International Journal of Control, Automation and Systems 16(3) (2018) 1459-1468 ISSN:1598-6446 eISSN:2005-4092

http://dx.doi.org/10.1007/s12555-017-0393-7 http://www.springer.com/12555

Robust Finite-time Extended Dissipative Control for a Class of Uncertain

Switched Delay Systems
Hui Gao, Jianwei Xia*, and Guangming Zhuang

Abstract: This paper investigates the problem of finite-time extended dissipative analysis and control for a class
of uncertain switched time delay systems, where the uncertainties satisfy the polytopic form. By using the average
dwell-time and linear matrix inequality technique, some sufficient conditions are proposed to guarantee that the
switched system is finite-time bounded and has finite-time extended dissipative performance, where the H∞ , L2 −L∞ ,
Passivity and (Q, S, R)-dissipativity performance can be solved simultaneously in a unified framework based on the
concept of extended dissipative. Furthermore, a state feedback controller is presented to guarantee that the closed-
loop system is finite-time bounded and satisfies the extended dissipative performance. Finally, a numerical example
is given to demonstrate the effectiveness of the proposed method.

Keywords: Average dwell-time, extended dissipative, finite-time, switch, time delay.

1. INTRODUCTION weighting matrices, the extended dissipative covers some

In the last decades, switched systems have received
considerable attention for their significant application in
various fields [1–5]. Switched system is a special class of
hybrid system, which consists of a finite number of sub-
systems and a logical rule that orchestrates switching be-
tween them. It is well known that time delay exists widely
in many practical systems and may cause undesirable sys-
tem performance or even instability [6]. Thus, great atten-
tion has been paid to switched delay systems in literatures
[7–10]. For example, for several switched delay systems,
the problem of stability and robust stabilization were con-
sidered in [8, 9]. [7] focused on the study of H∞ control. It
should be noted that, all the above literature related to sta-
bility of switched systems focus on Lyapunov asymptotic
stability, which is defined over an infinite time interval.
However, in practice, the transient performance of a sys-
tem is also of great significant. And numerous finite-time
analysis and control for different switched systems were
presented in recent literatures [11–14]. Specially, finite-
time boundedness and L2 -gain analysis for switched de-
lay systems was investigated in [11], robust finite-time H∞
control was addressed in [12], etc.
On the other hand, an important concept named ex-
tended dissipative was firstly proposed by Zhang in [15].
The novel feature of the concept is that by adjusting
well-known performance indices such as H∞ performance,
L2 − L∞ performance, Passivity performance and (Q, S,
R)-dissipativity performance. Recently, the concept of ex-
tended dissipative concept has been effectively applied to
several neural networks in reports [16–19]. To the best
of our knowledge, the extended dissipative concept has
not been applied to switched systems yet, and based on
above discussion, finite-time analysis for switched sys-
tems is worth researthing. Can we design the controller
for a class of switched time delay systems to make the
closed loop systems be finite-time extended dissipative?
This question is interesting and has not been investigated
yet, which motivates our current study.
This paper is organized as follows: In Section 2, Pre-
liminaries and Problem Statement are formulated. In Sec-
tion 3, finite-time boundness and finite-time extended dis-
sipative performance for switched delay systems are ad-
dressed, meanwhile, state feedback controllers are pro-
posed in terms of a set of linear matrix inequalities. In
Section 4, a numerical example is provided to show the
effectiveness of the proposed approach. Finally, conclu-
sions are given in Section 5. The most notable contribu-
tions of this paper can be summarized as follows: 1) The
concept of extended dissipative is successfully applied to
switched systems. 2) We addressed the problem of finite-
time extended dissipative analysis and control for a class

Manuscript received July 4, 2017; revised September 19, 2017; accepted October 10, 2017. Recommended by Associate Editor Young Ik Son
under the direction of Editor Duk-Sun Shim. This work was supported by Natural Science Foundation of China (No. 61573177, 61603170
and 61773191); the Natural Science Foundation of Shandong Province for Outstanding Young Talents in Provincial Universities under Grant
(ZR2016JL025).

Hui Gao, Jianwei Xia, and Guangming Zhuang are with the School of Mathematics Science, Liaocheng University, Liaocheng 252000, China
(e-mails: gaohui194011@sina.com, njustxjw@126.com, zgmtsg@126.com).
*Corresponding author.

⃝ICROS,
c KIEE and Springer 2018
1460 Hui Gao, Jianwei Xia, and Guangming Zhuang

of delayed uncertain switched systems, where the uncer- Assumption 4: For ∀α ≥ 0, µ ≥ 1, ∀t ∈ [0, Tf ], we
tainties satisfy the polytopic form. 3) We create a novel have
designed controller method to deal with the polytopic un-
certainties. eα t µ Nσ (0,t) ≤ b,

