Beruflich Dokumente
Kultur Dokumente
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/3022474
CITATIONS READS
1,245 632
2 authors, including:
Pascal M Gahinet
The MathWorks, Inc
61 PUBLICATIONS 11,567 CITATIONS
SEE PROFILE
All in-text references underlined in blue are linked to publications on ResearchGate, Available from: Pascal M Gahinet
letting you access and read them immediately. Retrieved on: 31 August 2016
358 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 41, NO. 3, MARCH 1996
Abstract-This paper addresses the design of state- or output- numerically. However, global convergence is not guaranteed,
feedback H , controllers that satisfy additional constraints on and the method is restricted to the state-feedback case.
the closed-loop pole location. Sufficient conditions for feasibility It has been noticed recently that H , synthesis can be
are derived for a general class of convex regions of the complex
plane. These conditions are expressed in terms of linear matrix
formulated as a convex optimization problem involving lin-
inequalities (LMIs), and our formulation is therefore numeri- ear matrix inequalities (LMI) [14], [22], [30]. These LMIs
cally tractable via LMI optimization. In the state-feedback case, correspond to the inequality counterpart of the usual H , Ric-
mixed H , / H , synthesis with regional pole placement is also cati equations. Because LMIs intrinsically reflect constraints
discussed. Finally, the validity and applicability of this approach rather than optimality, they tend to offer more flexibility
are illustrated by a benchmark example.
for combining several constraints on the closed-loop system
[7]. This important advantage of LMI techniques has been
I. INTRODUCTION exploited in [8] to handle H , design with pole clustering
ANY classical control objectives such as disturbance in the region Re (s) < -a where a is a positive scalar. This
attenuation, robust stabilization of uncertain systems, or paper extends these earlier results to a wide variety of stability
shaping of the open-loop response can be expressed in terms regions.
of H , performance and tackled by Ha-synthesis techniques Note that the LMI formulation of constrained H , opti-
[lo], [12]. Since it only involves solving two Riccati equa- mization is appealing from a practical standpoint. Indeed,
tions, H , synthesis has a low complexity comparable to that LMI s can be solved by efficient interior-point optimization
of linear-quadratic-Gaussian (LQG) synthesis [ 111. However, algorithms such as those described in [291, 1281, [61, 1341,
H , design deals mostly with frequency-domain aspects and and [27]. Moreover, software like MATLABS LMI Control
provides little control over the transient behavior and closed- Toolbox 1151 is now available to solve such LMIs in a fast
loop pole location. In contrast, satisfactory time response and and user-friendly manner.
closed-loop damping can often be achieved by forcing the This paper is organized as follows. Section I1 introduces
closed-loop poles into a suitable subregion of the left-half the concept of an LMI region as a convenient LMI-based
plane (see, for instance, the discussion in [I]). In addition, representation of general stability regions. By stability region,
fast controller dynamics can be prevented by prohibiting large we mean any subregion of the open left half of the complex
closed-loop poles (often desirable for digital implementation). plane. Section I11 discusses state-feedback H2/H, synthesis
One way of simultaneously tuning the H , performance and with pole clustering in arbitrary LMI regions. In Section IV,
transient behavior is therefore to combine the H , and pole- these results are fully extended to the output-feedback case.
placement objectives. Finally, the validity and practicality of this approach are
While many authors deal with root clustering in the context demonstrated by an example drawn from the literature in
of linear-quadratic regulation [13], [20], 1251, 1331, only a few Section V.
papers consider the design of H , controllers with additional
specifications on the closed-loop poles. Moreover, most papers 11. LMI FORMULATION
OF POLE-PLACEMENT OBJECTIVES
on this topic are restricted to the state-feedback case [4], [36]. This section discusses a new LMI-based characterization
This can be explained by the difficulty of incorporating pole- for a wide class of pole clustering regions as well as an
placement constraints in the standard state-space approach extended Lyapunov Theorem for such regions. Prior to this
to H , synthesis [ll]. Except for very special regions 1211, presentation, we briefly recall the main motivations for seeking
there seems to be no systematic way of modifying the two pole clustering in specific regions of the left-half plane. It
algebraic Riccati equations to enforce specific root-clustering is known that the transient response of a linear system is
objectives. Note that an attempt in this direction is made in related to the location of its poles [24], [I]. For example,
1361. There, the H , and pole-placement constraints are turned the step response of a second-order system with poles X =
into a system of coupled Lyapunov and Riccati equations, and -<w, fjwd is fully characterized in terms of the undamped
a Lagrange multiplier formulation is used to solve this system natural frequency wn = (XI, the damping ratio (, and the
Manuscript received August 12, 1994; revised August 18, 1995. Recom- damped natural frequency W d (see, e.g., [24, pp. 327-3541).
