Copyright© 1997, American Institute of Aeronautics and Astronautics, Inc.
A97-37112
ROBUST LPV CONTROL WITH BOUNDED PARAMETER
RATES
S. Bennani* D.M.C. Willemsent
Faculty of Aerospace Engineering
Section Stability and Control
Delft University of Technology,
Downloaded by PRINCETON UNIV. LIBRARY on September 30, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.1997-3641
E-mail: S.Bennani@LR.TUDelft.NL
Abstract
This paper discusses an approach to the control of
Linear Parametrically Varying (LPV) systems that
can take the rate of parameter variations into ac-
count and also guarantees robustness against para-
metric and dynamic uncertainties. To illustrate the
technique we consider a missile control problem that
has been extensively studied in the literature. For
this highly nonlinear model, the objective is to de-
sign a controller with guaranteed performance ro-
bustness over a given operating range.
Introduction
The classical approach to gain scheduling relies on
the interpolation of controllers designed for frozen
parameters as, e.g., the operating conditions. This
procedure, even if seemingly working well in prac-
tice, does not take the time-variations of the in-
volved parameters into account. In particular, one
*Ph.D. Candidate
tFlight Control Engineer, DLR German Aerospace Re-
search Establishment, Institute of Flight Mechanics, Braun-
schweig and DASA Daimler-Benz Aerospace Airbus, Flight
Mechanics Departement, Hamburg, Germany
* Assistant Professor
Copyright ©1997 by the American Institute of Aeronau-
tics and Astronautics, Inc. All rights reserved.
1080
AIAA-97-3641
C.W. Scherer*
Mechanical Engineering
Systems and Control Group,
Delft University of Technology,
E-mail: C.Scherer@MR.TUDelft.NL
cannot provide a priori stability and performance
guarantees, as shown in * for linear systems that
depend on a time-varying parameter. Such systems
are called linear parametrically varying (LPV). In
the early nineties > it has been observed that the
techniques of robust control can be generalized to
arriving at a systematic design procedure for such
LPV systems. These solutions, however, lead to
the desired performance guarantees even if the rate
of change of the parameters is unbounded. A re-
fined approach which will be pursued in this paper
takes bounds on this rate of change into account
19, 1, 18, 15, 6, 9 jn addition, our technique allows
to incorporate robustness properties into the design
procedure similarly to > .
The paper is structured as follows. We first provide
the description of the uncertain LPV system. Then
it is shown how the structural knowledge about the
uncertainties is reflected in suitable classes of so-
called multipliers or scalings. This leads to the anal-
ysis characterization of stability and robustness of
the LPV systems in terms of a scaled differential
Bounded Real Lemma. For LPV controller synthe-
sis, we apply a linearizing transformation of the con-
troller parameters > . Introducing basis func-
tions and gridding the parameter set will result, for
fixed scalings, in finitely many linear matrix inequal-
ities that can be readily solved . If optimizing as
 Copyright© 1997, American Institute of Aeronautics and Astronautics, Inc.

well over the scalings, we have to resort to a D-K- together with (2), where
like iteration .
The theory is illustrated on a missile benchmark A(p) + B(p)Dc(p)C(p) B(p)Cc(p)
problem as studied in 14 and ^. In contrast to Bc(p)C(p) Ac(p)
, which considers the nominal performance LPV + B(p)Dc(p)D(p)
problem, we also address robust performance issues. Bc(p)D(p)
Starting from a given nonlinear model, we obtain an = [ H(p) + E(p)Dc(p)C(p) E(p)Cc(p)
uncertain LPV representation accessible for design.
Then the design specifications are translated into • E(p)Dc(p)D(p) ].
suitable weighting functions as in the //oo-approach.
We end up with an interconnection structure and
Downloaded by PRINCETON UNIV. LIBRARY on September 30, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.1997-3641

