International Journal of Engineering and Advanced Technology (IJEAT)

ISSN: 2249-8958 (Online), Volume-6 Issue-4, April 2017

Robust Analysis with Controller Design of

Forward-Velocity Dynamics of UAVs in Close
Formation Flight
Johnson Y, Imthias Ahamed T P

Abstract: Study of multi-UAVs (Unmanned Aerospace Vehicles) Wind and aero-perturbed conditions through the leader-
in close formation flight has received wide attention due to the follower aerodynamics cross coupling is also discussed.
significant advantages in resource mapping at greater swaths, 3- Section-IV describes the control and tracking of leader
D imaging etc. The control design study and its analysis is
carried out for two aerodynamically non-identical UAVs in a dynamics using PID and robust 𝐻∞ controllers, design of
leader-follower pattern of flight and it can be extended for multi weights in 𝐻∞ controller, the robust stability and
UAVs. The modeling includes the effect of leader-trailing-wing performance analysis of leader and follower forward velocity
vortex on the follower, both for nominal and perturbed system dynamics under various atmospheric conditions such as
dynamics. Robust controller design to maintain the relative nominal and perturbed, with respect to pole-zero diagrams,
velocity between two UAVs in longitudinal plane close formation
followed by some concluding remarks (section-V).
flight under nominal, wind and aero-perturbed condition is the
major objective of this article. The forward velocity control of the
leading vehicle and its tracking by a follower are performed by a II. RELATED WORKS
PID controller and then compared with that of a robust 𝑯∞
Several studies have been done in the flight control of
controller.
UAVs. The most important among them are constraint forces
Keywords: Formation flight, UAV, PID, robust H-infinity, approach [1], 3D potential field approach [2], intelligent
tracking and control, wing vortex.
management control approach [3], co-operative control [4,
5], vision based scheme [6] and constrained adaptive back
I. INTRODUCTION
stepping approach [7]. There are several tracking control
schemes available in literature. Proud, A. et al. had done
T he application domain of Unmanned Aerospace formation control on two aircrafts using PID feedback [8].
Vehicles (UAVs) has extended from reconnaissance to Vanek and Balint introduced practical approach to real-time
scientific data collection in academic research. Multi UAVs trajectory tracking of UAV formations using model
in close formation flight can enhance its ability to collect predictive control [9]. Saffarian and Fahimi present a
data more efficiently. A formation flight problem in modified leader-follower framework for achieving the
longitudinal plane focuses on the tracking issues of formation and propose a model predictive nonlinear control
longitudinal states of forward velocity, distance, angle of
algorithm [10]. A novel decentralized control design
attack and the pitch dynamics etc. of the leader vehicle in a
procedure is developed by Haibo Min which guarantees
leader-follower system. Among all the above states, the state
which mainly focused on this paper is the forward velocity. obstacle avoidance and collision [11]. Dogan et al. applied
Here the modification of the induced drag component on the non linear control approach [12] for formation
follower dynamics due to the wing vortex effect of the leader reconfiguration.
is also taken for the proposed controller design and analysis. The major beneficial effect of the close formation flight is
When the leader vehicle is subjected to heavy atmospheric the induced drag reduction. Lot of studies have been reported
wind effects in its flight, simultaneous perturbations may dealing with drag reduction approach. C. F. Chichka, J. L.
pass over to the follower dynamics. Unlike a conventional Spaeyer, C. Fanti and C. G. Park conducted a vast study
PID controller, a robust controller can maintain the control about peak seeking control for drag reduction in formation
and tracking with negligible error even in the presence of flight [13]. Drag reduction in extended formation flight [14],
external perturbations. This paper is organized as follows. simulation of aerodynamics cross-coupling vortex effects
After a brief literature review (section-II), the mathematical [15], computational aerodynamics of flexible wings for
modeling of UAV close formation dynamics is presented MAV [16], formation flight with greatest fuel savings
(section-III). The simulation and the analysis of trailing wing benefits for the trail aircraft [17] are some of the significant
vortex effect of the leader upon follower under nominal, studies dealing with wing vortex effects. The effect of heavy
wind (more than 10m/s) upon vehicles during flight faces a
Manuscript published on 30 April 2017.
* Correspondence Author (s) challenging problem in modeling. The simulation of wind
Johnson Y*, PhD Research Scholar, Department of Electronics and effects with flight data analysis has been done in [18]. The
Electrical Engineering, University of Kerala, Thiruvananthapuram studies referred above do not establish robustness to large
(Kerala)-695034, India. E-mail: [email protected]
Imthias Ahamed T P, Department of Electronics and Electrical
parameter uncertainties that are associated with such
Engineering, Thangal Kunju Musaliar College of Engineering, Kollam vehicles.
(Kerala)-691005, India.