mended by Associate Editor, A Vicino. By constraining X to lie in a prescribed region, specific
The authors are with INRIA Rocquencourt, 78153 Le Chesnay, Cedex,
France. bounds can be put on these quantities to ensure a satisfactory
Publisher Item Identifier S 0018-9286(96)02100-9. transient response. Regions of interest include a-stability
0018-9286/96$05.00 0 1996 IEEE
CHILALI AND GAHINET: H , DESlGN 359
Consider now the region S ( a , T , 0) defined in (1) and take Let us first review some simple LMI regions. Considering
Q = T = 0. It was shown in [ 3 ] that the eigenvalues of A lie functions of the form
in the sector S ( 0 , 0, 0) if and only if there exists a positive
definite matrix P such that
( W B A ) P + P ( W ~ A<)0~ (1 1)
+
shows that conic sectors ax b < - 1 ~ 1 , vertical half planes
where z+a < 0 or x+a > 0, vertical strips hl < IC < hz,horizontal
strips IyI < w,as well as ellipses, parabolas, and hyperbolic
sectors are all LMI regions.
To derive more complex regions, observe that the class of
On the other hand, S(0, 0, B ) is an LMI region with charac- LMI regions is invariant under set intersection. Specifically,
teristic function given two LMI regions D I and Vz and their associated
characteristic functions fDl and f D 2 , the intersection D =
+
sin S(z X) cos O(z - Z) D1 n Dz is also an LMI region with characteristic function
+
cos O(Z - z ) sin S(z Z )
fVl"D2 = Diag (fD1, (16)
From Theorem (2), A has its poles in S(0, 0, e) if and only f-02).
if there exists X > 0 such that As a result, the class of LMI regions includes all polygonal
sin B(AX + X A T ) cos B(AX - X A T ) regions that are convex and symmetric with respect to the
(cosS(XAT - A X ) sin S(AX + X A T ) real axis. Indeed, any such region can be obtained as the
intersection of conic sectors, vertical strips, andor horizontal
or equivalently strips. An illustrative example is shown in Fig. 2. Here, the
polygonal region ( A , B , C, D , E , F, G) is the intersection
(W 8 A)Diag ( X , X ) + Diag ( X , X ) ( W 8 A)T < 0. (14) of the half plane on the left of the line ( D E ) ,the horizontal
strip delimited by the lines ( B C ) and ( G F ) , and the two
Compared to (1l), this last condition gives additional infor-
conic sectors C T F and BXG. Since any convex region can
mation on the structure of P. It is also better from an LMI op-
be approximated by a convex polygon to any desired accuracy,
timization perspective since the number of optimization vari-
we conclude that LMI regions are dense in the set of convex
ables is divided by four when replacing P by Diag ( X , x). regions that are symmetric with respect to the real axis.
Another consequence of the intersection property (16) is
C. Extent of the Class of LMI Regions that simultaneous clustering constraints can be expressed as a
Theorem 2.2 answers our need for a characterization of system of LMI's in the same variable X without introducing
stability regions that is affine in the A matrix. The convenience any conservatism.
of LMI regions for synthesis purposes will be illustrated in Corollary 2.3: Given two LMI regions D1 and Dz, a matrix
the next two sections. Yet the case for LMI regions would A is both D1-stable and D2-stable if and only if there exists
be incomplete if this class of regions was insufficiently large. a positive definite matrix X such that M D ~ ( AX) , < 0 and
In the remainder of this section, we show that LMI regions MD,(A, X ) < 0.
