Information Sciences 179 (2009) 3697–3710
Contents lists available at ScienceDirect
Information Sciences journal homepage: www.elsevier.com/locate/ins
Robust H1 filter design for uncertain fuzzy neutral systems q Jun Yang a,b,*, Shouming Zhong b, Guihua Li a, Wenpin Luo c a b c
School of Computer Science, Civil Aviation Flight University of China, Guanghan, Sichuan 618307, PR China School of Applied Mathematics, University of Electronic Science and Technology of China, Chengdu, Sichuan 610054, PR China College of Science, Sichuan University of Science and Engineering, Zigong, Sichuan 643000, PR China
a r t i c l e
i n f o
Article history: Received 29 April 2008 Received in revised form 17 June 2009 Accepted 22 June 2009
Keywords: Takagi–Sugeno (T–S) fuzzy systems Neutral systems Uncertain systems H1 filtering Lyapunov–Krasovskii functional Linear matrix inequalities (LMIs)
a b s t r a c t This paper investigates the problem of robust H1 filtering for a class of uncertain Takagi–Sugeno (T–S) fuzzy neutral systems with time-varying delay and linear fractional parametric uncertainties. The aim is to design an asymptotically stable fuzzy filter ensuring asymptotical stability and prescribed H1 performance of the filtering error system for all admissible uncertainties. Based on the Lyapunov–Krasovskii functional, the Barbalat lemma and the LMI approach, delay-dependent sufficient conditions for the solvability of this problem are proposed in terms of LMIs, and an explicit expression of the desired H1 fuzzy filter is readily given when these LMIs are feasible. Finally, the H1 filter design problem for a neutral-type nonlinear distributed network (long line with tunnel diode) is considered in the simulation example to demonstrate the effectiveness and applicability of the proposed method. Ó 2009 Elsevier Inc. All rights reserved.
1. Introduction The H1 filtering problem is to design an estimator to estimate the unknown state combination via output measurement, which guarantees that the L2 -induced gain from the external disturbance to the estimation error is less than a prescribed level [2,11,23,24,31]. In contrast with the well-known Kalman filter, one of the main advantages of the H1 filtering is that it is not necessary to know exactly the statistical properties of the external disturbance, only assuming that the external disturbance has bounded energy. This advantage renders the H1 filtering approach very appropriate to some practical applications (see, e.g. [7,16,29]). In addition, the parameter uncertainty that is the source of instability or degradation of control performance often arises in a filter system, so recently the robust H1 filtering problem has been extensively studied (see, e.g. [17,31,32,40]). The T–S fuzzy model [34] has been shown to be a powerful tool for modeling complex nonlinear system. It is well-known that, by means of the T–S fuzzy model, a nonlinear system can be represented by a weighted sum of some simple linear subsystems and then can be stabilized by a model-based fuzzy control. Therefore, it provides a good opportunity to employ the well-established theory of linear system to investigate the complex nonlinear system. Over the past two decades, many issues related to stability analysis and control synthesis of T–S fuzzy systems have been reported (see, e.g. [4,18,22,25,35,48]). On the other hand, as a source of instability, time delay is often encountered in various engineering systems such as chemical processes, long transmission lines in pneumatic systems [12]. Recently, the T–S fuzzy system with time delay was introduced in [5]. During the past two decades, the study of the time-delay T–S fuzzy system has received much attention
q
This work was supported by the National Natural Science Foundation of China under Grant 60736029. * Corresponding author. Address: School of Computer Science, Civil Aviation Flight University of China, Guanghan, Sichuan 618307, PR China. E-mail address:
[email protected] (J. Yang).
0020-0255/$ - see front matter Ó 2009 Elsevier Inc. All rights reserved. doi:10.1016/j.ins.2009.06.024
3698
J. Yang et al. / Information Sciences 179 (2009) 3697–3710
(see, e.g. [10,30,38,42,45,46,49]). The most conspicuous of these works is the H1 filter design for T–S fuzzy delayed system (see, e.g. [10,42,45,49]). Moreover, it is also well-known that many practical delayed processes can be modeled as general neutral systems, which contain delays both in their states and in the derivatives of their states, such as circuit analysis, computer aided design, realtime simulation of mechanical systems, power systems, chemical process simulation, optimal control [19]; in particular, some practical delayed nonlinear processes have been modeled as general nonlinear neutral systems, such as vacuum-tube oscillation [27], the car-following problem [14], distributed networks – long lines with tunnel diodes [20], the dynamics of the growth of capital stock [9], and nonlinear fluid dynamics [1]. So the stability and stabilization analysis of (nonlinear) neutral systems have recently been extensively investigated (see, e.g. [8,13,15,26,33]). As mentioned above, with the T–S fuzzy model, a nonlinear neutral system can be represented as a weighted sum of some simple linear neutral subsystems; then it provides a good chance to make use of the well-established theory of linear neutral systems to investigate the complex nonlinear neutral systems. So the T–S fuzzy neutral system was recently introduced in [41], where both the stabilization and H1 control problems were studied by the LMI approach. In [44], via a descriptor system approach introducing free matrices into the original systems, a generalized delay-dependent sufficient condition for T–S fuzzy neutral systems to achieve H1 disturbance attenuation was given. Utilizing the Lyapunov–Krasovskii functional and the LMI approach, Li and Xu [21] has investigated robust H1 control for uncertain T–S fuzzy neutral systems with both discrete and distributed delays. Based on the Lyapunov–Krasovskii functional, the descriptor system approach and the LMI approach, Yang et al. [43] presented sufficient conditions for solvability of non-fragile H1 control problem for a class of uncertain fuzzy neutral systems. However, to the best of the authors’ knowledge, the robust H1 filtering problem has not been addressed for T–S fuzzy neutral systems with time-varying delay and linear fractional parametric uncertainties, which motivates the present study. This paper aims to design an asymptotically stable fuzzy filter ensuring asymptotical stability and prescribed H1 performance of the filtering error system for all admissible uncertainties. The main contribution of this paper lies in the following aspects. First, based on the Lyapunov–Krasovskii functional, the Barbalat lemma and the LMI approach, delay-dependent sufficient conditions for solvability of the robust H1 filtering problem are obtained. It is worth mentioning that it is of great advantage to use the Barbalat lemma instead of the operator method (a very popular approach to stability analysis and control synthesis for general neutral systems; see, e.g. [12]) to deal with the H1 filter problem for T–S fuzzy neutral systems, since a strict requirement that the fuzzy weighting functions be differentiable with bounded derivatives (this requirement is inevitable when the operator method is involved) has been overcome via the Barbalat lemma. Second, as an application of these theoretical results, the H1 filtering problem for a nonlinear neutral-type distributed network (long line with tunnel diode) is investigated by the T–S approach proposed in this paper. The rest of this paper is organized as follows. The main problem is formulated in Section 2, and sufficient conditions for the solvability of the robust H1 filtering problem are derived in Section 3. A simulation example is provided in Section 4, and some concluding remarks appear in Section 5. Notations. Rn and Rnm denote, respectively, the n-dimensional Euclidean space and the set of all n m real matrices. A > ðPÞB means that A B is positive (semi-positive) definite. I is the identity matrix with the appropriate dimensions. A represents the sum of the A and its transpose. ‘‘*” denotes the eleA1 ðAT ) denotes the inverse (transpose) of the matrix A. ~ ments below the main diagonal of a symmetric block matrix. k k denotes the Euclidean norm in Rn . L2 ½0; 1Þ denotes the space of square integrable functions on ½0; 1Þ and k k2 the L2 -norm. Cð½s; 0; Rn Þ is the family of continuous functions / from the interval ½s; 0 to Rn with the norm k/ks ¼ sups6h60 k/ðhÞk.