perform the iteration that is comparable to the D- Analysis of uncertain LPV systems
K procedure in //-synthesis 4. Finally, we validate
In this section we will provide an analysis result
the robust LPV controller by nonlinear simulations.
that characterizes robust stability and robust per-
formance for the uncertain LPV system (4), (2).
LPV design For that purpose we need to introduce scalings that
characterize the nature of the uncertainties Aj af-
fecting the plant in terms of integral quadratic con-
LPV systems straints (IQCs) 13. Hence, for each channel, we de-
The uncertain LPV system is described by fine a collection of matrices Qi, Si, Ri such that the
IQC
x = A(p)x + G(p)w + B(p)u
z = H(p)x + F(p)w + E(p)u (1)
y = C(p)x + D(p)w dt>0 (5)
Si
where, with a suitable partition of the signals w —
holds for T > 0 and for all signals Wi, Zi of finite
\WQ7" ,Wi
Fni i,!r ,..., w
1H T T
and z
k j1 ana [Z0T , Zj
— \y
y — VT , , . . , z
rkT!
j T , flip
tne
energy that are related by Wi — A^. As examples,
uncertainty enters as
we mention
• time- varying parametric uncertainties Wi(t) =
and Wo "-> Zo is the channel to describe the perfor- Zi(t), \8i(t)\ < 1:
mance specification (Figure 1). The parameter p(t)
and its rate of variation p(t) are assumed to be con- < 0 Ri = -Qi, = 0.
tained in the a priori given compact sets P and Pr
respectively. • dynamic uncertainty A./ : L2[0, oo) i-> L2[0,oo)
For controller design the parameter p(t) is assumed with gain not larger than 1:
to be on-line measurable. Hence, LPV controllers
take the form
= Ac(p)xc •Bc(p)y
= Cc(p)xc Dc(p)y (3) We take the I/2-gain of the channel WQ -4 ZQ as a
measure for performance. The I/2-gain of this chan-
such that the resulting closed-loop system is de- nel is bounded by the value 7 if the IQC (5) holds
scribed by with the fixed scaling

(4) Qo = -7/, flo = -J, S o = 0 .

