Sie sind auf Seite 1von 19

Accepted Manuscript

Finite-time H estimation for discrete-time Markov jump systems with time-


varying transition probabilities subject to average dwell time switching

Jun Cheng, Hong Zhu, Shouming Zhong, Qishui Zhong, Yong Zeng

PII: S1007-5704(14)00266-4
DOI: http://dx.doi.org/10.1016/j.cnsns.2014.06.006
Reference: CNSNS 3218

To appear in: Communications in Nonlinear Science and Numer-


ical Simulation

Received Date: 21 June 2013


Revised Date: 19 February 2014
Accepted Date: 1 June 2014

Please cite this article as: Cheng, J., Zhu, H., Zhong, S., Zhong, Q., Zeng, Y., Finite-time H estimation for discrete-
time Markov jump systems with time-varying transition probabilities subject to average dwell time switching,
Communications in Nonlinear Science and Numerical Simulation (2014), doi: http://dx.doi.org/10.1016/j.cnsns.
2014.06.006

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.
Finite-time H estimation for discrete-time Markov jump systems with
time-varying transition probabilities subject to average dwell time switching

Jun Chenga,, Hong Zhua , Shouming Zhongb , Qishui Zhongc , Yong Zenga
a School of Automation Engineering, University of Electronic Science and Technology of China, Chengdu, Sichuan 611731,
P.R.China
b School of Mathematical Sciences, University of Electronic Science and Technology of China, Chengdu, Sichuan 611731,

P.R.China
c School of Aeronautics and Astronautics, University of Electronic Science and Technology of China, Chengdu, Sichuan

611731, P.R.China

Abstract

The time-varying character of transition probabilities is considered as finite piecewise homogeneous. This
paper studies the problem of finite-time H estimation for a class of discrete-time Markov jump systems with
time-varying transition probabilities subject to average dwell time switching, sufficient conditions ensuring
the Markov jump systems to be finite-time bounded and H filtering finite-time boundness are established.
Based on the results of finite-time boundness and average dwell time, the system trajectory stays within
a prescribed bound. Finally, an example is provided to illustrate the usefulness and effectiveness of the
proposed method.
Keywords: H finite-time stability; Time-varying transition probabilities; Markov jump systems; H
control; Average dwell time.

1. Introduction

As is well known, Markovian jump systems (MJSs) are regarded as a powerful framework to model the
dynamic systems subject to random abrupt variations. Up to now, MJSs have received increasing attention
owing to their strong applications in both theoretical and practical fields, such as networked control systems,
manufacturing systems [1-5]. It should be noted that, in finite operation modes, MJSs represent a special
class of stochastic hybrid systems can switch from one to another at different time, for instance, component
failures, sudden environmental disturbance and abrupt variations of the operating points. As a crucial
factor, the transition probabilities (TPs) in the Markov chain determine system behavior and corresponding

This work was partly supported by the National Basic Research Program of China (2010CB732501), the China Postdoctoral

Science Foundation (2012M521683, 2013T60848), the National Basic Research Program of China (2010CB732501), the Governor
of Guizhou Province Talent Funds (2012), the Program for New Century Excellent Talents in University (NCET-10-0097), CSC
(No.201306070065) and YBXSZC20131006.
Corresponding author.

Email address: jcheng6819@126.com (Jun Cheng)

