaries, less
Training System (JTCTS) is to provide an at-sea formidable but also less well defined. Simulation
combat training capability for an expeditionary technology will play an increasingly important
force and a land-based capability for air crew role in requirements definition, testing, training
training. It will include both a fixed capability and tactics, and doctrine development. Advanced
which will be used to reinstrument existing air distributed simulation provides an unprecedented
combat training ranges and a mobile capability. opportunity to explore new concepts, command
One of the objectives of the JTCTS program is to and control systems, and for joint development of
provide realistic threat representation via data links tactics among the NATO member nations.
to uplink air defense and other threat information. Experimentation is needed to mature the
The cost of maintaining and operating large technology and fully explore the concepts. The
aggressor aircraft units and ground-based air challenge is to be innovative in taking advantage
defense threats is becoming prohibitive for of the powerful simulation tools that are becoming
training purposes. This will be the first large available.
scale training application of distributed simulation
35-7

g
35-:
 35-9

IkkAutrnkW^^^j!^

u
*-■
a> QOC
j-T
CD CO
■a sz
o CO
a>
di
»IWTHllWWmfSP™ » o

1_
imii *2
o
o
(0 c 0)
a> Q
c
o 'S (0 i
y-. ■-■■'.'.v™ (0
0)
w CD
o "U , *•*

Position
0. o (0
0)
6
lli§§il:i r
' en
W

E5

a
35-10

Q
w
H
13 en
PQ
1—1
< HJ HJ i
1
S
£
H HJ s
^ 0
o
£
s
ELITY
DIUM

5 w
£ W Q
s HJ hJ ffi PC
fe
0
hJ
2E
^^1
<
1

D
o <
s
HIGH
FIDELF

C/5 ffi ffi S J 1II

3 s
V

i
CO (X)
H
l-l

HJ
<
II

z
So o (3
VELOP

ERATH
QUIRE
FINITI

STING

STING
STEM

z
h—1
H
h s?
>H w w S w PH LÜ tf < O w
5c^D D H O H H hDQ
 36-1

THE SYNTHETIC ENVIRONMENT - THE ULTIMATE DEFENCE SIMULATION?

Peter Beckett
British Aerospace Defence Ltd
Military Aircraft Division
Warton, Preston PR4 1 AX
United Kingdom

SUMMARY G Assess the impact of Force Structure changes

References to 'Defence Synthetic Environments' are becoming O Assess alternative methods of warfare
more common with the passage of time. At present, the term D Develop doctrine, tactics and mission plans
means different things to different people driven largely by C Rehearse joint missions
their viewpoint within the defence business. Indeed, to confine □ Train force commanders and personnel
the concept to the defence business is already narrowing the G Develop and assess products, needs and concepts
vision according to some. There are envisaged opportunities Q Allow concurrency between customer, user and provider
for applying similar techniques in commerce, medicine,
The human content of the SE is viewed as one of the essential
transport and so on. In fact any large system is a potential
ingredients. There are a number of ways in which the human
candidate. interacts with the SE including training for military decision
The defence business was, however, probably the first to adopt making, training for weapon system operation or investigating
the name and to start demonstrating what it means. In the UK, doctrinal options. In other parts of the SE, the human is
joint government and industry groups have been established to making product design decisions. Throughout the SE the
steer national activities for SE's. A number of early common theme of human interaction with computers, real
demonstrators are underway. In the USA, regular networking equipment and other people occurs.
demonstrations illustrate how this particular component of SE
Clearly, a tool which can meet all the above defined
is maturing. The general impression is one of commitment to
requirements has vast scope. Each user of the SE would
SE as a long term goal.
actually use subsets for his own purposes whether they be
The intention of this paper is to give a view of the extent of training, requirement capture or whatever. Because subsets of
Synthetic Environments (SE) and to discuss how it may be put a shared environment rather than independent sets would be
together and used to help the defence business. In doing this used, as is the tendency to date, a consistency occurs which
the role of Flight Simulation within SE is highlighted along widens the scope of computer modelling that can be
with the challenges which it faces. undertaken. The new levels of integrity produced by the SE
modelling process will significantly increase the value of the
The need for improvements throughout the defence business
results produced.
hardly needs justification here. Pressures created by falling
defence budgets, uncertain threats, shorter development times Simulation and modelling is already widely used within the
and heavy competition all point to the need for process defence business in many of the areas covered by the above
improvements right across the board. To embrace the definition. Currently the techniques are well developed in
opportunities offered by SE requires changes in all the parties certain areas but far less so in others. It is also carried out in a
involved including the customer. The resulting increase in fragmented rather than an integrated way. SE can be seen as
effectiveness and level of concurrency between all the the 'umbrella' which hosts this integration.
stakeholders should be a worthwhile payback.
OVERVIEW OF SE HIERARCHY
DEFINITION OF DEFENCE SYNTHETIC A way of visualising the hierarchy of processes to meet all the
ENVIRONMENT defined requirements is shown in Fig 1. These layers serve
Currently there are a number of variants on the definition, different purposes and in many cases have different users.
primarily because of the different viewpoints held by those They are arranged in this way to emphasise the 'vertical
defining it. In principle, though, they are similar. integration' or consistency between them which allows data
representing weapon systems to move freely between layers.
For the purposes of this paper, the following definition is
proposed. The upper levels of the SE tend to relate to user requirements
derived from defence commitments. This is where customer
A computing based environment which allows the risk
defined geopolitical scenarios would be input to the SE.
reduction of defence issues to take place through the
interaction of people, models and live equipment in order to: The strategic models are the highest level of wargaming -
probably at the campaign level. The next layer provides

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?'
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
36-2

indicative costs for such a strategy and identifies the logistic based on aggregated entities and computer generated entities
support implications. such as aircraft. It is probably inappropriate to include the
man-in-the-loop effects except as a result of the 'vertical
Moving down through the hierarchy, the models become more
integration' process discussed later.
specific weapon system simulations either as aggregated forces
or individual platforms. At this central level there are a SE for Force Structures and Defence Policy Analysis
number of uses for the SE including requirements capture, Vertical integration must continue to higher levels still, if the
performance evaluation and personnel training. full benefit of SE is to be obtained. Although it is not
envisaged that simulators would play a direct role in this high
As the lower levels are reached, the solutions to requirements
level use, it is worth considering how it operates in order to
are being produced. These can feed back to higher levels so
complete the picture. In this mode SE can address high level
that vital cost/performance trade-offs can be done. Flexibility
questions regarding strategies and the balance of forces
of movement through the levels is the key to the benefits of
against an overall commitment. Fundamentals such as doctrine
theSE.
can also be investigated. The users of the SE in this case are at
the highest levels of command. They use the SE to experiment
THE SE USERS
with interactions so that the sensitivities emerge and can be
In order to illustrate something of the degree of integration
understood. New concepts of operation can be tested assuming
which is required for SE, it is appropriate to consider the
either new or existing equipment.
different types of user and how they interact with the SE
processes. Much of the work is done using the SE for operational analysis
where conflicts of significant size can be modelled and the
SE for Forces Training and Mission Rehearsal
interaction of joint forces can be considered. The SE should
The core of this SE subset is shown in fig 2. The task is to use
also be capable of producing cost indicators for specified
the models and equipment within the SE to improve the
conflicts.
performance of personnel. This can take place at the operator
level where skills in using particular equipment are to be Whether or not models at this strategic level prove to be
acquired or at commander level where the training is aimed at reliable depends upon the assumptions upon which they are
decision making. based as well as Verification, Validation and Accreditation
issues (VV & A) discussed later.
The process assumes mission plans, doctrine and the current
definition of training procedures. The real-time simulations hi their simplest form, however, these may be little more than
and wargames are used by personnel and commanders to computerised battle tables allowing the thoughts of the user to
improve their various skills. During the process, measures of appear graphically. Even so, there is scope for data capture to
operator training effectiveness and mission effectiveness are take place which could possibly be later used to reinforce the
produced and analysed. Shortfalls in training effectiveness are assumptions feeding into other levels of modelling.
corrected by changes to training procedures. SE for Equipment Acquisition and Evolution
Using the SE for Mission Rehearsal is a special case of its use The same core processes are used for acquisition or evolution
for Training. The emphasis rather than on skills acquisition is of a weapon system capability as for the training application.
instead placed on familiarisation with a particular pre-defined Fig 3 shows that instead of concentrating on training
scenario and tactics which have been produced by a separate procedures and analysis, the measured outputs are concerned
Mission Planning activity (see next section). with weapon system capability. But the primary difference in
this use of SE is that the simulations must be able to accept
In order to produce a realistic training task for operational
changes or re-definitions rapidly and accurately in order to
training or mission rehearsal, it may be necessary to network
allow the options to be analysed for effectiveness.
the realtime models or equipment into large scenarios both to
increase the realism and/or to allow simultaneous training. This is shown by the 'Design Definition' input which may
well have many iterations during the assessment of a product.
Large networks of geographically separate simulators are often
In this way the SE can be used to rapidly converge on a
referred to as SE's. However, simulator networking is just one
defence solution to meet the given requirements.
of the many enabling technologies which are used within the
SE. Fig 3 simplifies the process by showing a single function
which derives costs, timescales and life cycle support
The linking process between wargames and simulators is an
implications from the design definition. In reality this is a
example of vertical integration. Ref 1 discusses the principles
substantial, industrially based part of the process which runs
of vertical integration.
in parallel with the vehicle or system performance assessments
SE for Mission Planning taking place on the simulators. By this means, not only is
It is unlikely that the 'what ifs' to be tried in order for mission system effectiveness being addressed but also cost
plans to be refined would be carried out with individual effectiveness, probably the parameter of greatest importance to
simulated entities such as manned simulators. More likely the SE users.
these trials would use the established wargame techniques
 36-3