z = +

1081
 Copyright© 1997, American Institute of Aeronautics and Astronautics, Inc.
Finally, we collect the scalings into block-diagonal
matrices as Q = diag(<2o,Qi,-- - > Q f e ) > R —
d\a,g(R0,Ri,.. . ,/Z fc ), and S = diag(Sb,5i,. . . ,Sk).
Now we are ready to provide the characterization
of robust stability and robust performance in terms
of the solvability of a so-called scaled differential
Bounded Real Lemma whose proof is straightfor-
ward **» 17> 15> ^. For notational convenience we
define sy(M) := M + MT.
Theorem 1 Suppose there exist smooth and
Downloaded by PRINCETON UNIV. LIBRARY on September 30, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.1997-3641
bounded functions X(p), Q(p), R(p), S(p) on P
such that
x(P) > o
* ]
* <0
R(p)H(p) R(p)\
(6)
holds for all p €. P and p 6 Pr . Then, for all param-
eter curves (p(t),p(t)) £ P x Pr and for all uncer-
tainties (2), the system (4) remains stable and the
of the performance channel is bounded by 7.
Here X'(p,p) is defined as
3=1
Hence we have to find a parameter dependent Lya-
punov function and parameter dependent scalings to
satisfy a differential linear matrix inequality > .
This generalizes the idea of using a constant Lya-
punov function 5> 2 and constant scalings for arbi-
trarily fast varying parameters.
To solve the inequalities numerically we choose con-
tinuously differentiable functions fi(p) ••• fi (p) and
search for the coefficients in the expansion
X(p)
Q(P) Qj
R(p) Rj
3=1
The resulting infinitely many LMIs are reduced to
finitely many inequalities by picking a finite number
1082
of points in P and Pr. If Pr is described as a con-
vex combination of finitely many vertices, it suffices
to choose the extreme points since the parameter p
appears linearly in (6) 19> *.
Controller synthesis
The synthesis problem consists of designing a con-
troller (3) that minimizes the robust performance
level 7 as characterized in Theorem 1. However, the
inequalities (6) are not linear in all the unknowns,
the Lyapunov function, the scalings, and the con-
troller parameters. It has been shown in 10> 15) 16
how the inequalities can be linearized, for fixed scal-
ings, by a suitable nonlinear transformation of the
controller parameters as follows: Denote the first
block rows of X(p) and X(p)~l as [ X(p) U(p) }
and [ Y(p) V(p) ] respectively, and introduce the
new controller parameters
K(p,p) = X(p)[A(p) + B(p)Dc(p)C(p)]Y(p)+
+U(p)Bc(p)C(p)Y(p) + X(p)B(p)Cc(p)VT(p)+
+U(p)Ac(p)VT(p) + X'(p,p)Y(p) + U'(p,p)VT(p)
L(p) = X(p)B(p)Dc(p) + U(p)Bc(p)
M(p) - Dc(p)C(p)Y(p] + Cc(p)VT(p)
N(p) = Dc(p).
(7)
Then synthesis inequalities are simply obtained by
performing the substitutions
XA + LC K
XA AY + BM
A + BNC
XG + LD (H + ENC)T
G + BND (HY + EM)T
X I X' 0
X
I Y 0 -Y'
in (6) where we dropped the dependence on p and p
for notational simplicity.
As explained for analysis, we can introduce basis
functions and grid the parameter set to end up with
finitely many inequalities. However, the resulting in-
equalities are still nonlinear in the new variables and
the scalings together. Hence we have to resort to a
D-K-\\ke iteration scheme that proceeds as follows:
 Copyright© 1997, American Institute of Aeronautics and Astronautics, Inc.
Start with the uncertainty scalings Qi(p) = — /,
Ri(p) = I, Si(p) = 0 and iterate the following
two steps until the performance level cannot be im-
proved:
1. Fix the scalings and minimize 7 over
X (p), Y(p) and the transformed controller
parameters K(p,p), L(p), M(p), N(p).
2. Fix the controller parameters K(p,p), L(p),
M(p), AT(p) and minimize the performance level
7 over X(p) and the uncertainty scalings Qi(p),
Downloaded by PRINCETON UNIV. LIBRARY on September 30, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.1997-3641
), Si(p) as described for analysis.
Suppose the iteration stops with X(p), Y(p),
K(p,p), L(p), M(p), N(p). Then one simply needs
to choose nonsingular smooth and bounded func-
tions U(p), V(p) satisfying
Y(p)X(p) + V(p)UT(p) = I (8)
in order to calculate the controller by solving (7) for
Ac(p,p), Be(p)t Cc(p), Dc(p).
Since Ac(p,p) depends on p, one needs to measure
not only the parameter value p(t) itself but also its
rate of variation p(t) to implement the resulting con-
troller. To avoid this undesired structure, we choose
K(p) independent of p and either X(p) or Y(p) in-
dependent of p similarly to > . Exploiting the free-
dom in the choice of U(p) and V(p) allows to con-
struct a controller that depends on p only:
• If X(p) is parameter dependent and Y is con-
stant, choose U(p) = 7 — X(p)Y and V = I.
Taking derivatives in (8) reveals
(9)
such that the terms in (7) that depend on p
indeed drop out.
• If X is constant and Y(p) is parameter depen-
dent, choose U = I and V(p) = I - Y(p)X.
This implies X'(p,p) = 0, U'(p,p) = 0 such
that, again, the variable p disappears in (7).
Note that this restriction to constant X(p) or Y(p)
certainly introduces conservatism, with the benefit
of a simpler controller implementation.
1083
To further speed up the computation, we finally re-
mark that we performed all calculations after elimi-
nating the transformed controller parameters along
standard lines by using the projection lemma, and
by using scalings with S(p) = 0 1. If columns
of Ky and Ku form bases of the the nullspaces of
[ B^ 0 ET ] and [ C D 0 ] respectively, the
reduced synthesis inequalities then read as
X I
(10)
I Y
sy(XA) + XG HT
GTX Q FT Ky < 0, (11)
H F -R-1
sy(AY) -Y' G YHT
K. GT Q FT Ku < 0. (12)
HY F -R-1
The iteration scheme proceeds as follows:
1. Start the iteration without scalings in the un-
certainty channels.
2. For fixed scalings Q an R, minimize 7 (in the
scaling of the performance channel) over the
constraint LMIs (10)-(12). Reconstruct the
controller parameters.
3. For fixed controller parameters, minimize 7 over
the scalings Q and R using (6). Go to step 2.
The missile control problem
For the application we have chosen a missile bench-
mark problem that has been extensively studied in
11
> > 19 > 8 and is particularly suited for address-
ing gain scheduling as well as robustness issues. The
problem is to design a longitudinal autopilot for a
tail-fin controlled missile providing normal accelera-
tion tracking over a large range of speed and angle-
of-attack. In order to arrive at a design model (in
section missile control strategy), the exact problem
 Copyright© 1997, American Institute of Aeronautics and Astronautics, Inc.

