=]- ,.

- j

Techlical Report 546

DETERMINATION OF MOTION AND VISUAL

SYSTEM REQUIREMENTS FOR FLIGHT
"TRAINING SIMULATORS
Robert K. Heffley. Warren F. Clement, Robert F. Ringland,
Wavne F. Jewell, Henry R. Jex and Duane T. McRuer
Systems Technology, inc.

and

Vernon E. Carter
Northrop Corporation

ARI FIELD UNIT AT FORT RUCKER, ALA9AMA

i-1

I DTIC
, I•ELECTE•

~JUL 29 1982

U.S. Army

Research Institute for the Behavioral and Social Sciences

August 1981
Approved for public release; distribution unlimited.

41, 82 07 29 036
-"Mm
U. S. ARMY RESEARCH INSTITUTE
FOR THE BEHAVIORAL AND SOCIAL SCIENCES

A Field Operating Agency under the Jurisdiction of the

Deputy Chief of Staff for Personnel

L. NEALE COSBY
JOSEPH ZEIDNER Colonel, IN
Technical Director Commander

Research accomplished under contract

to the Department of the Army

Systems Technology, Inc.

NOTICES-

DISTRIBUTION: Primary di-stribution of this report has been made by ARI.

Please address correspondence concerning distribution of reports to: U.S.

Army Research Institute for the Behavioral and Social Sciences, ATTN.
PERI-TST, 5001 Eisenhower Avenue, Alexandria, Virginia 22333.

FINAL DISPOSITION: This report may be destroyed when it is no longer

needed. Please do not return it to the U.S. Army Research Institute for
the Behavioral and Social Sciences.

NOTE: The findings in this report are not to be construed as an official !

Department of the Army position, unless so designated by other authorized
documents.

__________ _____ ~- A
 UNCLASSIFIED
SECURITY CLASSIFICATION OF THIS PAGE 'W"en Date Entered) _

REPORT DOCUMENTATION PAGE i BORE COMNSLRUCTIONS

F
1. REPORT NUMBER No. RECIPIENTS CATALOG NUMBER
BGOVTACCESSION
Technical Report 546 0065 ______________
4. TITLE (and Subtitle) S. TYPE OF REPORT & PERIOD COVERED

DETERMINATION OF MOTION AND VISUAL SYSTEM Final Report

REQUIREMENTS FOR FLIGHT TRAINING SIMULATORS
" ~STI
6.PERFORMING
TR 546ORG. REPORT NUMBER

7. AUTHOR(s) 8. CONTRACT OR GRANT NUMBER(e)

Robert K. Heffley, Warren F. Clement, Robert F.
Ringland, Wayne F. Jewell, Henry R. Jex, and
Duane T. McRuer, Systems Technology, Inc. MDA 903.-80-C-0235
9. PERFORMING ORGANIZATION NAME AND ADDRESS 10. PROGRAM ELEMENT. PROJECT. TASK
"Systems Technology, Inc. AREA & WORK UNIT NUMBERS

2672 Bayshore Frontage Road, Suite 505 2Q263744A795

Mountain View, California 94043
It. CONTROLLING OFFICE NAME AND ADDRESS 12. REPORT DATE
Ro IDefenseenaon1.
Defense Supply Service - Washington August
UBE 1981
FAE

Washington, DC 20310 251

14. MONITORING AGENCY NAME a ADDRESS(If dlf.erart from Controlling Office) 15. SECURITY CLASS. (of thie report)

Department of the Army

US Army Research Institute (ARI) Field Unit UNCLASSIFIED
Ft. Rucker, Alabama 36362 IS&. DECLASSIFtCATION/DOWNGRADING

16. DISTRIBUTION STATEMENT (of this Report)

Approved for public release; distribution unlimited.

17. DISTRIBUTION STATEMENT (of th. aLbetrect entered in Block 20. If different from Report)

18. SUPPLEMENTARY NOTES

19. -KEY WOROS-(Continue on reveree eade If neceesary iad Identify by block nzmblr)
-Training Simulators Flight Simulation Army Aircraft
Pilot Training Training Devices Army Training
Transfer of Training Simulation Visual Perception
Flight Training Simulator Fidelity Motion Perception

20. A
A*TRACT f C W.-r*vWv
,atfnsu abb If nrcemsy and idoetify by block nu*beir)
• Fidelity requirements for Army flight trainink simulators, are explored
-using a manual control theory approach. The firsi step is to define
simulator fidelity" in operational terms which provide a basis for each
of the subsequent steps. This definition is accompan -d-Sy--a taxonomy-.ofL,
-measureable fidelity parameters. The next step, also of a preparatory
nature, is the analysis of Army flight training missions. It describes
how specific flight tasks and piloting techniques can ben terms

-DD I AN 73 -EDITION OF I NOV6GS IS OltSOLETE

1 SECURITY CLASSIFICATIOII OF IflS PGE (When Date Entered)

Vi UNCLASSIFIED
SECURITY CLASSIFICATION OF THIS PAGE(Wlrhn Date Entfamd)

Item 7 (Cont'd)
Vernon E. Carter, Northrop Corporation

Item 20 (Cont'd)

compatible with- feedback control theory. Pilot modeling techniques are

then discussed, first in terms of pilot control and then in terms of pilot
perception. Next, armed wiLh compatible descriptions of fidelity, the
training context, and pilot behavior, a procedure is described for studying
visual and motion stimuli. It is found, however, that there are serious
gaps in the experimental data base; and this precludes the systematic
execution of this procedure. Because of the lack of data, it is not
possible to accomplish -fully the original objectives and, therefore, a
formal bookkeeping scheme is outlined to guide the investigation of
fidelity requirements. Conclusions and recommendations are then drawn.
As aids to the reader, an executive summary and glossary of terms are
provided.

Accession For°V

IIDTIC GRA&I
NTIS TAB
1nannouxacd
ti
J-ustiricttio

iDi-t r ib i oný__..___-

Avail',bilitY Codes
"-- -lAvai 8and/or -

,i 1specilnA

iiUNCLASSIFIED
SECURITY CLASSIFICATION OF THIS PAGE(When Data Emored)
 Techmicol Report 546
A

DETERMINATION OF MOTION AND VISUAL

SYSTEM REQUIREMENTS FOR FLIGHT
TRAINING SIMULATORS
Robert K. Hoff ley, Warren F. Clement, Robert F. Ringland,

Wayne F. Jewell, Henry R. Jex and Duane T. McRuer

Systems Technology, Inc.

and
S ~Vernon E. Carter
Northrop Corporation

t. Submitted by:
AR IEDCharles A. Gainer, Chief
ARI FIELD UNIT AT FORT RUCKER, ALABAMA

Approved by:
Edgar M. Johnson, Director
SYSTEMS RESEARCH LABORATORY

U.S. ARMY RESEARCH INSTITUTE FOR THE BEHAVIORAL AND SOCIAL SCIENCES
"5001 Eisenhower Avenue, Alexandria, Virginia 22333
Office, Deputy Chief of Staff for Personnel
Department of the- Army

August 1981

' Army Project Number. Aircrew Performance and Simulation

S~2Q263744A795

Approvd for public release, distribution unlimited.

Ii
ARI Research Reports and TeChnical Reports are intended for sponsors of
R&D tasks and for other research and military agencies. Any findings ready
for implementation at the time of Iublication are presented in the last part
of the Brief. Upon completion o, a major phase of the task, formal recom-
mendations for official action nforonliy are conveyed to appropriate military
agencies by briefing or Disposition Form.

- '5,
 FORAN ORD

r; - "'the
The Army of
mission Research Institute
providing timelyField Unit at
research andFort Rucker, Alabama,
development in
support has

aircrew training for the US Army Aviation Center. Research and development
are conducted in-house, augmented by contract research as required. This
research report documents contract work performed as a part of the Field
Unit's thrust in flight simulation developmenL.

The development of validated requirements for flight simulator Sneci-

fications is an area of intense interest for both DoD and industry. In
this report, an approach is developed which will translace requirements
based in human perceptual capabilities to requirements compatible with
engineering specifications.
This work is responsive to Project 2Q263744A795, to the US Army

Aviation Center, Fort Rucker, and to the US Army Project Manager for
Training Devices (PM-TRADE).

JOS H ZEI R
Tec ical Director

J.

77j
 DETERMINATION OF MOTION AND VISUAL SYSTEM REQUIREMENTS FOR FLIGHT TRAINING
SIMULATORS

BRIEF

r Requirement:

At present, there is no qtlantitative methodolog•i for statement of

flight simulator visual and motion systems requirements in terms of train-
ing objectives. This study uses available data relating visual and motion
senses to pilot closed loop control and to spatial orientation and develops
such a methodology.

Procedure:

First, fidelity is operationally defined in terms of the simulator's

ability to induce the pilot trainee to output those behaviors and behavior
patterns known to be essential to control and operation of the actual
aircraft in performance of a specific task. From this definition, a
control theoretic model of simulation training is developed consisting of a
closed loop of three interconnected processes: the pilot's perception of
task environment and simulator states, his control behavior technique, and
the simulator's response. Then, the existing body of literature on simu-
lation and simulation requirements is surveyed for data relevant to pilot
perception and pilot technique.

Findings:

is A control theoretic approach to training simulator requirements

is shown to be a potehtfally powerful tool. But it is also shown that many
areas of pilot control technique and dynamic perception are not suffi-
ciently quantified. Li
Utilization:

The model developed here and the indicated areas in which additional,
more quantitative research is needed will serve as a basis for an alter-
native approach to investigation of simulator visual and motion system
requirements.

- vii
 ACKNhnLoDGEMENTS

The work reported herein was performed for the US Army Research
Institute Field Unit at Fort Rucker, Alabama, under Contract MDA 903-
80-C-0235. The Systems Technology, Inc., project engineer was Mr. Warren
F. Clement and the Army Contracting officer's technical representative was
Dr. William R. Bickley. Dr. Kenneth D. Cross of Anacapa Sciences, Inc.,
and Mr. Vernon E. Carter of Northrop were project consultants to Systems
Technology, Inc. The final report manuscript was prepared by Mrs. Sharon
A. Duerksen of Systems Technology, Inc. Work on the project was begun in
February 1980 and completed in July 1981.

The authors gratefully acknowledge the contributions made by Dr.

Kenneth D. Cross of Anacapa Sciences, Inc., and Dr. William R. Bickley of
the US Army Research Institute Field Unit. Their efforts greatly aided in
bridging the disciplines of control engineering and training psychology.

%I
iIx
Sii
Il
£ • •
 TABLE OF CONTENTS

Page

EXE CUTTIVE SUMMARY 1

Backg round 1

Objectives and Scope of This Study 1

Technical Approach 2

Definition of Fidelity 2

Army Training Missions 3

Pilot Modeling Techniques 4

Analytic Procedure to Predict Cue Stimulus 4

Simulator Fidelity Bookkeeping 5

Research Needed 5

Conclusions and Recommendations "5

"INTRODUCION 7

Background 7

Objectives and Scope 9

Technical Approach 10

DEFINITION OF SIMJLATOR FIDELITY 12

An Operational Definition 14

A Taxonomy of Measurable Fidelity Parameters 32

"implications for Detdrmining Fidelity Requirements 49

xiF
 Page

ANALYSIS OF ARMY FLIGilT TiRAINING MISSIONS 51

Army Training Objectives to be Considered 51

Guidelines for Describing Task Loop Structure 59

Specific Task Analyses 69 4

PILOT MODELING TECHNIQUES 80

Pilot Control Modeling 80

Perceptual Hodeling 105

ANALYTIC PROCEDURE FOR DETERMINING NECESSARY VISUAL AND

MOTION CUE STIMILI 140

Analytic Procedure 141

Observations on Fidelity Analysis 146

BOOKKEEPING METHODS FOR SIMULATOR FIDELITY 154

Piloting Technique Versus Pilot Task 156

Cues Available Versus Training Device 156

Aerodynamic Feedbacks Versus Aircraft Type 158 1

Essential Loops Versus ,Piloting Task and Technique 158

Essential Loops Versus Skill Level for a Given Task 160

Essential Cues Avai? able. Versus Essential

Loops Required 162

Training Capability by Task Versus Training Device 164

Macro Detail Cues Available Versus Micro Detail
Features Required 164.

xii
 Page

RESEARCH NEEDED TO DEFINE SIMULATOR REQUIREMENTS 168

Preparatory Analysis of the Piloting Task 171

The Array of Measurement Tools 177

Measurement of Piloting Technique and

Pilot Perception 183

Examples of Piloting Technique and Perceptual

Measurements and Interpretation of
Experimental Results 186

Motion Perception Summary 204 t"

Summary for Planning and Collecting Measurements 206

CONCLUSIONS AND RECOMMENDATIONS 210

GENERAL ALPHABETICAL BIBLIOGRAPHY 214 A

BIBLIOGRPHY OF PILOTING TECHNIQUES 223

BIBLIOGRAPHY OF VISUAL PERCEPTION 226

GLOSSARY 228

i• xiii
 LIST OF TABLES

Number Page

1 Examples of Motion and Visual Perception Functions 23

2 A Summary Definition of Simulator Fidelity 33

3 Simulator-Fidelity-Related States 38

4 State Transfer Relationship Characteristics,

[• Motion-Visual Analogies 46

5 Army Flight Training Objectives 53

6 Control Technique Taxonomy for Fixed-Wing-and

Rotary-Wing Aircraft 55

7 Visual Field Information Requirements: Orientation 117

8a Visual Field Information Requirements: Position 118

8b Examples of Visual Position Cues Provided to the

Approaching Pilot by the Geometry or Delineation
of Visual Elements 119

9 Visual Field Information Requirements: Translational

Velocity 120

10 Summary of a Sequential Pattern Perception Theory 122

11 Analytic Procedure for Determining Minimal Level

of Visual and Motion Cue Stimuli Needed to Train 142

I, Control Theory Applied to the landing Maneuver 151

1-3'- A List of Bookkeeping Forms for Simulator Fidelity 155 !

14 Examples of Cues Available in a Hypotheticcl I
Simulator Device 157

15 Aerodynamic Effects 159

16 Example of Piloting Technique_ Versus Essential Loops 161

17 Example of Essential Cues Available Versus

V Essential Loops Required- 163

xiv

"Y
_ _ _ _ __ _ _ _ _-__ _ _ _ _

- - - - - - - - - -- - **- -
 "umber Page

18 Training Objectives Requiring Flight Quantification

of Piloting Technique 169

19 Summary Characteristics of-Pathways in Perceptually

0Centered Mdel of Human Behavior 174
20 Task Breakdown 175

21 Task Analysis of Unmask/Remask Maneuver 176

22 Representative Examples of Available Data 185

S23 A Checklist for Planning anc Collecting Measurements

4 of Pilot Behavior 207

LIST OF FIGURES

Number Page

1 Tyi-pical Training Simulator Apparatus 18

2 Typical. Representation of Simulator Apparatus for the

Purpose of Fidelity Analysis 20 I
3 Comparison of Behavioral Parameters Adopted for Motion
Drive Logic to Those Adopted for Aircraft 25

4 Comparison of Behaviorial Parameters Adopted for Visual

Field Drive Logic to Those Adopted for Aircraft 26 :

5 Components of Simulator Fidelity 36

6 Effective Pilot Time Delay (Teff) Versus Motion Drive

Natural Frequency (wM) 43

7 Pilot-Vehicle Diagram of Acceleration/Deceleration

Maneuver - Utility Helicopter 60

8 Translation of a Verbal Task Description to a Pilot-

Vehicle Loop Structure (Helicopter Approach to Hover) 61 VI
9 Translation of a Verbal Task De~cription to a Pilot-
Vehicle Loop Structure (Straight Climb on Instruments) 62

10 Common Loop Structure, Hovering Flight

Task - Helicopter 63

xv
Nuaber Page

11 Common Loop Structure, Basic Flight Tasks

Fixed-Wing, Frontside Technique 64

12 Simplified Heading Regulation Task Loop Structure

with Quantification of Pilot-Vehicle Elements 68

13 Bpnk Angle Regulation Task and Piloting Technique 70

14 Heading Regulation Task and Piloting Technique 71

15 Heading Change Task and Piloting Technique 72

16 Pitch Attitude Regulation Task and Piloting Technique 73

17 Altitude Regulation Task and Piloting Technique 74

18 Altitude Change Task and Piloting Technique 75

19 landing Maneuver Task and Piloting Technique 76

20 Normal Acceleration/Deceleration Task and

Piloting Technique 77
21 Decelerating Approach to Ho ar Task and
Piloting Technique 78

22 NOE Quickstop Deceleration Task and Piloting Technique 79

23 Structural Isomorphic Model of Man-Machine System 85

24 Algorithmic (Linear Optimal Control) Model of

Man-Machine System 86

25 Flow Diagram for SOP Operations 89

26 Compensatory Pilot Model 96

27 Pursuit Pilot Model 99

28 Comparison ,of Pursuit and Compensatory Model

Describing Functions 101

29 Refinement of Central Process Models for Perception

of Displayed Signals (Non-Scanned Cases) 103 T

30 Threshold Model and Threshold Values for Each Axis 108-

31 Circular Pad Appearance in Vertical Landing Approach at

Constant Rate of Descent and Horizontal Deceleration From
Left Seat of XC-_142 Cockpit 129

xvi
 I Number
32 inference of Aiming Point Location From the Apparent
Page

"Motion Streamers" Associated with a Circular Pad in a

Vertical Landing Approach at Constant Rate of DE•scent
and Constant -Horizontal Deceleration 130
S33 Circular Pad Appearance in Vertical landing Approach at

Constant Rate of Descent and Horizontal Deceleration with 4

Reduced Visual Range From Left Seat of XC-142 Cockpit 132

34 RunwJay Appearance in a Short Landing Approach at 9 deg

Glide Angle and Constant Ground Speed From Left Seat
"ofXC-142 Cockpit 134

J 35 Runway Appearance in a Short Landing Approach at 9 deg

Glide Angle and Constant Ground Speed with Reduced
Visual Range From Left Seat of XC-142 Cockpit 135 4

36 Geometry for Perception of Time-Advanced

lateral Deviation 137

37 The Need for Altitude Rate Cues Depending Upon

Aircraft Type and Piloting Technique 153

38 Array of Measurement Analysis Tools 178

39 The Identification Process 182

40 Comparison of Deceleration Profiles Between Analytical

Model and Flight Test Data 187

41 Boundaries of Sway-Axis Washout Filter Parameters

{
[' Which Delineate the Pilots' Impressions of Realism
e From Combined Roll and Sway Motion Cues 200,'

43 Summary of Pilot Commentary for Bank and Stop

Maneuvers and Roll Tracking 202

xI.

,,1xvi
 LIST OF ABBREVIATIONS .

ADF Automatic direction finder

AGARD Advisory Group for Aerospace -esearch and Development

ARI Army Research Institute

ATM Aircrew Training Manual _Z1

CGA Ground controlled approach

CGI Computer generated image

CRT Cathode ray tube j

DG Directional gyro t
EPR Engine pressure ratio -

HD Heave damping mode

IAS Indicated airspeed

ICAO International Civil Aviation Organization

K, kt Knot(s)

IFR Instrument flight rules

IMC Instrument meteorological conditions

LAMARS Large Amplitude Multimode Aerospace Research Simulator

LD Lateral damping -mode I

LSF Lateral specific force

NASA National Aeronautics and Space Administration

NIPIP Non-Intrusive Pilot Identification Program

NOE Nap of the earth

0CM Optimal control model |,

rad Radians

xviii-
 RPM, rpm Revolutions per minutes

SAS Stability augmentation system

SCAS Stability and control augmentation system

SD Speed damping mode

sec Seconds

SFTS Synthetic flight training system

SOP Successive organization of perception

S.P.R.R. Southern Pacific Rail Road

STOL Short takeoff and landing

TV Television

UCLA University of California at Los Angeles

VASI Visual approach slope indicator

VFR Visual flight rules

VMC Visual meteorological conditions

VOR Very high frequency omnirange

VTOL Vertical takeoff and landing

II

S~xix
 LIST OF SYMBOLS

A (1) Subjective landing area size; (2) apparent distance of

vanishing points from the principles of visual perspective

A Controlled element gain for () state variable

A Peak lateral specific force

ax Axial specific force

ay (1) Lateral specific force, (2) specific side force

a Lateral specific force

CL Centerline
CLV
d Vertical flight path excursion

e System error

e Error rate

g Gravity constant

gy Spurious motion cues

h Altitude

hVertical velocity

Ix Roll moment of intertia

Pitch moment of inertia

iz Yaw moment of inertia

i System input

J Imaginary operator

K or k Gain
L() Roll stability or control derivative with respect to
LLr Roll stability derivative with respect to yaw rate

Lr Roll stability derivative with respect to bank

M( Pitch stability or control derivative with respect to ()

xx
 Mq Pitch stability derivative with respect to pitch rate

m Aircraft mass

p Roll rate

q Pitch rate

R Range

RRange rate

R Deceleration

r Yaw rate

S Ground range, size j

s Laplace operator

Tý Effective closed loop flight path lag

TI Pilot lag time constant A

TL Pilot lead time constant

TR Roll mode time constant

Ts Spiral mode time constant

SHelicopter pitch damping time constant v

T Effective closed loop airspeed lag

Tyd Helicopter yaw damping time constant

To Effective closed loop pitch attitude lag

ToI Pitch numerator time constant

TeT2 Pitch numerator time constant

!TS Effective closed loop bank angle lag

T Roll numerator time constant

U Airspeed

U0 Trim airspeed

V Speedp4

IJxx
 VG Ground speed

V Minimum control speed

VM Never exceed speed

Maximum angle of climb speed

L w (1) Runway width; (2) z-axis perturbation velocity
component

X X-axis force

Speed damping stability derivative

X() X-axis stability or control derivative with respect to ()

x Distance along x-axis

7c Controlled element transfer function

Yc(jW) Controlled element describing function

Ycab Simulator cab translational acceleration

Yp (1) pilot transfer function, (2) subjective lateral

deviation

YPC Tursuit command loop transfer function

Y Pursuit error loop transfer function

Y~i Pursuit input transfer function

YPm Pursuit state measurement loop transfer function

Yp(OW) Pilot describing function

YpYc(OW) Pilot-vehicle describing function

"Yv Lateral damping stability derivative

Y y-axis force stability or control derivative with respect
C) too
yLateral
Y' displacement

Z() z-axis force stability or control derivative with respect

to ()
SZw Heave damping stability derivative

xxii
a Angle of ettack

aaero Aerodynamic sideslip angle

(1) Actual depression from horizon, (2) nomihal glide

slope angle

A Incremental quantity

Control deflection

G Pitch attitude

0 Pitch rate

S8c Pitch ettitude command

1 Backscattering coefficient

'IV Eyeball position

V (1) Actual angle of centerline with respect to vertical,
(2) approach line up centerline perspective angle, (3)
K
nozzle deflection

7t 3.14159..

a (1) Real part of Laplace operator, (2) Koschmeider's

meteorological extinction coefficient
STime delay

Tc Computing delays
Te Effective pure time delay

Teff Effective pilot time delay

* Bank angle

Heading

Wc Crossover frequency

Wd Dutch roll frequency I

Motion drive natural frequency

wPL LLateral phugoid frequency-V

to•p Phugoid frequency

xxiii
 ••sp Short period frequency

Wy (1) High pass break frequency, (2) sway washout filter

frequency

.a ,ateral control (ailerou)P

-beam Simulator support beam

c (1) controlled element,

control (2) command, (3) collective -4

cab- Simulator cab

h Height

e (1) Error, (2) elevator control

Si Lag

L Lead

m Measurement

* 0 Objective

*. P (1) Pilot, (2) perceived

ppkl Peak

R Range

r Rudder control

U Airspeed

T Thrust control

t Threshold,

v Visual

x Along aircraft fuselage reference line

y Perpendicular to fuselage reference line in lateral plane

xxiv

S- _ - ._-- . ,-...-, :- =-- - .. : _-.• .•

 z Perpendicular to fuselage reference line in vertical plane
Pitch attitude

Heading, yaw

xIi

LLi
.L

1j

• iI

* I

K Iv
 EXECUTIVE SUMMARY

Visual and wotion simulator systems should be designed to address the

specific training objectives which are of importance to the mis'sions of
Army aviators, but training objectives do not influence simulator designs
as they should. There is presently little rationale for setting simulator
specifications with regard to the specific training objectives of those
simulators, and it is necessary to rely heavily upon past experience. If
an existing simulator with given motion and visual system characteristics
provides a successful transfer of training, then it is assumed that the
same specifications- should be used for the next simulator even though some
of the specifications could be superfluous. Unfortunately there are no

rules to tell us what to do if training was not successful. Was the. de-
ficiency in the motion or visual system? What must be done to correct
that deficiency? What must be provided for new aircraft -or new training
objectives? These questions lead us to the objectives of this study.

OBJECTIVES AND SCOPE OF THIS STUDY

The main objective of this study is to develop a methodology for ap-

proaching training simulator fidelity in terms which will ultimately J
permit rational system specification. A secondary objective has been to
make use of available data where possible in order to exercise the meth-
odology adopted and to set simulator fidelity requirements where it is
possible. The scope of this study includes motion and visual fidelity

considerations for a wide range of Army training objectives. These in-

clude fixed- and rotary-wing Army aircraft, operations in visual and
instrument meteorological conditions, undergraduate through continuation

ip -
 ?'1!

training, and recognition of critical flight phases which include nap-of-

the-earth navigation, and weapon delivery.

TECHNICAL APPROACH

The approach taken in this study is

to formulate an operational def-
inition of fidelity which decomposeg the various aspects of fidelity which
include training objective or task, piloting technique, pilot perception,
and simulator model software and simulator hardware response. This de-
composition yields a framework quantified in common terms which, in this
case, are based on control theory ideas. The ultimate benefit of this
approach is that the same terms used to describe the various components of
the training scenario are, in fact, compatible with the terms used to
write engineering specifications for simulator components. In effect
there is a common frame of reference for system performance and fidelity
measurement. Another important byproduct of this technical approach is
that a number of natural constraints are automatically imposed as a result
of quantification of the task and aircraft type. Further constraints are
imposed by recognition of basic ideas from manual control theory. Thei
ultimate reward is an analytic formulation of the task-pilot-aircraft
system which describes the basic mechanisms of :ra-ining to fly an aircraft
as well as pilot performance with respect to desired flight tolerances.

DEFINITION OF FIDELITY

This effort began with ,the development of an operational definition of

simulator fidelity which sets the stage for all subsequent steps in i
achieving the study objectives. Briefly stated in verbal terms,

2
 "Simulator fidelity is the degree to which characteristics
of perceivable states induce correct psychomotor and
cognitive control strategy for a given task and envi-
ronment. This leads to s'ecial consideration of essential
feedback loops required to execute a task and the esse-
tial cues provided by the simulator or training device
which support those essential loops."

This definition of fidelity can also be depicted in a graphical form which

clearly identifies the ideas of objective fidelity and perceptual fidel-
ity. The implication is that it may not be necessary to have a highly
veridical situation for skill development except in terms of the induced
piloting technique.

AIfM TRAINING MISSIONS

A statement of fidelity is unavoidably conditional upon the training

mission or task being addressed. It is necessary first to specify the
various training- missions to be considered for simulation, and this is
done through direct reference to training literature and training syllabus
material, supplemented by discussions with instructor pilots. This in-
cludes aircrew training manuals, aircraft flight manuals, descriptions of
special missions such as helicopter gunnery or nap-of-the-earth flight.
It also involves reviewing training material from other branches of the A
military as well as from civilian aviation. Guidelines are presented
describing how to translate the training literature into a closed-loop.
system structure which is compatible with the engineering descriptions of
the aircraft and the pilot psychomotor behavior. A number of task an-
alyses are provided which illustrate the closed-loop nature of the task
and provide some idea of the ranges of numerical quantification.

r 3 U

___ _! __.J
 PILOT MODELING TECHNIQUES

One of the central themes of the technical approach is to use manual

control theory in pilot modeling techniques to aid in the analysis of
simulator fidelity requirements. Some of the basic ideas in pilot model- -

ing have been well developed; and it is possible to predict the nature of -

pilot behavior, espectally psychomotor, given a good description of tho

task and of the vehicle. One of the more difficult aspects of pilot I
modeling is handling perceptual mechanisms, especially where visual in-
formation is involved.

ANALYTIC PROCEDURE TO PREDICT CUE STIYLUS

With the ideas developed under the foregoing topics definition of

*• fidelity, description of -Army training missions, and pilot modeling tech-
-niques - it is possible to devise a procedure which informs us of the
cues required to perform training missions and whether those cues are, in
fact, available in a given training device or simulator. This is a guide ii
to discovering what cues are required and what cues may be lacking. Exe-
cution of this procedure requires a rather full numerical quantification
of the task, piloting technique, vehicle and simulator response, and pilot
perception. At this stage we find that the fragmentary nature of the
available- quantification of these characteristics prevents us from- an

effective analysis of visual and motion system requirements in general.

Nevertheless it begins to point out what is needed in terms of additional
research.
L

li
4 Ii•

-- - - - - - -- -- - - ----- - - -- - - - - - - - - - - - - - - - -
 SIDJLATOR FIDELITY BOOKKEEPING

This step is a formalization of a procedure to take stock of simulator

-fidelity requirements. We specify a series of matrices which are com-
prised of basic data describing piloting technique versus task, cues
available for various training devices, aerodynamic influences *of the
aircraft involved, and a series of matrices of constructed information
which lead us to an ultimate statement of the training capability of par-
ticular devices and the motion and visual cue detail which must be
specified when constructing a new device.

RESEARCH NEEDED

This study demonstrates how simulator fidelity requirements can be

specified in a rational way to address the wide range of Army training
missions. At the same time, however, we find repeatedly that there are
serious gaps in the basic experimental information required to exercise
the rational analytic procedure. These -gaps lie primarily in the pilot-
centered areas involving piloting technique for specific tasks and pilot
perception of motion and visual cues. In effect the data required are
those which would fill out the basic simulator fidelity bookkeeping ma-
trices previously described. The kinds of measurements to acquire these
data are, by and large, feasible and can be comprised of several different
approaches which can be applied simultaneously to any given flight or
simulator research experiment.

V
(X)NCLUSIONS ANDl RECOMINDATIONS
A
The application of control theory to training simulator fidelity of-
fers a poteutially powerful tool. All of the task, physical, and

5
 SIMJLN-TOR FIDELITY BOOKKEEPING

This step is a formalization of a procedure to take stock of simulator

fidelity requirements. We specify a series of matrices which are com-

prised of basic data describing piloting technique versus task, cues

-available for various training devices, aerodynamic influences *of the

aircraft involved, and a series of matrices of constructed information

which lead us to an ultimate statement of the training capability of par-

ticular devices and the motion and visual cue detail which must be
A specified when constructing a new device.

RESEARCH NEEDED

This study demonstrates how simulator fidelity requirements can be

specified in a rational way to address the wide range of Army training
missions. At the same time, however, we find repeatedly that there are
serious gaps in the basic experimental information required to exercise
the rational analytic procedure. These gaps lie primarily in the pilot-

centered areas involving piloting technique for specific tasks and pilot
perception of motion and visual cues. In effect the data required are

those which would fill out the basic simulator fidelity bookkeeping ma-

trices previously described. The kinds of measurements to acquire these

data are. by and large, feasible and can be comprised of several different
approaches which can be applied simultaneously to any given flight or

simulator research experiment. ii

CONCLUSIONS AND RECOMBENDATIONS

The application of control theory to training simulator fidelity -of-

fers a potentially powerful tool. All of the task, physical, and

Iz
pilot-centered constraints are brought together in a way which can produce
a meaningful basis of analysis. However, what confounds the execution of
this procedure is a lack of quantification in some of the pilot-centered
areas of control technique and perception. At the same time it is fea-
sible to go after these needed measurements. This can be done over the
long term using targets of opportunity or over a shorter term with delib-
erate, well-planned research programs involving both flight vehicles and
simulators.

In laying out a rational approach to determination of training simu-

lator fidelity requirements, one of the important byproducts is a
unification of training literature which describes various piloting tasks

and the established ideas of manual control theory. This unification is

useful in describing and predicting pilot behavior given quantification of
the task. These ideas, combined from the training and engineering coMr-
munities, provide some interesting implications for both of these
communities. For example it may be possible to enhance pilot training
methods through use of some of the piloting technique measurement methods
suggested. These are ways of describing how a pilot is carrying out a
particular task as opposed to how well certain standards are met, i.e., it
focuses on the mechanisms of performing a task. From the engineering
point of view, the formalized lists of tasks and task descriptions which
are presented in the training literature provide some useful views of how
better to formulate design objectives, e.g., if the piloting task is suf-
ficiently well defined, then it should be possible to develop clear
rationale in areas such as aircraft flying qualities and performance.

I
|•.
 'I ý
INTRODUCTION

•I NLA(ECROUND

Design specification for training simulators involves answering ques-

tions such as: How much motion travel? - How good a visual scene? Much

effort has been expended in trying to answer these questions with few
1
definitive results, yet economic pressure to use training simulators in

place of actual flight training steadily increases.

Attention frequently focuses on simulator fidelity as being the cul-

prit in many simulator training problems; but, as s' --te have pointed out:

"There is no compelling relationship between training effectiveness and

fidelity/realism.. And to the extent that this is a true statement, we
may want to be careful not As AGARD Working
to overdo realism aspects.
SGroup 10 points out, the question of fidelity involves perceptual aspects;

and requirements for these perceptual aspects are not well known with 51I

respect to training3 . This is an area, therefore, wtiich needs study. It

IAdams, Gerald H., (Ed.), Flight Simulator Training. Volume 1:

1964-1973 (A Bibliography with Abstracts), NTIS/PS-77/0145, March 1977.

Habercom, Guy E., Jr., (Ed.), Flight Simulator Training. Volume 2:

1974-February 1980 (Citations from the NTIS Data Base),
NTIS/PB80-806425, March 1980.

Anon., 50 Years of Flight Simulation. Conference Proceedings.

Session 1 through 3, 1979.

Puig, Joseph A., William T. Harris, and Gilbert L. Ricard, -Motion in

Flight Simulation: An Annotated Bibliography, NAVTRAEQUIPCEN IH-298,
July 1978.

Training In-Flight and Emergency Procedures, AGARD-AG-248, June 1980.

3 Key, David L., (Ed.), Fidelity of Simulation for Pilot Training, AGARD
Advisory Report No. 159, October 1980.

7 .i

-. -
 7--i - 'AZ

is of particular interest to focus on the issue of visual simulation

fidelity in connection with the difficult training objectives involving
nap-of-the-earth (NOE) operations4.

The Army Research Institute (ARI) Field Unit at Ft. Rucker, Alabama,
is engaged in conducting flight simulator research in three areas:

1. Requirements - the definition of training device require-

ments to meet the training objectives with emphasis on
visual and motion systems.

2. Evaluations/Validation - in terms of training transfer,

does the flight simulator meet its design goals? Further,
do the Visual systems provide sufficient information to
train the Army mission?

3. Utilization - research is in progress to determine the

training effectiveness of the Army's synthetic flight
training system (SFTS).

In recent rears, the Ft. Rucker Field Unit has been engaged in near-term
research to answer some questions in the three categories defined above.
In some cases, information is not sufficient to provide complete and ac-
curate answers, particularly in the areas of visual and motion system
fidelity.

In order to -approach training simulator fidelity in a rational way, it

is necessary to develop some cause and effect relationships. For
example: How does the pilot respond to stimuli, What happens to the air-
craft, and How is the task carried out? For the most part simulator

V Roscoe, Stanley N., Halim Ozkaptan, and Aaron Hyman, Review of Flight

Training Technology, U. S. Army Research Problem Review 76-3, July

1976.

Gainer, Charles A., and Dennis J. Sullivan, Aircrew- Training

Requirements for Nap-of-the-Earth Flight, U. S. Army Research
Report 1190, August 1976.

4 -.----
___ ___ ____
.,- ___ ~4 .-
 training literature 5 does not concentrate on piloting strategy or tech-
nique, rather on systems performance (scoring, tracking dispersions,

success/failure). Without an understanding of the pilot's actions, we are

dealing with a distinctly closed-loop system on an open-loop basis, that
is to say, we know just the command and the final response. But this
'4 limited knowledge of the command and the response confounds the analysis
of the factors affecting pilot actions.

There are,- however, established methods for dealing with the mechan-
isms of pilot behavior, especially'- on a psychomotor level. McRuer and -1

Krendel 6 outline a theory of manual control which gives a rationale for

i how a pilot operates, adjusts, and -fits into an overall pilot-vehicle
relationship. Various approaches to this theory include not only clas-
Ssical but also modern control theory techniques. Also there is an
existing manual control theory data base, although it is somewhat limited
in terms of completeness of piloting task. Thus there is a role for con-
trol theory in formulating a complete picture of the pilot in the loop.

, OBJECTIVES AND SCOPE

In view of the background outlined above and especially the 'motivation

provided by AGARD Working Group 107, the objectives of this study are to
identify simulator requirements for training by using available data on
motion and visual senses in combination with maiiual control theory. There
will be no original data collection involved; rather, a review of the

Adams, 1977, op cit.

Habercom, 1980, op cit.'

6 McRuer, D. T., and E. S. Krendel, Mathematical Models-of Human Pilot
Behavior, AGARD-AG-188, January 1974.

Key, 1980, op cit.

9,~~:
literature and an expression of a -methodology for addressing simulator I
requirements will be presented. The specific tasks to be addressed in
this study include: I

1. Definition of the term "simulator fidelity" and develop-

ment of a taxonomy of measurable parameters of simulator
fidelity.

2. Development of procedure for analytically determining the

type and quality of visual and motion cues required to
train flight skills in a simulator

3. Development of a methodology to define visual and motion

cue requirements by training objectives

4. Determination of commonality among training objectives in

terms of fidelity requirements for visual and motion
systems

5. Recommendations for research to acquire additional empiri-

cal data where it is deemed to be necessary.

The scope of this study includes a full range of .,my aircraft including
fixed-wing and helicopters, skill levels from undergraduate through con-
tinuation- training, both instrument and visual meteorological conditions,
and a variety of tasks including the critical scenarios involving NOE, and
weapon delivery.

TE CBNICAL APPROACH

The technical approach applied to this project is reflected by the

specific organization of this report. We begin by defining simulator
fidelity in operational terms which provides a basis for each of the sub-
sequent steps. This definition is accompanied by a taxonomy of measurable
fidelity parameters. The next step, also of a preparatory nature, is the

analysis of Army flight training missions. Here we describe how specific

flight tasks and piloting techniques can be cast in terms compatible with

10
 feedback control theory. Pilot modeling techniques are then discussed,
first in terms of pilot control and then in terms of pilot perception.
Next, armed with compatible descriptions of fidelity, the training con-
text, and pilot behavior, we describe a procedure for studying visual and
motion stimuli. Finding that there are serious gaps in the basic experi-
mental data, however, prevents the systematic execution of this
procedure. Thus, because of the lack of experimental data, we outline a
formal bookkeeping scheme to guide our investigation of fidelity require-
ments. This leads us finally to a discussion of the kind of research
needed and the applicable. measurement methods for defining fidelity re-
quirements. Conclusions and recommendations are then drawn. As aids to
the reader, an executive summary and a glossary of terms are provided.

The means to achieving the previously listed objectives is to apply

established data on human operator behavior, aircraft response, sensory
perception, and piloting technique with regard to the issues of simulator
fidelity and pilot training. A new area to be addressed consists of the
definition of piloting task (training objective) in terms which are com-
patible with the quantification of the pilot-vehicle system.

Because of the potential complexity due to the compounding effects of

several components (tasks, piloting technique, piloting perception, air-
craft model, and simulator response) it is desirable to minimize the

complexity of each of these components; but simplification is a worthwhile

goal, anyway, since it helps to isolate and emphasize the very important
parameters and features. Fortunately there are many examples which help
to show the way for system simplification.

A key feature of the technical approach is to acknowledge the bounds

and constraints imposed by the physics of the vehicles, the physiology of .
the pilot sensors, and by the rules and criteria of control theory (con-
trollability, response, bandwidth, stability, settling time, etc.). This
greatly aids in understanding what is important or essential to
training. In fact, a major objective of the technical approach is to
identify the essential behavior in terms of essential loops and the cor-
responding essential cues which support these loops.

V1
 Mr'
1- L

DEVIPKTION OF SIMJLATOR FIDELITY

This section addresses the topic of simulator fidelity in two steps:

"* First, definition of simulator fidelity in a manner

which is useful for the determination of motion and
visual system characteristics necessary for flight
,training.

"* Second, translation of this definition of fidelity into

a taxonomy of measurable fidelity parameters.

These two steps, taken together, provide the basis of the subsequent final
report topics.

The underlying approach used in the execution of this entire effort is

to formulate simulator requirements using a combination of (a) manual
control theory, (b) knowledge of human perceptual mechanisms, (c) defini-
tions of specific Army flight missions and tasks, and (d) conventional
engineering mathematical models of simulator components. The unifying
element will be the use of compatible mathematical terms for each of these
four components. -

I
With regard -to fidelity, we shall begin by reviewing some current
notions of fidelity which, in turn, will be incorporated into a special
definition of fidelity suited to our approach. This definition will then
lead us to consider parameters which are explicitly related to fidelity. 1]
It will be noted that these parameters are fundamentally separate from the
more implicit fidelity parameters commonly used to specify simulator Sys-
tem requirements. We will thus distinguish between "explicit" and
"implicit" fidelity in this section.

The strength of the approach to be applied to simulator fidelity re-

quirements is its potential to cover many diverse flight tasks and
environments, various stages of skill development, and the several modali-
ties used by the pilot. The approach can provide rational answers to [
12
 questions of simulator fidelity where there is adequate quantification of
critical elements, and it can aid in identifying where there are gaps in
such quantification.

Finally, before taking up the matter of simulator fidelity, the fol-

lowing ideas are offered for prefatory consideration. Each of these

notions which plays a role in fidelity will be further discussed and de-
veloped in the sections to follow.

* Transfer of training from simulator to flight can be

enhanced by identifying the pilot's organization of
perception and technique in flying tasks and by similar
organization of perception and behavior in the
simulator.