© The Authors. Published by Blue Eyes Intelligence Engineering and

Sciences Publication (BEIESP). This is an open access article under the
CC-BY-NC-ND license http://creativecommons.org/licenses/by-nc-nd/4.0/

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number D4895046417/17©BEIESP and Sciences Publication (BEIESP)
Journal Website: www.ijeat.org 96 © Copyright: All rights reserved.
Robust Analysis with Controller Design of Forward-Velocity Dynamics of UAVs in Close Formation Flight

All these studies are suffered from the lack of 𝑦̅ 𝑧̅

𝑦1 = and 𝑧1 = are named as the non-dimensional
𝑏 𝑏
computational efficiency and easiness in the controller
formation geometry parameters. Here, 𝑦̅ and 𝑧̅ are the
design approach. In this paper, the existing robust control
relative distances in lateral and vertical directions
toolboxes in MATLAB. 𝜋
Are appropriately utilized to solve these multidimensional respectively between the vehicles. The value of 𝑦1 is since
4
problems and hence reduce the controller design 𝑦̅ is taken as 𝑏 ′ for achieving optimal drag reduction
complexities in the existing literature. formation geometry and the value of 𝑧̅ is taken as one-tenth
of ̅𝑦.
III. UAV CLOSE FORMATION MODELING BASED The change in follower angle of attack due to upwash is
ON AERODYNAMIC COUPLING
𝑊𝑢𝑤
The UAV in flight creates spiral like disturbances in its ∆∝𝑓 = tan−1
𝑉𝑙
vicinity. The cylindrically shaped spiral disturbances are The drag on follower is changed due to the lift vector
projected backward perpendicular to its wings on two sides.
The air moves in a circular pattern in these disturbance rotation ∆𝐷𝑓 = 𝐿𝑓 tan ∆𝛼𝑓 𝐿𝑓 = 𝑞𝑆𝑓 𝐶𝐿𝑓
pockets originating from the wing tips, enlarging the circles
and dissipate over time gradually. Another disturbance (𝑆𝑓 - follower wing area; 𝐶𝐿𝑓 - lift coefficient of follower,
effect is seen parallel to the wings moving up circularly 𝐿𝑓 - lift on follower wings). Drag during Formation Flight
towards UAV body. This disturbance phenomenon is called
(DFF) is the difference between the induced drag and its
horseshoe vortex. The air movement between the vertical
incremental change due to formation. 𝐷𝑓𝑓 = 𝐷𝑓 − ∆𝐷𝑓
disturbances at the two wing tips is downwash whereas in
𝐷𝑓 = 𝑞𝑆𝑓 𝐶𝐷𝑓 (𝐶𝐷𝑓 - drag coefficient (for follower). When
the outside vicinity of horseshoe vortex, it is a combination
of sidewash and upwash. The aerodynamic interaction ′′𝐶𝐷𝑓 ′′ is treated in close formation, it is defined by a new
between the vehicles in CFF affects the total flight term, 𝐶𝐷𝑓𝑓 .
behaviour. The reduction in the induced drag on follower
𝐷𝑓𝑓 𝑊𝑢𝑤
occurs when it took advantage of the upwash effect from the 𝐶𝐷𝑓𝑓 = = 𝐶𝐿𝑓 The incremental change in non-
𝑞𝑆𝑓 𝑉𝑙
wings of the leader. This is the advantage that a typical CFF
system demands. The upwash and sidewash of the air vortex dimensional drag coefficient ∆𝐶𝐷𝑓 = 𝐶𝐷𝑓 − 𝐶𝐷𝑓𝑓
from leader wing tips affect the flight profile of follower. The incremental change in the non-dimensional lift
The required fuel that drives follower in-formation is less as
compared to single flight. coefficient ∆𝐶𝐿𝑓 = 𝐶𝐿𝛼𝑓 ∆𝛼 𝐶𝐿𝑓𝑓 = 𝐶𝐿𝑓 + ∆𝐶𝐿𝑓