Nσ (0,t) denote the switching number of σ (t) over (0,t),

2. PRELIMINARIES AND PROBLEM and b a positive number.
STATEMENT Definition 1: [11] Given three positive constants
c1 , c2 , Tf with c1 < c2 , a positive definite matrix R and a
Consider the following uncertain switched system with switching signal σ (t), assume that µ (t) ≡ 0, ∀t ∈ [0, Tf ],
time delay: switched system (1) is said to be finite-time bounded with
respect to (c1 , c2 , R, Tf , σ ), if ∀t ∈ [0, Tf ],
ẋ(t) = Aσ (t) x(t) + Bσ (t) x(t − h(t)) + Dσ (t) w(t)
∫ t sup { xT (θ )Rx(θ ), ẋT (θ )Rẋ(θ )} ≤ c1
+ Eσ (t) u(t) + Gσ (t) x(s)ds, −τ ≤θ ≤0
t−r(t)
⇒ x (t)Rx(t) ≤ c2 .
T
z(t) = Fσ (t) x(t),
x(t0 + θ ) = φ (θ ), ∀θ ∈ [−τ , 0], (1) If the condition above holds with w(t) ≡ 0, ∀t ∈ [0, Tf ], the
system is said to be finite-time stable.
where x(t) ∈ Rn is the state vector, u(t) the control input, Definition 2 [11]: For any T2 > T1 ≥ 0, let Nσ (T1 , T2 )
w(t) the exogenous disturbance which belongs to L2 [0, ∞) denotes the switching number of σ (t) over (T1 , T2 ). If
, z(t) ∈ Rn the output. The switching signal σ (t) : [0, ∞) 7→
M = {1, 2...l} is a piecewise continuous function, where T2 − T1
Nσ (T1 , T2 ) ≤ N0 +
l is the number of subsystems, σ (t) = i means that the τa
ith subsystem is activated. For each σ (t) = i, the matrix holds for τa > 0 and an integer N0 ≥ 0, then τa is called
parameters Ai , Bi , Di , Ei , Fi , Gi belong to the following un- an average dwell-time. Without loss of generality, in this
certainty polytype: paper we choose N0 = 0.
p Definition 3 [15]: For given matrices ψ1 , ψ2 , ψ3 and ψ4
[Ai , Bi , Di , Ei , Fi , Gi ] = ∑ θ j [Ai j , Bi j , Di j , Ei j , Fi j , Gi j ], satisfying Assumption 3, system (1) is said to be extended
j=1 dissipative if the following inequality holds for any Tf ≥ 0
p and all µ (t) ∈ L2 [0, ∞):
∑ θ j = 1, θ j ≥ 0, (2)
∫ Tf
j=1
J(t)dt − sup zT (t)ψ4 z(t) ≥ 0, (4)
0 0≤t≤T f
where θ j ( j ∈ {1, 2...p}) are time-invariant uncertainties,
φ (θ ) is the initial condition, h(t), r(t) denote the time- where J(t) = zT (t)ψ1 z(t) + 2zT (t)ψ2 w(t) + wT (t)ψ3 w(t).
varying delay and satisfying Lemma 1 [16]: For any real matrices a, b of appropri-
ate dimensions, we have 2aT b ≤ aT a + bT b.
0 ≤ h(t) ≤ hm , ḣ(t) ≤ ĥ < 1, 0 ≤ r(t) ≤ rm . (3)
Lemma 2 [13]: For any positive definite symmetric
Assumption 1 [11]: For a given time constant Tf , the matrix N ∈ Rn×n , scalar τ > 0 and a vector function x(·) :
external disturbance satisfies [−τ , 0] → Rn the following integral inequality is satisfied
∫ ∫ ∫
∫ t t t
Tf
−τ xT (s)Nx(s)ds ≤ − xT (s)dsN x(s)ds.
w (t)w(t)dt ≤ d, d ≥ 0.
T
t−τ t−τ t−τ
0

Assumption 2: For a given time constant Tf , the state 3. MAIN RESULTS

vector x(t) is time-varying and satisfies the constraint
3.1. Finite-time boundedness analysis
∫ Tf Consider the following unforced switched system with
xT (t)x(t)dt ≤ k, time delay
0

where k is a fixed sufficient large constant number. ẋ(t) = Aσ (t) x(t) + Bσ (t) x(t − h(t)) + Dσ (t) w(t)
∫ t
Assumption 3 [15]: Matrices ψ1 , ψ2 , ψ3 , ψ4 satisfy the + Gσ (t) x(s)ds,
t−r(t)
following conditions:
1) ψ1 = ψ1T ≤ 0, ψ3 = ψ3T > 0, ψ4 = ψ4T ≥ 0; z(t) = Fσ (t) x(t),
2) (∥ψ1 ∥ + ∥ψ2 ∥)ψ4 = 0. x(t0 + θ ) = φ (θ ), ∀θ ∈ [−τ , 0]. (5)
 Robust Finite-time Extended Dissipative Control for a Class of Uncertain Switched Delay Systems 1461

 (l j)

Theorem 1: Consider system (5). For given positive ϕ11 P̃i j Bil P̃i j Dil P̃i j Gil
scalars α , ĥ, hm , rm , if there exist positive definite sym-  ∗ −(1 − ĥ)Q̃i j 
(l j)  0 0 
Θi = ,
metric matrices P̃i , Q̃i , M̃i and matrices Ni and Y jl with  ∗ ∗ −Ni 0 
Y j j = Y jTj , j = 1, 2..., p, l = j, ..., p with appropriate dimen- ∗ ∗ ∗
M̃
− rmi j
sions, such that (l j)
ϕ11 = −α P̃i j + P̃i j Ail + ATil P̃i j + Q̃i j + rm M̃i j ,
( j j)
Θi < Y j j , j = 1, 2...p, (6)
( jl) (l j) and
Θi + Θi < Y jl +Y jlT , j < l, j, l = 1, 2...p, (7)
  λmin (Pi ) = λ1 , λmax (Pi ) = λ2 , λmax (Qi ) = λ3 ,
Y11 Y12 · · · Y1p
 ∗ Y22 · · · Y2p  λmax (Mi ) = λ4 , λmax (Ni ) = λ5 . (11)
 