not only include a wide variety of typical clustering regions, Proof: Simply observe that M D ~ ~ D X~ )( A= , Diag
but also form a dense subset of the convex regions that are (MD~(A X ,) , M D ~ ( AX, ) ) since D1 n D2 has the charac-
symmetric with respect to the real axis. In other words, LMI teristic function Diag ( f D , , f D , ) . Applying Theorem 2.2 to
regions cover most practical needs for control purposes. D1 n D2 completes the proof. lw
CHILALI AND GAHINET H , DESIGN 361
To illustrate the implications of this property, consider the Let (&, B,l, Ccl,, Dclm) and ( 4 1 , Bcl, G i z , DCiz)
vertical strip defined by denote realizations of T,,- and T,,,. We first examine
each specification separately. From Theorem 2.2, the pole-
V = {X + jy E C : -hi < x < -ha < 0). (17) placement constraint is satisfied if and only if there exists
From Corollary 2.3, a matrix A has all its eigenvalues in D if XZ, > 0 such that
and only if there exists a single X > 0 such that [QklXZ, + PklAclXZ, + PlkXZ,GIllk,llm < 0. (24)
AX+XAT+2h2X < 0, AX+XAT+2hlX > 0. (18) Meanwhile, the H , constraint is equivalent to the existence
In comparison, (4) for the region (x - h l ) ( z- hz) < 0 reads of a solution X , > 0 to the LMI
A 2 X + X ( A 2 ) T+ 2AXAT + 2(hl+ h z ) ( A X + X A T )
+ 4hlh21 < 0. (19)
AclX, + X,A: Bci
-I
Dei,
x,::,
DClm
-7'1
) < 0. (25)
Another example is the region S ( a , T , 0) defined by (1). This result is known as the Bounded Real Lemma (see, e.g.,
This clustering region is difficult to express with classical [2] and 1311). Consider finally the H Z objective. Recall that
representations. In contrast, Theorem 2.2 yields the following llT,,,II; = Trace(Cc12PCzz), where P is the solution of
LMI characterization for S ( a , T , 8): the Lyapunov equation A,lP + +
PAS B,$z = 0. Hence
llTwz,lli < Trace(C,lzX2C:2) for any X Z > 0 such that
AX + X A T + 2aX <O
+ +
AclX2 XZAZ B,lB: < 0. Equivalently, llTw,,IIi<
(i;% -rX
n x ) <0 Trace ( Y )whenever the symmetric matrices X Z and Y satisfy
+
sin B(AX X A T ) cos B(AX - XAT)
cos B(XAT - A X ) sin B(AX +XAT)
This follows from observing that this region is the intersection (note that (27) is equivalent to Y > Cc12X2Czz).
of three elementary LMI regions: an a-stability region, a disk, Our goal is to minimize the Hz norm of T,,, over all state-
and the conic sector S(0, 0, 6') discussed in Section 11-B. feedback gains K that enforce the H , and pole constraints.
From the previous discussion, this is equivalent to minimizing
111. STATE-FEEDBACKHz/H, Trace ( Y )over all matrices XZ,,X,, X2, Y ,and K satisfying
DESIGNWITH POLE PLACEMENT (24)-(27). While this problem is not jointly convex in the
variables (XZ,,X,, X Z ,Y , K ) , convexity can be enforced
This section discusses state-feedback synthesis with a com- by seeking a common solution
bination of H2/H, performance and pole assignment specifi-
cations. Here, the closed-loop poles are required to lie in some x = XZ, = X , = x2 >0 (28)
LMI region V contained in the left-half plane. Results are first
to (24)-(27). Admittedly with some conservatism, we are then
derived in the nominal case and then extended to uncertain
left with computing
systems described by a polytope of models. Unconstrained
mixed H2/H, synthesis is considered in [23], where an := inf {Trace ( Y ): X , Y, K satisfy (24)-(27)
J(TWz2)
LMI-based synthesis procedure is proposed. These results are with X Z , = X , = X2 = X} (29)
extended in [4] to include pole placement in a disk. The
discussion below refines the results of [4] and extends them [note that X > 0 is contained in the inequality (27)].
to the larger class of LMI stability regions. The auxiliary performance J(T,,,) is an upper estimate of
Consider a linear time-invariant (LTI) system described by the optimal H2 performance subject to the H , and pole-
placement constraints.
i ( t )= A z ( t )+ B l w ( t ) + B 2 ~ ( t ) Seeking a single Lyapunov matrix X that enforces multiple
+
&(t) = CCo..(t) D,lW(t) + DcoZU(t) constraints has been used in other problems such as the
+
% ( t )= Czx(t) D 2 4 ) (23) quadratic stabilization of a polytope of plants (see [7] for a
good survey). When D is the open left-half plane, the auxiliary
and let T,,- (s) [respectively, Twzz(s)] denote the closed-loop performance J(T,,,) coincides with that introduced in [5]
transfer function from w to z , (respectively, from w to 22) and successfully used in [23] for mixed H 2 / H , synthesis.