Preprint submitted to CNSNS June 6, 2014


performance. Over the past few decades, many important issues have been studied in the literature [6-9]. In
reality, it is worth mentioning that most of the aforementioned results based on the implicit assumption that
the complete knowledge of TPs are available. However, in the most cases, the TPs of MJSs and networks are
not exactly known, it is difficult to obtain precisely all the TPs, thus, it is of great significance to investigate
partly unknown TPs. Very recently, many attracting results with partially unknown TPs have been reported
[10-18]. It should be mentioned that, most of the available analysis results are assumed that the Markov
chain in the underlying systems are time-invariant.
However, in practice, the assumption is not always true. The time-varying Tps challenges the traditional
approaches of MJSs. Through the character of time-varying, the TP matrix of Markov chain changes
depending on time. Recently, in [20], a class of time-varying Markov chains subject to switching signal have
been introduced. Inspired by [20], the system is regarded as a switched system. Moreover, in the area of
switched systems, the results are always obtained with either dwell time (DT) method or average dwell time
(ADT) technique [19-22].
In practical process, the behavior of systems over a fixed finite time interval may be received a great
deal of attention, for example, in the presence of saturations, large values of the states are not acceptable.
For some Lyapunov asymptotically stable system such as robot control system, it seems that the transient
characteristics are more important. Therefore, we are more interested in the unacceptable values whether
the system states exceed the prescribed bound in a fixed finite time interval. To discuss this transient
performance of control dynamics, in 1961, Dorato proposed the concept of finite-time stability in [23] and
the problem of finite-time stability has been accordingly studied in the literature [23-33]. It should be
noted that there are still some related issues to be solved, to the best of our knowledge, the finite-time
H estimation for a class of discrete-time Markov jump systems with transition probabilities subject to
average dwell time switching have not been studied. The problem is interesting but also challenging, which
motivates the current study.
In this paper, we aim to solve the finite-time H estimation problem for a class of discrete-time Markov
jump systems with time-varying transition probabilities subject to average dwell time switching. The finite-
time stability is an independent concept from Lyapunov stability and can be affected by switching behavior
significantly, thus it deserves our investigation. By selecting the appropriate Lyapunov-Krasovskii functional,
sufficient conditions ensuring the Markov jump systems to be finite-time bounded and H filtering finite-
time boundness are established. The finite-time boundness (FTB) criteria can be tackled in the form of
linear matrix inequalities (LMIs). Finally, an example is introduced to illustrate the effectiveness of the
developed techniques.
Notations: Throughout this paper, we let P > 0(P 0, P < 0, P 0) denote a symmetric positive
definite matrix P (positive-semi definite, negative definite and negative-semi definite). For any symmetric
matrix P , max (P ) and min (P ) denote the maximum and minimum eigenvalues of matrix P , respectively.
2
Rn denotes the n-dimensional Euclidean space and Rnm refers to the set of all n m real matrices. The
identity matrix of order n is denoted as In . represents the elements below the main diagonal of a symmetric
matrix. The superscripts  and 1 stand for matrix transposition and matrix inverse, respectively.

2. Preliminaries

Given a probability space (M, F, P ) where M , F and P represent the sample space, the algebra of events
and the probability measure defined on F , respectively. In this paper, we consider the following Markovian
jump systems over the space (M, F, P ) described by


x(k + 1) = Ark x(k) + Brk (k),



y(k) = Crk x(k) + Drk (k), (1)




z(k) = Lrk x(k) + Ldrk (k),

where x(k) Rn is the discrete state vector of the system, z(k) Rm is the controlled output, (k) Rq
is the disturbance input which satisfies
N

 (k)(k) < d. (2)
k=0

Let the random form process {rk , k 0} be the Markov stochastic process taking values on a finite set
( )
N = {1, 2, . . . , N } with transition rate matrix = {ij k }, i, j N , namely, for rk = i, rk+1 = j, N
governs the switching among the different system modes with the following transition probabilities:

( )
Pr(rk+1 = j | rk = i) = ij k ,

( ) N ( ) ( )
where ij k 0, (i, j N ), and j=1 ij k = 1 for all i N . Here, ij k is now a function of k . By
k , we mean that the transition probabilities are time-varying. Meanwhile, k is assumed to be a piecewise
constant function of time k. Set N contains N modes of system (1) and for rk = i N , the system matrices
of the ith mode are denoted by Ai , Bi , Ci , Di , Li and Ldi , which are considered to be real known with
appropriate dimensions.
Furthermore, the transition probabilities matrix is defined by

( ) ( ) ( )
11k 12k 1Nk

(k ) (k ) (k )
21 22 2N
(k ) =
.. .. ..

.. .
. . . .

( ) ( ) ( )
N 1k N 2k k
N N

In addition, k is a high-level ADT switching signal to determine the time-varying property. Moreover,
k is a given initial condition sequence. For simplicity, m denoting k is a piecewise constant function of
3
time, which takes its values in a finite set M {1, 2, , M }. At an arbitrary time k, may be dependent
on k or x(k), or both, or other logic rules. For a switching sequence k0 < k1 < k2 < , is continuous
from right everywhere and maybe either autonomous or controlled. When k [kl , kl+1 ), we say that the
kl th transition probabilities matrix is active and therefore the trajectory xk of system (1) with the kl th
transition probabilities matrix.
In this paper, we estimating the signal z(k) in system (1) solved by a filter is of the form

xf (k + 1) = Af rk ,k xf (k) + Bf rk ,k y(k),
(3)

zf (k) = Cf rk ,k xf (k) + Df rk ,k y(k),

where xf (k) is the filter state vector, y(k) is the input of the filter. For all rk = i N , k = m M,
Af i,m , Bf i,m , Cf i,m and Df i,m are filter gain to be determined. It is seen that the considered filter in (3)
is variation-dependent and mode-dependent.
Let k = [x (k), xf (k)], ek = z(k) zf (k) and k = (k). Then, the filtering error dynamics can be
described as

k+1 = Ark ,k k + B rk ,k k ,
(4)

ek = C rk ,k k + Drk ,k k ,

where for all rk = i N , k = m M,



Ai 0 Bi
Ai,m = , B i,m = ,
Bf i,m Ci Af i,m Bf i,m Di

C i,m = [Li Df i,m Ci , Cf i,m ], Di,m = Ldi Df i,m Di .