0 The task objective, the dynamics of a vehicle, and the

realities of the environment dictate the behavior of
the pilot and organization of perceptibn.
0- Cognitive processes determine what is desired or re-
quired by the pilot to accomplish the task objective.

0 Psychomotor and cognitive actions tend to be a com- 3

parison of what is with what is desired followed by a
commensurate action to achieve the desired.

* The control strategy and- technique of the pilot is

induced by the dynamics of the vehicle being con-
trolled. The actual aircraft is the baseline
vehicle - the simulator is to some degree a distortionI
of the aircraft and the environment.

* The pilot reacts to the dynamics of the vehicle only if

they are detectable.

0 Training manual task descriptions and standards along I

with descriptions by skilled pilots can be interpreted
as mathematical control laws. I
* Training is manifest by a successive organization of
perception and the evolution of control strategy in
pilot technique. It requires formulation of pilot
objectives, detection of dynamic quantities, and appro-
priate action of controls.

f 13
 v ý!_4

AN OPERATIONAL DEFINITION

Some Notions of Fidelity

Any consideration of fidelity connected with the acquisition of a

simulator system has, in the past, focused more on what the siilator does 4

than how the pilot responds. This is understandable since the mechanical,
electrical, and computational specifications must ultimately be defined

for the simulator fabricator. Unfortunately there is nothing in such

specifications, even if providing adequate fidelity in one simulator, that
would guarantee the same degree of fidelity in another. It is widely

accepted that the adequacy of a simulator is highly conditional upon ve-

hicle type, flight task, flight environment, and pilot skill level. Such
conditionality, in practice, generally precludes the simple extrapolation
of simulator fidelity requirements from one system to another.

The first step of- this study is to derive a working definition of

simulator fidelity which ':.ill serve as a basis for quantification of fi-

delity parameters. This can be accomplished by refining and expanding
some existing notions regarding fidelity. More specifically, we shall
show how some perceptual and behavioral, aspects can be quantified for the
purposes of specification.

We shall begin by considering some simulator fidelity ideas which are

useful for our purposes. The AGARD Working Group I08 presents a discus-

sion of fidelity which distinguishes two main "types" of fidelity:

objective fidelity and perceptual fidelity.L

Objective fidelity (or, perhaps more precisely, engineering fidelity 9)

is the degree to which the simulator reproduces measurable aircraft states

or conditions. In terms of motion fidelity, perfect engineering fidelity 3

Key, Ibid.

9 Sinacori, John B., Piloted Aircraft Simulation Concepts and Overview,

-NASA CR-152200, March 1978.

14
 would correspond to a one-to-one duplication of inertial-based displace-
ments, velocities, and accelerations in each axis of freedom.

Perceptual fidelity is the degree to which subjects perceive the simu-

lator to duplicate aircraft states or conditions. This type of fidelity
is pilot-centered and includes both psychological and physiological ef-
fects. We shall not, however, concede that perceptual fidelity is either
unmeasurable or unquantifiable. In fact, our technical approach is based I
largely upon ultimately quantifying or describing how to quantify percep-
tual effects.

__iI To the extent that the human operator's perception can be explained inl

rational terms, it is possible to merge the ideas of engineering and per-

ceptual fidelity. For example, since the human vestibular system -can be
described in terms of effective washouts, lags, and- thresholds, then it is

possible to apply the same objective- engineering metrics as one does to a

mechanical motion base platform, an electrical network, or an airplane
equation of motion. We shall develop this idea shortly.

I .Ancther aspect of fidelity which needs to be addressed is that of

induced pilot control strategy and technique. A recently convened NASA
I Advisory Subcommittee 1 0
defines simulator fidelity as the adequacy of
perceptual effects and their consequent pilot response behavior [i.e.,
control strategy and technique] induced by the simulator. Furthermore
this is attached to a specified task environment. The issue of control
strategy and technique is, of course, central to learning and skill de-

. velopment. If the simulator cannot induce correct technique, -then its

role in training Is questionable. At the very least, failure of a simula-
I tor to induce certain features of correct technique should be duly noted.

It is further suggested that the pilot response induced by a training

simulator for a given task should normally include the. errors committed

11 10 Ad Hoc Advisory Subcommittee, Avionics,, Control and Human Factors.

515

~ 1r
while learning to perform the task in the aircraft. CarterII suggests the
use of the term "error fidelity" to formalize the idea of evaluating the
fidelity of a simulator for a specific task in terms of the similarity
between the errors occurring in the simulator and those occurring in the
aircraft. This concept appears to provide an additional framework for the
systematic determination of fidelity requirements which has not been ex-
plored in past studies. According to this concept a simulator has high
error fidelity for a given task when:

1. Students tend to make the same errors in the simulator

that they make when learning the same task in the
aircraft.

2. The relative frequency of the different errors asso-

ciated with the task in the simulator is approximately
the same as in the aircraft.

3. The effect of each student error on system perfor-

mance, flight path, and maneuver outcome in the
simulator is the same as in the aircraft.

Error fidelity can have a very significant impact on transfer of

training. Student errors which do not have the same effect on system
performance and maneuver outcome in the simulator as they do in the air-
craft can lead to serious negative transfer of training. For example, a
student control error which would cause the aircraft to depart from
controlled flight and result in an undesirable maneuver outcome in the
aircraft must be made to have the same effect in the simulator. Error
fidelity thus represents an important simulator design goal.

Carter, V. E., and C. A. Semple, Specific Behavioral Objectives for

VF-121 Training in Basic Air Combat Maneuvers. Volume 3,
(CONFIDENTIAL), NOR 76-52, March 1976.

16

oI
"- -, - . . -.

An Overview of the Pilot in the Simulator

We have mentioned several aspects of fidelity including:

0 Objective fidelity

* Perceptual fidelity

* Induced pilot control strategy and technique "

* Error fidelity

Our approach is to quantify each of the above to the extent possible and
to form a model of the combined pilot-simulator which can then be compared
to a. like model of the combined pilot-aircraft." We shall explain this
approach in several steps.

First, consider the diagram of a typical training simulator apparatus

as, depicted in Fig. 1. Three main components are shown, the digital com-
puter, the visual field synthesizer, and the cab. Inside the cab the
pilot is provided with information based on motion, instruments, audio,
and outside visual scene. Pilot behavior is then manifested by control
actions, which feed back to the computer. Setting the pilot aside, all
the components in Fig. 1 have widely accepted means of quantification.
That is, the so-called engineering fidelity of each can be stated in rea- K
sonably direct terms. In fact, such terms form some of the commonly used
fidelity parameters which are used to specify system performance.

-A

17

- -,*
 f.q

CC

[ IK
Lo
-A 4-

4-' C'n

_r I - -

r4-

- 18
Relevant training simulator examples include the Army's UH-l requirement
12
document and UI1-60 specification1 3 .

The key to providing a more wholly rational scheme of expressing fi-

delity depends on how we address the block labeled "pilot" in Fig. 1.
This is, in fact, precisely the matter of handling perceptual and induced
behavior aspects of fidelity.

Induced pilot behavior consists of making adjustments in piloting

technique within a given overall structure of piloting technique. The
adjustments can consist of loosening or tightening-up the regulation of

flight parameters (attitude, airspeed, etc.), making gentle or aggressive

maneuvers, and making use of the learned response of the aircraft by em-
ploying increasingly more precognitive control actions. A convenient
means of viewing the adjustment of pilot behavior is outlined by McRuer
14
and Krendel . Based on substantial experimental research, rules for
pilot psychomotor behavior are expressed in terms which are compatible
with the "controlled element" - the simulator or the actual aircraft, as
the case may be. We shall cover this more thoroughly in the section on
pilot modeling, but it is useful to present some overview at this point.

Figure 2 shows an expanded pilot block and labels it as a "structural

isomorphic model." That is to say that each important functional aspect

of the pilot is accounted for in the model. Functions include the sensory
(perceptual) mechanisms, the central elements (cerebrospinal system),
neuromuscular actuation system, and the controlled element (parts of the
aircraft which are closely tied to the human operator dynamics such as the
cockpit manipulators). It is important to note that, having considered

Radder, Preston W., Capt., Department of the Army Approved Qualitative

Materiel Requirement (QMR) for a Synthetic Flight Training System
(SFTS) (Rotary Wing), 28 August 1972.

Schalow, P. S., Specification for UTTAS Helicopter Synthetic Flight

Training System, Device 2B38, Naval Training Equipment Center Report
No. 2222-1152, 30 October 1975.
14 McRuer and Krendel, 1974, op cit.

19
 I Iz

I CL

14 o

F Iz

Io w
It INo

L L7J LIj- L-$

r I

II _Z_
L--

20 I -
 the structural isomorphic model, we are able to reduce its complexity to a
lower level by using an effective system pilot model.

The specific structure of the piloting technique used to carry out a

given task is already well established. That Is. there are prescribed
functions for each aircraft control and these functions are conveyed to
trainees at an early stage by instructors and by training manuals. We
shall rely heavily on the latter medium in order to construct formal
analytic models of piloting technique appropriate for each training ob-
jective. An example of this is given in the section to follow.

The modeling of perception requires consideration of sensory dynamics,

cerebral processing of sensor information, and division of attention among
multiple sensors and sensor channels. In the following discussions, how-
ever, we shall limit ourselves to the first two aspects.

Human perceptual mechanisms and their dynamics have been described in

15
a number of sources. The motion perception models described by Ormsby
are convenient and adequate for our purposes. A number of sources go on
to deal with how motion perception is used in the performance of certain
basic task components.16 Visual perception is more complicated, how-
ever. Hennessy, et al.,17 describes in considerable detail the many
dimensions of visual perception and emphasizes its importance in simulator
fidelity research topics. But visual perception can also be handled in
ways compatible with our closed-loop approach. Namely, visual perception
may be quantified and tied directly to pilot behavior during specific

15e

S15 Ormsby, C. C., Model of Human Dynamic Orientation, MIT Ph.D. Thesis,
January 1974.
16 Hofmann, L. G.,
and Susan A. Riedel, Manned Engineering Flight
Simulation Validation. Part I: Simulation Requirements and Simulator
i Motion System Performance, AFFDL-TR-78-192, February 1979.
17
Hennessy, Robert T., Dennis J. Sullivan, and Herbert D. Cooles,
Critical Research Issues and Visual System Requirements for a V/STOL
Research Simulator, Canyon Research Group, Inc., Final Technical
Report No. RTH-0180 40-3053, 14 June 1979.

21
tasks 1 8 . Additional background and insights into visual perception, its
measurement, and its modeling will be discussed in the section on pilot
modeling; but Table 1 gives several examples of how motion and visual
perceptual mechanisms can be modeled for use in determining simulator
fidelity.

An additional fidelity notion we shall mention at this point is that

of pilot model adjustment based on vehicle dynamics. According to manual

-
19
control theory1, the pilot tends to adapt his control behavior in a way
which complements the dynamics of the vehicle simulator or aircraft.
The specific rules for adaptation are reasonably well understood and have
been verified experimentally on many occasions. We shall dwell upon this
matter also in a subsequent section.

For now, however, we would propose that one measure of fidelity is how
well the simulator induces pilot control adaptation suitable for the ac-
tual aircraft. -This would be evident from direct measurement of features
involved in the pilot model and by analytically comparing features so
measured in a simulator to features so measured in an actual aircraft as
shown in Fig. 3 (for motion stimuli) and Fig. 4 (for visual stimuli).
There are several techniques available for measuring pilot psychomotor
behavior. A general discussion of these techniques is given

18 Sinacori, 1978, op cit.

Clement, Warren F., Robert K. Heffley, and Wayne F. Jewell, Field-of-
View Requirements for Approach and Landing of V/STOL Aircraft, NADC
77240-07, August 1978.

Heffley, Robert K., "A Model for Manual Decelerating Approaches to

Hover," Proceedings of the Fifteenth Annual Conference on Manual
Control, AFFDL-TR-79-3134, 1979, pp. 545-554.

Heffley, R. K., T. M. Schulman, R. J. Randle, Jr., and Warren F. '

Clement, Analysis of Airline Landing Data Comparing Pilot Training in

a Simulator Versus Actual Flight, Forthcoming NASA Technical

19
Memorandum, 1981.

McRuer and Krendel, 19v

1974, op cit.

22

~~;y
- -22•/
9

TABLE 1

EXAMPLES OF MOTION AND VISUAL

PERCEPTION FUNCTIONS

2 0
0 Perception of Rotational Motion

Response of Motion
Semicircular Canals Threshold Subjective
-Velocity s Velocity
'• + 0.)0 s 1s 1 b.

(Task Dependent)
2
* Perception of Translational Motion l

Response Motion
of Utricles Threshold Subjective
Force (s + 0.08) Acceleration
tis ý0•.o2)(o.ls 1)

(Task Dependent)

* Perception of Visual Height Subjective

Angle Ale-to-

ITransformation
Perception Unight
Perspective Threshold Scaling
Sub ective
Height
Actual Height, h a" p

'- h
hActual Subjective
Runway Width, w Runway
Width, wp

22
* Perception of Visual Range

Perceived Range Perception

Transformation Threshold
S• S Subjective
SActual Range, R Range, RL

Subjective
Landing Area
Size, A

20 Hofmann and Riedel, 1979, op cit.

21 Ibid.

22 Heffley, 1979, op cit.

23

--2 ,
 -b ------

TABLE 1 (Concluded)

L Perception of Visual Lateral Deviation

2 3

Subjective
Rotation Angle-to-
Angle Lateral
Perspective Perception Deviation
Transformation Threshold Scaling Subjective

Lateral Deviation, y Deviation, Yp

S~Subjective
Actual Range, R Height, UP
Actual Depression
From Horizon, 7
and Actual Angle of
Centerline With Respect
to Vertical, v

* Perception of Visual Bank Angle

Rotation
Angle
Perception
Threshold
Actual Perceived
Bank Angle Bank Angle

F Perception of Visual Pitch Attitude or Heading

Subtended
Angle
Perception
Actal Threshold Subjective
Pitch or Pitch Attitude

S~Subjective
Scaling
• Factor, k

-i _I
23 Clement, Beffley, and Jewell,.1978, op cit.

S~24
 0I w

4J 4

4r,
0 0) 0
4

0 -
0

94 0.d im

43 0.0

00

10 HO cs0

J
0..
0

14 0 0$-c
0 A Z

13

p
0V

. $J4
>.4
toE34
~ 04 0 0 04'0 4 0

40
ý42
r- 0
0 -
O
.2: 0

4 '

0z w 0
t OW0

c5 A.
C5~
34;*4

25.-0
 L4I

,4)04 1
LO 43 4 4

k 04.k
o0,-

0 C) 0

0 W 0 W

U)

00

C-O>
0 1 c. to 04
A 1 ) g,4 .

- C0 0

~1C4 0 4' v. Q 0 4'

0'0 0 W 43 0 /
z) u Id VC) U

43 04..,
2 443 - t 4 4
E $4

0) 0- k 05 0

0 ~
as 0

O0 0

> 0

0 c 0 04 4J

0 0

W '.4 .dW

0i1 rz.

< 41~ o
t

riw 0 9:
C0 C26

---- ~-43 0 24,1 op-.--.---

csi
 Si

by Clement and will be specialized more to our needs later in this¶

report.

"It should be pointed out that a large body of literature already

exists in which pilot measurements have been made for specific elements of
many tasks. A portion of this material has been compiled and will be
presented in the next section. Some cases considered address relatively
complete piloting tasks.

Statement of An Operational Definition of FidelityI

We have arrived at a point at which it is possible to set forth a

general definition of simulator fidelity which takes advantage of our
growing knowledge of the pilot's perceptual mechanisms, strategy, and
technique induced by the simulator, the dynamics of the simulator compo-
nents (electro-mechanical and electronic), and the specific flight tasks
of interest.

Note that-the means of viewing the simulator and the pilot, which is
described above, allows for extensive and direct quantification. Our
objective regarding fidelity is to establish a working definition which '

takes full advantage of such quantification.

Consider also that training is the development and refinement of a

suitable control loop structure- the specific means by which a task is
carried out. Further, training involves teaching the student to use per-
ceptual mechanisms appropriate to the given task.

Therefore an appealing approach to simulator fidelity is to focus on

"how the pilot carries out a particular task given the perception (or

24 Clement, Warren F., Robert K. Heffley, Wayne F. Jewell, and Duane T.

McRuer, Technical Approaches for Measurement of Human Errors. Final
Report, Systemsý Technology, Inc., Technical Report No. 1156-2, July
1980.

27

. - - - - - <1 - - - - - - - --. _ -
 No'

•~I~i
inferred perception) of necessary cues. Hence we would construct a quan-
titative comparison between simulator and the actual flight situation of
the combined induced strategy and pilot perception. This frees us from
the notion that perfect fidelity is a one-to-one correspondence between
simulator systems and the actual aircraft 2 5 _ a practical impossibility
anyway. Rather, perfect fidelity is characterized by the simulator pilot
behaving in a manner appropriate to the aircraft situation. These ideas
do not, in essence, vary from the various concepts of simulator fidelity
mentioned earlier.

We suggest, then, that fidelity is the specific quality of a simulator

that permits the skilled pilot to perform a given task in the same way
that it is performed in the actual aircraft. Execution of said task is
simply the closure of all loops made necessary by both the task require-
ments and the dynamics of the vehicle and subject to the information which
is available. In order to close loops on the required states, cues cor-
responding to the states themselves must at least be defined, perceived,
* and recognized in terms of essential cardinal abstractions from the pi-
lot's perceptual fields, here limited to visual and vestibulae. That cues
must be defined, perceived, and recognized implies first the requirements
that:

* The task variabler, have been defined for the pilot.

Task variables include the specific purposes, assign-
ments, and commands comprising the mission strategy,
the likely guidance media, the vehicle to be used, and
the likely disturbances and counteractions to be ex-
pected throughout the mission profile. Task variables
comprise all the system inputs and those vehicular
elements external to the pilot which enter directly and
explicitly into the pilot's assignment.

25Ths notion follows from the identical elements theory of transfer of

thorndike.

28

I
Second, this implies the requirement that:

0 The feedback (and zeedforward) cues essential to the

task can be (a) employed by the pilot and (b) dis-
covered by the ana st. These cues are called
"essential feedbacks" . The feedback cues actually
selected by the pilot will correspond to the states
which are both necessary and sufficient to satisfy the
4
guidance and control needs and certain pilot-centered
requirements.

The guidance and control needs are situation specific. Satisfaction

of these needs always involves the organization of perception and adoption
of task-centered outer loops, with the addition of subsidiary inner loops
and other axis crossfeeds as needed to promote the adoption of the outer
loops in accord with the following pilot-centered requirements2. The
feedback loops preferred are those which 2 8 :

26 McRuer, Duane, Irving Ashkenas, and Dunstan Graham, Aircraft Dynamics

and Automatic Control, Princeton University Press, Princeton, New
Jersey, 1973.
27 The Successive Organization of Perception (SOP) theory of skill
development is treated in McRuer and Krendel, 1974, op cit.

Krendel, E. S. and D. T. McRuer, "A Servomechanisms Approach to

Skill Development," Journal of the Franklin Institute, 269, 1960,
pp. 24-42.

McRuer, Duane T.,, Warren F. Clement, and R. Wade Allen, A Theory of

Human Error, Systems Technology, Inc., Technical Report No. 1156-1,
May 1980.

Sheridan has attributed the cognitive organizing activities

represented by SOP to a functional construct called the
""etacontroller" within the cerebrospinal portion of the nervous
system in Sheridan, T. B., "The Human Operator in Control H
Instrumentation," in R. H. Macmillan, et al., (Eds.), Progress in
Control Engineering, Academic Press, New York, 1962, pp. 141-187.
28 McRuer and Krendel, 1974, op cit.

- -- - 29 1
 1. Can be closed with pure gain equalization by the pilot.

2. Can tolerate a time delay which is characteristic of

the appropriate modality.

3. Require the least scanning activity to perceive tie

feedback cue.

4. Permit great latitude in the pilot's adopted

characteristics.

Third, this implies the requirement that:

*Te cues corresponding to tha essential feedbacks

should be represented by coherent patterns in the
perceptual fields which the pilot has learned (or will
learn) to recognize in flight. Each intrinsic pattern,
in turn, must be sufficiently coherent in situ to 1j
exceed the pilot's threshold of recognition.

Fourth, this implies the requirement that:

* The cardinal features which comprise the patterns

should present a perceived signal-to-noise ratio to
which the pilot is (or •i;l be) accustomed in flight.

Given the perceptual abilities of the pilot, there are four additional
requirements regarding dynamic changes in cues corresponding to dynamic
changes in the essential feedbacks. The change in cues or states must: V

Such coherent patterns have been called cardinal cues, rbstractions,

or features. Examjles are discussed in Ibid.

Clement, Field-of-View Requirements for Approach and Landing of V/STOL

Aircraft, 1978, op cit.

Gilinsky, A. S., "Perceived Size and Distance in Visual Space,"

Psychological Review, 58, 1951, pp. 460-482.

Gilinsky, A. S., "The Effect of Attitude Upon the Perception of Size,"

American Journal of Psychology, 68, 1955, pp. 173-192,

Hennessy, 1979, op cit.; as well as the bibliography ippended hereto.

S~30

L
L V.,
 I Be large enough to exceed the perceptual thresholds
(e.g., vestibular thresholds or visual acuity)

•1 Be quick enough to permit the closed loop bandwidths

required (e.g., motion lags or visual update)

• Be sufficiently distortion free to permit correct

compensation by pilot (e.g., washout not too fast, ... )I'

0 Be sufficiently noise free as not to require workload

for processing, filtering, or reconstructing patterns
of change (e.g., motion vibration level, picture jitter
or flicker should be minimized)

Hence we have tied fidelity directly to the pilot's use of perceived

states. The perceived states, in turn, caix be characterized in terms of:

Threshold
Quickness }

Distortion

Signal-to-noise ratio

Each of these characteristics is, in turn, directly quantifiable in a 1.

variety of ways. For example, motion threshold is directly related to
thresholds of the human vestibular system. Such tthresholds, although
somewhat task dependent, are nevertheless well resoarched quancities.
Quickness is most likely tied to the control bandwidth required for a
given task (e.g., pitch attitude usually involves a 1 to 2 rad/sec band-

width). Distortion may be as simple as specifying flatness of frequency

response which implies that the amplitude and shape of response are
adequate. Finally, signal-to-noise ratio relates to ease of detection and 4
can be established on an empirical basis. These four characteristics
provide a basis for tying together the notions of engineering fidelity and
perceptual fidelity.

31
 ~~1

It is important to recognize that the above concept of fidelity is

based simply on the consideration of usable cues for a specific task. It
is founded on the notion that pilot behavior and perception can be char-
acterized in terms which are compatible with the simulator on one hand and
the actual aircraft on the other. A summary deflnition of simulator fi-
delity is given in Table 2.

Let us now proceed to categorize quantifiable parameters which cor-

respond to the above characteristics of fidelity.

A TAXONOMY OF MEASURABLE FIDELITY PARAMETERS

General Classification Scheme

In order to obtain the objectives of this study it is necessary to

express the working definition of fidelity in terms of measurable fidelity
parameters. - In effect, we must encompass the notions of perceptual fi-
delity and induced pilot behavior into objective fidelity parameters.

The following is a taxonomy of fidelity parameters that is suffi-

ciently general. to incorporate previously developed fidelity metrics along
with those that must be used to extend perceptual and behavioral aspects
of fidelity.

The major compouents of our taxonomy include:

* States

0 Transfer relationships P

* Domains in which the above are expressed

The term "state" refers to a specific variable or dimension which is quan-

tifiable in the usual sense. Common aircraft states, for example, could

"32

- J
4 A. A

TABLE 2

A SUMMARY DEFINITION OF SIMULATOR FIDELITY

SIMILATOR FIDELITY:

The degree to which characteristics of perceivable

statee induce eorrect psychomotor and cognitive control
strategy for a given task and environment.
?K
WHEREIN:

CcryweCt strategy is locally defined in the training

environment.

Applicable states are chosen on the basis of speci-

fied loop structure essential for the task.

Cha•acteristics of states are determined by their

role in inducing correct control technique (as defined
in the training environment); i.e., quantification of
loop structure adjustments (tightness, compensation).
1
Several domains can be used to express characteris-
tics of applicable states in terms of convenient
fidelity parameters. -

The ultimate proof of training simulator fidelity is

reflected in terms of transfer of training.

33
WI

-•-- ~33 •;

--------------------------
-------------------------
 include attitude, airspeed, altitude, heading, and track along with their

respective longitudinal and lateral -directional controls. "Transfer re-

lationships" between states and pilot responses describe, among other
things, how perceived states or "Cues" issue from actual states and how
control actions vary with perceived statess. Characteristics of transfer
relationships include thresholds, quickness of response, distortion of
response, and noise characteristics. Finally, the "domains" in which the
states and transfer relationships could be expressed include temporal,
frequency, or statistical domains. We shall develop these explanations
further, but let us continue with a few more general comments on fidelity
parameters.

One additional distinction we can make at this point is between "im- I

* plicit" and "explicit" fidelity parameters. An implicit parameter carries
a certain implied level of fidelity as a rather indirect or incidental
consequence. An example might be pitching motion travel limits which, by
themselves, do not guarantee an adequate level of motion to perform a
particular task; yet without a given range of travel, motion fidelity
could be a physical impossibility. The comprehensive simulator system
31
specifications are examples of implicit fidelity parameters. While they
are admittedly necessary to construct a simulator, such implicit param-
eters do not, by themselves, convey or guarantee a level of fidelity.
Explicit fidelity parameters, on the other hand, do address fidelity very
directly because they characterize piloting technique and pilot perception V
with respect to relevant states and for a particular task. The approach
32
to simulator fidelity taken by Hofmann and Riedel deals with explicit

30 Mathematically linear transfer relationships are called "transfer

functions" of the complex operator, s = a + jw. Quasi-linear
approximations of nonlinear transfer relationships are called
"describing functions" of the imaginary operator, jw, in the frequency -
domain.
31 Schalow, 1975, op cit.

"32Hofmann and Riedel, 1979, op cit.

34

C *•J."- .
fidelity parameters in that pilot perception is included in the analysis
of simulator motion.

A concise way of illustrating the above ideas is given in Fig. 5 which

shows a single loop comparison of pilot action in a simulator versus that
in an actual aircraft. In each instance the input is pilot command and
the output is percei%-ier pilot response. The major task objective is
keeping the two sets of pilot behavioral quantities matched. This char-
acterization can apply to the psychomotor actions of maintaining attitude,
airspeed, altitude, or to the more cognitive actions of tuning radios,
setting flaps, and raising landing gear. Each of these actions would have
its own loop quantification and could involve cross-coupling among various
loops (subtasks).

Each of the blocks shown in Fig. 5 can be expressed mathematically.

The most convenient means of expression is a frequency domain transfer
function (or describing function).

The overall pilot-aircraft or pilot-simulator dynamics can be studied

in either closed-loop or open-loop (i.e., the feedback loop artificially
cut) terms. This provides a limited view of system performance. And
because the pilot is a compensating element, it is likely that overall
performance will appear to be similar between the aircraft and simula-
tor. Not apparent are the fidelity aspects as reflected in the piloting
technique and pilot perception elements. The point is: task performance,
alone, does not indicate level of fidelity.

Implicit fidelity is represented mainly by the block labeled "simula-

tor response" and, to an extent, by the "aircraft model." Both of these
blocks are commonly included in the concepts of objective or engineering
fidelity and may therefore. be alternatively qualified as representing
extrinsic fidelity.

Explicit fidelity includes the combination of "piloting technique" and

"pilot perception." The level of explicit simulator fidelity is charac-
terized by how well these blocks compare between the simulator and
aircraft and may therefore be alternatively qualified as representing
fidelity which is intrinsic to the training objective.

35
 E-1

rq 43

S• .- • oto-

o-4 4V
.g ILI la)
O-ý C

00

0
t-I-
o
H4

E-IH co

to a,

4--'

o0

4,,

Tr-40
CIS 01)as )

00 :->- r- :0

E2 to o0

0 _ -- 36
a a)__________

PL4 ca .

+6
+
 States "

The term "state" refers to any of the variables which describe air-
craft (or simulator) operation. This would naturally include attitude,
heading, airspeed, altitude, angle of attack, engine torque, etc., along
with various flight and engine controls. "State" could also refer to. time
derivatives or integrals of each variable above.

A list of fidelity-related states for both motion and visual modali-

ties is summarized in Table 3.

The states which relate to fidelity are those which are involved in a
particular mission phase or task. To find these states we would consider
"a block diagram of the piloting technique loop structure implied by, say,
"a training manual task description. Examples will be considered shortly.

Transfer Functions Between States

Our term "transfer functions" can express the functional dependence of

one state upon another and can include the implicit fidelity relationships
of the simulator model and the simulator response (motion or visual). But
more interesting are the transfer functions which describe pilot percep-
tion and pilot .behavior (control technique). In fact it is the
S~quantification of these relationships which ultimately can lead to quan-

tification of fideliLy requirements.

Any particular transfer function can be considered in terms of at ]

least four characteristics. We shall continue to address the particular
four which were mentioned in-connection with our definition of fidelity.

Threshold. Beginning with the aspect of threshold we would propose

that a state is not usable if the pilot cannot detect a change in that
state. Cab motion which is less than the vestibular system thresholds is

-- "37
[ TABLE 3

SIMULATOR-FIDELITY-RELATED STATES

MODALITY CORRESPONDING STATES

VESTIBULAR PERCEPTION ROTATIONAL VELOCITIES

(of motion taken with three components

respect
resprenc to an inertial
toramer SPECIFIC FORCES (translational
reference frame) acceleration)

-three components
(sensed states are subject to threshold,
washout, and lag)

VISUAL PERCEPTION ROTATIONAL POSITION

(of position or motion roll and pitch attitudes plus heading

taken with respect to an
earth fixed reference frame
or with respect to another rates of change of above aircraft states
* aircraft.)
TRANSLATIONAL POSITION
horizontal and vertical transverse
positions plus range

0 TRANSLATIONAL RATES A
horizontal and vertical transverse rates
plus range rate

(The visually perceived states are also

subject to threshold, lag, and washout but
in different amounts from the vestibular
states)

38

#; *-
useless, and visual motion less than the visual system resolution or the
pilot's visual acuity is likewise useless. It may be necessary to either
enhance the system (e.g., by adequate motion travel or increased visual
system resolution) or enhance the pilot's perceptual power (e.g., by pro-
viding a g-seat or corrective lenses).

Analytical tools for dealing with thresholds are described in Graham

and McRuer 33 and are compatible with the frequency domain methods commonly
used with other pilot-vehicle elements.

Quickness. Quickness of response in either the motion or visual sys-

tem has a direct impact on how tightly the pilot can regulate any of the
states connected with a particular task. For example, if there is a large
delay in the change of pitch attitude as presented to the pilot, then it
is necessary for the pilot to make slower, more moderate pitch attitude
control adjustments. Otherwise there is the likelihood of overcontrolling
to the point of producing divergent pilot-induced oscillation in pitch
attitude.

There is, of course, a basic limit on quickness of required simulator

response which is set by the dynamics of the aircraft being simulated.
The quickness of response inherent in the aircraft also sets a fundamental
upper bound on the tightness of regulation which can be achieved by the
pilot. Depending slightly upon the handling qualities of the aircraft,
the following levels of quickness in terms of closed loop bandwidth are
common34:

Graham, Dunstan, and Duane McRuer, Analysis of Nonlinear Control

Systems, McGraw Hill, New York, 1961.

Ringland, R. F., R. L. Stapleford, and R. E. Magdaleno, Motion Effects

on an IFR Hover Task - Analytical Predictions and Experimental
Results, NASA CR-1933, November 1971.

39
 0 Pitch and Roll Attitudes 1 rad/sec

* Vertical Flight Path 0.2 rad/sec

T Lateral Flight Path and Airspeed 0.1o rad/sec t

II
additional motion or visual system delays, Hence, of the groups listed
above, attitude control would be expected to be more susceptible to a lack
of quickness in the simulator. Lateral flight path and airspeed would be
less critical.

The sources of lags and delays in simulator systems are numerous and,
to a degree, generally additive in their net effect. The effects of var-
ious sources of delay within real-time digital computing systems for
flight simulators have been examined analytically by Heffley, et al., 5
The objective of the examination was to explore useful simulation fidelity
metrics and procedures for obtaining them. Emphasis is placed therein on
digital computing systems involving two computers operating in series or
within feedback loops in which there may be several forms of delays along
with the complications arising from multirate or multiloop operation.
Particular examples of delay discussed by Heffley, et al., include the
following:

• Transport delay

* Data skewness (in time)

• Algorithmic delay

Heffley, Robert K., Wayne F. Jewell, Richard R. Whitbeck, and Ted M.

Schulman, The Analysis of Delays in Simulator Digital Computing
Systems. Volume One: Formulation of an Analysis Approach Using a
Central Example Simulator Model, NASA CR-152340, February 1980.
36 Ibid.

40

- -
-- -- --.--- -IIIII__I__I__ ..... . .
 * Frame slip (two or more digital processors operating at

nearly the same frame rates)

* Data exchange between two or more digital processors

a Multiloop or multirate architecture.

Visual flight simulator digital computing systems usually employ a

"host" processor to represent the mathematical model of the aircraft and a
"satellite" processor to generate the dynamic external visual field seen
by the pilot in response to aircraft motions. Such an arrangement cor-
responds exactly to that of two digital computers operating in series
within feedback loops wherein the human pilot also participates as a ser-
ies element. Thus all of the delay phenomena described by Heffley, et
a!.,37 as well as the multirate aspects handled by Whitbeck, et al.38 are
potential candidates for degrading the fidelity of the visual flight
simulation employing digital computer-generated models of both aircraft
and external visual field.

There is evidence that differences in the relative quickness of visual

and motion modalities play a role in inducing effective pilot time de-
lay an important feature in the psychomotGz behavior of the pilot.

"When other [than visual] modalities are available, such

as rotary motion cues from a moving base simulator or
actual aircraft, certain of the visual workload
requirements can be reduced. In the case of rotary
motions greater than semicircular canal threshold levels, I
the low-frequency lead generation requirements are
reduced 3 In• essence, the rotary motion cues permit the
pilot to close an inner loop akin to that of a rate

37 Ibid.

38 Whitbeck, Richard F. and Dennis G.J. Didalensky, Multi-rate Digital

Control Systems with Simulation Applications, AFWAL-TR-80-3101,
Vol. 1, Sept. 1980.
39 Stapleford, Robert L., Richard A. Peters, and Fred R. Alex,
Experiments and a Model for Pilot Dynamics with Visual and Motion
Inputs, NASA CR-1325, May 1969.

4I
 gyro. The net effect Is to reduce tne effective whole-
task time delay by about 0.15 sec (which also happens to
be the time delay incremen% required to develop a first-
order low-frequency lead)."

e41
Figue
borowedfro
6 Sinco ) shows the effective pilot time delay
versus motion drive bandwidth. Note that lack of motion response can
contribute as much as a 2:1 effect in determining effective pilot delay.
42
Baron et al. also addresses this aspect of simulator fidelity.

Deficiencies in simulator response can sometimes be offset by trading

off lags in the basic vehicle (aircraft model) with lags associated with
the simulator motion or visual systems. For example, a lag in thc motion
drives could be offset, to some degree, by quickening the aircraft model
response in either the control system or aerodynamic equations of mo-
tion. Similar fixes can be applied to the visual system. In general, all
features - even nonlinear ones, can be expressed in useful terms in the
frequency domain. Instead of rise time we would refer to bandwidth, or
instead of washout time constant we would refer to washout frequency.
Furthermore, pilot control strategy can be concisely stated in frequency.
domain terms. The manual control principles 43 are given in the frequency
domain.

Summative statistical analysis permits additional freedom to express

performance in terms of probability of occurrence or exceedence. But
there are dangers in relying too heavily upon statistical measures for

40 Clement, W. F., D. T. McRuer, and R. H. Klein, "Systematic Manual

Control Display Design," Guidance and Control Displays, AGARD CP-96,
February 1972.

41 Sinacori, J. B., The Determination of Some Requirements for a

Helicopter Flight Research"Simulation Facility, Systems Technology,
Inc., Technical Report No. 1097-1, September 1977.
42 Baron, Sheldon, Roy Lancraft, and Greg Zacharias, Pilot/Vehicle Model
Analysis of Visual and M)tion Cue Requirements in Flight Simulation,
NASA CR-3312, October 1980.
43 McRuer and Krendel, 1974, op cit.

42

-'- -S.-,'

4-f-,
 XFi.'Cd
base 0.7

0.4

I(D C.3

(sec)

0.2 0.19 sec , 4 o60

~ . rad/sec'

0.1-

0 10 30 6o

(rad/sec)

Figure 6. Effective Pilot Time Delay (Teff) Versus

Motion Drive Natural Frequency (M)44

44 Sinacori, 1977, op cit.

43

+. . . . m+ • +, . . 1
-. -4 C
ý-simulator fidelity. As was pointed out earlier, simulator performance can
compare well with actual aircraft performance although the pilot may be
using differing control techniques or may have differing perceptions. A
better application of statistical analysis is perhaps in characterizing
how individual pilot technique, or pilot perception parameters (time or
frequency domain), are distributed for various subjects, levels of skill,
or over the period of a given task. Statistical parameters have a great .4
potential if applied to piloting technique as well as the usual leasures
of system performance. 4<

Distortion. The distortion of states as perceived by the pilot can be

responsible for inducing an incorrect application of control technique.
Consequently, the distortion characteristic is a critical element in simu-

lator fidelity.
Perhaps the most prevalent example of distortion is the level of mo-F
tion washout necessary for working within a restricted motion travel. Due
to washouts, sustained motion (angular rates or specific forces) is sub-
ject to gross distortion. As a result it is also necessary to scale down
motion commands and to use gravity forces via pitch and roll angles to
simulate long term specific forces. 4 5
Sinacori addresses motion dis-
tortion in a systematic way which includes presentation of supporting
experimental results. We shall shortly discuss further experimental
results which serve to define the respective roles of motion and vis-al
modalities. ,

Visual distortion aspects appear to be far more subtle than for mo-
tion. It is fair to say that the simulator system characteristics which
contribute to visual state distortion are not well understood at this
time. It is possible, however, to infer the amount of distortion by ob- 1.
serving pilot behavior, and an example involving range perception will be

45 Sinacori, The Determination of Some Requirements for a Helicopter

Flight Research Simulation Facility, 1977, op cit.

44
presented in a later section. The work of Roscoe, et al, 4 6
concerning
visual accommodation effects on apparent size and hLight should be con-
sidered with regard to the issue of distortion.

Signal-to-Noise Ratio. State noise is distinct from state distortion

on the basis of randomness or lack of signal coherence. Non1e can be
generated by the pilot (this is often called remnant) or can result from
system disturbances and uncertainties. This contributes to pilot workload
by obscuring information content. This is more of a concern in the visual
modality.

A summary of simulator-fidelity-related characteristics is given in

Table 4.

Domains

The characteristics discussed previously can be expressed in a variety

of ways but, generally speaking, will fall into one of the following:

"" Time domain

"* Frequency domain

46i
46 Roscoe, Stanley N., "When Day is Done and Shadows Fall, We Miss the
Airport Most of All," Proceedings of the llth NTEC/Industry
Conference, NAVTREQtIPCEN 11-306, Naval Training Equipment Center,
Orlando, Florida, November 14-16, 1978.

Roscoe, Stanley N., Ground-Referenced Visual Oriertat-ion with Imaging

Displays: Final Report, AFCSR-79-4, November 1979.

Randle, Robert J., Stanley N. Roscoe, and John C. Petitt, Effects of

Magnification and Visual Accommodation on Aimpoint Estimation in
Simulated Landings with Real and Virtual Image Displays, NASA TP 1635,
October 1980.

45
 TA BLE 4

STATE TRANSFER RELATIONSHIP CHARACTERISTICS

MOTION-VISUAL ANALOGIES

CHARACTERISTICS VESTIBULAR VISUAL

Threshold Vestibular thresholds Static resolution,

dynamic resolution,
line width

Quickness (lag,
delay, initial
response
bandwidth)
Motion bandwidth effective
throughput, delay, effective
lag
Visual update,
effective lag Li
Distortion (shape Flatness of frequency response, Perceived range,
of response, washout time constant size, height factors
correlation among
multiple stimuli)
imp1 ies Morion travel, velocity, Field of view,
sufficient acceleration depth of field,
amplituide maximum brightness,
contrast

Signal-to-Noise Vibration, rumble Acuity, detail,

(cleanliness) contrast, jitter,
flicker

S~I

46

I....,.
Each of the above domains offers potential benefits in terms of conven-
ience of measurement, availability of data, and compatibility of
description between the human operator and the simulator system,

The time domain is usually the least abstract and it is possible to

easily handle such features as thresholds, response, cr decay times.
Furthermore pilot control technique or control strategy can be easily cast
in terms of time domain differential equations or finite difference equa-
tions47 Working strictly in the time domain becomes a disadvantage when.

we attempt to work with the overall pilot-vehicle system.

The frequency domain, on the other hand, enables us to use powerful

analytic tools while combining the various elements in a simulator or
actual aircraft system.

Fidelity Versus Training Effectiveness

The value of a training simulator is typically measured in terms of

its ability to reduce total cost-to-train. Its ability to reduce training
costs is, in turn, a function of (a) its training effectiveness, and
(b) the relative cost of simulator and aircraft training. Training effec-
tiveness or "transfer of training" has mcst often been measured in terms
of (a) transfer percent, i.e., the percent of time normally required in
the aircraft which can be eliminated as a result of simulator training;
and/or (b) the cmulative transfer effectiveness ratio (CTER) defined by
Roscoe 4 8 as the ratio of time saved in the aircraft to time 13ent in the

7 Heffley, Robert K., and Wayne F. Jewell, Development of a CTOL

Piloting Technique Measurement Scheme for a Real-Time Simulator
Environment, NASA CR-152294, July 1979.
48 Roscoe, S. N., "Incremental Transfer Effectiveness," Human Factors,
13, 1971, pp. 561-567.

