You are on page 1of 8

TUB AERONAUTICAL JOURNAL

Vol. 108, No. 1081, March 2004, pp. 117 124

Dynamics and Control of Different Aircraft


Formation Structures
Fabrizio Giulietti*
University of Bologna, 47100 Forli, Italy
and
Giovanni MengalP
University of Pisa, 56122 Pisa, Italy
The problem of aircraft formation dynamics and control is investigated from the viewpoint of
formation architecture. Three different, formation structures, Leader-Wingman, Virtual Leader and
Behavioral approaches are introduced. A comparative study is made using an unified approach through a
suitable control law. The formation systems are analyzed on a quantitative basis and objective results are
made available for the designer. The trade-off between system performance and complexity is indicated. A
complete nonlinear simulation involving a flight-path change and a heading change manoeuvre is
discussed. Results show the superiority of a Behavioral approach to maintain close formations of vehicles.

N omenclature n — load factor


A = state matrix
t — time
B = input matrix
Cr>a = zero-lift drag coefficient x, y,z — inertial frame components
CD = drag coefficient V — velocity vector
CL = lift coefficient
5, £ — error vector components
D — drag force
d = relative distance vector
Kv,K^,K-, = trajectory gain components
Ka — position gain matrix e = error vector
Kx = trajectory gain matrix f,g,h — non-linear dynamics functions
Kdx. Kdv, Kdz = position gain components
r = position vector
L = lift force
P = Lyapunov matrix u - input vector

S = wing surface x — state vector


T = engine thrust
P,Q,R = angular velocity components
V = airspeed
Q ~ cross-product matrix
Y = diagonal matrix
R = rotation matrix iff, J, — heading, flight-path and roll angles
TTwv = local vertical to wind-axes rotation matrix p = air density
7} = inertial frame
a? — angular velocity
Tv = local vertical frame
— = desired
Tw — wind-axes frame
</ = dynamic pressure — time derivative
g = gravitational acceleration T — transposed
k = induced drag factor
G = Formation geometry center
m = aircraft mass
’Post Doctoral Fellow. Department of Electronics, Computer Science i — i—th aircraft
and Systems. E-mail: fabrizio.giulietti@mail.ingfo.unibo.it
T
Associate Professor, Department of Aerospace Engineering. E- mail: r = reference
g.mengali@ing.unipi.it

