Applied Mathematics and Computation 187 (2007) 1228–1237

www.elsevier.com/locate/amc

Delay-dependent robust stability and H1 control

for uncertain discrete-time switched systems
with mode-dependent time delays
a,b,*
Yuan Gong Sun , Long Wang a, Guangming Xie a

a
Center for Systems and Control, College of Engineering, Peking University, Beijing 100871, China
b
Department of Mathematics, Qufu Normal University, Qufu, Shandong 273165, China

Abstract

In this paper, we consider the problems of robust stability, robust stabilization and H1 control via memoryless state
feedback for uncertain discrete-time switched systems with mode-dependent time delays. Based on linear matrix inequal-
ities (LMIs), a descriptor model transformation of the system and a switched Lyapunov function, new delay-dependent
criteria are established which are not contained in known literatures. Numerical examples are presented to illustrate the
effectiveness of the theoretical results.
 2006 Elsevier Inc. All rights reserved.

Keywords: Discrete-time switched system; H1 control; Linear matrix inequality (LMI); Switched Lyapunov function; Delay-dependent
criteria

1. Introduction

By a switched system, we mean a hybrid dynamical system that is composed of a family of continuous-time
or discrete-time subsystems and a rule orchestrating the switching between the subsystems. Switched systems
have gained a great deal of attention mainly because many real-world systems such as, for example, chemical
process, transportation systems, computer controlled systems and communication industries can be modeled
as switched systems.
In the last two decades, there has been increasing interest in stability analysis and control design for switched
systems (see [1,3,4,7,11–23] and references therein). However, most of these results are aimed at switched sys-
tems without uncertainties and time delays. Very recently, Xie et al. [20] studied the problems of stability, sta-
bilization and H1 control via memoryless state feedback for uncertain discrete-time switched systems without
time delays. To the best of our knowledge, it seems to us that few people have studied such problems for uncer-
tain discrete-time switched systems with mode-dependent time delays. This has motivated our research.

*
Corresponding author. Address: Department of Mathematics, Qufu Normal University, Qufu, Shandong 273165, China.
E-mail address: [email protected] (Y.G. Sun).

0096-3003/$ - see front matter  2006 Elsevier Inc. All rights reserved.
doi:10.1016/j.amc.2006.09.053
 Y.G. Sun et al. / Applied Mathematics and Computation 187 (2007) 1228–1237 1229

We are here concerned with the problems of robust stability, stabilization and H1 control via memoryless
state feedback for uncertain discrete-time switched systems with mode-dependent delays. It has been recog-
nized that time delays, which are the inherent features of many physical process, are the big sources of insta-
bility and poor performances. In this paper, we will establish delay-dependent criteria in terms of linear matrix
inequalities (LMIs) for the robust stability, stabilization and H1 control of uncertain discrete-time switched
systems with mode-dependent delays. The main idea of our method is inspired by Fridman’s recent work
[5,6,10], where a descriptor system approach is introduced for stability analysis and controller synthesis of
deterministic delay systems. The advantage of the descriptor system approach lies in that it considerably
reduces the conservatism entailed in the previously developed transformation methods since it is based on
an equivalent ‘‘descriptor form’’ representation of the system and requires bounds for fewer cross terms.
We extend the descriptor system approach to uncertain discrete-time switched systems with mode-dependent
delays by constructing a switched Lyapunov function [4]. The results obtained in this paper are not contained
in our recent paper [16].
The paper is organized as follows: The problem is formulated in Section 2. Section 3 deals with delay-
dependent criteria for the robust stability and robust stabilization. Section 4 is devoted to derive the LMI-
based condition for robust H1 control. Remarks and numerical examples are presented in Section 5 to illus-
trate the theoretical results, and the conclusions are drawn in Section 6.

2. Model description and preliminaries

Throughout this paper, if not explicitly stated, matrices are assumed to have compatible dimensions. The
notation M > 0 (<0) is used to denote a positive (negative) definite matrix. I denotes an identity matrix. k Æ k
stands for the Euclidean norm of vectors. The notation * represents the elements below the main diagonal of a
symmetric matrix.
Consider a discrete-time, hybrid system with N nodes:

xðk þ 1Þ ¼ A1 ðk; rðkÞÞxðkÞ þ A2 ðk; rðkÞÞxðk  sðk; rðkÞÞÞ þ B1 ðk; rðkÞÞwðkÞ þ B2 ðk; rðkÞÞuðkÞ;
zðkÞ ¼ C 1 ðk; rðkÞÞxðkÞ þ C 2 ðk; rðkÞÞxðk  sðk; rðkÞÞÞ þ D1 ðk; rðkÞÞwðkÞ þ D2 ðk; rðkÞÞuðkÞ; ð1Þ
xðsÞ ¼ /ðsÞ; s ¼ s; . . . ; 1; 0;