A. Close Formation Modeling

Where 𝐶𝐿𝑓𝑓 is the value of 𝐶𝐿𝑓 when UAVs are in
The upwash effect produces reduced drag on follower.
formation. Both lift and drag effects are taken for the
The corresponding longitudinal stability derivatives of the
determination of new stability coefficients, even though
follower are proportionately affected. A.W. Proud had
relative change in lift is very small compared to that of drag.
elaborated and derived the aerodynamic close formation
Moreover, since 𝐿 >> 𝐷, drag is influenced by the rotation
effects quantitatively on a typical leader-follower system
effect significantly than lift.
[19]. Some important equations useful for the present
analysis are taken from [19] and are described in the B. Longitudinal Simulation and its Analysis
following paragraphs. The flight profile simulation of the proposed formation
The approximated wing span of the follower for an flight system is done in two phases: nominal and perturbed.
elliptical lift distribution on a wing using horseshoe vortex In nominal phase, externally applied disturbances are
𝜋
approach is 𝑏 ′ = 𝑏. The symbol 𝑏 ′ is the effective wing completely absent which is an ideal case. But the induced
4
span. The average induced upwash due to leader’s vortices wing vortex effects from the leader are taken care of.
from both disturbance filaments originating from the Perturbed phase is more like a real situation in which a
leader’s wings is, limited range of change is applied in the values of some
selected longitudinal or lateral stability and control
Ґ𝑙 𝑦12 +𝑧12 +𝜇2
2
(𝑦1 −𝜋⁄4) +𝑧12 +𝜇2 derivatives. The selection of stability derivatives is based on
𝑊𝑢𝑤 = (ln [ 2 ] − ln [ 2 2 ]) their sensitivity or response to a particular open loop
4𝜋𝑏 (𝑦1 −𝜋⁄4) +𝑧12 +𝜇2 𝑦1 +𝑧1 +𝜇2
follower plant dynamics WVU YF-22 aircraft is taken as
here 𝜇 2 is the correction factor to account for the physical the leader & CONDOR HE-UAV as the follower for the
viscosity effects and Ґ𝑙 is the vortex strength of leader wing present simulation. The geometrical and aerodynamic data
tips which is given by
2𝐶 𝑉 𝑏
Ґ𝑙 = 𝐿𝑙 𝑙 𝑙 Where 𝑏𝑙 - for leader and follower are taken from references [20] and
𝜋𝐴𝑅𝑙 [21] respectively. The follower UAV, namely, Condor
leader wing span; 𝐴𝑅𝑙 - leader aspect ratio; 𝐶𝐿𝑙 - Hybrid Electric (HE) aircraft was designed and constructed
leader lift coefficient; 𝑉𝑙 - leader velocity. The formation by a Colorado-based UAV design firm named CL Max
velocity of the system is assumed as leader velocity in Engineering.
subsequent analysis. If 𝐶𝐿𝑙 value is not given, it is calculated
𝑐̅
by 𝐶𝐿𝑙 = 𝐶𝐿0𝑙 + 𝐶𝐿𝛼𝑙 𝛼 + 𝐶𝐿𝑞𝑙 𝑞 + 𝐶𝐿𝛿𝑒𝑙 𝛿𝑒
2𝑉𝑙

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number D4895046417/17©BEIESP and Sciences Publication (BEIESP)
Journal Website: www.ijeat.org 97 © Copyright: All rights reserved.
 International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249-8958 (Online), Volume-6 Issue-4, April 2017