VL = Virtual Leader

I
arc 2 GIULIETTI AND MENGALI
W Ti-Di , . (
Vi =-----------------1- g sin ytIntroduction3) architecture. The approach is inspired by that proposed by
Aircraft formation control has become an active research area Beard ct al. for spacecraft formation control. 17 An important
• in recentg years.
ni&iTupi (
In a manner similar to migratory birds, aircraft in result is that in this way different formation systems arc
Vi ccsmay
a formation q* experience a substantial
4) drag reduction using compared on a quantitative basis and objective conclusions arc
q»the
= vortex upwash created
(rii cosept by the leading( aircraft. 1 This in turn
— cosqi) made available for the designer. In particular the tradeoff
implies a potential fuel reduction with 5) significant benefits for between system performance and complexity is pointed out.
where military and commercial employments. Further important Although the stability of the formation dynamics is guaranteed
aspects motivating
a Li the growing attention towards this subject arc only for the linearized model, all the simulations arc performed
as follows.ni =------- (
It is believed that aircraft formation control will play in a nonlinear framework and without simplifying assumptions.
a fundamental role in future aerospace 6) scenarios, where From this viewpoint the paper represents an improvement over
unmanned air vehicles will be required to maneuver in swarms existing approaches where linear aircraft dynamics or
fori surveillance, reconnaissance and rescue purposes in hostile formation planar models arc employed. 5,7
environments.2 Finally, autonomous formation flight systems arc
widely investigated due to their close connection ( to the aerial
refuelling problem.3’4 '
Most of the existing literature concerning the problem of aircraft Formation Modeling
5-9
formation control and/or aerodynamic interference
modelling,10-12 assumes a fixed formation structure. By contrast, A three dcgrccs-of-freedom point mass model is chosen to
in this paper the aircraft formation control is analyzed using represent each aircraft in the formation. This simplified model,
different formation arrangements in order to establish a link involving only the aircraft’s slow states, is well suited for
between system performance and complexity. In particular, we dealing with trajectory tracking problems and position keeping
concentrate on Leader-Wingman, Virtual Leader and Behavioral autopilots.
approaches, that represent the three main structures reported in Reference lYaines
the literature. 9'13114 In the Leader-Wingman structure the first
aircraft in the formation is designated as the leader, whereas the Aircraft dynamics arc described with the aid of two right-
remaining aircraft (wingmen) arc viewed as followers. While the hand reference frames: A local vertical frame Ty and a wind-
leader keeps a prescribed trajectory, the followers refer their axes frame Tw ■ Both have their origins in the instantaneous
position to the other aircraft in the formation. More precisely, position of the aircraft. It is assumed that 7v has axes parallel
they keep a fixed relative distance from the neighboring aircraft to an inertial frame 7}. As usual, the axis xw of 7U- is directed
and, in doing so, they form a chain. Although a rear aircraft along the velocity vector V and the axis zw lies in the plane of
usually exhibits a poorer response than its reference due to error symmetry of the aircraft.
propagations, for its simplicity the Leader-Wingman structure is The transformation between the vertical frame Ty and the
widely employed in control and management of multiple vehicle wind-axes frame Tw is defined through the following rotation
formations.5-9 In the Virtual Leader structure, instead, each matrix (let. c = cos, s = sin):
aircraft receives the same information, that is, the trajectory of
the V^rtu^TCcadciTT^Tiirlat'tcr inay -Kca-rcal aircraft or, more cqc^> cqs-^ sq
commonly, an ideal point in the formation that aircraft must TTwv = —cip tnp — stp sq c<p c^ — S(p sq stpcq
track. The advantage of a Virtual Leader structure is twofold: 1) c$ s^> c<pcq
all aircraft in the formation exhibit the same transient due to the s<p s^» — c<p sq —s<p cip — cep
absence of error propagations and, 2) the formation behavior is where xp.q and tp correspond to aircraft heading, flight-path
prescribed by simply specifying the behavior of the Virtual and roll angles. For convenience we choose the flight-path
Leader. The disadvantage is that there is no explicit feedback to angle q with the sign reversed with respect to the usual
the formation. Indeed, each aircraft has no information about its notations. This means, for instance, that the aircraft rate of
distance from the followers and therefore may not be able to climb is given by —V sin q, where V is the aircraft velocity.
avoid collisions. A quite different strategy is represented by the Also, wo use the following notation: If w is an arbitrary vector,
Behavioral approach. In this structure, the control action of each its components [w] arc transformed from 7vv to Ty through the
aircraft is a weighted average of the control for each behavior. In relation
the context of aircraft formation control this idea, has been first
introduced by Anderson and Robbins. 13 Then, it has been further [w]w = Twv [w]v (2)
exploited by Giulietti ct al.15 with the introduction of an
imaginary point in the formation called the Formation Geometry Aircraft Equations of Motion
Center (FGC). Each aircraft in the formation is required to The equations of motion of i—th aircraft in the formation
maintain a prescribed distance from the FGC. The position of
this point depends on the relative distances between all aircraft in
the formation. This allows each aircraft to have the capability of
sensing other vehicles’ movements from the nominal position
and to maneuver in order to reestablish the formation geometry.
This characteristic is very similar to the natural behavior of
migratory birds.15 The strength of this approach is that there is an
explicit feedback to the formation so that problems of collision
avoidance arc easily taken into account. Another important
advantage is that aircraft in a formation tend to cooperate with
each other so that there arc no points of failure in the structure.
On the other hand, the behavioral approach is more complex
than the others as it requires exchanging more information.
In this paper the above formation structures arc analyzed in a is the load factor, L, and Di arc lift and drag forces, is the
unified framework through a suitable control system engine thrust, m, is the aircraft mass and g is the gravitational
acceleration. Using the aircraft drag polar, the drag coefficient
CjOi is given by
CDi = CDoi + kiClt (7)
s
where is the lift coefficient, Ci>Oi’ zero-lift drag coefficient
and ki is the induced drag factor. The aerodynamic
’0 - Qi
= Ri 0 -Pi GHILIETTI AND MENGALI 3
Pi
L 0
where is the angular velocity of the wind-axes frame relative to
is the cross-product matrix an
d the local vertical frame Tv- To stress the fact that vectors arc
Note that: evaluated in the wind-axes frame of the i— th aircraft, Eq. (13)
T^v — Twy(<pi, 7i, ipi) ( is written as:
19)
and
Twrv = 1Twv(<p, 7, ip) (
where Twv is given by Eq. (1), one has: 20) W w, = »■. - [di “ 41W 1 P’] w,
(14>
( where
21)
and (15)
[V]„r=TWrv[V]v (
22) ( [Pi, Qi, Ri]T.
[v]w,-R[v]wv
23)
where ( Pi^(pi + tpi sinqi (16)
R — 1Twiv'n'itrv 24)
and Qi = —ji cos cp-i + ipi cos qi sin ipi (17)
(
[V]w, = [V, 0, of 25) Ri — ipi cos 7i cos <pi + 7i sin <pi (18)
FigureF1r +Definition ( actual (A) and desired
dt = Fi of geometric quantities with Assume that the desired orientation of the i—th aircraft
(£>) positions of i— th aircraft 10) coincides with that of the frame Twr and that the corresponding
and
Fr + di = Ti ( desired velocity Vi equals the velocity of the reference point V.
Defining the rotation matrices:
from which we 11)
get: drag force is computed as:
di — di — n — ri
(
12) 2
n _ c.r -Lt (T^”bg)
-C'i — Qi Di + Ki - g
(8)

