exponential image filter design for uncertain takagi–sugeno fuzzy systems with time delay
TRANSCRIPT
-
8/10/2019 Exponential Image Filter Design for Uncertain TakagiSugeno Fuzzy Systems With Time Delay
1/15
Engineering Applications of Artificial Intelligence 17 (2004) 645659
Exponential H1 filter design for uncertain TakagiSugeno fuzzysystems with time delay$
Shengyuan Xua,, James Lamb
aDepartment of Automation, Nanjing University of Science and Technology, Nanjing 210094, P.R. ChinabDepartment of Mechanical Engineering, University of Hong Kong, Pokfulam Road, Hong Kong
Received 23 September 2003; received in revised form 10 August 2004; accepted 10 August 2004
Available online 21 September 2004
Abstract
This paper considers the problem of robustH1filtering for state-space TakagiSugeno fuzzy systems with time delays and norm-
bounded parameter uncertainties. The problem we address is the design of an exponentially stable filter ensuring an exponential
stability and a prescribed H1 performance of the filtering error system for all admissible uncertainties. In terms of certain linear
matrix inequalities (LMIs), sufficient conditions for the solvability of this problem are obtained. When these LMIs are feasible, an
explicit expression of a desired H1filter is also given. Finally, a simulation example is provided to demonstrate the effectiveness and
applicability of the proposed design approach.
r 2004 Elsevier Ltd. All rights reserved.
Keywords: Fuzzy systems;H1 Filtering; Linear matrix inequality; Robust filtering; TakagiSugeno models; Time-delay systems; Uncertain systems
1. Introduction
In the past decades, there has been an increasing interest in the study of fuzzy systems in the TakagiSugeno (TS)
model since TS fuzzy models provide an effective way in representing complex nonlinear systems (Takagi and Sugeno
(1985)). In the fuzzy control setting, the TS fuzzy-model-based control technique has become popular recently. TS
fuzzy systems are defined by a set of IFTHEN fuzzy rules whose consequents are deterministic nonlinear dynamical
systems. Many control and filtering issues related to TS fuzzy systems have been studied in the literature. For
instance, the stability analysis and stabilization synthesis problems for these fuzzy systems were investigated in Cao et
al. (1997),Johansson et al. (1999),Tanaka and Sugeno (1992), and the references therein. In the case when parameter
uncertainty appears in a TS fuzzy system, the robust stability problem was addressed in Lee et al. (2001), where
several sufficient conditions for the robust stability were obtained in terms of linear matrix inequalities (LMIs); based
on these, the robust stabilization problem was solved and robust controllers were designed by the so-called parallel
distributed compensation (PDC) scheme (Tanaka et al., 1996). The observer design for TS fuzzy systems can be foundinKim (2002)andTanaka et al. (1998)and the H1 control problem was investigated inLee et al. (2001).
Since time delay is often a source of instability and encountered in various engineering systems such as chemical
processes, long transmission lines in pneumatic systems (Hale, 1977), the study of time-delay systems has received
much attention during the past years and various topics on time-delay systems have been addressed ( Hale, 1977;Jeung
ARTICLE IN PRESS
www.elsevier.com/locate/engappai
0952-1976/$ - see front matter r 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engappai.2004.08.012
$This work is supported by RGC HKU 7103/01P, the Foundation for the Author of National Excellent Doctoral Dissertation of P.R. China
under Grant 200240, the National Natural Science Foundation of P.R. China under Grant 60304001 and the Fok Ying Tung Education Foundation
under Grant 91061.Corresponding author. Tel.: +852-28598982; fax: +852-28585415.