2. Problem formulation In this section, a class of uncertain T–S fuzzy neutral systems is considered. For each i ¼ 1; 2; . . . ; r (r is the number of plant rules), the ith rule of the T–S fuzzy model is represented as follows: Plant Rule i: IF n1 ðtÞ is M i1 ; n2 ðtÞ is M i2 ; . . . ; np ðtÞ is M ip , THEN
_ sðtÞÞ þ Ei wðtÞ; _ xðtÞ ¼ Ai ðtÞxðtÞ þ Bi ðtÞxðt sðtÞÞ þ C i ðtÞxðt yðtÞ ¼ Di ðtÞxðtÞ þ F i ðtÞxðt sðtÞÞ þ Hi wðtÞ; zðtÞ ¼ Li xðtÞ; xðtÞ ¼ /ðtÞ;
t P 0;
ð1Þ
t P 0;
ð2Þ
t P 0;
ð3Þ
t 2 ½s; 0;
ð4Þ n
where n1 ðtÞ; n2 ðtÞ; . . . ; np ðtÞ are the premise variables, and each M il ðl ¼ 1; 2; . . . ; pÞ is a fuzzy set. xðtÞ 2 R is the state; yðtÞ 2 Rm is the measured output; zðtÞ 2 Rp is the signal to be estimated; wðtÞ 2 Rq is the exogenous disturbance input, which is assumed to be an arbitrary signal in L2 ½0; 1Þ; /ðtÞ 2 Cð½s; 0; Rn Þ is the initial condition, and sðtÞ is the time-varying delay satisfying
0 6 sðtÞ 6 s < 1:
ð5Þ
Moreover, Ei ; Hi and Li are known real constant matrices; Ai ðtÞ; Bi ðtÞ; C i ðtÞ; Di ðtÞ and F i ðtÞ are system matrices with appropriate dimension and admissible linear fractional parametric uncertainties, that is, these system matrices satisfy
J. Yang et al. / Information Sciences 179 (2009) 3697–3710
3699
½Ai ðtÞ Bi ðtÞ C i ðtÞ Di ðtÞ F i ðtÞ ¼ ½Ai þ DAi ðtÞ; Bi þ DBi ðtÞ; C i þ DC i ðtÞ; Di þ DDi ðtÞ; F i þ DF i ðtÞ ¼ ½Ai Bi C i Di F i þ Li DðtÞ½N0i N1i N2i N3i N4i ;
DðtÞ ¼ ½I EðtÞJ1 EðtÞ;
ð6Þ ð7Þ
T
I JJ > 0;
ð8Þ
where Ai ; Bi ; C i ; Di ; F i ; Li ; N 0i ; N 1i ; N 2i ; N 3i ; N 4i and J are known real constant matrices with appropriate dimensions, and EðtÞ is a matrix function satisfying
EðtÞET ðtÞ 6 I:
ð9Þ
Remark 1. The time delay satisfying (5) may be either constant or time-varying, and either differentiable or nondifferentiable. Remark 2. The uncertainty DðtÞ satisfying (7)–(9) is referred to as a linear fractional parametric uncertainty. The uncertainty in (7) is well defined, since the condition (8) guarantees that I EðtÞJ is invertible for all EðtÞ satisfying (9). This class of uncertainties has been extensively investigated in the existing literatures (see, e.g. [48] and the references therein). Note that when J ¼ 0, DðtÞ reduces to the well-known norm-bounded parametric uncertainty. Applying a center-average defuzzier, product inference and singleton fuzzifier, the dynamic fuzzy model in (1)–(3) can be represented by
_ xðtÞ ¼
r X
_ sðtÞÞ þ Ei wðtÞg; hi ðnðtÞÞfAi ðtÞxðtÞ þ Bi ðtÞxðt sðtÞÞ þ C i ðtÞxðt
ð10Þ
hi ðnðtÞÞfDi ðtÞxðtÞ þ F i ðtÞxðt sðtÞÞ þ Hi wðtÞg;
ð11Þ
hi ðnðtÞÞLi xðtÞ;
ð12Þ
i¼1 r X
yðtÞ ¼
i¼1
zðtÞ ¼
r X i¼1
where
Qp M il ðnl ðtÞÞ Qp hi ðnðtÞÞ ¼ Pr l¼1 ; i¼1 l¼1 M il ðnl ðtÞÞ
i ¼ 1; . . . ; r
ð13Þ
in which Mil ðnl ðtÞÞ is the grade of membership of nl ðtÞ in Mil , and nðtÞ ¼ ðn1 ðtÞ; . . . ; nr ðtÞÞ. For convenience, call hi ðnðtÞÞ ði ¼ 1; 2; . . . ; rÞ the fuzzy weighting functions. It follows from (13) that
hi ðnðtÞÞ P 0 ði ¼ 1; 2; . . . ; rÞ and
r X
hi ðnðtÞÞ ¼ 1:
ð14Þ
i¼1
For notational simplicity, hi is used to represent hi ðnðtÞÞ in the following description. Now, consider a full-order fuzzy filter as follows: Filter rule i: IF n1 ðtÞ is M i1 ; n2 ðtÞ is M i2 ; . . . ; np ðtÞ is M ip , THEN
^x_ ðtÞ ¼ Afi ^xðtÞ þ Bfi ^xðt sðtÞÞ þ C fi ^x_ ðt sðtÞÞ þ K fi yðtÞ; ^zðtÞ ¼ Lfi ^xðtÞ; ^xðsÞ ¼ 0;
ð15Þ
s 2 ½s; 0;
where i ¼ 1; 2; . . . ; r; ^ xðtÞ 2 Rn and ^zðtÞ 2 Rp ; Afi ; Bfi ; C fi ; K fi and Lfi are filter gain matrices to be determined. Via the fuzzy weighting functions defined by (13), the overall fuzzy filter can be represented by
P ^x_ ðtÞ ¼ hi fAfi ^xðtÞ þ Bfi ^xðt sðtÞÞ þ C fi ^x_ ðt sðtÞÞ þ K fi yðtÞg; r
^zðtÞ ¼
i¼1 r P
ð16Þ hi Lfi ^xðtÞ:
i¼1
Let
xf ðtÞ ¼ ½xT ðtÞ; ^xT ðtÞT ;
zf ðtÞ ¼ zðtÞ ^zðtÞ;
then the filtering error dynamics from systems (10)–(12) and (16) can be represented by
3700
J. Yang et al. / Information Sciences 179 (2009) 3697–3710