under state-feedback control U = K z . For a prescribed closed- Similarly, without the H , constraint, the problem (29) reduces
loop H , performance y > 0, our constrained mixed H z / H , to the auxiliary cost minimization problem proposed in [20]
problem consists of finding a state-feedback gain K that: and discussed in [33] in the case of pole placement in a
places the closed-loop poles in some LMI stability region disk. Finally, the same technique is used in [4] for H z / H ,
V with characteristic function (6); synthesis with root clustering in a disk. However, [4] combines
guarantees the H , performance llTWrm ,1 < y; the resulting conditions into a single LMI at the expense of
minimizes the H2 performance JITwZ211z subject to these additional conservatism. Note that the performance J(T,,, )
two constraints. cannot be improved by using dynamic state-feedback control.
362 E E E TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 41, NO. 3, MARCH 1996
This is proven in [23] when 2, is the open left-half plane, and where the matrices A and B take values in the matrix polytope
this result extends to arbitrary LMI regions by essentially the
same argument. f / N N \ N
for the closed-loop system, (30)-(32) readily follow from Conversely, any solution ( X , L ) of this LMI system satisfies
(24)-(27) together with the change of variable L = K X . +
MD(A B K , X ) < 0 as readily seen when forming the
weighted sum of the LMI' s (37) with nonnegative coefficients
Note that the constrained minimization of Theorem 3.1 is a P I , . . . , p ~Hence
. (37) and (38) are necessary and sufficient
standard LMI problem of the form for quadratic D-stabilizability. When specialized to a sector,
these conditions are exactly those derived in [3]. Note that
Minimize cTx subject to the LMI constraint L ( z ) < 0. LMI conditions for quadratic H , and H2 performance over
(34) (36) are obtained similarly by writing (30)-(32) at each vertex
In particular, it is readily solved with LMI optimization of the polytopic plant.
software such as the LMZ Control Toolbox [15].
The constrained H , problem under consideration can be An additional difficulty in the output-feedback case is that (24)
stated as follows. Given an LTI plant with and (43) now involve nonlinear terms of the form BOKCX
that can no longer be removed by the change of variable
+
k ( t )= A I G ( ~ ) B l w ( t ) +B24t) L := RKX used in Section 111. This has been a long-
~ ( t=Cie(t)
) +
Diiw(2) D i 2 4 t ) + standing obstacle to the derivation of LMI conditions for
y(t) =Cze(t) + D 2 1 4 t ) + D224t) (39) multi-objective output-feedback synthesis. Fortunately, these
nonlinearities can also be eliminated by some appropriate
an LMI stability region D,and some H , performance y > 0, change of controller variables, albeit a more sophisticated one
find an LTI control law U = K(s)y such that: than in the state-feedback case. This change of variable was
The closed-loop poles lie in D. introduced in [16] and is the core of our approach. Note that
l\Twzllm< y where Twz(s) denotes the closed-loop related ideas can be found in [26].
transfer function from w to z . Let A E Rnx" and 0 2 2 E R P 2 x X and z , let k be the
Without loss of generality, it is assumed that 0 2 2 = 0. This controller order ( A K E R"'). As in the state-feedback case,
amounts to a mere change of variable in the controller matrices the change of controller variables is implicitly defined in terms
and considerably simplifies the formulas. of the (unknown) Lyapunov matrix X . Specifically, partition
The LTI controller K ( s ) can be represented in state-space X and its inverse as
form by
(41) i CK
AK
+
:= C K M ~ D K C ~ R
+ S B ~ C K M (46)
:= N A K M ~ N B K C ~ R +
+S(A +B ~ D K C ~ ) R .
~
(AclXCo+~,A;
B:
Ccl xm
Be1
-yI
Dc1
x&
DZ
-71
) <0 (43)
First observe that there is no loss of generality in assuming
k 2 n. Suppose indeed that k < n and let X E V.Then another
solution of (44) is ( X , I?(s)) where X := Diag ( X , I n - k ) >
(this amounts to a mere rescaling of X,). As earlier, this 0 and K ( s ) is the full-order controller with realization
problem is not tractable unless we require that the same
Lyapunov matrix X > 0 satisfy both (24) and (43). We
therefore restrict our attention to the following suboptimal
AK :=Diag(Ak, XIn-k),
BK := (7)
formulation of H , synthesis with pole placement:
Find X > 0 and a controller K ( s ) G Q K that satisfy (24) Assume k 2 n and partition X > 0 as in (45). The
and (43) with X = X , = X , . (44) Sherman-Morrison matrix inversion formula gives S = ( R -
364 E E E TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 41, NO. 3, MARCH 1996
"I
Q22 <o
necessity part, (48)-(50) are derived from X > 0, (24), and
(43) by congruence transformations involving I I 2 . Since 112 is
invertible, we can undo these congruences to conclude that X
with the shorthand notation and see (53) at the bottom of the and K ( s ) solve (44). E
1I
-50 ............ ...........................