 .. ..  ≤ 0, (8)
 ∗ ∗ . . 
Then switched system (5) is finite-time bounded with re-
∗ ∗ ∗ Ypp spect to (c1 , c2 , d, R, Tf , σ ).
(λ2 + hm eα hm λ3 + rm eα rm λ6 )c1 + λ7 d < c2 λ1 e−α Tf , Proof: Choose the Lyapunov-Krasovskii functional
(9) candidate as

the average dwell-time satisfies V (t) = Vσ (t) (t) = Vi (t) = V1i (t) +V2i (t) +V3i (t),
(12)
τa > τa∗
Tf lnµ where
= ,
ln(λ1 c2 )
  V1i (t) = xT (t)P̃i x(t),
 − ln[(λ2 + hm eα hm λ3 + rm eα rm λ4 )c1 + λ5 d] ∫ t
− α Tf V2i (t) = eα (t−s) xT (s)Q̃i x(s)ds,
t−h(t)
∫ 0 ∫ t
where µ > 1 satisfies V3i (t) = eα (t−s) xT (s)M̃i x(s)dsd ε , (13)
−rm t+ε
P̃i < µ P̃j , Q̃i < µ Q̃ j , M̃i < µ M̃ j , ∀i, j ∈ M. (10)
in which α is a given scalar and P̃i , Q̃i , M̃i are positive def-
And inite matrices to be determined.
1 1 1 1 1 1
Taking the derivative of V (t) with respect to t along the
P̃i = R 2 Pi R 2 , Q̃i = R 2 Qi R 2 , M̃i = R 2 Mi R 2 , trajectory of system (5) yields

where V̇1i (t) = 2xT (t)P̃i ẋ(t),

p p
V̇2i (t) = α V2i (t) + xT (t)Q̃i x(t)
[Pi , Qi , Mi ] = ∑ θ j [Pi j , Qi j , Mi j ], ∑ θ j = 1, θ j ≥ 0,
j=1 j=1 − eα h(t) (1 − ḣ(t))xT (t − h(t))Q̃i x(t − h(t))

and ≤ α V2i (t) + xT (t)Q̃i x(t)

1 1 1 1 1 1
− (1 − ĥ)xT (t − h(t))Q̃i x(t − h(t)),
P˜i j = R 2 Pi j R 2 , Q˜i j = R 2 Qi j R 2 , M˜i j = R 2 Mi j R 2 .
V̇3i (t) = α V3i (t) + rm xT (t)M̃i x(t)
∫ t
Furthermore, − eα (t−s) xT (s)M̃i x(s)ds
 ( j j)
 t−r(t)
ϕ11 P̃i j Bi j P̃i j Di j P̃i j Gi j ≤ α V3i (t) + rm xT (t)M̃i x(t)
 ∗ −(1 − ĥ)Q̃i j  ∫
( j j)  0 0 
Θi = , t
 ∗ ∗ −Ni 0  − xT (s)M̃i x(s)ds
M̃ t−r(t)
∗ ∗ ∗ − rmi j
( j j)
ϕ11 = −α P̃i j + P̃i j Ai j + ATij P̃i j + Q̃i j + rm M̃i j ,
  by Lemma 2, it is easy to obtain that
( jl)
ϕ P̃il Bi j P̃il Di j P̃il Gi j
 11  ∫
( jl)  ∗ −(1 − ĥ)Q̃il 0 0  t
Θi =  , − xT (s)M̃i x(s)ds
 ∗ ∗ −Ni 0  t−r(t)
∗ ∗ ∗ − M̃rmil ∫ t ∫ t
1
( jl)
≤− xT (s)dsM̃i x(s)ds.
ϕ11 = −α P̃il + P̃il Ai j + ATij P̃il + Q̃il + rm M̃il , rm t−r(t) t−r(t)
1462 Hui Gao, Jianwei Xia, and Guangming Zhuang

∫ t
Thus + eα (t−s) wT (s)Ni w(s)ds
tk
V̇ (t) − α V (t) − wT (t)Ni w(t) ≤ X T (t)Θi X(t), =e α (t−0)
µ Nσ (0,t)V (0)
∫ t
where + eα (t−s) µ Nσ (s,t) wT (s)Ni w(s)ds
[ ∫t ]T 0
X(t) = xT (t) xT (t − h(t)) wT (t) t−r(t) xT (s)ds , α t Nσ (0,t)
<e µ V (0)
∫ t
meanwhile, we obtain from (6)-(8) that + µ Nσ (0,t) eα t wT (s)Ni w(s)ds
0
p p−1 p ∫ Tf
Θi = ∑
( j j)
θi2j Θi + ∑ ∑
( jl) (l j)
θ(i j) θ(il) (Θi + Θi ) <eα Tf µ Nσ (0,Tf ) [V (0) + wT (s)Ni w(s)ds]
0
j=1 j=1 l= j+1
p p−1 p <eα Tf µ Nσ (0,Tf ) [V (0) + λmax (Ni )d]. (16)
< ∑ θi2jY j j + ∑ ∑ θ(i j) θ(il) (Y jl +YlTj )
From Definition 3, we know
j=1 j=1 l= j+1
 
Y11 Y12 · · · Y1p Tf
 ∗ Y22 · · · Y2p  Nσ (0, Tf ) < ,
  T τa
= ∆ .. ..  ∆ ≤ 0, (14)
 ∗ ∗ . . 
such that
∗ ∗ ∗ Ypp
lnµ
V (t) < e(α + τa )Tf [V (0) + λ5 d]. (17)
where
  On the other hand,
Θ11 P̃i Bi P̃i Di P̃i Gi
 ∗ −(1 − ĥ)Q̃i 0 0 
Θi =  , 1 1

 ∗ V (t) > xT (t)P̃i x(t) = xT (t)R 2 Pi R 2 x(t)