x_ f ðtÞ ¼
ðRÞ :
r X r X i¼1
zf ðtÞ ¼
r X r X i¼1
hi hj fAij ðtÞxf ðtÞ þ Bij ðtÞxf ðt sðtÞÞ þ Cij ðtÞx_ f ðt sðtÞÞ þ Eij wðtÞg;
ð17Þ
hi hj Lij xf ðtÞ;
ð18Þ
j¼1
j¼1
where
Aij ðtÞ ¼ Aij þ DAij ðtÞ; Bij ðtÞ ¼ Bij þ DBij ðtÞ; Cij ðtÞ ¼ Cij þ DCij ðtÞ; Ai Bi 0 DAi ðtÞ 0 0 Aij ¼ ; DAij ðtÞ ¼ ; ; Bij ¼ K fj Di Afj K fj DDi ðtÞ 0 K fj F i Bfj Ci 0 DBi ðtÞ 0 DC i ðtÞ 0 DBij ðtÞ ¼ ; DCij ðtÞ ¼ ; Cij ¼ ; K fj DF i ðtÞ 0 0 C fj 0 0 Ei ; Lij ¼ ½ Li Lfj : Eij ¼ K fj Hi The robust H1 filtering problem addressed in this paper is formulated as follows: given the uncertain T–S fuzzy neutral system in (1)–(4) and a prescribed level of noise attenuation c > 0, design an asymptotically stable fuzzy filter in the form of (15) such that, for all admissible uncertainties, (i) when w 0, the filtering error system ðRÞ is robustly asymptotically stable; (ii) for a prescribed scalar c > 0, under zero initial conditions, ðRÞ satisfies
kzf k2 6 ckwk2
ð19Þ
for any nonzero w 2 L2 ½0; 1Þ. The filtering error system ðRÞ is said to be robustly asymptotically stable with disturbance attenuation level c if (i) and (ii) are satisfied. 3. Main results In this section, an LMI approach is developed to solve the robust H1 filtering problem for the uncertain T–S fuzzy neutral systems formulated in the previous section. Before proceeding, recall the following lemmas which will be used throughout the proofs. Lemma 1 [37]. For the fuzzy weighting functions hi ðnðtÞÞði ¼ 1; 2; . . . ; rÞ defined by (13), any vector variable fðtÞ and matrices Mij ði; j ¼ 1; 2; . . . ; rÞ with appropriate dimension, let D
VðtÞ ¼
r X r X i¼1
hi ðnðtÞÞhj ðnðtÞÞfT ðtÞMij fðtÞ;
ð20Þ
j¼1
then VðtÞ < 0; 8fðtÞ–0; if
Mii < 0;
i ¼ 1; 2; . . . ; r;
and
1 1 Mii þ ðMij þ Mji Þ < 0; r1 2
1 6 i < j 6 r:
Lemma 2 [39] (Barbalat lemma). If a function f ðtÞ is uniformly continuous and limt!1 limt!1 f ðtÞ ¼ 0.
Rt 0
f ðsÞds exists and is finite, then
Lemma 3 [6]. Suppose that matrices Mi 2 Rnm ; i ¼ 1; 2; . . . ; r, and a semi-positive-definite matrix P 2 Rnn are given, then r X i¼1
!T hi ðnðtÞÞMi
P
r X j¼1
! hj ðnðtÞÞMj
6
r X
hi ðnðtÞÞMTi PMi ;
i¼1
where hi ðnðtÞÞ; i ¼ 1; 2; . . . ; r; are defined by (13). Lemma 4 ([47,48]). Suppose that DðtÞ is given by (7)–(9), and matrices M ¼ M T ; L and N of appropriate dimensions are given. Then the following statements are equivalent:
3701
J. Yang et al. / Information Sciences 179 (2009) 3697–3710
(i) the inequality
M þ LDðtÞN þ NT DT ðtÞLT < 0 holds for all EðtÞ satisfying (9); (ii) for any scalar d > 0,
2
3 NT 7 dJ T 5 < 0:
M dL 6 4 dI
dI
Theorem 1. For a prescribed scalar c > 0, the filtering error system ðRÞ is robustly asymptotically stable with disturbance attenuation level c if there exist matrices P > 0; X lk ð1 6 l 6 k 6 3Þ such that the following conditions are satisfied
2
X 11
X 12
6 X¼4
X 22
X 13
3
7 X 23 5 P 0;
ð21Þ
X 33
Mii ðtÞ < 0; i ¼ 1; 2; . . . ; r; 1 Mij ðtÞ þ M ji ðtÞ M ii ðtÞ þ < 0; r1 2
ð22Þ 1 6 i < j 6 r;
ð23Þ
where
2 6 6 6 6 M ij ðtÞ ¼ 6 6 6 6 4
M 11 ij ðtÞ
M12 ij ðtÞ
sX 22 X 23
!
ATij ðtÞP
ATij ðtÞP þ PCij ðtÞ
BTij ðtÞP sX 33 2P
BTij ðtÞP P þ PCij ðtÞ
PCij ðtÞ
3
7 7 7 7 7; 7 7 7 PEij ðtÞ 5 0 PEij ðtÞ
!
PEij ðtÞ
1 6 i 6 j 6 r;
c2 I
with !
!
T M11 ij ðtÞ ¼ PAij ðtÞ þsX 11 þ X 13 þLij Lij ;
T M 12 ij ðtÞ ¼ PBij ðtÞ þ sX 12 X 13 þ X 23 :
Proof. Choose a Lyapunov–Krasovskii functional candidate for the system ðRÞ as follows:
VðtÞ ¼ V 1 ðtÞ þ V 2 ðtÞ þ V 3 ðtÞ;
ð24Þ
where
V 1 ðtÞ ¼ xTf ðtÞPxf ðtÞ; 2 3T 2 xf ðhÞ X 11 Z t Z h 6 7 6 V 2 ðtÞ ¼ 4 xf ðh sðhÞÞ 5 4 0 hsðhÞ x_ f ðsÞ Z 0Z t V 3 ðtÞ ¼ x_ Tf ðsÞX 33 x_ f ðsÞdsdh: s
X 12 X 22
X 13
32
xf ðhÞ
3
76 7 X 23 54 xf ðh sðhÞÞ 5dsdh; x_ f ðsÞ X 33
tþh
The time derivative of VðtÞ along the trajectory of the system ðRÞ is given by
_ VðtÞ ¼ V_ 1 ðtÞ þ V_ 2 ðtÞ þ V_ 3 ðtÞ:
ð25Þ
Differentiating V 1 ðtÞ along ðRÞ yields
V_ 1 ðtÞ ¼ 2xTf ðtÞPx_ f ðtÞ ¼
r X r X i¼1
!