where q, — jpV)2 is the dynamic pressure, p is the atmospheric


density and Si is the wing surface. Variations in aerodynamic
forces due to the trailing vortices for each aircraft have been
computed with a distributed horse-shoe vortex technique. 1,18
Such effects are implicitly contained in the expressions of
force coefficients— ------------------------------------ -—
Introducing the state vector Xi = [Vi, ipi, 7i]T and the input Using (21) and (22) gives the desired velocity vector in the
vector Ui = [Ti, m, (pi]1', the aircraft equations of motion (3)- reference frame Fwf
(5) can be written in standard form as:
Xi = f,[x,,Ui) 0)

Position Kinematics

One of the main issues in a formation flight control system


is to provide each aircraft with the ability to maintain a
prescribed distance from a certain reference point G. This may Substituting Eq. (23) into (14) and dropping for simplicity the
coincide with the formation leader (cither real or virtual), a square brakets, the position dynamics of the i—th aircraft
neighboring aircraft, or a certain point witliin the aircraft relative to the reference point in the frame arc:
formation.
To establish a suitable mathematical model, assume that F£ di = Vi-IRV-fii (di-di) +di (26)
and rr arc the positions of i—th aircraft and reference point
with respect to the origin O of the inertial frame. Let di be the Note that, if we introduce the state vector of the reference
relative actual position between G and the i—th aircraft, while system x = [V, -<p, 7] , the position dynamics of the i—th
the corresponding desired position is represented by an aircraft with respect to the reference point may also be wit- ten
overbarred symbol. Also assume that an orthogonal frame Twr in compact form as:
is “attached” to the reference point G whose orientation is
defined through Euler angles <p, 7, ip and lot V be the velocity di = ff(xi,x,di,di,Ui) (27)
of G. With reference to Fig. 1 the following relations arc
obtained:

Formation Control
This section outlines the formation control system design.
It is assumed that standard autopilot loops have been closed
around each aircraft in the formation. Our aim is to define
a formation control law that is able to fulfil simultaneously
two different tasks: 1) trajectory tracking, and 2) position I
Taking the derivative of (12) with respect to the Twt frame keeping. In the following, a suitable control vector in the
yields: form
di — di + Wi x [di — di) = Vi — Vi (13) U£ —— h>[Xi, X, di, di ) (28)
4 GIULIEITI AND MENGA1J