In order to more precisely describe the main objective, we introduce the following definitions and Lemmas
for the underlying system.

Definition 2.1. [30] The discrete-time linear system

x(k + 1) = Ax(k), k N.

is said to be finite-time stability (FTS) with respect to (c1 , c2 , R, N ), where R > 0 is a positive definite
matrix, 0 < c1 < c2 and N N , if x (0)Rx(0) c1 x (k)Rx(k) < c2 , k {1, 2, . . . , N }.

Definition 2.2. [30] The discrete-time linear system

x(k + 1) = Ax(k) + G(k), k N.

subject to an exogenous disturbance (k) satisfy (2), is said to be finite-time boundedness (FTB) with respect
to (c1 , c2 , R, d, N ), where R > 0 is a positive definite matrix, 0 < c1 < c2 and N N , if x (0)Rx(0) c1
x (k)Rx(k) < c2 , k {1, 2, . . . , N }.
4
Definition 2.3. [30] For > 0, 0 c1 < c2 , R is a positive definite matrix and N Z+ , system (4) is
said to be H finite-time boundness with respect to (c1 , c2 , d, , R, M ), and following condition should be
satisfied:
N
 N

ek ek < 2 k k (5)
k=0 k=0

N
under zero initial condition, it holds for all nonzero : k=0 k k < d.

Definition 2.4. [31] For any k k0 and any switching signal (l), k0 l < k, let N (k0 , k) denote the
number of switchings of (k), if

k k0
N (k0 , k) N0 + (6)
a

holds for N0 0 and a > 0, then a is called the average dwell time and N0 is the chatter bound.

Remark 1 It has been analyzed in [16] that N0 > 1 gives switching signals with ADT and N0 = 1
corresponds exactly to those switching signals with DT. Moreover, as an extreme case, a 0 implies that
the constraint on the switching times is almost eliminated and the resulting switching can be arbitrary [17].
Therefore, as a typical set of switching signals with regularities [18], the ADT switching covers the DT
switching and arbitrary switching and is relatively general.

Lemma 2.1. [14] (Schur Complement) Given constant matrices X, Y , Z, where X = X  and 0 < Y = Y  ,
then X + Z  Y 1 Z < 0 if and only if

X Z Y Z
< 0 or < 0. (7)
Y X

3. Finite-time H performance analysis

First, we will focus on finite-time bounded analysis of system (4).

Theorem 3.1. System (4) is finite-time boundness with respect to (c1 , c2 , d, R, M ), if there exist matrices
Pi,m > 0, (l = 1, 2, 3), S > 0, i, j N , m, n M, and scalars 1, > 1, s > 0, (s = 1, 2, 3), such that

P i,m P i,m Ai,m P i,m B i,m

Pi,m 0 < 0, i N , m, n M, (8)

S

1 1
Pi,m 1 Pi,m R 2 R I
< 0, < 0, i N , m, n M, (9)
1
1 R Pi,m
Pi,m Pi,n , i, j N , m, n M, (10)
5
1 c2
> 1. (11)
M ( 2 c1 + d3 )

and the average dwell time of the switched discrete-time signal (k) satisfies
M ln
a > a = . (12)
ln 1 ln 2
where
N

P i,m = m
ij Pj,m ,
j=1

1 1
Pi,m = R 2 Pi,m R 2 ,

3 = max (S),

1 = ( 1)c2 1 M + d3 ,

2 = c1 2 ( 1) + d3 .

Proof. We consider the following stochastic Lyapunov-Krasovskii functional:

V (k , rk , k ) = k Prk ,k k , (13)

We have

V (k+1 , rk+1 , k ) V (k , rk , k )  (k)S(k)



=k+1 Prk+1 ,k k+1 V (k , rk , k )  (k)S(k)


=k+1 m 
ij Pj,m k+1 k Pi,m k (k)S(k)


jN
 (14)
k (Ai,m P i,m Ai,m Pi,m )k + 2k Ai,m P i,m B i,m k

+ k (B i,m P i,m B i,m S)k



 
A P A
i,m i,m i,m Pi,m A P
i,m i,m i,mB
=k 
k .
B i,m P i,m B i,m S

where

k = [k , k ] .