where x(k) 2 Rn is the P system state, u(k) 2 Rm is the control input, w(k) 2 Rp is the disturbance input which
1
belongs to l2 ¼ fwðkÞj k¼0 wT ðkÞwðkÞ < 1g, and z(k) is the controlled output. rðkÞ : Z þ ¼ f0; 1; 2; . . .g !
N ¼ f1; 2; . . . ; N g is the control signal. s(k, r(k)) : Z+ ! {1, 2, . . .} denotes the time-varying delay of the sys-
tem involved in mode r(k). Here we assume that s(k, r(k)) is bounded for k P 0. Let h = maxkP0{s(k, 1),
s(k, 2), . . . , s(k, N)}. / : {h, h + 1, . . . , 0} ! Rn represents the initial condition. For each i 2 N, the system
matrices are assumed to be uncertain and satisfy:
     
A1 ðk; iÞ A2 ðk; iÞ B1 ðk; iÞ B2 ðk; iÞ Ai1 Ai2 Bi1 Bi2 H i1
¼ þ F ðk; iÞ½Ei1 Ei2 Ei3 Ei4 ; ð2Þ
C 1 ðk; iÞ C 2 ðk; iÞ D1 ðk; iÞ D2 ðk; iÞ C i1 C i2 Di1 Di2 H i2
where Aij, Bij, Cij and Dij (j = 1, 2) are constant matrices that describe the ith nominal mode, Hij (j = 1, 2) and
Eil (l = 1, 2, 3, 4) are given constant matrices which characterize the structure of the uncertainty, and F(k, i) are
uncertainties satisfying FT(k, i)F(k, i) 6 I for k 2 Z+. One reason for assuming that the system uncertainty has
the structure given in (2) is that a linear interconnection of a nominal plant with the uncertainty F(k, i) leads to
the structure of the form (2). The other comes from the fact that uncertainties in many physical systems can be
modeled in this manner, e.g. satisfying ‘matching conditions’.
We are interested to establish the delay-dependent robust stability, robust stabilization and robust H1 con-
trol for system (1) using the LMI technique [2]. In order to prepare for a precise formulation of our results, we
introduce the following definitions.
Definition 1. The uncertain hybrid system (1) with u(k) = w(k) = 0 is said to be uniformly robust stable (URS)
if for any  > 0, there is a d() > 0 such that for all admissible uncertainties and arbitrary switching signal r(k),
suph6s60k/(s)k < d implies kx(k)k <  for all k P 0. In addition, if there is a d 0 > 0 such that for all admissible
1230 Y.G. Sun et al. / Applied Mathematics and Computation 187 (2007) 1228–1237

uncertainties and arbitrary switching signal r(k), suph6s60k/(s)k < d 0 implies x(k) ! 0 as k ! 1, then system
(1) with u(k) = w(k) = 0 is said be be uniformly robust asymptotically stable (URAS).

Definition 2. System (1) with w(k) = 0 is uniformly robust asymptotically stablilizabe, if there exists a
memoryless state feedback controller in the form of u(k) = Kr(k)x(k) such that the closed-loop system (1) with
w(k) = 0 is URAS.

Definition 3. The uncertain hybrid system (1) is said to be uniformly robust asymptotically stablilizabe with c-
disturbance attenuation if there exists a memoryless state feedback controller in the form of u(k) = Kr(k)x(k)
such that the closed-loop system (1) with w(k) = 0 is URAS, and for any disturbance w(k) 2 l2 and for the
initial condition /(k) = 0, k 2 {h, h + 1, . . . , 0}, the response z : Z+ ! Rq satisfies
X1 X
1
T
zT ðkÞzðkÞ < c2 wðkÞ wðkÞ:
k¼0 k¼0

The following lemmas plays an important role in our later development.