E-mail address: [email protected] (S. Xu).
http://www.elsevier.com/locate/engappaihttp://www.elsevier.com/locate/engappai -
8/10/2019 Exponential Image Filter Design for Uncertain TakagiSugeno Fuzzy Systems With Time Delay
2/15
et al., 1998;Xu et al., 2001a, b). Very recently, the TS fuzzy systems with time delays were introduced in Cao and
Frank (2000), in which some stability results were developed in both the continuous and discrete cases, and stabilizing
controllers and observers for such time-delay fuzzy systems were also designed by using the LMI approach. The robust
H1 control problem for time-delay fuzzy systems were considered in Lee et al. (2000), in which an LMI approach to
designing output feedback controllers were developed.
On the other hand, the problem ofH1filtering, which is concerned with the design of an estimator ensuring that the
L2-induced gain from the noise signal to the estimation error is less than a prescribed level, has received muchattention in the past years. Compared with the conventional Kalman filtering, the H1 filter technique has several
advantages. First, the noise sources in the H1 filtering setting are arbitrary signals with bounded energy or average
power, and no exact statistics are required to be known ( Nagpal and Khargonekar, 1991). Second, the H1 filter has
been shown to be much more robust to parameter uncertainty in a control system. These advantages render the H1filtering approach very appropriate to some practical applications. When parameter uncertainty arises in a system
model, the robust H1 filtering problem has been studied, and a great number of results on this topic have been
reported (de Souza et al., 1993;Jin and Park, 2001;Xie et al., 1996;Xu and Chen, 2003). In the case when parameter
uncertainty and time delays appear simultaneously in a system model, the robust H1filtering problem was dealt with
inde Souza et al. (2001)andXu and Van Dooren (2002)via LMI approach, respectively. The corresponding results for
uncertain discrete delay systems can be found in Xu (2002). However, it is noted that for TS fuzzy systems with
parameter uncertainty and time delays, the robust H1filtering problem has not been addressed, which is still open and
remains unsolved; this motivates the present study.
In this paper, we deal with the problem of exponential H1 filter design for uncertain TS fuzzy systems with time
delays. The parameter uncertainties are assumed to be time-varying but norm-bounded. The problem we address is to
design an exponentially stable filter which guarantees that the error system is exponentially stable and the L2-induced
gain from the noise signal to the estimation error is below a prescribed level. It is worth pointing out that the advantage
of a filter with exponential stability in comparison with that with asymptotic stability lies in that the former provides
fast convergence and desirable accuracy in terms of reasonable error covariance of the filtering process (Reif and
Unbehauen, 1999). Sufficient conditions for the solvability of this problem are obtained in terms of certain LMIs. A
desired filter can be constructed through a convex optimization problem, which can be solved by using standard
numerical algorithms, and no tuning of parameters is required (Boyd et al., 1994). Finally, a simulation example is
provided to demonstrate the effectiveness of the proposed approach.
Notation: Throughout this paper, for real symmetric matrices Xand Y; the notation XXY (respectively, X4Y)
means that the matrix XY is positive semi-definite (respectively, positive definite). I is an identity matrix with
appropriate dimension. The superscript T represents the transpose. L20;1is the space of square-integrable vectorfunctions over0;1:The notationj j refers to the Euclidean vector norm, while k k2 stands for the usual L20;1
norm. We use lminandlmax; respectively, to denote the minimum and maximum eigenvalue of a seal symmetric
matrix. Matrices, if not explicitly stated, are assumed to have compatible dimensions.