∗ −Ni 0 
∗ ∗ ∗ − M̃ i ≥ λmin (Pi )xT (t)Rx(t) = λ1 xT (t)Rx(t), (18)
rm
Θ11 = −α P̃i + P̃i Ai + ATi P̃i + Q̃i + rm M̃i , V (0)
≤ λmax (Pi )xT (0)Rx(0)
and
+ hm eα hm λmax (Qi ) sup {xT(θ )Rx(θ ), ẋT (θ )Rẋ(θ )}
−τ ≤θ ≤0
∆ = [θi1 I, θi2 I, · · · , θip I], α rm
+ rm e λmax (Mi ) sup {xT(θ )Rx(θ ), ẋT (θ )Rẋ(θ )}
−τ ≤θ ≤0
Therefore, from (12) to (14), we can obtain that α hm
≤ [λmax (Pi ) + hm e λmax (Qi ) + rm eα rm λmax (Mi )]c1
V̇ (t) − α V (t) − w (t)Ni w(t) < 0.
T
(15) ≤ [λ2 + hm eα hm λ3 + rm eα rm λ4 ]c1 . (19)
Integrating (15), it can be obtained from (10) and (15) that, From (17)-(19), we have
for ∀t ∈ [tk ,tk+1 ),
∫ t xT (t)Rx(t)
α (t−tk ) α (t−s) T
V (t) <e V (tk ) + e w (s)Ni w(s)ds V (t)
tk
∫
≤ (20)
t λ1
<eα (t−tk ) µ V (tk− ) + eα (t−s) wT (s)Ni w(s)ds [λ2 + hm eα hm λ3 + rm eα rm λ4 ]c1 + λ5 d α Tf Tτ f
tk < e µ a.
λ1
<eα (t−tk ) µ [eα (tk −tk−1 )V (tk−1 )
∫ (21)
tk
+ eα (tk −s) wT (s)Ni w(s)ds]
tk−1 By virtue of (9), we have
∫ t
+ eα (t−s) wT (s)Ni w(s)ds ln(λ1 c2 ) − ln[(λ2 + hm eα hm λ3 + rm eα rm λ4 )c1 + λ5 d]
tk
α (t−0) − α Tf > 0.
<e µ Nσ (0,t)V (0)
∫ t1
+ µ Nσ (0,t) eα (t−s) wT (s)Ni w(s)ds Considering (10), we have
0
∫ t2 Tf ln(λ1 c2 ) − α Tf
+ µ Nσ (t1 ,t) eα (t−s) wT (s)Ni w(s)ds + · · · τa
<
ln(µ )
, (22)
t1
∫ tk ln[(λ2 + hm eα hm λ3 + rm eα rm λ4 )c1 + λ5 d]
+µ eα (t−s) wT (s)Ni w(s)ds − , (23)
tk−1 ln(µ )
 Robust Finite-time Extended Dissipative Control for a Class of Uncertain Switched Delay Systems 1463

−α T f
ln[ (λ2 +hm eα hmλλ13c+r
( j j)
2e
αr ] Ξ11
m e m λ4 )c1 +λ5 d
= . (24)
ln(µ ) = −α P̃i j + P̃i j Ai j + ATij P̃i j + Q̃i j − FiTj ψ1 Fi j + rm M̃i j ,
( jl)
Substituting (24) into (21) yields Φi
 ( jl)

xT (t)Rx(t) Ξ11 P̃il Bi j P̃il Di j − FiTj ψ2 P̃il Gi j
 
[λ2 + hm eα hm λ3 + rm eα rm λ4 ]c1 + λ5 d α Tf  ∗ −(1 − ĥ)Q̃il 0 0 
= ,
<[ ]e ,  ∗ ∗ −ψ3 0 
λ1
[ ] ∗ ∗ ∗ − M̃rmil
λ1 c2 e−α Tf
= c2 . ( jl)
(λ2 + hm e m λ3 + rm eα rm λ4 )c1 + λ5 d
α h Ξ11
Therefore, system (5) is finite-time bounded with respect = −α P̃il + P̃il Ai j + ATij P̃il + Q̃il − FiTj ψ1 Fi j + rm M̃il ,
to (c1 , c2 , d, R, Tf , σ ) from Definition 1. The proof is com- Φi
(l j)
pleted. □  
(l j)
Ξ11 P̃i j Bil P̃i j Dil − FilT ψ2 P̃i j Gil
 ∗ −(1 − ĥ)Q̃i j 
3.2. Finite time extended dissipative analysis  0 0 
= ,
Theorem 2: Consider system (5). For given posi-  ∗ ∗ −ψ3 0 
tive scalars α , ĥ, hm , rm , b, if there exist positive definite ∗ ∗ ∗
M̃
− rmi j
symmetric matrices P̃i , Q̃i , M̃i and matrices Y jl with Y j j = (l j)
Y jTj , j = 1, 2..., p, l = j, ..., p with appropriate dimensions, Ξ11
such that = −α P̃i j + P̃i j Ail + ATil P̃i j + Q̃i j − FilT ψ1 Fil + rm M̃i j ,
1
P̃i − FiT ψ4 Fi > 0, (25) and
b
( j j)
Φi < Y j j , j = 1, 2, ..., p, (26) λmin (Pi ) = λ1 , λmax (FiT Fi ) = λ6 , λmax (ψ2T ψ2 ) = λ7 ,
Φi
( jl)
+ Φi
(l j)
< Y jl +Y jlT , j < l, j, l = 1, 2, ..., p, λmax (ψ3 ) = λ8 .
(27)
  Then the system is finite-time bounded and satisfies the
Y11 Y12 · · · Y1p extended dissipative performance.
 ∗ Y22 · · · Y2p 
  Proof: Choose the same Lyapunov-Krasovskii function
 .. ..  ≤ 0, (28)
 ∗ ∗ . .  as in (12)-(13), similar to the proof of Theorem 1, we have
∗ ∗ ∗ Ypp
V̇ (t) − α V (t) − J(t) ≤ X T (t)Φi X(t)
the average dwell time satisfies
Tf lnµ where
τa > τa∗ = , [ ∫t ]T
ln(λ1 c2 ) − ln[λ6 k + (λ7 + λ8 )d] − α Tf X(t) = xT (t) xT (t − h(t)) wT (t) t−r(t) xT (s)ds ,
(29)
where by (26)-(28), we can obtain
1 1 1 1 1 1 p p−1 p
P̃i = R 2 Pi R 2 , Q̃i = R 2 Qi R 2 , M̃i = R 2 Mi R 2 ,
Φi = ∑ θi2j Φi +∑ ∑
( j j) ( jl) (l j)
θ(i j) θ(il) (Φi + Φi )
j=1 j=1 l= j+1
we denote
p p−1 p
p p
< ∑ θi2jY j j + ∑ ∑ θ(i j) θ(il) (Y jl +YlTj )
[Pi , Qi , Mi ] = ∑ θ j [Pi j , Qi j , Mi j ], ∑ θ j = 1, θ j ≥ 0, j=1 j=1 l= j+1
j=1 j=1
 