FLIGHT SIMULATION WITHIN SE further worsening in cues if the DIS links are over large
Earlier discussions have indicated the roles of SE in which distances because of increased latency effects. Today's large
'traditional' simulation including Flight Simulators may play a DIS exercises in the USA can probably be represented by the
direct part. Not surprisingly, these are the same roles which point X. There are also many smaller scale networks including
they currently play but as isolated activities. Considering the those in the UK which probably lie at Y which are aimed more
roles as part of the SE process does, however, raise several at engineering assessments rather than tactical training and
major integration issues. These integration challenges which have the opportunity for improved cues.
are discussed below will need to be resolved before flight
With the passage of time and an increase in bandwidth
simulation can play its full role in SE.
becoming available, the lines may well move downwards as
Networking Technology - Horizontal Integration indicated by EF which shows a future equivalent of line AB.
To many people, simulator networking is synonymous with
In summary, at any given time there will exist network
SE. Previous sections of this paper show that networking is
technology which gives certain player interaction fidelity
just one of the contributing technologies. Before using
characteristics depending upon how it is loaded. The challenge
networks, the case for it must be carefully considered. Local
is to stay within the constraints for the task in hand and the
area networks currently give the most satisfactory solutions
conclusions trying to be drawn. Fig 4 should ideally be used
but are not always possible. Wide area Networks are set up
by defining the fidelity and size of conflict required and then
often because legacy simulators and their operators and users
checking whether the technology can support it. If this is
are geographically remote. There have been examples where
unachievable then careful iterations of both parameters should
large numbers of forces have interacted (Ref 2). Most
be made until it is achievable, always taking care that
demonstrations take place using the Distributed Interactive
sufficient fidelity will result.
Simulation (DIS) protocols defined by IEEE 1278. However,
current communications technology will only give the quality Irrespective of technology constraints, the important question
of interaction suitable for limited tactical training in large is always: how large a conflict should be simulated. Clearly for
conflicts since each operator is receiving outside world training purposes, large groups of players are claimed to be
feedback cues restricted by network bandwidth limitations and beneficial.
possibly latency. This resultant loss to fidelity is particularly For engineering assessments, excessively large conflicts may
intrusive for simulations of fast moving vehicles such as make clear analysis and conclusions difficult. On the other
fighter aircraft. hand, too small a conflict may make for a too benign situation
For a given set of aircraft manoeuvres, there is a relationship to drive out the important issues. Choosing the scenario to
which links the rate of transmission of DIS Protocol Data match the purpose is and always will be another challenge.
Units (PDU's) to the size of dead-reckoning thresholds which Verification, Validation and Accreditation (W & A)
have been specified. Generally speaking, the smaller the The concept of using models and simulation for defence
threshold, the more PDU's are produced. At the same time, applications is nothing new. The benefit offered by SE is that
however, the fidelity with which aircraft is represented within this modelling is fully integrated, consistent and more
the network is increased accordingly. Thus for a given extensive than previous work which tended to be fragmented.
bandwidth capability on the network, conflicts with smaller All computer models need to be used with caution to ensure
numbers of players (e.g. tens) could produce the same network that they are appropriate for the purpose. A full understanding
loading as a larger (e.g. hundreds) conflict but at a higher of their applicability is essential before they are used. The
level of interaction fidelity because of the lower dead processes for ensuring this are referred to as VV & A.
reckoning thresholds. Interaction fidelity could probably be Verification is the process of ensuring that the computer
close to that achieved with local area networks where each model is a true representation of what the modeller intended.
new entity state is transmitted as soon as it is generated as Validation is the process of ensuring that the model is an
long as the number of interacting players is not excessive. The adequate representation of the real-world system for the
large scale demonstrations regularly taking place in the USA purpose intended. Accreditation is the acknowledgement by an
tend to use larger dead-reckoning thresholds in order to avoid authority that a particular model has been cleared for a
excessive network loading. Smoothing algorithms are then particular application.
included to mask discontinuities but these can really only give
the illusion of fidelity. Similarly, if simulators are interacting Concerns about the inappropriate use of models beyond their
over large geographical distances, the latency effects will also intended scope are highlighted by the scale of modelling
serve to degrade interactions. Again, predictors can be used within SE. The effectiveness of a weapon system is an
with limited success. emergent property of its interaction with its environment and
other players. In order to have confidence in the emergent
Fig 4 attempts to plot the situation qualitatively. AB shows properties produced by models of the weapon system and its
that with locally situated simulators using DIS the fidelity of environment, an authority will need to endorse the VV & A of
interaction cues decreases as the dead-reckoning thresholds the individual components of the SE as well as the VV & A of
are changed to allow for network bandwidth. Line CD shows a the overall system produced when all the models and the
36-4