3
8 0
i /
/
..........................
i
;.............
:
-50
lo-* 1o 1o2 10
radlsec radlsec
10
b
c
-5 ,3
-10)
0 10 20 30 0 10 20 30
Sec SeC
Fig. 3. Bode plots and impulse responses.
Remark 4.7: Using the change of variable (46) and the The design was performed for k = 0.5. Applying the
simplification (28), it is possible to derive suboptimal LMI LMI conditions of Theorem 4.3, we obtained the following
solutions for a wide variety of multi-objective output-feedback controller for p = 0.2:
problems. For instance, one can combine H2 and H , perfor-
mance, passivity constraints, regional pole assignment, spec- K ( s )= -
+ +
(72s 11.52)((s - 5.58)2 43.82)
ifications on the settling time and overshoot, etc. [9], [32].
+ + + +
( s 2.1)(s 28)((s 36)2 362)
In addition, practical experience indicates that the restriction Fig. 3 shows the Bode plots of the controller and open loop
(28) is not overly conservative. as well as the responses y ( t ) and u ( t ) for a unit impulse
disturbance w. The dashed line corresponds to k = 0.5, the
solid line to k = 1, and the dotted line to k = 2. Note that
V. EXAMPLE the 15 s settling time is robustly achieved and that the control
This example is adapted from the benchmark problem in magnitude lu(t)(does not exceed 10.
[35] and illustrates our LMI approach to output-feedback H ,
synthesis with regional pole assignment. Consider a two-mass- VI. CONCLUSIONS
spring system with We have introduced a new LMI characterization for general
$1 +k ( z 1 - z2) = U convex subregions of the complex plane and demonstrated its
$2 +k(z2 - 5 1 ) = W (57) usefulness for H , synthesis with closed-loop pole clustering
constraints. In the state-feedback case, a systematic LMI
{ r = 52
approach to mixed H 2 / H , synthesis with pole placement
where w is a disturbance and y is the controlled output. The in LMI regions has been presented. These results have then
stiffness k is uncertain and ranges in [0.5, 21. Using output- been fully extended to output-feedback H , synthesis with
feedback control U = K ( s ) y ,our goal is to minimize the effect pole assignment in arbitrary LMI regions.
of the disturbance w on the output y. Admittedly with some degree of conservatism, these re-
To this end, we seek to minimize the H , norm of the sults offer numerically tractable means of performing multi-
closed-loop transfer from w to (:v)
while assigning the objective and/or constrained controller design. This LMI ap-
closed-loop poles in the intersection of the half-plane z < proach has been implemented in the LMI Control Toolbox and
-a = -0.25 with the disk of radius T = 60 centered at tested on a variety of problems with satisfactory results.
the origin. The parameter p is used to compromise between
the control effort and the disturbance rejection performance. APPENDIX
Meanwhile, the a-stability and disk constraints are used to Lemma A.1: Let X E C n x n be a Hermitian positive
enforce a settling time of about 15 s for the impulse response definite matrix. Then its real part Re (X) is a real symmetric
and to prevent fast controller dynamics, respectively. positive definite matrix.
366 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 41, NO. 3, MARCH 1996
Proof: From X = Re (X) + j Im ( X ) and X = X H , D. Arzelier, J. Bernussou, and G. Garcia, Pole assignment of linear
it follows that Re(X) is symmetric and lm(X) is skew- uncertain systems in a sector via a Lyapunov-type approach, IEEE
Trans. Automat. Contr., vol. 38, pp. 1128-1131, 1993.
symmetric. As a result, wTXv = 1 Re ( X ) vfor any v E R, R. Bambang, E. Shimemura, and K. Uchida, Mixed H 2 / H , control
whence Re (X) = Re (X)> 0. H with pole placement: State-feedback case, in Proc. Amer. Contr. Con$,
1993, pp. 2777-2179.