Lemma 1 ([2], Schur complement). Let M, P, Q be given matrices such that Q > 0. Then
 
P M
< 0 () P þ M T Q1 M < 0:
M T Q

Lemma 2 [19]. Let D, E and F be matrices with appropriate dimensions. Suppose FTF 6 I, then for any scalar
k > 0, we have

DFE þ ET F T DT 6 kDDT þ k1 ET E:

Lemma 3 [9]. Assume that d(k) : Z+ ! {1, 2, . . .} and d(k) 6 d, where d is a positive integer, then for any posi-
tive-definite matrix P 2 Rn·n and vector function x(s) : {d, d + 1, . . .} ! Rn, we have
! !
X
k1 X
k1 X
k1
d T
x ðsÞPxðsÞ P T
x ðsÞ P xðsÞ ; k 2 Rþ :
s¼kd s¼kdðkÞ s¼kdðkÞ

3. Robust stability and stabilization

In this section, the discussion is restricted to the case w(k) = 0, namely,

xðk þ 1Þ ¼ A1 ðk; rðkÞÞxðkÞ þ A2 ðk; rðkÞÞxðk  sðk; rðkÞÞÞ þ B2 ðk; rðkÞÞuðkÞ;
ð3Þ
xðsÞ ¼ /ðsÞ; s ¼ s; . . . ; 1; 0:
Theorem 1. System (3) with u(k) = 0 is uniformly robust asymptotically stable, if for each i 2 N, there exist
constants ki > 0, matrices Pi, Q > 0, Ri and Si, i 2 N, such that the following LMIs hold,
2 T
3
X1 X2 RTi Ai2 RTi H i1 ðEi1 þ Ei2 Þ
6 7
6  X3 S Ti Ai2 S Ti H i1 0 7
6 7
6   Q 0 Ei2T 7 < 0; i; j 2 N; ð4Þ
6 7
6 7
4    ki I 0 5
    k1
i I

where
T
X1 ¼ RTi ½Ai1 þ Ai2  I þ ½Ai1 þ Ai2  I Ri þ P j  P i ;
T
X2 ¼ P j  RTi þ ½Ai1 þ Ai2  I S i ; X3 ¼ P j þ h2 Q  S i  S Ti :
 Y.G. Sun et al. / Applied Mathematics and Computation 187 (2007) 1228–1237 1231

Proof. First we write (3) with u(k) = 0 in the equivalent descriptor form
        
I 0 xðk þ 1Þ I I xðkÞ 0 0 xðk  srðkÞ Þ
¼ þ :
0 0 yðk þ 1Þ A1 ðk; rðkÞÞ  I I yðkÞ A2 ðk; rðkÞÞ 0 yðk  srðkÞ Þ

That is

xðk þ 1Þ ¼ xðkÞ þ yðkÞ;
0 ¼ ½A1 ðk; rðkÞÞ  IxðkÞ  yðkÞ þ A2 ðk; rðkÞÞxðk  srðkÞ Þ;

which can be further rewritten in the form with delay in the variable y(k):
8
< xðk þ 1Þ ¼ xðkÞ þ yðkÞ;
>
kP
1
ð5Þ
: 0 ¼ yðkÞ þ ½A1 ðk; rðkÞÞ þ A2 ðk; rðkÞÞ  IxðkÞ  A2 ðk; rðkÞÞ
> yðlÞ:
l¼ksðk;rðkÞÞ

Next, we construct a switching Lyapunov function V(k) as

X
0 X
k1
V ðkÞ ¼ xT ðkÞP rðkÞ xðkÞ þ h y T ðkÞQyðlÞ: ð6Þ
h¼hþ1 l¼k1þh

Let the node at times k and k + 1 be i and j, respectively. That is, r(k) = i and r(k + 1) = j for any i; j 2 N.
Along the solution of (3) with u(k) = 0 and using Lemma 3, we have

X
k1
DV ðkÞ ¼ V ðk þ 1Þ  V ðkÞ ¼ xT ðk þ 1ÞP j xðk þ 1Þ  xT ðkÞP i xðkÞ þ h2 y T ðkÞQyðkÞ  h y T ðlÞQyðlÞ
l¼kh

6 2xT ðkÞP j yðkÞ þ xT ðkÞðP j  P i ÞxðkÞ þ y T ðkÞðP j þ h2 QÞyðkÞ

! !
X
k1 X
k1
T
 y ðlÞ Q yðlÞ : ð7Þ
l¼ksðk;iÞ l¼ksðk;iÞ

Due to (5), we have

( )
X
k1
2½xT ðkÞRTi þ y T ðkÞS Ti  yðkÞ þ ½A1 ðk; iÞ þ A2 ðk; iÞ  IxðkÞ  A2 ðk; iÞ yðlÞ ¼ 0: ð8Þ
l¼ksðk;iÞ
Pk1
Set vðkÞ ¼ l¼ksðk;iÞ yðlÞ and combine (7) with (8), we get

