Accepted Manuscript

Non-fragile finite-time extended dissipative control for a class of

uncertain discrete time switched linear systems

Jianwei Xia, Hui Gao, Mingxin Liu, Guangming Zhuang,

Baoyong Zhang

PII: S0016-0032(18)30126-1
DOI: 10.1016/j.jfranklin.2018.02.017
Reference: FI 3342

To appear in: Journal of the Franklin Institute

Received date: 22 August 2017

Revised date: 8 February 2018
Accepted date: 19 February 2018

Please cite this article as: Jianwei Xia, Hui Gao, Mingxin Liu, Guangming Zhuang, Baoyong Zhang,
Non-fragile finite-time extended dissipative control for a class of uncertain discrete time switched linear
systems, Journal of the Franklin Institute (2018), doi: 10.1016/j.jfranklin.2018.02.017

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service
to our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and
all legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Non-fragile finite-time extended dissipative control for a

¶
class of uncertain discrete time switched linear systems
Jianwei Xia∗ Hui Gao† Mingxin Liu‡ Guangming Zhuang§ Baoyong Zhang¶

Abstract

T
In this paper, the issues of finite-time extended dissipative analysis and non-fragile control are in-
vestigated for a class of uncertain discrete time switched linear systems. Based on average dwell-time

IP
approach, sufficient conditions for the finite-time boundedness and finite-time extended dissipative per-
formance of the considered systems are proposed by solving some linear matrix inequalities, where using

CR
the concept of extended dissipative, we can solve the H∞ , L2 −L∞ , Passivity and (Q, S, R)-dissipativity
performance in a unified framework. Furthermore, two form of non-fragile state feedback controllers
are designed to guarantee that the closed-loop systems satisfy the finite-time extended dissipative per-

US
formance. Finally, simulation example is given to show the efficiency of the proposed methods.

Keywords: Extended dissipative, finite-time, switch system, average dwell-time

AN
1 Introduction
M

Switched system is a hybrid system which consists of several subsystems and a switching signal that
orchestrates the switching among them. It has attracted considerable attention and has been widely
studied in past decades [1, 2]. Stability is a basic topic for switched systems, and lots of researchers
ED

have shown great interest toward Lyapunov asymptotic stability, which is defined over an infinite time
interval [3–9]. Meanwhile, in practice, the system behavior over a fixed finite time interval is also important
in many practical applications, such as neural networks, chemical engineering processing, robot control
PT

and missile systems. Recently, a great deal of finite-time analysis and control for different switched systems
were presented in the literatures [10–18] with different effective methods. For example, finite-time H∞
control of switched systems with mode-dependent average dwell time method was discussed in [10], the
CE

problem of adaptive finite-time control for a class of uncertain switched nonlinear systems using neural
networks method was investigated in [11], a common Lyapunov function and backstepping method were
AC

proposed to investigate the finite-time control for stochastic switched systems under arbitrary switching
in [12], etc.
As is well known, in general, the designed controller are assumed to be exactly implemented. However,
in many industrial applications, the imprecision occurring in controller implementation is always inevitable,
∗
School of Mathematics Science, Liaocheng University, Liaocheng 252000, P. R. China. Corresponding Author.
Email:[email protected]
†
School of Mathematics Science, Liaocheng University, Liaocheng 252000, P. R. China. Email:[email protected]
‡
College of information science and engineering, Yanshan University, Qinhuangdao 066004, China. Email: li-
[email protected]
§
School of Mathematics Science, Liaocheng University, Shandong 252000, China. Email: [email protected]
¶
School of Automation, Nanjing University of Science and Technology, Nanjing 210094, P.R. China Email: baoy-
[email protected]
1
 ACCEPTED MANUSCRIPT

which results from actuator degradation and numerical roundoff errors and so on. Thus, a controller which
is insensitive to uncertainties in their coefficients and be able to provide sufficient tuning margins is desired
to be designed, which leads to the study of non-fragile control in recent years [19–22]. Consequently, some
study of non-fragile control for switched systems was considered in recent years [23,24]. It should be noted
that these aforementioned works are only concerned with the continuous switched systems, just few paper
deal with the discrete cases.
On the other hand, an effective concept named extended dissipative was proposed by Zhang in [25],
where some well-known performance indices such as H∞ performance, L2 − L∞ performance, Passivity
performance and (Q, S, R)-dissipativity performance can be derived through tuning weighting matrices of
extended dissipative. The analysis and synthesis of extended dissipative has been applied to some neural
networks in recent few reports [26–29]. It is noted that the above mentioned systems are continuous

T
cases, [30] extend the concept to the discrete type cases. Naturally, could this concept be applied to the

IP
non-fragile control for discrete switched linear systems? To the best of our knowledge, this issue has not
been fully investigated yet, which motivates our current study.

CR
We will apply the extended dissipative concept to finite-time analysis and synthesis for a class of
uncertain discrete time switched linear systems in the paper. The organization of this paper is as follows. In
Section 2, Some basic Preliminaries and Problem formulations are introduced. In Section 3, the finite-time

US
boundness and finite-time extended dissipative for mentioned switched systems are proposed, where two
form of non-fragile state feedback controllers are also designed by solving a set of linear matrix inequalities.
In Section 4, simulation example is provided to show the effectiveness of the proposed approach. Finally,
AN
conclusions are given in Section 5.
Notation: The notations used in this paper are standard. Rn denotes the n−dimensional Euclidean
space, M T represents the transpose of the matrix M . The notations X > 0 is used to denote a symmetric
M

positive definite matrix. λmin (P ), λmax (P ) denote the minimum and maximum eigenvalue of matrix P
respectively. I is the identity matrix with appropriate dimensions and Z+ = {1, 2 · · ·}.
ED

2 Preliminaries and problem formulations

Consider the following uncertain discrete time switched linear system:
PT

x(k + 1) = (Aσ(k) + ∆Aσ(k) (k))x(k) + (Bσ(k) + ∆Bσ(k) (k))w(k) + Cσ(k) u(k),

z(k) = Dσ(k) x(k), k ∈ Z+ . (1)
CE

where x(k) ∈ Rn is the state vector, u(k) is the control input, w(k) is the exogenous disturbance, z(k) ∈ Rn
is the output. The switching signal σ(k) : Z+ 7→ M̃ = {1, 2...l} is a piecewise constant function, where
AC

l is the number of subsystems, σ(k) = i means that the ith subsystem is activated. Ai , Bi , Ci , Di , i =
1, 2 · · · l are known real constant matrices with appropriate dimensions. The parameter uncertainties
∆Ai (k), ∆Bi (k) are assumed to be of the form:
h i h i
∆Ai (k) ∆Bi (k) = Mi ∆i (k) N1i N2i , (2)

where Mi , N1i , N2i are known real constant matrices, ∆i (k) is unknown time-varying matrix satisfying
∆Ti (k)∆i (k) ≤ I.
In this paper, we will consider the following two form non-fragile state feedback controllers:

u(k) = (Ki + ∆Ki )x(k), (3)