Proof: Proof of Theorem 2.2 (Suficiency): Let X be any D. S. Bemstein and W. M. Haddad, LQG control with an H, per-
eigenvalue of A, and let U E 6be a nonzero vector such that formance bound: A Riccati equation approach, IEEE Trans. Automat.
v H A = XuH. Using the identity Contr., vol. 34, pp. 293-305, 1989.
S. Boyd and L. El Ghaoui, Method of centers for minimizing generd-
ized eigenvalues, Lin. Alg. Applicat., pp. 63-111, 1993.
S. Boyd, L. El Ghaoni, E. Feron, and V. Balakrishnan, Linear Matrix
lnequalities in Systems and Control Theory, vol. 15. Philadelphia:
it is immediate that MD(A,X) < 0 and X > 0 imply SIAM, 1994.
M. Chilali and P. Gahinet, H , design with an a-stability constraint,
~ D ( X ) < 0, or equivalently X E D.Hence, A is D-stable in IFAC Proc., 1994.
when (8) holds. __ , Multi-objective output-feedback synthesis via LMI optimiza-
tion, in Proc. IFAC, 1996.
(Necessity): Suppose that A is D-stable. We must establish J. C. Doyle and G. Stein, Multivariable feedback design: Concepts for a
the existence of a real matrix X > 0 such that M n ( A , X ) < 0. ClassicaUmodem synthesis, IEEE Trans. Automat. Contr., vol. AC-26,
To this end, it is useful to extend the function M n to complex pp. 4-16, 1981.
J. C. Doyle, K. Glover, P. Khargonekar, and B. Francis, State-space
matrices A E G n X nand X = X H E 6 as follows: solutions to standard H2 and H , control problems, IEEE Trans.
Automat. Contr., vol. 34, pp. 831-847, 1989.
M n ( A , X ) := a @ X + @ @ ( A X )+ PT @ ( A X ) H . B. Francis, A course in H , control theory, No. 88, Lecture Notes
in Control and Injonnation Sciences. New York: Springer-Verlag,
1987.
First consider the case where A is a diagonal matrix A = K. Furuta and S. B. Kim, Pole assignment in a specified disk, IEEE
Diag (Xi) with A, E D.It is easily verified that Trans. Automat. Contr., vol. AC-32, pp. 423427, 1987.
P. Gahinet and P. Apkarian, A linear matrix inequality approach to H ,
control, Int. J. Robust Nonlinear Contr., vol. 4, pp. 421448, 1994; also
M n ( A , I) = UT Diag ( f n ( X i ) ) U (59) in Proc. CDC, 1993, pp. 656-661.
P. Gahinet, A. Nemirovski, A. Lanb, and M. Chilali, The LMI Control
where U is some permutation matrix. Consequently, &(A, Toolbox. The Mathworks, Inc., 1995; also in Proc. CDC, 1994, pp.
2038-2041.
X) < 0 holds for X := I . P. Gahinet, Explicit controller formulas for LMI-based H , synthesis,
In the general case, let A be the diagonal matrix of the to appear in Automatica, 1996.
G. Garcia and J. Bemussou, Pole assignment for uncertain systems in
eigenvalues of A (counting multiplicities in the character- a specified disk by state-feedback, IEEE Trans. Automat. Contr., vol.
istic polynomial). From the previous discussion, we have 40, pp. 184-190, 1995.
Mn(A, I ) < 0. Working with the Jordan canonical form of A. Graham, Kronecker Product and Matrix Calculus with Applications.
New York Wiley, 1981.
A, we can construct a sequence of invertible matrices Tk such S. Gutman and E. I. Jury, A general theory for matrix root clustering
that limk,, TLIATk = a [for instance, take Tk = (i y ) in subregions of the complex plan, IEEE Trans. Automat. Contr., vol.
AC-26, pp. 853-863, 1981.
if A = ( :)I. Since M D (Y,I ) is a continuous function of W. M. Haddad and D. S. Bemstein, Controller design with regional
pole constraints, IEEE Trans. Automat. Contr., vol. 37, pp. 54-61,
Y . it follows that 1992.
J. J. Hench, C. He, V. Kucera, and V. Mehrmann, Dampening con-
trollers via a Riccati equation approach, Res. Rep. 1835, 1995.
lim MD(!!FAT~,I ) = Mn(A,I) < 0. (60) T. Iwasaki and R. Skelton, All controllers for the general H ,
k+Cc
control problem: LMI existence conditions and state-space formulas,
Automatica, vol. 30, pp. 1307-1317, 1994.