specifications are first given in section performance
description. These are based on the missile model
which is defined in section missile model.
The missile model
The non-linear state equations of the missile are
These coefficients are valid for the missile traveling
between Mach = 2 and Mach = 4 at an altitude of
20,000 ft. Typical maneuvers for this missile result
in angle-of-attack values ranging between —20 and
+20 degrees. Hence the approximation cos(a) « 1
is legitimate. Then (13) simplifies to

d = /, (a, qt St M) = B (M,M)-Hr(13) (15)

mu mV
<7 = W) (14)
One way to obtain an LPV model for the missile is to
parametrize the set of all equilibrium models. For
Downloaded by PRINCETON UNIV. LIBRARY on September 30, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.1997-3641

with any angle-of-attack a 6 [0, 20] and Mach number

M 6 [2, 4], the fin deflection and pitch rate
a angle-of-attack (rad)
pitch rate (rad/s)
<S(a,M) = -— |ama3 , 2 ,„
bma2 - cm(7 - ——)a .
Cn(a,S,M)0.7p0M2S (Ibs) an
My Cm(a,S, M)0.7p0M2Sd (ft - Ibs) Fz(a,6,M)
8 tail fin deflection (rad) <7(a,M) - -
mV
Po 973.3 lb/ft2 (static pressure at 20,000 ft)
S 0.44 ft 2 (reference area) lead to an equilibrium of (13)-(14). The specific nor-
d 0.75 ft (diameter) mal force nz is measured by an accelerometer placed
m 13.98 slugs (mass of missile) at the the center of gravity of the missile. It is
u V cos(a) ft/s (speed along missile center line) defined as nz = |^- where W = mg. For conve-
V Mss ft/s (velocity of the missile) nience we use the shorthands Ka = °^p°s5, Kq —
2:2
M 2 — 4 (Mach number of the missile) ^i» Kn = ifo2^- The Jacobi linearization
ss 1036.4 ft/s (speed of sound at 20,000 ft) of the missile dynamics is then given as
fy 182.5 slug - ft 2 (pitch moment of inertia)
9 32.2 ft/s2 (acceleration due to gravity)
nz normal acceleration of the missile (per g).
The aerodynamic nonlinearity and parameter de- where
pendence in the missile model are reflected in the
normal force and moment coefficients Cn(a, 6, M)
KaMdCn(a,0,M)/da 1\
>l(o, M) =
and Cm(a,6,M) respectively. Taking the missile KqM2dCm(a,Q,M)/da 0 )
symmetry into acount it suffices to consider the pos-
itive values of the angle-of-attack. The aerodymanic B(a,M) = [ Kq'M2dCnm(Q,6,M)/dS
coefficients are then given by 0 1
C(a,M) =
KnM2dCn(a,0,M)/8a 0
Cn(a, S, M) - ana3 bna2 cn(2 + 4f)a + dn5
Cm(a,S, M) = - Sf )a + dmS 0
D(a,M) =
KnM2dCn(0,6,M)/da
where the polynomial coefficients are
We end up with a family of linearized system that
an = +0.000103 deg~3 am = +0.000215 deg-3 are parametrized by p = (a,M). For a particular
bn = -0.009450 deg~2 bm = -0.019500 deg-2 parameter value p in the allowable parameter set,
1
cn = -0.169600 deg-1 Cm = +0.051000 deg" the LPV dynamics are called frozen and reflect a
dn = -0.034000 deg"1 dm = -0.206000 deg"1. local linearization of the missile dynamics.