On the other hand, by Schur complement and (8), it yields that

V (k+1 , rk+1 , k ) < V (k , rk , k ) + k Sk . (15)

From which we can get that


k1

V (k , rk , k ) < kkl V (kl , rkl , kl ) + ks s Ss . (16)
s=kl

6
Note that

V (kl , rkl , kl ) V (kl , rkl , kl1 ), (17)

and
l 1
k
V (kl , rkl , kl1 ) kl kl1 V (kl1 , rkl1 , kl1 ) + kl s s Ss . (18)
s=kl1

It yields that
l 1
k
V (kl , rkl , kl ) kl kl1 V (kl1 , rkl1 , kl1 ) + kl s s Ss . (19)
s=kl

From (15)-(19) we can get that


k1

V (k , rk , k ) kkl1 V (kl1 , rkl1 , kl1 ) + ks s Ss
s=kl

kkl1
V (kl1 , rkl1 , kl1 ) + k dmax (S)

2 kkl2 V (kl2 , rkl2 , kl2 ) + k dmax (S) + k dmax (S)


(kk0 )/a
kk0  kk0 (20)
kk0 a V (k0 , rk0 , k0 ) + dmax (S)k a l s
s=1
k/a
kk0 1
kk0 a  (k0 )Prk0 ,k0 (k0 ) + k dmax (S)
1
k/a
kk0 1
kk0 a sup {max (Pi,m )}x (k0 )Rx(k0 ) + k dmax (S)
iN ,mM 1
 
M k/a 1
M a c1 sup {max (Pi,m )} + d3 .
iN ,mM 1
Moreover,

V (k , rk , k ) k Prk ,k k inf {min (Pi,m )}k Rk , (21)


iN ,mM

Then we can obtain


 
M M
 k/a 1
k Rk < a c1 sup {max (Pi,m )} + d3 . (22)
inf iN ,mM {min (Pi,m )} iN ,mM 1

From (9), we have

1 R < Pi,m < 2 R 1 I < Pi,m < 2 I. (23)

Since supiN {max (Pi,m )} 2 , inf iN {min (Pi,m )} 1 and by condition (11), we can obtain
 
M M k/a 1
k Rk < a c1 2 + d3 = c2 . (24)
1 1
7
Then the system is finite-time bounded with respect to (c1 , c2 , d, R, M ).
Remark 2 Just like [28], the concepts of Lyapunov stability and finite-time stability are different. The
Lyapunov stability is largely known to the control characteristic in infinite time interval, while the latter
focuses the bounded analysis of the controlled states within a finite time interval.
Remark 3 It is noted that when N = {1}, the time-varying transition probabilities Markovian jump system
(1) reduces to a homogeneous Markovian jump system, which has been fully investigated. Thus, homoge-
neous Markovian jump system can be regarded as a special case of time-varying transition probabilities
Markovian jump system. Therefore, the time-varying transition probabilities Markovian jump systems are
more general and complex than homogeneous Markovian jump systems.
Remark 4 The definition of finite-time stability in this paper is different from [33]. In [33], continuous
autonomous systems state trajectory converges to the equilibrium in a finite-time interval. However, in this
paper, we concerned the state of system does not exceed a certain threshold during a specified time interval.
In the following section, we will investigate the weighted H performance for system (4).

Theorem 3.2. System (4) is finite-time boundness with respect to (c1 , c2 , d, R, M ) and a weighted H
performance , if there exist matrices Pi,m > 0, i, j N , m, n M, scalars 1, > 0, > 1, 1 > 0,
2 > 0, such that the following holds

P i,m 0 P i,m Ai,m P i,m B i,m


I C i,m Di,m
< 0, i, j N , m, n M, (25)

Pi,m 0

2 I

Pi,m Pi,n , i, j N , m, n M, (26)


1 c2
> 1. (27)
M ( 2
2 c1 + d )

and the average dwell time of the switched discrete-time signal (k) satisfies

M ln
a > a = . (28)
ln 1 ln 2

where
N

P i,m = m
ij Pj,m ,
j=1

1 = ( 1)c2 1 M + d 2 ,

2 = c1 2 ( 1) + d 2 .

Proof. We now consider the H performance of system (4).