Y11 Y12 · · · Y1p
and  ∗ Y22 
 · · · Y2p  T
1 1 1
P˜i j = R Pi j R , Q˜i j = R Qi j R , M˜i j = R Mi j R .
1 1 1
=∆  .. ..  ∆ ≤ 0, (30)
2 2 2 2 2 2
 ∗ ∗ . . 
And ∗ ∗ ∗ Ypp
( j j)
Φi where
 ( j j)
  
Ξ11 P̃i j Bi j P̃i j Di j − FiTj ψ2 P̃i j Gi j Φ11 P̃i Bi P̃i Di − FiT ψ2 P̃i Gi
   ∗
 ∗ −(1 − ĥ)Q̃i j 0 0  −(1 − ĥ)Q̃i 0 0 
= , Φi = 
 ∗
,
 ∗ ∗ −ψ3 0  ∗ −ψ3 0 
M̃
∗ ∗ ∗ − rmi j ∗ ∗ ∗ − M̃rm
i
1464 Hui Gao, Jianwei Xia, and Guangming Zhuang

Φ11 = −α P̃i + P̃i Ai + ATi P̃i + Q̃i − FiT ψ1 Fi + rm M̃i , and

∫ Tf
lnµ
V (t) < e(α + τa )Tf J(s)ds,
and 0

for ψ1 ≤ 0, we can obtain

∆ = [θi1 I, θi2 I, · · · , θip I].
∫ Tf ∫ Tf
Therefore, from (30), we have that J(s)ds ≤ [2zT (s)ψ2 w(s)
0 0
V̇ (t) − α V (t) − J(t) < 0, + wT (s)ψ3 w(s)]ds,

following the proof line of (16), it is easy to obtain the and

following inequality ∫
lnµ Tf
∫ t V (t) <e(α + τa )Tf [ [2zT (s)ψ2 w(s)
V (t) < eα t µ Nσ (0,t)V (0) + eα (t−s) µ Nσ (s,t) J(s)ds, 0
0 + wT (s)ψ3 w(s)]ds],
under zero initial condition V(0)=0, it can be calculated
that so we get
∫ t V (t)
V (t) < eα t µ Nσ (0,t) J(s)ds, xT (t)Rx(t) ≤
0 λ1
lnµ ∫
by Assumption 4 we have e(α + τa )Tf Tf T
< [ [2x (s)FiT ψ2 w(s)
∫ t
λ1 0
V (t) 1 T
J(s)ds > > x (t)P̃i x(t) > 0, + wT (s)ψ3 w(s)]ds],
0 b b
considering inequality by Lemma 1 we have
∫ Tf 2xT (s)FiT ψ2 w(s) ≤ xT (s)FiT Fi x(s) + wT (s)ψ2T ψ2 w(s).
J(t)dt − sup zT (t)ψ4 z(t) ≥ 0,
0 0≤t≤T f
By (37) we can obtain that
∫ Tf
when ψ4 = 0, it is apparent that 0 J(t)dt ≥ 0, when ψ4 >
V (t)
0, it can be obtained by Assumption 3 that ψ1 = 0, ψ2 = xT (t)Rx(t) ≤
λ1
0, ψ3 > 0, so lnµ ∫
∫ ∫ e(α + τa )Tf Tf T
t t < [ [x (s)FiT Fi x(s)
J(s)ds = wT (s)ψ3 w(s)ds, λ1 0
0 0
+ wT (s)ψ2T ψ2 w(s) + wT (s)ψ3 w(s)]ds]
thus for ∀t ∈ [0, Tf ], lnµ
e(α + τa )Tf
∫ ∫ < [λ6 k + (λ7 + λ8 )d].
Tf t 1 λ1
J(s)ds > J(s)ds ≥ xT (t)P̃i x(t) > 0,
0 0 b
Considering (36), we have
by (32) we have
∫ xT (t)Rx(t) ≤ c2 .
Tf 1
J(s)ds ≥ xT (t)P̃i x(t) ≥ xT (t)FiT ψ4 Fi x(t)
0 b Thus the proof is completed.
= zT (t)ψ4 z(t),
3.3. Robust finite-time extended dissipative control
so we get Consider system (1), under the controller µ (t) =
∫ Tf Kσ (t) x(t), the corresponding closed-loop system is given
J(t)dt − sup zT (t)ψ4 z(t) ≥ 0. by
0 0≤t≤T f

ẋ(t) = (Aσ (t) + Eσ (t) Kσ (t) )x(t) + Bσ (t) x(t − h(t))

Thus the proof of extended dissipative is completed. □ ∫ t
Next, we proof finite-time boundedness. Following the + Dσ (t) w(t) + Gσ (t) x(s)ds,
above proof, we have t−r(t)

∫ t z(t) = Fσ (t) x(t),

V (t) < eα t µ Nσ (0,t) J(s)ds, x(t0 + θ ) = φ (θ ), ∀θ ∈ [−τ , 0]. (31)
0
 Robust Finite-time Extended Dissipative Control for a Class of Uncertain Switched Delay Systems 1465