4_I %r
simulator. Carter 4 9 illustrated the hypothetical relationships between
these parameters and stressed the importance of including both parameters
in pilot training system cost-effectiveness models, noting that transfer
percent is a measure of training effectiveness while CTER is a measure of
50
training efficiency. More recently, Bickley has shown that both of
these parameters and the interaction between them can be modeled by means

of a single differential equation, an approach which facilitates the cal-

culation of the most cost-effective mix of simulator and aircraft
training.

An implicit assumption in the present approach to determining fidelity

requirements is that the training effectiveness of a given simulator will
be highly correlated with the ability of the simulator to induce pilot
behavior which is similar to that exhibited in the aircraft; i.e., the
higher the similarity between simulator and aircraft behavior the greater
the potential for high transfer of training. Regardless of how fidelity
is defined, simulator fidelity is a necessary, but not a sufficient, con-
dition for training effectiveness. This is due to the significant impact
on transfer of such factors as the instructional features incorporated in
the device and the way in which the device is used. Thus the present
approach, after developing a sufficiently accurate model of pilot behavior
in the aircraft, will enable the analyst to predict whether the fide! ty
of a given simulator is sufficient to at least permit a high degree of
transfer. It is not possible, however, to predict the actual aircraft
hours which can be replaced or time required in the simulator without
accurately modeling the effects of such factors as the non-fidelity-
related design features and method of utilization. It should be noted,
however, that the ability to make accurate statements about whether a

Carter, V. E., "Training Effectiveness Analysis," Appendix XV, Future

Undergraduate Pilot Training System Study: Final Report, Aeronautical

Systems Division, Air Force Systems Command, WPAFB, Ohio, March 1971.
50 l;ickley, William R., "Optimizing Simulator-Aircraft Training Mixes,"
Proceedings of the 2nd Interserv1ce/Industry Training Equipment
Conference, November 1980.

48
given simulator will provide the essential cues required to permit a high
degree of transfer is, in itself, a worthwhile objective.

IMPLICATIONS FOR DETERMINING FIDELITY REQUIREMENTS

The approach to simulator fidelity described above establishes a point

of departure for the remaining sections. At the same time we will refine
and expand the ideas presented here.

The single most important notion in our discussions of simulator fi-

delity is that fidelity, or the lack thereof, is most directly tied to

0 Pilot perception of states involved in a particular

task, and

* Pilot behavior in terms of control actions in carrying

out the task.

A less direct measure of fidelity is the simulator response (motion or

visual), per se. Fidelity of simulator response is seen not with ref-
erence to mimicking the "real world" but with reference to inducing pilot
perception and control technique. Nevertheless, simulator response or
performance must ultimately be addressed when expressing simulator con-
struction standards. Hence parameters which indicate fidelity of
perception or behavior must, in turn, be converted to system performance
needs. This step will be considered in later sections.

Another major implication of the above approach to simulator fidelity

is that pilot behavior in terms of specific task loop structure needs to
be quantifiable and measurable. Both are realizable. The means of quan-
tification has been extensively developed. Measurement of pilot behavior
is less perfected, but there are a number of methods to measure pilot
actions, and these methods are continually being improved. One area where
measurements need to be more fully exploited, however, is in quantifying

49
and cataloging pilot behavior in full task situations rather than in per-
forming partial tasks. Furthermore, the process of making such
measurements - both in flight and in simulators -will itself expose and
illuminate concepts of simulator fidelity. Specific suggestions for mea-
eurements will be the objective of a later section.

Our next step, however, is to consider how to quantify the context in

which simulator fidelity must be considered - the training missions them-
selves. Hence Army training missions and their transfation into closed
pilot-vehicle loop structure'will be considered in the next section.

- -N - -
50

'i
 The value of understanding the training mission and piloting task was
emphasized in the earlier section. The goal now is to develop further
that understanding in specific terms which are useful for the ultimate
objective prescribing levels of simulator fidelity needed to train.

We shall begin this section with a definition of the training missions

to be considered and follow it with a description of how to formulate
analytical descriptions of flight tasks and associated piloting tech-
nique. The latter is based on manual control theory and a recognition of
how pilots nominally operate aircraft based on written training material
as well as discussions with pilots regarding piloting technique for speci-
fic tasks. A secondary concern is to link control strategies with the
vehicle dynamics, information available to the pilot, control character-
istics, and external disturbances. Beyond this we also want to consider
skill development in terms of compensatory, pursuit; and precognitive
behavior (i.e., the successive organization of perception or SOP).

ARMY TRAIFTNG OBJECTIVES TO BE CONSIDERED

The procedure for accomplishing the above begins with a review of the
various Army training missions and piloting tasks to be considered. This
determining and de-
will be followed by a discussion of guidelines for
scribing task loop structure. Finally we shall consider a number of
specific task analyses which pertain to Army flight training missions.
All of this will be preparatory to the subsequent section dealing with
analytic tools for describing motion and visual fidelity.

The scope of this study covers a wide spectrum of Army training mis-
sions including both rotary- and fixed-wing aircraft with the level of
training spanning undergraduate through continuation training. Of

51

S.---.-- ----- -: .L''... A

 particular interest are the critical flight phases which include nap-of-
the-earth navigation, weapons delivery, terrain following, and low-level
maneuvering. Operating environments to be considered include both visual
meteorological conditions (VMC) and instrument meteorological conditions
(IMC). The aspect of our approach which makes consideration of such a
wide spectrum of conditions feasible is that there is the ability to
divide these training missions into basic components and to find areas of
commonality for simulator fidelity requirements.

A basic list of Army flight training objectives (i.e., fundamental

flight tasks) was compiled based upon a review of the Aircrew Training
Manuals (ATMs) for the various Army aircraft types such as utility fixed
wing, utility helicopter, cargo helicopter, etc 5 1 . Table 5 shows this
list without specific regard to aircraft type. The table is a grouping of
individual piloting tasks considered to be basic training components ac-
cording to the ATMs. The list was also ciosschecked against the
helicopter field manual52 the basic undergraduate training syllabus for

LI
51 Anon., Aircrew Training Manual. Utility Helicopter, Department of the
Army, TC 1-135, October 1978.

* Anon., Aircrew Training Manual. Attack Helicopter, Department of the

Army, TC 1-136, October 1978.

Anon., Aircrew Training Manual. Observation Helicopter, Department of

the Army, TC 1-137, October 1978.

Anon., Aircrew Training Manual. Cargo Helicopter, Department of the

Army, TC 1-139, October 1980.

Anon., Aircrew Training Manual. Surveillance Aircraft, Department of

the Army, TC 1-144, October 1978.

Anon., Aircrew Training Manual. Utility Aircraft, Department of the

Army, TC 1-145, October 1980.
52 Anon., Rotary Wing Flight, U. S. Army Field Manual No. 1-51,
16 April 1979.

52

77ill
( TABLE 5

ARMY FLIGHT TRAINING OBJECTIVES3

BASIC FLIGHT LOW ALTITUDE OPERATIONS (CONCLUDED)

* Straight and Level NOE Flight

0 Climb/Descent Unmask/Remask
0 Level Turns 0 Quickstop
* Climb/Descending Turns Evasive Maneuvers
* Acceleration/Deceleration
Traffic Pattern WEAPON DELIVERY
Slow Flight
Stalls Hover Fire
Running Fire
HOVERING 0 ACM

Takeoff to Hover INSTRUMENT FLIGHT

* Hover
Hover Checks Takeoff
Hover Turns * Level Flight
Forward Hover * Turns
Land from Hover • Timed Turns
Hover Out of Ground Effect * Climbs/Descents
Confined Area e Climb/Descending Turns
Pinnacle/Ridgeline * Acceleration/Deceleration
Slope 0 Autorotation
VOR Navigation
TAKEOFF ADF Navigation
Holding
Normal Takeoff Unusual Attitude Recovery
Maximum Performance Navaid Approach
Short Field GCA Approach
Obstacle Clearance Tactical Instrument Takeoff
Terrain Flight Takeoff Tactical Instrument Approach

APrROACH/LANDING EMERGENCIES

0 Normal Approach/Landing Hover Autorotation

Steep Approach * Standard Autorotation
Shallow Approach Standard Autorotation With Turn
Go Around Low-Level Autorotation
Short Field Hydraulic Malfunction
Obstacle Clearance Anti-Torque Malfunction
Terrain Flight Approach Engine Failure at Altitude
VASI Approach Engine Failure at Hover
Flight at V (Single Engine)
LOW ALTITUDE OPERATIONS Single EngineM Landing

0 Terrain Flight Navigation Single Engine Go Around

Low Level Flight Engine Failure at Takeoff
Contour Flight Engine Failure During Approach

53 Bullets indicate those tasks which are at least partially quantified.

53

'.- * *-
 Army rotary-wing pilots 5 4 and, for more specialized tasks, the helicopter
gunnery manuals 5 5 and the NOE task analyses of Gainer and Sullivan5 6 .

The major groupings for Table 5 imply either a common piloting loop
structure (as in cases of basic flight or hovering tasks), a class of
maneuvers (such as takeoff and approach and landing), a basic environment
(low altitude operations or weapon delivery), or a basic limitation on
modality of pilot information (as in instrument flight tasks). This
classification system is not a rigid, well-defined form; but it is a com-
plete and convenient checklist of flight training objectives to be
considered in our analyses.

Another list which is convenient to consider at this stage is a con-

trol technique taxonomy for fixed- and rotary-wing aircraft as given in
Table 6. This shows a list of conLrol techniques which are more or less
independent of the piloting tasks which they support. For example control
techniques for pitch and roll control are common to nearly every piloting
task or training objective regardless of the aircraft type, the environ-
mental condition, or the level of training. Likewise heading, speed, and
flight path control are common to most tasks. Note that the control tech-
q niques are grouped first according to their place in the overall loop
structure, i.e., whether they are inner loo-s, intermediate loops, or
outer loops; second, the aircraft axis; and finally, whether they are
S~related to a basic compensatory level or pursuit level of operation (the

distinction will be discussed further in this section). One of our ob-

jectives will be to quantify control technique using existing data from
direct pilot technique measurements and, further, to generalize these data
to as many of the piloting tasks, shown in Table 5, as is possible.

4 54
Anon., Program of Instruction for 2C-15A/2C-10OB-B, Officer/WO Rotary
Wing Aviator Course, U. S. Army Aviation Center, Ft. Rucker, Alabama,
SSI/MOS: 15A/100B, February 1978.

Anon., Helicopter Gunnery, Department of the Army, TC 1-4,

September 1976.
56 Gainer and Sullivan, 1976, op cit.

54
 TABLE 6

CONTROL TECHNIQUE TAXONOMY FOR FIXED-WING AND ROTARY-WING AIRCRAFT

Comupensatory Level Techniques

I. ATTITUDES (Inner loops)

A. Pitch (perceived horizon, artificial horizon, pitch director,

"right-side-up," etc.)

1. Normal: Pitch + Stick

B. Roll (perceived horizon, artificial horizon, roll director,
"right-side-up," etc.)

1. Normal: Roll + Lateral Stick

2. "Rudder to Bank": Roll + Pedal (applicable at high AOA)

C. Yaw (runway line-up, compass, D G etc.-)

1. Normal: Yaw + Pedal (short final approach, hover)

II. FLIGHT PATH (Intermediate loops)

A. Vertical (altitude, sink rate, flight path angle, ILS glide

slope, VASI, etc.)

1. "Frontside": Flight Path + Pitch Command (applicable for

5V>V)
2. "Backside": Flight Path + Throttle (applicable for V < Vx)
3. "Forward Slip": Flight Path + Slip (to steepen approach
path)

B. Lateral (heading, turn rate, drift, VOR, localizer, etc.)

1. "Bank-to-turn": Turn Rate + Roll Command (usually

coordinated)
2. "Bank-to-Translate": Drift + Roll command (slip)
3. "Yaw to Turn or Translate": Drift + Yaw Command (skid)

C. Fore and Aft (applicable only to hovering vehicles)

1. Normal: Surge + Pitch Command (deceleration, hover)

2. "Direct x-Force": Surge + x-Force (e.g., where nozzle
deflection variable)

55

14[I
 TABLE 6 (Continued)

III. FLIGHT CONDITION (Outer loop or trim)

A. Speed (airspeed, groundspeed, angle of attack, stall horn, etc.)

1. "Frontside": Speed + Throttle (applicable for V > Vx)

2. "Backside": Speed + Pitch Command (applicable for V < Vx)

B. Engine State (rpm, manifold pressure, EPR, etc.)

1. Throttle: Thrust + Throttle

2. Governor: Thrust + RPM Command (can be a configuration

state also)

C. Symmetry (lateral acceleration, ball, sideslip, etc.)

1. Normal: lateral g + Pedal (for coordination or trim)

2. "Wing-Low": Lateral g - Roll Command (for x-wind, engine-
out, etc.)

IV. CONFIGURATION (Open loop or very loose loop)

A. High Drag (flaps, gear, speed brake, etc.)

1. "Flight Condition Adjustment": Speed + Flaps, Gear, Speed

Brake
2. "Flight Phase Appropriatp": Checklist + Flaps, Gear

B. Augmentation (SAS, SCAS, autopilot, flight director)

.. Workload Relief: Off + On

2. "Normal Operation": On (normal operating mode, safety of

flight, etc.)

C. Failure (engine, airframe, systems, etc.) H

"1. "Direct Action"
2. "Ignore" (not important, unaware, etc.)

56

IY
 TABLE 6 (Concluded)
Pursuit Level Techniques
(Feedforward loop structure involving coitrols,
states, errors, and disturbances)

I A. Pitch

V. "Avoid Extremes": Pitch + Stick (stall avoidance, VNE

avoidance, etc.)

B. Roll

1. "Avoid Extremes": Roll + Lateral Stick

C. Yaw

1. "Collective-to-Pedal": (counter yaw)

II. FLIGHT PATH

A. Vertical

1. "Flare" Height + Attitude (or collective)

2. "Flaps-to-Pitch" crossfeed (to counter height excursion)
3. "Flaps-to--Throttle" (to counter height excursion)

B. Lateral

1. "Coordinated Turn": Bank + Pedal

2. "Cross-Control": Bank + Pedal (to correct x-wind drift)

III. FLIGHT CONDITION

A. Speed
l."Pitch-to-Throttle" Crossfeed (frontside operation - to
counter speed change)

2. "Throttle-to-Pitch" Crossfeed (backside operation)

3. "Flaps-to-Pitch" Crossfeed
4. "Flaps-to-Throttle" Crossfeed (to counter speed excursion)
5. "Chop the Throttle" (short final approach)

B. Engine State

1. Collective-to-Throttle (to counter rpm change)

C. Symmetry

"1. "Throttle-to-Pedal" Crossfeed (counter propeller asymmetry)

57

i-1
 A review of the existing manual control literature reveals that a
quantification in terms of piloting tasks and piloting technique is lack-
ing. Where pilot techn'que has been directly measured in manual control
theory ten,'s, it has generally been associated with: (a) inner loop pi-
loting task'.. such as regulation of pitch and/or roll attitude, by means of
their respective controls; (b) simultaneous roll attitude and yaw rate
regulation by means of two controls57; (c) inner- and outer-loop regula-
tion of height by means of a single control 58
; or (d) inner and outer loop
*regulation of speed and position by means of a single control9, al based
* on instrument flight techniques. 60
A few measurements have been taken for
* an overall piloting task in which inner loops, euch as attitudes, are used
in direct support of outer loops, such as position, altitude, and head-
ing. Thus, even though the inner loop measurements which do exist are
* widely applicable to many of the training missions and piloting tasks
listed in Table 6, we lack the important quantification of the outer loop
behavior, especially in critical tasks such as nap-of-the-earth operation,
approach and landing, or air-to-air combat; and these are areas where

Stapleford, R. L., D. T. McRuer, and R. E. Magdaleno, Pilot Describing

Function Measurements in a Multiloop Task, NASA CR-542, August 1966.
58 Weir, D. H., and D. T. McRuer, Pilot Dynamics for Instrument Approach
Tasks: Full Panel Multiloop and Flight Director Operations,
NASA CR-2019, May 1972.

Stapleford, R. L., S. J. Craig, and J. A. Tennant, Measurement of

Pilot Describing Functions in Single-Controller M-ltioop Tasks,
NASA C7-1238, January 1969.

Heffley and Jewell, 1979, op cit.

Jewell, Wayne F., and Ted M. Schulman, A Pilot Control Strategy H

Identification Technique for Use in Multiloop Control Tasks, Systems
Technology, Inc., Technical Report No. 1153-2, August 1980. IA
59 Clement, Warren F., R. Wade Allen, and Dunstan Graham, Pilot
Experiments fc: a Theory of Integrated Display Format (Final Report),
JANAIR Report No. 711107, October 1971.

"60Ringland, Stapleford, and Magdaleno, 1971, op cit.

58
__ _ __ t
i
defhintion of visual and motion simulator fidelity requirements is prob-
ably of most interest.

JIt will be the objective of the remainder of the section to develop

the quantitative data which are available for piloting tasks and piloting
technique to show where these data are applicable to Army flight training
missions and to demonstrate ways of extending or extrapolating existing
data to cover the various areas of interest to the Army.

GUIDELINES FOR DESCRIBING TASK LOOP STRUCTURE

Let us now consider a simple basic set of guidelines and rules for
describing task loop structure for a given statement of the task and some
understanding of the vehicle dynamics and operating environment. We shall
gain from this a useful point of perspective for viewing the specific task
analyses to be presented in the subsequent subsection and beyond that, to
lead into the discussion of analytic tools for describing fidelity in the
next section.

Simply stated, the primary rationale for describing task loop struc-
ture is first to define the overall piloting objective and, second, to
discover how the piloting objective might be successfully executed working
within (a) the constraints of the information available, (b) the aircraft
dynamics, and (c) the workload limitations of the pilot. In general, to
satisfy all of these constraints, the e'cecution of a primary task objec-
tive requires the use of a multiloop structure for which there are nested
intermediate loops (series loop structure) accompanied by a similar set of
nested loops for each additional axis of control (parallel loop struc-
ture). Samples of this general structure are shown in Figs. 7 through
11. Figure 7 shows an acceleration/deceleration maneuver in a utility
helicopter. The basis for this control structure is the task description

599

"M- -
=
1 4j

> ?I
-4

45.

"04

0 r-
00

-I

60-
zii
-~ j ~---~-"--~
A
-X * . -U
- -;-
~~~~~rn
0I

o0 0o~ 000

0.

0 > v

CL~

V) 0
$.4

CL

.5.
V4

0
15-

0 0

d .4-d

100

Z1.4
go

-z) 00

"J
-2-2 -

Li L 0

.4~ ~~~ Ld
.
-a
v5Ii05
E

61 o 0 L
 oc

z$4

0~
0

CC I
a 00
- 40 -

4J.

to

I 0

I~I-

Il o

p zz.

zz-4

445 t- ZA z '. .4 ;r
a -.

~
.4 - 0 6 2 . >
 - I

-l
o

E!-1
000

ci 5-

41

( bD

100

SAS

IL'

63
 412

:r4•
I to

Li i °

* .1-i
0
o ii
** - * W 0
-• * cc
• o0
F4

cc

N.-I

~I-IE-I6:

__.......___________________________________

64•
• I
 64
provided for Army aviators in the aircrew training manual. Figure 8
similarly describes the pilot-vehicle loop structure which can be inferred
from a literal interpretation of another training manual task descrip-
6
tion. 5 In this case an approach to hover is considered. Note that t;he
structure is similar to the previous case except that range perception is
involved and manual control of rotor rpm is addressed. Figure 9 shows the
inferred loop structure for a straight climb on instruments 6 6 which is
equally applicable to both helicopters and fixed-wing aircraft. Addi-
tional examples of common loop structure are shown in Figs. 10 and 11 for
helicopter hovering and fixed-wing frontside operation, respectively.

In each of the above examples only the general loop structure is de-
scribed - there Is no quantification of the various pilot or vehicle
elements. Quantification is required, however, if these diagrams are to
be used for determination of simulator fidelity requirements. Pilot-
vehicle loop structure can be quantified by direct measurement of piloting
technique and aircraft response; however, estimates can be made by direct
literal interpretation of pilot training manual descriptions. 67 Some
examples are shown below.

tJ

64 Anon, AIM, Utility Hellcopter, 1978, op cit.

65 Anon., Basic Helicopter Handbook, 1965, op cit.

66 Anon., Jeppesen Commercial Instrument Manual, 1977, op cit.

67 Heffley, Robert K., and Ted M. Schulman, Derivation of Human Pilot

Control Laws Based on Literal Interpretation of Pilot Training

i
Literature, AIAA Paper No. 81-1822, 1981.

65
 To illustrate the concept of flight task quantification, consider a
simple example - heading regulation during cruising flight. The in-
structions for executing a small amplitude level turn are:

"As a guide for turns of 30 deg or less, the bank anfe

should approximate the number of degrees to be turned."U

This implies the following piloting technique relating bank angle command,
ý0 to heading error, le:

ýc =•e

Or, in block diagram form the piloting technique is:

(Commanded or (Heading
Desired Error) (Commanded
Heading) + *e Bank Angle)

4m (Measured Heading)

mI

Note that this piloting technique is not dependent upon vehicle dynamics
or flight condition; it is invariant.

Combining the above piloting technique with the vehicle dynamics in-
volved in turns (i.e., turn rate, ', equals bank angle, 4, times the
kinematic ratio of gravity acceleration to airspeed, g/U) we obtain the

68 Anon., Instrument Flying, Air Force Manual, AFM 51-37, 15 August 1979.

66
-71

simple pilot-vehicle feedback loop structure shown in Fig. 12. It should

be noted that the numerical piloting technique indicated here could be
inserted directly into the heading regulation blocks in several of the
task loop structures shown previously (e.g., in Figs. 7, 8, 9, and 11).
Also note that the basic form of the loop structure in Fig. 12 follows

that of Fig. 5 in which basic fidelity components are identified.

Other .echniques can also be applied to defining specific piloting

tasks and training objectives. Besides direct measurement of piloting
technique and interpretation of training literature, Carter 6 9 has found
that considerable insight can be gained by asking instructor pilots not
only how but how not to perform specific tasks or maneuvers (i.e., po-
tential errors). In fact the latter appears to be substantially better in
provoking a good task analysis than the former. While this procedure may
not aid directly in numerical definition, it can be useful in refining or
testing a basic task loop structure hypothesis.

Studying the errors in the performance of actual flight tasks can also
be a useful technique in identifying cues used by the pilot. more spe-
cifically, errors frequently occurring in the simulator which never or
rarely ever occur in the aircraft can lead to the identification of real
world cues to correct performance that is missing or distorted in the
simulator. For example, the tendency of students to always pull up late
in a Northrop simulation of the F-4J barrel roll attack was traced to a
probable target minification effect, apparently caused by the eye's ten-
dency to return to a resting accommodation in the absence of textural
detail, as desribed70
as described by Roscoe0. Conversely, the faf ture of a simulator

69 Carter and Semple, 1976, op cit. It should be noted that this

involved development of a task description for a complicated fixed-
wing air combat maneuver- the "barrel roll attack."
70 Spring, W. G., "Simulation Development and Preparation," Volume 4 of
Experiments to Evaluate Advanced Flight Simulation in Air Combat Pilot
Training, NOR 76-52, Northrop Corporation, March 1976.

Roscoe, S. N., 1978, op cit.

67
 -~ N'

IA
Z 4-') W

44J

CH !d
0 to
oW
mJ 0
4' -Uq

0H
00
41
:ii
cicis

om o ctj csj4
0'~ CJS ) 4t
4
0\ -4O

0 O'\*

0- PA 04V
~d0 0 3

t
00
bo 0~O 4' r

4-') ;9 0
4-)'

+2IL me~ 0o
Uq

r 0 0

68i
 W M WO"

to induce an error which normally occurs in the aircraft can help pinpoint
artifactual cues which do not exist in the real world or to identify mis-
sing or inadequately simulated distracting cues and workload factors which
make the task unrealistically easy in the simulator in order to correct
; performance.
Important clues to how piloting technique might be adjusted in terms

of control gains or control compensation can be obtained from manual con-

trol theory. This subject will be discussed at length in the next section
of the report.

It should be emphasized that all available sources should be utilized I"

in the construction of quantitative models of training objectives in terms
of flight tasks and piloting technique. No single source is likely to
provide a complete picture.

SPECIFIC TASK ANALYSES

Few examples exist of even partially quantified pilot-vehicle loop

structures for full-task analyses. A review of the literature shows an
emphasis on loop structures for partial task analysis, and this is limited
mainly to inner-loop aspects. The following Figures 13 through 22 depict
some of those few cases where relatively complete outer loop tasks have
been considered and, to some degree, quantified. Note that each task is
described first verbally and then in terms of a feedback control block
diagram. In each case an attempt is made to focus only on the primary
task loop. In the interest of simplicity, supporting or secondary loops
are omitted (e.g., for a speed change maneuver only the longitudinal
control axis is addressed - supporting pitch attitude control and control
of other axes are not shown explicitly.)

69
 r - --

Verbal Description

Adjust bank angle by applying appropriate pressure to the

lateral cyclic stick. V'znitor roll response using either the
actual horizon or the cockpit artificial horizon.

Feedback Control Description

Stick Deflection
Stick Force Ailercn Deflection
...... . Flight T k (Lateral Swashpl.te)

F P-iotnnr Technique P-V Inner Loop Vehicle Outer Loop

Desired Bk-nk e (I TTs)/ ActualBank

Adjustments made Control Vehicle roll

on the basis of spring-damper Stick-to-surface response properties
other roll attitude properties gearing including roll
response parameters including time constant,
and the next pilot's arm spiral divergence,
exterior flight task dutch roll, and
(e.g., heading adverse yaw
regulation)

Figure 13. Bank Angle Regulation Task and Piloting Technique I

70

___________________________
 77-7

Verbal Description7 1

When a deviation from the desired heading occurs, refer

to the attitude indicator and smoothly establish a definite
angle of bank which will produce a suitable rate of returnm
As a guide, the bank attitude change on the attitude
indicator should equal the heading deviation in degrees not
to exceed 30 deg. (For example, if the heading deviation is
10 deg, then 10 deg of bank would produce a suitable rate of
correction.) This guide is particularly helpful during
instrument approaches at relatively slow airspeeds. At
higher true airspeeds, a larger angle of bank may be required
to prevent a prolonged correction. A correction to a heading
deviation of 2 deg to 5 deg may be accomplished by
application of rudder.

Feedback Control Description

Comnanded Bank Actual Bank

ii PiloýTingjec]Lique P-V Inner Loop Vehicle Outer Loop

• jI, I._ T 1 1L
Des LUiL-------

Heading Compensation Turn Rate

K 2 I dsg/deg Effective Bank Angle Lag
(closed loop)
Tz I=see

Figure 14. Heading Regulation Task and Piloting Technique

S71 Anon.,
_1

Instrument Flying, 1979, op cit, p. 2-13.

"71

... " ... .

 Verbal Description 72

To enter a turn, you should refer to the 'attitude indicator while

applying smooth and coordinated control pressures to establish the
desired angle of bank. Bank control should then be maintained
throughout the turn by reference to the attitude indicator. Cross-
check the heading indicator or turn needle to determine if the angle of
bank is satisfactory. Trim may be helpful during prolonged turns to
assist in aircraft control.

To roll out of a turn on a desired heading, a lead point must be

used. The amount of lead required depends upon the amount of bank used
for the turn, the rate the aircraft is turning, and your rollout
rate. As a guide, a lead point of approximately 1/3 the angle of bank
may be used. With experience and practice a consistent rate of rollout
can be developed. A lead point can then be accurately estimated for
any combination of angle of bank and rate of turn. Make a note of the
rate of movement of the heading indicator during the turn. Estimate
the lead required by comparing this rate of movement with the angle of
bank and the rate of rollout.

Feedback Control Description

PrcgiieCot=rnded Bazwk Actual Bank Tu•rnrate
Precognitive

SFliht T!ask

Turn:i Technique

DeV Inner Loop Vehicle Outer Loop

,,
then `roout
•~ o.,
I
1
+
-o,,i
s I
| "I
1
5
I' I
;F~llul.'Headin~g

il
e dg

D e.re. -i' ____ , __ __-__

seadingI

Co=pensatory Effective Cineatic Relationsbip

Turn Rollout Bank Angle Lag for a Coordinated Turn
i T 1 sec -- a Function Only of
Decision to Start TAirspezd
Rollout from Turn

Figure 15. Heading Change Task and Piloting Technique

72 Ibid, p. 2-13.

72
 Verbal Description

"Adjust pitch attitude by applying appropriate pressure to the

longitudinal cyclic stick (or trim button). Monitor pitch response
using either the actual horizon or the cockpit artificial horizon.

= ]Feedback Control Description

Stick Deflection
Stick Force Elevator Deflection
_ _ _g S Flight Task
a(Longitudinal 12 Swasbplato)

F PilotimsTechnicue P-V Inner Loop Vehicl Outer Loop-

. . [.'IC
I * V
O+ Oe OU1T~s)e- s K

j tfue I 1 /-j e _ _1 _ ..

Desired s)kIto-surtual

Ajust11n6s Fade Conventional

Athe other pitch daAper-ineitia including short

attitude rasion properties period and phugoid
;•+ $parameters and Including
the next exterior pilot's aun
• I flight task (e.g.,
- ,•speed regulation)

Figure 16. Pitch Attitude Regulation Task and Piloting Technique

°" 73

+ • + +~~ -- - - + _•- - . +• +
 7
Verbal Description 3

When a deviation from the desired altitude occurs,

determine a rate of vertical correction and apply a power
change to correct back to the desired altitude. The
correction rust not be too large, resulting in the aircraft
"overshooting" the desired altitude, nor should it be so
small that the correction is unnecessarily prolonged. As a
guide, the power change should produce a rate of vertical
-alocity approximately twice the value of the altitude
Zeviation. For example, if the aircraft is 100 feet off the
desired altitude, a 200 feet per minute rate of correction
would be a suitable amount. By knowing the present rate of
climb or descent and the results to be expected from a power tJ
change, you can closely estimate how much to change the
power. The adjusted power must be held constant until the
rate of correction is observed on the vertical velocity
indicator. If it differs from that desired, then further
adjustment of the power is required.

Feedback Control Description

Co-andeed vertical velocity Actual vertical velocity

_________ Flight Tae: -

Piloting, Technicue P Inner_Los vehicle outer iToc-a

h 11 l I h
hcj I +{"• • I ift/secl
-f
I
1+T S
II
Desirei Actual
hei gt
e cigh t I -I I / I

/
Compensation gain
_ - / _

equals: Effective flight path

200 ft/min lag, T.A2 2 sec
100 (Note that the closed bandwidth
is .03 rad/sec.)

Figure 17. Altitude Regulation Task and Piloting Technique

Ibid, p. 2-11.

74
 Verbal Description'74

Before entering the climb or descent, decide on a power setting and estimate the amount of
pitch attitude change required to maintain the airspeed. Normally, the pitch and power changes
are made simultaneously.

The power change should be smooth, uninterrupted, and at a rate comensurate with the rate
of pitch change. In some aircraft, even though a constant throttle setting is maintained, the
power may change with altitude. 1herefore, it may be necessary to occasionally cross-check the
power indicator(s).

While the power is being changed, rcfer to the attitude indicator and smoothly accomplish
the estimated pitch change. Since smooth, slow power applications will also produce pitch
changes, only slight control p:essures are needed to establish the pitch change. Additionally,
very little trim change is raquired since the airspeed is constant. With a moderate amount of
will remain
practice, the pitch and power changes can be properly coordinated so the airspeed
within close limits as the climb or descent is entered.
Ten
Upon approaching the desired altitude, select a predetermined leveloff lead point.
percent of the vertical velocity in feet is a good estimate for the level-off lead point. At
the level-off lead point, smoothly adjust the power to an approximate setting required for level
flight and simultanecusly change the pitch attitude to maintain the desired altitude.

Feedbaek Control Description

Co~anded vertical velocity

Preco 'tiv ]
clio/descent Actual vertical velocity
initiation

Piloting Technique

c 5O~sgn(b)

•cc
airb/P-VInnerLoops Vebicle Outer Loop
descent*,-I- - 1
If~~h~5Ott desce..11
them leveloff 1 +Tlj - J•i S A.ld
t-, If :hJ<--
eOf

Altua ft i !~ - I . II I A
L-------------------------.. ..- _

Compensatoryr levcloff
based cn smoth tnsiton Effective vertical velocity lag
o cimb/descent T 2 sec
Decision to h
start leveloff
= climb/descent
fmAec

Figure 1B. Altitude Change Task and Piloting Technique

I7 Ibid, p. 2-11.

75

i,
Verbal Description
75

In a typical approach, the pitch attitude is 3 to 4 deq

for the DC-10. As speed is decreased to threshold speed, the

pitch attitude will increase about I deg. landing flare is
normally initiated at approximately 30 to 40 ft above the
runway surface. In the hypothetical case, if the airplane is
flared to a zero rate of descent with idle thrust and a speed
of just under ;hbreshold at touchdown, the pitch attitude will
be 8 to 9 deg. However, with a typical low rate of descent
at touchdown, the pitch attitude will normally be 7 to
8 deg. Landing with a I 50
attitude approximately degflap
over setting
that fordecreases
35 flap. the pitch
There is :

ample tail ground clearance for a normal 35 to 50 flap

approach anJ landing, even with the main landing gear struts
fully compressed and flat tires. Fuselage contact with the
runway will not occur until approximately 14 deg pitch
attitude. ?

Feedback Control Description7 6

Coczmaded Pitch Actul Pitch

Vertical Velocity

"r----" ----- -Oute

Fe£h - t[ _ ngt~s~ Heighte-

h~~ II . .SActual
Hegh '2 Height

Height Co=pensation Airfrae

Y.ý.l dee/ft Flight Path 14r,
Vcrticýl Velc=ity (heave da:pig)
Co-,pensation Effective _ 1kt
d Pitch Attitu4e T,0 = U -X1.8 see (ZC-io)
Kh
Te =..5 to 1.2 see

Figure 19. Landing Maneuver Task and Piloting Technique

Anon., United Airlines, DC-10 Flight Manual, Training and Reference,

1973-1974, United Airlines Flight Training Center, Stapleton
International Airport, Denver, Colorado, October 17, 1977, p. 4-16.

76 Beffley, Schulman, Randle, and Clement, 1981, op cit.

76

77-1
Verbal Description

Effect a speed change Maneuver by simultaneously chang-

ing pitch attitude and offsetting flight path upset by suit--
able use of collective control. Stabilize on the desired new
speed through appropriate use of pitch attitude.

Feedback Control Description

Desi Attitude Actual Attitude Acceleration

Flight Ta-k
P-V ._ er Loop Vehicle!Outer Loop
1
| _ otITeeI.ni ue

. ,__
U + j
1T
_]
1sec
s 0
_____
1
+/TI I S e pi ,
U

egk S....3 Efetv Pic1a

Pilot Co=pensat.-on
1z =.o - xg e
(efciv-uerlo Efecie=
Ef.0(1c ic Lage
K,.3-)u!e Speed Damping
Xu= .020I I-iU
(effective outer loop

crossover frequency
.07 rad/sec)

Crossfeed to
collective
"control

Figure 20. Normal Acceleration/Deceleration Task and Piloting Technique

77
 - 'a

Verbal Description

Fly a descending, decelerating visual approach

terminating in a 40 ft hover over a landing pad. Avoid
abrupt maneuvers. No approach guidance includes only
standard aircraft instruments normally used for visual
approaches.

7
Feedback Control Description 7

Co:anded spAed Accaal seed

-L___]___ __- F ... t Task-j _

"Pilotintechniave P-V Inne Vehicle Outr _Lao

Desired R**2~ I i + Te -T

R Pilot Perception
I f
1
Perceived1+ /
Range

Visual perception of range

A =500 ft Range conpe-nsation
Note that hover over the .25 /see
pad, the range perception Effective speed lag
transfer function is unity. foame_-d by airspeed ,
and attitude ýeedbacks
TU--- sec

"a 3

Figure 21. Decelerating Approach to Hover Task and Piloting Technique

[ 7 Heffley, "A Model for Manstal Decelerating Approaches to Hover," 1979,

op cit.

78

r-

-a- . -- - - - -
 7i

Verbal Descriptin

roHalt all forward motion with respect to the terrain as

-rapidly
•' rotor) orasexcessive
possible without ground contact
increase in height (of the tail

Feedback Control Description

Cor~anded Pitch Actua.-l Pitch Speed V.

* ~~~. ~Flight - Taz __ _ __ _ __ _

SPiting Te"hniqe • -/t Loop Vehicle\ou o

--
Desre Acts _I

(Ix;•.reieIntrlo

Raage cL anthe d Ra
I deg/rt SpCcZ Ccu:pensation I
Ku = deg/kt zffective
Pitch Attitude Lrg
T9 =.3 se
(Im-~ol~ms precise control of
Though not shown, the piloting attitude and is critical to
technique for controflng height success of Lr.ný=er,)
is as crucial as the deceleration
per SO.

Figure 22. NOE Quickstop Deceleration Task and Piloting Technique

79

- - - -- _________________-----------~---------~-'----I-
__. _____..... _______________I . . . .. ..... ..
 PILOT MODELING TECHNIQUES

Our next step is to focus on pilot modeling techniques in view of the

previous discussions of similator fidelity and- the training missions to be

considered. In -this section we shall concentrate on how the pilot's con-

trol strategy and perception can be modeled.

PILOT CONTROL MODELING i

Pilot control refers not only to the general piloting technique (as I
discussed already) but also ty the adjustment of compensation, selection
of available cue information, and adoption of higher levels of control" •-1
organization. Pilot control can involve a whole "bag of tricks" which is
aimed at accomplishing the given task with adequate performance and rea-
sonable workload. The -topics which are discussed in- the following pages
K
will touch, upon some of the more important ideas of pilot control and
bring us to a point at which perceptual effects can be addressed.

Generality of Pilot Modeling

A casual survey of pilot modeling techniques would seem to indicate

that it is necessary to restrict ourselves to linear, continuous, and V
single-loop system models. Yet when we carefully observe the actions of -•

human
linear,
psychomotor
sometimes
and cognitive

attention, sometimes discrete,

more or
behavior in
less continuous,
and always multiloop.
flight,
but
it is
subject
clearly non-i
to
The simple explana-
divided
.
V
tion of this dichotomy is that first-order effects can often -be
effectively and -conveniently addressed through use of simplified modeling
approaches. Only for special cases is it really- necessary to complicate

80
 UAI

model features and assumptions, with the examples of thresholds, limiters,

variable gains, and cognitive logic illustrated in the previous section.

In fact it is really a matter of necessity that pilot modeling be

approached using minimal complexity -only enough to do the job. And the
"Job" is to gain insight, detect key system parameters, discover sensitive
areas, and, in our specific case, to determine motion and visual system
characteristics needed to train pilots. It only impedes progress to in-
troduce unimportant pilot model features (or task, vehicle, or disturbance
model features) even though many of these features are well known. How-
ever let us regard simplification in a far more positive light.
The important and fortunate reality of most pilot modeling applica-

tions is that simple models frequently describe pilot behavior within the
limits of measurement. Further the same simple models follow the common-
sense dictates of good control design, minimization of effort, economy of
essential information transfer, and realization of acceptable (but not
necessarily optimal) performance to accomplish the task objectives.

Linearization of pilot behavior is an important simplifying assumption

for mathematical analysis, but it does not prevent us from introducing any
important nonlinear effects. Such effects are merely modeled by linear
model forms 7 8 . One common nonlinearity is the pilot's threshold of in-
difference to the task 79
objective9, an effect easily modeled using a
(reduced average) linear gain and a remnant.

Continuous system modeling can be applied under certain conditions

even though an inspection of pilot behavior might reveal clear discrete
steps in control movement (e.g., throttle) or sampling of visual informa-
tion (e.g., instrument scanning or head-up/head-down action). Discrete
behavior can frequently be described in continuous terms through either
simple modeling of cognitive switching or by introduction of effective .

78 Graham and McRuer, 1961, op cit.

The perceptual threshold is a lower bound on the threshold of

indifference.

81
system lag which is related to aveiage intervals of discrete activity
(e.g., the discrete behavior of a digital computer can often be accounted
for using a first-order lag having a time constant equal to the computer
frame time). It is also feasible to treat the human controller as a low-
rate sample data system for low frequency outer loops such as control of
airspeed with throttle. 8 0

From the standpoint of constructipg minimal parameter models it is

frequently desirable to subsume inner loops where possible and, certainly,
to neglect any axes which are not directly coupled to the control axis of
concern. However, single loop system modeling is the least necessary
simplifying assumption, because muiltiloop analysis procedures are well
established and easy to handle w~ith and without computational aids.

With these preliminary statements, let us now develop some notions

about modeling control behavior taking full advantage of insights afforded
by a simplified, continuous model description.

Continuous Controller Models

Analysis and measurement (both objective performance, and subjective

opinion and commentaryy of piloting technique demonstrate that the more
important and demanding aspects of the pilot's task are often more con-
tinuous in nature than discrete. That is, the pilot performs in a
reasonably continuous fashion in response to continuous changes in the
aircraft's situation with respect to a mission objective. Thus tight
regulation of helicopter position in a hover and maintenance of a pre-
scribed flight path on approach are examples of continuous tasks performed
over a period of time in response to continuously changing situations. If

-80 Heffley and Jewell, Development of a CTOL Piloting Technique

Measurement Scheme for-a Real-Time Simulator Environment, 1979, op
cit.
the continuous task is too demanding of pilot skill, he will be hard
pressed to perform other more discrete aspects of his job; e.g., communi-
cation with the tower, tuning radios, weapons selection, etc.