8
Define

Jk = ek ek 2 k k . (29)

Following the similar line as before, we have

V ( k+1 , rk+1 , k ) V (k , rk , k ) + Jk

 
Ai,m P i,m Ai,m + C i,m P i,m C i,m Pi,m
 
Ai,m P i,m B i,m + C i,m P i,m Di,m (30)
k  
k .
B i,m P i,m B i,m + Di,m P i,m Di,m 2 I
By using Schur complement, (25) and (30) guarantee that

V (k+1 , rk+1 , k ) V (k , rk , k ) + Jk < 0. (31)

Applying (31) recursively gives


k1

V (k , rk , k ) < kk0 V (k0 , rk0 , k ) kl1 J(l). (32)
l=k0

It follows from (31) and (32) that


k1

V (k , rk , k ) kks V (ks , rks , k ) kl1 J(l)
l=ks
k1

kks V (ks , rs , ks1 ) kl1 J(l)
l=ks

s 1
k k1

kks ks ks1 V (xks1 , rs1 , ks1 ) ks l1 J(l) kl1 J(l)
l=ks1 l=ks

ks1 1

kks1 2 ks1 ks2 V (ks2 , rs2 , ks2 ) ks1 l1 J(l)
l=ks2

s 1
k k1
 (33)
kks ks l1 J(l) kl1 J(l)
l=ks1 l=ks


s 1
k
kk0 N (k0 ,k) V (k0 , rk0 , k0 ) kks N (k0 ,k) ks l1 J(l)
l=k0
2 1
k k1

kk2 N (ks ,k) k2 l1 J(l) kl1 J(l)
l=ks l=ks
k1

=kk0 N (k0 ,k) V (k0 , rk0 , k0 ) N (l,k) kl1 J(l).
l=k0

Under zero condition, one has V (k0 , rk0 , k0 ) = 0 and V (k , rk , k ) 0, thus, one can easily have
N
 N

el el 2 l l . (34)
l=0 l=0

9
Thus it is concluded by Definition 2.3 that system (4) is finite-time boundness with an H performance
. The proof is completed. 

4. Finite-time H filter

Theorem 4.1. system (4) is finite-time bounded with respect to (0, c2 , d, R, M ) and a weighted H per-
formance , if there exist real matrixes Xi,m > 0, Yi,m > 0, G1i,m > 0, G2i,m , Af i,m , B f i,m , C f i,m , Df i,m ,
and scalars 1, > 0, > 1, > 0, such that (i, j) N N , (m, n) M M, i = j, m = n

P1i,m P2i,m
>0 (35)
P3i,m
  
G1i,m Ai + B f i,m Ci Af i,m B f i,m Di
i,m 0      
Xi,m Ai + B f i,m Ci Af i,m B f i,m Di


I Li Df i,m Ci C f i,m Ldi Df i,m Di
< 0, (36)

P1i,m P2i,m 0

0
P3i,m
2 I

P1i,m P2i,m P1i,n P2i,n
, (37)
P3i,m P3i,n
1 1
0 < R 2 Pi,m R 2 < I, (38)

2 d < c2 M . (39)

with the average dwell time of the switching signal satisfying

M ln
a > a = . (40)
ln 2 ln ln d

where

P 1i,m G1i,m G1i,m P 2i,m G3i,m Yi,m

i,m = .

P 3i,m Yi,m Yi,m

c2 M
= ( 1) + 2 d,
0
1 1
0 = max {max (R 2 Pi,m R 2 )}.
iN ,mN

Moreover, if (4) has a feasible solution, then the admissible filter as follows


Af i,m = G1  1
2i,m Af i,m G2i,m G4i,m , Bf i,m = G2i,m B f i,m ,
(41)
Cf i,m = C f i,m G 
2i,m G4i,m , Df i,m = D f i,m .

10
Proof. Using the fact that
1
(P i,m Gi,m ) P i,m (P i,m Gi,m ) 0. (42)

We can obtain
 1 
Gi,m P i,m Gi,m P i,m Gi,m Gi,m . (43)
1 1
Pre- and post- multiplying (25) with diag{P i,m Gi,m , I, I, I} and diag{(P i,m Gi,m ) , I, I, I}, respectively.
We have


P Gi,m Gi,m 0 Gi,m Ai,m Gi,m B i,m
i,m

I C i,m Di,m
< 0. (44)

Pi,m 0

2 I
Partition Gi,m as

G1i,m G2i,m
Gi,m = . (45)
G3i,m G4i,m

It follows from (36) that G4i,m is invertible. Denote



I 0 P P2i,m
K= , H = diag{K, I, K, I}, HPi,m H = 1i,m . (46)
0 G2i,m G14i,m P3i,m