Hence MD(TF1ATk,I) < 0 for k large enough. Pick such P. P. Khargonekar and M. A. Rotea, Mixed H z / H , control: A
a k and let T := T k . Using the identity (A @ B ) ( c@ D )= convex optimization approach, IEEE Trans. Automat. Contr., vol. 36,
(AC)@ ( B D ) [181 and the definition (7), it is easily verified pp. 824-837, 1991.
B. C. Kuo, Automatic Control Systems. Englewood Cliffs, NJ: Prentice-
that Hall, 1982.
Y. Liu and R. K. Yedavalli, Linear quadratic control with stability
degree constraint, Syst. Contr. Lett., vol. 21, pp. 181-187, 1993.
I. Masubuchi, A. Ohara, and N. Snda, LMI-based controller synthesis:
A unified formulation and solution, Int. J. Robust Nonlinear Contr.,
Together with Mn(T-AT, I ) < 0, this shows that submitted, 1994.
M n ( A , X ) < 0 for X := T T H > 0. A. Nemirovski and P. Gahinet, The projective method for solving linear
matrix inequalities, Math. Programming Ser. B to appear, 1996.
To complete the proof, observe that even though X = T T H Y. Nesterov and A. Nemirovski, An interior-point method for general-
is not necessarily real, its real part also satisfies Re ( X ) > 0 ized linear-fractional problems, Math. Programming Ser. B, to appear,
and M n ( A , Re(X)) = Re(MD(A, X ) ) < 0 as a conse- 1996.
-, Interior Point Polynomial Methods in Convex Programming:
quence of Lemma A.l and A real. Theory and Applications. Philadelphia: SIAM, 1994.
A. Packard, Gain scheduling via linear fractional transformations, Syst.
Contr. Lett., vol. 22, pp. 79-92, 1994.
REFERENCES C. Scherer, The Riccati inequality and state-space H,-optimal con-
trol, Ph.D. dissertation, Univ. Wnrzburg, Germany, 1990.
[ l ] J. Ackermann, Robust Control: Systems with Uncertain Physical Param- -, From LMI analysis to multichannel mixed LMl synthesis: A
eters. London: Springer-Verlag, 1993. general procedure, in Syst. Contr. Lett., submitted, 1995.
[2] B. D. 0 . Anderson and S. Vongpanitlerd, Network Analysis and Syn- N. Sivashankar, I. Kaminer, and P. Khargonekar, Optimal controller
thesis: A Modern Systems Theory Approach. Englewood Cliffs, NJ: synthesis with regional stability constraints, in Proc. CDC, 1993, pp.
Prentice-Hall, 1973. 110-1 15.
CHILALI AND GAHINET H , DESIGN 361
[34] L. Vandenberghe and S. Boyd, Primal-dual potential reduction method Pascal Gahinet (M90) graduated from the Ecole
for problems involving matrix inequalities, Math. Programming Ser. Polytechnique in 1984 and from the E.N.S.T.A.,
B, vol. 69, pp. 205-236, 1995. Paris, France, in 1986 and received the M.S. degree
[35] B. Wie and D. Bernstein, Benchmark problem for robust control in electrical and computer engineering and the Ph.D.
design, J. Guidance Contr., vol. 15, pp. 1057-1059, 1992. degree in system theory, both from the University
[36] R. K. Yedavalli and Y. Liu, Hoo-control with regional stability of California, Santa Barbara, in 1987 and 1989,
constraints, Automatica, submitted 1993. respectively.
He has been with the French National Research
Institute in Computer and Control Sciences since
1990 and has taught control at E.NS.T.A. since
Mahmoud Chilali was born in Casablanca, Mo- 1991. He has also served as a consultant for
rocco, in 1966. He received engineer diplomas from Thomson-CSF and ACrospatiaIe. His research interests include robust control
the Ecole Polytechnique and the Ecole Nationale theory, linear matrix inequalities, numerical linear algebra, and numerical
SupCrieure de Techniques AvancCes, Paris, France, software for control. He is coauthor of the LMZ Control Toolbox for use
in 1990 and 1992, respectively. He is currently at with MATLAB.
the Institut National de Recherche en Informatique
et Automatique, France
His research interests include robust and multi-
objective control and LMI optimization.