2
 ACCEPTED MANUSCRIPT

Case 1: ∆Ki is with the norm-bounded additive form:

∆Ki = G1i J1i (k)G2i , (4)

Case 2: ∆Ki is with the norm-bounded multiplicative form:

∆Ki = L1i J2i (k)L2i Ki , (5)

where Ki are the controller gains and G1i , G2i , L1i , L2i are known real constant matrices, J1i (k), J2i (k) are
unknown time-varying matrix satisfying:

T
T T
J1i (k)J1i (k) ≤ I, J2i (k)J2i (k) ≤ I. (6)

IP
Assumption 1 [16] The external disturbance satisfies the constraint
t
X

CR
wT (k)w(k) ≤ d, ∀t ∈ Z+ ,
k=0

where d is a positive number.

Assumption 2 For ∀α ≥ 1, µ ≥ 1, ∀M > 0, we have

US
AN
αM µNσ (0,M ) ≤ h,

where Nσ (0, M ) denotes the switching number of σ(k) over (0, M ), and h is a positive number.
M

Assumption 3 The state vector satisfies the constraint

ED

t
X
xT (k)x(k) ≤ H, ∀t ∈ Z+ ,
k=0
PT

where H is a positive scalar.

Definition 1 [16] Given three positive constants c1 , c2 , M with c1 < c2 , a positive definite matrix R and
CE

a switching signal σ(k), assume that µ(k) ≡ 0, ∀k ∈ [0, M ], if

xT (0)Rx(0) ≤ c1 ⇒ xT (k)Rx(k) ≤ c2 , ∀k ∈ [1, M ],

AC

then switched system (1) is said to be finite-time bounded with respect to (c1 , c2 , R, M, σ). If the above
condition holds with w(k) ≡ 0, ∀k ∈ [1, M ], the system is said to be finite-time stable.

Definition 2 [16] For any T2 > T1 ≥ 0, let Nσ (T1 , T2 ) denotes the switching number of σ(k) over
(T1 , T2 ). If

T2 − T1
Nσ (T1 , T2 ) ≤ N0 +
τa
holds for τa > 0 and an integer N0 ≥ 0, then τa is called the average dwell-time and N0 is the chatter
bound. Without loss of generality, in this paper we choose N0 = 0.
3
 ACCEPTED MANUSCRIPT

Definition 3 [30] For given real matrices ψ1 = ψ1T ≤ 0, ψ2 , ψ3 = ψ3T > 0 and ψ4 = ψ4T ≥ 0 satisfying
(kψ1 k + kψ2 k)ψ4 = 0, system (1) is said to be extended dissipative if the following inequality holds for any
w(k) ∈ L2 [0, ∞):
M
X
J(k) ≥ sup z T (k)ψ4 z(k) , ∀M > 0, (7)
k=0 0≤k≤M

where J(k) = z T (k)ψ1 z(k) + 2z T (k)ψ2 w(k) + wT (k)ψ3 w(k).

Lemma 1 [31] Assume X, Y be real matrices of appropriate dimensions, they satisfy 2X T Y ≤ X T X +

Y TY .

T
Lemma 2 [32] For any real positive scalar ς > 0, we assume x, y and X, Y, Z and W > 0 be real vectors
and matrices with appropriate dimensions, and F (t) be function matrices satisfying F T (t)F (t) ≤ I, the

IP
following inequalities are always hold:
(1)X T F (t)Y + Y T F T (t)X ≤ ς −1 X T X + ςY T Y,

CR
(2)2xT X T F (t)Y y ≤ ς −1 xT X T Xx + ςy T Y T Y y,
(3) if W − ς −1 Y Y T > 0, then we have

3 Main results
US
(X + Y F (t)Z)T W −1 (X + Y F (t)Z) ≤ X T (W − ς −1 Y Y T )−1 X + ςZ T Z.
AN
3.1 Finite-time boundedness analysis
Consider the following unforced uncertain discrete time switched linear system:
M

x(k + 1) = (Aσ(k) + ∆Aσ(k) (k))x(k) + (Bσ(k) + ∆Bσ(k) (k))w(k),

z(k) = Dσ(k) x(k), k ∈ Z+ . (8)
ED

Theorem 1 Consider system (8), if there exist positive scalars , µ > 1, αi ≥ 1 and positive definite
symmetric matrices Pi , Qi with appropriate dimensions, such that
PT

 
TN
−αi Pi + N1i TN
N1i Pi ATi 0
1i 2i
 
 ∗ TN
−Qi + N2i 2i Pi Bi
T 0 
  < 0, (9)
 −Pi Pi Mi 
CE

 ∗ ∗ 
∗ ∗ ∗ −I
AC

Pi < µPj , ∀i, j ∈ M̃ , (10)

αM (λ2 c1 + λ3 d) < λ1 c2 , (11)
the average dwell-time satisfies
M lnµ
τa ≥ τa∗ = , (12)
ln(λ1 c2 ) − ln(αM (λ2 c1 + λ3 d))
and
1 1
α = max {αi }, λ1 = min (λmin (P̃i )), λ2 = max (λmax (P̃i )), λ3 = max (λmax (Qi )), P̃i = R− 2 Pi R− 2 ,
∀i∈M̃ ∀i∈M̃ ∀i∈M̃ ∀i∈M̃

then switched system (8) is finite-time bounded with respect to (c1 , c2 , M, R, d, σ).
4
 ACCEPTED MANUSCRIPT

Proof Choose the Lyapunov-Krasovskii functional candidate as

V (k) = Vσ(k) (k) = xT (k)Pσ(k) x(k). (13)

Assume σ(ki ) = i, for ∀k ∈ [ki , ki+1 ), we have

Vi (k + 1) − αi Vi (k) − wT (k)Qi w(k) = xT (k + 1)Pi x(k + 1) − αi xT (k)Pi x(k) − wT (k)Qi w(k)