The wings are of Eppler 210 airfoil configuration while its The available and estimated parameters given in Table-III
tail is configured with NACA 0009 airfoil. Due to modeling & IV are used to find the following aerodynamic quantities
limitations, the NACA 2412. using the formulae described in the beginning of this
Airfoil was substituted in place of the Eppler 210 since the session. These numerical calculations are crucial for
performance difference between the two is insignificant. understanding the aerodynamic interaction between the
The follower aerodynamic stability and control derivatives vehicles and the controller design in the following
are estimated and are given in Table-II [20, 21]. Its simulation.
developed controllers and other related accessories are not The calculated quantities under nominal conditions are:
considered for this academic simulation. The nominal lift coefficient of leader, 𝐶𝐿𝑙 = 0.118; vortex
aerodynamic profile given in terms of state space data is strength, Ґ𝑙 =2.38 𝑚2 /𝑠 ; average induced upwash,
performed at straight and level flight. 𝑊𝑢𝑤 =0.08; change in AOA, ∆∝𝑓 = 0.11 deg; follower lift,
The longitudinal dynamics equation for a single UAV in 𝐿𝑓 = 1803.6 N; change in drag, ∆𝐷𝑓 = 3.43 N; drag, 𝐷𝑓
state-space form portrays the multivariable coupling among =18 N; the Drag during Formation Flight (DFF), 𝐷𝑓𝑓 = 14.6
longitudinal states of forward velocity(u),angle of attack(𝛼), N; change in lift, ∆𝐿𝑓 = 0.0342 ; the Lift during Formation
pitch rate(q) and pitch angle(𝜃) with two inputs[22]. The
Flight (LFF), 𝐿𝑓𝑓 = 1803.63 N; modified drag coefficient of
terms are explained in Table-VII. For the leader, only one
input, i.e, the stabilator (elevator), is used and the follower by vortex effect, 𝐶𝐷𝑓𝑓 = 0.01.
corresponding data in steady and level conditions is given. The achieved reduction in drag, as per calculation,
For the follower, both elevator and throttle are used together is 20%. Thus the calculated modified drag coefficient of
having MIMO dynamics structure and its numerical data is follower is from the upwash wing vortex effects since the
also given. upwash is responsible for the angle of attack (AOA)
variation of follower's wings.
𝑢̇ 𝑋𝑢 𝑋𝛼 𝑋𝑞 𝑋𝜃 𝑢 𝑋𝛿𝑖1 𝑋𝛿𝑖2 (ii) Perturbed Aerodynamic Profile
∝̇ 𝑍𝑞 𝑍𝜃 ] [𝛼 ] + [ 𝑍𝛿𝑖1 𝑍𝛿𝑖2 𝛿𝑖1
[ 𝑞̇ ] = [ 𝑢 𝑍𝛼
𝑍 ][ ] The perturbations considered for the present
𝑀𝑢 𝑀𝛼 𝑀𝑞 𝑀𝜃 𝑞 𝑀𝛿𝑖1 𝑀𝛿𝑖2 𝛿𝑖2 simulation are (a) the atmospheric wind and (b) the variation
𝜃̇ 0 0 1 0 𝜃 0 0 in aerodynamic derivatives or coefficients.
Effect of Wind
𝑢̇𝑙 −0.2835 −23.0959 0.00 −0.17 𝑢𝑙 20.168
𝛼̇𝑙 The flight trajectory and the longitudinal stability
0.00 −4.1172 0.7781 0.00 ∝𝑙 0.5435 ] [𝛿 ]
= [ ] [ ] + [ derivatives of the UAV dynamics are changed due to its
𝑞𝑙̇ 0.00 −33.8836 −3.5729 0.00 𝑞𝑙 −39.08 𝑠
[ 𝜃𝑙̇ ] 0.00 0.00 1.00 0.00 𝜃𝑙 0.00 effect. The wind magnitude is limited to 5m/s. The effective
𝑢𝑓1 ̇ 𝑢 velocity contains the wind component. Assume that wind
−0.047 −0.039 0 −32.2 𝑓1 0 0.15
𝑤𝑓1̇ −0.374 −4.592 50.8 0 ] [𝑤𝑓1 ] + [−35.7 0 ] [𝛿𝑒 ]
= [ flows in the same direction of UAVs. The effective velocity
𝑞𝑓1 ̇ 0.0015 −0.7248 −2.62 0 𝑞𝑓1 −66.1 0 𝛿𝜏
̇ 0 0 1 0 𝜃𝑓1 0 0
is the summation of two quantities since the wind is
𝜃
[ 𝑓1 ]
assumed to have no lateral components. The aerodynamic
The follower longitudinal dynamics data contains parameters which depend on velocity are changed. The
dimensional stability and control derivatives. These calculated values of variables under perturbation are (the
derivatives are functions of dynamic pressure which by suffix ‘p’ in the notations of variables represent perturbed
themselves are functions of both local atmospheric density values):
𝑚2
(altitude) and velocity. The given follower CONDOR HE- 𝑉𝑙𝑝 = 47𝑚/𝑠; Ґ𝑙𝑝 = 2.67 ; 𝑊𝑢𝑤𝑝 = 0.09; ∆∝𝑓𝑝 = 0.109
𝑠
UAV undergoes a steady level flight of velocity 17m/s at a deg ; 𝐿𝑓𝑝 = 2258.6 N; ∆𝐷𝑓𝑝 = 4.3 N;
height of 366m. When it enters into formation with the
𝐷𝑓𝑝 = 22.58𝑁; 𝐷𝑓𝑓𝑝 = 18.28𝑁; ∆𝐿𝑓𝑝 = 0.043 N; 𝐿𝑓𝑓𝑝 =
leader which flies steadily with 42m/s velocity at a level of
2258.64 N; 𝐶𝐷𝑓𝑓𝑝 =0.0105.
120m above ground, two factors are used to modify the
present data in order to achieve close formation mechanism The velocity induced dynamic pressure variations have no
with dependent aerodynamic interactions. One is the ratio of effect in the leader lift coefficient since the value of the term
two velocities and the other one is the ratio of local 𝐶𝐿𝑞𝑙 is zero in this case. So the relative changes in lift
atmospheric densities. The density ratio is very small as characteristics due to the velocity increment are very small
compared to the velocity ratio and can assume to be compared to drag effects.
neglected in practical considerations. All the other non- Effect of changes in aero coefficients
dimensional stability and control derivatives and geometric Variations in aerodynamic stability derivatives usually
parameters like mass, wing area, chord length etc. will not occur during flight from their pre-flight estimated values. It
undergo any variations in these transformations. The may occur either in leader or follower dynamics or both. As
transformed [A] and [B] matrices of the follower at the a typical case study, consider a typical follower longitudinal
leader height and speed level are: forward velocity dynamics with elevator input plant, .
𝑢
𝛿𝑒
−0.11 −0.09 0 −32.2 0 0.815 From fig.3, The most sensitive three parameters in order are
𝐴 = [ −0.87 −10.7 118.36 0] ; 𝐵 = [ −193.6 0 ] 𝐶𝑚∝ , 𝐶𝑚𝛿𝑒 and 𝐶𝐿𝑞 .
0.0035 −1.69 −6.12 0 −358.98 0
0 0 1 0 0 0