DV ðkÞ 6 eT ðkÞUði; jÞeðkÞ;

where eT(k) = [xT(k) yT(k) vT(k)], and

2 3
U1 U2 RTi A2 ðk; iÞ
6 7
Uði; j; kÞ ¼ 4  U3 S Ti A2 ðk; iÞ 5 ð9Þ
  Q

with
T
U1 ¼ RTi ½A1 ðk; iÞ þ A2 ðk; iÞ  I þ ½A1 ðk; iÞ þ A2 ðk; iÞ  I Ri þ P j  P i ;
T
U2 ¼ P j  RTi þ ½A1 ðk; iÞ þ A2 ðk; iÞ  I S i ; U3 ¼ P j þ h2 Q  S i  S Ti :
1232 Y.G. Sun et al. / Applied Mathematics and Computation 187 (2007) 1228–1237

Using the uncertain condition (2), we have

2 3 2 T 3
U1 U2 RTi Ai2 Ri H i1
6 7 6 7
Uði; j; kÞ ¼ 4  U3 S Ti Ai2 5 þ 4 S Ti H i1 5F ðkÞ½Ei1 þ Ei2 0  Ei2 
  Q 0
2 T
3
ðEi1 þ Ei2 Þ
6 7 T
þ4 0 5F ðkÞ½H Ti1 Ri H Ti1 S i 0; i; j 2 N; ð10Þ
ETi2
where U1 and U2 are taken from U1 and U2 by replacing A1(k, i) and A2(k, i) with Ai1 and Ai2, respectively. In
order to prove DV(k) < 0 for k P 0, it suffices to prove that U(i, j; k) < 0. By Lemma 2, a sufficient condition
guaranteeing U(i, j; k) < 0 is that there exist positive constants ki such that
2 3 2 T 3 2 3
U1 U2 RTi Ai2 Ri H i1 ðEi1 þ Ei2 ÞT
6 7 1 6 7 6 7
4  U3 S Ti Ai2 5 þ ki 4 S Ti H i1 5½H Ti1 Ri H Ti1 S i 0 þ ki 4 0 5½Ei1 þ Ei2 0  Ei2  < 0; i; j 2 N:
  Qm 0 ETi2
ð11Þ
Applying Schur complement (Lemma 1), we have that (4) is equivalent to (11). Therefore we have DV(k) < 0
for all k P 0. Using the standard Lyapunov stability theory we have that system (3) with u(k) = 0 is URAS.
This completes the proof of Theorem 1. h
The remaining of this section is devoted to design a robust memoryless state feedback controller in the form
of u(k) = Kr(k)x(k) such that the resulting closed-loop system is URAS.
Theorem 2. If for each i 2 N, there exist constants ki > 0 and matrices Xi,U > 0, Yi and Zi such that the
following LMIs hold:
2 3
X i P1 0 0 P2 X i þ Z Ti hZ Ti
6 7
6  P3 Ai2 U H i1 0 Y Ti hY Ti 7
6 7
6   U 0 UETi2 0 0 7
6 7
6 7
6    ki I 0 0 0 7 < 0; ð12Þ
6 7
6     1
ki I 0 0 7 7
6
6 7
4      X j 0 5
      U
where i; j; m 2 N, and
P1 ¼ X i ðAi1 þ Ai2  IÞT þ K e T BT  Z T ;
i i2 i
e T ET ; P3 ¼ Y i  Y T :
P2 ¼ X i ðEi1 þ Ei2 ÞT þ K i i4 i
Then system (3) is uniformly robust asymptotically stabilizable with state feedback u(k) = Kix(k), where
e i X 1 .
Ki ¼ K i

Proof. Define the Lyapunov functional as (6) and assume that r(k) = i, r(k + 1) = j, i; j 2 N. Similar to the
proof of Theorem 1, we have that the sufficient condition for uniform robust asymptotic stability of system
(3) with feedback u(k) = Kix(k) is
2 3 2 T 3
W1 W2 RTi Ai2 Ri H i1
Wði; j; kÞ ¼ 4  W3 S Ti Ai2 5 þ 4 S Ti H i1 5F ðk; iÞ½ Ei1 þ Ei2 þ Ei4 K i 0 Ei2 
  Q 0
2 T
3 ð13Þ
ðEi1 þ Ei2 þ Ei4 K i Þ  
6 7 T
þ4 0 5F ðk; iÞ H Ti1 Ri H Ti1 S i 0 < 0;
ETi2
 Y.G. Sun et al. / Applied Mathematics and Computation 187 (2007) 1228–1237 1233