2. Definition and problem formulation
The continuous TakagiSugeno fuzzy dynamic model with parameter uncertainties and time delays is described by
the following fuzzy IFTHEN rules (Cao and Frank, 2000;Lee et al., 2000):
Plant Rule i: IF s1t is mi1 and s2t is m i2 and . . . and sgt is mig THEN
_xt Ai DAitxt Adi DAditxt t Diot; (1)
yt Ci DCitxt Cdi DCditxtt Eiot; (2)
zt Lixt; (3)
xt ft; t2 t; 0; i 1; 2; :. . .; r; (4)
where mij is the fuzzy set and r is the number of IFTHEN rules; xt 2 Rn is the state; yt 2 Rm is the
measured output; zt 2 Rp is the signal to be estimated; ot 2 Rq is the noise signal which is assumed to be an
arbitrary signal in L20;1;t40 is a scalar representing the time delay of the fuzzy system; s1t; s2t; ; sgtare the
premise variables. Ai;Adi;Ci;Cdi;Di;EiandLiare known real constant matrices; DAit; DAdit; DCitand DCdit
are real-valued unknown matrices representing time-varying parameter uncertainties, and are assumed to be
ARTICLE IN PRESS
S. Xu, J. Lam / Engineering Applications of Artificial Intelligence 17 (2004) 645659646
-
8/10/2019 Exponential Image Filter Design for Uncertain TakagiSugeno Fuzzy Systems With Time Delay
3/15
of the form
DAit DAdit
DCit DCdit
M1i
M2i
FitN1i N2i; i 1; 2; :. . .; r; (5)
where M1i; M2i; N1i and N2i are known real constant matrices and Fi: N ! Rl1l2 are unknown time-varying
matrix function satisfying
FitTFitpI: (6)
The parameter uncertainties DAit; DAdit; DCit and DCdit are said to be admissible if both (5) and (6) hold.
By using a center-average defuzzier, product inference, and singleton fuzzifier, the dynamic fuzzy model in (1)(4)
can be represented by
_xt Xri1
histfAi DAitxt Adi DAditxt t Diotg; (7)
yt Xri1
histfCi DCitxt Cdi DCditxtt Eiotg; (8)
zt Xri1
histLixt; (9)
xt ft; t2 t; 0; (10)
where
hist $istPr
j1 $jst; $ist
Ygj1
mijsjt;
st s1t s2t sgt;
in which mijsjt is the grade of membership ofsjt inmij: Then, it can be seen that
$istX0; i 1; 2; :. . .; r;Xrj1
$jst40
for all t. Therefore, for all t,
histX0; i 1; 2; :. . .; r; (11)
Xrj1
hjst 1: (12)
Throughout the paper, we adopt the following exponential stability concept.
Definition 1. The fuzzy system (7) is said to be exponentially stable if there exist constants c40 and d40 such that
jxtjpc suptpyp0
jfyjedt;
whenot 0:
Similar to the fuzzy observer design (Cao and Frank, 2000), we now consider the following fuzzy filter for the
estimate ofzt:
Filter Rule i: IF s1t is m i1 ands2t is m i2
and . . . and sgt is mig THEN
_xt Afixt Bfiyt; (13)
zt Lfixt; i 1; 2; :. . .; r; (14)
ARTICLE IN PRESS
S. Xu, J. Lam / Engineering Applications of Artificial Intelligence 17 (2004) 645659 647
-
8/10/2019 Exponential Image Filter Design for Uncertain TakagiSugeno Fuzzy Systems With Time Delay
4/15
where xt 2 Rn and zt 2 Rp; Afi; Bfi and Cfi are matrices to be determined. Then, the overall fuzzy filter can be
inferred by
_xt Xri1
histAfixt Bfiyt; (15)
zt Xri1
histLfixt: (16)
Let
et xtT xtTT; ~zt zt zt: (17)
Then, the filtering error dynamics from systems (7)(9) and (15) and (16) can be described by
~S: _et Xri1
Xrj1
histhjst ~Aijtet ~AdijtHett ~Dijot; (18)
~zt
Xr
i1 Xr
j1
hist hjst~Lijet; (19)
where
~Aijt ~Aij D ~Aijt; ~Adijt ~Adij D~Adijt; (20)
~AijAj 0
BfiCj Afi
; ~Adij
Adj
BfiCdj
; ~Dij
Dj
BfiEj
; (21)
H I 0; ~Lij Lj Lfi; (22)
D ~Aijt DAjt 0
BfiDCjt 0
; D ~Adij
DAdjt
BfiDCdjt
: (23)
The robustH1filtering problem to be addressed in this paper is formulated as follows: given the uncertain time-delayTS fuzzy system (7)(9) and a prescribed level of noise attenuationg40;determine an exponentially stable filter in the
form of (13) and (14) such that the following requirements are satisfied:
(I) The filtering error system ~S is exponentially stable;
(II) Under zero initial conditions, ~S satisfiesk~zk2ogkok2 (24)
for any nonzero o 2 L20; 1 and all admissible uncertainties.