(i). Nominal Aerodynamic Profile

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number D4895046417/17©BEIESP and Sciences Publication (BEIESP)
Journal Website: www.ijeat.org 98 © Copyright: All rights reserved.
Robust Analysis with Controller Design of Forward-Velocity Dynamics of UAVs in Close Formation Flight

IV. CONTROLLER DESIGN AND ITS ANALYSIS Table. I Desired Step Response Specifications for
Control and Tracking
The controllability and observability of the system
matrices are checked and it is assured that the system Serial
Time Domain Specifications Value
is both controllable as well as observable. The evaluation No
of control and tracking characteristics can be done better by 1 Rise time < 1 sec
considering the required specifications either in time or 2 % rise 90
frequency domain. The step response specifications in time 3 Settling time < 2 sec
domain desirable to the performance analysis are given in 4 % settling 3
Table-III. The first option selected for this moment is the 5 % overshoot 10
classical PID controller since it is known for well damped 6 % undershoot 2
response with minimum errors and its design simplicity.
Table II Geometrical Data of UAVs
A. PID Controller
The classical PID controllers are used at first to control
various states of leader and follower dynamics
independently and to track the leader states accurately under
nominal and perturbed conditions. If the results got with
PID are not satisfied with the required specifications, some
other control action having more robustness can be used
instead for the same purpose. PID parameter tuning is
performed with optimization technique. The closed loop unit
step responses of PID controlled leader forward velocity
dynamics plant under nominal and perturbed conditions are
given in fig.1. With the parameter values of 𝑘𝑝 =
−0.47, 𝑘𝑖 = 0.43 and 𝑘𝑑 =0.09, the achieved time response
specifications for leader forward velocity control under
nominal conditions are: rise time (𝑡𝑟 ) = 0.82 sec; settling
time (𝑡𝑠 ) =1.8 sec and overshoot = 8% . Its closed stability
is also checked and found to be stable with a gain margin
(GM) of [email protected] rad/s and a phase margin (PM) of
[email protected] rad/s. So PID is enough to control the given leader
forward velocity state in nominal conditions. But in one
perturbed condition (with changing 𝐶𝐿𝑞 ), unbounded
oscillations had resulted. In that case the stability of the TABLE-II Longitudinal Stability Derivatives
system is checked and found to be unstable with GM of
[email protected] rad/s and PM of [email protected] rad/s. So PID fails
at this moment. It is noticed that the PID designed for
nominal conditions is no longer able to maintain the
performance and robustness requirements for the given
perturbed conditions [24]. So some robust control
techniques should be applied for getting precise solution.

Fig.1: Unit Step Response of PID Controlled Leader

Forward Velocity Dynamics Under Nominal and
Perturbed Conditions.