where i; j 2 N, and
W1 ¼ RTi ½Ai1 þ Ai2 þ Bi2 K i  I þ ½Ai1 þ Ai2 þ Bi2 K i  IT Ri þ P j  P i ;
W2 ¼ P j  RTi þ ½Ai1 þ Ai2 þ Bi2 K i  IT S i ; W3 ¼ P j þ h2 Q  S i  S Ti :
Using Schur complement and Lemma 2, we have that a sufficient condition for (13) is that there exist positive
constants ki such that
2 T 3
W1 W2 RTi Ai2 RTi H i1 ðEi1 þ Ei2 þ Ei4 K i Þ I 0
6 T T 7
6  W3 S i Ai2 S i H i1 0 I hI 7
6 7
6   Q 0 ETi2 0 0 7
6 7
6 7
6    ki I 0 0 0 7 < 0; ð14Þ
6 7
6     1
ki I 0 0 77
6
6 7
4      P 1
j 0 5
      Q1
where i; j; m 2 N,
T
W1 ¼ RTi ½Ai1 þ Ai2 þ Bi2 K i  I þ ½Ai1 þ Ai2 þ Bi2 K i  I Ri  P i ;
T
W2 ¼ RTi þ ½Ai1 þ Ai2 þ Bi2 K i  I S i ; W3 ¼ S i  S Ti :
On the other hand, noting that
" #    T
W1 W2  2I 0  2I 0
¼ W Ti þ W i;
 W4 Ai1 þ Ai2 þ Bi2 K i  I I Ai1 þ Ai2 þ Bi2 K i  I I
where
  " #
Pi 0 P 1 0
Wi ¼ and hence W 1
i ¼
i
:
Ri Si S 1 1
i Ri P i S 1
i

Set X i ¼ P 1 1 1 e i ¼ K i X i . Multiply (14) by diag½ðW 1 ÞT ; U ; I; I; I; I; I; I

i , Y i ¼ S i , Zi = YiRiXi, U = Q and K i
1
and diag½W i ; U ; I; I; I; I; I; I on the left-hand side and the right-hand side, respectively, we have (12). This
completes the proof of Theorem 2. h

4. H‘ controller synthesis

This section is devoted to synthesizing a memoryless controller given in the form u(k) = Kr(k)x(k) that sta-
bilizes system (1) and guarantees the closed-loop system reaches the noise attenuation level c.
Theorem 3. If for each i 2 N, there exist constants ki > 0 and matrices Xi, U > 0, Yi and Zi such that the
following LMIs hold:
2 3
X i K1 0 0 K2 0 K3 X i þ Z Ti hZ Ti
6 7
6  K4 Ai2 U Bi1 0 H i1 0 Y Ti hY Ti 7
6 7
6   U 0 UC 0 UETi2 0 0 7
6 i2 7
6 7
6    c2 I DTi1 0 ETi3 0 0 7
6 7
6  0 7 ð15Þ
6    I H i2 0 0 7 < 0;
6 7
6      ki I 0 0 0 7
6 7
6       k1 0 0 7
6 i I 7
6 7
4        X j 0 5
        U
1234 Y.G. Sun et al. / Applied Mathematics and Computation 187 (2007) 1228–1237

where i; j 2 N,
K 1 ¼ P1 ; e T DT ;
K2 ¼ X i ðC i1 þ C i2 ÞT þ K K 3 ¼ P2 ; K4 ¼ P3
i i2

and P1, P2, P3 are the same as in (12). Then system (1) is uniformly robust asymptotically stablilizabe with c-
e i X 1 .
disturbance attenuation, and the state feedback gain is K i ¼ K i

Proof. Rewrite system (1) with u(k) = Kr(k)x(k) in the equivalent descriptor system

8
>
> xðk þ 1Þ ¼ xðkÞ þ yðkÞ;
>
>
>
> 0 ¼ yðkÞ þ ½A1 ðk; rðkÞÞ þ A2 ðk; rðkÞ þ B2 ðk; rðkÞÞK rðkÞ  IxðkÞ þ B1 ðk; rðkÞÞwðkÞ
>
>
>
>
>
> P
k1
>
< A2 ðk; rðkÞÞ yðlÞ;
l¼ksðk;rðkÞÞ
>
>
>
> zðkÞ ¼ ½C 1 ðk; rðkÞÞ þ C 2 ðk; rðkÞÞ þ D2 ðk; rðkÞÞK rðkÞ xðkÞ þ D1 ðk; rðkÞÞwðkÞ
>
>
>
>
>
> P
>
>
k1
>
: C 2 ðk; rðkÞÞ yðlÞ
l¼ksðk;rðkÞÞ