hi hj fxTf ðtÞ PAij ðtÞ xf ðtÞ þ 2xTf ðtÞP½Bij ðtÞxf ðt sðtÞÞ þ Cij ðtÞx_ f ðt sðtÞÞ þ Eij wðtÞg:
ð26Þ
j¼1
Taking the derivative of V 2 ðtÞ gives
V_ 2 ðtÞ ¼
Z
2 t
xf ðtÞ
3T 2
X 11
X 12
X 13
32
xf ðtÞ
3
! 76 7 X 23 54 xf ðt sðtÞÞ 5ds 6 xTf ðtÞðsX 11 þ X 13 Þxf ðtÞ x_ f ðsÞ X 33 Z t ! x_ Tf ðsÞX 33 x_ f ðsÞds; þ 2xTf ðtÞðsX 12 X 13 þ X T23 Þxf ðt sðtÞÞ þ xTf ðt sðtÞÞðsX 22 X 23 Þxf ðt sðtÞÞ þ
6 7 6 4 xf ðt sðtÞÞ 5 4 tsðtÞ x_ f ðsÞ
X 22
ts
ð27Þ
3702
J. Yang et al. / Information Sciences 179 (2009) 3697–3710
where the above inequality holds from the fact (5). V_ 3 ðtÞ is given by
V_ 3 ðtÞ ¼ sx_ Tf ðtÞX 33 x_ f ðtÞ
Z
t ts
x_ Tf ðsÞX 33 x_ f ðsÞds:
ð28Þ
On the other hand, the following equality holds from (17):
½x_ Tf ðtÞ x_ Tf ðt sðtÞÞ2P
( r X r X i¼1
) hi hj ½x_ f ðtÞ þ Aij ðtÞxf ðtÞ þ Bij ðtÞxf ðt sðtÞÞ þ Cij ðtÞx_ f ðt sðtÞÞ þ Eij wðtÞ
¼ 0:
ð29Þ
j¼1
Using Lemma 3 twice, the following inequality holds: r X r X
zTf ðtÞzf ðtÞ ¼
i¼1
!T hi hj Lij xf ðtÞ
j¼1
r X r X
! hl hk Llk xf ðtÞ
6
r X r X i¼1
l¼1 k¼1
hi hj xTf ðtÞLTij Lij xf ðtÞ:
ð30Þ
j¼1
_ Adding the terms on the left side of (29) to VðtÞ, it follows from (25)–(28) and (30) that
_ VðtÞ þ zTf ðtÞzf ðtÞ c2 wT ðtÞwðtÞ 6
r X r X i¼1
hi hj fT ðtÞM ij ðtÞfðtÞ;
ð31Þ
j¼1
where
fðtÞ ¼ ½xTf ðtÞ; xTf ðt sðtÞÞ; x_ Tf ðtÞ; x_ Tf ðt sðtÞÞ; wT ðtÞT : Taking into account (22), (23) and (31), it follows from Lemma 1 that
_ VðtÞ þ zTf ðtÞzf ðtÞ c2 wT ðtÞwðtÞ < 0:
ð32Þ
When wðtÞ ¼ 0, (32) implies that
_ VðtÞ < 0;
8fðtÞ–0;
so
_ VðtÞ 6 ekxf ðtÞk2
e > 0. Integrating both sides of the above inequality from 0 to t gives
for a sufficiently small
VðtÞ 6 Vð0Þ e
Z
t
kxf ðsÞk2 ds
0
and thus
Z
t
kxf ðsÞk2 ds 6
0
therefore, limt!1
Rt 0
Vð0Þ
e
;
kxf ðsÞk2 ds exists and is finite. By Lemma 2 (Barbalat Lemma), one has
limt!1 kxf ðtÞk ¼ 0: Therefore, the system (R) is robustly asymptotically stable when wðtÞ ¼ 0. When wðtÞ–0, introduce
JðTÞ ¼
Z
T
½zTf ðtÞzf ðtÞ c2 wT ðtÞwðtÞdt;
0
where the scalar T > 0. Noting the zero initial conditions, it can be verified that for any nonzero w 2 L2 ½0; 1Þ and T > 0,
JðTÞ ¼
Z 0
T
_ ½VðtÞ þ zTf ðtÞzf ðtÞ c2 wT ðtÞwðtÞdt VðTÞ 6
Z 0
T
_ ½VðtÞ þ zTf ðtÞzf ðtÞ c2 wT ðtÞwðtÞdt:
It thus follows from (32) that JðTÞ < 0 (8 T > 0), which implies that kzf k2 6 ckwk2 for any nonzero w 2 L2 ½0; 1Þ. The proof is complete. In what follows, based on the LMI approach, we can now give the main results on the solvability of the robust H1 fuzzy filtering problem. Theorem 2. For a prescribed scalar c > 0, the robust H1 filtering problem for the uncertain T–S fuzzy neutral system in (1)–(4) is solvable if there exist scalars dij > 0 ð1 6 i 6 j 6 rÞ and matrices Z > 0; Y > 0; Xlk ð1 6 l 6 k 6 3Þ; Aj ; Bj ; Cj ; Kj ; Lj ðj ¼ 1; 2; . . . ; rÞ, with
3703
J. Yang et al. / Information Sciences 179 (2009) 3697–3710
" Xlk ¼
X11 lk
X12 lk
X21 lk
X22 lk
# ð1 6 l 6 k 6 3Þ;
;
such that the following LMIs hold:
2
X11 6 X¼4
X12 X22
X13
3
7 X23 5 P 0; X33
ð33Þ
Z Y > 0;
ð34Þ
Nii < 0;
ð35Þ
i ¼ 1; 2; . . . ; r;
and
2 6 6 6 6 6 6 6 6 6 6 Nij ðrÞ ¼ 6 6 6 6 6 6 6 6 6 4
N11 N12 N13 N14 ij ðrÞ ij ðrÞ ij ðrÞ ij ðrÞ
N15 ij ðrÞ
N22 N23 N24 ij ðrÞ ij ðrÞ ij ðrÞ
N16 N17 ij ðrÞ ij ðrÞ
0
N33 N34 ij ðrÞ ij ðrÞ
44 ij ðrÞ
45 ij ðrÞ r 2 r1 I
N
N19 Nij1;10 ðrÞ ij ðrÞ
N26 ij ðrÞ
0
0
0
0
0
N37 ij ðrÞ
0
0
0
47 ij ðrÞ
N35 ij ðrÞ N
N18 ij ðrÞ
46 ij ðrÞ
0
0
0
0
0
0
0
0
dij I
N67 ij
0 0 ðr 1ÞI
0 0 0 2I
0 0 0 0 2I
N
c
N
dij I
3 7 7 7 7 7 7 7 7 7 7 7 < 0; 7 7 7 7 7 7 7 7 5
1 6 i < j 6 r;
ð36Þ where
2 6 6 6 6 6 6 6 6 6 6 Nij ¼ 6 6 6 6 6 6 6 6 6 4
N11 N12 N13 N14 ij ij ij ij
N15 ij
N16 ij
0
N18 ij
N19 ij
0
N26 ij
0
0
0
N33 N34 ij ij
N35 ij
0
0
N38 ij
N39 ij
N44 ij
N45 ij
0
N47 ij
N48 ij
N49 ij
0
N22 N23 N24 ij ij ij
2
c I
0 dij I
0 0
N68 ij
0 0
dij I
0 dij I
dij J T 0 dij I
Nij1;10
3
7 0 7 7 7 0 7 7 7 0 7 7 0 7 7; 0 7 7 7 0 7 7 7 0 7 7 0 5
I
with
2
N
11 ij
N12 ij N14 ij N13 ij N16 ij N19 ij N23 ij
3 ! YAi þ ATi X þ DTi KTj þ ATj 5 þ sX11 þ X13 ; ! ZAi þ Kj Di YBi YBi þ sX12 X13 þ X32 ; ¼ ZBi þ Kj F i þ Bj ZBi þ Kj F i " # ðATi Z þ DTi Kj T þ ATj Þ þ YC i ATi Y þ YC i ; ¼ ATi Y þ Z T C i þ Cj ðATi Z þ Di T KTj Þ þ ZC i " T # Ai Y ATi Z þ DTi KTj þ ATj YEi 15 45 ; ¼ N ¼ ¼ N35 ij ij ¼ Nij ; T T ZEi þ Kj Hi Ai Y Ai Z þ DTi KTj " # YLi 0 NT0i NT3i 18 48 ; ¼ dij N ¼ ¼ N38 ij ij ¼ Nij ; ZLi Kj Li NT0i NT3i " # ! LTi LTj YLi 1;10 39 49 ; N22 ¼ Nij ¼ Nij ; Nij ¼ ¼ ij ¼ sX22 X23 ; T ZLi Li " T # " # Bi Y BTi Z þ F Ti KTj þ BTj NT1i NT4i 24 26 ¼ Nij ; Nij ¼ dij ; ¼ BTi Y BTi Z þ F Ti KTj NT1i NT4i ¼4
!