= [(Ai + ∆Ai (k))x(k) + (Bi + ∆Bi (k))w(k)]T Pi [(Ai + ∆Ai (k))x(k) + (Bi + ∆Bi (k))w(k)]
−αi xT (k)Pi x(k) − wT (k)Qi w(k) ≤ X T (k)Θi X(k), (14)

where
h iT

T
X(k) = xT (k) wT (k) ,

IP
and
" #

CR
(Ai + ∆Ai (k))T Pi (Ai + ∆Ai (k)) − αi Pi (Ai + ∆Ai (k))T Pi (Bi + ∆Bi (k))
Θi = ,
∗ (Bi + ∆Bi (k))T Pi (Bi + ∆Bi (k)) − Qi

by Schur complement, Θi < 0 is equivalent to

Πi = 


−αi Pi
∗
0 US
(Ai + ∆Ai (k))T


−Qi (Bi + ∆Bi (k))T  < 0,
AN
∗ ∗ −Pi−1

here we consider the parameter uncertainties in the form of (2), by Lemma 2, we have
   
M

−αi Pi 0 (Ai + ∆Ai (k))T −αi Pi 0 ATi

   
Πi =  ∗ −Qi (Bi + ∆Bi (k))T  =  ∗ −Qi BiT  +
∗ ∗ −Pi−1 ∗ ∗ −Pi−1
ED

   
0 T
N1i
h i h i
   T  T
PT

 0  ∆i (k) N1i N2i 0 +  N2i  ∆i (k) 0 0 MiT <

Mi 0
CE

     
−αi Pi 0 ATi T
N1i h i 0 h i
   T  −1  
 ∗ −Qi BiT  +   N2i  N1i N2i 0 +   0  0 0 MiT (15)
AC

∗ ∗ −Pi−1 0 Mi

Pre- and post-multiplying (9) by diag{I, I, Pi−1 , I}, by Schur complement, we have
 
TN
−αi Pi + N1i TN
N1i ATi
1i 2i
 TN 
 ∗ −Qi + N2i 2i BiT  < 0, (16)
∗ ∗ −Pi−1 + −1 Mi MiT

then, we can easily obtain from the above discussion that Θi < 0. Furthermore, we have

Vi (k + 1) < αi Vi (k) + wT (k)Qi w(k), ∀k ∈ [ki , ki+1 ). (17)

5
 ACCEPTED MANUSCRIPT

Since α = max {αi }, we can deduce that

∀i∈M̃

k−1
X
k−ki
Vi (k) ≤ α Vi (ki ) + αk−s−1 wT (s)Qi w(s), ∀k ∈ [ki , ki+1 ). (18)
s=ki

Assume σ(ki ) = i, σ(ki − 1) = j, we have

V (ki ) = Vσ(ki ) (ki ) = Vi (ki ) = xT (ki )Pi x(ki ). (19)

Considering (10), we have

Vσ(ki ) (ki ) = xT (ki )Pi x(ki ) ≤ µxT (ki )Pj x(ki ) = µVj (ki ) = µVσ(ki −1) (ki ). (20)

T
Let N < M τa be the number of switching over [0, M ], For any k ∈ [0, M ], the switching sequence is denoted

IP
as k1 , k2 , · · · , km , m ≤ N, based on (19) and (20), using the iterative method, we can obtain that
1 −1
kX 2 −1
kX

CR
k N N k−s−1 T N −1
V (k) ≤ α µ Vσ(0) (0) + µ α w (s)Qσ(0) w(s) + µ αk−s−1 wT (s)Qσ(k1 ) w(s) + · · ·
s=0 s=k1
k−1
X
αk−s−1 wT (s)Qσ(km ) w(s).
+

Therefore
s=km
US (21)
AN
k−1
X
k N T N
V (k) ≤ α µ x (0)Pσ(0) x(0) + µ λ3 αk−s−1 wT (s)w(s)
s=0
M
X −1
M

1 1 1 1
≤ αM µN xT (0)R 2 (R− 2 Pσ(0) R− 2 )R 2 x(0) + αM µN λ3 wT (s)w(s)
s=0
≤ αM µN λ2 xT (0)Rx(0) + αM µN λ3 d ≤ αM µN (λ2 c1 + λ3 d). (22)
ED

On the other hand,

1 1 1 1
Vσ(k) (k) = xT (k)Pσ(k) x(k) = xT (k)R 2 (R− 2 Pσ(k) R− 2 )R 2 x(k) ≥ λ1 xT (k)Rx(k). (23)
PT

It follows from (22) and (23) that

M
CE

T αM µ τa (λ2 c1 + λ3 d)
x (k)Rx(k) < .
λ1
Using (12), one obtains
AC

xT (k)Rx(k) < c2 .

Then, the switched system (8) is finite-time bounded with respect to (c1 , c2 , M, R, d, σ) with Definition 1.
The proof is completed.

Remark 1 Multiply Lyapunov functions and average dwell time methods are used in the proof of Theorem
1 to analyze the finite-time boundedness for a class of discrete switch systems, which have been proved the
effective tools to obtained stability results with less conservatism [15–17].

If we ignore the effect of the exogenous disturbance w(k), the finite-time stability condition can be
easily obtained in the following Corollary.
6
 ACCEPTED MANUSCRIPT

Corollary 1 Consider system (8) with w(t) = 0. If there exist positive scalars µ > 1, αi ≥ 1 and positive
definite symmetric matrices Pi with appropriate dimensions, such that
 
TN
−αi Pi + N1i T
1i Pi Ai 0
 
 ∗ −Pi Pi Mi  < 0, (24)
∗ ∗ −I

Pi < µPj , ∀i, j ∈ M̃ , (25)

αM λ2 c1 < λ1 c2 , (26)

the average dwell-time satisfies

T
M lnµ

IP
τa ≥ τa∗ = , (27)
ln(λ1 c2 ) − ln(αM λ2 c1 )

CR
and
1 1
α = max {αi }, λ1 = min (λmin (P̃i )), λ2 = max (λmax (P̃i )), P̃i = R− 2 Pi R− 2 , (28)
∀i∈M̃ ∀i∈M̃ ∀i∈M̃

US
then switched system (8) is finite-time stable with respect to (c1 , c2 , M, R, σ).

Proof The proof is similar to that of Theorem 1, it is omitted here.

AN
3.2 Finite time extended dissipative analysis
Theorem 2 Consider system (8). For given matrices ψ1 , ψ2 , ψ3 , ψ4 , if there exist positive scalars h, , µ >
M

1, αi ≥ 1 and positive definite symmetric matrices Pi with appropriate dimensions, such that
 
−αi Pi − DiT ψ1 Di + N1i
TN T T
1i −Di ψ2 + N1i N2i Pi Ai
T 0
 
ED

 ∗ −ψ3 + N2iTN
2i Pi BiT 0 
  < 0, (29)
 ∗ ∗ −P P M 
 i i i 
∗ ∗ ∗ −I
PT

Pi < µPj , ∀i, j ∈ M̃ , (30)

1
Pi − DiT ψ4 Di > 0, ∀i ∈ M̃ ,
CE

(31)
h
the average dwell-time satisfies
M lnµ
AC

τa ≥ τa∗ = , (32)
ln(λ1 c2 ) − ln(λ4 H + (λ5 + λ6 )d) − M lnα

and

α = max {αi }, λ1 = min (λmin (P̃i )), λ2 = max (λmax (P̃i )), λ4 = max (λmax (DiT Di )),
∀i∈M̃ ∀i∈M̃ ∀i∈M̃ ∀i∈M̃
− 12 − 21
λ5 = λmax (ψ2T ψ2 ), λ6 = λmax (ψ3 ), P̃i = R Pi R .