environment interact. When models may be produced by represents equipment or operator behaviour under certain
different sources or be legacy models, the role of SE authority conditions.
raises VV & A issues on an unprecedented scale. Strengthening and formalising the process by which upward
This issue is further highlighted if the network is being used integration takes place will further strengthen the role of both
to evaluate an envisaged system which does not yet exist. the simulators and wargames in the SE.
Where the system already exists, there is at least a reference A case can be argued for validating the wargame and its data
with which to compare the simulation. assumptions by cross checking its behaviour with an identical
Playing the devil's advocate, a question which might be asked conflict but with one or more of the entities being coupled in
is: are these the real properties of this particular sensor on this real-time to manned simulators. If the same kind of results are
particular platform or just an artifact of mismatched software obtained then it should confirm that the upward integration
models communicating over an inadequate network? process was valid. The wargame could then be confidently
used without the need for on-line manned simulators.
Rapid Evaluation of Designs
A whole sector of SE briefly referred to earlier is the ability Another benefit of real time vertical integration between
for a requirement to be converted into a design and then into a manned simulators and wargames could be the higher levels of
performance model in a short timescale. This is essential to realism which the simulator operator could experience since
allow various options to be assessed. A large portion of the SE he would become part of a realistic sized conflict. However it
concept takes the risk out of the production process by rapidly must be remembered that for many training or evaluation
costing the product and estimating the lead time. It also purposes, small conflicts are perfectly adequate (for example
considers the life cycle costs and logistics issues. initial flying training).
With such a system, there are increased opportunities for more Some wargames, particularly those representing land battles,
design interactions to take place which should ultimately lead rely on aggregation and de-aggregation to represent large
to a better product. Even with the faster iteration period forces. Attempting to couple a manned simulator into such an
proposed by SE, the high payback during this phase may mean aggregation of entities presents a further set of problems.
that more time is spent in this activity than currently. The relatively low numbers of aircraft used in conflicts
Because there may be many options to be assessed by flight compared with the numbers of land vehicles probably means
simulation, a process must exist to take the design definition that flight simulators can avoid this particular problem.
and convert it into a simulation reliably and quickly with the
appropriate traceability to ensure configuration control. Take CONCLUSIONS
for example a new flight control system or cockpit display The fully integrated SE is a long term vision. Many of the
format. Traditionally, there have been significant delays in components are here now. Integration can begin. When the
transferring these to training simulators and plenty of tool is in place, it will help all those involved in the defence
opportunity for mistakes to be introduced. The SE can only process to be more effective. What it will not do is to replace
function effectively if the implementation process is rapid and intellectual power. We will need the same amount of thought
rigorous using an appropriate quality management system. hours to reach the same decisions. But the work supporting
There are a number of methods for achieving this for avionic those decisions should be made easier with a higher degree of
systems under development within industry at the moment. confidence and more visibility of the outcome.
This capability must be extended to cover all aspects of the There will, of course, be tilings you can do in the SE which
weapon system through the use of high-level languages, you couldn't do before such as the large conflict simulations.
autocode generators, computer networks, data standards and These are part of the SE process, however, not the whole.
so on if real-time simulation is not to become a weak link. Flight Simulation as we know it today has the challenge of
Vertical Integration adapting to take its place in the SE. Most of the issues centre
Higher level wargames often deal with large conflicts around integration whether it be horizontal networking,
primarily to allow such parameters as attrition rates to be vertical integration or integration with design definitions.
calculated by repeated runs of a scenario with weapon system Pervading the whole process are the V,V and A issues and the
performance tolerances included using Monte-Carlo methods. organisation to support them.
The same high level models could be used for training
commanders in decision making. The assumptions upon which REFERENCES
the players' behaviour is based must come from analysis Conference Proceedings
and/or manned simulations of smaller subsets of the conflict. 1. Clements, R.D. and Beautement, P., "Vertical Integration -
Feeding these assumptions into the wargames from the Networking to the Campaign Level", RAeS Conference
manned simulators is one example of vertical integration. The Nov 94.
process already takes place to an extent within industry and 2. Scribner, K.W., "Distributed Interactive Simulation and the
the research establishments. The data can consist of anything War Breaker Zen Regard Simulation, a participant's
from lookup tables to transfer function equations which perspective", RAeS Conference Nov 94.
 36-5

REQUIRED
Customer Defined Geopolitical EFFECTIVENESS
Scenarios INPUTS
Strategic Models e.g.
WEAPON PERFORMANCE
TRAINING & READINESS
JOINT DOCTRINE
Costs & Logistics
Operational Models, Wargames

Simulations & Live Equipment

Product Definition
Component Design
Cost models, Manufacturing
Models

SOLUTION
CONCEPTS
e.g.
DESIGN ANALYSIS
TECHNOLOGY DEMONSTRATIONS
RISK & COST VISIBILITY
TOOLS, METHODS FOR DEVELOPMENT1

FIG 1. THE SYNTHETIC ENVIRONMENT HIERARCHY

36-6

MEASURES OF MISSION
EFFECTIVENESS

WARGAMES INCL.
AGGREGATED COMMANDERS
FORCES

DOCTRINE, VERTICAL
MISSION PLANS, INTEGRATION
TACTICS PROCESS

TRAINING
PROCEDURES
(MODIFYABLE)
REAL-TIME
PERFORMANCE
SIMULATIONS, LIVE
EQUIPMENT
PERSONNEL

"
MEASURES OF
TRAINING
Fig. 2 Simplified SE Process for PERFORMANCE
Training or Mission Rehearsal

MEASURES OF
EFFECTIVENESS

t
WARGAMES INCL.
AGGREGATED COMMANDERS
FORCES

<
DOCTRINE, VERTICAL
MISSION PLANS, INTEGRATION
TACTICS PROCESS

i DESIGN DEFINITION
(MODIFYABLE)
REAL-TIME
PERFORMANCE
SIMULATIONS, LIVE
PERSONNEL
EQUIPMENT ''
COST, TIMESCALE,
' LOGISTICS
SUPPORT
MEASURES OF
Fig. 3 Simplified S.E. Process PERFORMANCE
for Equipment Acquisition

Interaction
Fidelity

better

Number of DIS participants

Fig. 4 Network characteristics

 37-1

Estimating The Training Effectiveness of

Interactive Air Combat Simulation

Wayne L. Waag, Ph.D. and Herbert H. Bell, Ph.D.

Aircrew Training Research Division

Armstrong Laboratory
6001 S. Power Rd
Mesa, AZ 85206-0904
USA

SUMMARY employment of classified systems, communications, and

tactics. These factors combine to limit the opportunities for
This paper is concerned with the general problem of measuring training combat tasks at both individual and team levels.
the value of simulation for combat mission training. There are
a number of engineering efforts currently attempting to In developing its multiship simulation program, the Armstrong
develop multi-player, virtual simulations that will allow Laboratory's Aircrew Training Research Division, in
soldiers, sailors, and pilots to interact with one another in a cooperation with the Air Combat Command, surveyed over
synthetic battlefield for combat mission training. This paper 300 mission ready (MR) pilots and air weapons controllers
will briefly discuss the continuation training environment that (AWCs) to identify continuation training needs (2,3).
simulation must effectively complement and the various Responses to these surveys were surprisingly similar no matter
approaches for obtaining training effectiveness data for the respondent's experience level, unit, or weapon system. The
estimating the training payoff of these efforts. It will then consensus is that it is difficult to train the pilot and AWC to
summarize the results of recent efforts conducted by the make full use of the weapon system as part of a combat team.
Armstrong Laboratory to assess the value of combat mission Table 1 shows the combat training areas most frequently
simulation for continuation training of pilots. Although the mentioned as needing improvement.
results of these studies indicate high user acceptance for
simulation and improved performance during the course of Table 1. Mission Activities Most Frequently Mentioned As
simulator-based training, transfer of training data has yet to be Requiring Additional Training
obtained.
Multibogey, four or more
1. INTRODUCTION All-aspect defense
Reaction to surface-to-air missiles
The United States Air Force (USAF) spends a great deal of Dissimilar air combat tactics
money to develop and maintain the combat proficiency of its Four-ship tactics
pilots. Most of this combat-oriented training is conducted at Reaction to air interceptors
the operational unit as part of its continuation training Employment of electronic countermeasures
program. The basic instructional media for continuation Chaff/flares employment
training are the aircraft, the environment in which it operates,
and the post-mission debrief. Together they provide an on-the-
These mission areas involve the very tasks for which in-flight
job training environment built around the opportunities for in-
training is most likely to be constrained by the factors
flight training. In-flight training opportunities, however, are
mentioned above. If anything, the negative impacts of these
limited by many factors (1). These factors include: peacetime
factors on training will increase in the future. Therefore, we
training rules, resource limitations, technical constraints, and
must develop other training approaches that will maintain the
security restrictions. Each of these factors places restrictions
readiness of our combat air forces. Simulation is one such
or imposes unnatural constraints on training. Peacetime
approach (4). In particular, distributed interactive simulation
training rules impose altitude and weather restrictions, limit
seems especially promising since it offers the potential
use of communications jamming, permit limited weapons
interactivity that characterizes the combat environment.
firings, and require a minimum separation between aircraft.
Resource limitations restrict the number of aircraft available
Because of the high cost of flight simulators and the potential
for training, the number of flying hours available, and the size
consequences of inadequate training, one would assume there
of the training ranges. Technical constraints limit the use of
is an extensive research base establishing the value of training
electronic warfare systems, prevent practice against an
combat tasks in simulators. It is not unreasonable to ask
integrated air defense system, and limit the measurement of
questions such as: Was the simulator training effective? Can
combat performance. Security restrictions prevent full
it be improved? How frequently is it needed? Is simulation