YAi
1 6 i 6 j 6 r;
3704
J. Yang et al. / Information Sciences 179 (2009) 3697–3710
N33 ij ¼
2
4 N44 ij ¼
2Y
2Y
2Z !
þ sX33 ;
N34 ij ¼
YC i
ðYC i þ C Ti Z þ CTj Þ
ZC i
!
3 5;
Y þ YC i
Y þ YC i
Y þ ZC i þ Cj Z þ ZC i " # NT2i ; N47 ¼ d ij ij NT2i
;
T T N68 ij ¼ diagfdij J ; dij J g;
and
mn
mn
Nij þ Nji 1 Nmn ; ð1 6 i < j 6 r; m ¼ 1; . . . ; 4; n ¼ 1; . . . ; 5Þ; ii þ r 1 2 " T # N0i NT3i 0 NT0j N T3j 0 N16 ðrÞ ¼ d ; ij ij NT0i NT3i 0 NT0j N T3j 0 " # 1 1 1 þ 12 YLi 0 þ 12 YLi 12 YLj 0 YLj 17 47 r1 r1 2 1 1 ¼ N37 Nij ðrÞ ¼ 1 1 ij ðrÞ ¼ Nij ðrÞ; þ 2 ZLi Ki þ 12 Kj Li þ 12 ZLi 12 ZLj 12 Ki Lj 12 ZLj r1 r1 r1 " # " T # Li LTj LTi LTi 18 19 ; Nij ðrÞ ¼ ; Nij ðrÞ ¼ LTi LTi " T # " T # N1i NT4i 0 N T1j NT4j 0 Lj LTi 26 ; ; N1;10 ðrÞ ¼ N ðrÞ ¼ d ij ij ij LTj NT1i NT4i 0 N T1j NT4j 0 " # 0 0 NT2i 0 0 NT2j T T 46 ; N67 Nij ðrÞ ¼ dij ij ¼ diagfdij J ; . . . ; dij J g: 0 0 NT2i 0 0 NT2j Nmn ij ðrÞ ¼
Furthermore, the H1 fuzzy filter is given in the form of (15), with the following parameters:
½Afj ; Bfj ; C fj ¼ S1 ½Aj ; Bj ; Cj Y 1 W T ;
K fj ¼ S1 Kj ;
Lfj ¼ Lj Y 1 W T ;
j ¼ 1; 2; . . . ; r;
ð37Þ
where S and W are any nonsingular matrices satisfying
SW T ¼ I ZY 1 :
ð38Þ
Proof. Let
Li 0 0 e e ðtÞ ¼ DðtÞ L ij ¼ ; H ¼ ½I ; 1 6 i 6 j 6 r; D 0 K fj Li DðtÞ 0 Li e 0i ¼ N0i ; N e 1i ¼ N 1i ; e N Li ¼ ; i ¼ 1; 2; . . . ; r; N3i N4i 0
0 ;
then it can be verified that
e ðtÞ½N ~ 0i H; N e 1i H; ½DAij ðtÞ; DBij ðtÞ ¼ e L ij D DCij ðtÞ ¼ eL i DðtÞðN2i HÞ; i ¼ 1; 2; . . . ; r:
1 6 i 6 j 6 r;
Hence (22) in Theorem 1 can be rewritten as
2
X11
6 6 6 6 Mii ðtÞ ¼ 6 6 6 4 2
ATii P
ATij P þ PCii
BTii P sX 33 2P
BTii P P þ PCii ! PCii
X12 !
sX 22 X 23
PEii
3
7 7 7 7 7 7 7 PEii 5 0 PEii
c2 I 3 e e P L ii P Li 6 0 " #" # 0 7 7 e 6 e 0i H N e 1i H 0 N 0 0 0 7 D ðtÞ 6 þ 6 Pe L ii Pe Li 7 7 0 6 0 0 0 N2i H 0 DðtÞ 4 P e L ii P e Li 5 0 0 3 2 T ~T H N 0i 0 7" 6 T ~T #" # 0 7 eT 6 H N1i e L Tii P 0 e L Tii P e L Tii P 0 0 7 D ðtÞ 6 þ6 0 < 0; 0 7 7 6 0 DT ðtÞ eL Ti P 0 eL Ti P eL Ti P 0 4 0 HT N T2i 5 0 0
i ¼ 1; 2; . . . ; r;
ð39Þ
3705
J. Yang et al. / Information Sciences 179 (2009) 3697–3710
where !
!
X12 ¼ PBii þ sX 12 X 13 þ X T23 :
X11 ¼ PAii þsX 11 þ X 13 þLTii Lii ; Let
EðtÞ 0 e EðtÞ ¼ ; 0 EðtÞ
eJ ¼
J
0
0
J
;
then
"
# # #" #!1 " "e e ðtÞ e eJ 0 I 0 EðtÞ 0 EðtÞ 0 D 0 ¼ ; 0 I DðtÞ 0 0 EðtÞ 0 EðtÞ 0 J " #" #T " #" #T ~ eJ 0 eJ 0 e EðtÞ 0 EðtÞ 0 < I; 6 I: 0 EðtÞ 0 J 0 J 0 EðtÞ
ð40Þ
ð41Þ
Observing (40) and (41), and using Lemma 4 and the Schur complement, one can see that (39) holds if and only if, for scalars dii > 0; i ¼ 1; 2; . . . ; r, the following inequality holds:
2 6 6 6 6 6 6 6 6 6 6 6 6 M ii ¼ 6 6 6 6 6 6 6 6 6 6 6 6 4
M 11 ii
M12 ii
ATii P
M14 ii
PEii
eT dii HT N 0i
0
Pe L ii
Pe Li
M22 ii
BTii P
BTii P
0
~T dii HT N 1i
0
0
0
M 33 ii
M 34 ii
PEii
0
0
Pe L ii
Pe Li
M 44 ii
PEii
0
dii HT N T2i
P e L ii
P e Li
c2 I
0
0
0
0
dii I
0
dii~J T
0
dii I
0
dii J T
dii I
0
dii I
LTii
3
7 0 7 7 7 7 0 7 7 7 0 7 7 7 0 7 7 < 0; 7 0 7 7 7 0 7 7 7 0 7 7 7 0 7 5
i ¼ 1; 2; . . . ; r;
ð42Þ
I
where !
!