Pre- and post-multiplying the left hand side of (44) by H and H , respectively, it yields that

 
 G1i,m Ai + G2i,m Bf i,m Ci
  G2i,m Af i,m G G1i,m Bi + G2i,m Bf i,m Di
i,m
0   G2i,m G1
4i,m G3i,m Ai    4i,m   

G2i,m Af i,m G
4i,m G3i,m Bi + G2i,m Bf i,m Di
+G2i,m Bf i,m Ci


I Li Df i,m Ci Cf i,m G
4i G2i,m Ldi Df i,m Di < 0,

0

P1i,m P2i,m

P3i,m 0

2 I
(47)

where

P 1i,m G1i,m G1i,m P 2i,m G3i G2i,m G4i,m G2i,m
i,m = .
P 3i,m G2i,m G4i,m G2i,m G2i,m G4i,m G2i,m
Denote

Xi,m = G2i,m G1 
4i,m G3i,m , Yi = G2i,m G4i,m G2i,m ,

Af i,m = G2i,m Af i,m G 


4i,m G2i,m , B f i,m = G2i,m Bf i,m ,
(48)

C f i,m = Cf i,m G 
4i,m G2i,m , D f i,m = Df i,m .

11
From (47) and (48), we can easily obtain (36). Moreover, we can find


Af i,m = G1  1
2i,m Af i,m G2i,m G4i,m , Bf i,m = G2i,m B f i,m ,
(49)
Cf i,m = C f i,m G 
2i,m G4i,m , Df i,m = D f i,m .

Therefore, if (36) hold, system is finite-time boundness with a prescribed H performance index and
average dwell time of the switching signal. The proof is completed. 
2
Remark 5 In many actual applications, the minimum value of min is of interested. In theorem 3.2, with
a fixed and , min can be obtained through following optimization procedure

min 2

s.t. (35)-(39)

In theorem 4.1, as for finite-time stability and boundness, once the state bound c2 is not ascertained, the
minimum value c2 min is of interested. With a fixed and , and define 1 = 1, 2 = , then the following
optimization problem can be formulated to get minimum value c2 min

min 2 + (1 )c2

s.t. (35)-(39)

where is weighted factor, and [0, 1].

5. Illustrative example

Example 1. Consider system (1) with two operation modes and the following data

0.88 0.05 0.8 0.16    
A1 = , A2 = , B1 = 0.7 1.3 , B2 = 1.1 0.9 ,
0.4 0.72 0.8 0.64
     
C1 = 0.2 0.1 , C2 = 0.1 0.2 , D1 = 0.35, D2 = 0.1, L1 = 1.8 0.2 ,
 
L2 = 0.5 0.5 , Ld1 = 1.88, Ld2 = 1.98.

The piecewise-constant transition probabilities are given as



0.1 0.9 0.8 0.2
1 = , 2 = .
0.1 0.9 0.9 0.1

Choosing c1 = 0, N = 2, M = 2, R = I, d = 0.01. For subsystem 1, when = 1.01, through Theorem


4.1, the value 1 min = 2.9707. For subsystem 2, when = 1.01, we have 2 min = 2.7810. The variety of
has little effect on the value of min . For given the initial state [0.5, 1.2], the solution trajectory of the
system (4) is show in Figure 1. Then we can obtain the optimal bound of c2 with different value of a in

12
X Y Plot
1.5

0.5

Y Axis
0

0.5

1.5
1.5 1 0.5 0 0.5 1 1.5 2
X Axis

Figure 1: Phase plane plot

10

6
a

1
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2
c
2

Figure 2: The optimal bound of a with different value of c2

each subsystems, as shown in Figures 2. The state responses of system (4) is shown in Figure 3-4. We can
2
obtain that the value of min = 7.7340 and a = 4.9852. By solving the matrix equalities in Theorem 4.1,
we have the following controller gain


1.136 1.123 0.019  
Af 1,1 = , Bf 1,1 = , Cf 1,1 = 0.689 1.187 , Df 1,1 = 6.492.
0.415 0.411 0.007

4.512 4.235 7.631  
Af 2,1 = , Bf 2,1 = , Cf 2,1 = 1.668 1.124 , Df 2,1 = 4.803.
3.999 3.754 6.764

2.522 1.552 1.018  
Af 1,2 = , Bf 1,2 = , Cf 1,2 = 0.868 1.469 , Df 1,2 = 6.492.
0.925 0.565 0.373
13
X Y Plot