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number D4895046417/17©BEIESP and Sciences Publication (BEIESP)
Journal Website: www.ijeat.org 99 © Copyright: All rights reserved.
 International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249-8958 (Online), Volume-6 Issue-4, April 2017

B. 𝐇∞ Controller state. 𝜔 (desired closed-loop bandwidth) : 90; M (desired

The basic object in 𝐻∞ control is a stable and proper bound on 𝐻∞ norm): 1; A (desired disturbance attenuation
1 1
transfer function. 𝐻∞ denotes the space of bounded analytic inside bandwidth): (for leader velocity control),
1500
(for
1706
functions in the open Right Half Plane (RHP). 𝐻∞ controller follower velocity control) and
1
(for leader velocity
helps to select a transfer function with least 𝐻∞ norm. In 510

graphical perspective, it is actually the minimization of the tracking)

peak in bode magnitude plots. Examples are the Sensitivity The Sensitivity (S) & Complementary sensitivity (T)
function, S(s) and the Complementary sensitivity function, functions for tracking forward velocity dynamics (under
T(s). Unlike the classical LQG control which minimizes the perturbation with 70% of 𝐶𝑚𝛿𝑒 value) are shown graphically
closed loop peak for known input signals, H∞ control deals as Singular Value (SV) plots in fig.3. The shape of S and T
with unknown inputs, resulting in ‘flat’ Bode magnitude verified the rules regarding its required shape.
plots. Loop-shaping techniques are used to impose particular
shapes to the output responses.
The given 𝐻∞ control problem is formulated to find a
robust controller for the generalized plant such that
∞ −norm of the transfer function relating d to z is
minimum. The robust 𝐻∞ controller is chosen for the control
and tracking because it can maintain stability not only in
nominal plants, but for real plants also [25]. A
𝐻∞ controlled plant with appropriate frequency dependent
weights is shown as a block diagram in fig.6. The transfer
functions of plant and weights should be proper and stable,
otherwise the stabilizability conditions of augmented matrix
(plant transfer function with weights) along with the gain Fig. 3: Singular Value plot showing S & T.
will not be satisfied. The weights are designated as W1, W2
The leader and follower output responses obtained in
and W3. W1, known as sensitivity weight or performance
various conditions with the help of newly designed 𝐻∞
weight, penalizes the error signals. The control weight W2
controllers are given in figures from fig.4 to fig.6. The
penalizes the control signal. The output signal is penalized
forward velocity of the leader UAV is controlled within less
by complementary sensitivity weight or stability weight W3.
than 0.1 seconds as shown in the top part of fig.4. When the
simulation is repeated with 30% change in the parameter 𝐶𝐿𝑞
in the forward velocity plant transfer function, only 0.02
seconds is delayed in its control in the perturbed state from
the nominal one. In the follower forward velocity control
simulation, the time delays of only 0.01 and 0.03 seconds
can be noticed with 30% change in 𝐶𝑚𝛿𝑒 and 𝐶𝐿𝑞 parameters
[C–controller; P–plant;r - reference;e – error;u – plant input;y – actual
output;z1 – desired error signal;z2 – desired input signal;z3 – desired respectively in the forward velocity transfer function. An
output signal] enlarged view of a small section of the top part of fig.4 is
Fig.2: 𝑯∞ - Controlled Leader Tracking Dynamics Plant. given in fig.5. This leader forward velocity in nominal
For multivariable (MIMO) transfer matrices, 𝐻∞ norms are conditions can be successfully tracked by a perturbed
described with the concept of singular values. follower UAV with a small time delay of 0.02 sec, as shown
in fig.6.
The well known parameter selection rules for 𝐻∞ control
weights are reproduced below [26].
(a) For achieving better performance i.e. good
disturbance rejection, design W1 such that it should be small
inside desired control bandwidth. So it requires small value
of sensitivity function (S) in low frequency region. Keep in
mind that 1/W1 should have the general shape of ‘S’. The
sensitivity function S can be suitably shaped by properly
designing the weighted norm ‖W1(s). S(s)‖∞ .
(b) In order to achieve good stability margin i.e.
robustness, select W3 such that it should be small outside
desired control bandwidth. So it requires small value of
complementary sensitivity function (T) in high frequency Fig.4: Unit step response of (a) leader (top) and (b)
region. The weight 1/W3 should have the general shape of follower (bottom) 𝑯∞ controlled forward velocity
‘T’. For robust stability requirement, the infinity norm of the dynamics under various conditions
weighted ‘T’ function should be equal or less than one, i.e.,
‖W3(s). T(s)‖∞ ≤ 1. Based on the design rules (a) & (b)
above, the weights are designed such that the desired
parameters have the following values for forward velocity

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number D4895046417/17©BEIESP and Sciences Publication (BEIESP)
Journal Website: www.ijeat.org 100 © Copyright: All rights reserved.
Robust Analysis with Controller Design of Forward-Velocity Dynamics of UAVs in Close Formation Flight

occurred. With the development of tracking 𝐻∞ controller

the perturbed follower advances by 0.01 second in its
tracking of the leader compared to the case of without
tracking controller. The comparative study detailed above is
summarized in Table-IV.