These observations motivate a description of the pilot's behavior

which can account for the continuous aspects of the pilot's overall
task. It is readily observed, for example, that with some aircraft he has
a diffizult time; with others he -an fly without significant subjective
effort - in both cases we are supposing that the task goals (e.g., regu-
lating to a desired flight path) are identical. The difference, in this
example, is related to the nature of the aircraft's responses to the pi-
lot's controls. What is required is a mathematical model which can
account for such influences on pilot opinion and behavior as well as being
predictive of pilot continuous task performance. If, in addition, the
model is consistent with what is known of the pilot's perceptual, cogni-
tive, and neuromuscular capabilities, considerable confidence is gained in
applying it to situations which have not been precisely replicated in
previous experiments.

In the continuous piloting task, the pilot is performing as part of a

closed loop control system consisting of himself and the aircraft he is

flying, and acted upon by environmental influences. His control activity
and the external environment cause aircraft motions which he perceives
either directly (i.e., from his field of view outside the cockpit and from
his vestibular, proprioceptive, and kinesthetic senses) or indirectly from
panel-mounted or head-up displays. He responds to these motions by appro-
priate deflections of the control manipulators in the cockpit so as to
achieve task goals -minimize path error, maneuver to a new heading, etc.

Classical and Modern Control Theory

Control system theory offers means by which pilot behavior in closed

loop systems can be modeled, measured, and analyzed. "Classical control
theory" models the pilot as a combination of sensing, organizing,

-83
 PF~

equalizing, monitoring, and actuating elem.ents (the choice of words comes

from control technology) which are therselves arranged or organized in
such a way as to be consistent with what is known of these human capabili-
ties and limitatýons as -well as with the measurable performance of the
aircraft and pilot in combination. "Modern control theory" accounts for
they possible complexity of this organization by describing the pilot in
terms of observers, estimators, control laws, and effectors - the terms
are likewise those of the control systems specialist. In either case the
behavior being described is the same and either model can be used succe's-
fully to the extent that it adequately accounts for pilot behavior as
noted in past experiments.

Clpssical control theory employs isomorphic structural models of pilot

behavior, whereas modern control theory employs algorithmic models of
pilot behavior 8 1 .

An isomorphic structure refers to having a form much like that of the

human operator or the operator's organizational structure. Isomorphic can
apply to neuromuscular, sensory, and equalization functions such as shown
in rather general terms in Fig. 2382. Taken on a larger scale, it can
also apply to a basic task-dependent loop structure as demonstrated ear-
lier in Figs. 8 and 9 for two common aircraft maneuvers.

The algorithmic psychomotor behavior structure supplies (e.g.,

Fig. 24) the various organizational units which are, in turn, identified
or measured. by any suitable identification method -- parametric or nonr-

parametric, time or frequency domain.

An algorithmic model structure is, in some ways, an abstraction of

psychomotor behavior and is based on the notions of optimal control and
optimal estimation, i.e., modern control theory. Typically this form cf
3

81 McRuer, Duane, "Human Dynamics in Man-Machine Systems," Niemi (Ed.), A

Link Between Science and Application of Automatic Control, Proceedings
of Triennial World Congress of IFAC, 4, Pergamon Press, Elmsford, New
York, 1979.
82 Ibid.

84
- . -2
0 0
0 E

lb c +

I
0

W
+

44,

CC

0 E-0

C ~Cc
0
o
IL H

Cs 0

0 0

* "0

CA % CO
 rw T

0+

020

E --

P4
U00

00

2-

- .1-4

86
 ! model expresses the human operator's adaptive control (motor) behavior as
an optimal controller which makes use of all system states and controls in
such a way as to minimize some form of cost function. Those state var-
iables which are assumed to be perceived are operated on by an optimal
estimation process (Kalman filter) in order to generate the needed states
for the control process.
Three areas of difficulty of the modern control theory algorithmic :
83
model approach regarding psychomotor behavior are given in McRuer
These are, briefly stated:

0 The human operator must contain essentially complete

knowledge of the man-machine characteristics, i.e., be

a complete internal model. Although this might be
plausible at the precognitive level of skill develop-
ment, it is incompatible with what we know about the
H
compensatory level. "-

"v Identification from experimental data is difficult.

* A cost function appropriate to a particular task must
be available.

Some other consid=rations regarding the use of optimal control models

(0CM) include:

"* The ovetall pilot-vehicle-disturbance system can be

modeled in compatible mathematical terms, but the OCM
formulation does not easily permit the simplification
advantage of frequency partitioning or decoupling of
tasks.

F "" Application of the OCM procedure is

tion of error, but there is
tied to minimiza-
no allowance for non-
optimal piloting -techniques which may be based on
rules-of-thumb or other training techniques.
ILY

0 The 0CM pilot tends to predict "pursuit" behavior, a

stage generally corresponding to a high level of skill
j
development, but other stages may not be easily
addressed.

83 Ibid.

87
 Nevertheless algorithmic models may be appropriate y and successfully
applied to represent the human operator's cognitive perceptual organizing
and activity-supervising processes. The process is the Successive Organi-
zation of Perception 84
(SOP) developed by Krendel and McRuer and McRuer,
86
et al. ,85 and summarized in McRuer and Krende1 . Most of the observed
"continuous" manual control behavior falls into relatively few arche-
typical categories from which logical criteria zan be employed to select
the most appropriate organizational category, given a task, an environ-
ment, and the circumstances attending the operator. One model for the
perceptual organization process would be an active off-line supervisory
monitor which identifies the conditions that currently exist, selects and
activates some most appropriate organizational structures for behavior,
monitors the result, and reselects a new or modified- organizational struc-
ture when necessary or when further information is identified as a result Y
of the initial operations. Appropriately this has been termed the meta-
87
coto 88
control system8. A simplified diagram of such a metacontroller is
given in Fig. 25. Other preliminary work on algorithmic models for these
cognitive perceptual organizing and activity-supervising processes are
89
given in McRtier, et a1 . Thus algorithmic models should be used where
they are best suited (for logical functions), while classical isomorphic
structural models are used where they are most efficient (for well-defined

184

84 Krendel and McRuer, 1960, op cit.

85 McRuer, D. T., et al., New Aporoaches to Human-Pilot/Vehicle Dynamic

Analysis, AFFDL-TR-67-150, February 1968.
86 McRuer and Krendel, 1974, op cit.

87 Metacontrol = the human's activity-supervising control, transcending

the various directly involved systems such as the perceptual, central,
and neuromuscular systems (from G:eek "meta" meaning "involved with
changes").
88 Sheridan, 1962, op cit.

89 McRuer, et al., 1968, op cit.

88 T:
 i•-

SITUATION IDENH'FICAT]OX '

SELECTION OF APPROPRIATEPUT-
PATHWAY(S)

'II

SELECTION OF
RESPON{SE
UNIT FROM
REPERTOIRE

'I1
[EXECUT.ION

S~OF:RESPONSE
1

•. HUMAN OPERATOR OUTPUT

-• OFF-LINE SUPERVISOR;

MONITOR RESULTS,
, I

*" l RE-IDENTIFICATION OR MODE SELECTION

FEEDBACKS TO PRIOR BLOCKS

Simplified Metacontrol1¢r for Successive

Organization of Perception
Figure 25•. Flow Diagram for SOP Operations

89

1
 continuous tracking or stimulus-response situations which are subject to
divided attention).

In what follows the classical model is chosen to represent the psycho-

motor activities of the pilot. The initial emphasis in this model is on
the pilot's response to visually-derived information -this is where the
data base is largest. Later the model will be amended to account for the
pilot's ability to sense and use motion information in flying tasks.

Pilot Compensation and Adjustment

The mrthernatical models which quantify pilot behavior in continuous

control tasks take into account two kinds of system requirements:

0 Guidance and control requirements which are related to-

system stability and the capability of following a
desired path or executing a desired maneuver.

* Pilot-centered requirements which express the abilities

and limitations of the human pilot.

The firsc ae. of raquirements comprises those whicn depend upon the
task being performed and are independent of the fact that the pilot is
human. Aircraft angular and linear motions are sensed and used in closed H
feedback loops to develop, through appropriate weighting and equalization,
the throttle, control stick, pedal, and collective (if the aircraft is a
helicopter) deflections that guide and control the aircraft. The choice
of which aircraft motion quantities to use and what their relative
weighting and equalization must be- to achieve system stability and path
following capability are all guidance and control requirements which would
, •'exist even if the pilot were to be replaced by a complex and elaborate
machine.
The second set of requirements comprises those--which arise because the

pilot is indeed a human being. In spite of his capabilities. he has

90
 S1
certain limitations- he can perform only a limited number of tasks at
the same time. The kind and degree of equalization applied to the
perceived vehicle motions is likewise limited. Further his control
activity is not completely correlated with the aircraft motions - it
contains uncorrelated activity- noise to the control engineer, remnant
to the human performance researcher. In short the pilot-centered
requirements are largely those which are imposed by human limitations9 0 .

Among these limited capabilities is the ability to generate lead

equalization based on visually perceived aircraft motions. This capabil-
ity forms an essential part of the human perceptual basis for vehicular
guidance, especially in the vicinity of the ground plane. There the rela-
tive velocity between helicopter and terrain may be so low as to inhibit
the perception of "streamers" which otherwise play a more significant role
in visual judgment of aircraft motion 91 . If the motions were presented on
a panel instrument, a human factors engineer would think of adding "quick-
ening" to the display-equivalent to one form of lead equalization-

90:
90 Exceptions are those pilot-centered requirements which are imposed by
human capabilities such as adaptation and learning, both of which are
at present hardly imitated by machines in their infancy. Learning
expresses a human organizational capability in successive encounters
with the same environment, whereas adaptation expresses a human
organizational capability in an encounter with a new environment.
91 Gordon, D. A., "Perceptual Basis of Vehicular Guidance," Public Roads,
34, 1966, pp. 53-68.

Calvert, E. S., "The-Theory of Visual Judgments in Motion and Its

Application to the Design of Landing Aids for Aircraft," Transations
of the Illuminating Engineering Society, 22, 1957.

"Havron, M. Dean, Information Available from Natural Cues During Final

Approach and Landing, Human Sciences Research, Inc.,
Report HSR-RR-62/3-MK-X, March 1962.
Naish, J. M., "Control Informat:ion in Visual Flight," Proceedings of
the Seventh Annual Conference on Manual Control, NASA SP-281,
June 1971 (also McDonnell Douglas Paper No. 5921).
Grunwald, A. J., and S. J. Merhav, "Vehicle Control by Visual Field

Cues," IEEE Transactions, SMO-6, 1976, pp. 835-845.

91
thereby relieving the pilot of this burden by matching, in a limited
sense, the requirements of the task to the pilot's capabilities. At the
same time this change In the display might obscure from the pilot that
which he calls status or situation information. In this example the state
of the motion variable is obscured to an extent depending upon the quick-
ening added to the signal.

Another means of providing lead equalization is to allow the pilot to

sense the aiircraft motions via his vestibular senses - the semicircular
canals and utricles. Observation of pilots in simulators suggests that
lead equalization generated by this means comes at less cost (subjective
workload) than that derived purely on the basis of visual information- a
given level of task performance in a moving base simulator comes easier
than if the simulator were on a fixed base.

All of this is well known to the pilot training community. What is

important for present purposes is the ability to -quantify this behavior in
a mathematical model. In fact there is a heirarchy of such models which
vary in their elaboration of detail, in their application to specific
questions, and so on. Here we shall present a basic model which illus-

trates most of the concepts involved. For the moment we shall ignore the
vestibular aspects and consider visually-perceived information only.

The human pilot, when treated as a set of elements in a larger feed-

back control system cokprised of pilot and aircraft, can be modeled as
follows:

* A set of describing functions expressed in terum of a

number of parameters.

* A set of adjustment rules for the parameters which

depend upon the task and the controlled element
(aircraft).

* An additive noise, or remnant, in the pilot's output

which accounts for those portions of his response not
correlated with his input; i.e., not accounted for by
the describing functions operating on the input
"signals."

92
 I 4
The term describing function comes from automatic control theory. De-
scribing functions are used to express the linear properties of nonlinear J
13
elements. The term's use in the context of hunan response description
serves to emphasize that a complex, nonlinear element is being approxi-
mated linearly for purposes of quantification in the particular situation.

The describing function is a frequency dependent function which, in

its most complete form, contains gains, perceptual and indifference thres-
holds, time delays, equalizers, and neuromuscular dynamics. Tho
perceptual and indifference thtesholds are of paramount importance in the i
present context; i.e., determining minimal levels of visual and motion
stimuli to train flight skills in a si-mulator. Usually, however, the
thresholds are higher-order effects that can either be ignored when the
inputs are large or accounted for approximately by using decreased average
pilot gain. The neuromuscular system dynamics are based on very high
frequency data, and can be approximated at the mid-range of frequencies of
interest in flying as a first-order lag or, even more simply, as an in--
crement in the time delay.

Croscover Model

The adjustment riles for pilot describing function parameters have

been derived from experimental data for a wide variety of single-loop and
mualtiloop control tasks. These data show that a relatively simple ration-
ale exists for the equalization adopted by the pilot. It is called the
"crossover model" 9 2 .

According to tJis model, given a controlled element having describing

function Yc, the pilot adjusts his describing function, Yp,, in each loop
- such that the open-loop combined describing function, YpYYC approximates a
frequency-dependent function of the form:

92 McPaer and Krendel, 1974, op cit-.

~ &93
 mw~
CL

YpYc
yy s c- z
(1)
S|

in the vicinity of the gain crossover frequency, wc' for that loop 9 3 This
form should be more correctly written to include a pure time delay, Te)
which represents the pilot's effective neuromuscular lag as well as any
high frequency vehicle response lags.

To accomplish this end result, the pilot must tailor his describing
function, Yp, to the specifics of the control situation. Y is expres-
I: p
sible in terms of the following parameters:

"S Time delay, T, which, accounts for the latencies due to

perception, interpretation, and neuromuscular
actuation.

"* Pilot gain, K , which determines the amount of control

correction for a perceived level of state error.

"" Equalization, which tailors the form of the pilot's

response to suit a given vehicle's handling dynamics.

The crossover model (Eq. 1) states that if the describing function,

Y has more lag than the
CI describing function, K/s, in the vicinity of the
desired gain crossover- frequency, wc, then some pilot lead equalization
(anticipation), will be required;, while if Yc is more like K than K/s,
some pilot lag equalization will be used to achieve good pilot/vehicle
system response and performance.' The parameters, wc and Te, in the cross-

93 The describing function -JIs or sometimes "K/s"

is a mathewatical form
representing in integration with respect to time. A system having
such dynamic properties is called a "rate system;" i.e., a st&p input
results in a constant rate of change output, at least over a
reasonable spectral range. One-example of a K/s system would be the
airspeed response due to a pitch attitude change in a helicopter a
unit nose-down pitch change results in a nearly-constant forward
acceleration over a large speed change,

94
r

over model depend on the frequency content (bandwidth) of the disturbance

or desired path (maneuver) input and on the pilot equalization. Data show
that the crossover frequency is greatest and the pilot time delay is least
when his equalization is a low-frequency lag. At the other extreme, Wc iJ

the least and Te the greatest when the pilot must generate low-frequency
lead. In fact the major "cost" of equalization is this increase in time
delay. Both Te and wc will vary with pilot skill and level of attention
to the control task. Experimentally-based values for the crossover model
parameters for several dynamic forms and input bandwidths are given in
94
McRuer and Krendel as well as elsewhere.
The complete pilot description includes the remnant. The major

sources of this "noise" in the pilot's output appear to be caused by non-

stationarity in his behavior, divided attention, and low frequency lead
generation within the visual modality. When his output is expressed as a
power spectrum, the remnant can be considered as pilot-induced broadband
random noise added to the input-correlated portion of the signal. For
aircraft control situations involving reasonable flying qualities, the
remnant will be small relative to that part of his response involved in
making a maneuver or in regulating against an external disturbance.

Levels of Skill Development

A
The simple crossover model given above is diagrammed as shown in
Fig. 26. It describes pilot behavior in single-loop compensatory control
"tasks and in the outermost loop of multiloop compensatory -tasks. In this
system structure, the pilot is controlling a single task-related response
variable with a single control by operating only on the perceived error.
•, A typical example of such behavior would be regulation of vehicle attitude

U in a gusty environment as represented by a fixed base simulator (no

Ibid.

95 -
 TIM

ControlledIj
Pilot Element

I-i. *1

rv
Figrure 26. Compensatory Pilot Model

It

tii 96
 motion). The pilot cannot preview what is coming and can only respond to
the developing error as it becomes manifest on his displays or in the
simulated visual scene.
The data base which quantifies the single loop compensatory pilot
model consists primarily of describing function measurements of well-

trained pilot, subjects. This qualification significantly reduces the

variability which would otherwise be quite prevalent in the data base.
Pilots exhibit quite remarkable uniformity of compensatory behavior when
the task goals are well defined and the pilot is well trained. There are,
of course, individual differences in ability. Some individuals can
achieve somewhat better performance than the average through reduced time
delay and increased lead generating capability.

The compensatory model is fundamental from a variety of standpoints.

Complex multiple-loop tasks involving several controls and responses can
be represented by repeated applications of the crossover model. More
important for purposes of the present discussion is the fact that the
model represents the earliest stage of psychomotor skill development-in
the beginning, the pilot regulates against errors and is unable to take
advantage of other indicators of system behavior available to him in a
rich perceptual environment. Finally it represents the form of behavior
to which the pilot reverts when confronted with an unfamiliar or stressful
situation wherein more skilled behavior and/or perceptual patterns break
down for one reason or another.
Higher levels of psychcmotor skill are possible when the pilot can
take advantage of additional information available to him. These can be
represented by more complex pilot models. There is a heirarchy of such i
F models used to represent various levels of psychomotor skill develop-
ment. These have been organized according to the degree of information
exploitation and formalized in the theory of SOP. IV this organization,
the compensatory model occupies the lo-est position, followed by the pur-
suit, and finally the precognitive models representing the highest level
of skill. !.

97
 Figure 27 is a simplified diagram of the pursuit model, so-called be-
cause it typifies the classic situation where target motion (the input
command) is perceived as separate and distinct from the aiming error. If
ouite familiar with the response properties of the controlled element (his
own aircraft), he can operate only on this input and generate control
activity which will result in responses having very little error. In
effect his describing function (operating only on the command) is given by
1/Y. such that YPiYcSY~1. At the same time he will respond to the
residual errors with the appropriate compensatory behavior, represented in
Fig. 27 by Y

If, in addition, the pilot is aware of the controlled element's re-

sponses, as distinct from the error or the command, he may find it useful
to operate on this information as well to improve system performance. In
Fig. 27 this is represented by Ym The most typical example is the
pm
pilot's use of vehicle motion, sensed through vestibular, proprioceptive,
and kinesthetic senses, to improve upon system performance. Because the
information is perceived through a sensory modality other than visual,
this model,- a special case of the pursuit model, is referred to as a mul-
timodality pilot model.

Unfortunately, quantification of the multimodality pursuit model, the

data base is limited to two separate visual and motion channels, Y and
YPm, Y
respectively. ~95e et al.,
Stapleford, suggest a model in which YPM
takes on the known properties of the vestibular sensing apparatus of the
human in those cases where the controlled element requires lead equal-
ization. The pilot need not derive his lead from the visual stimuli alone
and (need- not) incur the resultant "costs" associated with visual lead-
increased time delay, higher remnant levels, and so on.

For purely compensatory tasks, i.e., wherein the simpler structure of

available. These
Fig. 26 is assumed, moving base data are more readily
data suggest that with those controlled elements requiring lead

~ 91Stapleford, Peters, and Alex, 1969, op cit.

98

__ __ __ _
 Pilot

{ Controlled,
S~Element
1"1
Command + 1 Error CotorRsos
II

Figure 27. Pursuit Pilot Model

99

- -S-99-
equalization the effect of the motion is to reduce the pilot time delay
and increase the available crossover frequency. The continuous task
performance is thereby improved (increasing crossover frequency, W., means

reduced error in following pilot commands or regulating against distur-

bances). Pilot opinion tends to be more favorable, indicating less
subjective workload with motion perception. Yn sun these data are con-
sistent with the more limited data pertaining to the two-channel
multimodality model.

Pursuit level of operation can sometimes be detected by certain de- =

scribing function signatures. Hess96 discusses this by exploring
analytically the basic pursuit model (Fig. 27) and relating it to various
laboratory measurements which reveal pursuit versus compensatory be-
havior. The nature of pursuit effects is illustrated in Fig. 28 in terms
of the combined- pilot-vehicle open-loop describing function amplitude and
phase characteristics. Note that the symptoms of-pursuit behavior are the
low frequency amplitude "droop" and reduced phase lag. At crossover the
open-loop pilot-vehicle describing functibn remains close to the form

K/s. Only at low frequencies do the pursuit tendencies appear. This

suggests how to. detect and measure higher stages of SOP and gives some
idea of where it is of some consequence spectrally.

At this point let us make two observations which relate to the aboveI
pursuit-level symptoms. First, optimal control pilot models, when used to
97
predict pilot behavior such as described by Curry, et al., , tend to
yield a pursuit-level describing function signature. Second, the role of
cockpit motion discussed by Hess98, whether predicted by optimal or "[
classical methods, tends to be related to low frequency pursuit behavior
and not to basic compensatory actions at or near the crossover frequency.

96 Hess, Ronald A., "Pursuit Tracking and Higher Levels of Skill

Development in the Human Pilot," Sixteenth Annual Conference on Manual
Control, MIT, Cambridge, Massachusetts, May 5-7, 1980.

97 Curry, R. E., W. C. Hoffman, and L. R. Young, Pilot Modeling for

Manned Simulation. Volume I, AFFDL-TR-76-124, December 1976.

98 Hess, 1980, op cit.

100
 -7j

4 0r - COMPENSATORY
--- PURSUIT

20

>- 0

-201I Ii i

1'; -.100

-200

-300

-400j.

-500

.1 1 10
w..,radls

Figure 28. Comparison of Pursuit and Compensatory

99
ModelB
Describing Functions. (Taken From Hess )

Ibid.

101 -
 Finally, the highest level of psychomotor skill is described by the
precognitive model. This model represents those skilled control activi-
ties which resemble preprogrammed or self-generated behavior. The pilot
acts as though the only information needed, presuming full familiarity
with the input command and the controlled element, is the triggering
stimulus which tells him when to begin and his sensations of his control
movements.

From the standpoint of training, it is the "full familiarity" with the

task which must be acquired -- in part through simulated aircraft mo-

tions. Even though performance in a particular task is known to be

precognitive in nature (for well-trained pilots) and thus a task where
precise replication of stimuli in all sensory modalities is not necessary
for task performance, it is nonetheless required that the stimuli be
available to facilitate skill acquisition through the compensatory and
pursuit phases of the SOP sequence. In this respect it can be argued that
the training simulator has fidelity requirements in excess of those needed
in many research simulators- at least if proficiency levels approaching
those of trained pilots are to be reached. In the research simulator the
pilots are more often than not well up on the training curves for basic
and secondary flying skills.

As a consequence motion fidelity requirements analysis using mathe-

matical models of the human pilot will probably concentrate on pursuit
models. These models provide the means by which the several training task
scenarios can be analyzed for the potential involvement of motion in im-
proving task performance.

Perceptual Effects on Control

Our current approach, illustrated in Figs. 7 and 29, is to separate

the operator's perceived states into integral, proportional, rate, and
sometimes acceleration signals, each derived and weighted by central ner-
vous system processes. These processes are, however, imperfectly

102
 a. Physical Model for a Typical Perceptual Channel Showing Separation
of Display, Perceptual, and Equalizing Pareme-ters

DIS PLA~Y CENTRAL PROCESS-

V15IOcci-cnc,1 dhannet s.hownl
E~je Mcollj sitrndur -for Yac lIcifeoldci.1Fri
(From Sclar.4r.
11.
and~ Vibration)

c~~-PERCEPTION
DISLA EQUALIZATION

Yd()Cotinpulrg DcIk~' -

PAR~AM.ETER
RSeqc
(Parcmielcr~s c/cpcncl ota J;5?6y charneferstac,. actn parafovcal oingle)Clcac.s
ractims Delcciion Process Delays I-C Control C-aeiuas K (4)
Lo~qs S4ai;c, 0 Dyramwijk Art.41 ThrelcsI-r-, A Cornuirfin5 Delays -re~f)
Resolui~orn (QoaMIZ04ion Blur 4 Saiurofioni Limits, B Processin.g Noisr and
Lu~aieObservcttiok1 ?4osc 'rime V~ir.yinq Poroameicrs