is used to guarantee that each aircraft in the formation has the Leader-Wingman structure is:
following dynamics:
di — g(xi,Xi-i,di,di,Ui)
= —Kx(wi - ®) - Kd(di - di) (29) Xi = f(Xi,Ui) (39)
Hi — h ( X i, X i — 1, di, dri)
where
0
0 where the subscripts i — 1 indicates the neighboring aircraft.
K* (30)
0 0 Virtual Leader Structure
and (31) The Virtual Leader structure can be viewed as a variation of
[Kdx 0 0
the Leader-Wingman structure. Each aircraft is expected to
Kd = 0 0 0 follow an imaginary point in the formation (the Virtual
0 KG. Leader), while keeping a prescribed relative distance from it.
arc positive definite gain matrices. The main difference with respect to the Leader-Wingman
To get the required control efforts a dynamic inversion ap- structure consists in that each aircraft receives the same
proach is employed. More precisely, the commanded rates in trajectory information i.c., the Virtual Leader position. A
Eq. (29) arc set equal to the actual rates in Eqs. (3)- (5) and a mathematical model for this system is given by:
solution is found for thrust, bank angle, and load factor. From
flight-path and heading equations we obtain respectively: dj — g(xi.xvL-di.di.Ui)
Xi = f{xi,Ui) _ (40)
nicostpi = — \K^(yi — 7) 4- Kdz(dZi - dZi)\ 4- cos7,- = ft 9 «i = h{xi,XvL,di, di)
(32) where the subscript VL indicates the Virtual Leader.
and
Behavioral Approach (Formation Geometry Center Structure)
m sin tpi - — [-K^i -tp)~ Ka^dy, - dyi)] cos^i = f2
The Behavioral approach is a quite different strategy with
(33) respect to those described previously. In this ease each aircraft
Note that the ratio between Eq. (33) and Eq. (32) provides the is required to keep a relative distance from an imaginary point
required bank angle command: (the Formation Geometry Center) whose dynamics depends on
all the aircraft positions in the formation. This allows each
Pi — arctan j (34) aircraft to “perceive” the movement of the other vcliicles from
the nominal position. In the presence of disturbances, for
while the sum of the squares of Eqs. (33) and (32) gives the instance, if an aircraft loses its position, the others sense a
required load factor: change in the formation structure. As a result, they depart
momentarily from the prescribed trajectory, manoeuvring so as
to reconstitute the formation gcomctry.TIic mathematical
= \/ fi2 + /22 (35) model for the generic aircraft in a na-aircraft formation is given
The thrust command is obtained from the following equation: by:

-Kv(Vi - V) - Kdx(dXi -dXi) = Ti^Di -g^i (36) di — g (a>-i, x G, d,, di, u-t)
Xi = f(Xi,Ui) (41)
Solving for the thrust produces:
Ui — h(xi ,X(;,di, di)
Tj = — miKv{Vi — V) — miKdx(dXi — dXi) — rmg sin7,
(37) where XG = [VG VG 7«]r is the state vector of the Formartion
where Di is computed through Eq. (8). Clearly, Eqs. (34), (35) Geometry Center. It is assumed that:
and (37) arc equivalent to (28). To summarize, the closed loop
dynamics of the i—th aircraft arc given by [see Eqs. (9), (27) Vb = 4- VG h
and (28)]: ipa — arctan^c/dro) (42)
di = g(xi,x,di,di,Ui) 7G — arcsin^o/Vo)
in = f(xi,Ui) _ (38)
Ur,—— h(a?^, d„. di) where, by definition
~XG 'Vi cos 7» cos tpi
~ Vi cos 7i sin ipi (43)
VG . Visinq-i J
The form of the closed loop dynamics (38) is quite general and ZG.
encompasses different formation structures. Some of the most Eq. (43) shows that all aircraft contribute to establishing the
significant structures arc now described. Geometry Center position.
Note that the equations of motion for the three formation
structures have the same form as expressed in Eqs. (38). This
allows us to treat the stability problem in an unified
Formation Structures framework. This subject is discussed in the next section.
Leader-Wingman Structure
The Leader-Wingman is a commonly employed structure
for control and management of multiple-vehicle formation. The Formation Stability
leader aircraft directs the whole formation while keeping a
prescribed trajectory. The other aircraft (wingmen) have to We turn our attention to the closed loop dynamics (38)
maintain a given distance from the neighboring ones. The aiming to investigate its stability properties. With the aid of
mathematical model for the generic aircraft in the the Eqs. (26) and (29) the mathematical model may be re writ-
GIULIETTI AND MENGALI 5