V. CONCLUSION
The close formation flight UAV system consists of two
aerodynamically non-identical UAVs. Their longitudinal
stability and control derivatives are determined from the
flight profile and geometric data. The trailing wing vortex
Fig.5: Enlarged view of a small section of top part of effect of the leader upon follower dynamics under nominal
Fig.4. and wind and aero-perturbed conditions are numerically
calculated. The forward velocity dynamics is particularly
focused and their individual control and tracking issues are
tuned with a conventional PID and robust 𝐻∞ controllers and
their performance are compared. The robustness property of
𝐻∞ at the cost of high control input is also revealed.

REFERENCE
1. Zou, Y. and Pagilla, P.R., “Distributed Formation Flight Control
Using Constrained Forces”, J. of Guidance, Control and Dynamics,
Vol.32, No.1, 2009
2. Paul, T., Krogstad, T.R. and Gravdahl, J.T., ‘Modeling of UAV
Formation Flight Using 3D Potential Field’, Simulation Modeling
Practice and Theory,Vol.16, Issue 9,pp.1453-1462,2008
3. Innocenti, M., Guilietti, F. and Pollini, L., ‘Intelligent Management
Fig.6: Unit step response of leader nominal forward Control for Unmanned Aircraft Navigation and Formation keeping’,
velocity and its tracking by follower under perturbed RTO AVT Course at Belgium, May, 2002
condition. 4. How, J., King, E. and Kuwata, Y., ‘Flight Demonstration of
Cooperative control for UAV Teams’, AIAA 3rd Unmanned
Table IV: The Comparative Study Unlimited Technical Conference, Workshop and Exhibit, Illinois,
2004
5. Yoon S., Bae. J, and Kim. Y, Cooperative Standoff Tracking of a
Moving Target using Decentralized Extended Information Filter,
KSAS (Korean Society for Aeronautical and Space Science) Spring
Conference, Gyeongju, Republic of Korea, Apr 2011
6. Kim H.J., Kim. M, Lim. H, Park. C, Yoon. S, Lee. D, Oh. G, Park. J
and Kim. Y, ‘Fully-Autonomous Vision- based Net-Recovery
Landing System for a Fixed-Wing UAV’,IEEE/ASME Transactions
on Mechatronics, accepted for publication, 2013.
7. S. Yoon, S. Park, and Y. Kim, ‘Constrained Adaptive Backstepping
Controller Design for Aircraft Landing in Wind Disturbance and
Actuator Stuck’,International Journal of Aeronautical and Space
Sciences, Vol. 13, No. 1, pp. 101-116, 2012.
8. Proud, A., Pachter, M., and DAzzo, J. J., ‘Close Formation Control’,
Proceedings of the AIAA Guidance, Navigation, and Control
Conference, AIAA - 99 – 4207, pp. 123 1 - 1246, Portland, OR,
C. Comparative Study and the Stability Analysis August, 1999.
9. Vanek and Balint, ‘Practical approach to real-time trajectory tracking
The comparative stability analysis of the overall system Of UAV formations’ ,Proceedings of the American Control
dynamics will help to get better concepts about system Conference,2005.
stability and control. The performance comparison can be 10. Saffarian, M. and Fahimi, F., ‘Control of helicopters formation using
non-iterative nonlinear model predictive approach, Proceedings of
done by comparing the responses in fig.1 and fig 4(a) along IEEE American Control Conference, pp. 3707-3712, Seattle,
with fig.5 in terms of the settling time behaviour in the time Washington, June 2008
domain characteristics of the leader forward velocity 11. Min,H., Decentralized UAV formation tracking flight control using
trajectories in nominal and perturbed (30% change in 𝐶𝐿𝑞 ) gyroscopic force, IEEE International Conference on Computational
Intelligence, 2009.
conditions with PID and 𝐻∞ controllers. The leader with 12. Dogan,A. and Venkataramanan,S., ‘Nonlinear control for
PID in closed loop requires 1.8 seconds to settle its forward Reconfiguration of Unmanned Aerial Vehicle Formation’, J. of
velocity step response which is within the specified time of Guidance, Control and Dynamics, Vol 28, No 4(2005)
13. Chichka D.F.,Speyer J.L.,Fanti C. and Park C.G., Peak-seeking
2 seconds. But in 30% 𝐶𝐿𝑞 condition the unstable Control for Drag Reduction in Formation Flight, AIAA Journal of
unbounded oscillations did not lead to proper settling. Guidance, Control and Dynamics,University of California,2002.
𝐻∞ controlled leader requires only 0.1 second to settle in 14. Ning S.A., Aircraft Drag Reduction through Extended Formation
nominal conditions as compared to 1.8 seconds of the PID Flight,PhD thesis,Dept of Aeronautics and Astronautics,Stanford
University, August 2011.
counterpart. But in the perturbed conditions 𝐻∞ produces
astonishing performance with settling time of only 0.12
second against the unbounded unstable behaviour with the
PID counterpart. In the case of follower similar results had