Paper presented at the AGARD FVP Symposium on "Flight Simulation — Where are the Challenges?",
held in Braunschweig, Germany, from 22-25 May 1995, and published in CP-577.
37-2

worth the costs? All of these questions reflect the need to Stage 5. Extrapolation to Combat Environment. The last stage
evaluate the potential benefits of distributed simulation for of the evaluation process attempts to answer the question of the
combat-oriented training. military value of training. As might be expected, an empirical
approach is not amenable for this question. Rather, a modeling
2. MEASURING TRAINING BENEFITS approach is recommended as a means of extrapolating from
simulator-based training to a combat environment. An
An immediate question becomes exactly how to evaluate the example of such an approach is provided by Deitchman (6) in
benefits of simulation-based training as a means of improving an attempt to project the impact of training into a central
combat mission performance. Bell and Waag (5) have European type of wartime scenario. In that case, arbitrary
proposed a five-stage sequential model which is briefly estimates were used to represent the potential impacts of
summarized below. training. However, data from a systematic evaluation program,
which recorded performance as a function of training, could
Stage 1. Utility Evaluation. The objectives of the initial stage easily be substituted into constructive models at the
are to (a) evaluate the accuracy or fidelity of the simulation engagement level and the results fed into the higher level
environment; and (b) to gather opinions from users concerning mission and campaign models. For example, training
the potential value of the simulation for specific training effectiveness data might show that survival is increased by an
applications. average of 25% as a result of simulator-based training. Using
constructive simulations, the relative impact of such changes
Stage 2. Performance Improvement. The objective of the could be assessed in operational terms.
second stage of the evaluation is to determine the extent to
which performance improves during the course of training 3. F-l 5 ADVANCED AIR COMBAT SIMULATION
within the simulation environment. The major challenge
during this stage of the evaluation is to ensure that there is a In concert with this model, the Armstrong Laboratory has been
proper means of establishing that performance has indeed gathering data over the past few years attempting to establish
improved as a result of the training. This requires the the value of simulation for air combat training. In 1988, a
development of mission scenarios that are flown before and program was initiated with the Tactical Air Command (now
after the training that are similar to but not identical to Air Combat Command) to evaluate multiship air combat
missions flown during training. It also requires the training using commercially available contractor facilities. In
development and use of measures whereby improvements in all, two utility evaluations and one simulator performance
performance can be meaningfully reflected. improvement study were conducted as part of this project.

Stage 3. Transfer to Alternative Simulation Environment. The These efforts used the McDonnell Aircraft Simulation facility
question of generalizability now is raised-does training in St. Louis, Missouri. This simulation system was designed to
transfer to another environment? While the acid test is usually support engineering development. Its design and equipment
considered to be transfer to the air, it is our view that a more typify the full mission simulator facilities developed by
logical intermediate step involves demonstrating transfer to aircraft manufacturers in the late 1980s. Figure 1 shows the
other simulation environments. Recall that one of the primary principle components of this system. EachF-15C cockpit was
justifications for multiplayer air combat simulation is the located in a 40-ft diameter dome which provided the pilot with
ability to practice certain events under conditions that are a nearly full field of view. Each simulator had high fidelity
generally not available in peacetime training environments. aerodynamic, engine, avionics, communication, sensor, and
Because of safety restrictions, security considerations, rules of weapon simulations. Other components included additional
engagement, etc., peacetime exercises will always be limited in aircraft (either digital or manned), digitally controlled surface-
terms of their situational fidelity. For this reason, it is to-air-threats, exercise control, debrief, and data record. A
essential that transfer be demonstrated to another simulation more detailed description of the basic simulation system is
environment in which a wartime environment can be created. available (7).

Stage 4. Transfer to Flight Environment. If positive transfer to Utility Evaluations. Two utility evaluations were conducted.
a simulated wartime environment has been shown, the next In the first evaluation (8), 42 mission-ready F-l 5 pilots and 16
stage is to show transfer to the air. Unfortunately, such a AWCs received four days of training. The training unit was
transfer test is limited by the large number of peacetime the team comprised of two pilots (lead/wingman) plus the
restrictions that characterize current flight operations. For this AWC. This team flew a variety of combat missions against an
reason, a smaller sample of combat tasks would most likely opposing force comprised of four to eight adversaries plus the
have to be selected for evaluation. To whatever extent adversary AWC.
possible, the transfer test should represent a highly controlled
flight environment wherein performance data can be gathered Upon completion of training, pilots rated the value of both
easily. their "unit training" and the "simulation training" for 41 air-to-
air tasks. The pilots felt that simulator training was much
better than their current unit training for many air combat tasks
 37-3

CompuScene CompuScene

1.00 T
DPretest
Ethernet
( F"15 0.80 ■• ■ Posttest
( F"15 1 ^Simulations
1
V Simulation )
0.60 ■-

J
HSD- Gould 32/97 •HSD 0.40 ■ -
0.20 ■-
0.00
AWC IBL
Success Survival Efficiency

Gould 32/97 — HSD — Threats Figure 2. Comparison of Pre/Post Test Performance

4. F-15 SITUATION AWARENESS STUDY

Figure 1. McDonnell Douglas Simulation Facility
In 1991, the US Air Force Chief of Staff posed a series of
including multibogey, chaff and flares employment, all-aspect
questions concerning situation awareness (SA). First of all,
defense, use of electronic countermeasures and counter-
what is SA? Can it be objectively measured? Is SA learned or
countermeasures, communications jamming, and work with the
does it represent a basic ability or characteristic that some
AWC. These tasks were also rated high in "need for additional
pilots have and others do not? From a research standpoint,
training'1 prior to the start of simulator training. On the other
these questions translate into issues of measurement, selection,
hand, tasks such as air combat maneuvering (ACM), visual
and training. The Armstrong Laboratory was subsequently
lookout, gun employment, and basic fighter maneuvering
tasked with providing research answers to these questions. A
(BFM) were rated as better trained in their in-flight
research investigation was initiated that had three goals: first,
continuation training program than in the simulation. Air
to develop and validate tools for reliably measuring SA;
weapons controllers, however, rated all tasks as better trained
second, to identify basic cognitive and psychomotor abilities
in the simulation environment. Open-ended opinion data were
that are associated with pilots judged to have good SA; and
also gathered, the results being quite positive toward the
third, to determine if SA can be learned, and if so, to identify
training.
areas where cost-effective training tools might be developed
and employed. An overview of the investigation can be found
A second evaluation, was conducted using the same procedure
in McMillan, Bushman, and Judge (9).
but with a larger sample of pilots and AWCs (3). This
evaluation produced essentially the same results. Based on the
The general approach was to first develop criterion measures
high user acceptance demonstrated during these utility
of SA based upon performance ratings collected within an
evaluations, Air Combat Command continued this program
operational flying environment. The results of this part of the
under its own sponsorship.
study can be found in Waag and Houck (10). These measures
were necessary for two reasons. First, they would serve as
Performance Improvement. In the third study, again using the
criterion measures against which to validate a battery of basic
same facility, in-simulator learning was also shown, in
ability tests considered relevant to SA, thereby addressing the
addition to positive user opinion. Subjects consisted of 16
question of basic human abilities. The results of this part of
teams, each team being made up of two pilots and an AWC.
the study can be found in Carretta and Ree (11). Second, these
Each of the elements flew controlled offensive and defensive
measures would serve as a means of selecting a sample of
scenarios "before" and "after" three days of intensive
pilots who would participate in a simulation phase of the
simulation training. Digital data as well as videotapes of
effort. During that phase, simulated air combat mission
displays used for replay and debriefing purposes were archived
scenarios were developed for assessing SA and objective
for later analysis.
measures of performance gathered in an attempt to determine
those characteristics that distinguish pilots with good SA.
Preliminary analyses reveals that post-training mission
These data would be used to identify areas where training tools
performance is significantly higher than pretraining
might be developed. We now summarize the results of the
performance. Figure 2 shows the mean value of several pre-
third phase of the program, namely the use of simulation as a
and post-training mission performance indicators for defensive
tool for measuring and training SA. The complete findings are
counterair missions. The data clearly indicate that the
presented in Waag, Houck, Greschke and Raspotnik (12).
probability of mission success (i.e., no enemy strikers to
target) and F-15 survival increased during the course of the
Method. A total of 40 MR F-15 pilots, who were flight lead-
simulator training (p < .05). In addition, weighted exchange
qualified served as subjects. An additional 23 MR F-15 pilots
ratios, reflecting the efficiency of mission accomplishment,
served as wingmen throughout the data collection. The
also increased as a function of training (p < .05).
simulated combat missions were flown using the Armstrong
37-4