3.5

2.5

Y Axis
1.5

0.5

0 0.5 1 1.5 2 2.5 3 3.5


X Axis

Figure 3: The state of system (4)

3.5

2.5

1.5

0.5

0
0 1 2 3 4 5 6 7 8 9 10

Figure 4: xk Rxk


1.318 0.438 0.011  
Af 2,2 = , Bf 2,2 = , Cf 2,2 = 1.713 1.248 , Df 2,2 = 4.813.
1.169 0.389 0.010
Remark 6 Note that in Example 1, we choose the fixed values for c1 and M , and optimise over value c2 .
Similarly, we can fix c2 and look for the maximum admissible c1 guaranteeing the Finite-time bounded of
system (4).

6. Conclusions

In this paper, we examine the problems of finite-time H estimation for a class of discrete-time Markov
jump systems with time-varying transition probabilities subject to average dwell time switching. Based on
the analysis result and average dwell time method, the static state feedback finite-time boundness is given.

14
Although the derived results are not in LMIs form, we can turn it into LMIs feasibility problem by fixing
some parameters. It is possible to extend the main results obtained in this paper to the practical systems
such as failures of controllers or actuator saturation. Further research directions would include the extension
of our main results to more complex discrete-time systems with time-varying delays. Details will be reported
in our follow-up work in the future.

7. Acknowledgment

The authors would like to thank the associate editor and the anonymous reviewers for their detailed
comments and suggestions. This work was partly supported by the National Basic Research Program
of China (2010CB732501), the China Postdoctoral Science Foundation (2012M521683, 2013T60848), the
National Basic Research Program of China (2010CB732501), the Governor of Guizhou Province Talent
Funds (2012) and the Program for New Century Excellent Talents in University (NCET-10-0097), the
Program for New Century Excellent Talents in University (NCET-10-0097), CSC (No.201306070065) and
YBXSZC20131006.

References

[1] N. Krasovskii, E. Lidskii, Analysis and design of controllers in systems with random attributes, Automatic Remote Control
22 (1961) 1021-1025.
[2] Z. Shu, J. Lam, S. Xu, Robust stabilization of Markovian delay systems with delay-dependent exponential estimates.
Automatica 46 (2006) 2001-2008.
[3] B. Chen, Y. Niu, H. Huang, Output feedback control for stochastic Markovian jumping systems via sliding mode design,
Optimal Control Applications and Methods 32 (2011) 83-94.
[4] H. Gao, Z .Fei, J. Lam, B. Du, Further results on exponential estimates of Markovian jump systems with mode-dependent
time-varying delays, IEEE Transactions on Automatic Control 56 (6) (1) (2011) 223-229.
[5] M. Luo, S. Zhong, Passivity analysis and passification of uncertain Markovian jump systems with partially known transition
rates and mode-dependent interval time-varying delays, Computers and Mathematics with Applications 63 (2012) 1266-
1278.
[6] Z. Wu, P. Shi, H. Su, J. Chu, Asynchronous l2 -l filtering for discrete-time stochastic Markov jump systems with randomly
occurred sensor nonlinearities, Automatica (2014) http://dx.doi.org/10.1016/j.automatica.2013.09.041.
[7] S. Hu, D. Yue, J. Liu, H filtering for networked systems with partly known distribution transmission delays, Information
Sciences 194 (2012) 270-282.
[8] L. Zhang, E. Boukas, Mode-dependent H filtering for discrete-time Markovian jump linear systems with partly unknown
transition probabilities, Automatica 45 (2009) 1462-1467.
[9] L. Zhang, E. Boukas, Stability and stabilization of Markovian jump linear systems with partly unknown transition prob-
abilities, Automatica 45 (2009) 463-468.
[10] L. Zhang, E. Boukas, H control for discrete-time Markovian jump linear systems with partly unknown transition prob-
abilities, International Journal of Robust and Nonlinear Control 19 (2009) 868-883.
[11] A. Filardo, Business-cycle phases and their transitional dynamics, Journal of Business and Economic Statistics 12 (3)
(1994) 299-308.