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number D4895046417/17©BEIESP and Sciences Publication (BEIESP)
Journal Website: www.ijeat.org 101 © Copyright: All rights reserved.
 International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249-8958 (Online), Volume-6 Issue-4, April 2017

15. Saban,D.,Whidborne,J.F. and Cooke A., Simulation of Wake Vortex

Effects for UAV in close formation flight, Aeronautical Journal,
Volume 113, Issue 1149,pp. 727-738,2009.
16. Shyy,W.,Lian Y.,Tang J.,Liu H.,Trizila P.,Stanford P.,Bernal
L.,Cesnik C.,Freedmann P. and Ifju P., Computational Aerodynamics
of Low Reynolds Number Plunging, Pitching and Flexible Wings for
MAV Applications, Proceedings of the 46th AIAA Aerospace
Sciences Meeting and Exhibit 7 - 10 January 2008.
17. Morgan M.T., A Study in Drag Reduction of Close Formation Flight
accounting for trim control flight positions and Dissimilar
Formations. ,thesis,M.S.Aeronautical Engineering,Air force Institute
of Technology, March,2005.
18. Lewis T.A., Flight Data Analysis and Simulation of Wind Effects
during Aerial Refuelling. ,thesis, M.S. Aerospace Engineering,
University of Texas,Arlington,May,2008.
19. Proud, A.W., “Close Formation Flight Control”, M.S. Thesis in
Electrical Engineering, Air Force Institute of Technology, Air
University, March, 1999.
20. Campa,G,Gu,Y.,Design and Flight Testing of Non-linear Formation
Control Law,Science Direct,Control Engineering Practise 15,pp-
1077- 1092,2007.
21. Christopher G., “Modeling, Simulation and Flight test for Automatic
Flight Control of the Condor Hybrid-Electric Remote Piloted
Aircraft,” Thesis, M.S. Aeronautical Engineering, Presented to the
faculty, Departments of System Engineering, Aeronautics and
Astronautics, Air University, March, 2012
22. Nelson, R.C., “Flight Stability and Automatic Control”, Tata
McGraw-Hill Publishing Company Ltd., New Delhi, 2 nd edition,
2007.
23. Johnson Y. and Dasgupta S., “Robust Hurwitz Stability and
Performance Analysis of H-Infinity Controlled Forward-Velocity
Dynamics of UAVs in Close Formation Flight Using Bounded Phase
Conditions in a Kharitonov Framework’, Int. J. Inst. Eng. India,
Springer Publications, Series. C, vol 95, Issue 3, pp 223-231, July
2014.
24. Johnson Y. and Dasgupta S., “Control and Tracking of Longitudinal
Dynamics of UAVs in Synchronized Motion,” Conference
Proceedings, IEEE, International Conference on Control
Communication and Computing (ICCC), Trivandrum, 2013
25. Scherer C., “Theory of Robust Control”, Delft University of
Technology, The Netherlands, April 2001
26. Johnson Y. and Dasgupta S., “Pitch Attitude Dynamics Control and
Tracking of UAVs in Close Formation Flight,” Int. Conf. on
Emerging Trends in Electrial Engineering (ICETREE-14), Kollam,
Kerala, India, Conference Proceedings, Elsevier Publications, 2014.

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number D4895046417/17©BEIESP and Sciences Publication (BEIESP)
Journal Website: www.ijeat.org 102 © Copyright: All rights reserved.