Figure 3 Desired trajectory

Figure 2 Formation arrangement


ramp-like flight-path change is produced over t G [40,451 s.
Second, a ramp-like heading change of —60° is generated over
t G [70,100] s. Figures 4-9 show the simulation results. Figure
ten as 4 illustrates the z—axis distance error of the aircraft in a
di-^di = -Qi(di-di) + Vi-RV (44) Leader-Wingman structure during the flight-path change
manoeuvre. It points out the performance limits of such a
Xi — —Kd(di — di) - Kx(xi — x)-tc (45) structure due to the error propagations along the vehicle chain.
Clearly the front aircraft has a better response than the rear
Eqs. (44)-(45) arc useful for simulations, but in order to prove aircraft. Figures 5-7 compare the three formation structures by
the system stability, a linearized version of Eq. (44)- is showing the y—axis distance error for each aircraft during the
employed. More precisely, if 0 represents the trim value and heading change manoeuvre. Note that the Behavioral and the
A(-) the generic first order variation, the linear approximation Virtual Leader structures have the same characteristics for all
of Eq. (44) is given by the three aircraft in the formation, while, once again, the
Leader-Wingman structure has a degrading performance from
Ad, - &di = y(Axi - Ax) (46) the front, to the rear of the formation.
It is interesting to see what happens when a system failure
where it has been assumed that Qp — 0aX3 and occurs. Suppose for instance that the central aircraft (Aircraft 2
in Figure 2) cannot keep the prescribed position in the
0 O'
formation. In our example this happens when its feedback with
% 0 (47) the distance measurement fails, or Kd = 0. The simulations arc
0 Vo. shown in Figures 8, 9 and 10.
is a positive definite diagonal matrix. Correspondingly, Eqs. Figure 8 illustrates the Leader-Wingman ease. As anticipated,
(44)-(45) become the Aircraft 2 cannot keep its distance relative to Aircraft 1
and, in fact, the position error shows up after the heading
5i = change manoeuvre and propagates in the formation. Although
(48) Aircraft 3 has no steady-state relative error, it docs have an
& = -Kd5i - ®
absolute position error because it looks at Aircraft 2.
Let us now concentrate on Figure 9 dealing with a Virtual
where Si = Adi — Adj, = Ax, — Ax. The stability of the Structure. The time histories look similar to those of the
linearized system (48) around the trim condition is guaranteed Leader-Wingman ease but an important difference exists. As in
by the following theorem. the previous ease Aircraft 2 cannot keep its position and
accumulates a position error after the heading change
Theorem 1 Define the error vector ei = [<5^, and manoeuvre. However, now Aircraft 3 (which now looks at the
suppose S € £2(0,00) A LoofO.oo). Then, —> 0 as t — Virtual Leader trajectory) has no absolute position error. This
>• is an important point because it says that each aircraft exhibits
00. the same transient. At the same time it also implies that a
Virtual Structure formation docs not possess a real formation
Proof Sec Appendix. feedback and, therefore, a protection against collisions cannot
be guaranteed.
Simulation Results As Figure 10 shows, the situation in the Behavioral structure
The nonlinear equations (38) have been employed in com- case is quite different. Here the capability exists of sensing
paring the properties of the three formation structures. Each each aircraft movement from the prescribed position. As a
formation is made of three F16 aircraft models and the general result, when Aircraft 2 takes a wrong position, Aircraft 1 and 3
arrangement is shown in Figure 2. The aircraft parameters arc manoeuvre so as to remove the position error with respect to
summarized in Table 1. The simulations deal with a trajectory their common reference (the FGC point). Therefore the
tracking task. Starting with a velocity of 200 m/s and a height Behavioral approach is the best way to obtaining close vehicle
of 6000 m each aircraft is required to perform a flight-path formations and allows both adequate trajectory tracking and
change and a heading change manoeuvre. The commanded geometry formation keeping. On the other hand it needs more
trajectory is represented in Figure 3 and is described as follows. information because each aircraft has to know all the aircraft
First, a ramp-like flightpath change of —6° is applied over t e states.
[25,30]s. The value 7 = —6° is then kept until t = 40s when a
+6°
6 GlULIETTl AND MENGALI

Figure 4 Leader-Wingman Structure: z—axis distances during flight- Figure 7 Aircraft 3: y—axis position error within the three structures
path change manoeuvre

time (a)

Figure 5 Aircraft 1: y—axis position error within the three structures Figure 8 Leader-Wingman structure formation: position error along y
—axis

firns (s)