15
[12] L. Zhang, Y. Leng, L. Chen, Y. Zhao, A BRL for a Class of discrete-time Markov jump linear system with piecewise-
constant TPs, Preprints of the 18th IFAC World Congress Milano (Italy) (2011) 8699-8704.
[13] F. Takeuchi, US external debt sustain ability revisited: Bayesian analysis of extended Markov switching unit root test,
Japan and the World Economy 22 (2) (2010) 98-106.
[14] S. Boyd, L. Ghaoui, E. Feron, V. Balakrishnan, Linear Matrix Inequalities in System and Control Theory, SIAM, Philadel-
phia, 1994.
[15] S. Hu, D. Yue, J. Liu, H filtering for networked systems with partly known distribution transmission delays, Information
Sciences 194 (2012) 270-282.
[16] Q. Zhu, J. Cao, Robust exponential stability of Markovian jump impulsive stochastic Cohen-Grossberg neural networks
with mixed time delays, IEEE Transactions on Neural Networks 21 (2010) 1314-1325.
[17] Z. Wu, H. Su, J. Chu, State estimation for discrete Markovian jumping neural networks with time delay, Neurocomputing
73 (10-12) (2010) 2247-2254.
[18] D. Zhang, L. Yu, Q. Wang, C. Ong, Estimator design for discrete-time switched neural networks with asynchronous
switching and time-varying delay, IEEE Transactions on Neural Networks and Learning Systems 23 (5) (2012) 827-834.
[19] L. Chen, Y. Leng, H. Guo, P. Shi, L. Zhang, H control of a class of discrete-time Markov jump linear systems with
piecewise-constant TPs subject to average dwell times witching, Journal of the Franklin Institute 349 (2012) 1989-2003
[20] P. Bolzern, P. Colaneri, D. Nicolao, Markov jump liner systems with switching transition rates: Mean square stability
with dwell-time, Automatica 46 (6) (2010) 1081-1088.
[21] L. Zhang, E. Boukas, P. Shi, Exponential H filtering for uncertain discrete-time switched linear systems with average
dwell time: a dependent approach, International Journal of Robust and Nonlinear Control 18 (11) (2008) 1188-1207.
[22] V. Phat, Switched controller design for stabilization of nonlinear hybrid systems with time-varying delays in state and
control, Journal of the Franklin Institute 347 (1) (2010) 195-207.
[23] X. Huang W. Lin, B. Yang, Global finite-time stabilization of a class of uncertain nonlinear systems, Automatica 41 (5)
(2005) 881-888.
[24] C. Qian, J. Li, Global finite-time stabilization by output feedback for planar systems without observable linearization.
IEEE Transactions on Automatic Control 50 (6) (2005) 549-564.
[25] X. Luan, S. Zhao, F. Liu, H control for discrete-time Markov jump systems with uncertain transition probabilities,
IEEE Transactions on Automatic Control 58 (6) (2013) 1566-1572.
[26] X. Luan, P. Shi, F. Liu, Finite-time stabilization for Markov jump systems with Gaussian transition probabilities, IET
Control Theory and Applications 7 (2) (2013) 298-304.
[27] H. Liu, Y. Shen, X. Zhao, Finite-time stabilization and boundedness of switched linear system under state-dependent
switching, Journal of the Franklin Institute 350 (3) (2013) 541-555.
[28] J. Cheng, S. Zhong, Q. Zhong, H. Zhu, Y. Du, Finite-time boundedness of state estimation for neural networks with
time-varying delays, Neurocomputing 129 (2014) 257-264.
[29] X. Luan, F. Liu, P. Shi, Finite-time filtering for non-linear stochastic systems with partially known transition jump rates,
IET Control Theory and Applications 4 (5) (2010) 735-745.
[30] J. Cheng, H. Zhu, S. Zhong, Y. Zeng, L. Hou, Finite-time H filtering for a class of discrete-time Markovian jump
systems with partly unknown transition probabilities, International Journal of Adaptive Control and Signal ProcessingDOI:
10.1002/acs.2425.
[31] J. Cheng, H. Zhu, S. Zhong, Y. Zhang, Y. Li, Finite-time H control for a class of discrete-time Markov jump systems
with partly unknown time-varying transition probabilities subject to average dwell time switching, International Journal
of Systems Science, http://dx.doi.org/10.1080/00207721.2013.808716.
[32] J. Cheng, H. Zhu, S. Zhong, Y. Zeng, X. Dong, Finite-time H control for a class of Markovian jump systems with

16
mode-dependent time-varying delays via new Lyapunov functionals, ISA Transactions 52 (2013) 768-774.
[33] S. Bhat, D. Bernstein, Finite-time stability of continuous autonomous systems, SIAM J. Control Optim. 38 (3) (2000)
751-766.

17
Z,

The time-varying transition probabilities mean that the transition probabilities are
varying but invariant within an interval.
The system trajectory stays within a prescribed bound.
The number of switches in a finite interval is bounded.

Das könnte Ihnen auch gefallen