1084
 Copyright© 1997, American Institute of Aeronautics and Astronautics, Inc.
The specifications
The specifications to be achieved by the controller
have to hold over the whole Mach range [2, 4].
Therefore, the system should globally provide nor-
mal acceleration command tracking features, with
rise-time not greater than 0.35 s, overshoot not
greater than 10 %, and steady state error not greater
than 1 %. The measurements available for control
are the normal acceleration nz, the pitch rate q and
the Mach number M. During a typical maneuvre the
angle-of-attack ranges within |a| < 20 degrees, the
Downloaded by PRINCETON UNIV. LIBRARY on September 30, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.1997-3641
tail-fin deflection rate should not exceed 25 deg /s
per commanded g-level.
As very strong simplifications in the missile mod-
eling have been made, we take the robustness is-
sue originating from the uncertainty in the aerody-
namic coefficients Cn and Cm into account. The un-
certainity levels considered are ACn = ±10 % and
A<7m = ±25 %.
The controller provides fin commands Sc that are
processed through second order actuator dynamics
2
given b y G^s = ,aa+2C , tta a > with natural fre-
quency a) = 150 rad/s and damping C = 0.7. To
avoid exciting unmodeled high frequency dynam-
ics, the multiplicative input uncertainty weighting
is placed at the actuator.
Control strategy
To realize the specifications over the prescribed op-
erating range, the missile dynamics are reformu-
lated into an uncertain parameter varying system
representation. The LPV system with parameter
p = (a, M) has now an uncertain part arising from
the perturbations in the aerodynamic coefficients. A
further strategy, which might not be the best one is
to view the angle-of-attack in the parameter vector p
as uncertain. One one side we chose this strategy to
illustrate the method and furthermore because the
angle of attack a. is not measurable in the problem
specification. Another approach would be to use a
as a schedule parameter and and use an observer
to implement in the schedule as was shown in .
However, the resulting uncertainty model would not
be anymore challengeing enough to test the effect of
scaling. Hence, the angle-of-attack a and the uncer-
1085
tainties in the aerodynamic coefficients Cm and Cn
are pulled out of the system and rescaled to [—1, +1].
The resulting uncertainty structure in the missile dy-
namics is then A u = diag((5a/2, ACn, A Cm ).
The control architecture for the missile problem
is depicted in figure 2. The tracking specifica-
tion has been translated into an ideal accelera-
tion model that the closed loop system should
match. The ideal model comes from ™ and is
WuC*) := 1"+wX$ for which the allowable er-
ror is weighted as Wper{(s) := ° 1 1 . The low
frequency gain of Wperf is 300 to bound the track-
ing error by 0.33 %. The high frequency gain is
chosen to be 0.5 in order to limit the overshoot to
be less than 5 %. To reflect the tail-fin deflection
and deflection rate limits of 20 degrees and 25 de-
grees/s per g respectively, the filters Wg = ^ and
W$ = •%£ have been chosen accordingly. Finally,
noise filters Wnl = 0.001, and Wn2 = 0.001 are used
to reflect the measurement imperfections in pitch
rate and normal acceleration.
Results
The design of the LPV controller
In this section we will use the synthesis LMIs (6)
as derived previously. Solving the LMIs is done
via basis functions and gridding of the parameter
space. For the missile control problem, the Mach
number M is the remaining parameter for schedul-
ing. The angle-of-attack a, the aerodynamic coeffi-
cients Cn and Cm and the actuator were all assumed
uncertain. As the Mach number M can be placed
in an LFT linearly, we choose a basis function as
/i(M) = M (see also 1). For the function X(M),
e.g., we thus have
The other functions depending on the parameter M
are Y(M), Q(M), K(M), L(M), M(M), N(M),
and have the same structure as X(M).
The grid of the parameter set M € [2, 4] consisted
of five points equally spaced between Mach = 2
 Copyright© 1997, American Institute of Aeronautics and Astronautics, Inc.
and Mach — 4. As the Mach number will de-
crease from M = 4 to M = 2 in five seconds in
the non-linear simulations, the parameter rate was
taken \M\ < 0.5/s. Further, we use a block diagonal
scaling matrix Q = diag (Qo> • • • > <?4) arising from
the uncertainty and the performance channels: the
matrix Q\ of dimension 2x 2 for the uncertain a, the
two scalar blocks Qz, Qa for the uncertainty in Cn,
Cm, the scalar block Q4 for the dynamic actuator
uncertainty, and the 3 x 3 block Q0 for the perfor-
mance specification. In the first iteration the seal-
Downloaded by PRINCETON UNIV. LIBRARY on September 30, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.1997-3641
ings are set to unity. Once convergence is achieved,
the large LMIs (6) are solved for the last scaling Q
that was found in the iteration.
In the section on the analysis of uncertain LPV sys-
tems it was explained that choosing either X or Y
constant led to a controller that does not need a mea-
surement of the parameter rate. For the missile con-
trol problem, both options were tested. The scheme
using X constant and Y parameter dependent con-
verged faster than the one where X was parameter
dependent and Y constant. In 14 steps a 7-value
of 2.50 was reached whereas the other option only
reached the value of 3.87 after 20 iteration steps. In
figure 3 the achieved 7-value is set out against the
number of iteration steps for both processes.
The final controller synthesis was carried out on the
choice (Xo, Y(M)) where 7 = 2.5 was achieved. The
scaling matrix Q of the last iteration is used to syn-
thesize the controller using the full LMIs (6). The
achieved 7-value was 2.51 and the test on a denser
parameter grid (with twice the density) gave 7 val-
ues between 2.22 and 2.45. The achieved perfor-