b. Revised Central Process Model, Including Perceptual Processes,

Showing Each State Chaannel Represented by a Describing Function
Plus Related Noise
To14ik Oirder.Lead 6en

- ~~~ ~ ~ RT )' ~ k Denotes adjustieneft

ThA. &.2.sec. -~ ptin o~ro

nr(f) JL- Mo~ooTh~csholcls

Blur Llinijis

S Rollo

Motion ~DISPLACEMENT -~Cont~ud~rol uCuo

Dsl~r.er.Aýi htMc~

Foveal LUný%s

Nlonc.et 1RTl:6IAL(Le. 9j

Yh (A,;

Percraved
S'I.ates
I Pzcpual
Remnant
Penceptuol
0-:r~b~ncS FUICtiDoa
'klivsResidusal
Pe~rbVnC RAGOai Central ahcA NevromuLsculatr
I(Threskliz. LrWaili) (Lca4*,jLgs.,eC.) Tvrar:.3.?WiS.SIci

Figure 29. Refinement of Central Process Models for Perception

Of Displayed Signals (Won-Scanned Cases)

103
 understood at present. But the lack of determinism in portions of the
human operator models under certain conditions is reflected by correspond-
ing variations in the stochastic "remnant" (e.g., observation and motor
noise). The "remnant" forms an-essential part of the contemporary, quasi-
linear pilot models. It is attributable to time-varying time delays,

scanning and sampling processes, neuromuscular ("motor") noise, tracking

L style, and idiosyncratic effects. Among the sources neuronuscular noise

isthe least important one when investigating multiaxis visual display-

related effects. We expect that the adverse perceptual effects of

inadequate luminance, resolution, contrast, and symbol motion (threshold
and blur limits in Fig. 29a) will be reflected in increased observation
noise, as well as in attenuation and time delay in the operator's
describing function.

By careful experimentation during the last several years we, and

others, have found that such closed-loop perceptual effects can be repre-
sented by a nonlinear random-input "perceptual describing function" plus a
"processing remnant spectral gradient," both of which depend on the dis-
play and pe~rceptual characteristics. On the other hand the remnant
spectrum has been found to be nearly invariant when it is normalized by
the perceived state displacement and rate - especially the latter. Hence
the processing remnant is represented as a multiplicative noise process
acting on each perceived state 100

When passed around a tracking control loop, the high frequency noise
components in the control signal are usually attenuated by the controlled
element, and any quasi-periodic components are dispersed, so that the
resulting perceptual remnant appears broadband and has stochastically
stationary properties. This fact permiits a great economy of computa-
iton. Because most of the display and perceptual effects tend to have
f,'equency invariant describing functions, and because we can determine
only overall effects from any one measurement, it is customary to lump
these perceptual remnant effects with the operator's adjustable state-

~4

i00 McRuer and Krendel, 19i'4, op cit.

104

__________ _____________________
weightings (e.g., rate, displacement, integral) so as to -yield an
effective set of gains for the operator as shown in Fig. 29b. For our
purposes here, however, we will keep the perceptual describing function
and remnant separated from the equalization, so as to permit a closer tie-
in with associated displayed variables; and because we plan to include -the
pilot's scanning, sampling, and- reconstruction effects on the remnant as
well.
1

PERCEPTUAL MODELING

Notion Perception

The material in this section describes and applies the multimodality

pilot model to the question of motion fidelity requirements in a training
simulator. Motion is necessary for pilot-vehicle performance on the high-
est levels: it "unloads" the pilot's visual modality by opening up
another sensory pathway for control of the simulated aircraft and moni-
toring of its dynamic behavior.

Distorted motion may compromise skill development, at least for inex-

?erienced pilots, because the pilot has no reservoir of past flight
experience upon which to assess the motion "leasons" taught by the simu-
lator. For experienced pilots the fidelity can be compromised more; but,
even here, he will not be able to acquire the highest level of skill that
can be acquired in flying the aircraft.
Threshold effects are it1rtant in modeling sensory processesI01.
Given a threshold effect, it is assumed that motions whose magnitudes fall

below the threshold level will not be perceived by the pilot. There are
two types of thresholds modeled here. The first is a sensory threshold,

101 Hofmann and Riedel, 1979, op cit.

105
it arises from physical limitations of the organ itself. This zhreshold
type is evident for the semi-circular canals they are unable to sense
angular velocities of magnitude less than 0.035 rad/sec1 02
.

The second type of threshold is the "indifference" threshold, men-

tioned. previously. This threshold type is evident for specific force
sensing. Under normal workload, pilots appear unable to detect specific
forces of magnitude less than approximately 0.1 g 1 0 3 . It is important to
note that phrase "under normal workload" in discussions of indifference
thresholds. If a subject is asked to concentrate on determining when the
sensation of specific force begins, the sensory threshold is found to be
104
approximately 0.005 g _ the otoliths are very sensitive to stimu-
lation. If the subject is given a task to perform in addition to
indicating the onset of a specific force, this "indifference" threshold is
found to be much higher than the sensory threshold -approximately 0.1 g,

as stated above. In the simulation of actual -flght scenarios, it is

assumed that the indifference threshold will be oper ýive; since the
-pilot's primary task will not be mere detection of the motion cues but
rather an actual flying task. The indifference threshold may be visual-
ized not as a sensory limitation but as the result of an information
processing allocation decision made in the central nervous system to
weight the primary tasks associated with flying the aircraft more heavily
than the task associated, with specific force sensing. Although this al-
location process is not at all well understood, the resulting threshold
effect is easily modeled.

102 Young, L. R. "Current Status of Vestibular System Models,"

Antomattca, 5, 1969, pp. 369-383.
103 Roark, Marvin, and Andrew Junker, "The Effects of Closed Loop .
Tracking on a Subjective Tilt Threshold in the Roll Axis,"
Proceedings of the Fourteenth Annual Conference on Manual Control,
University of Southern California, April 25-27, 1978.
104 Young, 1969, op cit.

106
 The perceptual model includes the sensory threshold for angular
velocity and the indifference threshold for specific force in the appro-
priate paths. Figure 30 depicts the model used to implement these
thresholds as a mathematical functions. Also the threshold level for each
axis is indicated.

[ An alternate,
perception is
but indirect, method of treating thresholds in motion
to use a "divided attention" model of the pilot's sensors.
This technique is involved in the optimal control model employed by Baron,
et al., 105 and Levison and Junker 0 to match rms performance of experi-
mental data obtained from motion simulators. The technique assumes that
the pilot spends fractions of time sampling various sensors (e.g., visual,
vestibular, tactile). These time fractions are then used to scale noise
models of the sensor outputs, and thus to predict rms performance. The
method is indirect in that rantlom noise alone is not a sufficient model
for a threshold either physically or mathematically. It is also not clear

how one systematically sets the values for the various time fractions.

Unfortunately- high fidelity motion is very expensive to obtain in

ground simulators because of the large linear travel required in the mo-
tion base. This suggests that the training simulator can accomplish only
part-task training, i.e., for tasks involving motion with high levels of
fidelity, the motion amplitudes are necessarily small. For full task
training, distortion in the motion sensations must be accepted in a ground
simulator, particularly in the military application where the accelera-
-tions can be relatively high as compared to the civilian application.
This suggests the simulator to be capable of training to a lower level of
proficiency, or as a procedural trainer, a "refresher" for experienced
pilots. The novice should transition to the aircraft before inappropriate
precognitive skills in "flying" the full task simulator are acquired.

S105 Baron, Lancraft, and Zacharias, 1980, op cit.

106 Levison, W. H., and A. M. Junker, "Use of a Tilt Cue in a Simulated

Heading Tracking Task," Fourteenth Annual Conference on Manual
Control, NASA Conference Publication 2060, November 1978.

-1,.-I0"

#2
 2 "PT
Up

T
PT PT
2
T(Ps) P PT sin (p /PT)

o) Presumed Form of Indifference Thrseho/d

AXIS
Pitch
THRESHOLD VALUE
2 deg/sec
K
Surge 0.1 g

Roll
Sway 2 deg/sec
0.1 g

Yaw 2 deg/sec
Heave 0.0 g

Figure 30. Threshold Model and Threshold

.Values for Each Axis
(Borrowed from Hofmann and Riedel 107)

107 Hofmann and Riedel, 1979, op cit.

108
 It is convenient to categorize the kinds of flight motions in three

groupsl08 according to the difficulty and expense of replicating them ac-

curately on the ground. By "accurately" it is implied that the
differences between simulator motions and flight motions are undetectable
in the objective results of training with motion in the simulator. Sub-
jectively the differences may be detectable, although acceptable to the
pilots. The criteria for judgment of accuracy in motion replication
include:

1. The pilot-vehicle task performance in the simulator

and in the aircraft are the same.

2. Pilot task adaptation (pilot model equalization,

gains, etc.) are the same in the simulator and in the
aircraft.

3. Perceptual conflicts and disorientation episodes occur

in similar situations, with similar frequency and fcr
the same reasons in the simulator and in the. aircraft.

The three categories of flight motion are as follows:

The first comprises those low level, typically high frequency motions
which are not manifest in visual displays or simulations of an outside
visual scene, and which make modest demands on linear motion base
travel. These motions provide a "background" and contribute a sense of
realism to the simulated task but are not used in the closed loop control K
sense. They typically are related to the simulated flight environment:
rotor slap, buffet, very low level turbulence (ignorable from the stand-
point of control), etc. These motions may be used intermittently as
indicators of the simulated flight condition. In any event the fidelity

18Caro considers two basic kinds of motion in a similar discussion of

motion fidelity needs. The reader should consult: Caro, Paul W.,
"PlatformiMotion and Simulator Training Effectiveness," Proceedings
of the 10th NTEC/Industry Conference, NAVTRAEQUIPCEN IH-294, Naval
Training Equipment Center, Orlando, Florida, November 15-17, 1977.

109
 requirements are relatively eacily met. Because visual display cor-

respondence is not required, the reproduction at mid-range to high

frequencies is not critical.

The second category includes somewhat larger amplitude motions which

are ased in a closed loop control sense, and whose correspondence with the
visual scene must be sensibly exact (i.e., no detectable discrepancies).
The linear motions are still modest - the angular motions are probably
less than * 5 deg. The engineering fidelity requirements are relatively
high but need not extend to large travel. Typical tasks include attitude
regulation in light turbulence, perhaps even hover if the visual scene
display exhibits negligible delay and the simulated vehicle is well
damped. Fidelity at this level is adequate for most regulatory tasks but
not for maneuvers.

The third category, besides requiring close correspondence between the

visual scene and the motions, also requires large linear -travel of the
motion, base. These motions are necessary to replicate faithfully sensa-
tions in discrete maneuvers and are used therein while acquiring skill or
as a "background" for monitoring the response to a precognitive man-
euver. They are also used as triggering stimuli (e.g., engine failure).
This category of fidelity has only been approached in certain research
simulators, and typically only for a restricted class of maneuvers. Even
so, reproduction of the motion environment at this level is required if
the pilot trainee is to progress through all levels in the Successive H
Organization of Perception for psychomotor skill acquisition in maneuvers.

Meeting the accuracy criteria for the third category of motions is

impractical, except in flight, and extremely difficult even for the second
category. It is therefore apparent that any practical Eimulator will
compromise these criteria, probably to a substantial degree in the mili-
tary context of large accelerations and high psychomotor skill
requirements.

The acthodology outlined herein can identify the difierences in pilot

- task adaptations which will result, presuming that training is continued
until a stable, high proficiency performance level is reached in the

110
 simulator. The model is based upon data taken for well trained pilot
subjects performing in continuous control tasks. It is somewhat less
* suitable for predicting adaptations requiied for the performance of
d iscrete maneuvers; although, even here, the same modeling principles can
be applied for analysis - only the data base is deficient.

However, learning dynamics, as influenced by typical simulator motion

distortions, are not predicted by the analysis technique. Additional
research Is required in this area to establish an appropriate empirical
data base. The objectives of such research would be to quantify the pro-
ficiency level attained versus simulator time as a function of:
H
* Visual scene delay

* Motion distortion (various descriptors)

L * Pilot background

Plight
M Task (including controlled element dynamics).

The resulting data base would allow quantification of training effec-

tiveness of a particular simulator by identifying the point of diminishing
returns - that point beyond which further exposure is counterproductive
or not cost effective fo= the novice pilot. Such quantification of
training effectiveness would be fully compatible with the experimental
9
approaches demonstrated by both HolmanI0 and Bickley!0.

SlC Holman, Garvin L.. "Suitability-for-Training Evaluation of the CH-47

Flight Simulator," Proceedings of the llth NTEC/Industry Conference,
NAVTRAEQUIPCEN IH-306, Naval Training Equipment Center, Orlando,
Florida, November 14-16, 1978. -

Bickley, William R.,

1 1980, op cit.
II0I

__ _ __ III _ __II
 A suggested task for such a research program is that of hovering a

helicopter over a spot in moderate turbulence. To build a simulator cap-

able of emulating this task is, in itself, a significant challenge to the
state of the art; although the difficulties in replicatinF the motion are
not considered to be insurmountable.

Visual Perception

The purpose of this section is to develop and to quantify the needed

rationale for establishing training simulator visual fidelity require- K

ments. The analytical technique is unique and involves the dynamic

modeling of: (a) the training objective (including the pilot's control
activities); (b) the perceptual mechanisms of the pilot as affected by
motion and visibility limitations; and (c) the resulting apparent geo-
metric forms, locations, angular sizes, and angular velocities relative to
perceptual thresholds of information elements within the visual field
which provide guidance and control cues to the pilot. These are combined
in the identification of those areas of the field of view which are es-.
sential for the perception of the cues for guiding and controlling the
aircraft to accomplish the training objective. What results is a syn-
thesis ofl the disciplines of perception, guidance, and control based on
external vi3ual spatial and temporal cues. It then becomes possible to
assess the fidelity of visual simulation in terms of the distortion,
delay, suppression, omissicn, and/or occlusion of essential cues for a

particular trairing objective.

Summary of Resources. Comprehensive expositions of the bases for

understanding human visual perception of spatial and temporal surroundings

112
 were presented by GibsonIII in narrative form with propositions, axioms,
and geometrical graphics which can be tested experimentally. Apparently
motivated in part by a concurrent (circa 1950) survey of the status of
12
research in visual perception by GrahamI , Gilinsky1 13
developed and
validated a quantitative formulation of visual size and distance percep-
tion. Gordon discussed human space perception -mathematically in the
"context of the environmental geometry around a moving eye114 and set forth
the perceptual -basis of vehicular guidanceI15 in the vicinity of fhe
16
ground plane. RoscoeI has more recently addressed size perception as a
function )f visual accommodation and found that, in general, eyes focus
only well enough for the required discrimination.

%Thedevelopment of visual aids for conventional commercial aircraft

approach guidance motivated Calvert (also circa 1950) to extend the theory
of visual judgments of motion1 17. Further insight for quantifying flight
guidance and control by visual cues has been supplied by HavronI 1 8 ,

i Gibson, J. J., The Perception of the Visual World, Houghton Mifflin,

Boston, 1950, Chapters 6 and 7.
112 Graham, C., "Visual Perception," In A Handbook of Experimental
Psychology, (S. S. Stevens, Ed.), John Wiley, New York, 1951,
pp. 868-920.
113 Gilinsky, "Perceived Size and Distance in Visual Space, 1951, op cit.
"Gilinsky, "The Effect of Attitude Upon the Perception of Size, 1955,
op cit.
) 114 Gordon, D. A., "Static and Dynamic Visual Fields in Human Space
Perception," Journal of the Optical Society of America, 55, 1965,
pp. 1296-1303.
1.15 Gordon, "Perceptual Basis of Vehicular Guidance," 1966, op cit.

116' Roscoe, 1979, op cit.

S117 Calvert, 1957, op cit.

S118 Havron, 1962, op cit.

113
"Raish'1 9 , and Grtn-aI And Merhav 1 2 0 These visual cues include, among
others, apparent motion of single points and apparent size, orientation,
and motion of groups of points which may be fixed in the external pan-
orama, but which appear framed by the windshield boundary; perspective
distortions of known geometrical shapes on the ground; perceived rates of
change of lineal and areal dimensions of individual objects and their
differences from the corresponding rates of change of their apparent.back-
ground; and perceived changes in uniform ground textures.

Notion Perspective

The apparent motion of single points which are fixed in inertial space
gives rise to the "streamers" 1 2 1 of point images, and the apparent notion
of inertially-fixed point sets gives rise to the phenomenon of "motion
pese lie"22 123.
perspective,1 created by the streamers of the point set

119 Naish, 1971, op cit.

120 Grunwald and Merhav, 1976, op cit.

121 If individual fix points in inertial space are continually observed

in uni-directional monocular viewing with the line of sight fixed (or
continually moving) in a moving frame, the locus of each fix point is
perceived as a continuous curve, called the "streamer." Streamers K
correspond to the traces created by luminous fix points on the plate
of a camera with open shutter and with the optical axis oriented in
some specified relation to the moving frame. See Alex, Fredric R.,
Geometric Foundation of Motion Perceptive, Systems Technology, Inc.,
Working Paper No. 170-3, January 1967.
122 When observing in uni-directional monocular ,viewing a point set on a
ground plane fixed in inertial space, a set of correlated streamers
is generated which constitute the "expansion pattern" of the region
defined by the point set. At any instant this expansion pattern is
seen to emanate from an individual motionless "focus of expansion"
located at the intersection of the instantaneous velocity vector of
the moving frame with the ground plane of the inertial space. The
expansion pattern of the streamer set (in relation to the motionless
point) is perceived as the "notion perspective" of the subspace
spanned by the observed point set. See Ibid.

S123 Ibid.

114r
£r to true range S~related
with the aid of a characteristic measure of

range known asS~perceived

the apparent distance of vanishing points from
of visual S~~the-principles
perspective1.

: ! Perspective distortions of known geometrical shapes on the ground (or

between the observerandapparent background) can be related toa

Schanges inteperceived oretto fgroups ofpoints wihmay b

fixed in the external panorama. In turn changes in the perceived orien-

tation of groups of fixed points can be related to changes in the o_

observer's position relative to the object defined by the observed point

set.

Perceived rates of change of lineal and areal dimensions of individual

objects can be related to the observer's velocity and acceleration rela-
tive to the object dpafned by the observedgat.poihta
inheether o he
perceived rates of change are judged to be caused primarily by the ob-
server's motion or by the object's motion will depend in part on
differences in perceived rates of change of the object's motion from the
corresponding rates of chanee of the apparent background.

Finally perceived gradients in uniform ground textures give the ob-

server a continuous impression of distance; and perceived changes in
uniform ground textures can be related to changes in slant range, height,
and the relative bearing of the observer's line of sight. It then follows
that perceived rates of change in uniform ground texture can be related to

124 Perceived size, S , and perceived range, R , are related to objective

true size, So, anx true range, R, respectively by:

S R
S R 1 + R/A

where A is the apparent distance of vanishing points from the

principles of visual perspective (i.e., the apparent distance of V
objects at an optically infinite distance from the observer.) See
Gilinsky, "The Effect of Attitude Upon the Perception of Size," 195',
op cit.

115

- ---
closure rate, vertical velocity, and relative angular velocity of the line
of sight, respectively.

Those portions of the theory of visual judgments of motion which co.n-

cern the acquisition of position, velocity, and acceleration information
in both orientation and translation will be applicable to the present
study; furthermore, our interest in the essential cues within the field of
view will necersarily focus on the training objectives in low altitude
operations, including navigation, autorotation, landing, and nap-of-earth
and hovering flight where the apparent relative size of the visual ele-
ments is great but where the relative velocity between aircraft and K
landing pad may sometimes be so low as to inhibit the perception of
"streamers" which otherwise play a more significant role in conventional
visual flight guidance at low altitudes. Consequently for the present
study we shall borrow from the theory of visual perception chiefly those
essential geometrical constructs aud ralationaiso of stati perspective
which describe relative orientation and position between aircraft and
visual elements, some of which are summarized in Tables 7 and 8. Where
the relative velocity between aircraft and visual elements is sufficiently
grent, we shall incorporate those essential geometrical constructs of
motion perspective in cable 9 which are vested in the theory of streamers.

Streamers offer the observer the means to discern the point toward
which a vehicle is moving (aiming point) from a two-dimensional abstrac-
tion of the outside world as projected on the retina. The probable basis
for the perception of visual directional information is the expansion
pattern of the panorama about the "fixed point" toward which the aircraft
is moving. In the parafoveal streamer theory1 2 5 directional information
is obtained by the observer, in motion, by making use of the objects in
the field of view which appear to move along paths radiating from the
fixed point called the "focus of expansion" in the perspective. nhe
vectors tangent to these paths are the "streamers," and the pattern formed

125 Gordon, "Perceptual Basis of Vehicular Guidance," 1966, op cit.

Calvert, 1957, op cit.

116
1-4

z
aA

to I.

C C -, , ..
0. 4 *0 -L v'

41 0. .4 0 0 - :ý -,
S 0t0 - C6a C Vc I

z ~~-.. c..- ,. .b0 1 co .41It -0

E- 0.~ 1 C . .- . C ~ C 0

*-0 -.. O.0C C .0 .0 2 :1

.4IL 5. to5.4

t4 9C W C -C

117H
 * TABLE 8a.
VISUAL FIELD INFORMATION REQUIREM~ENTS: P(6SITION

OUTDR LOOP PRIMARY COU(FROL RESVUSE

ID A7PPROPRIATE- RATI:XLE FOR
r5S~EN)"i T RUISFER CLOSURE OF CWOSEr) LOOP YOM
JEOl?~ot~ -
F E FDOAZ" FUNCTIOfl CTHjR jOOpS I&U?/-ATOR PATIO r~

x C ~ees~az for

FPositio
U
Lon--g
al
Positio _________
AT-tu-
be
-
e
eW; aqr

_____
SO~~
a
.h
(0
Ig
S

________landing_
-x
apprzazh vzd-
ance, se
z:ia
ly at IOU
speeda, in
hovering, and

Paer 6, 5 - :bab SDo

% recessaty7 for
be R. pproach,
-- 0-15Tbabr
beb Izhovering and
launding guid-uJ
lvert~al 66~
0. or CO 2&6 wher 0aviergu atdg idacel ln
Position B.x ;qArgCo 70
1(o)
orluin bT '6.6 gP Cos 78e6
A -0go.6 - N , - e TL

ST'

118c-s LI-f-
- ~ 7:;:C;~>-p'...~ -
 4-

64

v, 4,
.4' vi v~
2 S O
x
to 0 w
E-4
0~rd
E4

0 L

0 ~/0

J44

H 0

8450

>~ 4,4'
-4.
' C
g- 0,
00 ý 9. "'-4
0

Ut

4,
N 0
VC CV

~ 4

0 r-04

0 "
0 4, 4
ao oj
-. Ns
I:4,

dD
,,4,
Z.4 4,>

4,1 , v

4 Q4a
Q 4
41 w~
c0 4
__
___ __

119
ft 141
 Ci)

&-4

CC
1
rj0 cu

-0
I
o".30
c t*

10

I.,~~~ CC CC11-~
-- 9

v 0., C1 .0 Z.

'-40

a..-

01
f VIf itffII

0 I 11 c

zz

>u IC

1-I C. *120
 by them is the point toward which the observer is moving. Sometimes this
point can be observed directly as when the observer's speed is high,
weather is clear, and objects are close. Otherwise the location of the
point must be inferred by extrapolation of the streamer pattern character-
istics of various types of objects and ground texture.
S. Ringland, et al., 126
presents an analysis showing how the interpre-
tation of motion perspective geometry will enable the observer to
anticipate changes in the future course of his motion. When present and
recognized, these essential visual elements from motion perspective, in
turn, will enable an observant cont-ller to provide first- and second-
order visual lead compensation127 of his controlled element without the
customary intensive psychomotor workload which accompanies visual antici-

pation of low frequency motions.

Pattern Recognition

Perfect recognition by the pilot of the visual elements which are

necessary for guidance and control of the aircraft in low altitude opera-
tions is an idealization of reality even though in most low altitude tasks
the experienced pilot must recognize changes in, very familiar memorized
geometric patterns. Memorizing and recognizing a pattern involve cogni-
tive activities which can be considered analogous to meraorizing and
repeating a sequence of psychomotor activitites. Table 10 presents the
summary of a sequential pattern perception and recognition theory from

126 Ringland, Robert F., Warren F. Clement, and Henry R. Jex, Factors in
Determining Visual and Motion Fidellt- Requirements for Trainins.
Simulators, Systems Technology, Inc., Working Paper No. 1162-4,
December 1980, Revised March 1981.
127 Second order lead compensation here is akin to the adoption of
pursuit behavior measured in: Allen, R. Wade, and Duane McRuer, The
Man/Machine Control Interface - Pursuit Control, Systems Technology,
Inc., Paper No. 243, April 1979.

121

j$ . . - 5oS -g
 _4

TABLE 10. SUMMAAY OF A SEQUENTIAL PATTERN PERCEPTION THEORY

SUXXAY OF A SEQUE1IIAI. PATTERN•Z R•KIS AOlt COhNWECTIOXSWITH PERCEPTUAL

ERtCEPTIOC CES"TfEREDMID
?: fOGNITION THEORY CEKTERED AltD OTHER MODELS
"(Prom
Noton

1) Memorizing a pattern i. the process of constructing 1) Closed cyclic nature of feature network.
oan Internal representation of the pattern In.memory,
"
in the form of a sequential feature network. a
closed network of memory traces recording the fee-
tures of the pattern and the attention shifts
required to pass from feature to feature across Zhe
visual field.

2) Becognizing a pattern is the p-ocess of finding in 2) Closed-loop process of recognition: "matchln" _

swmory a feature network which matches tne pattern, proceeds at the compensatory level In the most
the matching being carried oitt sequentially feature unfmiliar situations.
by feature.
129
3) The attention shifts from feature to feature may 3) Consistent with Sanders' flndlngs , internal
take the form of saccadic eye movements or of in- attention shifts proceed at neural speeds.
ternal attention shifts, according to the angular
displacement involved.
4) During recog. ition the ma.tching process Is guided by 4) -Matching- is aided by short-term secoryv whj.¢h
the feature -etwork. which directs attention fro% is consistent with Sperling's findingsli
featurn to feature of the pattern. Peripheral vision may also guide the matching
process at the pursuit-level in ýore familiar
situations.

5) The directed ;sture of the matching process (note in 5) Consistent with Fackvorth's fings that
4) sa the key to the recornition of patterns in the visual noise causes tunnel vision *

6)
presence of noise and clutter. The feature network
directs attention to the features of the pattern,
while avoiding the noise and clutter.

Memorizing and recognizing a pattern are seen to be 6) Consistent with succe s i ve organization of
U
closely analogous to memorizing and repeating a perception (SOP) theory
conventional sequence of behavior, each being an
alternating sequence of sensory and motor
activities.

7) Thus habit produces the acan-path, a habitually 7) Inaracterized by great determinism.

preferred path followed from feature to feature
through the feature network and, correspondingly.
across the visual field. This path differs from
*• person to person and from pattern tt' pattern, but is
fixed and characteristic for a given person viewing
a given pattern.

8) Under conditions in which attention shifts must take 8) Contrast these findings with the apparent lack
the form of eye movements, the development of the of determinism in Instrument scanning under
scan-path during memor~zation of a pattern has been IFR reported by Fitts, et al.; Veir aTIn6,
1 eln;
experimentally demonstrated. Its use In subsequent Clement, et al., and Clement, et al.,3 end
recognition awaits co.afirmation. their antecedents. L

128 Noton. 1970, op c.t.

129 Sanders, A. F., 'Some Aspects of the Selective Process In the Functional "leual Field,. Ergonomics, 13,
1970, pp. 101-117.

130 Sperling, C.. 'Successive Approximations to a Model for Sh-vt Term Memory,' Act& Psy-hologic, 13.,1970.

131 Ihcktorth, N. R., -Visual Noise Causes Tunnel Visior., Psychone'jrological Sýiences, 3_ 1965, pp. 67 and 68.

132 McRuer and Krendel, 1974, op cit.

133 Fitts, Jones. atd Milton, 1950, op cit.

Vair and Zlein, 1970, op cf.t.

Clement, W. F., at 4l., Aplication of lknual Control Teor, to the Development of P1ight Director
viThlaw
5,mtrms for STO0. Aircraft, fart 11: )W/ti-As Sampllng, P!Ilot Wtrkload, and Diploy Integration- System
-thnology, Inc., lhchnlcel isport No. 1011-2, 1974.

COemnt, U. F., at al., A Meesmination of Pilot Ee -Movement Data, Systems Technology. Inc., Technical
Memorandum No. 263-A, iev. 28 February 1967.

122

I
 "Noton134 together with some remarks and connections with the SOP theory
and other findings which have been found to be useful in characterizing
visual perception and cognitive activities. In terms of the successive
organization of perception theory 135 and the sequential theory of pattern
perception in Noton1 3 6 , the experienced pilot is probably always operating
at least at the pursuit level of pattern perception. At the pursuit level
of pattern perception, it may be reasonable to assume that recognition is
conditioned only upon detection with fairly high probability, say 0.95,
and insignificant additional time delay over that already accounted for in
the effective perceptual-motor delay, Tel of the crossover model for the
A; compensatory level of control, unless head movements are required to scan
the functional visual field as defined, for example, by Sanders' selective
process 137

In contradistinction, the student pilot may not yet have memorized all
ofthe changing geometric patterns in the visual elements which are neces-

sary for guidance and control of the aircraft in low altitude

operations. Thus the student pilot may be operating at the compensatory
level of p~ttern perception as well as the compensatory level of con-
trol. At the compensatory level of pattern perception, recognition may be
conditioned upon detection with a lower probability than 0.95 and addi-
tional time delay over that already accounted for in the effective
perceptual-motor delay, Tel of experienced pilots. Experiments to test I
this hypothesis with a student pilot population have yet to be conducted.

134 Noton, D., "A Theory of Visual Pattern Perception," IEEE

STransactions, SSC-6, 1970. pp. 349-357.
1
135 McRuer and Krendel, 1974, op cit.
136
137 Noton, 1970, op cit.

137 Sanders, 1970, op cit.

A4 123

•-i
 Sanders has shown in a decade of visual research at the TNO in the
13
Netherlands • that a "scanning controller" function exists which directs
the motibns of the eye and head to points of interest in the visual

field. For our purposes, "points of interest" in the functional visual

field would correspond to the locations of visual elements and geometric
patterns which are necessary for guidance and control of the aircraft. We
have called these points of interest essential "cues" for brevity. For
cues at angles from the foveal axis of less than about 20 deg, the eye
need not, but may, be moved; for angles between 20 and 60 deg only the eye
is moved in one dominant saccade at a slew rate on the order of
300 deg/sec. For points-of-interest further than 60 deg, the head is
moved as well, thus requiring more time to complete the saccade.

Numerous experiments 1 3 9 , using eye- and head-point-of-regard instru-

mentation, have shown that both the eye and head tend to move in a series
of very sharp sac~ades (steps) while changing one's point of regard. The
resulting image blur during a saccade is suppressed internally by a feed-
forward signal from the eye movement controller, so that pe.'ception is
briefly lost during each saccade. Thus the net effect on the perception
of a cue's form or motion is a short average delay due to scanning

i ,

138
138 Ibid.

139 Fitts, P. M., R. E. Jones, and J. L. Milton, "Eye Movements of

Aircraft Pilots During Instrument~-Ilanding Approaches," Aeronautical
Engineering Review, Vol. 9, No. 2, Feb. 1950, pp. 24-29.

Weir, D. H., and R. H. Klein, The Measurement and Analysis of Pilot

Scanning and Control Behavior During Simulated Instrumented
Approaches, NASA CR-1535, June 1970.

Young, L. R., L. Stark, and C. Kupfer, Physiology of the Visual

Control System, NASA CR-238, June 1965.

124
 140.
saccades.140 Consequently our divided or
intermittent attention models
141
for perception , coupled with GilirF~ky's models for the perception of
size and distance 1 4 2 and streamer theory1 4 3 , should be adequate for
application in the guidance and control Analysis for the present study.

Perception of the changing geometric properties, however, is condi-

tioned upon the perceived photometric properties, such as illuminance,
reflectance, and color, which will be altered by environmental attenuation
and scattering. At the relatively short visual ranges to the visual ele-
ment from hovering and near-hovering positions, scattering is the

predominant effect at night. Scattering acts to reduce contrast and to

desaturate colos4 Backscatter from landing lights on the aircraft,

fintermittent attenuation from precipitation and the possibility of re-

fiected glare from flood ligh-s provide a complex distributior. of Veiling
luminance in the external field of view at night. Scattering and at-
tenuation increase with the density of precipftation (e.g., rain, sea
spray, fog) and the range of the observer.

S140
The details oi the intentional movement dynamics of both the head and
• eye are fairly well known and are quite complex in detail (e.g.,,
Ibid). However, because the saccades are rapid (less than 0.06 sec
for eye-only motionr), for purposes of analysis and evaluation the
effect can be represented by a delay in perception of displayed
• &tates. This delay can be lumped with a typical detection delay of
•:• About 0.02 sec.

McRuer and Krende!, S•141

1974, op cit.

S142Gii-insky,."Perceived
Size and Distance in Vir~ua! Space," 1951,
Op cit.

• Gilinsky, "The Effect of" Attituie Upon the Perception of Size," 1955,
op cit.

Goron 1965, op
ct.
S•
144Middleton, W. E. Knowles, "Vision Through the Atmosphere,2' Canada:
S• 1'952.
University of Toronto,

? :12'5
 Visibility Effects

Luminant source attenuation effects, which predominate in daylight,

14 5
have been described by the Koschmeider theory . The effects of back-
scatter and glare at night cause the inherent contrast of a luminant
source to attenuate more rapidly than is described by the (exponentially)
linear relationship with range identified as Koschmeider's law14 6 . The
effects of backscatter and glare can, however, be imbedded in a modifica-
tion of Koschmeider's law by representing the exponent as a truncated
4:
power series in range, R147

Allen and McRuer 1 4 8 have coupled the modified Koschmeider theory with
the modified Blackwell-Davies contrast thresholds149 in studies of how the

145 5Ibd.

146 Frenk,-R. S., D. E. Skaar, and .. A. Tennant, "Driver's Visual

Detection," SAE Paper 720142 presented at the SAE Automotive
Engineering Congress, Detroit, Michigan, 10-14 January 1972.
147 The observed luminance contrast CR, at range, R, is given by

C - Cexp (-R- XR .)
where CO = inherent source or target contrast
So
W Koschmeider's meteorological extinction coefficient
X - backsca&tering coefficient

148 Allen, R. W., and D. T. HcRuer, "The Effect of Adverse Visibility on

-liver treering Performance in an Automobile Simulator, SAE Paper
770239, March 1977.
•" ~149
149Davies, E. B., Contrast Thresholds for Air to Ground Vision, RAE
Technical Report 65089, A4ril 1965.

Blackwell, H. R., "Contrast Thresholds of the Human Eye," J. of the

Optical Society of America, No. 36, 1946, pp. 624-643.

Blackwell, H. R., and D. W. McCready, Jr., "Foveal Detection

Thresholds for Various Durations of Urget Presentation," Minutes and
Proceedings of NAS-NRC Vision Committee, November 1952,
ACSIL/53/4405.

226
 visual segt.'et can affect perceptual cues for guidance and control of
surface vehicles. The same technical approach ha. been applied by Clement
1 50
and Heffley to estimate the minimum effective visual range for guidance
and ccntrol of VTOL operations at night.
15 1
Jewell, et al., describe a simple method for measuring and comr-
paring, in the closed-loop context involving the human operator, various
techniques of obtaining computer-generated images (CGI). The proposed
method is independent of the manufacturers' hardware and software specifi-
cations and allows different CGIs to be compared in situ on an absolute
scale as well as back to back. Furthermore the method also offers a ra-
tional means for evaluating hardware and/or software changes to extant CGI
systems and to their host computation systems which provide the aircraft
mathematical model. For exa=ple a carefully designed experiment using the
method described in Jewell, et al,152 can reveal how the effects of asyn-
chronous data transfer and lead prediction or 9moothing compensation in
the CGI software affect the performance of the human operator in accom-
plishing a specific training objective.

The optimal control and estimation methods demonstrated by

Wewerinke153; Baron, et al., 154 and Zacharias and L;vI.son155 offer a

111
S150 Clement, Warren F., and Robert K. Heffley, Some Effects of Adverse
* Visibility on Threshold Properties of the Pilot's Perception in VTOL
Approaches to Non-Aviation Ships, Systems Technology, Inc., Working
Paper No. 1115-3, May 1978.
151 Jewell, Wayne F., and Warren F. Clement, A Simple Method for
Measuring the Effective Delay in Simulator Digital Computing Systems
Involving Manual Control, Systems Technology, Inc., Working Paper

12No. 1156-4, December 1980.

15 Ibid.
Wewerinke, P. H., Visual Scene Perception Process Involved in the
Fianual Approach, NLR-TR-78130U, 1978.
154 Baron, Lancraft, and Zacharias, 1980, op cit.
155 Zacharias, Greg L., and William H. Levison, "A obdel-Based Procedure
for Determining Visual Cue Requirementsi" 1981 IMAGE II Conference
Proceedings, AFHRL, Scottsdale, Arizona, 10-12 June 1981.

127
means of calculating the relationships between the geometric pattern
details and the quality of state information used by the pilot. Care must
be exercised, however, to insure that the "optimal" solutibns really do
reflect correct or realistic piloting technique and task performance, It
may be advisable to simply use such methods to establish likely starting
points for pilot-vehicle models which are then studied in order to reveal
the fundamental first-principles effects. The concern is that the mathe-
matical complexity usually associated with optimal control and estimation
procedures might cloud the insight necessary to Identify basic cause-
effect relationships. (These same comments also apply to use of optiral
control techniques in conjunction with motion fidelity.)

Sensitivity of Visual Cues in Low Altitude Operations. We have al-

ready presented in Tables 7, 8, and 9 some examples of key roles for
visual cues in low altitude operations. In this .topic we shall introduce
a measure of sensitivity which affects the pilot's perception of those
essential geometric properties of visual elements which provide position
information upon which he, in turn, takes action to control. This measure
of sensitivity is also summarized in Table 8.

The need for a wide parafoveal - even peripheral - awareness of

position and velocity cues is apparent from the appearance of the landing
pad in Fig. 31. This is a polar picture plane representation of the for-
ward field of view from the left seat of the XC-142 in a level
156
attitude . The appearance of a ctrcular landing pad is shown as the
aircraft proceeds on a constant vertical rate of descent approach at con-
stant horizontal deceleration. Notice that even under visual conditions
the tracking of the aiming point 1 5 7 with respect to the pad is a pursuit

156 See Roberts, Edward 0., External Visibility Criteria for VTOL
Aircraft, AFFDL-TR-67-27, March 1967, for the method of
representation.
157 A geometric construction for inferring the approximate location of
the moving aiming point is illustrated in Fig. 32. Notice that the
aiming point will always pursue the pad from a shallower depression
angle, i.e., 'from "above" the apparent pad. until arrival over the
spot ior hover.

128

g-*- - -~
 Polar picture plane representation of forward hemispheric field of view

Hover altitude = pad radius 2Th

-Irate of descent
=
Zenith' a. M horizontal deceleration

T. Wh/a - normalizing unit of time

Elapsed time interval between successive

apparent locations of pad is Ta

Pad at Pad a

one-half hover
hover . altitude

Figure 31. Circular Pad Appearance in Vertical Landing

Approach at Constant Rate of Descent and Horizontal
Deceleration From Left Seat of XC-142 Cockpit

129
 Zenith i

To locate aiming point, construct lines EP and X'P tangent

to their respective loci of extreuities of the- pad's major
axis EE'. The intersectiona P is the aiming point from
which all apptrent motion atreaawers will originate.

[ ~ Locus of extremities of the

Semimajor axis corresponding
with the elliptic apparation
Sf -thepad-
-- - - itantaneous
AiingPoint t a

-1
I

atorepndn topaoajrers E

~one-a~lf
a~tit~de / *A

ta a
,~& A

ligure3.Ineec
Foeh ofAmnPitLoainFo te

Apparent "Motion Streanmrs" Aesociated with a Circular

Pad in a Vertical L~anding Approach at Constant Rate of
Descent and Constant Horizontal Deceleration

130
task until the pad is occAx'ed by the nose. in this aircraft, which has a
rather generously depres:4d field of view, the pad is marginally visible
throughout the critical portion of the conversion to hover. This profile
is most prone to pad overshoot when a transitio-v from instrument flight
rules to visual flight rules is required.

The appearance of a circular landing pad with reduced visual range is

shown in Fig. 33 for the same type of approach as in Fig. 31. The mdximum
slant visual range is assumed equal to 17 pad radii (approximately
700 ft). For example if the rate of descent, h = 10 ft/sec, and the hori-
zontal deceleration, ax = 5 ft/sec2 , the normalizing unit of time, Ta 2

sec, and the pad radius in Figs. 31, 32, and 33 will be 40 ft. Note that
for this case the maximum slant visual range will be 680 ft, which cor-
responds roughly to Runway Visual Range for ICAO Category III-A. Notice
how little time (5 sec) Is available after acquiring visual contact with
the pad to establish pursuit tracking of the inferred aiming point with
respect to the pad before the pad is occulted by the nose. Thereafter
inference of the aiming point from motion streamers is hampered by the
occultation unless other ground texture besides a partial glimpse of pad
outline becomes visible. Yet the pilot must continue to pursue the
inferred center of the pad with the inferred aiming point until coinci-
dence is achieved, whence the pure descent can be arrested in hover over
the pad.

When the angular regions in the field of interest defined by Sanders'

selective process '(p. 123 herein) are considered, it is clearer why pilots
may experience significant workload in acquiring external visual cues
close to the pad. Experimental work reported in Gordon 1 5 8 , while applied
to automobile driving, confirms that the use of small aperture viewing
(without correspondingly improved viewing resolution) is reported to cause
increased workload, suggesting that denial of the peripheral cues may lead

158 Gordon, 1966, op cit.

131
;Polar picture plane representation of forward hemispheric field of view

Max slant visual range A 17 pad rqdii

lover altitude - pad radius -2ah

h - rate of descent
a- horizontal deceleration
_na ft h/ax = normalizing unit of time

Elapsed time interval between successive

apparent locations of pad is Ta

Figure 33. dircular Pad Appearance in Vertical Landing

Approach at Constant Rate of Descent and Horizontal..
Deceleration with Reduced Visual Range From Left Seat of
XC-142 Cockpit

132
 Melt-

to significant degradation in the perceptive structure, a point also ad-

15 9
dressed by Roscoe .

Figure 34 shows a short runway approach viewed from inside the same
aircraft. A constant glide slope angle of 9 deg is maintained. Notice
how the depression angle of the runway threshold at a constant 9 deg below
the horizon coincides with the fixed aiming point and offers a .compensa-
tory cue for maintaining position on the glide slope. Furthermore the
perspective angle of the line-up centerline or the asymmetry of the
(lighted) runway outline will provide a compensatory lateral displacement V
cue as shown in the inset at the lower left corner of Fig. 34. The tran-
sition to a level of skill higher than compensatory is thus postponed
until the flare point is approached or until the minimum go-around deci-
sion altitude is reached. The runway appears in the pilot's tunnel field
of view throughout most of the approach. This constant-angle profile
offers superior potential for compatible instrumefit-to-visual transitions. '

If the maximum slant visual range be limited to 1000 ft (2/3 of the

runway length in Fig. 34), the appearance of the runway will be as shown
in Fig. 35. Although the (lighted) approach line-up centerline will
appear long before the runway threshold outline, perception of the
perspective angle of the line-up centerline for the purpose of lateral
guidance will be compromised by the absence of an external local hori-
zon. After the threshold appears, its depression angle will be useless as
a glide slo: ! position cue unless a collimated artificial horizon is
displayed (head-up) in registration with the (invisible) local horizon.
Therefore, in addition to itg vital role as an attitude reference, a col-
limated artificial horizon perfectly registered with the local horizon is
essential to the abstraction of line-up and glide slope displacement in-
formation after transition to visual contact with the ground, if the
external local horizon is occluded.

159 Roscoe, 1979, op cit.

133
 py,~

Polar picture plane representation of forward hemispheric field of view

Flare altitude A 33 ft Runway length - 1500 ft

Runvay width - 100 ft

(I
ZI'?ithi-

Elapsed time Interval between successive

apparent locations of the runway is 200/VG
"where VC Is ground speed i. ft/unit time

iV
Hortixon
4A

i/

a\ ay ..
at one-half £jdi"
"flare altitude

" 'Inset S•'•. *symmetry

showing pezaspective of outline std centerline
Inclination with respect to a line perpendicular to the

horizon when aircraft is displaced right of upprosch zenterli

borison .

havay
b

C. approach

Figure 34. Rwivay Appearance in a Short Landing Approach at

i•.of 9 deg Glide Angle and Constant Ground Speed From Left Seat
iC12Cockpit,

1314.
 * -mw
,.

Poler picture plaue representationa of forward 'hemispheric field of viev

HMat slant visual range - 1000 ft Runway length - 1500 ft

Flare altitude 33 ft Runway width - 100 ft

_Zenith

Elapsed time interval between successive

apparent locbzions of the runway ig 20 0 !Vc
where V is ground speed in ft/unit time

GA

' /I KA

Uiway
at one-half
flare altitude J..
1 kzdir

Inset showing perspective asymmetry of outline and centerline

inclination with respect to a line perpendicular to the
horizon when aircraft is displaced right of approach centerline
Norigon
Runway
Outline

approach

Figure 35. Runway Appearance in a Short Landing Approach at 9 deg