Figure 6 Aircraft 2: y—axis position error within the three structures Figure 9 Virtual structure formation: position error along y-axis
Parameter Value GIULIETTI AND MENGAU 7

Wing area, S 27.87 m2 where V is given by Eq. (47) and


Wing span, b 9.14 m
A
=[-L -KJ • B=[J
Aspect ratio, AR 3
Lift curve slope, CL„ 5.3 rad"1
Zero-Lift drag coefficient, CD„ 0.015

<50’
Induced drag factor, k 0.1118
Mass, m 11336 kg

Velocity gain, Kv 5 s"1 Recall that /G, and Kd arc both diagonal and positive definite
Flight-path gain, Ky 3 s"1 matrices [see Eqs. (30)—(31)]. Wc first prove the stability of
Heading gain, K^, 3 s'1 e£=Ae£ (51)
af-axis distance gain, Kx 0.5 s’2
To this end wc introduce[el
theV(Kd)-
Lyapunov
1
] candidate V(et) — | ef
y-axis distance gain, Kv 0.03 m"1 s'1 (52)
z-axis distance gain, Kz 0.03 m"1s"1
and c > 0 is a scalar. Note that P is positive definite for small
values of e since
p _Tl 0 IT I 01
Figure 10 Behavioral structure formation : position error along y—
[O I JL 0 Y{Kdyl\ [eKaY-1
axis
(53) and
(Kd)~l = di^(i/Kdx,l/Kdy,l/KdJ and (V)“l = diag (1,1/Vo, 1/Vo).
Differentiating V wc get:
V = -efQCi (54)
where
Q=-1(ATP + PA) (55)
Wc claim that Q is positive definite. In fact, as Q = QT, Q may be
partitioned as (sec c.g. Zhou ct al.1’’)

Q12I
Q22J

1 Q12Q22 I [Qu Q12Q22QT2 Q22] /]


0 1
JL 0 (56)

It is easily checked that (Q^ 1)7" = Q22 because Q22 — Y[KX


(Kd)-1 — ef] is a diagonal matrix. Accordingly, Eq. (56) is in
the form Q — NMNT and Q is positive definite if and only if M
Table 1 Simulation data is positive definite. Now,
+eF
Conclusions M— ®
L0 (57)
The problem of aircraft formation dynamics and control has
been investigated from the viewpoint of the formation
architecture. Three different, formation structures have been where F is a diagonal matrix whose entries contain the entries
compared in an unified manner. A suitable control law has also of Kd, Kx and Y. For instance, it may be checked that
been described and analyzed. Our approach is physically
intuitive, simple to handle and is guaranteed to be stable at 1 KyKdx
F(l,l) =
least in the linearized case. An important contribution of the 4 eKdx - Kv
paper is that different formation structures may be analyzed on
a quantitative basis and objective results arc made available for and so on. It is concluded that Af (and hence Q) is positive
the designer. In this way the trade-off between system definite for sufficiently small e. This means that V is a
performance and complexity may be fully exploited. A Lyapunov function, the system (51) is stable, and ||e,|| —> 0 as
nonlinear simulation of a flight-path change and a heading t —> 00. Returning now to Eq. (49), wc sec that it involves
change manoeuvre has been illustrated. The results show the the extra term S G £2(0,00) n £00(0,00). However this H term
superiority of the Behavioral approach for it guarantees both docs not affect the stability of (49) as a consequence H of
trajectory tracking and formation keeping with no propagation Lemma I in Ref. 17.
errors. Exploiting the Formation Geometry Center concept, it
also avoids collisions because each aircraft senses the others’ 1
References
Breukenberg, M. and Hummel, D., “Aerodynamics, Performance and
movements. On the other hand, the Behavioral approach is Control of Airplanes in Formation Flight,” 17th 1CAS Congress
more complex because each aircraft needs knowledge of all Proceedings, Vol. 2, September 1990, pp. 1777 1794.
2
the other state vectors. Singh, S. N., Chandler, P., Schumacher, C., Banda, S., and Pachter,
M., “Nonlinear Adaptive Close Formation Control of Unmanned Aerial
Vehicles.” Dynamics and Control, Vol. 10, No. 2. April 200JUpp. 179 194.
3
Bfey, A. W., West, M. G., Lea, K. A., and Jouina’a, M., “Lateral
Appendix: Proof of Theorem 1 Aerodynamic Interference Between Tanker and Receiver in Air-to-Air
Refuelling,” Journal of Aircraft, Vol. 30, 1993, pp. 705 710.
Consider the closed loop dynamics (48) and define the error 4
Bloy, A. W. and Jounia’a, M., “Lateral and Directional Stability
vector e.j = [5f, £^]T- One has: Control in Air-to-Air Refuelling,” Journal of Aerospace Engineering, Vol.
209, 1995, pp. 299 305.
— A — Bi (49)
8 GIULIETTI AND MENGALI