mance level given by the propsed scheme is com-
parable to results given in ^ (7 = 3.13) and 8
(7 = 3.855) with the difference that in our design pa-
rameter rate boundness and uncertainties are taken
into account. Since the angle of attack was not
measured but viewed as uncertain it increases the
achievable 7 level. On the other hand the parameter
rate information reduces the achievable L2 induced
norm, therefore compensating conservatism intro-
duced by uncertainty effects and the fact that less
sensor information has been used (measurements of
p and a).
1086
Simulations
The non-linear simulations of the LPV controlled
missile are depicted in the figures 4 to 7. The maneu-
ver, during which the Mach number varies as shown
in figure 7, consists of a series of acceleration step
commands as depicted in figure 4. The accelera-
tion command response of the LPV controlled mis-
sile has a rise time that is less than the prescribed
0.35 s. The steady state error is within the required
bounds. Overshoot characteristics are also within
the limits. Only the step command from 30 g to
-15 g causes a 3 % overshoot violation. As a remedy
one could try to redesign the weightings. A possible
choice to enhance damping of the acceleration re-
sponse is to increase the weight on the fin rate filter
Wg since maximum fin rate is by far not reached in
the non-linear simulation. Also the performance fil-
ter could be adjusted to further punish the overshoot
(increase high frequency gain of the filter). However,
we left the filters W\d and WpeT{ the same as in
in order to be able to compare the results. The
LPV controller synthesized here has a larger over-
shoot, but it is faster than the one in 19. Moreover
the LPV controller of shows angle-of-attack ex-
ursion beyond —20 degrees while also commanding
four times more fin deflection rate during the —45 g
command. Figure 5 and 6 shows that our controller
remains within the limits which might be interpreted
as a rather conservative design in the sense that it
doesn't use the available fin deflection to enhance
performance. It should be noted that the missile in
19
runs along a slightly different Mach trajectory.
Finally, to demonstrate the robustness properties,
figure 8 shows the acceleration command responses
for all combinations of the aerodynamic uncertain-
ties. As can be seen from the figure, overshoot in
the 45 g step is the most sensitive to uncertainties,
while the other performance characteristics seem to
behave well.
Conclusions
We have shown a controller synthesis technique for
Linear Parameter Varying (LPV) that takes the
boundedness of the parameter variation rates into
 Copyright© 1997, American Institute of Aeronautics and Astronautics, Inc.
account. This technique gives guaranteed stabil-
ity and performance levels. Moreover, robustness
against uncertainties has been incorporated via the
use of scalings. Using basis functions and grid-
ding, the synthesis problem is reduced to an iter-
ation of solving finitely many linear matrix inequal-
ities. The method was applied and tested on a mis-
sile benchmark problem. The non-linear simulations
have shown that the proposed method is successful
in achieving the desired performance and robustness
goals. Further research should be directed towards
Downloaded by PRINCETON UNIV. LIBRARY on September 30, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.1997-3641
finding systematic procedures to choose the basis
functions in the presented approach.
Acknowledgment
We would like to thank Dr. P. Apkarian for making
his software available to us.
References
[1] Apkarian, P. and R.J. Addams (1995). Ad-
vanced gain-scheduling techniques for uncertain
systems, preprint.
[2] Apkarian, P. and P. Gahinet (1995). A con-
vex characterization of gain-scheduled //oo-
controllers, IEEE Trans. Automatic Control,
40, 853-864.
[3] Athans, M. and J.S. Shamma (1992). Gain-
scheduling: potential hazards and possible
remedies, IEEE Control Systems Magazine, 12,
101-106.
[4] Balas, G.J., J.C. Doyle, K. Glover, A. Packard,
and R. Smith (1993). fi-Analysis and Synthe-
sis TOOLBOX, For Use with MATLAB. The
Mathworks Inc.
[5] Decker, G. (1993). Quadratic stability and per-
formance of linear parameter dependent sys-
tems. PhD thesis, University of California at
Berkeley, Berkeley, USA.
1087
[6] Becker, G. (1996). Additional results on pa-
rameter dependent controllers for LPV systems,
IFAC World Congress, vol. G, pp. 351-356,
1996.
[7] Gahinet, P., A. Nemirovski, A. Laub, and
M. Chilali (1994). LMI Control Toolbox. The
Mathworks Inc.
[8] Helmersson, A. (1995). Methods for robust gain-
scheduling. PhD thesis, Linkoping University,
Sweden.
[9] Lee, L.H., M. Spillman, Control of slowly vary-
ing LPV systems: An application to flight con-
trol. Guidance Navigation and Control Confer-
ence, AIAA paper no. 96-3805, Jul. 1996.
[10] Masubuchi, I., N. Suda and A. Ohara (1995).
LMI-based controller synthesis: a unified for-
mulation and solution, Proc. ACC, 3473-3477.
[11] Packard, A. and G.J. Balas (1992). Time-
varying controllers for missile autopilots, 1st
IEEE Conference on Control Applications,
Dayton, OH.
[12] Packard, A. (1994). Gain scheduling via linear
fractional transformations, System and Control
Letters, 22, 79-92.
[13] Rantzer, A. and A. Megretski (1994). Sys-
tem analysis via integral quadratic constraints,
Proc. IEEE Conf. Decision Contr., 3062-3067.
[14] Rugh, W.J., R.A. Nichols and R.T. Reichert
(1993). Gain scheduling for H^ controllers: a
flight control example, IEEE Transactions on
Control Systems Technology, 1, 69-78.
[15] Scherer, C.W. (1995). Mixed H^/H^ control.
In: A. Isidori (Ed.), Trends in Control, A Euro-
pean Perspective, Springer-Verlag, Berlin, 173-
216.
[16] Scherer, C.W., P. Gahinet and M. Chilali
(1995). Multi-objective output-feedback con-
trol, preprint.
 Copyright© 1997, American Institute of Aeronautics and Astronautics, Inc.