Then the system is finite-time bounded and satisfies the extended dissipative performance.

7
 ACCEPTED MANUSCRIPT

Proof Choose the Lyapunov-Krasovskii functional candidate as

V (k) = Vσ(k) (k) = xT (k)Pσ(k) x(k). (33)

Assume σ(k + 1) = σ(k) = i, we have

Vi (k + 1) − αi Vi (k) − J(k)
= xT (k + 1)Pi x(k + 1) − αi xT (k)Pi x(k) − z T (k)ψ1 z(k) − 2z T (k)ψ2 w(k) − wT (k)ψ3 w(k)
= [(Ai + ∆Ai (k))x(k) + (Bi + ∆Bi (k))w(k)]T Pi [(Ai + ∆Ai (k))x(k) + (Bi + ∆Bi (k))w(k)]
−αi xT (k)Pi x(k) − xT (k)DiT ψ1 Di x(k) − 2xT (k)DiT ψ2 w(k) − wT (k)ψ3 w(k)
≤ X T (k)Φi X(k), (34)

T
where

IP
h iT
X(k) = xT (k) wT (k) ,

CR
and
" #
Φ11 (Ai + ∆Ai (k))T Pi (Bi + ∆Bi (k)) − DiT ψ2
Φi =
∗
US
(Bi + ∆Bi (k))T Pi (Bi + ∆Bi (k)) − ψ3
,
AN
Φ11 = (Ai + ∆Ai (k))T Pi (Ai + ∆Ai (k)) − αi Pi − DiT ψ1 Di ,

by Schur complement, Φi < 0 is equivalent to

 
M

−αi Pi − DiT ψ1 Di −DiT ψ2 (Ai + ∆Ai (k))T

 
Ψi =  ∗ −ψ3 (Bi + ∆Bi (k))T  < 0,
∗ ∗ −Pi−1
ED

similar to the proof of Theorem 1, we have

   
−αi Pi − DiT ψ1 Di −DiT ψ2 ATi T
N1i h i
PT

   T 
Ψi <  ∗ −ψ3 BiT  +   N2i  N1i N2i 0
∗ ∗ −Pi−1 0
CE

 
0 h i
 
+−1  0  0 0 MiT ,
AC

Mi

On the other hand, pre- and post-multiplying (29) by diag{I, I, Pi−1 , I}, by Schur complement, we have
 
−αi Pi − DiT ψ1 Di + N1i
TN T T
1i −Di ψ2 + N1i N2i ATi
 TN 
 ∗ −ψ3 + N2i 2i BiT  < 0, (35)
−1 −1 T
∗ ∗ −Pi +  Mi Mi

then, we can easily derive that Φi < 0, note that α = max {αi }, then we have
∀i∈M̃

Vi (k + 1) < αVi (k) + J(k), (36)

8
 ACCEPTED MANUSCRIPT

Let N < M τa be the number of switching over [0, M ], For any k ∈ [0, M ], the switching sequence is denoted
as k1 , k2 , · · · , km , m ≤ N, considering (30), we have

Vσ(ki ) (ki ) ≤ µVσ(ki −1) (ki ), (37)

based on (36) and (37), using the iterative method, we can obtain

1 −1
kX 2 −1
kX
k N N k−s−1 N −1
V (k) ≤ α µ Vσ(0) (0) + µ α J(s) + µ αk−s−1 J(s) + · · ·
s=0 s=k1
k−1
X k
X
+ αk−s−1 J(s) = αk µN Vσ(0) (0) + µNσ (s,k) αk−s−1 J(s). (38)
s=km s=0

T
Therefore, under zero initial condition with (38), we have

IP
k
X
0 ≤ V (k) ≤ µNσ (s,k) αk−s−1 J(s). (39)

CR
s=0

It follows from (39) that

0 ≤ V (k) ≤
k
X

s=0
µ Nσ (s,k) k−s−1
α
US J(s) < α µ k Nσ (0,k)
k
X

s=0
J(s). (40)
AN
Then we get

X k
V (k)
0 ≤ k N (0,k) < J(s). (41)
α µ σ
M

s=0

Setting k = M , we can obtain that

ED

X M
V (M )
0 ≤ M N (0,M ) < J(s), (42)
α µ σ s=0

by assumption 2, we have
PT

M
V (M ) X
0≤ < J(s). (43)
h
CE

s=0

Considering inequality
M
X
AC

J(k) − sup z T (k)ψ4 z(k) ≥ 0, (44)

k=0 0≤k≤M

when ψ4 = 0, it is apparent that

M
X
J(k) > 0, (45)
k=0

when ψ4 > 0, by (31) we have

M
X V (M ) xT (M )Pσ(M ) x(M )
J(k) > = > xT (M )DiT ψ4 Di x(M ) = z T (M )ψ4 z(M ), (46)
h h
k=0
9
 ACCEPTED MANUSCRIPT

so we get
M
X
J(k) − sup z T (k)ψ4 z(k) ≥ 0. (47)
k=0 0≤k≤M

The proof of extended dissipative is completed. Next, we proof finite-time boundedness. Following the
above proof, we have
k
X
V (k) < αk µNσ (0,k) J(s), (48)
s=0

and
k
X
lnµ
V (k) < e(lnα+ τa )k
J(s), (49)

T
s=0

IP
for ψ1 ≤ 0, we can derive that
k
X k
X

CR
J(s) ≤ (2z T (s)ψ2 w(s) + wT (s)ψ3 w(s)), (50)
s=0 s=0

and

US
k
X
(lnα+ lnµ )k
V (k) < e τ a (2z T (s)ψ2 w(s) + wT (s)ψ3 w(s)), (51)
s=0

so we have
AN
lnµ k
V (k) e(lnα+ τa )k X
T
x (k)Rx(k) < < (2z T (s)ψ2 w(s) + wT (s)ψ3 w(s)). (52)
λ1 λ1
s=0

Considering
M

k
X k
X
(2z T (s)ψ2 w(s) + wT (s)ψ3 w(s)) = (2xT (s)DiT ψ2 w(s) + wT (s)ψ3 w(s)), (53)
ED

s=0 s=0

by lemma 1 we can obtain that

2xT (s)DiT ψ2 w(s) ≤ xT (s)DiT Di x(s) + wT (s)ψ2T ψ2 w(s), (54)

PT

then we can get that

lnµ k
V (k) e(lnα+ τa )k X
CE

T
x (k)Rx(k) < < (2z T (s)ψ2 w(s) + wT (s)ψ3 w(s))
λ1 λ1
s=0
(lnα+ lnµ )k k
X
e τa
< (xT (s)DiT Di x(s) + wT (s)ψ2T ψ2 w(s) + wT (s)ψ3 w(s))
AC

λ1
s=0
(lnα+ lnµ )k lnµ
e τa e(lnα+ τa )M
< [λ4 H + (λ5 + λ6 )d] < [λ4 H + (λ5 + λ6 )d].
λ1 λ1
(55)

Therefore, it can be easily verified with (32) that

xT (k)Rx(k) ≤ c2 . (56)

Then, the system is finite-time bounded and satisfies the extended dissipative performance with Definition
3. The proof is completed.