Laboratory's multiship simulation facility (MULTERAD) The same two subject-matter experts (SMEs) were used
located at Williams AFB (WAFB), AZ. The major throughout the year-long data collection effort. Upon
components of the simulation system are shown in Figure 3. completion of the mission, they discussed each engagement,
These components represent independent subsystems operating and completed a consensus performance rating scale consisting
as part of a secure distributed simulation network. This local of the 24 behavioral indicators of SA related to F-15 mission
area network was connected to the air weapons controller performance. A variety of other data were also gathered and
simulator (AESOP) at Brooks Air Force Base (BAFB), TX by archived, including mission events and outcomes, digital data
a dedicated T-l telephone line. Additional details concerning passed over the network, videos used for debriefing, eye
the basic simulation architecture and components are available movement data recorded on the last mission, and finally,
(13,14,15). "critiques" of the simulation and opinions regarding its
potential for training. Two types of user opinion data were
gathered. First, pilots rated the training benefit for various
MULTIRAD Simulation pilot experience levels. And second, pilots completed an
Configuration for SA Study open-ended questionnaire pertaining to the overall value of the
simulation and how it might best be used.
Maual Image
Generator

WILLIAMS AFB
Findings. The results of the ratings of potential training
360 Degree
Vlaual
160 Degree
MBUll
benefits are provided in Figure 4. These data clearly indicate
Diaplay Dlepliy
that positive opinions were expressed by the study participants
on the value of this type of simulation for training. The
potential training was considered beneficial for all levels of
Automated Exerclae
qualification. It is of interest to note that training was
Threat
Engagement
Control
Station
considered highly beneficial for four-ship flight leads, despite
syatem
Su 27, MIG-27, SA-4,6,0
AESOP the fact that the MULTIRAD simulation facility provided
SIMULATION
MULTIRAD SIMULATION FACILITY FACILITY training for only a flight lead and wingman. As expected,
higher benefit ratings were given to pilots upgrading into a
Figure 3. Armstrong Laboratory Simulation Facility given qualification level.
The manned flight simulators consisted of two F-15C
simulators and two F-16 simulators. The F-15C simulators had
Extremely Beneficial
high fidelity aerodynamic, engine, avionics, radio, sensor, and
weapons simulations. Each F-15C simulator was equipped
Highly Beneficial
with an out-the-window visual display system covering
approximately 360 deg horizontal by 200 deg vertical. The
Somewhat Beneficial 3
external visual scene was created using computer-generated
imagery. The lower fidelity, manned F-16 simulators played
the role of enemy aircraft in conjunction with computer-
controlled adversaries. The visual and electronic signatures of Not Beneficial
these F-16 simulators were modified so that they appeared as 2-SHp 4SKp
the appropriate threat aircraft. Each F-16 simulator was Lead Lead
equipped with a single channel of out-the-window visual
Riojrt edification of Trainee
imagery covering approximately 45 deg horizontal by 45 deg
vertical. A manned AWC provided the F-15C pilots with
appropriate threat information and warnings. Depending upon Figure 4. Rated Benefits of Simulation Training
the availability of qualified AWCs and equipment status, the
AWC was either located at WAFB or BAFB. In either case, the Opinions expressed in the open-ended questionnaire were also
AWC had a realistic simulation of the appropriate AWC quite positive. Although qualitative, they provide additional
console and communicated with the F-l 5C pilots by radio. insight into the potential focus of training using multiship
simulation and how it might be employed. In particular,
The primary approach taken toward the measurement of SA mention was made of using such training as a means of
was through scenario manipulation and observation of enhancing both situation assessment and decision-making
subsequent performance. A week-long SA "evaluation" skills. It was also frequently noted that there was tremendous
exercise was constructed that consisted of 9 sorties with 4 value in learning flight leadership and resource management
engagements per sortie. Sorties were arranged in a building skills. In terms of the location of such simulation, the
block manner. Over the week, engagements increased in overwhelming consensus was that they would be of most value
complexity in terms of numbers of adversaries, enemy tactics, within the operational units. This was not too surprising since
lethality of ground threats, AWC support, etc. each unit now has the operational version of the cockpits used
in the present investigation. However, they are stand-alone
and non-visual, and as such their training capability is fairly
 37-5

limited. In contrast, the networking of such devices within a (MDT2). MDT2 provides a common, virtual environment that
realistic combat environment increases the potential greatly. is being used to support training research involving collective
The bottom line from the utility data is that the participants tasks.
considered multiship simulation as a tool with high training
potential.

It should be pointed out that it was never the intent, at the

outset of the study, to demonstrate performance improvements.
It must be emphasized that the sole purpose was to develop a
set of simulation scenarios that could be used to assess SA
within a combat environment. As such, normal training
interventions were not permitted. For example, during the
debrief, pilots were permitted to only view their own in-
cockpit displays and not the planned view display. Moreover,
the two SMEs were not permitted to provide any type of
feedback to the pilots regarding their performance. However,
data from the ninth mission did permit some comparison since
identical scenarios had been flown earlier in the week. The Figure 6. Multiservice Distributed Training Testbed
ninth mission was designated the "eye track" mission in which
eye movement data were recorded. Currently, MDT2 is focusing on the planning and execution of
close air support (CAS) at the engagement level. The goal is to
Two scenarios, a 2 V 2 defensive counterair (DCA) mission establish a virtual training environment. The participants in
and a 2 V 4 offensive counterair (OCA) mission, were flown this virtual training environment will be soldiers, marines, and
during the middle of the week and then again on the last pilots executing their unique combat tasks in virtual simulators
mission. A comparison of performance is presented in Figure at their individual service training sites. This environment will
5. In both cases, performance on the last mission was allow collective training to occur which involves both unit and
improved. However, only the 2 V 2 DCA mission was found task force components.
to be statistically significant. When such data are coupled with
the very strong pilot opinions that they had received valuable The initial training utility evaluations of MDT2 were
training, it seems reasonably safe to conclude that learning had conducted in May of 1994 and February of 1995. For the
occurred over the week. February evaluation, four sites were interconnected using
Distributed Interactive Simulation Protocol 2.0, version 3.
These sites were the Institute for Defense Analysis in
Alexandria, Virginia; Mounted Warfare Testbed at Ft. Knox,
■ First Engagement
Kentucky; the Manned Flight Simulation Facility at Patuxent
a Last Engagement River, Maryland; and the Armstrong Laboratory at Mesa,
Arizona. This simulation network is illustrated in Figure 6,
while information regarding each of the sites is summarized in
Table 2.
Table 2. Simulation Testbed Components