[17] Scorletti, G. and L. El Ghaoui (1995). Im-

proved linear matrix inequality conditions for
gain-scheduling, Proc. 34th IEEE Conf. on De-
cision and Control, 3626-3631.
[18] Willemsen, D.M.C. (1996). Linear matrix in-
equalities in robust control. Master's thesis,
Delft University of Technology, Department
of Aerospace Engineering, Delft, The Nether-
lands.
[19] Wu, F. (1995). Control of Linear Parameter
Downloaded by PRINCETON UNIV. LIBRARY on September 30, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.1997-3641

Varying Systems. PhD thesis, University of Cal-

ifornia at Berkeley, Berkeley, USA.

0 2 4 6 6 10 12 14 16 18 20
number of iterations

Figure 3: Achieved performance for two choices of

basis functions

Figure 1: LPV system with uncertainty and a per-

formance channel

£i
3 [lif—1
\3 Li—u

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Figure 4: Normal acceleration nz for the com-

manded acceleration scenario nc of the LPV con-
trolled missile.
Figure 2: The controller synthesis interconnection
structure

1088
 Copyright© 1997, American Institute of Aeronautics and Astronautics, Inc.

4
Downloaded by PRINCETON UNIV. LIBRARY on September 30, 2014 | http://arc.aiaa.org | DOI: 10.2514/6.1997-3641

3.8

3.6

3.4

L3.2

i 3

| 28

2.6

2.4

2.2

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

lime (>)

Figure 5: Angle-of-attack a for the commanded ac-

Figure 7: The parameter trajectory M(t).
celeration scenario nc of the LPV controlled missile.

1.5 2 2.5 3 3.5

trne(s) 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Figure 6: Fin deflection rate 6 for the commanded

Figure 8: Command response for all combination of
acceleration scenario nc of the LPV controlled mis-
perturbed aerodynamics
sile.

1089
American Institute of Aeronautics and Astronautics