10
 ACCEPTED MANUSCRIPT

Remark 2 By tuning the weighting matrices, the extended dissipative inequality contains some well-known
performance indices as special cases:
(1)H∞ performance: ψ1 = −I, ψ2 = 0, ψ3 = γ 2 I, ψ4 = 0 ;
(2)L2 − L∞ performance: ψ1 = 0, ψ2 = 0, ψ3 = γ 2 I, ψ4 = I ;
(3)Passivity performance: ψ1 = 0, ψ2 = I, ψ3 = γI, ψ4 = 0 ;
(4)(Q, S, R)-dissipativity performance: ψ1 = Q, ψ2 = S, ψ3 = R − βI, ψ4 = 0.

3.3 Non-fragile finite-time extended dissipative control

Consider system (1), under the controller µ(k) = (Kσ(k) + ∆Kσ(k) )x(k), the corresponding closed-loop
system is given by

T
x(k + 1) = (Aσ(k) + ∆Aσ(k) (k) + Cσ(k) Kσ(k) + Cσ(k) ∆Kσ(k) )x(k) + (Bσ(k) + ∆Bσ(k) (k))w(k),

IP
z(k) = Dσ(k) x(k), k ∈ Z+ . (57)

CR
For the additive form uncertainty ∆Ki = G1i J1i (k)G2i , we have the following theorem.

Theorem 3 For given matrices ψ1 , ψ2 , ψ3 , ψ4 , if there exist positive scalars h, , δ, µ > 1, αi ≥ 1 and



−αi Pi − DiT ψ1 Di + N1i
∗
T N + δGT G
1i
T

−ψ3 + N2i
US
positive definite symmetric matrices Pi and matrices Yi with appropriate dimensions, such that

T

TN
2i
T
2i 2i −Di ψ2 + N1i N2i Pi Ai + Yi Ci
Pi BiT
T 0
0
0
0



AN
 
 Pi Mi Pi Ci G1i 
 ∗ ∗ −Pi  < 0,
 
 ∗ ∗ ∗ −I 0 
∗ ∗ ∗ ∗ −δI
M

(58)

Pi < µPj , ∀i, j ∈ M̃ , (59)

ED

1
Pi − DiT ψ4 Di > 0, ∀i ∈ M̃ , (60)
h
PT

the average dwell-time satisfies

M lnµ
τa ≥ τa∗ = , (61)
ln(λ1 c2 ) − ln(λ4 H + (λ5 + λ6 )d) − M lnα
CE

and

α = max {αi }, λ1 = min (λmin (P̃i )), λ2 = max (λmax (P̃i )), λ4 = max (λmax (DiT Di )),
AC

∀i∈M̃ ∀i∈M̃ ∀i∈M̃ ∀i∈M̃

− 12 − 21
λ5 = λmax (ψ2T ψ2 ), λ6 = λmax (ψ3 ), P̃i = R Pi R . (62)

Then the system (57) is finite-time bounded and satisfies the extended dissipative performance under the
non-fragile controller µ(k) = (Kσ(k) + ∆Kσ(k) )x(k), where the controller gains can be given by Ki =
(Pi−1 Yi )T .

Proof Replacing Ai + ∆Ai (k) in the left side of (34) with Ai + ∆Ai (k) + Ci Ki + Ci ∆Ki , we can obtain
that
" #
Ξ11 (Ai + ∆Ai (k) + Ci Ki + Ci ∆Ki )T Pi (Bi + ∆Bi (k)) − DiT ψ2
Φi = , (63)
∗ (Bi + ∆Bi (k))T Pi (Bi + ∆Bi (k)) − ψ3
11
 ACCEPTED MANUSCRIPT

where

Ξ11 = (Ai + ∆Ai (k) + Ci Ki + Ci ∆Ki )T Pi (Ai + ∆Ai (k) + Ci Ki + Ci ∆Ki ) − αi Pi − DiT ψ1 Di , (64)

by Schur complement, Φi < 0 is equivalent to

 
−αi Pi − DiT ψ1 Di −DiT ψ2 (Ai + ∆Ai (k) + Ci Ki + Ci ∆Ki )T
 
Υi =  ∗ −ψ3 (Bi + ∆Bi (k))T  < 0, (65)
−1
∗ ∗ −Pi

here we consider the parameter uncertainties in the form of (2), by Lemma 2, we have
 
−αi Pi − DiT ψ1 Di −DiT ψ2 (Ai + Ci Ki )T
 

T
Υi =  ∗ −ψ3 BiT  (66)
∗ ∗ −Pi−1

IP
   

CR
0 T
N1i
h i h i
   T  T
+  0  ∆i (k) N1i N2i 0 +  N2i  ∆i (k) 0 0 MiT (67)
Mi 0



0


h i


GT2i

 T
+  0  J1i (k) G2i 0 0 +  0  J1i
h US
(k) 0 0 GT1i CiT
i
(68)
AN
Ci G1i 0

 
−αi Pi − DiT ψ1 Di −DiT ψ2 (Ai + Ci Ki )T
M

 
≤ ∗ −ψ3 BiT  (69)
−1
∗ ∗ −Pi
ED

   
T
N1i 0
h i h i
 T  −1   T
+  N2i  N1i N2i 0 +   0  0 0 M i (70)
PT

0 Mi

   
GT2i h i 0 h i
CE

   
+δ  0  G2i 0 0 + δ −1  0  0 0 GT1i CiT , (71)
0 Ci G1i
AC

Pre- and post-multiplying (58) by diag{I, I, Pi−1 , I, I}, by Schur complement, we have
 