Define the Lyapunov functional as (6) and assume that r(k) = i, r(k + 1) = j, i; j 2 N. Similar to the proof of
Theorem 1, we have that

U ðkÞ :¼ DV ðkÞ þ zT ðkÞzðkÞ  c2 wT ðkÞwðkÞ ¼ gT ðkÞNði; j; kÞgðkÞ; ð16Þ

where gT(k) = [xT(k) yT(k) zT(k)wT(k)], z(k) is defined as in Theorem 2, and

2 3 2 3
N1 N2 RTi A2 ðk; iÞ RTi B1 ðk; iÞ NT4
6  N3 S Ti A2 ðk; iÞ S Ti B1 ðk; iÞ 7 6 0 7
6 7 6 7
Nði; j; kÞ ¼ 6 7þ6 7½N4 0  C 2 ðk; iÞD1 ðk; iÞ;
4   Q 0 5 4 C 2 ðk; iÞ 5
T

   c2 I DT1 ðk; iÞ

where

T
N1 ¼ RTi ½A1 ðk; iÞ þ A2 ðk; iÞ þ B2 ðk; iÞK i  I þ ½A1 ðk; iÞ þ A2 ðk; iÞ þ B2 ðk; iÞK i  I Ri þ P j  P i ;
T
N2 ¼ P j  RTi þ ½A1 ðk; iÞ þ A2 ðk; iÞ þ B2 ðk; iÞK i  I S i ;
N3 ¼ P j þ h2 Q  S i  S Ti ; N4 ¼ C 1 ðk; iÞ þ C 2 ðk; iÞ þ D2 ðk; iÞK i :

In order to prove U(k) < 0 for k P 0, it suffices to show that N(i, j;k) < 0 for i; j 2 N and k P 0. Using Schur
complement, we have that N(i, j;k) < 0 is equivalent to
2 3
N1 N2 RTi A2 ðk; iÞ RTi B1 ðk; iÞ NT4
6 7
6  N3 S Ti A2 ðk; iÞ S Ti B1 ðk; iÞ 0 7
6 7
6 T 7
6   Q 0 C 2 ðk; iÞ 7 < 0; i; j 2 N:
6 7
6 7
4    c2 I DT1 ðk; iÞ 5
    I
 Y.G. Sun et al. / Applied Mathematics and Computation 187 (2007) 1228–1237 1235

From (2), that is

2 3 2 3
N1 N2 RTi Ai2 RTi Bi1 NT4 RTi H i1
6 7 6 T 7
6  N3 S Ti Ai2 S Ti Bi1 0 7 6 7
6 7 6 S i H i1 7
6 T 7 6 7
6   Q 0 C i2 7 þ 6 0 7F ðk; iÞ½ Ei1 þ Ei2 þ Ei4 K i 0 Ei2 Ei3 0
6 7 6 7
6    c2 I DTi1 7 6 7
4 5 4 0 5
    I H i2
2 3
ðEi1 þ Ei2 þ Ei4 K i ÞT
6 7
6 0 7
6 7  
6 7 T
þ6 ETi2 7F ðk; iÞ H Ti1 Ri H Ti1 S i 0 0 H Ti2 < 0;
6 7
6 ETi3 7
4 5
0

where i; j 2 N, N1 , N2 , N4 are taken from N1, N2, N4 by replacing At(k, i), B2(k, i), Ct(k, i), D2(k, i) with Ait, Bi2,
Cit, Di2 (t = 1, 2), respectively. Similar to the proof of Theorem 2, we have that (15) implies N(i, j;k) < 0. From
(16), we can easily obtain DV(k) < 0 for w(k) = 0, which means that system (3) with w(k) = 0 is uniformly ro-
bust asymptotically stabilizable. On the other side, we have that zT(k)z(k)  c2wT(k)w(k) 6 DV(k). Summing
it from k = 0 to l and letting l ! 1, we have zT(k)z(k)  c2wT(k)w(k) 6 V(0) = 0. This completes the proof of
Theorem 3. h

5. Remarks and numerical examples

Remark 1. Our results given here are more effective for the case when the switching signal is restricted. If we
impose some restrictions on the switching signal, we usually do not need to resolve N2 LMIs. For example,
consider the periodic switching r(i) = i + 1 for i = 1, 2, . . . , N  1 and r(N) = 1. In this case, indexes i,j in
Theorems 1–3 have the following relation j = i + 1 for i = 1, 2, . . . , N  1 and j = 1 for i = N. Thus, we only
need to solve N LMIs in this case.