Glide Angle and Constant Ground Speed with Reduced Visual Range
From Left Seat of XC-142 Cockpit

135

- __ _ -
 :+ •++•... . .. . . --.. . " -.- -

Considerations for Vertical Displacement Control. Expres-

13ions are shown in the extreme right column 9f Table 8b, for
the sensitivity of the line of sight deoression angle, 8, to

ii changes in ground pzsition, x, and height, h. The advantage

of the constant path sngle apprcach to the runway or pad is
The compensatory guidance and control feature of the runway
or pad appearance at constant depression angle. The superior
altitude sensitivity of pad depression cngle on shallow path

L angle approaches is evident from the expression for 30/l(h/S)

inTable 8b where the sensitivity decays as the square of the
cosine of the depression angle of the line of sight to the
pad. The altitude sensitivity is quite ecceptable over the
range of relatively shallow approach angles.

Ln contrast, the sensitivity of lateral and longitudinal

ground position cues while in hover and near trAnsition to
hover is unfortunately best at large depression angles near
the nadir. This underscores the need to keep the simulated
range st large depression angleg. Forward field of view at

these angles is nearly impossible to provide in aircraft.

V
Hover position sensitivity perceived by the pilot falls

off to about half its maximum value as the line of sight

rises to 45 deg depressi-on. This two-fold change in ground
position sensitivity while on VFR is most important when
teaching a student to transition from a hover posltiv dis-
play to a head-up display. We have seen a va-xfity of
simulated plan position indicators which present to:e nadir
view of the pad in relation to the aircraft or the aircraft
in relatien to the pad. While using such an instru•ment, the
pilot has the advantage of skaximum longitudinal and lateral
displacement sensitivity over the ground, yet when he seeks
the transition to visual contact, his line of eight depres-
sion angle to the hangar or pad may be much less than
45 deg. On acquiring the hangar or pad !isually, the pilot
may discover Zhat the displacement sensitivity is less thaa
half that to which he was accustomed on instruments. Thu:-:
the pilot is ssaddenly required to adapt his loop gain to the
change in perceived displacenent sensitivity. Although a
three- or four-fold Increase in pilot gain is qbite possible,
the IFR/VFr: trancit~on poirt is a very unfair place to re-
quire it, since the pilot is preoccupied with vehicle trim
changes and with the transition to hover in an unforgiving
vehicle.

Donsiderations for Lateral Displacement Control. tbnýider

the pilot'u view of the landing pad centerline as seen frvtm a
point short of the pad and displaced to the left. Geometri-
ially the situation is as shoun in the left portion of
Fig. 36, with the pilot's eye vicw chcawn on the right. •Te

136

- - -- ~ -~ - ~ ::-7
 7 Nowinal Glide Slope Angle

V = Approach Line Up Centerline Perspective Angle

y = Lateral Displacement

h = Altitude

F F Line Up Centerline
12

Observer; 0 Horizon

JfF 7t~
Observer's View of
S~F.F2 Ground Range, Foresho:rtened Line Up C

Line Up F2

Figure 36. Geometry for Perception of Time-Advanced

Lateral Deviation

137
 sensitivity of the perspective angle, V, to changes in lat-
eral displacnment, y (with the latter normalized by the
ground range, S), is equivalent to the partial derivative,
i.e.,

3V 2
7 -cot Y sin V

This sensitivity is greatest at shallow depression angles

(small y) when v is near 90 deg, i.e., when nearest to being
lined up with the pad centerline. It falls to half of this
value for v - 45 deg and increases rapidly toward zero for
decreasing v. The strong lateral displacement or line-up cue :'
provided by the line-up centerline is also evident in the
inset in Figs. 34 and 35. A

One way in which the line-up centerline displacement cues

can provide for adoption of an advantageous lateral displace-
ment control loop-st ture on the part of the pilot is shcwn
by Ringland, et al.,L. The perspective angle of the line-
up centerline in combination with its time rate of chfnfe
provides an effective time-advanced lateral deviation 6
Preview is explicit in this case simply by viewing the per-
spective angle of the centerline and associating this angle
with the lateral position error that the vehicle ipould have
at some point ahead of the vehicle if it continued along its
current path. Other ways in which motion perspective "strea-
mers" can provide the equivalent of time-advancec lateral
deviation are also illustrated by Ringland, et al.. where
the ground range to the point of regard is given by S, which
is approximately equal to the relative closing velocity, VG,
multiplied by the time, T0 , to travel to that point. The
time advance, T-, provides a perceptual preview which results
in a pure-leaX equalization in the effective controlled
ele
K/s ent
formdynamics. This, deviation
of the lateral in turn,dynamics
offsets atthe
low undesirable
frequency. il

160 Ringland, it al., L981, op cit. 1

161 Weir, David B., and Duiane T. McRuer, Conceptual.ization of Overtaking
an? Passing on Two-lane Rural Roadsa, Franklin Institute Research
Laboratory, Technical Report No. 1-193, December 1967.
162 Ringland, at al., 1981, op cit.

138 .11

•- ..... " . . . . . . . .. . ..... ... . ..-

 Concluding Remarks. The point of all this is to demonstrate that the
pilot's perceptions of changes in altitude, h, cross range, y, and ground
range, S, errors are functions of where and how the visual field elements
are located in his field of view (3, y, •, and V in Table 8), in this
instance, all on the (approximately horizontal.) plane representing the
earth's surface. Further, these functions can be explicitly quantified.
For training obje tives in addition to the landing approach, the objective
oi our research is to -develop similar measures of visual element changes
with vehicle motions, and various disturbances; and to relate these mea-
sures of change to the pilot's ability to perceive or discriminate the
change. It will then be possible to assess the fidelity of visual simula- L
tion in terms of the distortion, delay, suppression, omission, and/or
occlusion of essential cues for a particular training objective.

.•'I'
1

:•-f 139

' I
f^ .
 ARALYTIC PRO(ZDURE FOR DETEIMINIOG NECESSARY
VISUAL AND MDTION CUE STIMULI

The pilot uodel concepts outlined in the preceding section form the
basis for the fidelity requirements analysis. The methodology uses the
analytical techniques of the control engineer to formulate plausible sys-
tem structures consistent with human pilot characteristics expressed in

quantified, mathematical terms. These structures can then be investigated

for their sensitivity to distortion or disruption of the visual and motion
inputs to the pilot. The results are interpreted in terms of fidelity

requirements for pilot training.

The procedure itself does not require inordinate mathematical manLpu-

lation and computation. Rather it is the judicious application of
modeling techniques discussed previously which requires considerable

thought before a plausible and persuasive formulation is achieved. As

discussed previously, one of the prime sources of information for this
step is the pilot himself. Th~s does not necessarily imply pilot surveys
with thei- tabulations of subjective instrument usage versus task,
aircraft type, and training background; although all of these have their
place. More important are discussions with pilots on the techniques they

use, or think they use, to execute particular flying tasks. Again, as

Carter and Semple 163 point out, pilots also should be queried about the

possible incorrect performance of a particular maneuver. These verbal

descriptions are admittedly fallible, yet can often be translated into the
system loop structures discussed previously which can then be examined for
consistency with existing pilot model data. The analyst thereby avoids

laborious tuning of -a performance index or exhaustive consideration of

every possible" structure in favor of the -most likely."

S163 Carter and Semple, 1976, op cit.

140
 ANALYTIC PROCEDURE

A concise statement of the analytic procedure is given in Table 11.

Let us proceed through this list with a brief discussion of each item and
then consider a specific example.

Compile a Quantitative Description

of Components (Step 1)

The fivst step in the procedure is to gather the material necessary to

permit quantitative definition of the key variables for a particular fly-
ing task. These include the task itself, the vehicle dynamics, the visual
scene kinematics (as appropriate), the display dynamics, the disturbances
(and vehicle response properties thereto), and the commands (desired man-
euvers, flight paths, etc.). The three longitudinal degrees of freedom 16
4 (motion in the vertical plane) and the three lateral-directional degrees
of freedom1 6 5 (motion in the horizontal plane) are normally uncoupled from 1•
one another (neither influences the other) or nearly so, making possible a
separation of the system dynamics into two smaller analytical and more
manageable pieces. Only very special cases of fixed- or rotary-wing
aircraft require coupled longitudinal and lateral.-directional dynamics
16 6
involving six or more degrees of freedom.

164 Actually there is only one "longitudinal" degree of freedom. This

misnomer has come into deliberate and pervasive use for the sake of
brevity; because the longitudinal, vertical, and pitching degrees of
freedom are always coupled.

Te third S165
"lateral-directional" degree of freedom is roll and,
although not stated, is always coupled with the lateral and
directional motions.

166 Additional degrees of freedom can be contributed by structural

flexibility in the case of fixed-wing aircraft or by the rotor in the
case of rotary-wing aircraft.

141
 ~ i-:

"4-

TABLE II

ANALYTIC PROCEDURE FOR DETERMININC MINIMA-L LEVEL

OF VISUAL AND MOTION CUE STIMULI NEEDED TO TRAIN

(1) Compile a quantitative description of common errors: task, vehicle,

piloting technique, and disturbances.

(2) Identify command (outermost) loops for each task component.

(3) Determine whether vehicle dynamics permit successful task execution

wIth only the command loops closed.

-- If so only command loops are essential, therefore go to

Step 5.

(4) Determine which additionacl loops are necessary to permit successful

task execution--these additional loops also become essential.

(5) For each essential loop, look up the catalogued fidelity potential
for a specific candidate medium and determine whether this potential
permits the required loop closure.

- If not then simulator training will be compromised for that

particular loop.

- Catalog will include most common or likely student errors

142

--

X I
 In executing this first step, it is exceedingly helpful to have the
benefit of experience in selecting information. Some areas are well de-
fined and documented while others will rzquire extrapolation of data or
estimations requiring good engineering judgment. Perhaps the most crucial
aspect is strict economy of system parameters without losing the ability
to include all critical system effects. Two approaches can be taken:

i. Start with an overdefined system, then weed out

unnecessary complication

or 2. Start with an underdefined system, then add

Scomplication as required. 3

The latter is preferred because it will most quickly provide some degree
of "solution" in order to get the analysis process underway. The hazard
in this may be the willingness to accept immediately the first solution
without qualification.

Identify Comiand Loops for

Each Task Component (Step 2)

The next step is to establish the likely system loop structure re- I
quired to regulate against disturbances assuming information for the
outermost loops. This is usually visually-derived information only. This
may be relatively easy (pilot input strictly governed by explicitly dis-
played quantities) or more difficult (visual information derived from the
pilot's out-of-cockpit view in a multiple loop task). Past experience
plays an important role here; but clearly the pilot cannot respond to
0Z- something if he c ann•o t perceive it; and to control a particular degree of ,'

freedom (e.g., pitch attitude, vertical velocity) generally means that

of it or something closely related to it must be possible.

143
 In most cases the command (outermost) loops will consist of either
displacements or velocities. The only acceleration command loop which is
obvious is normal acceleration command for fixed-wing air combat man-
euvering, and even this may not be appropriate for helicopters.

Determine Whether Vehicle Dynamics Permit

Successful Task Execution with Only the
Commnd Loops Closed (Step 3)

The pilot equalization required for the loop structure is established

according to the guidance, control, and pilot-centered requirements out-
lined earlier. System stability is foremost, followed by system
performance and workload. Performance is often expressed by loop c"osure

bandwidth or crossover frequency (which can -be shown to be near-

equivalents) in the absence of specific information on the disturbance
being regulated against or the maneuver precision required. At the end of
this step, one usually has the basic system model for the no-motion input
case where the task is purely one of regulating system error (no command
preview, implying YP 0 in Fig. 27; no motion implying Y, 0 in
Fig. 27) using only the visual modality - see Figs. 26 and 29. In rare
cases the analysis may show that successful task execution is possible
with only the command loops closed, because the equalization requirements
do not involve lower frequency anticipation or lead and the performance
demands are not great. If this be the case, we can jump to Step 5.

Determine Which Additional Loops are

Necessary to Permit Successful Task
ftecution (Step 4)

If the equalization requirements are extreme (particularly for lower

frequency lead) or tightly constrained by performance demands, the vehicle
will be troublesome, difficult to fly, and will rate poorly in terms of

144
 pilot opinion or workload for the task under consideration. Even when
less extreme, the need for lead equalization signals for potential utility
of motion inputs to the pilot. In either case this classification by lead
equalization requirements using the compensatory pilot model identifies
those flying tasks whenever additional visual information or motion is
potentially useful - particularly in the first two phases of the SOP
sequence.

Depending upon the task and the skill level of interest, there may be
various solutions to this step. For example it may be found that the
addition of an intermediate visual velocity feedback loop is all that is
necessary to provide adequate pilot-vehicle damping for the execution of a
task at the compensatory level. Or it may be necessary to utilize ves-
tibular feedback in order to boost the task execution to a pursuit
level. Unfortunately we presently lack adequate quantification of tasks
and piloting technique versus skill level to perform this step with con-
fidence, but it is a crucial step to discovery of simulator fidelity needs
and therefore will be addressed under research objectives.

The purpose of this step is to elaborate the compensatory loop struc-

f tures with additional parallel paths responsive to sensed visual rate
information or sensed angular and linear accelerations at the pilot loca-
tion in the aircraft. Current knowledge suggests that equalization
possibilities in the vestibular channels are restricted to the known dy-
namics of these revised pathways coupled with an adjustable gain. The
appropriate choice is that gain which allows the lead in the parallel
visual channel to be eliminated. The resultant loop structure is known to
be capable of somewhat greater crossover frequencies for the system as a
whole should the nature of the regulatory task demand it.

The evidence for such structure is most substantial for the angular
acceleration motion inputs, somewhat less so for linear accelerations.
This is primarily because control of an aircraft's angular motions is
fundamental to virtually any flying task and considerable experimental
data have been gathered. On the other hand, the pilot's use of proprio-
ceptively perceived linear accelerations appears to be quite task-specific
(e.g., flare of the aircraft prior to touchdown). The pilot's awareness

145
of linear accelerations (and by tentative inference, his use of these
sensations) depends in part on his past experience. A background of heli-
copter flying generally implies a greater awareness of linear
accelerations than does the background of the fixed-wing pilot.

For Each Essential Loop, Look up the Cataloged

Fidelity Potential for a Specific Candidate
Medium and Determine Whether This Potential
Permits the Required Loop Closures (Step 5)

In this step we, compare what we have determined analytically to be

necessary with what a given device is capable of. In effect it is the
matchup between essential loops required and essential cues available.
This presumes the existence of a "cue catalog" for motion and visual de-
vices - a source -Whi&.ch&does not really exist exet in fragmentary form.

Nevertheless if there were a strict accounting of available cue in-

formation it would be possible to predict the fidelity potential of a
given device in terms of the fidelity parameters listed earlier
(threshold, quickness, distortion, and signal-to-noise ratio). Knowledge
about the fidelity potential of devIces-in-being, in turn, will make it
possible to set requirements for future devices.
:'A

OBSERVATIONS ON FIDELITY ANMLYSIS

The experimental work of Ringland, et al. ,67 provides a number of

lessons. The experiment consisted of- an extremely difficult VTOL hovering
task using panel instruments wherein the presence or absence of simulator
motion was an experimental variable. That the task was difficult is

167 Ringland, Stapleford, and Magdalano, 1971, op cit.

146
 attested to by the performance variability among the several experimental
conditions. Normally system performance remains approximately fixed and
pilot opinion (interpreted as subjective workload) changes. One result,
surprisirg at first, was that the elimination of linear motion improved
performance relative to the case where both linear and angular motion
K (presumably most "representative" of flight) were present. The elimina-
tion of the linear motion resulted in unrealistic tilting sensations which
nevertheless helped the pilot control the simulated vehicle, thus
constituting what training psychologists term an "irrelevant cue to
correct performance." (Sometimes an irrelevant cue to correct performance
is a good thing if it doesn't become a crutch.) Tight attitude control
was required by the nature of the task for system stability, and, without-
these sensations, attitude information was available only visually and
more difficult to obtain because of other scanning demands.

The elimination of linear motion in this experiment resulted in a

device quite like the World War II vintage Link trainers used for instru-
ment flight traLning. P~lots trained %tth this device commented that the
real aircraft felt less sensitive or more sluggish relative to the
Strainer. Pi~ots back for a "refresher" after extended flight duty

typically felt the trainer to be too sensitive. The false, attitude-

proportional tilting sensation in the trainer provided additional
information not available in the actual aircraft. Subjectively the
S~trainer was more sensitive in its angular motion responses to the controls
because an additional source of attitude information was made available to

the pilot.

Another major factor in the experimental design was the elimination

of motion 'washouts," those intentional motion distortions introduced to
maintain the simulator's linear travel within the bounds of the simula-
tor. Instead the magn'itude of both the linear and angular motions were
reduced relative to the real world, to keep linear motions within simu-
lator limits, resulting in accelerations at or below the known thresholds

170 Ibid.

1.47

• ....
 .- . .

of the pilot's vestibular senses for much of the time. Thus the perfor-
mance differences due to motion were considerably attenuated.

These results are quite important for motion simulation. The princi-
pal means of reducing the linear travel requirements in motion simulators
is the use of one or more variants on the "residual tilt" scheme, a
tilting of the simulator cab at sub-vestibular threshold rates to produce
linear acceleration sensations while avoiding the linear travel otherwise
required. This technique amounts to an intentional distortion of the
motion sensations. Whether or not it is acceptable for training in a
particular task depends upon the magnitude of the distortion relative to

vestibular thresholds. If the distortion is comparable to or higher than

the thresholds, analysis following the lines laid down above can determine
if the unrealistic cue will influence the pilot equalization adopted. If
it does, the training effectiveness is compromised because the pilot
equalization differs from that adapted in the real world task.

Motion distortions greater than vestibular thresholds and counter to

visual information have the potential of disorienting the trainee to the
point where training effectiveness is compromised. This comes about
because the pilot has two conflicting sources of information regarding
which way is up, or how far he is tilted. His eyes tell him one thing but
his vestibular senses tell him another - even to the point where the
polarity of the angular rates perceived by the two senses disagrees!

Avoidance of perceptual conflict leads to requirements for large simu-

lator cab linear motion travel. In general there will be amplitudes of
motion for a given simulator above which unacceptable perceptual conflict
and improper pilot adaptation results. This is particularly true for the
pilot who is not simulator-wise and therefore not aware of such con-
flicts. The research'simulator pilot does not expect realistic motion and
is not as deeply affected when he does not get it.

Another perceptual conflict which can be introduced by the simulator

is due to time discrepancies between the onset of motion perceived via
vestibular senses and that perceived visually. Delays between control
inputs and motion responses are a separate problem, even in the actual

148

LLI
K aircraft!
greater than 0.1
The pilot will not be able to tolerate time discrepancies much
sec without complaint. This is
tasks demanding maximal performance on the part of the pilot, e.g.,
particularly true for
target
tracking, where the crossover frequencies to be attained are relatively
high.

At this point it is reasonable to argue that even distorted motion may

I be of some benefit, provided that the simulator training does not proceed
too far. The simulator training will become non-productive when the
trainee begins to accept the distorted motion as normal, and to develop
precognitive levels of skill on the simulated task. Past this point the
trainee will have to unlearn the simulator to learn the airplane.

to establish when this will occur. Here the metho-

The problem is
dology outlined above cannot help us because the model does not embody the
dynamics of the SOP sequence, i.e., the psychomotor skill acquisition
process. This deficiency is a major one, for without being able to pre-
dict the speed of learning for a given flying task as represented on a
simulator with given motion fidelity limitations, one cannot establish, a
priori, its training effectiveness. Considerable analyses of well
documented simulator-to-aircraft transfer of training experiments
involving flying tasks using moving base simulators are required in this
area.

The jet transport flare and landing study cited previously 1 7 1 provides
us with a relevant case illustrating some aspects of the foregoing proce-
dure as they relate to visual fidelity. The model of the landing task, as
determined from flight measurement, indicated a level of closed-loop damp-
ing which was more than a simple pure gain feedback of altitude (the
command loop) alone could provide. A feedback of vertical velocity or its
kinematic equivalent was also required. The large amplitude vertical
motion simulator involved in this study did not induce the required level
of closed-loop damping. This analysis was a good example of failure to

171 Heffley, Schulman, Randle, and Clement, 1981, op cit.

149
i! i
reject the hypnthesis that the simulator differs from the aircraft in some
property essential to the flare and lending. He~ice there was an obvious
question of simulator fidelity. What essentidl cue was either lacking or
incorrect?

Analysis along the lines suggested in Table 12 presented some specific

possibilities for the apparent simulator deficiencies. Candidate explana-
tions included:

0 Excessive threshold for visual vertical velocity

infof~mation or its kinematif2 equivalent, vertical
flight path angle information

* Lach of pilot lead Gr anticipation of visual altitude

* Deficient motion (if motion can be used to enhance

closed-loop dampingi-

Exceosive sivulator
E delays or lags (visual update,
algorithmic, etc.,

SIncorrect aerodynamic heave damping (slope ) l

!it
coefficient versus angle of attack)

0 Combinations of the above

This list amounts to fidelity factors which could be checked in order to 4i

discover the deficient fidelity characteristics. Or, from another view-

point, this same list constitutes the features which should be measured
and cataloged for the particular simulator components used in order to
have predicted their usefulness for the landing task.

1722
172 For example, the focus of expansion (of streamers) itself may be
diffuse, the reduced field of view may inhibit extrapolatiou of
peripheral streamers to define the focus of exnansion, or perspective
distortion may contribute to misjudgment of the direction of flight.

150
 I
TABLE 12

~CON~TROL THEORY APPLIED 'O THE LANDING M4ANEUVER

il A LANDING CRITERION:

A "good" landing consists of a significant reduction in sink rate

executed suificiently fast to counter disturbances.

Control implications are therefore

(1) Adequate "closed loop damping ratio" or 'phase margin" or

(e.g., damping ratio > 0.7)

(2) Highest possible "natural frequency" or "bandwidth" or

2. "crossover frequency" (e.g. natural frequency a 0.4 fad Sec)

RAW DATA: h FLAt

REDUCED DATA: A

h(t) matched to second-order differential equation, i.e.,

d2h
for •d +A ý-+ B h - 0
dt
2 dt
"A
andB solved for best fit to raw data.

I •. •, PILOT-VERICLE INFLUENCES, ,

There are components of pilot and aircratt in both A and B. The

aircraft portions are easy to identify:

For A the aircraft component consists of "heave damping," a

1
predTctable quantity which can be estimated or measured
directly.

ii For B the aircraft

negligible quantity component consists
compared to of "ground effect,"
pilot actions.

After subtracting aircraft effects from A and B, the remainder

represents the piloting technique component. Rre specifically,
1
A Aa/c + Apilot

(Negligible)
B -B /+ Bpilot
a/c
In the case of the landing maneuver

The above criterion forces certain numerical values for A and B,

S~~and these are confirmed from the flight data. - -

Also we can characterize fidelity in terms of an equivalence

between flight and simulator, i.e.,

Good fidelity
•!Bflt pilot
_S Basn
pilot
Godfieit
in terms of
height --
o attitude
~ aim Good fidelity in terms of
Aflt Apim direction of flight jor sink
pilot
•= pilot rate) -. attitude

But the actual data show that ASim is nearly zero!

pilot

Therefore there is a deficiency in the direction-of-flight

loop, an essential loop for this aircraft.

151
 173
The results of the landing study can be extrapolated to other
height control tasks, piloting techniques, and aircraft types if we pre-
serve several constraints. These include basic control theory, the
physical laws of vehicle dynamics, and the perceptual pathways implied in
the landing situattmn, i.e., height and direction of flight. Figure 37
shows the results of such an extrapolation for ranges of Army aircraft,
both rotary- and fixed-wing, over their respective speed ranges and con-
trol techniques. Thn boundaries shown indicate the amount of height
.174 above4 which direction of flight cues are needed.
response lag1 The
Sconclusions from this plot would be that direction-of-flight cues are
required for low-speed helicopter flight (below translational lift) and
for low-speed fixed-wing aircraft with relatively high-wing loadings. On
the other hand, only altitude cues would be required for fixed-wing air-
craft with low-wing loading. Further the boundaries shown could be
adapted to any simulator for which direction-of-flight feedbacks were
required yet not exhibited by the pilot.

S173 effley, Schulman, Randle, and Clement, 1981, op cit.

174 Height response lag is primarily due to "heave damping', a

characteristic approximated by the dimensional stability derivative,
Zw. (See McRuer, et a!., 1973, op cit. for a definition'of this
parameter.) In fact the characteristic time lag due to heave damping
equals -1/Z2. Additional lag can also be added if there is
applicable control lag such as the time required to change attitude.

152
 TASK: PRECISE CONTROL. OF HEIGHT

LAG
IN
HEIGHT
RESPONSEY
(SEC)

2'2
DIRECTION-OF-FLIGHT CIJ'ES NEEDED FOR
:% EIGHT CONTROL VIA SQ.jk~TIVE

OH-6

DIRECTION-OF-LHT US
A4 NEEDED FOR HEIGHT CONTROL
VIA PITCH ATTITUDEi
ROTARY-
WING
RANGE OV-1 (high wing loading)

44
0-2(o iglaig
00 100 200
AIRSPEED QrT)

Figure 5.The Need for Altitucne Rate Cues Depending! Upon

Aircraft Type and Piloting Technique i
153
 -!",-

DOOKKEEPING METHODS FOR SIM4LATOR FIDELITY

Our treacment of simulator fidelity thus far has prepared us for the
next step - to establish a system~atic bookkeeping scheme for simulator
fidelity with respect to the array of Army training objectives. This step
will take full advantage of our definition of fidelity, our task analysis,

the modeling approaches for pilot behavior and perception, and the an-
alytic approach for cue stimulus needed to train.

SSmulator fidelity bookkeeping is out of necessity a multidtihensional

procedure because of the number of factors which must be observed, i.e.,
task, aircraft type, level of training, and environment. Our ultimate
goal is a tradeoff matrix of visual and motion fidelity characteristics
needed to train, but this is highly conditional because of the above fac-
tors. Therefore it is necessary to introduce intermediate Lables and

plots as part of oar bookkeeping scheme.

One list cf useful matrices is suggested in Table 13, and we shall

explain each in detail shortly. The value of arranging information in
these forms is that it compartmentalizes the data and maximizes its use.
Rather than acquiring basic data for each task, level of training, and
aircraft type, we have th. possibility of finding commonality among both

training objectives and training device requirements. Another advantage

is that we have, in this kind of list, a ready-made shopping list for the
gathering of missing data.

Note that the matrices listed in Table 13 are composed of

0 Basic information

0 Constructed information

Basic information must be acquired from past research and analysis and

from a good deal of future work. Constructed information is based upon

analytic procedure:: such as described in the previous section. Now let us

o 154

£1
 TABLE 13

A LIST OF BOOKKEEPING FORMS FOR SIMULATOR FIDELITY

* Pilotiug technique versus task (basic information)

* Cues available versus training device or simulator (basic

information)

0 Aerodynamic feedbacks versus aircraft type (basic

information)

* Essential loops versus piloting task and technique

(constructed information)

* Essential loops versus skill level for a Siven task

(constructed information)

& Essential cues available versu3 essential loops required for

a given aircraft, task, and triining device (constructed
• inf ormation)

* Training capability by task versus training device

(constructed information)

0 Macro-detail cues available versus micro-detail features

required (Basic and constructed Information but not rendily
obtainable at this time)

155

155

I . . .. .j
consider each of the entries in Table 13 and suggest examples of their
form and content.

PILOTING TECHNIQUE VERSUS PILOT TASK

This is a logical starting point - consiceration of how pilots oper- K

ate an aircraft in order to execute a given task or perform a mission. We
have a qualitative description in the training literature, in syllabus
material, and from pilot commentary. As stated previously there is little
quantification of piloting technique by task in terms which are useful for
analysis of fidelity requirements. The form of quantification would most
likely consist of a feedback control law formulation in terms of loop
gains or, more generally, in terms of loop bandwidths and closed-loop
damping (or phase margins). Examples were presented in the cataloged
Figures 13 through 22 at the end of the section on Army Missions. It is
essential that, for each pilot task to be addressed on a training
simulator, there be a well established quantitative description of that
task along with the necessary piloting technique required to execute it.

CUES AVAILANE VERSUS TRAINING DEVICE

This matrix deals with the perceptual fidelity potential of a specific

device. The objective in constructing such a matrix would be to take
stock of the special cue availability in
tems. Table 14 illustrates
exicting motion and visual sys-
how the matrix might be formulated.
angular
Cues P
ticuld be expressed in terms of convenient scate variables (e.g.,
Ond translational positions and rates for visual; angular and transla-
tional rates and specific farces for motion). The quantification of those
cues would consist of the iidelity parameters such as threshold, response,

distortion, and signal-to-noise ratio.

156
 0
041

0
041

4E-

CD
0 ba

1-40

0o0
~~Cl C)) C) C -4 4 .

E- C C C

0 CC) C)co V) CC
Q.~CC 44 4. . -4- 1-3 ý4- C

Co~0 . 0 0

0 0 CJ V) cn ,-4 I.T V4J -4 C) .-4

4-3.1 u- mC ca1

Q) 0 -- 4C 4

0) Q)
-
4r4
~CCAi1) > U4 co
- > 4- ,* w-

uo -4 44 -HW 14 4J
'I4-1 ca CC 41 4

w-41-J .1 CC

v4~ ý ) 01 to
00 .4 H-.- 44 41 4-IS-

rC rC. , 0CCU0 0 C: (1
0. v-i410 0. 00 '4

H -T3
C-Z

F44 157C 4
cc e ) 1 $

:3 >
to c-v4
 AERODYNAMIC FEEDBACKS VERSUS AIRCRAFT TYPE

f While

pilot feedback,
most command loops

a number
depend almost entirely

of supporting loops depend both upon pilot

upon the degree of

and

aerodynamic feedbacks. The nature of those aerodynamic loops, axis by

axis, are described in Table 15. Note that the table identifies specific
aircraft stability derivatives which can be evaluated for any given air-

craft and analyzed to find the net contribution of each derivative to the
task in question. To an extent, this kind of analysis has already been

performed for some Army helicopter1 Similar procedures could be per-

formed for the current inventory of Army fixed-wing aircraft.

ESSENTIAL LOOPS VERSUS PILOTING TASK AND TE(IN1QUE

Refer to Table 6, pp. 55, 56, for examples of pilot control techniques

which are more or less independent of the specific Army Flight Training

Objectives listed in Table 5, p. 53. Also refer to the block diagrams at

the end of the section entitled "Analysis of Army Flight Training Object-
ives" for examples of more task-specific piloting techniques and

corresponding essential control loops. In particular, note the three

speed-change maneuvers shown in Figs. 19, 20, and 21. .ch case some-
a-

what different essential loops and correspond! 1 cues are

involved. therefore important to search out n, e

It is -ps within
context.
avery specific task and aircraft
1a
7

173 Heffley, Robert K., Wayne F. Jewell, John M. Lehman, and Richard A.
Van Winkle, A Compilation and Analysis of Heliccpter Handling
Qualities Data. Volume One: Data Compilation, NASA CR 3144, August
1979.

Heffley, Robert K., A Compilation and Analysis of Helicopter Handling

qualities Data. Vclume Two: Data Analysis, NASA CR-3145, August
1979.

158
 TABLE 15

"AERODYNAMIC EFFECTS

t• Control Axis ".tability Primary

Derivatives Stability Derivative
Contributors- R rks

Roll Bank Angle -Ground effect, (Lf(h)) Usually stabilizing but negligible
ix 3u

Wind shear, Lr•.-- h Destabilizing

Poll Rate Roll damping, Stabilizing, a function of wing loading and, for
av
3i helicopters, rotor type

Pitch Pitch Rate Pitch damping, Mq Stabilizing, a function of tail geometry for
1 36 airplane and helicopter at high speed and rotor
type for helicopter at low speed

Heading Yaw Rate _Yaw damping, Nr Stabilizing

Surge Fore/aft Porition ;- Wind shear, X - 8 Destabilizing

Surge Velocity I 2-X Speed damping, Xu Parasite drag, negligible effect for airplane
m au at low speed

Sway Sway Velocity 7- Drift damping, Yv Always negl.gible, function of vertical fin

Heave Altitude Ground effect, Zh

G- Exponentially increases in vicinity of ground

ýu

159hr, 4--
 Table 16 provides another example for landing a fixed-wing light-twin
engine aircraft using attitude to control the direction of flight.

ESSENTIAL LOOPS VERSUS SKILL LEVEL

FOR A GIVEN TASK

One particularly crucial step in the proposed bookkeeping procedure is

to document skill level in terms of essential loop structure features. In
general we would expect to see a gradual tightening of control loops with
increasing pilot proficiency. This would be reflected in any quantifica-
tion of the overall closed-loop task as well as in piloting technique
gains. In addition there may be development of important control cross-

I174
feeds from one axis of control to another.
indicator of pursuit-level piloting technique.)

Heffley,
development
et al.,
in
and Jewell 175
the Navy carrier landing task in
(In fact this is

each describe an example of skill

typically an

terms of the development

of a pursuit-level technique for controlling flight path and speed. It is
pointed 'out that the use of a control crossfeed (or feedforward) from
throttle to pitch attitude can result in a dramatic improvement in flight
path bandwidth along with a reduction in pilot workload. Therefore the
development of this feature would represent an important objective in
skill development.
The reader should refer to Table 6, pp. 55, 56, for examples of
compensatory and pursuit skill level techniques which are more or less

cK
174 Heffley, Robert K., Warren F. Clement, and Samuel J. Craig, "Training
Aircraft Design Considerations Based on the Successive Organization
of Perception in Manual Control," Sixteenth Annual Conference on
Manual Control, MIT, ay 5-7, 1980, pp. 119-127.

175 Jewell, Wayne F. Identification of Multiloop Pilot Describing

Functions Obtained from Simulated Approaches to an Aircraft Carrier,
Systems Technology, Inc., Paper No. 289, presented at the 17th Annual
Conference on Manual Control, UCLA, June 16-19, 1981.

160
"_ it
 U!!
"U,TABLE 16

EXAMPLE OF PILOTING TECHNIQUE VERSUS ESSENTIAL LOOPS

i "Task Essential Loops Indifference

Threshold

Landing Altitude 2 ft
(fixed-wing,
light-twin) Direction of Flight (Vertical) I deg @ 0.25 rad/sec
using attitude
to control Drift 4 ft
flight path
Direction of flight (lateral) ?

Pitch Attitude 0.5 deg

1L

"'7-v

,p - r

161. _..,.•• • ._ . .• i ••.,•--. . . ._ ,_ ,. _.•

independent of the Specific Army Flight Training Objectives listed in
Table 5, p. 53. Also, refer to the block diagrams at the end of the
section entitled "Analysis of Army Flight Training Objectiyes" for more
specific examples of essential loops versus skill le':el for a given task.

ESSENTIAL CUES AVAILABLE VERSUS

ESSENTIAL LOOPS REQUIRED

The next set of features which should be cataloged are the various
essential cues which might be candidates for each essential loop in a
given task scenario. This information would be valuable in analyzing any
deficiencies in essential loop behavior. Table 17 presents some examples
for direction-of-flight and attitude.

Evidence for deficiencies in the simulation of essential cues has been

found in the following instances:

0 Visual height cues from ground texture (or lack there-

of)

• Visual direction-of-flight cues (six hypotheses are

given on p. 150 and in Fig. 35, p. 135)
* Visual range-to-go to a specific visual element
(pp. 187-191)

0 Visual image of rotor tip path plane

0 Delays in both visual and motion cues (p. 40-41)

* Distortion in motion cues (examples are given on

pp. 105-112, 146-149, and 194-206)

162
 TABLE 17

EXAMPLE OF ESSENTIAL CUES AVAILABLE VERSUS ESSENTIAL LOOPS REQUIRED

Essential Cue Candidates

Es sential Loops_______________

Visual Motion

Direction of 1) Streamer origin 1) ýuasi-integration of

Flight (vertical axis) -vertical specific force

2) Derived rate of
change of altitude

3) Vertical velocity

Attitude 1) Aircraft reference 1) Quasi-integration of

with respect to angular rate
horizontal (or
other earth
reference)

163

T 77>-
 TRAINING CAPABILITY BY TASK
VERSUS TRAINING DEVICE

An earlier matrix discussed was the cataloging of available cues for

extant training devices. It would also be useful similarly to catalug the
known training capabilities, by specific task, of existing training simu-
lator systems.

One known example of this kind of data is the airline landing maneuver
analysis previous described177.

MACRO DETAIL CUES AVAILABLE VERSUS

MICRO DETAIL FEATU-RES REQUIRED

One important kind of bookkeeping information for simulator fidelity

is the indexing *of "macro detail" cues in terms of "micro detail" fea-
tures. "Macro detail" refers to certain visual or motion cues in a gen-
eral sense; for example, the visual presentation of a height cue without
regard to which geometric features combine to provide height informa-
tion. The "micro detail" would be those individual features contributing
to the overall height cue- angles, texture, etc.

This is a particularly difficult issue to address because of the re-

dundancy in pattern recognition and the possible variation from one
individual to the next in extracting information from a given presenta-
tion. The work of researchers such as Wewerinke 1 7 8 represents attempts to
quantify micro detail information content using optimal estimation
methods. Such procedures may provide useful starting points, but the

Heffley, Schulman, Randle, and Clement, 19811, op cit.

178 Wewerinke, 1978, op cit.

I- 164
Ii________________
results must be validated or refined on the basis of experimental data
involving pilot subjects performing realistic flight tasks.
This matter is discussed in several places in this report. For ex-

ample motion cues in general are discussed on pp. 105-112. Motion cues in
support of lateral tracking are discussed on pp. 194-206, Visual range
perception is discussed on pp. 187-191. Hypotheses for deficiencies in
visual direction-of-flight cues are discussed on p. 150 and in Fig. 37,
1 9
p. 153. For a given training mission ' (e.g., helicopter NOE), training
objective (e.g., "attack target"), and training medium (e.g., ground-based
simulation) - it is possible to determine the tradeoffs among various
visual and motion fidelity measures and their training value. Other ex-
amples of this form of tradeoff would be as follows:

A
0 Target tracking range versus visual system resolution:

:& r

Maximn -- 7 Required Resolution vs Targets

Trackin- or vs Distance
Range

Safe Firing Distance

short " Bridges

TanksA
fine coarse
Visual Field Resolution

179 This case corresponds to Aircrew Function D-4.1, "Engagement: We,%'on

-elivery-Attack Targets," of Table 1 "NOE Operations" in Gainer and
Sullivan, 1976, op cit.

165
 0 Allowable sideslip alignment maneuvers versus motion system
lateral travel:
1C'O'f Sidestep Distance Desired

Spurious 2Oft
Motion
Cues .1MEE/ / Indff d'e re nc e
Threshold
Y \ While
Tracking

2,40 60 80 100

La,.-eral Trave! LimTits

v (ft) HS
0 Kinetosis versus field-of-viewl80

ALWAYSCINERAMA
_________(160*)
OFTEN •

CAR Sim
OCCASIONAIM. B ,enkA
UCLA
CAR S I U
OCCASIONALLY TV SANrRE

NEVFZ H ME

030' 60, 90r 12db 15Cr 180'

LATERAL VISUAL ANGLE SLJB1ENDED By SCREEN

Anecdotal Experience with Vertigo versus Screen/Size 1 8 1

When Viewing Driving Scenes From a Fixed-Base Situation

180 Jex, H. R., D. T. McRuer, R. W. Allen, and J. R. Hogge, Requirements

and Program Plan for Simulation Facility and Data Handling for GM
Design Staff Safety and Human Performance Group, Systems Technology,
Inc., Technical Report 2039-1, March 1974.
181 Ibid.

166

5 -
 Except for the kinetosis tradeoff plot, the above sketches are not
based on actual calculations, but have the correct trends and are typical
of numerous tradeoffs involving macro detail cues available and micro
detail features required.

iI

tA

167
 RESEAICH NEEDED TO DEFINE SIMULATOR REQUIREMENTS

it has 'been suggested repeatedly throughout the various sections of

this report that additional research must be directed toward the remaining
quantification of simulator fidelity components. Major gaps still lie in
the pilot perception and piloting technique functions, especially where
the use of outside visual information is involved. One of the major ob-
jectives of this report has been to describe ways of viewing sirulator
fidelity which would lead to the identification of missing data and to the
ultimate acquisition of those data. The organization of the report,
beginning with the definition of fidelity and leading through definition
of Army training missions, pilot modeling techniques, analytic procedures,
and bookkeeping procedures has, in fact, now led us to the point of
outlining topics for additional investigation in Table 18.

The fact that there are presently gaps in the data necessary to define
simulator fidelity requirements does not mean that those data cannot be
acquired. Various means do exist for obtaining the required information
and one of the main goals in this section will be to outline the various
methods for collecting empirical data needed to determine simulator
fidelity requirements. The examples given are based on some of the Army
aviation training missions, although the material presented is never-
theless applicable to many other areas of aircraft operations.

The ideas proposed are aimed at comprehensive quantification of how

the pilot operates the aircraft in the execution of a given mission or
task. The emphasis is placed on accumulating knowledge of pilot control
laws, sensory feedbacks, and decision making rather than on overt task
performance (i.e., precision of flight) or pilot opinion. This is in
keeping with the notion that training can be equated to correct and appro-
priate development of feedback loops needed for each stage of flight. We
hasten to add, however, that task performance and pilot opinion ratings
should also be collected. We believe the more kinds of data obtained, the
better the opportunity to gain insight and understanding.

168
 TABLE 18

TOPICS FOP ADDITIO-.A. -. ESEARC-H:

TRAINING OBJECTIVES REQUIRING FLIGjQUANTIrICATION
OF PILOTING .EC~F1I0JE •-

BASIC FLIGHT TASKS

Fixed Wing and Helicopter

VMC and IMC
Primary Skill Level Through Continuation Training
Slow Flight Through Cruise Speeds

FLIGHT TASKS S-HOVERING

Helicopter
I! VMC and IMC
Primary Through Continuation Training-

I TAKEOFF/APPROACH/LANDING

{
• Fixed Wing and Helicopters
VMC
Primary Through Continuation Training

1 83
LOW ALTITUDE OPERATIONS

Helicopter
VMC Day and Night

1
WEAPON DELIVERY 84

{ •Helicopter
-- VMC Day and Night

"EMERGENCIES

Fixed Wing and Helicopter

182

S182 See Table 5 for further breakdown into specific piloting tasks.
183 High priority.

184 High priority.

169
?I ..

.-- <<. --- .

 The earlier section which addressed the definition of sioulator fi-
delity described the key background concepts rnd definitions of terms of
vhich we make use in this report. Perhaps most important is the distinc-
tion between implicit fidelity and explicit fidelity, and it is the latter
which has been the more elusive to obtain or to quantify. But there are
methods - some old, some relatively new - which permit direct and mean-
ingful quantification of explicit fidelity aspects. Simply stated, these
are measurement methods which tell us how a pilot flies an airplane or
helicopter - the methods address the mechanisms of piloting behavior
rather than just the symptoms or results.

Much of the collection of empirical data suggested here is more easily

done on simulators, but many of the methods can effectively be made in

actual floghta- sometnmes even using uninstrumented aircraft. We shall

indicate where this is feasible.

The organization of this section consists of four main topics:

* Preparatory analysis of piloting tasks

* Measurement tools

* Measurement of piloting technique and pilot perception

* Examples of piloting technique and perceptual measure-

ments and interpretation of experimental results.

The discussion of these topics is followed by a summary.

170

_ _ _ _ __ _ _
PIEPARAMTOX ANALYSIS OF THE PILOTING TAS

Our overall approach to determining simulator fidelity requirements

hinges on first, fully understanding the individual task being performed
or trained. It is clear that the relative importance and quantitative
nature of the various motion and visual stimuli can vary vi-.ely. The
pilot measurement methods or procedures have to be custom-suited to -each
individual application.

A formal approach to analyzing piloting tasks has been covered in the

section describing the analytic procedure for determining necessary visual
and motion cue stimuli; however, we shall summari:e that approach for the
purpose of discussing methods of collecting data on pilot behavior. The
first step in task analysis should be to review wrltteni descriptions of
the nominal piloting task. For Army training mis~ions, the most sysrema-
tic description of piloting tasks is given in the aircrew training manuals- V
185