3. Main results
In this section, we will develop an LMI approach to solve the robustH1 filtering problem for uncertain time-delayfuzzy systems formulated in the previous section. We first give the following results which will be used in the proof of
our main results.
Lemma 1 (Xie and de Souza, 1992). Given any matrices X; Y andZ with appropriate dimensions such that Y40:Then
we have
XTZZTXpXTYXZTY1Z:
Lemma 2 (Wang et al., 1992). Let D; H; and Ft be real matrices of appropriate dimensions with Ft satisfying
FtTFtpI: Then, for any scalar e40 and vectors x and y with appropriate dimensions, we have
xTDFtHyyTHTFtTDTxp1xTDDTx yTHTHy:
The following theorem will play a key role in the design of desired fuzzy H1 filters.
ARTICLE IN PRESS
S. Xu, J. Lam / Engineering Applications of Artificial Intelligence 17 (2004) 645659648
-
8/10/2019 Exponential Image Filter Design for Uncertain TakagiSugeno Fuzzy Systems With Time Delay
5/15
Theorem 1. The uncertain system( ~S) is exponentially stable and(24) is satisfied if there exist matrices P40;and Q40;
and scalars eij40; 1pipjpr; such that the following LMIs hold for 1pipjpr:
Oij12~Lij ~Lji
T~Lij ~Lji 2HTQH Odij P ~Dij ~Dji P ~Mij P ~Mji
OTdij Xij 2Q 0 0 0
~Dij ~DjiTP 0 2g2I 0 0
~MT
ijP 0 0 ijI 0
~MT
jiP 0 0 0 ijI
2666666664
3777777775o0; (25)
where
OijP ~Aij ~Aji ~Aij ~AjiTP ij ~N
T
1i~N1i ~N
T
1j~N1j; (26)
OdijP~Adij ~Adji ij ~NT
1iN2i ~N
T
1jN2j; (27)
Xij ijNT2iN2i N
T2jN2j; (28)
~N1i N1i 0; ~Mij
M1j
BfiM2j
: (29)
Proof. Under the conditions of the theorem, we first show the exponential stability of system ~S: To this end, we
consider (18) with ot 0; that is,
_et Xri1
Xrj1
histhjst ~Aijtet ~AdijtHett: (30)
For this system, we choose a LyapunovKrasovskii functional as follows:
Vet etTPet
Z t
tt
esTHTQHes ds; (31)
where
et etb; b2 t; 0:
Using the expressions in (20) and taking the time-derivative ofV et along the solution of (30) give
_Vet 2Xri1
Xrj1
histhjstetTP~Aijtet ~AdijtHet t
etTHTQHet et tTHTQHett
1
2
Xri1
Xrj1
histhjstfetTP~Aijt ~Ajitet ~Adijt ~Adjitxtt
~Aijt
~Ajitet
~Adijt
~Adjitxt t
T
PetgetTHTQHet xt tTQxtt
1
2
Xri1
Xrj1
histhjstfetTP~Aij ~Ajiet ~Adij ~Adjixtt
~Aij ~Ajiet ~Adij ~AdjixttTPet 2etTHTQHet 2xt tTQxtt
etTPD~Aijt D ~Ajitet D~Adijt D~Adjitxtt
D~Aijt D ~Ajitet D ~Adijt D ~AdjitxttTPetg; 32
where the relationship