( jl)
Theorem 3: Consider system (31). For given posi- ϕ11 = −α Ril + Ai j Ril + Ril ATij + Ei jYi +YiT EiTj + Q̂il
tive scalars α , ĥ, hm , rm , b, if there exist positive definite + rm M̂il ,
symmetric matrices P̃i , Q̃i , M̃i and matrices Y jl with Y j j =
( jl)
Y jTj , j = 1, 2..., p, l = j, ..., p with appropriate dimensions, ϕ13 = Di j − Ril FiTj ψ2 ,
such that ( jl)
ϕ16 = Ei jYi + SiT − Ril ,
1
P̃i − FiT ψ4 Fi > 0, (32) ( jl)
ϕ22 = −(1 − ĥ)Q̂il ,
b
(l j)
( j j)
Πi < Y j j , j = 1, 2, ..., p, (33) Πi
 (l j) (l j) (l j)

Πi
( jl)
+ Πi
(l j)
< Y jl +Y jlT , j < l, j, l = 1, 2, ..., p, ϕ11 Bil Ri j ϕ13 Gil Ri j FilT ϕ16
 (l j) 
(34)  ∗ ϕ22 0 0 0 0 
   
 ∗ ∗ −ψ3 0 0 0 
Y11 Y12 · · · Y1p = ,
 ∗ Y22   ∗ ∗
M̃
∗ − rmi j 0 0 
 · · · Y2p   
 .. ..  ≤ 0, (35)  ∗ ∗ ∗ ∗ ψ1−1 0 
 ∗ ∗ . . 
∗ ∗ ∗ ∗ ∗ Si + SiT
∗ ∗ ∗ Ypp (l j)
ϕ11 = −α Ri j + Ail Ri j + Ri j ATil + Eil Yi +YiT EilT + Q̂i j
the average dwell time satisfies
+ rm M̂i j ,
Tf ln µ
τa > τa∗ = , (l j)
ϕ13 = Dil − Ri j FilT ψ2 ,
ln(λ1 c2 ) − ln[λ6 k + (λ7 + λ8 )d] − α Tf
(l j)
(36) ϕ16 = Eil Yi + SiT − Ri j ,
(l j)
we denote ϕ22 = −(1 − ĥ)Q̂i j ,
p p
[Pi , Qi , Mi ] = ∑ θ j [Pi j , Qi j , Mi j ], ∑ θ j = 1, θ j ≥ 0, and the matrices and parameters are defined as follows:
j=1 j=1

and 1 1 1 1
P̃i = R 2 Pi R 2 , Q̃i = R 2 Qi R 2 , M̃i = R 2 Mi R 2 ,
1 1

p p 1 1 1 1 1 1
[Ri , Q̂i , M̂i ] = ∑ θ j [Ri j , Q̂i j , M̂i j ], ∑ θ j = 1, θ j ≥ 0, P˜i j = R 2 Pi j R 2 , Q˜i j = R 2 Qi j R 2 , M˜i j = R 2 Mi j R 2 ,
j=1 j=1 −1 −1 −1 −1 −1
P̃i = Ri , P̃i Q̃i P̃i = Q̂i , P̃i M̃i P̃i = M̂i ,
where λmin (Pi ) = λ1 , λmax (FiT Fi ) = λ6 , λmax (ψ2T ψ2 ) = λ7 ,
( j j)
Πi λmax (ψ3 ) = λ8 .
 ( j j) ( j j) ( j j) 
ϕ11 Bi j Ri j ϕ13 Gi j Ri j FiTj ϕ16
 ( j j)  Then the switched linear system is finite-time bounded
 ∗ ϕ22 0 0 0 0 
  and satisfies the extended dissipative performance under
 ∗ ∗ −ψ3 0 0 0 
= , the controller µ (t) = Kσ (t) x(t). The controller gains can
 ∗ ∗
M̃
∗ − rmi j 0 0 
  be given by Ki = Yi Si−1 .
 ∗ ∗ ∗ ∗ ψ1−1 0 
∗ ∗ ∗ ∗ ∗ Si + SiT Proof: By (33)-(35), we get the following inequality
( j j)
ϕ11 = −α Ri j + Ai j Ri j + Ri j ATij + Ei jYi +YiT EiTj + Q̂i j p p−1 p
∑ θi2j Πi +∑ ∑
( j j) ( jl) (l j)
+ rm M̂i j , Πi = θ(i j) θ(il) (Πi + Πi )
j=1 j=1 l= j+1
( j j)
ϕ13 = Di j − Ri j FiTj ψ2 , p p−1 p
( j j)
ϕ16 = Ei jYi + SiT − Ri j , < ∑ θi2jY j j + ∑ ∑ θ(i j) θ(il) (Y jl +YlTj )
j=1 j=1 l= j+1
( j j)
ϕ22 = −(1 − ĥ)Q̂i j ,  
Y11 Y12 · · · Y1p
Πi
( jl)  ∗ Y22 · · · Y2p 
  T
  = ∆ .. ..  ∆ ≤ 0, (37)
ϕ11
( jl)
Bi j Ril ϕ13
( jl)
Gi j Ril FiTj ϕ16
( jl)  ∗ ∗ . . 
 ( jl)  ∗ ∗ ∗ Ypp
 ∗ ϕ22 0 0 0 0 
 
 ∗ ∗ −ψ3 0 0 0 
= ,
 ∗ ∗ ∗ − M̃rmil 0 0  where
 
 ∗ ∗ ∗ ∗ −1
ψ1 0 
∗ ∗ ∗ ∗ ∗ Si + SiT Πi
1466 Hui Gao, Jianwei Xia, and Guangming Zhuang

 
ϕ11 Bi Ri ϕ13 Gi Ri FiT ϕ16 where (39) is equivalent to Φi . The following proof is
 ∗ −(1 − ĥ)Q̂i 0 0 0 0  similar to that of Theorem 2, it is omitted here.
 