Location Simulators Role

Armstrong 2F-16 Close Air Support
2V2 DCA 2V4 OCA Laboratory
Type of Mission
1 Laser Scout and Target
Designator Designation
Naval Air OV-10 Forward Air Control
Figure 5. Effects of Practice on Observer SA Ratings Warfare Center
Mounted Warfare M-1A1 Ground Maneuver
5. MULTISERVICE DISTRIBUTED TRAINING TESTBED Testbed M-2 Battalion Tactical
Operations Center
The Armstrong Laboratory is currently working with the Army Institute for None Stealth Display
Research Institute and the Naval Air Warfare Center to Defense Analysis Data Record
develop a training testbed that can be used to assess the value
of Distributed Interactive Simulation for multiservice training.
The results of this training effectiveness evaluation indicated
This effort, initially sponsored by the Defense Modeling and
that mission-ready fighter pilots, airborne forward air
Simulation Office, links service-developed training systems
controllers, and ground combatants felt that the simulation
together to create a Multiservice Distributed Training Testbed
37-6

provided significant training benefits. In addition, trained particularly when these findings show no transfer effects or
observers and subject-matter experts monitored that fairly small transfer effects. The empirically obtained outcome
performance of selected mission tasks throughout the training for any transfer of training experiment is one of three possible
period. The data indicated that mission performance for each outcomes; positive transfer, no transfer, or negative transfer.
of the components, air and ground, increased over the four Similarly, the true effect of training is one of the same three
days of simulated combat. The high user acceptance and possibilities. The problem is to infer the true state from the
increase in performance are most likely due to the unique obtained outcome. This inferential process works quite well
ability that MDT2 provides for planning, execution, and review when statistically significant outcomes confirm expectations.
of close air support as an integral part of the ground Unfortunately, the inference process does not work as easily
commander's battle plan. when little or no transfer is obtained as a result of training.
Now the investigator must decide between two possibilities.
6. DISCUSSION Indeed, the training may have little of no effect on
performance. Or, the effects may be much larger, but because
The results obtained from these efforts provide strong user of methodological problems inherent in conducting transfer of
belief in the value of interactive air combat simulation. From training experiments, they are masked. Although we do not
the user's perspective, the data are very clear regarding the know the true effects of training, we generally attempt to
potential value of such simulation for training. Users "explain away" any lack of positive effects and attribute it to
consistently report that such simulations are an enhancement to these "methodological problems," especially if there are other
their current mission training. Although such subjective data such as expert opinion that suggest the training to be
evidence is often considered suspect from a scientific beneficial.
perspective, it is nevertheless an absolute prerequisite for
effective training. Unless there is user acceptance, the A good case in point is a study by Pohlmann and Reed (17)
resulting training will be of marginal value regardless of the that failed to show positive transfer effects for ACM training
device's inherent potential. in the Simulator for Air-to-Air Combat (SAAC). Do we
believe that simulator training does not improve air-to-air
In addition to the opinion data, there is evidence that performance? Probably not, since we have other evidence
performance did improve within the simulation environment. suggesting the training to be beneficial. This evidence
Performance improvements were demonstrated in the F-l 5 includes positive end-of-course opinions suggesting training in
Advanced Air Combat Simulation, the F-l 5 Situation the SAAC to be some of the best air-to-air training pilots had
Awareness Study, and the Multiservice Distributed Training ever received, in-simulator performance improvements (18),
Testbed. These data, combined with the fact that the study and positive transfer of training in another experiment (19).
participants expressed opinions to the effect that their The study failing to demonstrate positive transfer had one
proficiency had improved, leave little doubt that learning had potentially serious limitation in that instructor ratings were
occurred. used as a measure of performance. Such measures have been
shown to be quite insensitive in other air combat domains. For
Although the data clearly indicate that the end user expresses example, a study by Gray and Fuller (20) which demonstrated
very positive opinions toward the value of multiship simulation significant transfer of training in terms of bombing accuracy,
and that learning occurs, there still remains the issue of also used instructor ratings of performance in the air.
transfer of training. Does such training transfer to other Interestingly enough, the rating data showed no effects of
simulation environments (Stage 3 of the Evaluation Model) simulator pretraining despite large differences in objective
and does it transfer to the real world (Stage 4 of the Evaluation measures of weapons delivery. So it seems at least plausible
Model)? The data gathered in this study do not bear upon that the failure to show any effect in the Pohlmann and Reed
these issues. (17) study may have been due largely to the measures that
were used. For this reason and the fact that we have other
The question becomes, "are transfer of training data needed?" evidence suggesting the training to be valuable, we can make
While no one would argue the desirability of having such data, the case to simply "dismiss" these findings.
there are practical issues which seriously question the
advisability of conducting such studies. Lack of experimental At this point, we have a paradox emerging. On the one hand,
control, insufficient sample sizes, insufficient training time in we have made the argument that the transfer of training
the simulator, insufficient time for evaluating transfer in the evaluation is the only sufficient test for establishing training
air, insensitive measures, etc. are problems that plague the effectiveness. On the other hand, we have also shown that we
conduct of any transfer of training evaluation (16). In fact, one tend to dismiss those studies failing to demonstrate positive
can argue that it is virtually impossible to conduct a well- transfer when we have other data, which is usually expert
controlled transfer test within an operational military opinion, suggesting the training to be effective. In such
environment. instances we attribute the lack of positive transfer effects to
one or more of those "methodological problems" which always
This inability to adequately control such evaluations perhaps exist in the conduct of such evaluations within an operational
has its greatest impact on the interpretation of findings, military training environment. If we are willing to explain
 37-7

away our inability to demonstrate training effectiveness, the Williams AFB AZ: Air Force Human Resources Laboratory,
question becomes, "why conduct the transfer evaluation?" Operations Training Division.