> ^Pacliter, M., D’Azzo, J. J., and Proud, A. W-, “Tight Forma-
tion Flight Control,” Journal of Guidance. Control and Dynamics,
Vol. 24, No. 2, March-April 2001, pp. 246 254.
“Proud. A. W., Pachter, M., and D’Azzo, J. J., “Close Formation
Flight Control,” AIAA Guidance, Navigation and Control Confer-
encfi. AIAA Paper 99-4207, Portland, OR, August 1999.
’achter, M., D’Azzo, J. J., and Dargan, J., “Automatic Forma-
tidn Flight Control,” Journal of Guidance, Control and Dynamics,
Vol. 17, No. 6, November-December 1994, pp. 1380 1383.
s
Buzogany, L. E., Pacliter, M., and D’Azzo, J. .1., “Automated
Control of Aircraft in Formation Flight,” AIAA Guidance. Navi-
gation and Control Conference. Monterey, CA. August 1993, pp.
1349 1370.
9
Stankovic, S. S., Stanojevic, M., and Siliak, D. D., “Decentral-
ized Overlapping Control of a Platoon of Vehicles,” IEEE Transac-
tion on Control Systems Technology, Vol. 8, No. 5, September 2000,
pp. 816 832.
10
Gingras, D. R., “Experimental Investigation of a Multi-Aircraft
Formation,” AIAA Applied Aerodynamics Conference. AIAA Paper
99-3143, Norfolk, VA, June 1999.
11
Gingras, D. R., Player, J. L., and Blake, W. B., “Static and
Dynamic Wind Tunnel Testing of Air Vehicles in Close Proxim-
ity,” AIAA Atmospheric Flight Mechanics Conference and Exhibit,
AIAA Paper 2001-4137, Montreal, Canada, August 2001.
12
Nelson, R. C. and Jumper, E. J., “Aircraft Wake Vortices and
their Effect on Following Aircraft,” AIAA Atmospheric Flight Me-
chanics Conference and Exhibit, AIAA Paper 2001-4073, Montreal,
Canada, August 2001.
13
Anderson, M. R. and Robbins, A. C., “Formation Flight as a
Cooperative Game,” AIAA Guidance, Navigation and Control Con-
ference, AIAA Paper 98-4124, Philadelphia, PA, August 1998.
14
Giulietti, F., Pollini, L., and Innocenti, M., “Autonomous For-
mation Flight,” IEEE Control Systems Magazine, Vol. 20, No. 6,
December 2000, pp. 34 44.
1R
Giulietti, F., Pollini, L., and Innocenti, M,, “Formation Flight
Control: A Behavioral Approach,” AIAA Guidance, Navigation and
Control Conference, AIAA Paper 2001-4239, Montreal, Canada, Au-
gust 2001.
16
Van Tyne, J. and Berger, A. J., Fundamentals of Ornithology,
John Wiley, New York, 1976.
17
Beard, R. W., Lawton, J., and Hadaegli, F. Y., “A Coordination
Architecture for Spacecraft Formation Control,” IEEE Transactions
on Control Systems Technology. Vol. 9, No. 6, November 2001,
pp. 777 790.
1R
Giulietti, F., Napolitano, M., Capetta, R., and Innocenti, M.,
“’’Detailed Modeling of Multiple Aircraft within a Close Formation
Flight”,” Proceedings of AIAA Atmospheric Flight Mechanics Con-
ference, Monterey, CA, August 2002.
19
Zhou, K., Doyle, J. C-, and Glover, K., Robust and Optimal
Control, Prentice Hall, Upper Saddle River, NJ, 1996, pp. 37 38.

You might also like