Remark 2. All the results in this paper are delay-dependent. By using the resolver GEVP in LMI Tool Box of
MATLAB [8]. We can easily obtain an upper bound of the maximal time delay h guaranteeing the uniform robust
asymptotic stability and stabilization from Theorems 1 and 2. By Theorem 3, we can get an optimal noise
attenuation level c for fixed h. Similarly, an upper bound of h can also be obtained for the fixed noise atten-
uation level c.

Remark 3. For the problem of robust stability, stabilization and H1 control via output-feedback of uncertain
discrete-time switched systems with mode-dependent time delays, we will study it in our next paper.
In the sequel, we will present two examples to illustrate the effectiveness of our results.
Example 1. The system is described by (3) with u(k) = 0 and assumed to have two modes, i.e., N ¼ f1; 2g.
The system data are as follows:
   
0:8 0:2 0:1 0
A11 ¼ ; A12 ¼ ; E11 ¼ E12 ¼ 0:01I; H 11 ¼ 0:1I;
0 0:91 0:1 0:1
   
0:3 0 0:12 0
A21 ¼ ; A22 ¼ ; E21 ¼ E22 ¼ 0:01I; H 21 ¼ 0:1I:
0 0:58 0:11 0:11

Applying Theorem 1 to this example shows that the system is robust stable for h 6 5. That is, this system is
robust stable for any time delay s(k, i) satisfying 0 < s(k, i) 6 5 for k P 0 and i = 1, 2.
1236 Y.G. Sun et al. / Applied Mathematics and Computation 187 (2007) 1228–1237

Example 2. The system is described by (1) and assumed to have have two modes, i.e., N ¼ f1; 2g. The system
data are
   
1 0 0:02 0:05
A11 ¼ ; A12 ¼ ;
0 1:01 0 0:01
BT11 ¼ ½ 0 0:01 ; BT ¼ ½ 0:02 0:01 ;
  12  
0:08 0 0:02 0
C 11 ¼ ; C 12 ¼ ;
0 0:01 0 0:01
DT11 ¼ ½ 0:02 0 ; DT ¼ ½ 0:01 0:03 ;
  12  
0:96 0 0:03 0:01
A21 ¼ ; A22 ¼
0 1:03 0 0:02
BT21 ¼ ½ 0 0:03 ; BT ¼ ½ 0:01 0:02 ;
  22  
0:03 0 0:02 0
C 21 ¼ ; C 22 ¼ ;
0 0:05 0 0:04
DT21 ¼ ½ 0:01 0:06 ; DT22 ¼ ½ 0:02 0:03 ;
Eij ¼ 0:01I; ETil ¼ ½ 0:01 0 ; H ij ¼ 0:1I; i; j ¼ 1; 2; l ¼ 3; 4:
By Theorem 3, we have that for any time delay s(k, i) satisfying 0 < s(k, i) 6 10, c = 0.0771 is the smallest value
of c for which the system is uniformly robust asymptotically stabilizable with disturbance attenuation. The
corresponding stabilizing control law is given as
e 1 X 1 ¼ ½ 3:5972
K1 ¼ K 1:7040 ; e 2 X 1 ¼ ½ 0:0910 3:1552 :
K2 ¼ K
1 2

On the other hand, if we restrict c = 0.08, by Theorem 3, we get the upper bound h = 9 of time delays guar-
anteeing the uniform robust asymptotic stabilizability of the system with disturbance attenuation. The corre-
sponding stabilizing control law is given as
e 1 X 1 ¼ ½ 6:0300
K1 ¼ K 2:6857 ; e 2 X 1 ¼ ½ 0:7574 3:3116 :
K2 ¼ K
1 2

6. Conclusions

By using a type of switched Lyapunov function and the descriptor system method, we establish some new
delay-dependent criteria for robust stability, stabilization and H1 control of uncertain discrete-time switched
systems with mode-dependent time delays via a memoryless state feedback. All the results are given in terms of
LMIs and can be easily tested with efficient convex optimal algorithms. Numerical examples have been pre-
sented to demonstrate the effectiveness of the results obtained in this paper.

Acknowledgements

The first author was supported by the National Natural Science Foundation of China (10671105) and the
Natural Science Foundation of Shandong Province of China (Y2005A06). The second and the third authors
were supported by the National Natural Science Foundations of China (10372002, 60274001, 60404001,
60528007), National 973 Program (2002CB312200) and the National Key Basic Research and Development
Program (2002CB312200).