for various fixed- and rotary-wing aircraft types1. -These manuals pro-
vide the basis for sketching a nominal loop structure which may involve a
number of secondary piloting tasks as well as the primary one. An example
was shown in Fig. 8 for the helicopter approach to hover piloting task.
The primary feature to be noted in this kind of a task description is the
approximately parallel structure for the loops involved in each of the H
three major axes: axial, lateral, and vertical. Within each axis is a
-nested series loop structure. The outermost loop in each of the series

185 Anon., ATM, Utility Helicopter, 1978, op cit.

Anon., ATM, Attack Helicopter, 1978, op cit.

Anon., ATM, Observation Helicopter, 1978, op cit.

Anon., ATM, Cargo Helicopter, 1978, op cit.

Anon., ATM, Surveillance Aircraft, 1978, op cit.

Anon., ATM, Utility Airplane, 1978, op cit.

171
structures relates to the primary objective in that axis, and the inner
series loops act to support that primary objective. For example, in the
lateral axis, the primary objective is to maintain a track along the ap-
proach centerline; and this -is reflected in the outermost loop around
lateral position. In direct support of this is regulation of heading and,
in direct support of heading, Ps regulation of bank angle. A similar
organization can be observed in other axes and in other piloting' tasks.
Systematic recognition of such structure is thus the first step in analy-
sis of any given piloting task or mission.

The second step in our task aaalysis is to estimate the likely ranges
of aircraft dynamics or kinematic relationships involved in the task.
This is not inordinately difficult, since most fixed-wing or rotary-wing
aircraft of a given size and type tend to have very similar dynamic prop-
erties, especially if these are aircraft in an operational status and not
unusual research or experimental aircraft. For example, in an analysis of
the typical properties of single-rotor helicopters by Heffley1 8 6 , it is
shown that the major variation in dynamic features among various helico'p-
ters is primarily in attitude control characteristics. However, even this
aspect is highly predictable with knowledge of the basic rotor type (tee-
tering, articulated, or rigid). The main point to be made is that an
estimation of likely ranges of vehicle dynamics is not unduly tied to any
particular helicopter model. For the purposes of a preparatory analysis,
the differences bet-'een, cay, an OH-6 and a UH-l are not great.

The third step in our task analysis is to estimate the likely ranges
of bandwidth for each of the loops involved in the task. This can be
accomplished on the basis of relatively simple rules of manual, multiloop
control theory in combination Vith available rules of thumb or stated
performance objectives. Ir fact, a dependence on manual control theory
ideas is likely to be reletively low. A very good estimation of loop
bandwidths is possible by an engineering interpretation of nominal rules

186 Beffley, A Compilation of Helicopter Handling Qualities Data.

Volume Two: Data Analysis, 1979, op cit.

172 f
rV

fr
of thumb for each pil6ting task. Also bandwidths, once established for
-I one loop, are likely to be applicable to the same loop in other tasks. As
with the other steps in the preparatory analysis, it is really, only neces-
sary to obtain "ballpark". estimates for the various loop bandwidths.

•4 Another dimension of the task loop structure which might be important

to measure, and therefore estimate a priori, is the level -of successive
organization of perception (SOP). Simply stated, this refers to the de-
gree that the pilot is only reacting or whether there is significant
anticipation
[[ of results And prediction of vehle response. For example
turn -coordination can be handled on a "compensatory" level with rudder
applied only in response to miscoordination cues - "ball out of center"
or pilot perception of lateral g's. On the other hand the pilot can apply
rudder knowing "that's about what it takes" to coordinate- the turn. This
is anticipatory and based on learning how the airplane responds. Such
pilot actions might b( considered as "pursuit" or even- "precognitive"
-strategy - higher levels of -SOP. Some of the features of SOP are sum-
&. : marized in Table 19; and these are perhaps very important targets for
measurement.

Let us now consider how we might approach our preparatcry planning

phase for a particular Army training task, the unmask/remask maneuver.
First, we would take an overview in order to see- how the task interfaces
with other tasks. This may be Important in terms of defining fully enough
the overall task loop structure. Often any single task is really just
-• , constituting an inner loop for a subsequent exterior loop-another more 4
all-encompassing tasK. s-shown in Table 20, the unmask/remask task shows
up in various forms of engagement in the attack mission.- Wa would also -

want to consider other missions (e.g., utility or scout). 1

Next, focusing on the unmask/remask task itself, we would try to de-
velop th'e -kind. of -scenario shown in- Table 21. Note that this is really-
"* just a verbalization of 'each- subtask, - things -which we could otherwise
express -in terms of a closed loop block diagram as shown earlier in
"Fig. 37. The immediate and direct benefit of this kind of exercise is the
identification of -specific measurements such as how tight is regulation of
position (fore/aft, vertical, and lateral), what is the desired pop-up
l

173

•lz .- i~i~~ ..
 Increasing perceptual
motor load

Increasing rehearsal

HI Increasing bandwidth

. Increasing time delay

FExhpanding perceptual field

content end extension thereof

::72 0 1

•tW • o40 CS 4:
Q) rq2

I• a 0 Ell
40

Er4
0 0'

to 04
0 0
0)0 )
vt, a
" U
a)
V z0~0
FOS
Z,
430)-
(1) 1C

0 0 lod
0) 0 a)
0 0) 0D

"0'-4iO Ow-'P
4-1
CO

W 0 a) 14 4CPJCJ0

E- t 0 I I,

E-4~
 VH
I
pr

TABLE 20

F! TASK BREAKDOWN

(Summarized by Gainer and Sullivan1 8 7 )

Attack Mission Phase:

ii Enroute - Cruise NOE

Airspeed
Altitude
Heading
-Maintain mask
Maintain/monitor obstacle clearance
Determine position/performance intersection

-Engagement -Maneuver

iManeuver into pygiattack position-

* Unmask .(pop-up)-
* Remask
Perform evade drop-report
Perform evade dash

Engagement - Pre-At tack

Hover instrument check

0 Unmask
Target acquisition
* Remask
Engagement - Weapons Delivery

J,
Attack
fire SHovex
-Running fire
. Mask

Return to Base- Depart Maneuver Area

187 Gainer and, Sullivan, 1976, op cit.

| 188 Appearance of unmask/remask maneuv'er indicated by bullets.

S"• 17 5
 TABLE 21

TASK ANALYSIS OF UNMASK/REMASK MANEUVER

Implication on Test Plan, Measurements.

Discussion of Piloting Subtask and Data Analysis

Control of Fore and Aft Position:

As the pilot ascends toward the top of the xask, The task scenario shou•ld Involve a terrain or vegetation
whether a stand of trees or steeply sloping terrain, feature which induces forward translation during vertical
there is a tendency to translate forward. Hence there ascent and unmasking, e.g., a pyramidal shaped hill, bar-
is an absolute requirement to translate rearward by an rner, or tree. In order to gauge position relative to the
appropriate amount during the descent. Also, the terrain, a tip-path plane edge should be providei in the
pilot is likely to gauge position and ve:tical .path of outside visual scene. Direct measurements should be applied
the helicopter using the tip-path plane to help evalu- to control of fore and aft position vhich reveal the a.ount
ate tree-top clearance. of pilot regulation (bandwidth. Phase margin) and coordina-
tion with height (crossfeed of x-positlon comand with
vertical position).
Control of Vertica•l Position:

The pilot needs the atility to make a crisp bob-up Direct measurements should be applied to cou.mand and control
a bove the mask. hold altitude precisely for 5 to of vertical position and should indicate bandwidth and level
10 sec while perforn-ing the observation or weapon of compensation while unmasked.
delivery task, and descend smartly belou the aask.

Control of Lateral Pos'tion:

Woring the vertical ascent nd descent there is likely Measurement of lateral command and control should likewise
to be little or no lateral movement. However, follow- be applied but will focus on the time interval following
Ing rems~.ng, the pilot will want to translate remask and descent.
quickly to a new lateral position below the mask.
Regulation Against Atmorpheric Disturb~ances:

AtAospheric disturbances introduce a significant A simple random free-air turbulence model would probably
.ompllcation In the performance of the maneuver, suffice in forcing a position dispersion on the sirframe/
Turbulence can of" course, affect the attitude and control system combination. The major effect on pilot
position regulation task through direct action on the behavior, however, will be the introduction of a determinis-
vehicle. More Insidious is the effect of wind shear tic wind shear at tree-top height. The shear should be
a-, the tree top level or ridrellne. On one hand, a applied in bot.: directions, but it can-be asssumd that the
Stsalvlwnd forces the-helicopter toward the vegetation pilot has prior knowledge of which direction. Position
Sor terrain sask and thus creates a hazard during the bandwidth and compensation should be measured for varying
desacent. Perhaps a more serious condition is a head- amplitudes and directions of disturbances. Position of the
ind. however. During ,he unmasking the pilot might aircraft relative to the obstacle should be plotted for the
-trim Into the headina; but during remasking the vertical plane.
headwind quickly disappearci, forcing the vehicle into
the trees or ground. (A Navy counterpart to this
hazard has been observed for landing on the stern of a
ship.)

176
I

; '- " . . . . . - 2 "- . . .

 !4

trajectory (vertical command), and how much effort is exerted to counter

disturbances (gusts, wind shear, terrain.)

At this point we have developed a shopping list of things to measure

and, just as importantly, the context in which we want to make those mea-
surements. Now let us consider some measurement methods.

"THE ARRAY OF MEASUREMENT TOOLS

There is a vast array of measurement tools which are at the disposal

of the engineer and psychologist. In selecting these tools, the philo-
sophy should be to use as imany methods as is practical in order to gain
the broadest perspective and fullest degree of quantification possible.
Each flight or simulator situation to be studied should be regarded as a

unique target of opportunity which deserves scrutiny from as many points

of view or perspectives as is feasible.

In discVssing the various- kinds of measurement tools, we shall not

restrict ourselves to only simulator situations even though that is nor-
mally the easiest environment in- which to gather data. It is also
possible to obtain useful data from actual flight even without using in-
strumented aircraft, and it is important to note that -data taken in flight
4•. can carry more credibility than simulator -data, even though the flight K
data acquisition media may be substantially inferior in quality and
quantity. It should -also be noted that much of the data required to sat-
isfy our needs need not be overly precise. In some cases, we may require
only rough verification of our pre-experimental or preparatory task
analyses.

A Figure 38 shows an array of measurement analysis tools which are ap-

propriate for many simulator facilities. These tools would be most easily
implemented on -research simulator facilities, but could also be applied to
training simulator facilities. Note that the two sources of data are the
pilot and the simulator digital computer. Also, the kinds of data are
divided into routine versus non-routine (or novel) data.

177

.,- --- . .
 COOPER-HARPER RATING
PILOT
u
ORAL OR WRITTEN COMMENTARY

ON-LINE TIME HISTORIES

END OF RUN STATISTICS

SI
MAGNETIC

TAPES
TA-ES- ---- OFF-LINE REVIEWOR
"DATA ANALYSIS

PARAMETER IDENTIFICATION
SOLUTIONS OF PILOTING
TECHNIQUE

i •X-Y P!OTS PHASE-PLANES , CONTROL-STATE

PORTRAITS, CONTROL-CONTROL
OOTPORTRAITS,..
:> ;
0
0

z
•. VIDEO REPLAY FOR
S , • -PILOT DEBRIEFING
-• z APESOR ENGIN~EERIN~G ANlALYSIS

SFigure 38. Array of Measurement Analysis Tools

178

II
 Pilot-centered analysis tools can consist of the standard Cooper-
189
Harper rating and oral or written pilot commentary. Cooper-Harpe*.
ratings are really not of any particular value unless a pilot is exper-
lenced in giving them, and this usually requires a research or flight test

background, On the other hand, oral or written commentary is always pos-

sible and is relatively valuable so long as the pilot is sufficiently
articulate in describing aspects of the experiment; and even an inarticu-
late pilot will -be able to answer questions posed by the experimenter.
One main limitation of pilot commentary, however, is that it is likely to
be relatively qualitative. The pilot may be able to make reasonable esti-
mates on overall task performance but will probably not be able to provide
details of bandwidth or piloting technique. 190 Naturally the more
engrossed the pilot tends to be in a task, the less likely -will be the
ability for objective self-analysis. Nevertheless, pilot commentary has
the potential for providing increased insight when combined with other,
more quantitative and direct measurements.

Other routine data sources include gathering of on-line time histories

of state variables of interest and the summaries of those time histories
in terms .of end-of-run statistics. It is the latter of these which is
most frequently obtained from simulator experiments. The kinds of statis-
tics normally gathered include means and standard deviations of the
various states of interest, such as: control displacement, attitudes,
heading, airspeed, altitude, and any special display features which are
actively tracked such as glide slope, localizer, or flight director.
Unfortunately, these kinds of raw performance statistics tend to be rela-
tively invariant over a wide range of environmental conditions or from one
pilot to the next. This is likely due to the fact that most standards of
performance do not vary regardless of the situation. For example, even
though gusty air may tend to upset -or disturb aircraft attitudes, the

189 Cooper, George E., and Robert P. Harper, The Use of Pilot Rating in
the Evaluation of Aircraft Handling Qualities, NASA TN D-5153, 1969.

190 tarter and Semple, 1976, op cit.

179 j

f -'-
 -- - •' -. ' 74$

pilot will simply apply more effort to increase attitude bandwidth and
thereby retain the same standard of pitch attitude dispersion as was en-
joyed in calm air with relatively little effort. Thus the fact that
attitude dispersions are about the same in rough air as in calm air im-
plies that the pilot has substantially modified his behavior and is
working harder, but these dispersion statistics do not provide a direct
measure of that crucial difference. However there are ways to make that -
important direct measurement using additional statistical measures or raw
time history recordings of variables.

A relatively simple way to make bandwidth measurements for a given

variable, which is controlled by the pilot, is to divide the standard de-
viations of the time rate of change of that variable by the standard
deviation of the variable itself. In other words, the approximate band-
width of pitch attitude control is roughly proportional to the standard
deviation of pitch rate divided by the standard-deviation of pitch at-
titude. This method would show a significant adjustment in piloting
technique between calm air and very rough air even though the standard
deviations of attitude itself were nearly the same between the two
cases. It sho~uld alao be noted that this method can always be used as a
•,, simple, independent crosscheck of loop bandwidth even though a more so-
phisticated measurement technique is also being used.

Time history data are frequently believed to be of only limited

value. They are sometimes used to determine peak excursions or time be-
tween important events, but they are usually not regarded as being
particularly indicative of piloting behavior. There are, however, im--
portant indicators of piloting technique which are easily derived directly r
from time history data. The most prominent indicator is uo;ually the dam-
inant oscillatory mode which is readily visible in a variable which is
known to be regulated by the pilot. Again, taking pitch attitude as an
"example, the dominant mode that is normally apparent in a pitch attitude
time history, is directly related and, in fact, approximately equal to the
pitch attitude bandwidth or crossover frequency. Therefore we need merely
to measure the average period over several cycles of the dominant pitch
.attitude oscillation and divide this average period into 2w in order to

180

'I _--______________ ______

 -17

obtain a crossover frequency in units of radians per second. This same

technique can also be readily applied to bank angle regulation, altitude
regulation, and glide slope and localizer regulation. As with the statis-
tical method described previously, the dominant oscillatory mode method
can be used as either a primary clue -to loop bandwidth or it can be used
as an independent check on a more sophisticated measurement approach.

Another important -source of simulator -data can be magnetic tape re-

cordings of important task variables, which can be stored and reduced at a
later point in time. Magnetic tape data is, o- course, important where
data analysis must be done off-line because of either time or computing
limitations.

We show three possibilities of non-routine or novel data gathering

methods in Fig. 38. The first of these, parameter identification solu-
tions of piloting technique, is the most sophisticated and can involve a
significant amount of off-line data analysis. On the other hand, there
are also relatively simple yet effective procedu 's which can- do a good
job of quantifying specific features of piloting .-- hnique. In general
the degree of sophistication required in the identification algorithm will
depend upon how specifically the piloting technique loop structure can be

characterized. The most notable example of a simple identification scheme

191
is -the "NIPIP" algorithm . It has been applied successfully for single
and multiloop situations. The NIPIP scheme identifies the piloting tech-
nique sufficiently fast to detect significant changes in loop bandwidths
or in control strategy itself. Besides the NIPIP algorithm, other avail-
able schemes include the describing function analyzer and various maximum-
likelihood identification techniques. All are described by Clement, et
19 2
al., and a general schematic form for characterizing the identification
process is shown in Fig. 39.

1[91

191

192 Clement, Heffley,

192*
Jewell and Schulman, 1980, op cit.

Jewell, and McRuer, 1980, op cit.

181
 DISTURBANCE
INPUT

STIMULI HUMAN OPERATOR RESPONSESK

SUBJECT -

MEASUREMENTS/
FROM
HE /HUMAN OPERATOR_
SUBJECT MODEL STRUCTURE{

INTERPRETATION
jITERTIVE IJOF
-IDENTIFICATION~
RESULTS
SEARCH-MTO NUMERICAL
S PROCEDURE DEFINITION
) I___ -OF MODEL
STRUCTURE

LNG SOLUTION-j

Tigur.e 39- The Identificatioi Process

I

-182 {
 Another analysis tool not routinely used iE direct plotting of phase
planes, control-state portraits, and control-control portraits. Thes.f are
graphic representations of the relationships between selected pairs of
state variables in which time is not necessarily represented as a basic
independent paraueter. For example the closed-loop behavior of the pilot
performing a landing flare maneuver has been quantified using a direct
193
plot of sink rate versus altitude and pitch attitude versus altitude
The first of these plots can be interpreted in order to obtain the
effective closed-loop frequency and damping of the landing maneuver and
the second plot shows the specific control law being used by the pilot to
perform this maneuver. This kind of scheme involves minimal instrumen-
tation of either a simulator or aircraft and has, in fact, been used to
analyze the differences which result from training for the landing flare
194
maneuver in a simulator as opposed to an actual aircraft9.

The final and perhaps simplest of the non-routine analysis tools de-
scribed here is provision -for a video replay of a simulated or actual
flight maneuver for the purpose of either debriefing the pilot or for
direct engineering analysis. It has been observed that pilots, when re-
viewing a replay of their own flight, can adopt a more objective point of
view for self-analysis and may even substantially change commentary given
before that replay. This idea could be extended to include replay of the
simulator rather than just. video replay.

MEASUREMENT OF PILOTING TECHNIQUE

AND PILOT PERCEPTION

The direct measurement of piloting behavior, especially psychomotor,

has been a popular area of investigation for more than two decades. Much

193 Heffley, Schulman, Randle, and Clement, 1981, op cit.

194 Ibid.

183

9.
 ~1

of this activity has been founded upon the pilot-in-the-loop models which
have come about through the application of control system theory to the

human operator. Unfortunately investigation into the area has been

somewhat limited in comparison to what is needed for the relatively expan--
sive full-task context involved in pilot training. Review of available
data on piloting tecbnique shows that most investigators have concentrated
on inner-loop tasks, such as regulation of pitch and roll attitude. Far
less has been done in connection with intermediate- or outer-loop tasks.
Some of the reasons for this are that investigation of inner loops reveal
considerably more in terms of pilot compensation roles- and adoption of
higher levels of SOP. Also inner loops tend to involve a correspondingly
higher bandwidth and therefore are easier to measure than the lower ire-

[
qaency outer loops whose char3cteristics can change within a fraction of a
cycle of the predominant outer loop frequency.. (For example, on a final
approach segment the regulation of pitch attitude can involve tens of
cycles without significant change in piloting technique while the landing

flare is a maneuver executed over only one-quarter cycle of 'its

predominant natural frequency.) Table 22 contains representative examples
of available information on piloting technique. While a variety of tas::s
is listed, one would correctly 'infer that the data available are sparse in
terms of aircraft type, flight phase, and types of cues available. On the 4
plus side, however, the range of tasks shown provides a
of departure for a large portion of the tasks which are of interest to
reasonable point
1
Army training missions.
Measurement of pilot perception also has received considerable at-

S* tention but there appesr to be fundamental deficiencies in the perceptual

data available. As described earlier, motion thresholds appear far larger
under realistic task loading than in some laboratory situations 95. - I
Therefore it l necessary to congider making perceptual measurements while H
v pilot is performing an actual flight task which in'olves the modality in
auestion.

195L
Hofmann ane Riedel, 1979, op cit.

184
 TABLE 22

4; ~
REPRESENTATIVE EXAMPLES OF' AVAILABLE DATA

Examples of piloting technique and pilot perception measurements include:

* Attitude and position In hover (simulator - cockpit reference with and without motion references)
196
i
-
0 Height, pitch, and roll attitude regulation in cruise (simulator -. cockpit rteference)197

198
* Glide slope regulation (STOL dircraft simulator ~-cockpit reference)

*Coordinated turn (simulator-- outside visual field and motion reference)1 9

2 00
0Landing flare (STOL aircraft simulator - outside visual and motion references)

0 Glide slope~rind airspeed regul-ttion (jet tran.;port simulator cockpit reference) 20 1

202
* Foroatidn-keepinj In forwiard flight under IFR (helicopter - csckpit reference)

I Acceleratlon/Decele-ration (helicopter flight __ outside reference) 2 0 3

*Landing flare (in flight an 0 ~n jet transport training simulatcr, outside night visual field, cockpit

L
display, motion references)5

visual field, cockpit and motion references)

196i
16Ringland, Stapleford, and Hagdaleno, 1971, op cit.
197 Stapleford, McRuer, and Magdaleno. 1966, op cit.

Stapleford, Craisg, and Tennant, 1969, op-cit.

Hubh,Roger H., Samuel J. Craig. and Irving L. Ashkenas. Identification of -Minimum Acceptable
Characteristics for Manu.il STOL Flight Path Control. Volume III: Detailed Analyses and Tested Vehicle
Characteristics, FAA-RD-75-123, June 1976.
199
200
Jex, Jewell, and Magdaleno, 1979, op cit.
Heffley', Robert- I.,
1
Closed Loop Analysis of Manual Flare and Landing, AIMA Paper 74-834, August- 1974.

201 Reffley-and Jewell, 1979, op cit.

Jewell and Schulman, 1980, op cit.

.Weir, D. H., and Di. T. -McRuer, Pilot LDynamics for Instrumnent Ap1 roach Tasks: PFull-ranel Mtultiloop and
Flight Director Operations NASA CR-2019, May 1972.-
202 Clement, Allen, and Graham, 1971, op cit.
203 Heffley, A Model for Manual DELcelerating Approaches to Hover, 1979, op cit.

204 Heffley, Schulman, Randle, and Clement, 1981, op cit.

205 Hofmann and Riedel, 1979, op cit.

185

_ _ _ _ _ __ _ .f
EIAMPLES OF PILOTING TECHNIQUE AND
PERCEPTUAL MEASUREMENTS AND INTERPRETATION
OF EXPERIMENTAL RESULTS

Vioual Range Perception Throughout

the Decelerating Approach to Hover
in a Helicopter

The manually controlled decelerating approach to a hovering condition

in a helicopter has been described as time-varying maneuver for which
closed-form solutions of the linear differential equation describing the
range-dependent kinematics are not evident 2 0 6 . A slightly altered dif-
ferential equation has been formulated, however, by Heffley0, which
combines the crossover model of the pilot-vehicle combination (which is
described in the section on pilot modeling techniques) with the effects of
visual perception2,8 and yields a simple manual deceleration guidance law
which agrees well with in-flight measurements of the range-dependent
kinematics (Fig. 40), which accompany the pilot's control actions.
Although the visual manual deceleration guidance law is time-varying, it
permits closed-form solutions for speed, acceleration, -and time as
functions of range to the hovering point. One potential use of the
deceleration guidance law, which concerns us here, is as a simulator
validation tool by comparing simulator measurements with in-flight

206 Moen, Gene C., Daniel J.

DiCarlo, and Kenneth R. Yenni, A Parametric
Analysis of Visual Approaches for Helicopters, NASA TN D-8275,
December 1976.
207 Beffleyý A Model for Manual Decelerating Approaches to Hover, 1979,
op cit.
208 Palmer, Everett, and John Petitt, "A Measure of Psychological Realism
on a Visual Simulator," Journal of Aircraft, 14, May 1977,
pp. 421-422.

Gilinsky, "Perceived Size and Distance in Vizual Space," 1951, op

cit.
'i!
186

_ _ _ __ _ _-
 CL) 0 0

r P4 4)40C

00C (\1
01 0 -0 CO
4-) 0J
4) Je It U

o r9
0Cd-

04,
04

4)
.

CN 0
K
-A

I3 r 4i

/d 0

C)a 0

1870
d)~
measurements of the parameters A and k in the deceleration guidance law
((Fi8 . 40) while the helicopter is under visual manual control, In
addit"on the same ideas applied to the deceleration task in Fig. 40 can
alco be extended to vertical and lateral flight path guidance.

The key to describing (and measuring) the fidelity of the visual per-
spective (Fig. 40) is provided in Gilinsky20 9 where the psychological
measurements of apparent range and apparent size of essential cues in the
visual field are related to varl-as metrics of visual perspective. There
it is shown that perceived range, Rp, is related to t-rue range, R, by:

R
p 1+RIX

where the length A is a characteristic measure of perceived range known as

the apparent -distance of vanishing points from the principles of
perspective.

Likewise, perceived size, Sp, is related to objective true size, S0o1

by:

Sp Rp.
._ 1
S R I + R/A

The vslue of k in the- guidance law can be interpreted as the "crossover

frequency of the pilot-vehicle system, which represents the psychomotor
-bandwidth achieved by the pilot in the control task. Values of A and kP
210
identified by Heffley from the decelerating helicopter flight tests by-

209 Gilinsky, "Perceived Size and (Distance in Visual Space," 1951, op

210 Beffley, A Hodel for Manual Decelerating Approaches to Hover, 1979, *

Opcit.

188
 21 1
Moen are given in Fig. 40 for eventual comparison with corresponding
measurements from simulator tests.

Independent out-of-doors field measurements of A were made over twenty

years ago by an entirely different technique using comparative apparent
size judgments of two plain white isosceles triangles in daylight and re-
ported by Gilinsky212. One of the isosceles trianges, called the
"standard," was of constant physical size, but was viewed by the subjects
at ranges varying from 100 to 4000 ft. The physical size of the other
isosceles triangle was adjustable by the subjects, but the triangle re-
mained at a constant range, ro = 100 ft, and 36 deg to the right of the
direct line of sight to thr standard triangle in order to prevent simul-
taneous foveal viewing while the adjustment was being made to match the U
apparent size of the standard. The -experimental site was a fairly level
stretch of grassy terrain and the direct line of sight was parallel to an
inactive airport runway 5000 ft long.
Since the- adjustable triangle is always at range re, its perceived

size will be s P = + sr=A' where b is the adjusted (objective) true Z

size. The constant size triangle is viewed at varying ranges R, therefore
S0
its perceived size will be S I 0 where S is a constant. The
p 1 + RIl S0
subjects were instructed to adjust s so that Sp = S while using fi
binocular vision. The resulting objective size measurements are then
related by

A+ r r R
S A +R 1 AAL 1 ~A
0

211 ~Moen, 1976, op cit.

212 W
-ilinsky, "The Effect of Attitude Upon the Perception of Size," 1955,
op cit.

189

-L ! -,
The length A is thus the subjectively perceived range at which the size
ratio s/S 0 tends to vanish. The mean out-of-doors field value of A ex-
trapolated from the measurements 2 1 3 was 300 ft.

More recently, similar out-of-doors field measurements in daylight

have been repeated and compared with measurements derived from analogous
tests while the same subjects viewed collimated and uncollimated closed-
circtit TV monitors displaying the same out-of-doors tests. The results
for A have been calculated and are listed below based on data from Palmer,
2 1
et al.,

Out-of-doors, daylight 530 ft 4 A 4 680 ft

Collimated TV monitor, daylight 216 ft 4 A < 239 ft

Uncollimated TV monitor, daylight 66 ft 4 A 4 115 ft

These results for A imply that the collimation tended in part to com-
pensate for the distortion of the visual perspective associated with

I I
21

213 Ibid.

214 Palmer, Everett, Tamara Mitchell, and John Pettit, Angular Size-
Estimation: A Measure of -Simulator Visual Fidelity, (unpublished
-paper), No date, airca January 1979.

190

2 7 - '
 tA

direct viewing of the TV monitor 2 1 5 . The range of "out-of-doors" values

for A is approximately the same as the range of values for A estimated
from the helicopter deceleration flight tests in Fig. 41.

Other analogous measurement:s have been derived from tests wherein the
subjects viewed computer-generated imagery (CGI) consisting of calli-
graphic night visual scenes of an airport runway beside which the standard
and variable triangles were alternately presented for comparative judg-
21 6
ment. These results are reported in Palmer and Petitt , also for
collimated and uncollimated viewing; Again the results for A have been
calculated and are listed below based on data from Palmer and Petitt2i7 .

Collimated CGI, night scene 76 ft 4 A < 170ft

Uncollimated CGI, night scene 24 ft 4 A < 70 ft

Since the comparable out-of-doors night scene was not -tested for compari-
son, one is left- to speculate among hypotheses for the much lower ranges

215 See Kibort, Bernard- R., and Fred. J. Drinkwater III, A Flight Study of
Manual Blind Landing Performance Using Closed Circuit Television
Displays, NASA TN D-2252, May 1964, for results of flight tests of
blind landing performance using closed-circuit TV displays with
iconoscope lenses -having different focal lengths. The average error
in touchdown point varied- in linear proportion to the focal length of
the lens. Thus:

a) Angular magnification, as with a telescopic lens,, caused more

undershoots (angular magnification tends to increase A)

b) Duplication of the perspective caused no mean bias in touchdown

error

c) Angular reduction,- as with a wide;angle lens, caused more

overshoots (angular reduction tends to decrease A)
216 Palmer and Petitt, 1977, op cit.

217 ibid.

191 1
I,- -

. - 4- s-. -
for values of A. Again, however, the beneficial contribution of callima-
tion is apparent in increasing the range for A.

To summarize, the apparent distance, A, of vanishing points in the

visual perspective can be estimated from a variety of experimental tests
in flight and in simulators. The values of A so obtained offer a unique
measure of the fidelity of visual perspective for application to the role
of simulated visual devices in training. Our recommended technical ap-
proach for evaluating the psychomotor fidelity of the simulated visual
field also relies on the application of validated mathematical models of
human pilot behavior to determine the interactive- influence of the fol-
lowing attributes on overall simulator system validity:

I. The displayed var-YaboCes and control display associa-

tions required for the task from the likely loops
closed by the pilot to accomplishha given task (i.e.,
instruments used in IFR, visual cues and field of view
requirements in VFR);

2. The dynamic behavior required of the pilot (e.g.,

describing functions), and hence the piloting tech-
niques exhibited -in the given tasks for fixed-base
operations;
3. Effects of certain motions on the pilot dynamic be-
havior including cues likely to be utilized or
ignored; permissible dynamic lags and errors in the
presentation of simulated visual and motion cues to
the pilot;

4. Closed-loop system performance;.

5. Pilot commentary and ratings;

6. Excess manual control capacity, i.e., measures of task

workload or additional workload that could be accom-
plished; preferred combinations of displayed variables
which are compatible with the physical scanning work-
load constraint;

7. Scan patterns (for VFR) inc],-:ing proportions- of time

spent on each fixation within the visual field and
link fractions from fixation to -fixation; and (for
IFR) including -proportions of time spent on each
instrument and link fractions from instrument to
instrument using such tools as the Honeywell
oculometer.

192
Notion Perception in Target Tracking2 1 8

In establishing fidelity requirements for the simulation of cockpit

motion, consideration must be given to the effects of motion cues on:

"* Tracking

"* Failure detection Ij

With regard to tracking performance, it is generally more important to

have the rotational cues than the translational ones. If tracking per-
formance were the sole criterion, the translational motions might even be
eliminated altogether as long as the task did not require a translational

acceleration feedback which had no visual equivalent. Nevertheless one

must be cautious about providing only angular motion cues in a simulator
-(which are potentially useful to the pilot), but which arc not present in

actual flight without providing corresponding specific forces which

accompany translation.

On the other hand, the rotary motions should be faithfully reproduced,

at least over an appropriate frequency range. A reasonable high frequency
limit is 10 rad/sec. This is the approximate bandwidth of the vestibular
sensor and is considerably above any manual-control crossover frequen-

cies. For the low frequency limit, it does not appear necessary to go as
low as the vestibular sensor washout, roughly 0.1 rad/sec. A conservative

218 Stapleford, Peters, and Alex, 1969, op c ot.

Ringland, Stapleford, and Magdaleno, 1971, op cit.

Jex, Henry R., Wayne F. Jewell, and Raymond E. Magdaleno, "Effects of

Various Lateral-Beam-Motion Washouts on Pilot Tracking and Opinion in
the "LAMAR" Simulator,"-Proceedings of the Fifteenth Annual
Conference on Manual Control, AFFDL-TR-79-3134, 1979, pp. 244-266.

193
 lower frequency limit would be 0.5 rad/sec and even 1 rad/sec would be
reasonable.

Tracking requirements are also affected by controlled element dy-

namics. For an easy control task, one requiring little pilot lead
equalization, the effects of motion cues are considerably less than for a
difficult task, one requiring large pilot lead equalization. Fixed-base
results may be completely adequate, although slightly conservative, for a
vehicle with good handling qualities. On the other hand, fixed--base re-

sults for a vehicle with poor handling qualities or a marginally

controllable task will be overly conservative.

The following procedure will be used to estimate motion simulation

requirements for a -specific tracking situation:

0 Define the system- piloting task, vehicle dynamics,

displays, inputs, and disturbances

* Determine potential visual and motion feedbacks for

the task

* Analyze the flight situation using the Multimodality

i• Pilot model and, if necessary, the Multiloop Pilot
Model219T

Reanalyze with a variety of,12 imulator dynamics in- I

cluded (e.g., Jewell, et al., )
* Determine limits of simulator dynamics for acceptable
performance degradation relative to flight.

The second consideration affecting motion simulation fidelity require-

ments is failure detection. If the piloting task includes recovery from

219 McRuer and Krendel, 1974, op cit.

220 Jewell, Wayne-F., Robert L. Stapleford, and Robert Xt. Hef fley, rI
Computed Responses of Several Aircraft to Atmospheric Turbulence and
Discrete Wind Shears, NASA CR-152,181, February 1977, 1
"194
an aircraft, rotorcraft, or system failure, such as an engine or stability
augmentation failure, motion cues can •play an especially important
role. 2 2 1 The motions accompanying a failure can help greatly in the
pilot's timely detection of the failure. This is especially -true if the

visual modality is already heavily loaded with a demanding task. For

example, a hardover elevator due -to a pitch damper failure could be
detected by the -normal acceleration and pitch rate motion cues before

aoticeable effects were displayed on the flight instruments (such -as the
artificial horizon).
At the, present no general -requirements based on failure detection are
available. As a minimum, the motion should be enough to provide an un-
ambiguous clue to the failure. For example, to simulate a hardover yaw
damper malfunction, the simulator should have enough lateral travel so
that -the pilot can clearly separate the lateral acceleration cue accom-
panying the failure from those due to gusts. In many cases failure
detection may put the most stringent- requirements. on translational
motions.

Spurious Motion Cues

Another consideration affecting motion simulation fidelity require-

ments is realism- or false cues. Two specific problems which compromise
the -pilots' impressions of -realism are false translational accelerations
and washout effects on open-loop maneuvers. An example of the first would
be roll control in a simulator with' roll motion but no lateral travel.
When the subject rolled the simulator he would sense a proportional
lateral -acceleration because of gravity, whereas in an airplane- the
perceived, acceleration is generally very small -(i.e., the turn is
"coordinated"). Not only may the false cue affect the pilot's control

S-221
Caro, 1977, op cit.

195

÷ - 7 -- -
 S~12

behavior, but it will surely influence his subjective opinion of the

simulation realism. An example of the washout problem would be a pull-up
maneuver in a simulator with limited vertical travel. The initial
acceleration would be correct; but, because of the limited travel, it
would ýbe necessary to reverse the acceleration quickly. Washout
characteristics, which might be completely masked in a tracking task,
could become quite obvious in certain open-loop maneuvers.
•kD_•Several moving-base flight simulator experiments were recently per-

formed using roll and sway motions of the Large Amplitude Multimode
Aerospace Research Simulator (LAMARS) of the Flight Dynamics Laboratory at
Wright-Patterson Air Force Base, Ohio. The objectives of these experi-
ments were:

a. To tie in the roll-only results of four ex-

perienced pilots with previous results for four
well-trained nonpilot subjects.

To investigate effects of various lateral-beam-motion

"washout" filters designed to keep the lateral sway
Sb.
I within the * 10 ft of LAMARS travel. (Lateral beam
sway is used, within limits, to imitate the realisti-
-cally "coordinated" lateral motions of free-flight 4
roll manetrvers.)

The high-pass washouts- on lateral beam travel ( were of the

general second-order form:
• 2-

K s
Ybeam
2
Yfree flight s2+ 2wyaS + W
yy y

222 Jex, H. R., R. E. Magdaleno, and A. M. Junker, "Roll Tracking Effects

of G-Vector Tilt and Various Types of Motion Washout," Fourteenth
Annual Conference on Manual Control, NASA Conference Publication
20,November 1978.

196
where K. = attenuation factor, w - high-pans break frequency (rad/sec),

and closed loop damping -:atio, ýy = 0.70 (fited).

Values of K and wy were explored, from which example data will be

shown subsequently. A nonlinear (titae varying) washout waa also tested in
which wy was continuously adjusted in accordance with the smoothed magni-
tude of roll angle- so as -to permit correct cues for small roll activity,
while reducing the lateral beam travel peaks kor large roll angles.

Reshaping the forcing functions was also investigated and shown to reduce
travel requirements.

The pilot's taE-k was to follow an evasive (randomly rolling) target

while suppressing gust disturbances 2 23
. A two-independent-input technique
produced behavioral data (describing functions) and performance data
(error and control -scores), which revealed how pilots used the visual and
motion cues. Subjective data were also gathered -on the tracking task as
well as on limited "sidestep" maneuvrers.

The main results 2 2 4 show that:

I. 1he prent pilots and previous well-trained non-

pilots exhibited nearly identical behavior and
performance, implying universality of adaptation and
results.

2. The pilots' roll tracking behavior and performance

were not significantly affected by a variety of lat-
eral-sway washouts.

3. The nonlinear beam washout filter reduced the peak

lateral travels at the expense of occasionally greater
lateral-specific-force (ay) peaks, but otherwise did
not affect behavior or performance. It promises to
provide an adaptive washout which does not need to be

-223 Ibid. 11
224 Jex, Jewell, and Magdaleno, 1979, op cit.

5Je, Magdaleno, and Junker, 1978, op cit.

197
 iteratively fine-tuned to avoid hitting stope while
minimizing spurious washout artifacts. Additionally,
It should be especially useful during training, where
motion cue usage ic changing.

4. Both sidestep and random tracking maneuvers gave rise

to spurious lateral motion cues (the coordinated free-
flight case would have none) which were-characterized
as "out-of-phase," "like a student on the rudder
pedals," etc. Analysis showed these to be roughly
correlated by time- and frequency-response parameter:
related to sway washout gain, K., and frequency, w •
Combinations of K. and w were identified which pro-
vided the most acceptable impressions of roll and sway
N motion realism.

We shall now present some of the results which characterize the pilots'
judgments of "realism."

Although the pilots were encouraged to use their own words tc describe
* the effects of the motion cues, there was a certain amount of commonality
in the terms used by all the pilots. These are summarized below:

1. "Delayed side forces": These were side forces that were

* seemingly uncorrelated with the roll motion of the air-
craft. The• specific force, aaycab, was not completely
eliminated by translational acceleration, Ycab only
attefiuated and delayed by the sway axis washout filter.
Some pilots said this felt like a student kicking on the
rudder pedals.

2. "The leans": These were aide forces that wete perfectly

correlated with the roll motion of :he aircraft. The
pilots described "the leans" as a pressure either on their
knees or shoulders against the bulkhead of the 2ab when
they knew their aircraft was rolled either left or
right. Some pilots commented that when they were actively
involved in the roll -tracking task they did not notice
"the leans" but the "delayed side forces" could be

198
'.{{

~~ -
7 77T IFMP,
disconcerting. (Note here the conditional dependence of
the utricular threshold on task workload.)

3. "Change in the effective roll axis": The pilots felt that

the effective roll axis was above them for roll-only
motion. However for combined roll and sway motion the
pilots could discern changes in the effective roll axis
for various types of sway axis drive logic (i.e., various
combinations of K 1 and WY). This made the pilots feel as
if they were on the end of a variable-length pendulum as
Sand
Wy were changed.

4. "Change in stick sensitivity": Although not a consistent

comment, some pilots could discern changes in the effec- -
tive stick gain for various types of sway-axis drive
logic. This affected their impression of the task dif-
ficulty (e.g., "easier to fly now," or "more difficult to
track now").

The pilots' subjective impressions of the motion cues, as described

above, were used to define boundaries of acceptable combinations of the
parameters cf the sway-axis washout filter. The resulting "boundaries"
are summarized in the plot of K verus w shown in Fig. 41 (from Jex,

Jewell, and Magdaleno 226). The boundaries shown in Fig. 41 intentionally

appear nebulous for three reasons:

1. Pilot comments were not always repeatable, and many

times the pilots admitted that the changes in the
motion cues due to changing K and wy were very

226 Jex, Jewell, and Magdaleno, 1979, op cit.

199

? : - J-- -
 61"Roli-Swoy
-1 s + Washout Filters
- cbI.iJ..4

g- +

Roll Washout Roll Sway Washout Sway

clycob Ycab g9 cab

b) Comments Sway Washout Filter Frequency

Wylrtd/seci(e=.7l-
.2 4 .6 .8 1.0
"THE LEANS" '•
Gy

andsmall oy attenuation
to eth

correlated ray alosely

with at)

- Tro ctve y rol-nl due. NERANT

e- •..'•;."ACCEPTABLE"••.;"'"•';.••t

Fr toJexo
it) and the loans or dhch
yed side forces
>pressionso eaicsn predominat y o

•- '~~ULLY •:.- "•DELAYED SIDE-FORCE"

igrCOORDINATED
i.6
"• of
,• (oundarieg
W ashou
the-~
FitherPaaers1
(-l
f tiag t.7 1., 17`1" ( ay u ncorre l oted with )

tigure-41. Boundaries of Sway s Washout Filter Parameters

t •(Front Jex, Jewell, and Magdalena ) ghich ,Delineate the Pilots'
~Impressions of Realism
Si; From_ Combined Roll and Swa olo Iues

22 IbTid..

SI 200 ii
 4 77 ¾.'L'CZq'ý'''
7 "

subtle. Therefore only relative Judgments could be

-'rendered, and the pilots' subjective impressions of
the motion cues were a function of the starting points
of the , wy combination. The pilots were not told
which combination of Ky and W was being used, but
they were told when a change in the value of either KY
or wy was made. This experimental tecb,,ique was
adopted because it was very difficult for the pilots
to rate the motion cues on an absolute scale.

2. Pilot comments changed with the magnitude of the

target's randomly rolling motion. The pilots were
much more sensitive to changes in K. and/or wy for the
larger rolling amplitude than for the i iduced ampli-
tude. The difference in the pilot .- :mmentary is
probably due to an indifference threshold on specific
2 28
force (Roark and Junker report the ay indifference
threshold to be approximately 0.1 g).

3. Pilot comments changed with the task. This too was

probably related- to the p~lots' indifference thresh-
olds to specific force. For example, Fig. 42a

summarizes some pilot comments on a plot of peak ay

versus wy for Ky = 0.9. For bank and stop (sidestep) -

maneuvers the side forces become "disconcerting" twhen

w is greater than 0.4 rad/ sec (note that this is

where the ay peaks become greater than 0.1 g), but for
the tracking task with the reduced input the pilot
said "no difference" between wy - 0.3 and 1.0 rad/sec
(note that the ay peaks just reach 0.1 g for Wy = 1.0
rad/sec). A similar phenomenon occurred when Wy was

228 Roark and Junker, 1978, op cit.

201
A
 0 Bank and stop maneuvers.
0 Roll tracking.with "reduced" input

a) PEAK SPECIFIC FORCE VS. w yAT K, .9

Ayp
[3 "About the same"
"Very Uncoordinated"'
(large delayed side forces) K

(g) "~Side forces now 'O

2• Ay threshold
".2 sdisconcerting" threholdK
reported
"Less coordinated,
but not too bad"

"Feels 0 •- "No Difference"

C ae"Coordinated"
b) EASPCIICORCS
0 - , K FR
0 .5e .: ylradl sed, LO

b) PEAK SPECISIC FORCE VS. Ky FOR wo y.3

And So Ma"Feelv quite a bit ofreaning

(g) /-"Slight side forces detectable"

.2 "Feel leaning-but / "No difference"

not annoying"
"1//I e """..
" . . .I 10 g

Change" =-JCoordinated"
0-
0 .5 ~Kvlrad/ sec) L

Figure 4•2' Summary of Pilot Commentary for Bank

and Stop Maneuvers,-and Roll Tracking
 fixed and K varied, as shown in Fig. 42b. Also note
from Figs. 42a and 42b that for small values of K1
with w = 0.3 rad/sec the pilot complained about the
whereas for large " values of wy with K O 0.9
the pilot complained about "lagged side forces."

Finally one other important comment was the pilots' universal dis-
pleasure with hitting, the sway displacement limits. The adverse effects
of hitting displacement limits have been observed in other simulators
(e.g., Jewell, et al.,2)
229 and should be prevented by adopting nonlinear
motion drive logic.