Since there is currently no definitive data, the question of 9. McMillan, G.R., Bushman, J. & Judge, C.L.A. (1995)
training benefits of interactive air combat simulation is largely Evaluating pilot situational awareness in an operational
answered by ones' personal view of simulation and one's environment. InAGARD-CP-575, Situation Awareness:
willingness to generalize from previous investigations of Limitations and Enhancement in the Aviation Environment,
transfer in other domains. For the "believer," including the Neuilly-Sur-Seine, France: Advisory Group for Aerospace
authors of this paper, the evidence to date is strong enough to Research and Development.
warrant the conclusion that training will be effective. In fact,
given the previous transfer of training research that has already 10. Waag, W.L. & Houck, M.R. (1995) Development of
been conducted (5,16) there is little reason to suspect that such criterion measures of situation awareness for use in operational
training within a multiship simulation environment would not fighter squadrons. InAGARD-CP-575, Situation Awareness:
have a positive effect upon subsequent performance in the air. Limitations and Enhancement in the Aviation Environment,
Consequently, there is no compelling reason to conduct Neuilly-Sur-Seine, France: Advisory Group for Aerospace
transfer of training studies within the air combat environment. Research and Development.
However, for the "skeptic," no definitive evidence has been
presented and the question remains unanswered. 11. Carretta, T.R. & Ree, M.J. (1995) Determinants of
situational awareness in U.S. Air Force F-15 pilots. In
7. REFERENCES AGARD-CP-575, Situation Awareness: Limitations and
Enhancement in the Aviation Environment, Neuilly-Sur-Seine,
1. United States Air Force Scientific Advisory Board. (1992, France: Advisory Group for Aerospace Research and
December). Report of the Support Panel of the Air Force Development.
Scientific Advisory Board Ad Hoc Committee on Concepts and
Technologies to Support Global Reach - Global Power 1995 - 12. Waag, W.L., Houck, M.R., Greschke, DA. & Raspotnik,
2020. Washington, DC: Author. W.B. (1995) Use of multiship simulation as a tool for
measuring and training situation awareness. InAGARD-CP-
2. Gray, T.H., Edwards, B.J., & Andrews, D.H. (1993). A 575, Situation Awareness: Limitations and Enhancement in the
survey of F-16 squadron level training in PACAF (AL-TR- Aviation Environment, Neuilly-Sur-Seine, France: Advisory
1993-0041). Williams Air Force Base, AZ: Armstrong Group for Aerospace Research and Development.
Laboratory, Aircrew Training Research Division.
13. Gehl, T.L, Rogers, R.L, Miller, M.A., & Rakolta, J.
3. Houck, M.R., Thomas, G.S., & Bell, H.H. (1991). Training (1993). Interdependence of training utility and network
evaluation of the F-15 advanced air combat simulation (AL- performance using the Armstrong Laboratory Multiship
TP-1991-0047). Williams Air Force Base, AZ: Armstrong Research and Development System. In, Proceedings of 15th
Laboratory, Aircrew Training Research Division. Interservice/Industry Training Systems Conference, Orlando,
FL: National Security Industrial Association.
4. Alluisi, E.A. (1991). The development of technology for
collective training: SIMNET, a case history. Human Factors, 14. Platt, P. & Crane, P. (1993). Development, test, and
33, 343-362. evaluation of a multiship simulation system for air combat
training. In. Proceedings of 15th Interservice/Industry
5. Bell, H.H. & Waag, W.L. (1994). Evaluating the Training Systems Conference, Orlando, FL: National Security
Effectiveness of Flight Simulators for Training Combat Skills: Industrial Association.
A Review. Submitted to International Journal of Aviation
Psychology. 15. Rogers, B. (1992). Tactical air threat system for a
distributed simulation network. In, Proceedings of 14th
6. Deitchman, S.J. (1988). Preliminary exploration of the use Interservice/Industry Training Systems Conference, San
of a warfare simulation model to examine the military value of Antonio, TX: National Security Industrial Association.
training (IDA Paper P-2094). Alexandria, VA: Institute for
Defense Analysis. 16. Waag, W.L. (1981). Training effectiveness of visual and
motion simulation. (AFHRL-TR-79-72). Williams AFB AZ:
7. McDonnell, G.W., Broeder, R.F, & Cutak, R.J. (1989). Air Force Human Resources Laboratory, Operations Training
Multi-ship air combat simulation. In Proceeding of the 11th Division.
Interservice/Industry Training Systems Conference. Ft. Worth,
TX: National Security Industrial Association. 17. Pohlmann, L.D. & Reed, J.C. (1978). Air-to-air combat
skills: contribution of platform to initial training (AFHRL-TR-
8. Thomas, G.S, Houck, M.R. & Bell, H.H. (1990). Training 78-53). Williams AFB AZ: Air Force Human Resources
evaluation of air combat simulation (AFHRL-TR-90-30). Laboratory, Operations Training Division.
37-8

18. Robinson, J.C., Eubanks, J.L. & Eddowes, E.E. (1981).

Evaluation of pilot air combat maneuvering performance
changes during TAC ACES training. NellisAFBNV: USAF
Tactical Fighter Weapons Center.

19. Jenkins, D.H. (1982). Simulation training effectiveness

evaluation (TAC Project 79Y-001F). Nellis AFB, NV:
Tactical Fighter Weapons Center.

20. Gray, T.H. & Fuller, R.R. (1977). Effects of simulator

training and platform motion on air-to-surface weapons
delivery training (AFHRL-TR-77-29). Williams AFB AZ: Air
Force Human Resources Laboratory, Operations Training
Division.
 REPORT DOCUMENTATION PAGE
1. Recipient's Reference 2. Originator's Reference 3. Further Reference 4. Security Classification
of Document
AGARD-CP-577 ISBN 92-836-0024-X UNCLASSIFIED/
UNLIMITED

5. Originator Advisory Group for Aerospace Research and Development

North Atlantic Treaty Organization
7 rue Ancelle, 92200 Neuilly-sur-Seine, France

6. Title
Flight Simulation — Where are the Challenges?

7. Presented at/sponsored by
The Flight Vehicle Integration Panel Symposium,
held in Braunschweig, Germany, from 22-25 May 1995

8. Author(s)/Editor(s) 9. Date
Multiple April 1996

10. Author's/Editor's Address 11. Pages

Multiple 398

12. Distribution Statement There are no restrictions on the distribution of this document.
Information about the availability of this and other AGARD
unclassified publications is given on the back cover.

13. Keywords/Descriptors
Flight simulation Motion
Research projects Visual perception
NATO forces Design
Training simulators Models
Simulators Military aircraft
Display devices Helicopters
Pilot training

14. Abstract
Effective Flight Simulation is an important capability for NATO nations, and it will become even
more important in the face of reductions in defence budgets. This symposium reviewed the state
of the art in flight simulation, in order to identify weaknesses where additional research and
development are needed. Presentations dealt with simulation used for engineering and for
training, both on the ground and in-flight.
Session topics were:
— Visual and Motion Cueing and Requirements;
— Advances in Modelling;
— Simulation in Design and Development - Rotorcraft;
— Simulation in Design and Development - Fixed Wing and Systems;
— Simulation in Training;
— In-Flight Simulation;
— Future Applications.
The keynote address introduced the symposium, clearly establishing the importance of the topic
and its relevance to NATO's requirements.
A Technical Evaluation Report is also included.
Copies of papers presented at the Flight Vehicle Integration Panel Symposium held in
Braunschweig, Germany, 22-25 May 1995.
 NATO -^- OTAN