M 11 ii ¼ PAii þsX 11 þ X 13 ; !
M 22 ii ¼ sX 22 X 23 ;
T M 12 ii ¼ PBii þ sX 12 X 13 þ X 23 ;
M 33 ii ¼ sX 33 2P;
M34 ii ¼ P þ PCii ;
T M14 ii ¼ Aii P þ PCii ; !
M 44 ii ¼ PCii :
On the other hand, (34) implies that I ZY 1 is nonsingular. Therefore, there always exist nonsingular matrices S and W such that (38) holds. Now, let
"
P1 ¼
Y 1
I
WT
0
# ;
P2 ¼
I
Z
0 ST
and
P ¼ P2 P1 1 ; then
P¼
Z
S
T
S
;
where
¼ W 1 Y 1 ðZ YÞY 1 W T : Observe that
Z S 1 ST ¼ Y > 0; thus it follows from the Schur complement that P > 0. Let
e ¼ diagfY 1 ; Ig; Y
e P1 ; Y e P1 ; Y e P1 g; e ¼ diagf Y P 1 1 1
e TXP e: X¼P
3706
J. Yang et al. / Information Sciences 179 (2009) 3697–3710
Pre- and post-multiply LMI (35) by T e T e T e e diagfPT 1 Y ; P1 Y ; P1 Y ; P1 Y ; I; I; I; I; I; Ig
and its transpose, respectively; then, observing (37) and the above-introduced matrix variables P and X yielding (42), thus (22) follows. Similarly, LMI (36) implies (23). Therefore, by Theorem 1, the desired results follow immediately. The proof is complete. h Remark 3. To prove the negative-definiteness of the fuzzy summations (20), Proposition 2 in [28] provides a set of sufficient conditions that are progressively less conservative than those proposed in Lemma 1, when the number of plant rules is r P 4 and the complexity parameter is n P 4. However, using Proposition 2 is a sort of a ‘‘not-too-practical” way of obtaining less conservative conditions because the computational cost grows wildly with the complexity parameters there. Therefore, at the cost of dramatically high computational complexity, less conservative conditions can be derived via Proposition 2 than those presented in Theorem 2 by means of Lemma 1. Remark 4. Based on the LMI approach, Theorem 2 establishes delay-dependent sufficient conditions for the solvability of the robust H1 filtering problem for a class of uncertain T–S fuzzy neutral systems with time-varying delay and linear fractional parametric uncertainties. The desired H1 fuzzy filter is readily constructed by solving the LMIs in (33)–(36), which can be implemented by standard numerical algorithm [3]without requiring any tuning of parameters. Remark 5. For the uncertain delayed T–S fuzzy system in (1)–(4) without neutral terms, the robust H1 filtering problem has been investigated in [42]; so, in view of this, the results in this paper to some extent extend the corresponding results in [42].
4. Simulation example In this section, as an application of the theoretical results, the H1 filtering problem for a distributed network (long line with tunnel diode) is considered. The tunnel diode is a two-electrode device on the basis of semiconducting crystals having a very narrow potential barrier hindering the motion of electrons. Such diodes are wildly used in high-frequency amplifiers of electric oscillations and in other devices. The dynamic of the distributed network (long line with tunnel diode) is represented by the following nonlinear neutral functional differential equation (see, e.g. [18]).
_ K xðt _ hÞ ¼ ðaðtÞ 1ÞxðtÞ ð1 þ aðtÞÞKxðt hÞ þ ½xðtÞ Kxðt hÞ2 þ ewðtÞ; xðtÞ
ð43Þ
with 1
h ¼ 2lCC l ;
Z¼
pffiffiffiffiffiffiffiffiffiffi LC 1 ;
K ¼ ðZ R0 ÞðZ þ R0 Þ1 ;
a ¼ aZ; aðtÞ ¼ a þ DaðtÞ;
where l is the length of the conductor, L and C are the inductivity and capacity of the conductor per unit length, respectively, R0 is the resistance at the input, C l is the capacity at the output, a > 0 is a scalar, DaðtÞ is the perturbation of the coefficient a, and e is the coefficient of the exogenous disturbance input wðtÞ. In what follows, applying the idea of ‘‘using local sector nonlinearity in T–S fuzzy model construction”[36], the nonlinear system (43) will be represented by the T–S fuzzy system. For simplicity, it is assumed that xðtÞ; xðt hÞ 2 ½1; 1. Of course, any ranges for xðtÞ and xðt hÞ can be assumed to construct the T–S fuzzy model. Define nðtÞ ¼ xðtÞ Kxðt hÞ; then it follows from (43) that
_ K xðt _ hÞ ¼ ðaðtÞ 1ÞxðtÞ ð1 þ aðtÞÞKxðt hÞ þ nðtÞ½xðtÞ Kxðt hÞ þ ewðtÞ: xðtÞ Next, calculate the maximum and minimum values of nðtÞ under xðtÞ; xðt hÞ 2 ½1; 1:
max
xðtÞ2½1;1; xðthÞ2½1;1
nðtÞ ¼ 1 þ K;
min
xðtÞ2½1;1; xðthÞ2½1;1
nðtÞ ¼ ð1 þ KÞ:
From the maximum and minimum values, nðtÞ can be represented by
nðtÞ ¼ M 1 ðnðtÞÞ ð1 þ KÞ þ M 2 ðnðtÞÞ ð1 KÞ; where
M1 ðnðtÞÞ þ M 2 ðnðtÞÞ ¼ 1: Therefore, the membership functions are given by
M1 ðnðtÞÞ ¼
nðtÞ 1 þ ; 2ð1 þ KÞ 2
which are shown in Fig. 1.
M 2 ðnðtÞÞ ¼
1 nðtÞ ; 2 2ð1 þ KÞ
3707
J. Yang et al. / Information Sciences 179 (2009) 3697–3710
With the membership functions M 1 ðnðtÞÞ and M 2 ðnðtÞÞ, the nonlinear neutral system (43) with (2) and (3) can be represented by the following T–S fuzzy model. Plant Rule 1: IF nðtÞ is M1 , THEN
_ _ hÞ þ E1 wðtÞ; xðtÞ ¼ A1 ðtÞxðtÞ þ B1 ðtÞxðt hÞ þ C 1 ðtÞxðt yðtÞ ¼ D1 ðtÞxðtÞ þ F 1 ðtÞxðt hÞ þ H1 wðtÞ;
ð44Þ
zðtÞ ¼ L1 xðtÞ; Plant Rule 2: IF nðtÞ is M2 , THEN
_ hÞ þ E2 wðtÞ; _ xðtÞ ¼ A2 ðtÞxðtÞ þ B2 ðtÞxðt hÞ þ C 2 ðtÞxðt yðtÞ ¼ D2 ðtÞxðtÞ þ F 2 ðtÞxðt hÞ þ H2 wðtÞ;
ð45Þ
zðtÞ ¼ L2 xðtÞ; where
A1 ðtÞ ¼ aðtÞ þ K; B1 ðtÞ ¼ ðaðtÞ þ K þ 2ÞK; C 1 ðtÞ ¼ K; E1 ¼ e; A2 ðtÞ ¼ aðtÞ K 2; B2 ðtÞ ¼ ðK aðtÞÞK; C 2 ðtÞ ¼ K; E1 ¼ e: Assume aðtÞ ¼ a þ a sinðtÞ , then 5
A1 ðtÞ ¼ ða þ KÞ þ 0:2aDðtÞ;
B1 ðtÞ ¼ ða þ K þ 2ÞK 0:2aK DðtÞ;
A2 ðtÞ ¼ ða K 2Þ þ 0:2aDðtÞ;
B2 ðtÞ ¼ ðK aÞK 0:2aK DðtÞ;
Fig. 1. Membership functions M1 ðnðtÞÞ and M 2 ðnðtÞÞ.