 ∗ ∗ − ψ 0 0 0 
= ,
3 Remark 1: The above results could be easily extended
 ∗ ∗ ∗ − rm
M̃i
0 0 
  to the other systems and models, for example the neu-
 ∗ ∗ ∗ ∗ ψ1−1 0 
ral networks [22, 23], Markovian jump delayed systems,
∗ ∗ ∗ ∗ ∗ Si + SiT sampled-data systems [20], Lur’e systems [21], and so on,
ϕ11 = −α Ri + Ai Ri + Ri ATi + EiYi +YiT EiT + Q̂i + rm M̂i , which deserve further study.
ϕ13 = Di − Ri FiT ψ2 ,
4. NUMERICAL EXAMPLE
ϕ16 = EiYi + SiT − Ri ,
In this section, we present an example to illustrate the
and effectiveness of the controller design method.
∆ = [θi1 I, θi2 I, · · · , θip I], Consider system (31) with two subsystems, each sub-
system has two vertices to represent polytopic uncertain
pre- and post-multiply (37) by system:
  Subsystem 1:
I 0 0 0 0 −Ei Ki
 0 I 0 0 0  [ ] [ ]
 0  2 0 −2 0
 0 0 I 0 0 0  A11 = , B11 =
−1 −2
,
  3 3
 0 0 0 I 0 0  [ ] [ ]
0.3 0 1 −1
0 0 0 0 I 0 C11 = , D11 = ,
0 −0.2 2 1
[ ] [ ] [ ]
and its transpose, respectively, we have 3 −3 1 0.5 1 1
E11 = , G11 = , F11 = ,
  0 4 0.5 0.6 0 2
ϕ11 Bi Ri Di − Ri FiT ψ2 Gi Ri FiT [ ] [ ]
 ∗ −(1 − ĥ)Q̂i 0 0 0  2 −1 −1 0
  A12 = , B12 = ,
 ∗ ∗ − ψ 0 0  < 0, −2 3 −2 1
 3  [ ] [ ]
 ∗ ∗ ∗ − M̃i
0  0.2 0 −1 0
rm
C12 = , D12 = ,
∗ ∗ ∗ ∗ ψ1−1 0 −0.3 2 0.8
(38) [ ] [ ] [ ]
4 −1 0.7 0 2 1
E12 = , G12 = , F12 = .
1 6 1 0.5 0 3
ϕ11 = −α Ri +Ai Ri +Ri ATi +Ei Ki Ri +Ri KiT EiT + Q̂i +rm M̂i
pre- and post-multiply (38) by diag{P̃i , P̃i , I, I, I} and its Subsystem 2:
transpose, respectively, by Schur complement, we have [ ] [ ] [ ]
  2 0 −2 −1 1 0
ϕ11 P̃i Di − FiT ψ2 P̃i Gi A21 = , B21 = , C21 = ,
P̃i Bi 0 3 0 −2 0 −1
 ∗ −(1 − ĥ)Q̃i 0 0  [ ] [ ]
  < 0, (39) −1.5 −1 −0.7 0
 ∗ ∗ −ψ3 0  D21 = , E21 = ,
0 −0.9 0.2 1.4
∗ ∗ ∗ − M̃rm
i
[ ] [ ]
0.3 0 0.1 0
G21 = , F21 = ,
where ϕ11 = −α P̃i + P̃i (Ai + Ei Ki ) + (Ai + Ei Ki )T P̃i + Q̃i + −0.2 −1.1 0.4 1
[ ] [ ] [ ]
rm M̃i − FiT ψ1 Fi . Similar to the proof of Theorem 2, we 3 0 1 3 −3 2
A22 = , B22 = , C22 = ,
can obtain that 0 5 0 −2 0 3
[ ] [ ]
V̇ (t) − α V (t) − J(t) ≤ X T (t)Φi X(t), 0.5 0 0.5 −0.4
D22 = , E22 = ,
0 −0.7 0 2
where [ ] [ ]
0.7 0 0.3 0
[ ∫t ]T G22 = , F22 = .
X(t) = xT (t) xT (t − h(t)) wT (t) t−r(t) xT (s)ds , 0.4 0.3 0.2 3

Furthermore, we assume the uncertain parameters to be

Table 1. Matrices for each case. θi j = 0.5, i = 1, 2, j = 1, 2. And ĥ = 0.01, hm = 0.01,
rm = 0.01, α = 0.01. For extended dissipative control,
Analysis Ψ1 Ψ2 Ψ3 Ψ4 choose matrices for each case in Table 1:
L2 − L∞ performance 0 0 γ 2I I By solving the LMIs from (33)-(35) presented in Theo-
H∞ performance −I 0 γ 2I 0 rem 3, we can obtain the optimized variables of four per-
Passivity 0 I γ 0 formances in Table 2, and controller gain for each case in
Dissipativity −I I I −β ∗I 0 Table 3, respectively.
 Robust Finite-time Extended Dissipative Control for a Class of Uncertain Switched Delay Systems 1467