Γ11 −DiT ψ2 + N1i TN
2i ATi + KiT CiT
 TN 
 ∗ −ψ3 + N2i 2i BiT  < 0, (72)
−1 −1 T −1 T T
∗ ∗ −Pi +  Mi Mi + δ Ci G1i G1i Ci

where
Γ11 = −αi Pi − DiT ψ1 Di + N1i
T
N1i + δGT2i G2i . (73)

It can be easily obtained that Φi < 0. The following proof is similar to that of Theorem 2, it is omitted
here.
For the multiplicative form uncertainty ∆Ki = L1i J2i (k)L2i Ki , we have the following theorem.
12
 ACCEPTED MANUSCRIPT

Theorem 4 For given matrices ψ1 , ψ2 , ψ3 , ψ4 , if there exist positive scalars h, , δ, µ > 1, αi ≥ 1 and
positive definite symmetric matrices Pi and matrices Yi with appropriate dimensions, such that
 
−αi Pi −Pi DiT ψ2 Pi ATi + Yi CiT Pi N1iT Y LT
i 2i Ri Di
T
 
 ∗ −ψ3 BiT T
N2i 0 0 
 
 ∗ ∗ −R + M M T + δC L LT C T 0 0 0 
 i i i i 1i 1i i 
  < 0, (74)
 ∗ ∗ ∗ −I 0 0 
 
 ∗ ∗ ∗ ∗ −δI 0 
 
∗ ∗ ∗ ∗ ∗ ψ1−1

Pi < µPj , ∀i, j ∈ M̃ , (75)

T
IP
1
Pi − DiT ψ4 Di > 0, ∀i ∈ M̃ , (76)
h

CR
the average dwell-time satisfies
M lnµ
τa ≥ τa∗ = , (77)
ln(λ1 c2 ) − ln(λ4 H + (λ5 + λ6 )d) − M lnα

and US
α = max {αi }, λ1 = min (λmin (P̃i )), λ2 = max (λmax (P̃i )), λ4 = max (λmax (DiT Di )),
AN
(78)
∀i∈M̃ ∀i∈M̃ ∀i∈M̃ ∀i∈M̃
1 1
λ5 = λmax (ψ2T ψ2 ), λ6 = λmax (ψ3 ), Ri = Pi−1 , P̃i = R− 2 Pi R− 2 . (79)
M

Then the system (57) is finite-time bounded and satisfies the extended dissipative performance under the
non-fragile controller µ(k) = (Kσ(k) + ∆Kσ(k) )x(k), where controller gains can be given by Ki = (Pi−1 Yi )T .
ED

Proof Replacing Ai + ∆Ai (k) in the left side of (34) with Ai + ∆Ai (k) + Ci Ki + Ci ∆Ki we can obtain
that
" #
Ξ11 (Ai + ∆Ai (k) + Ci Ki + Ci ∆Ki )T Pi (Bi + ∆Bi (k)) − DiT ψ2
Φi = , (80)
PT

∗ (Bi + ∆Bi (k))T Pi (Bi + ∆Bi (k)) − ψ3

where
CE

Ξ11 = (Ai + ∆Ai (k) + Ci Ki + Ci ∆Ki )T Pi (Ai + ∆Ai (k) + Ci Ki + Ci ∆Ki ) − αi Pi − DiT ψ1 Di ,

by Schur complement, Φi < 0 is equivalent to

AC

 
−αi Pi − DiT ψ1 Di −DiT ψ2 (Ai + ∆Ai (k) + Ci Ki + Ci ∆Ki )T
 
∆i =  ∗ −ψ3 (Bi + ∆Bi (k))T  < 0, (81)
−1
∗ ∗ −Pi

here we consider the parameter uncertainties in the form of (2), by Lemma 2, we have
 
−αi Pi − DiT ψ1 Di −DiT ψ2 (Ai + Ci Ki )T
 
∆i =  ∗ −ψ3 BiT + (82)
−1
∗ ∗ −Pi

13
 ACCEPTED MANUSCRIPT
   
0 N T
h i 1i h i
   T  T
 0  ∆i (k) N1i N2i 0 +  N2i  ∆i (k) 0 0 MiT + (83)
Mi 0

   
0 h i KiT LT2i h i
    T T T
 0  2i J (k) L2i Ki 0 0 +  0 J
 2i (k) 0 0 L1i Ci (84)
Ci L1i 0

 
−αi Pi − DiT ψ1 Di −DiT ψ2 (Ai + Ci Ki )T
 
≤ ∗ −ψ3 BiT  (85)

T
−1
∗ ∗ −Pi

IP
   
0 T
N1i
h i h i

CR
   T 
+  0  0 0 MiT + −1  N2i  N1i N2i 0 (86)
Mi 0



0


h i


+δ  0  0 0 LT1i CiT + δ −1  0
US

KiT LT2i h

i
 L2i Ki 0 0 , (87)
AN
Ci L1i 0

Pre- and post-multiplying (74) by diag{Pi , I, I, I, I, I}, by Schur complement, we have

 
M

Ψ11 −DiT ψ2 + −1 N1i

TN
2i ATi + KiT CiT
 
 ∗ −ψ3 + −1 N2iTN
2i BiT  < 0, (88)
∗ ∗ −1 T T T
−P + Mi Mi + δCi L1i L1i Ci
ED

where
Ψ11 = −αi Pi − DiT ψ1 Di + δ −1 KiT LT2i L2i Ki + −1 N1i
T
N1i . (89)
PT

We can also easily obtain that Φi < 0. The following proof is similar to that of Theorem 2, it is omitted
here.
CE

Remark 3 It is noted that, just simple discrete switch model is proposed in this paper, there is no other
complex factors are not considered. Therefore, it should be interesting study to extend the correspond-
ing results to more complex systems in further research when the time delay terms [33–37], stochastic
AC

disturbance [38–40], etc. will be included.

4 Numerical example
In this section, we present an example to illustrate the effectiveness of the controller design method.
Example: Consider system (57) with two subsystems with parameters as follows
" # " # " # " # " #
1 2 1 0 2 0 0.3 0 0.1 0.1
A1 = , B1 = , C1 = , D1 = , M1 = ,
1 0 0 2 1 2 0 0.3 0 0.2

14
 ACCEPTED MANUSCRIPT
" # " #
0.1 0.2 0.4 0
N11 = , N21 = ;
0 0.2 0.2 0.2
" # " # " # " # " #
2 3 1 0 1 4 0.2 0 0.1 0
A2 = , B2 = , C2 = , D2 = , M2 = ,
0 2 1 3 0 2 0 0.2 0.3 0.1
" # " #
0.3 0.3 0.2 0.1
N12 = , N22 = .
0 0.4 0.2 0.3
For extended dissipative control, choose matrices in Remark 2 for each case in Table 1.