The nonlinear washout filter had the predicted attribute of preventing

the sway displacement from hitting the LAMARS limits, because the amount
of lateral travel used is extremely sensitive to w (recall that w is
self-adaptive for the nonlinear filter). Otherwise back-to-back compari-
sons of the linear and nonlinear washout filters with the same value of
revealed no consistent differences in the pilots' subjective impression of
the motion cues- The tracking scores obtained with the linear and non-
linear filters were virtually identical, and the pilot describing
functions were also the same. However the amount of lateral travel used
by the nonlinear- filter was usually 30 percent less than that used by the
linear
aninease
filter
finter
during
ii iefre tracking.y
roll Except for occasionally greater K
,peaks, this reduction in lateral travel was not otherwise accompanied 'by I
an increase in spcific side force, ay

229 Jewell, Wayne F., Warren F. Clement, Lt. Col. Thomas C. West, USA
(Ret.), and Dr. S. R. M. Sinclair, Powered-Lift Aircraft Handling
Qualities in the Presence of Naturally-Occurring and Computer-
Generated.Atmospheric Disturbances, FAA-RD-•79-59, May 1979.

203
"AI
 R7----,- ---

N)TION PERCZPTION SUMMARY

In summary, fidelity requirements for the simulation of cockpit motion

Fi depend on the psychomotor role of motion cues in tracking and failure
detection tasks as well as on the pilots' impressions of realism.

With regard to the pilot's tracking performance and behavior, it is

generally more important to reproduce correct rotational motion cues over
an appropriate frequency range which will be predicted from validated

analysis of the specific tracking situation using the Multimodality Pilot

-Model and, if necessary, the Multiloop Pilot Model. Nevertheless one must

be cautious about providing only rotational motion cues in a simulator

(which are potentially useful to the pilot) but which are not present in
actual flight without providing corresponding specific forces which
accompany translation.

The simulation of motions accompanying a failure will help greatly in

the pilot's timely detection of the failure. This is especially true if

the visual modality is already heavily loaded with a demanding task. At

the very least the motion should be sufficient to provide an unambiguous

clue to the failure. In many cases failure detection may put the most
demanding requirement on translational motions.

Two specific problems which compromise the pilot's impressions of

realism are false translational accelerations and washout effects on open-
loop maneuvers. Roll motion without sway motion provides an exaggerated
proportional gravitational component of lateral acceleration which is
unrealistic. An example of the washout problem is provided by a pull-up

maneuver in a simulator with limited vertical displacement. Although the

initial acceleration would be correct, it would be necessary to reverse
the acceleration unrealistically because of the limited travel.

Roll and sway motion cues have recently been investigated with the aid

of the Air Force Flight Dynarics Laboratory's Large Amplitude Multimode

Aerospace Research Simulator (LAMARS). Various linear and nonlinear sway
motion washout filters were designed and tested to keep the sway

204

41-
 I.V.

displacement within the 110 ft of LAMARS travel. The main results from
this investigation show that:

0 The pilots' roll tracking behavior and performance

were not significantly affected by a variety of lat-
eral-sway washouts.

* The nonlinear beam washout filter reduced the peak

travels at the expense of occasionally greater
lateral-specifc-force (a) peaks, but otherwise did

I. not affect behavior or performance. It promises to

provide an adaptive washout which does not need to be
iteratively fine-tuned to avoid hitting stops while
minimizing spurious washout artifacts. Additionally
it should be. especially useful during training where
motion cue usage is changing.

0 Both sidestep and random tracking maneuvers gave rise

I! to spurious lateral motion cues (the coordinated
free-flight case would have none) which were charac-
-terized as "out-of-phase," "like a student on the
rudder pedals , etc. Analysis showed these to be
roughly correlated by time- and frequency-response
:1 parameters related to sway washout gain, Ky, and
ffrequency, y. Combinations of K. and Wy were iden-

tified which provided the most acceptable impressions

of roll and sway motion realism.

•il •205

. - -
 SUMMARY FOR rLAflKENG AND COLLECTING MEASUREMENTS

The purpose of this section is to outline various methods for col-

lecting empirical data needed to determine simulator fidelity

requirements. As shown, this involves not only the measurement tools

themselves but also the preparation for their use and the Judicious in-
terpretation of their results.

Some of the princJples considered important to the data colilection

process include: V

0 Being sensitive to the task being examined, its con-

text, and its -components.

S Being eclectic in applying measurement techniques

I several approaches may produce insight and are not
necessarily much more expensive than just one.

U
Table 23 is offered as a check-list for setting out to obtain empirical
data whether from simulator or -flight.

2i
0

" I 206
l(
 TABLE 23

* A CHECKLIST FOR PLANNING AND COLLECTING MEASUREMENTS

OF PILOT BEHAVIOR

* Establish purpose, scope, and scenario

Elect part- or full-missloh simulation

Specify mission phases, events, environment

Organize responsibilities, procedures, tasks for each crew member

within each mission phase delineated by events

Specify Inputs, types of activity (e.g., cognitive or

psychomotor), outcomes, and outputs associated with each task

S Perform essential pre-experimental analysis

Prepare activity time line analyses for normal and emergency

operations together with likely alternatives for procedural errors
which are foreseen

Classify non-intrusive measurements for the purpose of identifying

errors, piloting techniques, unusual pilot actibns, and degraded
pilot rating
- Procedure-centered evaluation based on time-sequences of all
variables and events

-System performance-centered evaluation

Command-following bandwidth or latency and critical

exceedences

I Disturbance regulation bandwidth or latency and critical

exceedences
/ Safety; operational capability (distributions of state
variables)

- Hman operator-centered evaluation

4Pilot acceptance (distributions of state and control

variables)

/ Temporal averages of task-specific dynamic behavior among

crew members

20V

,- _ 'n.-t2 --
- - -•---
 TABLE 23 (Continued)

t Subjective ratings - appropriate workload indices for

full-mission simulation

- Objective workload correlates Useful for

part-task
-Psychophysiological correlates simulation.

(Note that objective workload correlates are useful for

"calibrating" subjective ratings and psychophyslological
correlates are useful event markers)

V Eye Point of regard: useful for event markers, temporal

and ensemble distributions of attention

Define measurement support and structure organization, and specify

formats and media for output variables to be measured and recorded

-Discrete outputs, events

Continuous signals to be sampled

-Continuous signals without sampling

- Closed-circuit video

-Audio communications
"-Hard
copy (e.g., subjective ratings and observers' notes)

Estimate likely parameter values for proper and improper executior

of activities within normal and emergency procedures

Dry run portions of experiment and refine measurement techniques

Specify output variables to be fitted by distributions from which

probabilities can be estimated for the purpose of safety analysis
verification and for interpretation in terms of decision analysis
and workload analysis

* Manage and monitor data acquisition during experiment

Check against pre-experimental analysis

Look for measurement deficiencies

Keep up to date with as many on-line measurements as possible

"R"elate ueasurements to commentary and observations -

"208
 TABLE 23 (Concluded)

S Post experimental analysis

Analyze interrelationships among

- Procedure-centered measurements

- System performance-centered measurements

- Operator-centered measurements

Identify or postulate sources of human error and workload

Perform planned statistical analyses (if any) and update

hypotheses

Refine behavioral models

Recommend improvements to measurement prqcedures

Organize and present results

0 General recommendations

Treat data as archival

Acquire as much numerical definition as is practical (may be

limited by storage and non-interference requirements)

Do not restrict data acquisition to the narrow objectives of the

experiment; it may serve someone else 10 years hence!

CI

209

~-_A
 C0NCLUSIMOWS AND RECOMMENDATIONS

,f Human operator control theory, combined with adequate quantitacive J`

determining of
description Army training
simulator fidelityobjectives, offers The
requirements. a powerful potential for
full exploitation of

this potential, however, must await the gathering of crucial data which
describe piloting technique and pilot perception during specific flight
tasks and operating environments.

few significant examples have been identified which indicate how to

analyze and interpret simulator fidelity questions and obtain quantitative
results and answers. These examples tend to be associated with outer-loop

piloting tasks in critical flight phases, such as nap-of-the-earth

maneuvering and fixed-wing landing flare.

An operational definition of simulator fidelity has been proposed in a

manner which unifies control theory and pilot training notions. Simulator
fidelity is considered to be the degree to which perceivable states are
present which are essential to inducing correct psychomotor and cognitive
behavior for a given task and envirorment. If training is the development
of essential feedback loop structure for a given task or training
objective, then we may considec fidelity as being reflected by the
essential cues which are available to the pilot to close the essential
loops.

The term "explicit fidelity" has been tied to the combination of pilut
perception and piloting technique exhibited in the simulator as compared
to that exhibited in an actual flight situation. Thus explicit fidelity
carries with it the ideas of perceptual fidelity. A more incidental kind
of fidelity can be associated with the actual simulator software and
hardware characteristics, those features which are normally considered to
come nnder the heading of objective or engineering fidelity. One
important result of this report has been to show how the explicit
simulator characteristics of piloting technique and pilot perception can
be q..antified.

210
 The first step in establishing simulator fidelity requirements is the
quantitative description of the piloting task or training objective. This
constrains the pilot-aircraft scenario so that control theory analysis
tools can be systematically applied. Much of this quantitative
description can be obtained from existing training documents such as the
ATMs.

The major missing elements which preclude an immediate, across-the- K

board analysis of fidelity needs are:

1. The systematic measurement of piloting technique in a

real-world environment.

2. Measurement of perceptual transfer functions under

appropriate task loadings.

3. Determination of how the micro-details of a modality

medium (especially visual) convey composite state
information.

The first two elements above would permit a systematic accounting of the
fidelity potential which is available in existing simulator or training
device motion and visual systems. With the addition of the third item
frcm the above list, it would be theoretically possible to predict
requirements for given training objectives.

Essential cues, the root of simulator fidelity requirements, issue

from essential loops; and, in many cases, these loops are easy to identify

t from task descriptions.

essential, fidelity
For such cases, where only the command loops are
requirements tend to be
requirements may not -be as clear for critical fidelity conditions where
loops intermediate between the command and innermost loops are also
relatively clear; but the

essential.

Essential loop structure can be inferred from relatively simplh

measurements made of actual flight maneuvers. In one example, dealing
with the landing maneuver for large jet transport aircraft, strong
J evidence was found for an essential direction-of-flight loop in addition

211
 to the mote obvious and essential altitude loop. When viewing thir same
maneuver on a particular training simulator there was little or no
evidence of this important direction-of-flight loop in most cases, and it
wa.3 preýsumed that there was a deficiency in the direction-of-flight cue
available to the pilot. While It has not yet been possible to discover
the exact source of the cue deficiency in this case (either motion or
visual) it has been possible to tabulate a bounded list of candidates for
the fidelity problem. The most plausible of the various alternatives in
this case was a deficiency in visual eirection-of-flight information due
to the restricted field of view. For thie fidelity potential exhibited in
this particular simulator, a sample extrapolation was made from the
airline jet transport to various Army fixed- and rotary-wing aircraft and
their respective piloting techniques.

A procedure was devised which would provide an analytic determination

of essential cue information required to accomplish a given training
objective, but systematic execution of this procedure would be seriously
hampered by lack of basic data describing piloting technique and pilot
perception in flight. The procedure itself involves reasonably simple and
common feedback system analysis tools.

A bookkeeping scheme has been suggested which would aid in gathering

basic data for use in simulator fidelity determination. This bookkeeping
scheme consists of a number of matrices which can be used to describe the
various components of simulator fidelity. This bookkeeping scheme has the
additional benefit of pointing out where there is commonality of training
objectives with regard to simulator fidelity. Furthermore it would permit
extrapolation to training objectives not yet defined or to aircraft types
not yet designed and built.

Additional research is required to define the missing elements which

were previously listed. This research centers around quantifing piloting
technique and pilot perception under appropriate task loading in flight
and in corresponding training simulators. These kinds of measurements
should constitute the next major steps in simulator fidelity research. It
should be recognized that this research. can be accomplished with various
levels of effort ranging from occasional. targets of opportunity to a full-

212
.x
i: blown, intensive research program intended to fill out large portions of

Sthe missing data. Clearly the time required to obtain meaningful

Ssimulator fidelity definition depends upon the intensity of the research
S~effort.

I--

II
r!I
--' ' ..2-r_
--=; .F
- ,. _b
 GENERAL ALPHABETICAL BIBLIOGRAPHY

Adams, Gerald H., (Ed.), Flight Simulator Training. Volume 1: 1964-1973

(A Bibliography with Abstracts), NTIS/PS-77/0145, March 1977.

Alex, Fredric R., Geometric Foundation of Motion Perspective, Systems

Technology, Inc., Working Paper No. 170-3, January 1967.

Allen, R. W., and D. T. McRuer, "The Effect of Adverse Visibility on

Driver Steering Performance in an Automobile Simulator," SAE
Paper 770239, March 1977.

Allen, R. Wade, and Duane McRuer, The Man/Machine Control Interface --

Pursuit Control, Systems Technology, Inc., Paper No. 243, April
1979.

Anon., Basic Helicopter Handbook, Federal Aviation Administration,

AC 61-31, 1965.

Anon., Helicopter Gunnery, Department of the Army, TCI-4, September 1976.

Anon., Jeppesen Commercial Instrument Manual, Jeppesen Sanderson Inc.,

Denver, Colorado, 1977.

Anon., United Airlines, DC-10 Flight Manual, Training and Reference,

1973-1974, United Airlines Flight Training Center, Stapleton.
International Airport, Denver, Colorado, October 17, 1977.

Anon., Program of Instruction for 2C-15A/2C-100B-B, Officer/WO Rotary Wing

Aviator Course, U. S. Army Aviation Center, Ft. Rucker, Alabama,
SSI/MOS: 15A/100B, February 1978.

Anon., Aircrew Training Manual, Utility Helicopter, Department of the

Army, TC 1-135, October 1978.

Anon., Aircrew Training Manual. Attack Helicopter, Department of the

Army, TC 1-136, October 1978.

Anon., Aircrev Training Manual. Observation Helicopter, Department of the

Army, TC 1-137, October 1978.

Anon., Aircrew Training Manual. Cargo Helicopter, Department of the Army,

TC 1-139, October 1980.

Anon., Aircrew Training Manual. Surveillance Aircraft, Department of the

Army, TC 1-144, october 1978.

Anon., 50 Years of Flight Simulation. Conference Proceedings. Session I

through 3, 1979.

214
 Anon., Rotary Wing Flight, U. S. Army Field Manual No. 1-51,
4 16 April 1979.

Anon., Instrument Flying, Air Force Manual, AFM 51-37, 15 August 1979.

Anon., Aircrew Training Manual. Utility Aircraft, Department of the Army,

TC 1-145, October 1980.

Baron, Sheldon, Roy lancraft, and Greg Zacharias, Pilot/Vehicle Model

Analysis of Visual and Motion Cue Requirements in Flight Slmula-
tion, NASA CR-3312, October 1980.

Bickley, William R., "Optimizing Simulator-Aircraft Training Mixes,"

* Proceedings of the 2nd Interservice/Industrv Training Equipment
Conference, Novevr--r 1980.

Rlackwell, H. R., "Contrast Thresholds of the Human Eye," Journal of the

Optical Society of America, 1946, pp. 624-643.

Blackwell, H. R., anti D. W. MeCready, Jr., "Foveal Detection Thresholds

'i$ for Various Durations of Target Presentation," Minutes and Pro-
ceedings of NAS-NRC Vision Committee, November 1952,
ACSIL/53/4405.

Calvert, E. S., "The Theory of Visual Judgments in Motion and Its Appli-
cation to the Design of Landing Aids for Aircraft," Transactions.
of the Illuminating Engineering Society, 22, 1957.

Caro, Paul W., "Platform Motion and Simulator Training Effectiveness,"

Proceedings of the 10th NTEC/Industry Conference, Naval Training
Equipment Center, Orlando, Florida, November 15-17, 1977,
NAVTRAEQUIPCEN 111-294.
Carter, V. E., "Training Effectiveness Analysis," Appendix XV, Future
Undergraduate Pilot Training System Study: Final Report,
Aeronautical Systems Division, Air Force Systems Command, WPAFB,
Ohio, March 1971.

Carter, V. E., and C. A. Semple, Specific Behavioral Objectives for VF-121

Training in Basic Air Combat Maneuvers. Volume 3, (CONFIDENTIAL),
NOR 76-52, March 1976.

Clement, W. F., et al., A Reexamination of Pilot Eye Movement Data,

Systems Technology,. Inc., Technical Memorandum No. 163-A, Rev.
28 February 1967.

Clement, Warren F., R. Wade Allen, and Dunstan Graham, Pilot Experiments
"for a Theory of Integrated Display Format (Final Report),
JANAIR Report No. 711107, October 1971.

215
 Clement, W. F., D. T. McRuer, and R. H. Klein, "Systematic Manual Control
Display Design," Guidance and Control Displas, AGARD CP-96, Feb-
ruary 1972.

Clement, W. F., et al., Application of Manual Control Display Theory to

the Development of Flight Director Systems for STOL Aircraft. H
Part II: Multi-Axis Sampling, Pilot Workload, and Display Inte-
gration, Systems Technology, Inc., Technical Report No. 1011-2,
¶ 1974.
Clement, Warren F., and Robert K. Heffley, Some Effects of Adverse. Visi-
bility on Threshold Properties f the Pilot's Perception in VTOL
A•roaches to Non-Aviation .Shins, Systems Technology, Inc.,
Working Paper No. 1115-3, May 1978.

Clement, Warren F., Robert K. Heffley, and Wayne F. Jewell, Field-of-View

Requirements for Approach and landing of V/STOL Aircraft,
NADC 77240-07, August 1978.

Clement, Warren F., Robert K. Heffley, Wayne F. Jewell, and Weane T.

McRuer, Technical Approaches for Measurement of Human Errors.
Final Report, Systems Technology, Inc., Technical Report
No. 1156-2, July 1980.

Cooper, George E., and Robert P. Harper, The Use of Pilot Rating in the
Evaluation of Aircraft Handling _Qalities, NASA TN D-5153, 1969.

Curry, R. E., W. C. Hoffman, and L. R. Young, Pilot Modeling for Manned

Simulation. Volume I, AFFDL-TR-76-124, December 1976.

Davies, E. B., Contrast Tresholds for Air to Ground Vision RAE Technical
Report 65089, April 1965.!

Eddowes, Edward E., and Wayne L. Waag, The Use of Simulators for Training
In-Flight and Emergency Procedures, AGARD-AG-248, June 1980.

Fitts, P. M., R. .E. Jones, and J. L. Milton, "Eye Movements of Aircraft

Pilots During Instrument-Landing Approaches," Aeronautical Engin-
eering Review, 9, February 1950, pp. 24-29.

Frenk-, R. S., D. E.-Skaar, and J. A. Tennant, "Driver's Visual Detection,"

SAE Paper 720142 presented at the SAE Automotive Engineering Con-
:1 gress, Detroit, Michigan, 10-14 January 1972.

Gainer, Charles A., and Dennis J. Sullivan, Aircrew Training Requirements

for Nap-of-the-Earth Flight, U. S. Army Research Report 1190,
August 1976.

Gibson, J. J., The Perception of the Visual World, Houghton Mifflin,

Boston, 1950, Chapters 6 and 7.

101
216
-I
Gilinsky, A. S., "Perceived Size and Distance in Visual Space," Psycho-
logical Review, 58, 1951, pp. 460-482.

Gilinsky, A. S., "The Effect of Attitude Upon the Perception of Size,"

American Journal of Psychology, 68, 1955, pp. 173-192.

Gordon, D. A., "Static and Dynamic Visual Fields in Human Space Percep-
tion," Journal of the Optical Society of America, 55, 1965, pp.
1296-1303.

Gordon, D. A., "Perceptual Basis of Vehicular Guidance," Public Roads, 34,

1966, pp. 53-68.

Graham, C., "Visual Perception," In A Handbook of Experimental Psychology,

(S. S. Stevens, Ed.), John Wiley, New York, 1951, pp. 868-920.

Graham, Dunstan, and Duane McRuer, Analysis of Nonlinear Control Systems,

McGraw Hill, New York, 1961.

Grunwald, A. J., and S. J. Merhav, "Vehicle Control by Visual Field Cues,"

IEEE Transactions, SMC-6, 1976, pp. 835-845.

Habercom, Guy E., Jr., (Ed.), Flight Simulator Training. Volume 2:

1974-February 1980 (Citations from the NTIS Data Base),
NTIS/PB80-806425, March 1980.

Havron, M. Dean, Information Available from Natural Cues During Final

Approach and landing, Human Sciences Research, Inc., Report
HSR-RR-62/3-MK-X, March 1962.

Heffley, Robert K., Closed Loop Analysis of Manual Flare and Landing,
AIAA Paper 74-834, August 1974.

Heffley, Robert K., "A Model for Manual Decelerating Approaches to Hover,"
Proceedings of the Fifteenth Annual Conference on Manual Control
AFFDL-TR-79-3134, 1979, pp. 545-354.

Heffley, Robert K., and Wayne F. Jewell, Development of a CTOL Piloting

Technique Measurement Scheme for a Real-Time Simulator Environ-
ment, NASA CR-152294, July 1979.

Heffley, Robert K., Wayne F. Jewell, John M. Lehman, and Richard A. Van
Winkle, A Compilation and Analysis of Helicopter Handling
Qualities Data. Volume One: Data Compilation, NASA CR 3144,
August 1979.

Heffley, Robert K., A Compilation and Analysis of Helicopter Handling

Qualities Data. Volume Two: Data Analysis, NASA CR 3145, August
1979.

217
 - 11,

Heffley, Robert K., Wayne F. Jewell, Richard R. Whitbeck, and Ted M.

Schulman, The Analysis of Delays in Simulator Digital Computing
Systems. Volume One: Formulation of an Analysis Approach Using a
Central Example Simulator Model, NASA CR-152340, February 1980.

Heffley, Robert K., Warren F. Clement, and Samuel J. Craig, "Training

Aircraft Design Considerations Based on the Successive
Organization oi• Perception in Manual Control," Sixteenth Annual
Conference on Manual Control, MIT, May 5-7, 1980, pp. 119-127.

Heffley, Robert K., Ted M. Schulman, Robert J. Randle, Jr., and Watren F.
Clement An Analysis of Airline Landing Data Based on Flight and
Training Simulator heasurements, Forthcoming NASA Technical
Memorandum, 1981.

Heffley, Robert K., Ted M. Schulman, Derivation of Human Pilot Control

Laws Based on Literal Interpretation of Pilot Training Literature,
AIAA Paper No. 81-1822, to be presented at the AIAA Guidance and
Control Conference, Albuquerque, New Mexico, August 19-21, 1981.

Hennessy, Robert T., Dennis J. Sullivan, and Herbert D. Cooles, Critical

Research Issues and Visual System Requirements for a V/STOL Re-
search Simulator, NAVTRAEQUIPCEN 78-C-0076-1, June 1979.

Hess, Ronald A., "Pursuit Tracking and Higher Levels of Skill Development
in the Human Pilot," Sixteenth Annual Conference on Manual Con-
trol, MIT, Cambridge, Massachusetts, May 5-7, 1980.

Hofmann, L. G., and Susan A. Riedel, Manned Engineering Flight Simulation

Validation. Part I: Simulation Requirements and Simulator Motion
System Performance, AFFDL-TR-78-192, February 1979.

Hoh, Roger H., Samuel J. Craig, and Irving L. Ashkenas, Identification of

Minimum Acceptable Characteristics for Manual STOL Flight Path
Control. Volume III: Detailed Analyses and Tested Vehicle Char-
acteristics, FAA-RD-75-123, June 1976.

Holman, Garvin L., "Suitability-for-Training Evaluation of the CH-47

Flight Simulator," Proceedings of the llth NTEC/Industry
Conference, NAVTRAEQUIPCEN IH-306, Naval Training Equipment
Center, Orlando, Florida, November 14-16, 1978.

Jewell, Wayne F., Robert L. Stapleford, and Robert K. Heffley, Computed

Responses of Several Aircraft' to Atmospheric Turbulence and Dis-
crete Wind Shears, NASA CR-152181, February 1977.

"Jewell, Wayne F., Warren F. Clement, Lt. Col. Thomas C. West, USA (Ret.),
and Dr. S. R. M. Sinclair, Powered-Lift Aircraft Handling Quali-
ties in the Presence of Naturally-Occurring and Computer-..enerated
Atmospheric Disturbances, FAA-RD-79-59, May 1979.
4J

218

Wit
Jewell, Wayne F., and Ted M. Schulman, A Pilot Control Strategy Identifi-
cation Technique for Use in Multiloop Control Tasks, Systems
Technology, Inc., Technical Report No. 1153-2, August 1980.

Jewell, Wayne F., and Warren F. Clement, A Simple Method for Measuring the
Effective Delay in Simulator Digital Computing Systems Involving
Manual Control, Systems Technology, Inc., Working Paper
No. 1156-4, December 1980.

Jewell, Wayne F., Identification of Multiloop Pilot Describing Function

Obtained From Simulated Approaches to an Aircraft Carrier, Systems
Technology, Inc., Paper No. 289, presented at the Seventeenth
Annual Conference on Manual Control, UCLA, June 16-19, 1981.

Jex, H. R., D. T. McRuer, R. W. Allen, and J. R. Hogge, Requirements and

Program Plan for Simulation Facility and Data Handling for GM
Design Staff Safety and Human Performance Group, Systems
Technology, Inc., Technical Report No. 2039-1, March 1974.

Jex, H. R., R. E. Magdaleno, and A. It. Junker, "Roll Tracking Effects of

G-Vector Tilt and Various Types of Motion Washout," Fourteenth
Annual Conference on Manual Control, NASA Conference Publication
No. 2060, November 1978.

Jex, Henry R., Wayne F. Jewell, and Raymond E. Magdaleno, "Effects of

Various Lateral-Beam-Motion Washouts on Pilot Tracking 2nd Opinion
in the "LAMAR" Simulator," Proceedings of the Fifteenth finnual
Conference on Manual Control, AFFDL-TR-79-3134, 1979, pp. 244-266.

Key, David L., (Ed.), Fidelity of Simulation for Pilot Training, AGARD
Advisory Report LNo. 159, October 1980.

Kibort, Bernard R., and Fred J. Drinkwater III, A Flight Study of Manual
Blind Landing Performance Usiný Closed Circuit Television Dis-
plays, NASA TN D-2252, May 1964.

Krendel, E. S., and D. T. McRuer, "A Servomechanisms Approach to Skill

Development," Journal of the Franklin Institute, 269, 1960,
pp. 24-42.

Levison, W. H., and A. M. Junker, "Use of a Tilt Cue in a Simulated

Heading Tracking Task," Fourteenth Annual Conference on Manual
Control, NASA Conference Publication 2060, November 1978.

Mackworth, N. H., "Visual Noise Causes Tunnel Vision," Psychoneurological

Sciences, 3, 1965, pp. 67 and 68.

McRuer, D. T., et al., New Approaches to Htuan-Pilot/Vehicle Dynamic An-

alysis, AFFDL-TR-67-150, February 1968.

219

- .- - - -- - - - ---- - - --
 McRuer, Duane, Irving Ashkenas, and Dunstan Graham, Aircraft Dynamics and
Automatic Control, Princeton University Press, Princeton, New
Jersey, 1973.

McRuer, D. T., and E. S. Krendel, Mathematical Models of IHutan Pilot Be-

havior, AGARD-AG-188, January 1974.

McRuer. Duane, "Human Dynamics in Man-Machine Systems," Niemi (Ed.), A

Link Between Science and Application of Automatic Control, Pro--
ceedings of Triennial World Congress of IFAC, 4, Pergamon Prese,
Elmsford, New York, 1979.

McRuer, Duane T., Warren F. Clement, and R. Wade Allen, A Theory of Human
Error, Systems Technology, Inc., Technical Report No. 1156-1,
May 1980.

Middleton, W. E. Knowles, "Vision Through the Atm6sphere," Canada: Ln.!-

versity of Toronto, 1952.

Moen, Gene C., Daniel J. DiCarlo, and Kennetn F. Yenni, A Parametric An-
alysis of Visual Approaches for Heli-opters, NASA TN D-8275,'
December 1976.

Naish, J. M., "Control Information in Visual Flight," Proceedings of the

Seventh Annual Conference on Manual Control, NASA SP-281, June
1971 (Also McDonnell Douglas Paper No. 5921).

Noton, D., "A Theory of Visual Pattern Perception," IEEE Transactions,

SSC-6, 1970, pp. 349-357.
Ormsby, C. C., Model of Human -Dynamic Orientation, MIT Ph.D. Thesis,

January 1974.

Palmer, Everett, and John Petitt, "A Measure of Psychological Realism on a

Visual Simulator," Journal of Aircraft, 14, May 1977, pp. 421-422.

Palmer, Everett, Tamara Mitchell, and John Pettit, Angular Size Estima- 41
tion: A Measure of Simulator Visual Fidelity, (unpublished paper)
No date, circa January 1979.

Puig, Joseph A., William T. Earrns, and Gilbert L. Ricard, Motion in

Flight Simulation: An Annotated Bibliography, NAVTRAEQUIPCEN
IH-298, July -1978.

Radder, Preston W., Capt. Department of the Army Approved Qualitative

Materiel Requirement (QMR) for a Synthetic Flight Training System
(SFTS) (Rotary Wing), 28 August 1972.

(' Randle, Robert J., Stanley N. Roscoe, and John C. Petitt, Effects of
Magnification and Visual Accommodation on Aimpoint Estimation in
Simulated Landings with Real and Virtual Image Displays, NASA
TP 1635, October 1980.

220

-SZ
Ringland, R. F., R. L. Stapleford, and R. E. Magdaleno, Motion Effects on
an IFR Hover Task - Analytical Predictions and Experimental Re-
sults, NASA CR-1933, November 1971.

Ringland, Robert F., Warren F. Clement, and H1enry R. Jex, Factors in De-
"termining Visual and Motion Fidelity Requirements for Training
Simulators, Syetens Technology, Inc., Working Paper No. 1162-4,
December 1980, Revised March 1981.

"Roark, Marvin, and Andrew Junker, "The Effects of Closed Loop Tracking on
a Subjective Tilt Threshold in the Roll Anis," Proceedings of the
Fourteenth Annual Conference on Manual Control, University of
Southern California, April 25-27, 1978.

Roberts, Edward 0., External Visibility Criteria for VTOL Aircraft,

AFFDL-TR-67-27,

Roscoe, S. N., "Incremental Transfer Effectiveness," Human Factors, 13,

1971, pp. 561-567.

Roscoe, Stanley N., Halim Ozkaptan, and Aaron Hyman, Review of Flight
Training Technology, U. S. Army Research Problem Review 76-3,
July 1976.

Roscoe, Stanley, N., "When Day is Done and Shadows Fall, We Miss the
Airport Most of All," Proceedings of the 11th NTEC/Industry
Conference, NAVTRAEQUIPCEN IH-306, Naval Training Equipment
Center, Orlando, Floridai November 14-16, 1978.
L
Roscoe, Stanley N., Ground-Referenced Visual Orientation with Imaging
Displays: Final Report, AFOSR-79-4, November 1979.

Sanders, A. F., "Some Aspects of the Selective Process in the Functional

Visual Field," Ergonomics, 13, 1970, pp. 101-117.

Schalow, P. S., Specification for UTTAS Helicopter Sy'nxthetic Flight Train-

ing System, Device 2B38, Naval Training Equipnent Center Report
No. 2222-1152, 30 October 1975.
Sheridan, T. B., "The Human Operator in Control Instrumentation," in R, H.
Macmillan, et al., (Eds.), Progress in Control Engineering, Aca-
demic Press, New York, 1962, pp. 141-187.

Sinacori, J. B., The Determination of Some Requirements for a Helicopter

Flight Research Simulation Facility, Systems Technology, Inc.,
Technical Report No. 1097-1, September 1977.

Sinacori, John B., Piloted Aircraft Simulation Concepts and Overview,

NASA CR-152200, March 1978.

Sperling, G., "Successive Approximations to a Model for Short Te•-iem-

Sory," Acta Psychologica, 13, 1970.

221

- -< - N "
 Spring, W. G., "Simulatlon Development and Preparation," Volume 4 of

Experiments to Evaluate Advanced Flight Simulation in Air Combat

Pilot Training, NOR 76-52, Northrop Corporation, March 1976.

Stapleford, R. L., D. T. McRuer, and R. E. Magdaleno, Pilot Describing

Function Measurements in a Multiloop Task NASA CR-542, August
1966.

Stapleford, R. L., S. J. Craig, and J. A. Tennant, Measurement of Pilot

Describing Functions in Single-Controller Multiloop Tasks,
NASA CR-1238, January 1969.

Stapleford, Robert I., Richard A. Peters, and Fred R. Alex, Experiments

and a Model for Pllot Dynamics with Visual and Motion Inputs,
NASA CR-1325, May 1969.

Weir, David H., and Duane T. McRuer, Conceptualization of Overtaking and

Passing on Two-Lqne Rural Roads, Franklin Institute Research Lab-
oratory, Technical Report No. 1-193, December 1967.

Weir, D. H., and R. H. Klein, The Measurement and Analysis of Pilot Scan-
ning and Control Behavior During Simulated Instrumented
Approaches, NASA CR-1535, June 1970.

Weir, D. H., and D. T. McRuer, Pilot Dynamics for Instrument. Approach

Tasks: Full Panel Multiloop and Flight Director Op2rations,
NASA CR-2019, Fay 1972.

Wewerinke, P. Ho, Visual

Apprcnch, NR-TR-78130U,
Scene Perception rrocess Involved in
1978.
the Manual
1i
Whitbeck, Richard F., and Dennis G, J. Didaleusky, Multi-Rate Digital
Control Systems with Simulation Appllcations. Volume I:
Tectnical Report, AFWAL-TR-80-3101, Volume 1, September 1980.

Young, L. R., "Current Status of Vestibular System Models," Automatica, 5,

1969, pp. 369-383.

Young, L. R., L. Stark, and C. Kupfer, Physiology of the Visual Control

System, NASA CR-238, June 1965.

Zacharias, Greg L., and William H. Levison, "A Model-Eased Procedure for
Determining .Visual Cue Requirerents," 1981 IMAGE Ii Conference
Proceedings, AFHRL, Scottsdale, Arizona, June 10-12, 1981. I

C 222 :
 BIBLIOGRAPHY OF PILOTING TECHNIQUES

Anon., Basis Helicopter Handbook, Federal Aviation Administration

AC 61-13, 1965.

Anon., Undergraduate Pilot Training Task Analysis, Phase II Report, Naval

Air Station, Corpus Christi, Texas, July 1973.

Anon., Technical Manual. Operator's Manual, Army Model AH-1CG Helicopter,

Department of the Army, TM 55-1520-221-10, 12 December 1975.

Anon., Gunnery Training for Attack Helicopters, Department of the Army,

TC 17-17, 31 December 1975.
:•Anon., Helicopter Gunnery, Department of the Army, TC 1-4, September 1976.

Anon,, Commercial Instrument Manual, Jeppesen Sanderson, Inc., 1977.

SAnon., Technical Manual. Operator's Manual, Army Model U-21A Aircraft,

Department of the Army, TM 55-1510-209-10, 25 March 1977. i1.

Anon., Technical Manual. Operator's Manual, Army AH-IS (Prod) Helicopter,

Department of the Army, TM 55-1520-236-10, 29 April 1977.

Anon., "Mountain Flying Skill and Know-How," U. S. Army Aviation Digest,

November 1977, pp. 30-32.

Anon., Technical Manual. Operator's Manual, Army Model OH-58A Helicopter,

Department of the Army, TM 55-1520-228-10, 7 April 1978.

Anon., Technical Manual. Operator's Manual, Army OH-58C Helicopter, De-

partment of the Army, TV.55-1520-235-10, 7 April 1978.

SAnon._, Flight Training Instruction, T.--4J, Part X, CQ Stage, Advanced

i•,• Strike, Naval Air Station, CNAT P-1545, July 1978.

Anon., Technical ,Manual. erator's Manualt OV-pD/RV-,D Aircraft Depart-

ment of the Army, TM 55-1510-213-90, 4 August 1978.
• ~Anon., Technical Manual. Operator's Manual. Army Model CH-47C Helicopter,
Department of the Army, TM 55-1520-227-10-2, 23 August 1978.

Anon., Aircrew Training Manual. AUtlitt Helicopter, Department of the

Army, TC 1-135, October 1978.

uon., Aircrew Training Manual. Attack Helicopter, Department of the

S~Army, TC 1-136, October 1978. '2

' ~223
 Anon., Aircrew T'raining Manual. (bservation Helicopter, Department of the
Army, TC 1-137, October 1978.

Anon., Aircrew Training Manual. Surveillance Aircraft, Department of the

Army, TC 1-144, October 1978.

Anon., Technical Manual. Operator's Manual, Army Model T-42A Aircraft,

Department of the Army, TN, 55-1510-208-10, 27 February 1979.

Anon., Rotary Wing Flight, U. S. Army Field Manual No. 1-51, 16 April
1979.

- Anon., Operator's Manual. Army ?kdels, UH-ID/H and EH-IH Helicopters,

Department of the Army, TM 55-1520-210-10, 18 May 1979.

Anon., Instrument Flying, Ai: Force Manual AFM 51-37, 15 August 1979.

Anon., "Keep That Tail Up!" U. S. Army Aviation Digest, January 1980,
pp. 24-26.

Anon., Flight Training Instructions, TA-4J, Parts VII and VIII, UG 43 00

0000 KQR, ON and WEP Stages, Naval Mtr Station, CNAT P-1543, February

Anon., Flight Training Instruction. TA 4J. Volume 1, Parts I and II, UG

43 00 00 00 KR, FAM Stage, Strike, Navkyal Air Station, CNAT P-1531,
March 1980.

Anon., Flight Trmining Instruction. UG 43 00 00 KL, BI, RI and AN Stages,

Striket Naval Air Station, CNAT P-1539, March 1980.

-Anon., Flight Training Instructions, TA-4J, Volume 5, UG 43 00 00 00 KR,

TALF and ACM Stages, Strike, 1980. Naval Air Station, CNAT P-1544,
March 1980.

Anon.j, Flight Training Instruction, TA-4J, Volume 3, Parts IV and V, Form V

and NF Stages, Advanced STRIKE, 1980, Naval Air Station, CNAT P-1561,"
April 1980.

Anon., Aircrew Training Manual. Cargo Halicopter, Department of the Army,

TC 1-139, October 1980.

Anon., Aircrew Training Manual. Utility Airplane, Department of the Army,

TC 1-145, October 1980.

Anon., Airman's Information Manual (AIM), Basic Flight Information and ATC
Procedures, Federal Aviation Administration, Jaiuary 1981.

Bickley, William R., Formulation and Evaluation of a herhodology for Pre-

dicting Required Hands-On Trainiag Following Simulator Training, (to
be published).

224
 i -

§ M __F_ _'IM

Cook, Charles E, CW3, "Tips for Instrument Flight Planning," U. S. Army

Aviation Digest, February 1980, pp. 41-44.
Gainer, Charles A., and Dennis Aircrew Task Analysis and
J. Sullivan,
Training Objectives for Nap-of-the-Earth Flight, U. S. Army Research
Institute for the Behavioral and Social Sciences, Research Memor-
andom 76-2, February 1976.

Holman, Garvin L., Training Effectiveness of the CH-47 Flight Simulator,

U. S. Army Research Institute for the Behavioral and Ficial Sciences,
Research Report 1209, May 1979.

Kershner, William K., The Instrument Flight Manual, The Instrument Rating,
•,. Second Edition, 1976.

Manningham, Dan, Staying Current. A Proficiency. Guide for Serious Pilots,

Ziff-Davis Publishing Company, New York, 1980.

John G. , III, STaylor,

Capt., "P~eware of the Vortex Ri•ng," '. S. Army Avia-
tion Digest, January 1978, pp. 45-47.

a.I-

.•1.

9F 77 I _M

22
 RIBLIOGRAPHY OF V•'3IJAL PERCEPTION

Alex, F. R., Geometric Foundation of Motion Perspective, Systems Tech-

nology, Inc., Working Paper No. 170-3, January 1967.

Baron, Sheldon, Ramal Muralidharan, and David Kleinman, Closed Loop Models
for Analyzing Engineering Requirements for Simulators,
NASA CR-2965, February 1980.

Baron, Sieldon, Roy Lancraft, and Greg Zacharias, Pilot/Vehicle Model

Analysis of Visual and Motion Cue Requirements in Flight Simula-
tion, NASA CR-3312, October 1980.

Borah, Joshua, Laurence R. Young, and Renwick E. Curry, Final Briefing,

Sensory Mechanism Modeling Study, Contract No. F33615-76-6-0039,
January 1976.- July 1977, and Initial Briefing, Additional Re-
search on Sensory 'echanism ?odelinZ Contract No. F33615-76-
6-0039, June 1977 - August 1978, Applied Science Laboratories,
Gulf + Western, 20 July 1977.

Bynum, J. A-., W. G. Matheny, J. E. Flexman, and R. K. Wilson, Test of a

Model of Visual Spatial Discrimination and Its Application to
Helicopter Hovering Control, Life Sciences, Inc., LSI-TR-73-1, V
November 1973.

Bynum, James A., Brian D. Shipley, and William C. White, "Helicopter Night
Low-Level Navigation: Study 1 - Checkpoint Identification," Pro-
ceedings of the 32nd Annual National V/STOL Forum of the American
Helicopter Society, 1976 (AHS Preprint 1024).

Calvert, E. S., "The Thecry of Visual Judgments in Motion and Its Appli-
cation to the Design of landing Aids for Aircraft," London,
Transactions of the Illuminating Engineering Society, 22, 1957.

Clement, W. F., Some Pilot-Oriented Considerations of the External Field

of View in Low-Visibility V/STOL Approaches, Systems Technology,
Inc., Working Paper No. 51G-29, April. 1968.

Ebeling, William C., "A New Approach to Visual Simulation," Journal of

Aircraft, 6, 1960, pp. 59-64.

Gibson, J. J., The Perception of the Visaal World, Boston, Houghton Mif-
flin, 1950, Chaps. 6-7.

Gordon, D. A., "Static and Dynamic Visual Fields in Human Space Percep-
tion," Joucnal. of the Optical Society of America, 55, 1965, pp.
1296-1303.

Gordon, D. A., "Perceptual Lasis of Vehicular Guidance," Public Roads, 34,

1966, pp. 53-68.

226

g - -- k _ _ __ _ _ __ _ _- _ _ _ __ _ _
Graham, C. H., "Visual Perception," In A Handbook of Eicperimental Psy-
chology, (Ed. S. S. Stevens), John Wiley, New York, 1951, pp. 868-
920.

Grunwald, A. J., and S. J. Merhav, "Vehicular Control by Visual Field

Cues," IEEE Transactions, SMC-6, 1976, pp. 35-845.

Haines, R. F., "A Review of Peripheral Vision Capabilities for Display

Layout Designers," Proceedings of the Society for Information
Display, 16, 1975, pp. 238-249.

Havron, M. Dean, Information Available from Natural Cues During Final

Approach and Landing, Human Sciences Research, Inc.,
Report HSR-RR-62/3-MK-X, March 1962.

Linton, Paul M., and Warren F. Clement, "Predicting Field of View Require-
ments for VSTOL Aircraft Approach and Landing," AGARD-CP-:268,
January 1980, pp. 40-1 through 40-17.

McColgin, F. H., "Movement Thresholds in Peripheral Vision, Journal of

the Optical Society of America, 50, 1960, pp. 774-779.

Naish, J. M., "Control Information in Visual Flight," Proceedings of the

Seventh Annual Conference on Manual Control, NASA SP-281, June
1971 (also McDonnell Douglas Paper No. 5921).

Noton, D., "A Theory of Visual Pattern Perception," IEEE Transactions,

SSC-6, 1970, pp. 349-357.

Roberts, Edward 0., External Visibility Criteria for VTOL Aircraft,

AFFDL-TR-67-27, March 1967.

Sanders, A. F., "Some Aspects of the Selective Process in the Functional

Visual Field," Ergonomics, 13_, 1970, pp. 101-117.

Semple, C. A., R. Hennessey, Mark Sanders, Bart Cross, Barry Beith, and
Michael McCauley, Aircrew Training Device Fidelity Peatures, Can-
yon Research Group, CRG TR-3041 B, February 1980.

Discriminations in Helicopter Control, HumRRO Professional Paper

No. 4-66, Presented to the Southwestern Psychological Association,

Arlington, Texas, 21-23 April 1966.

Wewerinke, P. H., Visual Scene Perception Process Involved in the Manual

Approach, NLR TR-78130, 13 October 1978.

"227

S'- '"'-"'--.-
---- %:-7- -ý*t
 GLOSSARY

The following definitions of terms are useful for understanding the

methodology for approaching simulator fidelity presented in this report.

hnhnduidth An indication of how quickly a system responds

to a command or disturbance. Bandwidth is us-
ually expressed in terms of frequency, i.e,, the
inverse of time, thus an alternative to expres-
sing system response in terms of bandwidth is to
use an effective response time or rise time.

Command Loop The outermost control loop necessary for the

execution of a given task or subtask. For ex-
ample, for a guidance task, the lateral command
K
loop would be heading; for a navigation task,
the lateral command loop would be track or
course deviation error.

Control (variable) Any of the variables which the pilot must man-
ipulate in order to fly the aircraft (or
simulator).

Crossover Frequency Numerically similar to bandwidth but, strictly

speaking, it is the frequency at which the open-
loop pilot-vehicle transfer flnction is equal to
unity. Crossover frequency can also be likened
to an effective loop gain.

Describing Ftuction A counterpart of the transfer function which is

restricted to the frequency domain and which may
be valid for only a limited range of ampli-
tudes. A describing function can be used to
portray any of the components of the pilot-
simulator or pilot-vehicle system.

Direction of Flight The directional aspect of a velocity vector can

include flight path angle, vertical velocity, or
any state variable combination equivalent to
them.

Disturbance Any unwanted variable which affects the aircraft

(or sitmlator) response.

228
 Essential Cue A cue which is required to close an essential
loop.

Essential Loop One of possibly several loops which are neces-

sary to perform a given task. A command loop is
an essential loop as are the various inner sup-
porting loops or control crossfeeds.

Equations of K~tion The set of mathematical relationships which

describe the behavior of a given system compo-
nent or the combination of several components.
A number of factors may be included in a set of

[ Explieit or
equations of motion such as Newton's second law
of motion, aerodynamic effects,
chomotor and cognitive behavior.

The specific and appropriate

and pilot psy-

behavior of a
Intrinsic Fidelity pilot which, if exhibited in a simulator
situation for a given task, would lead to suc-
cessful execution uf the same task in an actual
flight situation. Specifically, this involves
both the piloting technique and pilot perceptual
transfer functions essential for the task and is
therefore intrinsic to the task.

fttrinsic or See incidental fidelity.

Implicit Fideiity

Feedback The use of a portion of a system state variable

to influence a system control. The purpose of
feedback is usually to provide stability and to
minimize errors.

Feedfarinrd The direct application of a portion of a system

commgnd to a control. The purpose of a feed-
forward is usually to quicken response or offset
lags imposed by feedbacks. Another form of
feedforward is a crossfeed from one control to
another in order to reduce crosscoupling
effects.

Frequency Domain The expression of system dynamic properties with

frequency appearing as an independent var-
iable. For example, the quickness of a response
could be expressed in terms of "bandwidth,"
":natural frequency," eta.
Heave Damping An important aerodynamic characteristic whicb is
most commonly addressed in terms of the partial
derivative of specific z-force due to a unit
change in the component of aerodynamic velocity
along the z-axis, i.e., Zw. Heave damping

229

-- -- t-- .> k•-. - : *'- -- ••

 represents the basic flight path response of an
aircraft due to a rotation in attitude, activa-
tion of a direct z-force control, or the action
of a vertical gust component. Heave damping can
be estimated for any airplane or helicopter,
based only on gross weight and geometry of the
wing or rotor blades.

Incidental (or The counterpart of explicit fidelity involves

Extrinsic or Implicit) the specific der'iption of the simalator
Fidelity response and aircraft model which may contribute
to providing fidelity but which do not expressly
describe the piloting technique and pilot per-
ception whiLh have to be learned to perform the
task. Incidental fidelity can be roughly
equated to objective or engineering fidelity.

Inner Loop Usually a supporting loop for some task or

flight objective. The most common and the most
important inner loops are usually pitch attitude
and bank angle.

Intermediate Loop A supporting loop which is sometimes necessary

to bridge the gap between a piloting task and
the basic vehicle dynamici,. *For example, in a
navigation task, it is frequently necessary to
provide an intermediate guidance loop between
the outer navigation loop and the inner control
loop.

Kalman Filter A form of optimal estimation based on a quad-

ratic weighting of the uncertainty in
measurements relative to the variance of states.

Laplace Operator A way of representing tine derivstives or in-

tegrations with respect to time. The laplace
operator, s, is substantially equivalent to the
derivative operator, d/dt. Similarly, the in-
verse, lfs, is equivalent to fdt. Using Laplace
operators, linear differential equations can be
restated as transfex functions betw-.en controls
and state variables.

OJective Fidelixy Those characteristics of a training simulator

which can be quantified by direct measurement of
physical characteristics of hardware and soft-
ware.

Cuter loop A general term including command and inter-

mediate loops, In general an outer loop is more
influenced by the task description than by the
vehicle dynamics.

230
Perceptual Fidelity Characteristics of those aspects of a sitiulator
which the pilot uses to obtain essential cues or
characteristics of those behavioral aspects of
the pilots themselves.

Pilot Gain A numerical quantity which describes any of

several relationships between a given pilot cue
and the command or control actions resulting
from that cue.

Pilot Perception The transfer function which relates an actual

motion or visual quantity and the information
derived by the pilot about that quantity. Pilot
perception can involve threshold, response time,
distortion, or signal-to-noise aspects.

Piloting Technique The specific control laws exhibited by a pilot

which produce commands or control movements
based on the pilot's perception of stimuli.

State (Variable) Any of the dependent or independent variables

which describe the aircraft (or simulator) re-
sponse to controls and disturbances.

Time Domain The expression of aystem dynamic properties with

time appearing as an independent variable. For
example, the quickness of a response could be
expressed as a "rise time to 50 percent," a "lag
time constant," etc.

Training Objective Usually a specific piloting task or subtask

which is related to a specific aircraft or air-
craft type, operating environment, and involves
a particular skill level.

Transfer Function An operational mathematical expression of the

Aunctional dependence of one state variable upon
another or upon a control variable or a
disturbance.

231