Table 2. Optimized variable for each case. time switching,” Automatica, vol. 48, pp. 1132-1137, 2012.
[click]
L2 − L∞ performance H∞ performance [4] L. Lu, Z. L. Lin, and H. J. Fang, “L2 gain analysis for a
γmin
2
= 1 ∗ 10−7 γmin
2
= 1 ∗ 10−7 class of switched systems,” Automatica, vol. 45, pp. 965-
Passivity Dissipativity 972, 2009. [click]
γmin = 1 ∗ 10−7 βmax = 0.9999999 [5] X. D. Zhao, P. Shi, and L. X. Zhang, “Asynchronously
switched control of a class of slowly switched linear sys-
tems,” Systems Control Letters, vol. 61, pp. 1151-1156,
Table 3. Controller gain for each case. 2012. [click]
Subsystem 1 [6] M. J. Hu, Y. W Wang, and J. W. Xiao, “Positive observer
[ ]
design for linear impulsive positive systems with inter-
0.2985 0.0520
L2 − L∞ K1 = val uncertainties and time delay,” International Journal of
0.0445 0.2233 Control, Automation and Systems, vol. 15, no. 3, pp. 1032-
[ ]
0.3146 0.0503 1039, 2017. [click]
H∞ performance K1 =
0.0458 0.2340 [7] Y. Zhang, H. Zhu, X. Liu, and S. Zhong. “Reliable H∞ con-
[ ] trol for a class of switched neutral systems,” Complex Syst
0.4951 0.0457 Appl: Model. Control Simulat, vol. 14, no. S2, pp. 4-9,
Passivity K1 =
0.0356 0.3400 2007.
[ ]
0.4956 0.0449 [8] D. Liu, S. Zhong, X. Liu, and Y. Huang. “Stability analysis
Dissipativity K1 = for uncertain switched neutral systems with discrete time-
0.0353 0.3419
varying delay: a delay-dependent method,” Math Comput
Subsystem 2 Simulat., vol. 80, no. 8, pp. 28-39, 2009.
[ ]
−0.6627 −0.0232 [9] L. Xiong, S. Zhong, M. Ye, and S. Wu. “New stability and
L2 − L∞ K2 =
−0.0256 0.6866 stabilization for switched neutral control systems,” Chaos
[ ] Solitons Fract, vol. 42. no. 3, pp. 1800-11, 2009.
−0.6869 −0.0207
H∞ performance K2 = [10] R. Wang, P. Shi, Z. G. Wu, and Y. T. Sun, “Stabilization of
−0.0440 0.7234 switched delay systems with polytopic uncertainties under
[ ]
−0.5140 −0.0486 asynchronous switching,” Journal of the Franklin Institute,
Passivity K2 = vol. 350, pp. 2028-2043, 2013.
−0.0711 0.8044
[ ] [11] X. Lin, H. Du, and S. Li. “Finite-time boundedness and
−0.5388 −0.0392 L2 -gain analysis for switched delay systems with norm-
Dissipativity K2 =
−0.0669 0.8064 bounded disturbance,” Appl Math Comput., vol. 217(12),
no. 59, pp. 82-93, 2011.
[12] Z. Xiang, Y. Sun, and M.S. Mahmoud, “Robust finite-time
5. CONCLUSION H∞ control for a class of uncertain switched neutral sys-
tems,” Commun. Nonlinear Sci. Numer. Simul., vol. 17,
In this paper, we have investigated the problem of finite no. 4, pp. 1766-1778, 2012. [click]
time extended dissipative analysis and control of switched
[13] S. Wang, T. G Shi, L. X. Zhang, A. Jasra, and M. Zeng,
systems with time delay. Based on extended dissipative “Extended finite-time H∞ control for uncertain switched
performance, we can solve the H∞ , L2 − L∞ , Passivity and linear neutral systems with time-varying delays,” Neuro-
(Q, S, R)-dissipativity performance in a unified frame- computing, vol. 152, pp. 377-387, 2015. [click]
work. All the results are given in terms of linear matrix [14] S. Wang, M. Basin, L. X. Zhang, M. Zeng, T. Hayat,
inequalities (LMIs), numerical examples are provided to and A. Alsaedi, “Reliable finite-time filtering for impul-
show the effectiveness of the proposed method. sive switched linear systems with sensor failures,” Signal
Processing, vol. 125, pp. 134-144, 2016. [click]

REFERENCES [15] B. Y. Zhang, W. X. Zheng, and S. Y. Xu, “Filtering of

[1] H. Lin and P. J. Antsaklis, “Stability and stabilizability of
switched linear systems: a survey of recent results,” IEEE
Trans. Autom. Control, vol. 54, no. 2, pp. 308-322, 2009.
[click]
[2] D. Liberzon and A. S. Morse, “Basic problems in stability
and design of switched systems,” IEEE Control Systems
Magazine, vol. 19, pp. 59-70, 1999. [click]
[3] X. D. Zhao, L. X. Zhang, P. Shi, and M. Liu, “Stabil-
ity of switched positive linear systems with average dwell
Markovian jump delay systems based on a new perfor-
mance index,” IEEE Trans. Circuits Syst. I Reg. Pap., vol.
60, pp. 1250-1263, 2013.
[16] H. Shen, Y. Z. Zhu, L. X. Zhang, and J. H. Park, “Ex-
tended dissipative state estimation for Markov jump neural
networks with unreliable links,” IEEE Trans. Neural Netw.
Learning Syst., vol. 28, pp. 346-358, 2017.
[17] 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, vol. 64, pp.
113-128, 2016.
1468 Hui Gao, Jianwei Xia, and Guangming Zhuang
[18] J. W. Xia, G. L. Chen, and W. Sun, “Extended dissipative
analysis of generalized Markovian switching neural net-
works with two delay components,” Neurocomputing, vol.
260, pp. 275-283, 2017.
[19] H. L. Yang, L. Shu, S. M. Zhong, and X. Wang, “Ex-
tended dissipative exponential synchronization of complex
dynamical systems with coupling delay and sampled-data
control,” Journal of the Franklin Institute, vol. 353, pp.
1829-1847, 2016. [click]
[20] H. B. Zeng, K. L Teo, and Y. He, “A new looped-functional
for stability analysis of sampled-data systems,” Automat-
ica, vol. 82, pp. 328-331, 2017. [click]
[21] S. P. Xiao, X. Z. Liu, C. F. Zhang, and H. B. Zeng, “Fur-
ther results on absolute stability of Lur’e systems with a
time-varying delay,” Neurocomputing, vol. 207, pp. 823-
827, 2016. [click]
[22] H. B. Zeng, Y. He, P. Shi, M. Wu, and S. P. Xiao, “Dis-
sipativity analysis of neural networks with time-varying
delays,” Neurocomputing, vol. 168, pp. 741-746, 2015.
[click]
[23] S. P. Xiao, H. H. Lian, H. B. Zeng, G. Chen, and W. H.
Zheng, “Analysis on robust passivity of uncertain neural
networks with time-varying delays via free-matrix-based
integral inequality,” International Journal of Control, Au-
tomation and Systems, vol. 15, pp. 1-10, 2017.