Table 1: Matrices for each case

Analysis Ψ1 Ψ2 Ψ3 Ψ4

T
L2 − L∞ performance 0 0 γ2I I
H∞ performance −I 0 γ2I 0

IP
Passivity 0 I γ 0
Dissipativity −I I 10I − β ∗ I 0

CR
Firstly, we consider ∆Ki is with the norm-bounded additive form. We choose

G11 =
"
0.3 0.3
0 0.2
#
, G21 =
"
0.2 0
0 0.1
#
, G12 = US
"
0.3 0.2
0 0.2
#
, G22 =
"
0.3 0.1
0 0.3
#
,
AN
and αi ≡ 1.5,  = 0.1, δ = 0.1 in Theorem 3, by solving the LMIs presented in Theorem 3, we can obtain
the optimized variables of four performances in Table 2 and controller gain for each case in Table 3.

Table 2: Optimized variable for each case

M

Subsystem L2 − L∞ performance H∞ performance Passivity Dissipativity

1 2
γmin = 0.05902 2
γmin = 1.8786 γmin = 1.02816 βmax = 2.2669
ED

2 2
γmin = 0.5903 2
γmin = 1.35894 γmin = 0.8392 βmax = 8.4686

Table 3: Controller gain for each case

PT

Subsystem 1 2
" # " #
−0.4914 −1.5311 −1.5642 −0.0870
L2 − L∞ performance K1 = K2 =
CE

−0.4186 0.7694 −0.0098 −0.8619

" # " #
−0.2722 −1.1737 −1.4141 −0.0884
H∞ performance K1 = K2 =
−0.5829 0.7517 −0.0884 −0.8187
" # " #
AC

−0.1396 −1.1470 −2.8589 0.1882

Passivity K1 = K2 =
−0.6173 0.9892 0.2076 −0.7480
" # " #
−0.2529 −1.2104 −2.0131 0.0435
Dissipativity K1 = K2 =
−0.6147 0.8658 0.1575 −0.8007

Secondly, we consider ∆Ki is with the norm-bounded multiplicative form. We choose

" # " # " # " #
0.1 0 0.2 0 0.3 0 0.3 0
L11 = , L21 = , L12 = , L22 = ,
0.2 0.2 0.3 0.3 0.3 0.2 0.2 0.2
and αi ≡ 1.5,  = 0.1, δ = 0.1 in Theorem 4, by solving the LMIs presented in Theorem 4, we can obtain
the optimized variables of four performances in Table 4 and controller gain for each case in Table 5.
15
 ACCEPTED MANUSCRIPT

Table 4: Optimized variable for each case

Subsystem L2 − L∞ performance H∞ performance Passivity Dissipativity
1 2
γmin = 6.7498 2
γmin = 6.9405 γmin = 6.4369 βmax = 3.3662
2 2
γmin = 74.2711 2
γmin = 76.9558 γmin = 72.6476 βmax = −65.3072

Table 5: Controller gain for each case

Subsystem 1 2
" # " #
−0.7823 −0.8451 −0.7331 −0.3092
L2 − L∞ performance K1 = K2 =
0.0111 0.1946 −0.1934 −0.8248
" # " #
−0.7484 −0.8671 −0.7328 −0.3109
H∞ performance K1 = K2 =

T
−0.0250 0.2259 −0.1953 −0.8216
" # " #

IP
−0.7419 −0.8540 −0.7274 −0.3057
Passivity K1 = K2 =
−0.0224 0.2595 −0.1934 −0.8241
" # " #

CR
−0.7101 −0.8736 −0.7271 −0.3076
Dissipativity K1 = K2 =
−0.0546 0.2856 −0.1953 −0.8209

5 Conclusion
US
In this paper, we have investigated the problem of finite time extended dissipative analysis and non-
AN
fragile control for a class of uncertain discrete time switched linear systems. By employing average dwell
time approach, the problem of finite-time boundedness and finite-time extended dissipative analysis is
discussed. Based on extended dissipative performance, we can solve the H∞ , L2 − L∞ , Passivity and (Q,
S, R)-dissipativity performance in a unified framework. Two form of non-fragile state feedback controllers
M

are designed to guarantee the finite-time extended dissipative performance. A numerical example is also
provided to show the effectiveness of the proposed method.
ED

6 Acknowledgement
PT

This work was supported by Natural Science Foundation of China(No. 61573177 and 61773191), the
Natural Science Foundation of Shandong Province for Outstanding Young Talents in Provincial Universities
under Grant ZR2016JL025.
CE

References
AC

[1] D. Liberzon, A.S. Morse, Basic problems in stability and design of switched systems, IEEE Control
Systems Magazine, 19, 59-70, 1999.

[2] J. Daafouz, P. Riedinger, C. Iung, Stability analysis and control synthesis for switched systems: a
switched Lyapunov function approach, IEEE Trans Automat Control, 47(11): 1883-7, 2002.

[3] Z. Sun, S.Ge, Analysis and synthesis of switched linear control systems, Automatica, 41(2):181-95,
2005.

[4] D. Liberzon, Switching in Systems and Control, Boston: Birkhauser, 2003.

16
 ACCEPTED MANUSCRIPT

[5] C. Lien, K. Yu, Y. Chung, Y. Lin, L. Chung, J. Chen, Exponential stability analysis for uncertain
switched neutral systems with interval-time-varying state delay, Nonlinear Anal. Hybrid Syst, 3(3):
334-342, 2009.

[6] D. Ding, G.Yang, H∞ static output feedback control for discrete-time switched linear systems with
average dwell time, IET Control Theory Appl, 4(3):381-90, 2010.

[7] L. Zhang, P. Shi, Stability, L2 -gain and asynchronous H∞ control of discrete-time switched systems
with average dwell time, IEEE Trans Automat Control, 54(9): 2193-200, 2009.

[8] X. J. Su, P. Shi, L. Wu, Y. D. Song ,Fault detection filtering for nonlinear switched stochastic systems
IEEE Transactions on Automatic Control, 61(5): 1310C1315, 2016.

T
[9] X. J. Su, X. X. Liu, P. Shi, R. N. Yang, Sliding mode control of discrete-time switched systems with

IP
repeated scalar nonlinearity IEEE Transactions on Automatic Control, 62(9): 4604C4610, 2017.

CR
[10] H. Liu, X. Zhao, Finite-time H∞ control of switched systems with mode-dependent average dwell
time, J. Frankl.Inst, 351: 1301-1315, 2014.

[11] M.J. Cai, Z.R. Xiang, Adaptive neural finite-time control for a class of switched nonlinear systems,
Neurocomputing, 155, 177C185, 2015.
US
[12] S. P. Huang, Z. R. Xiang, Finite-time stabilization of a class of switched stochastic nonlinear systems
AN
under arbitrary switching, Int. J. Robust Nonlinear Control, 26: 2136-2152, 2016.