xtt Hett; (33)
ARTICLE IN PRESS
S. Xu, J. Lam / Engineering Applications of Artificial Intelligence 17 (2004) 645659 649
-
8/10/2019 Exponential Image Filter Design for Uncertain TakagiSugeno Fuzzy Systems With Time Delay
6/15
is used. By (5), it is easy to see that
D ~Aijt D ~Adijt ~MijFjt ~N1j N2j: (34)
Set
Mij ~Mij ~Mji;
N1ij~N1j
~N1i
" #; N2ij
N2j
N2i
" #;
Fijt Fjt 0
0 Fit
" #:
Then
FijtTFijtpI (35)
By considering (34) and (35) and applying Lemma 2, we have that for 1pipjpr;
etTP D ~Aijt D~Ajitet D ~Adijt D ~Adjitxt t
D ~
Aijt D~
Ajitet D ~
Adijt D ~
Adjitxtt
T
PetetTPMijFijt N1ijet N2ijxtt N1ijet N2ijxtt
TFijtTM
T
ijPet
p1ij et
TPMijMT
ijPet ijN1ijet N2ijxtt
T N1ijet N2ijxtt: 36
This together with (32) implies
_Vetp1
2
Xri1
hist2xtTGiixt
Xri;j1;ioj
histhjstxtTGijxt; (37)
where
xt et
xkt ; (38)
GijOij
1ij P
MijMT
ijP2HTQH Odij
OTdij Xij 2Q
24
35o0; 1pipjpr: 39
On the other hand, from (25), it is easy to see that for 1pipjpr;
Oij2HTQH Odij P ~Mij P ~Mji
OTdij Xij 2Q 0 0
~MT
ijP 0 ijI 0
~MT
jiP 0 0 ijI
2666664
3777775
o0;
which, by the Schur complement formula, implies that for 1pipjpr;
Gijo0:
This together with (37) gives
_Vetpajetj2
; (40)
where
a 12
minflminGij; 1pipjprg40:
Then, it can be shown that there always exists a scalar b40 such that
blmaxP abtlmaxHTQHebt 0: (41)
ARTICLE IN PRESS
S. Xu, J. Lam / Engineering Applications of Artificial Intelligence 17 (2004) 645659650
-
8/10/2019 Exponential Image Filter Design for Uncertain TakagiSugeno Fuzzy Systems With Time Delay
7/15
Now, by (31) and (40), we have
d
dtebtVet e
btbVt _Vet
pebt blmaxHTQH
Z ttt
jesj2 ds blmaxP ajetj2
:
Integrating both sides from 0 to T40 gives
ebTVeT Ve0 pblmaxHTQH
Z T0
ebt dt
Z ttt
jesj2 ds blmaxP a
Z T0
ebtjetj2 dt: (42)
Since Z T0
ebt dt
Z ttt
es 2 ds Z 0
t
jesj2 ds
Z st0
ebt dt
Z Tt0
jesj2 ds
Z sts
ebt dt
Z TTt
jesj2 ds
Z Ts
ebt dt
p
Z 0t
tebstjxsj2 ds
Z Tt0
tetstjxsj2 ds
Z TTt
tebstjxsj2 ds
tebt Z T
t
ebsjesj2 ds;
we have
ebTVeT pVe0 blmaxP abtlmaxHTQHebt
Z T0
ebtjetj2 dtpk suptpyp0
jfyj2; (43)
where
k max lmaxP; lmaxHTQH
:
Taking into account that
VeTXlminPjeTj2
:
This together with (43) implies that for any T40;
jeTjp
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffik
lminP
s sup
tpyp0
jfyjebT:
Therefore, by Definition 1, we have that system ( ~S) is exponentially stable.