References

[1] M.S. Branicky, Multiple Lyapunov functions and other analysis tools for switched and hybrid systems, IEEE Trans. Automat.
Control 43 (1998) 475–482.
[2] S. Boyd, L. El Ghaoui, E. Feron, V. Balakrishnan, Linear Matrix Inequalities in System and Control Theory, SIAM, Philadelphia,
PA, 1994.
 Y.G. Sun et al. / Applied Mathematics and Computation 187 (2007) 1228–1237 1237

[3] W.P. Dayawansa, C.F. Martin, A converse Lyapunov theorem for a class of dynamical systems which undergo switching, IEEE
Trans. Automat. Control 44 (1999) 751–760.
[4] J. Daafouz, P. Riedinger, C. Lung, Stability analysis and control synthesis for switched systems: a switched Lyapunov function
approach, IEEE Trans. Automat. Control 47 (2002) 1883–1887.
[5] E. Fridman, New Lyapunov–Krasovskii functionals for stability of linear retard and neutral type systems, Syst. Control Lett. 43
(2001) 309–319.
[6] E. Fridman, U. Shaked, A descriptor system approach to H1 control of linear time-delay systems, IEEE Trans. Automat. Control 47
(2002) 253–270.
[7] E. Feron, Quadratic stabilizability of switched systems via state and output feedback, Center for Intelligent Control Systems,
Massachusetts Institute of Technology, Cambridge, MA, Technical Report CICS P-468, February, 1996.
[8] P. Gahinet, V. Nemirovski, A.J. Laub, M. Chilali, LMI Control Toolbox for Use with Matlab, The Math Works, Natick, MA, 1995.
[9] K. Gu, V. Kharitonov, J. Chen, Stability of Time-Delay Systems, Birkhauser, Boston, 2003.
[10] X. Li, E. Fridman, U. Shaked, Robust H1 control of distributed delay systems with application to combustion control, IEEE Trans.
Automat. Control 46 (2001) 1930–1935.
[11] Z.G. Li, C.Y. Wen, Y.C. Soh, Stabilization of a class of switched systems via designing switching laws, IEEE Trans. Automat.
Control 46 (2001) 665–670.
[12] Z.G. Li, C.Y. Wen, Y.C. Soh, Observer-based stabilization of switching linear systems, Automatica 39 (2003) 517–524.
[13] D. Liberzon, J.P. Hespanha, A.S. Morse, Stability of switched systems: a Lie-algebraic condition, Syst. Control Lett. 37 (1999) 117–
122.
[14] D. Liberzon, A.S. Morse, Basic problems in stability and design of switched systems, IEEE Control Syst. Mag. 19 (1999) 59–70.
[15] A.S. Morse, Control Using Logic-Based Switching, Springer, London, 1997.
[16] Y.G. Sun, L. Wang, G. Xie, Delay-dependent robust stability and stabilization for discrete-time switched systems with mode-
dependent time-varying delays, Appl. Math. Comput. 180 (2006) 428–435.
[17] M.A. Wicks, P. Peleties, R.A. DeCarlo, Construction of piecewise Lyapunov functions for stabilizing switched systems, in:
Proceedings of IEEE Conference on Decision and Control, Lake Buena Vista, December 1994, pp. 3492–3497.
[18] M.A. Wicks, P. Peleties, R.A. DeCarlo, Switched controller synthesis for the quadratic stabilization of a pair of unstable linear
systems, Eur. J. Control 4 (1998) 140–147.
[19] Y. Wang, L. Xie, C.E. De Souza, Robust control of a class of uncertain nonlinear systems, Syst. Control Lett. 19 (1972) 139–149.
[20] D. Xie, L. Wang, F. Hao, G. Xie, LMI approach to L2 gain analysis and control synthesis of uncertain switched systems, IEE Proc.
Control Theory Appl. 151 (2004) 21–28.
[21] G. Xie, L. Wang, Controllability and stabilizability of switched linear systems, Syst. Control Lett. 48 (2003) 135–155.
[22] G. Xie, D. Zheng, L. Wang, Controllability of switched linear systems, IEEE Automat. Control 47 (2002) 1401–1405.
[23] G. Zhai, B. Hu, K. Yasuda, A.N. Michei, Disturbance attenuation properties of time-controlled switched systems, J. Franklin Inst.
338 (2001) 765–779.