[13] S. Wang, T.G Shi, L.X. Zhang, A. Jasra, M. Zeng, Extended finite-time H∞ control for uncertain
switched linear neutral systems with time-varying delays, Neurocomputing, 152: 377-387, 2015.
M

[14] H. Liu, Y. Shen, X.D. Zhao, Finite-time stabilization and boundedness of switched linear system
under state-dependent switching, J. Frankl.Inst, 350: 541-555, 2013.
ED

[15] Z. Xiang, Y. Sun, M.S. Mahmoud, Robust finite-time H∞ control for a class of uncertain switched
neutral systems, Commun. Nonlinear Sci. Numer. Simul, 17(4): 1766-1778, 2012.
PT

[16] X.Z Lin, H.B Du, S.H Li, Y. Zou, Finite-time boundedness and finite-time L2 -gain analysis of discrete-
time switched linear systems with average dwell time, J. Frankl.Inst, 350: 911-928, 2013.
CE

[17] W. Xiang, J. Xiao, H∞ finite-time control for switched nonlinear discrete-time systems with norm-
bounded disturbance, J Franklin Inst, 348(2): 331-52, 2011.
AC

[18] X. Y. Li, W. J. Mao, Finite-Time H Output Tracking Control for a Class of Switched Neutral Systems
with Mode-Dependent Average Dwell Time Method International Journal of Innovative Computing,
Information and Control, 13(3): 767-782, 2017.

[19] C.H. Lien, W.C. Cheng, C.H. Tsai, Non-fragile observer-based controls of linear system via LMI
approach, Chaos Solitons Fractals, 32: 1530-1537, 2007.

[20] S.L. Wo, Y. Zou, Q.W. Chen, Non-fragile controller design for discrete descriptor systems, J. Frankl.
Inst, 346: 914-922, 2009.

[21] G.H. Yang, J.L. Wang, Non-fragile H∞ control for linear systems with multiplicative controller gain
variations, Automatica, 37 (5): 727-737, 2001.
17
 ACCEPTED MANUSCRIPT

[22] Y. Ma, N. Gu, Q. Zhang, Non-fragile robust H∞ control for uncertain discrete time singular systems
with time-varying delays, Journal of the Franklin Institute, 351(6): 3163-3181, 2014.

[23] Y. H. Liu, Y.G. Niu, Y.Y Zou, Non-fragile observer-based sliding mode control for a class of uncertain
switched systems, Journal of the Franklin Institute, 351, 952-963, 2014.

[24] Z. W. Xu, H. Y. Su, R.Q. Lu, Z.G. Wu. Non-fragile dissipative filtering for 2-D switched systems,
Journal of the Franklin Institute, 354: 6234-6246, 2017.

[25] B.Y. Zhang, W.X. Zheng, S.Y. Xu, Filtering of Markovian jump delay systems based on a new
performance index, IEEE Trans.Circuits Syst.I Reg.Pap, 60: 1250-1263, 2013.

T
[26] J.Y. Xiao, Y.T Li, S.M Zhong, F Xu,et. al, Extended dissipative state estimation for memristive
neural networks with time-varying delay, ISA Transactions, 64: 113-128, 2016.

IP
[27] J.W. Xia, G. L. Chen, W. Sun, Extended dissipative analysis of generalized Markovian switching

CR
neural networks with two delay components, Neurocomputing, 260: 275-283, 2017.

[28] Tae H. Lee, Myeong-Jin Park, Ju H. Park, Oh-Min Kwon, Sang-Moon Lee, Extended Dissipative
Analysis for Neural Networks With Time-Varying Delays, IEEE Trans. Neural Networks and Learning
Systems, 25: 1936-1941, 2014.
US
[29] H.L. Yang, L. Shu, S.M. Zhong, X. Wang, Extended dissipative exponential synchronization of com-
AN
plex dynamical systems with coupling delay and sampled-data control,J. Frankl.Inst, 353: 1829-1847,
2016.

[30] H. Shen, Y.Z. Zhu, L.X. Zhang and J.H. Park, Extended Dissipative State Estimation for Markov
M

Jump Neural Networks With Unreliable Links,IEEE Trans. Neural Netw. Learning Syst, 28: 346-358,
2017.
ED

[31] L. Xie, C. E. de Souza, Robust H∞ control for linear systems with normbounded time-varying uncer-
tainty, IEEE Transactions on Automatic Control, 37(8): 1188-1191, 1992.
PT

[32] Y Wang, L. Xie, C.E. de Souza, Robust control of a class of uncertain nonlinear systems,Systems and
Control Letters, 19: 139-149, 1992.
CE

[33] B. Y. Zhang, J. Lam, S. Y. Xu, Relaxed results on reachable set estimation of time-delay systems
with bounded peak inputs, Int. J. Robust Nonlinear Control, 26: 1994-2007, 2016.

[34] B. Y. Zhang, J. Lam, S. Y. Xu, Stability analysis of distributed delay neural networks based on
AC

relaxed Lyapunov-Krasovskii functionals, IEEE Trans. Neural Netw. Learn. Syst., 26: 1480-1492,
2015.

[35] J. Xia, Ju H. Park, H. Zeng, Improved Delay-dependent Robust Stability Analysis for Neutral-type
Uncertain Neural Networks with Markovian jumping Parameters and Time-varying Delays, Neuro-
computing, 149, 1198-1205, 2015.

[36] W. Qian, M. Yuan, L. Wang, X. H. Bu, J. Q. Yang, Stabilization of systems with interval time-varying
delay based ondelay decompsing approach, ISA Transactions, 70:1-6, 2017.

18
 ACCEPTED MANUSCRIPT

[37] W. Qian, L. Wang, M. Z.Q.Chen, Local consensus of nonlinear multiagentsystems with

varying delay coupling, IEEE Transations on Systems,Man,andCybernetics: Systems, DOI:
10.1109/TSMC.2017.2684911.

[38] J.W. Xia, J.H.Park, H.B.Zeng, H.Shen, Delay-difference-dependent robust exponential stability for
uncertain stochastic neural networks with multiple delays, Neurocomputing, 140: 210-218, 2014.

[39] H. Mukaidania, H. Xu, Infinite horizon linear-quadratic Stackelberg games for discrete-time stochastic
systems, Automatica, 76: 301-308, 2017.

[40] Z. G. Wu, P. Shi, Z.Shu, H. Su, R. Lu, Passivity-based asynchronous control for markov jump systems,
IEEE Transactions on Automatic Control, 62:2020-2025, 2017.

T
IP
CR
US
AN
M
ED
PT
CE
AC

19