0.4
0.2
States
0
−0.2
−0.4
−0.6
−0.8
0
5
10
15
20
25
Time (s)
Fig. 2. State responses xðtÞðÞ and ^ xðtÞð Þ.
30
3708
J. Yang et al. / Information Sciences 179 (2009) 3697–3710
where DðtÞ ¼ sinðtÞ. For the convenience of the simulation, set the above parameters to
l ¼ 50 m; L ¼ 200 lH; C ¼ 3600 lF; C l ¼ 540000 lF; R0 ¼ 160 X; e ¼ 0:5;
D1 ðtÞ ¼ 0:6;
F 1 ðtÞ ¼ 0:4;
D2 ðtÞ ¼ 0:5;
F 2 ðtÞ ¼ 0:3;
H2 ¼ 0:2;
a ¼ 0:0025; L1 ¼ 1;
H1 ¼ 0:3; L2 ¼ 1:4:
In this example, the H1 performance level is specified to be c ¼ 0:8. Solving the LMIs in (33)–(36), the feasible solutions are obtained as follows:
; 0:3053 2:3934 0:3700 0:2748 ¼ ; 0:2748 2:7766
X11 ¼ X22
0:0828
0:3053
A1 ¼ 0:5893;
B1 ¼ 2:1055;
B2 ¼ 1:5002;
C2 ¼ 0:3438;
X12 ¼ X23
0:3434
0:4658
; 0:2420 2:9302 0:3779 0:1114 ¼ ; 0:0866 1:5939
C1 ¼ 0:4135;
X13 ¼
L2 ¼ 1:2722;
0:0530 1:5879 0:0839 0:0161 ¼ ; 0:0161 0:6586
;
X33
L1 ¼ 0:8666;
K1 ¼ 2:8736;
K2 ¼ 2:9178;
0:1393
0:3522
X ¼ 1:9908;
A2 ¼ 3:0182; Y ¼ 0:6525:
Therefore, by Theorem 2, the robust H1 filtering problem for system (43) is solvable. In order to construct the desired H1 fuzzy filter, choose
S ¼ 1:4321;
W ¼ 1:4321;
it can be verified that S and W satisfy the equality (38); thus by (37) the fuzzy filter gain matrices are given as follows:
Af 1 ¼ 0:4403;
Bf 1 ¼ 1:5733;
C f 1 ¼ 0:3090;
K f 1 ¼ 2:0065;
Lf 1 ¼ 0:9274;
Af 2 ¼ 0:3652;
Bf 2 ¼ 1:1210;
C f 2 ¼ 0:2569;
K f 2 ¼ 2:0374;
Lf 2 ¼ 1:3614:
0.5 0.45 0.4 0.35
y(t)
0.3 0.25 0.2 0.15 0.1 0.05 0
0
5
10
15
Time (s)
20
25
30
Fig. 3. Measured output yðtÞ.
0
zf(t)
−0.05
−0.1
−0.15
−0.2
−0.25 0
5
10
15
Time (s)
20
Fig. 4. Response of the error signal zf ðtÞ.
25
30
3709
J. Yang et al. / Information Sciences 179 (2009) 3697–3710 1
0.95
0.9
0.85
0.8
0
2
4
Time (s)
6
8
10
Fig. 5. H1 performance level c.
The simulation results of state responses of the plant and filter are given in Fig. 2, where the initial condition is zero, and the 1 exogenous disturbance input is wðtÞ ¼ 1þt 2 ðt P 0Þ, which belongs to L2 ½0; 1Þ; Figs. 3, 4 show the measured output yðtÞ and response of the error signal zf ðtÞ, respectively; the H1 performance level c is given in Fig. 5. From the simulation results, it follows that the designed H1 fuzzy filter meets the specified requirements. 5. Conclusion In this paper, the problem of robust H1 filtering for a class of uncertain T–S fuzzy neutral systems with time-varying delay and linear fractional parametric uncertainties has been investigated. By means of the Lyapunov–Krasovskii functional, the Barbalat lemma and the LMI approach, delay-dependent sufficient conditions for the solvability of the robust H1 filtering problem have been established, and the desired H1 fuzzy filter is readily constructed by solving a set of LMIs via standard numerical algorithm without requiring any tuning of parameters. Finally, the H1 filter design problem for the nonlinear distributed network (long line with tunnel diode) is considered in a simulation example to demonstrate the effectiveness and applicability of the proposed method. Acknowledgements The authors would like to thank the Editor in Chief and anonymous reviewers for their helpful and insightful comments for further improving the quality of this work. References [1] S.R. Barone, A new approach to some nonlinear fluid dynamics problems, Phys. Lett. A 70 (3) (1979) 260–262. [2] M. Basin, E. Sanchez, R. Martinez-Zuniga, Optimal linear filtering for systems with multiple state and observation delays, Int. J. Innov. Comput. Inform. Contr. 3 (5) (2007) 1309–1320. [3] S. Boyd, L. El Ghaoui, E. Feron, V. Balakrishnan, Linear Matrix Inequalities in System and Control Theory, SIAM Studies in Applied Mathematics, SIAM, Philadelphia, PA, 1994. [4] S.G. Cao, N.W. Rees, G. Feng, Analysis and design of a class of continuous time fuzzy control systems, Int. J. Contr. 64 (1996) 1069–1087. [5] Y.Y. Cao, P.M. Frank, Analysis and synthesis of nonlinear time-delay systems via fuzzy control approach, IEEE Trans. Fuzzy Syst. 8 (2000) 200–211. [6] Y.Y. Cao, Z.L. Lin, Y. Shamash, Set invariance analysis and gain-scheduling control for LPV systems subject to actuator saturation, Syst. Cont. Lett. 46 (2) (2002) 137–151. [7] B.S. Chen, C.L. Tsai, Y.F. Chen, Mixed H2 =H1 filtering design in multirate transmultiplexer systems: LMI approach, IEEE Trans. Signal Process. 49 (11) (2001) 2693–2701. [8] W.H. Chen, W.X. Zheng, Delay-dependent robust stabilization for uncertain neutral systems with distributed delays, Automatica 43 (1) (2007) 95–104. [9] E.N. Chukwu, Mathematical controllability theory of capital growth of nations, Appl. Math. Comput. 52 (2-3) (1992) 317–344. [10] C. Gong, B. Su, Robust H1 filtering of convex polyhedral uncertain tim-delay fuzzy systems, Int. J. Innov. Comput. Inform. Contr. 4 (4) (2008) 793–802. [11] M.J. Grimble, A.E. Sayed, Solution of the H1 optimal linear filtering problem for discrete-time systems, IEEE Trans. Acoust. Speech Signal Process. 38 (7) (1990) 1092–1104. [12] J.K. Hale, Theory of Functional Differential Equations, Springer, New York, 1977. [13] Q.L. Han, A descriptor system approach to robust stability of uncertain neutral systems with discrete and distributed delays, Automatica 40 (10) (2004) 1791–1796. [14] J. Harband, The existence of monotonic solutions of a nonlinear car-following equation, J. Math. Anal. Appl. 57 (2) (1977) 257–272. [15] Y. He, M. Wu, J.H. She, G.P. Liu, Delay-dependent robust stability criteria for uncertain neutral systems with mixed delays, Syst. Contr. Lett. 51 (1) (2004) 57–65. [16] J.C. Hung, B.S. Chen, Genetic algorithm approach to speech fixed-order mixed H2 =H1 optimal deconvolution filter designs, IEEE Trans. Signal Process. 48 (12) (2000) 3451–3461. [17] S.H. Jin, J.B. Park, Robust H1 filtering for polytopic uncertain systems via convex optimisation, IEE Proc. Contr. Theory Appl. 148 (2001) 55–59. [18] E. Kim, H. Lee, New approaches to relaxed quadratic stability condition of fuzzy control systems, IEEE Trans. Fuzzy Syst. 8 (5) (2000) 523–534. [19] V. Kolmanovskii, A. Myshkis, Applied Theory of Functional Differential Equations, Kluwer Academic Publishers, Netherlands, 1992.