7 RUE ANCELLE • 92200 NEUILLY-SUR-SEINE DIFFUSION DES PUBLICATIONS

FRANCE AGARD NON CLASSIFIEES
Telecopie (1)47.38.57.99 • Telex 610 176
Aucun stock de publications n'a existe ä AGARD. A partir de 1993, AGARD detiendra un stock limite des publications associees aux cycles
de conferences et cours speciaux ainsi que les AGARDographies et les rapports des groupes de travail, organises et publies ä partir de 1993
inclus. Les demandes de renseignements doivent etre adressees ä AGARD par lettre ou par fax ä l'adresse indiquee ci-dessus. Veuillezne
pas telephoner. La diffusion initiale de toutes les publications de l'AGARD est effectuee aupres des pays membres de l'OTAN par
l'intermediaire des centres de distribution nationaux indiques ci-dessous. Des exemplaires supplemental res peuvent parfois etre obtenus
aupres de ces centres (ä l'exception des Etats-Unis). Si vous souhaitez recevoir toutes les publications de l'AGARD, ou simplement celles
qui concernent certains Panels, vous pouvez demander ä etre inclu sur la liste d'envoi de Tun de ces centres. Les publications de l'AGARD
sont en vente aupres des agences indiquees ci-dessous, sous forme de photocopie ou de microfiche.
CENTRES DE DIFFUSION NATIONAUX
ALLEMAGNE ISLANDE
Fachinformationszentrum Karlsruhe Director of Aviation
D-76344 Eggenstein-Leopoldshafen 2 c/o Flugrad
BELGIQUE Reykjavik
Coordonnateur AGARD-VSL ITALIE
Etat-major de la Force aerienne Aeronautica Militare
Quartier Reine Elisabeth Ufficio del Delegato Nazionale all'AGARD
Rue d'Evere, 1140 Bruxelles Aeroporto Pratica di Mare
CANADA 00040 Pomezia (Roma)
Directeur, Services d'information scientifique LUXEMBOURG
Ministere de la Defense nationale Voir Belgique
Ottawa, Ontario K1A 0K2 NORVEGE
DANEMARK Norwegian Defence Research Establishment
Danish Defence Research Establishment Attn: Biblioteket
Ryvangs Alle 1 P.O. Box 25
P.O. Box 2715 N-2007 Kjeller
DK-2100 Copenhagen 0 PAYS-BAS
ESPAGNE Netherlands Delegation to AGARD
INTA (AGARD Publications) National Aerospace Laboratory NLR
Carretera de Torrejon a Ajalvir, Pk.4 P.O. Box 90502
28850 Torrejon de Ardoz - Madrid 1006 BM Amsterdam
ETATS-UNIS PORTUGAL
NASA Headquarters Estado Maior da Forca Aerea
Code JOB-1 SDFA - Centro de Documentacäo
Washington, D.C. 20546 Alfragide
2700 Amadora
FRANCE
ON.E.R.A. (Direction) ROYAUME-UNI
29, Avenue de la Division Leclerc Defence Research Information Centre
92322 Chätillon Cedex Kentigern House
65 Brown Street
GRECE Glasgow G2 8EX
Hellenic Air Force TURQUIE
Air War College
Scientific and Technical Library Mill! Savunma Baskanligi (MSB)
Dekelia Air Force Base ARGE Dairesi Baskanligi (MSB)
Dekelia, Athens TGA 1010 06650 Bakanliklar-Ankara
Le centre de distribution national des Etats-Unis ne detient PAS de stocks des publications de l'AGARD.
D'eventuelles demandes de photocopies doivent etre formulees directement aupres du NASA Center for AeroSpace Information (CASI)
ä l'adresse ci-dessous. Toute notification de changement d'adresse doit etre fait egalement aupres de CASI.
AGENCES DE VENTE
NASA Center for ESA/Information Retrieval Service The British Library
AeroSpace Information (CASI) European Space Agency Document Supply Division
800 Elkridge Landing Road 10, rue Mario Nikis Boston Spa, Wetherby
Linthicum Heights, MD 21090-2934 75015 Paris West Yorkshire LS23 7BQ
Etats-Unis France Royaume-Uni
Les demandes de microfiches ou de photocopies de documents AGARD (y compris les demandes faites aupres du CASI) doivent
comporter la denomination AGARD, ainsi que le numero de serie d'AGARD (par exemple AGARD-AG-315). Des informations
analogues, telles que le titre et la date de publication sont souhaitables. Veuiller noter qu'il y a lieu de specifier AGARD-R-nnn et
AGARD-AR-nnn lors de la commande des rapports AGARD et des rapports consultatifs AGARD respectivement. Des references
bibliographiques completes ainsi que des resumes des publications AGARD figurent dans les journaux suivants:
Scientific and Technical Aerospace Reports (STAR) Government Reports Announcements and Index (GRA&I)
publie par la NASA Scientific and Technical publie par le National Technical Information Service
Information Division Springfield
NASA Headquarters (JTT) Virginia 22161
Washington D.C. 20546 Etats-Unis
Etats-Unis (accessible egalement en mode interactif dans la base de
donnees bibliographiques en ligne du NTIS, et sur CD-ROM)

Imprime par le Groupe Communication Canada

45, boul. Sacre-Cceur, Hull (Quebec), Canada K1A 0S7
 NATO -0- OTAN
7 RUE ANCELLE • 92200 NEUILLY-SUR-SEINE DISTRIBUTION OF UNCLASSIFIED

FRANCE AGARD PUBLICATIONS

Telefax (1)47.38.57.99 • Telex 610 176

AGARD holds limited quantities of the publications that accompanied Lecture Series and Special Courses held in 1993 or later, and of
AGARDographs and Working Group reports published from 1993 onward. For details, write or send a telefax to the address given above.
Please do not telephone. , . ,. . T . . ,
AGARD does not hold stocks of publications that accompanied earlier Lecture Series or Courses or of any other publications. Initial
distribution of all AGARD publications is made to NATO nations through the National Distribution Centres listed below. Further copies are
sometimes available from these centres (except in the United States). If you have a need to receive all AGARD publications, or just those
relating to one or more specific AGARD Panels, they may be willing to include you (or your organisation) on their distribution list.
AGARD publications may be purchased from the Sales Agencies listed below, in photocopy or microfiche form.
NATIONAL DISTRIBUTION CENTRES
BELGIUM LUXEMBOURG
Coordonnateur AGARD — VSL See Belgium
Etat-major de la Force aerienne NETHERLANDS
Quartier Reine Elisabeth Netherlands Delegation to AGARD
Rue d Evere, 1140 Bruxelles National Aerospace Laboratory, NLR
CANADA P.O. Box 90502
Director Scientific Information Services 1006 BM Amsterdam
Dept of National Defence .,.„„,.„
Ottawa, Ontario K1A 0K2 NORWAY „,,,.. mt
Norwegian Defence Research Establishment
DENMARK Attn: Biblioteket
Danish Defence Research Establishment p Q gox 25
Ryvangs Alle 1 N-2007 Kjeller
P.O. Box 2715
DK-2100 Copenhagen 0 PORTUGAL
cr> A NT^C
tKAS
Estado Maior da Forca Aerea
aX?_ _ . _.. .. , SDFA - Centro de Documentacäo
O.N.E.RA. (Direction) Alfragide
29 Avenue de la Division Leclerc 2700 Amadora
92322 Chätillon Cedex
GERMANY SPAIN
Fachinformationszentrum Karlsruhe INTA (AGARD Publications)
D-76344 Eggenstein-Leopoldshafen 2 Carretera de Torrejön a Ajalvir, Pk.4
28850 Torrejön de Ardoz - Madrid
GREECE
Hellenic Air Force TURKEY
Air War College Milli Savunma Baskanligi (MSB)
Scientific and Technical Library ARGE Dairesi Baskanligi (MSB)
Dekelia Air Force Base 06650 Bakanliklar-Ankara
Dekelia, Athens TGA 1010
ICELAND UNITED KINGDOM
Director of Aviation Defence Research Information Centre
c/o Flugrad Kentigern House
Reykjavik 65 Brown Street
Glasgow G2 8EX
ITALY
Aeronautica Militare UNITED STATES
Ufficio del Delegato Nazionale all'AGARD NASA Headquarters
Aeroporto Pratica di Mare Code JOB-1
00040 Pomezia (Roma) Washington, D.C. 20546

The United States National Distribution Centre does NOT hold stocks of AGARD publications.
Applications for copies should be made direct to the NASA Center for AeroSpace Information (CASI) at the address below.
Change of address requests should also go to CASI.
SALES AGENCIES
NASA Center for ESA/Information Retrieval Service The British Library
AeroSpace Information (CASI) European Space Agency Document Supply Centre
800 Elkridge Landing Road 10, rue Mario Nikis Boston Spa, Wetherby
Linfhicum Heights, MD 21090-2934 75015 Paris West Yorkshire LS23 7BQ
United States France United Kingdom
Requests for microfiches or photocopies of AGARD documents (including requests to CASI) should include the word 'AGARD
and the AGARD serial number (for example AGARD-AG-315). Collateral information such as title and publication date is
desirable Note that AGARD Reports and Advisory Reports should be specified as AGARD-R-nnn and AGARD-AR-nnn,
respectively Full bibliographical references and abstracts of AGARD publications are given in the following journals:
Scientific and Technical Aerospace Reports (STAR) Government Reports Announcements and Index (GRA&I)
published by NASA Scientific and Technical published by the National Technical Information Service
Information Division Springfield
NASA Headquarters (JTT) Virginia 22161
Washington D.C. 20546 United States
United States (also available online in the NTIS Bibliographic
Database or on CD-ROM)

Printed by Canada Communication Group

45 Sacre-Cceur Blvd., Hull (Quebec), Canada K1A 0S7

ISBN 92-836-0024-X