3710
J. Yang et al. / Information Sciences 179 (2009) 3697–3710
[20] V. Kolmanovskii, A. Myshkis, Introduction to the Theory and Applications of Functional Differential Equations, Kluwer Academic Publishers, Netherlands, 1999. [21] Y. Li, S. Xu, Robust stabilization and H1 control for uncertain fuzzy neutral systems with mixed time delays, Fuzzy Sets Syst. 159 (20) (2008) 2730– 2748. [22] X. Liu, Q.L. Zhang, New approaches to controller designs based on fuzzy observers for T–S fuzzy systems via LMI, Automatica 39 (2003) 1571–1582. [23] W.M. McEneaney, Robust H1 filtering for nonlinear systems, Syst. Contr. Lett. 33 (1998) 315–325. [24] K.M. Nagpal, P.P. Khargonekar, Filtering and smoothing in an H1 setting, IEEE Trans. Auto. Contr. 36 (1991) 152–166. [25] S.K. Nguang, P. Shi, Fuzzy H1 output feedback control of nonlinear systems under sampled measurements, Automatica 39 (12) (2003) 2169–2174. [26] J.H. Park, Delay-dependent guaranteed cost stabilization criterion for neutral delay-differential systems: matrix inequality approach, Comput. Math. Appl. 47 (10-11) (2004) 1507–1515. [27] V.P. Rubanik, Oscillations of quasilinear systems having delay, Izd. Nauka, Moscow 1969 [in Russian]. [28] A. Sala, C. Ariño, Asymptotically necessary and sufficient conditions for stability and performance in fuzzy control: applications of Polya’s theorem, Fuzzy Sets Syst. 158 (2007) 2671–2686. [29] X. Shen, L. Deng, A dynamic system approach to speech enhancement using the H1 filtering algorithm, IEEE Trans. Speech Audio Process. 7 (4) (1999) 391–399. [30] X.N. Song, S.Y. Xu, H. Shen, Robust H1 control for uncertain fuzzy systems with distributed delays via output feedback controllers, Inform. Sci. 178 (22) (2008) 4341–4356. [31] C.E. de Souza, L. Xie, Y. Wang, H1 filtering for a class of uncertain nonlinear systems, Syst. Cont. Lett. 20 (1993) 419–426. [32] C.E. de Souza, R.M. Palhares, P.L.D. Peres, Robust H1 filter design for uncertain linear systems with multiple time-varying state delays, IEEE Trans. Signal Process. 49 (2001) 569–576. [33] X. Sun, J. Zhao, W. Wang, Two design schemes for robust adaptive control of a class of linear uncertain neutral delay systems, Int. J. Innov. Comput. Inform. Contr. 3 (2) (2007) 385–396. [34] T. Takagi, M. Sugeno, Fuzzy identification of systems and its applications to modeling and control, IEEE Trans. Syst. Man Cybern. SMC-15 (1) (1985) 116–132. [35] K. Tanaka, T. Ikeda, H.O. Wang, Robust stabilization of a class of uncertain nonlinear systems via fuzzy control: quadratic stabilizability, H1 control theory, and linear matrix inequalities, IEEE Trans. Fuzzy Syst. 4 (1) (1996) 1–13. [36] K. Tanaka, H.O. Wang, Fuzzy Control Systems Design and Analysis, J. Wiley, New York, 2001. [37] H.D. Tuan, P. Apkarian, T. Narikiyo, Y. Yamamoto, Parameterized linear matrix inequality techniques in fuzzy control system design, IEEE Trans. Fuzzy Syst. 9 (2) (2001) 324–332. [38] M. Wang, B. Chen, K.F. Liu, X.P. Liu, S.Y. Zhang, Adaptive fuzzy tracking control of nonlinear time-delay systems with unknown virtual control coefficients, Inform. Sci. 178 (22) (2008) 4326–4340. [39] L. Xie, Output feedback H1 control of systems with parameter uncertainty, Int. J. Contr. 63 (4) (1996) 741–750. [40] S. Xu, P. Van Dooren, Robust H1 filtering for a class of non-linear systems with state delay and parameter uncertainty, Int. J. Contr. 75 (2002) 766–774. [41] S. Xu, J. Lam, B. Chen, Robust H1 control for uncertain fuzzy neutral delay systems, Eur. J. Contr. 10 (2004) 365–380. [42] S.Y. Xu, J. Lam, Exponential H1 filter design for uncertain Takagi–Sugeno fuzzy systems with time delay, Eng. Appl. Artif. Intell. 17 (2004) 645–659. [43] J. Yang, S.M. Zhong, L.L. Xiong, A descriptor system approach to non-fragile H1 control for uncertain fuzzy neutral systems, Fuzzy Sets Syst. 160 (2009) 423–438. [44] J. Yoneyama, Generalized conditions for H1 disturbance attenuation of fuzzy time-delay systems, IEEE Int. Conf. Syst. Man Cybern. 2 (2005) 1736– 1741. [45] J. Yoneyama, Robust H1 filtering for fuzzy time-delay systems, IEEE Int. Fuzzy Syst. Conf. (2007) 1–6. [46] J. Yoneyama, Robust stability and stabilizing controller design of fuzzy systems with discrete and distributed delays, Inform. Sci. 178 (8) (2008) 1935– 1947. [47] B.Y. Zhang, S.S. Zhou, T. Li, A new approach to robust and non-fragile H1 control for uncertain fuzzy systems, Inform. Sci. 177 (2007) 5118–5133. [48] S. Zhou, G. Feng, J. Lam, S. Xu, Robust H1 control for discrete-time fuzzy systems via basis-dependent Lyapunov functions, Inform. Sci. 174 (2005) 197– 217. [49] S. Zhou, J. Lam, A. Xue, H1 filtering of discrete-time fuzzy systems via basis-dependent Lyapunov function approach, Fuzzy Sets Syst. 158 (2) (2007) 180–193.