Next, we show that for any nonzero o2 L20; 1; the uncertain system in ( ~S) satisfies (24) under the zero initial
condition. To this end, we introduce
JT
Z T0
~ztT ~zt g2otTot
dt; (44)
where the scalarT40:Noting the zero initial condition, it can be shown that for any nonzero o 2 L20;1and T40;
JT Z T
0
~ztT ~zt g2otTot _Vet dtVeTp
Z T0
~ztT ~zt g2otTot _Vet
dt; 45
where Vet is defined in (31). By some calculations, it can be verified that
~ztT ~zt g2otTot _Vet
Xri1
Xrj1
Xru1
Xrv1
histhjst husthvstetT ~L
T
ij~Luvet g
2otTot
2Xri1
Xrj1
histhjstetT P ~Aijtet ~AdijtHet t ~Dijot
etTHTQHet xttTQxt t: 46
ARTICLE IN PRESS
S. Xu, J. Lam / Engineering Applications of Artificial Intelligence 17 (2004) 645659 651
-
8/10/2019 Exponential Image Filter Design for Uncertain TakagiSugeno Fuzzy Systems With Time Delay
8/15
By (36), it can be seen that
2Xri1
Xrj1
histhjstetTP~Aijtet ~AdijtHet t ~Dijot
1
2Xr
i1X
r
j1
histhjstfetTP~Aijt ~Ajitet ~Adijt ~Adjitxtt ~Dij ~Djiot
~Aijt ~Ajitet ~Adijt ~Adjitxtt ~Dij ~DjiotTPetg
1
2
Xri1
Xrj1
histhjstfetTP~Aij ~Ajiet ~Adij ~Adjixt t ~Dij ~Djiot
~Aij ~Ajiet ~Adij ~Adjixt t ~Dij ~DjiotTPet
etTPD~Aijt D~Ajitet D ~Adijt D~Adjitxt t D ~Aijt D ~Ajitet
D~Adijt D ~AdjitxttTPetg
p1
2
Xri1
Xrj1
histhjstZtT ~OijZt; 47
where
Zt etT xt tT otT T
;
~Oij
Oij 1ij P
MijMT
ijP Odij P~Dij ~Dji
OTdij Xij 0
~Dij ~DjiTP 0 0
26664
37775:
Using Lemma 1, we have
Xr
i1 Xr
j1 Xr
u1 Xr
v1
histhjst husthvstetT ~L
T
ij~Luvet
1
4
Xri1
Xrj1
Xru1
Xrv1
histhjst husthvstetT~Lij ~Lji
T~Luv ~Lvuet
p1
4
Xri1
Xrj1
histhjstetT ~Lij ~Lji
T~Lij ~Ljiet: 48
Then, it follows from (46)(48) that
~ztT ~zt g2otTot _Vet
p1
2
Xri1
hist2ZtT ~GiiZt
Xri;j1;ioj
histhjstZtT ~GijZt; 49
where
~Gij ~Oij
12~Lij ~Lji
T~Lij ~Lji 2HTQH 0 0
0 2Q 0
0 0 2g2I
264
375; 1pipjpr:
Applying the Schur complement formula to the LMI in (25) results in
~Gijo0; 1pipjpr:
This together with (45) and (49) givesJTo0 for anyT40;which impliesk ~zk2ogkok2 for any nonzero o 2 L20;1:
This completes the proof. &
Now, we are in a position to give the main result on the solvability of the robust H1 filtering problem.
ARTICLE IN PRESS
S. Xu, J. Lam / Engineering Applications of Artificial Intelligence 17 (2004) 645659652
-
8/10/2019 Exponential Image Filter Design for Uncertain TakagiSugeno Fuzzy Systems With Time Delay
9/15
Theorem 2. Consider the uncertain time-delay fuzzy systems (1)(4),and letg40be a given scalar.Then the robust H1filtering problem is solvable if there exist scalars ij40; 1pipjpr; and matrices X40; Y40; Q40; Yi; Fi and Ci;
1pipr; such that the following LMIs
J1ij H1ij H2ij H3ij H3ji H4ij
HT1ij J2ij 0 0 0 0
HT2ij 0 2g2I 0 0 0
HT3ij 0 0 ijI 0 0
HT3ji 0 0 0 ijI 0
HT4ij 0 0 0 0 2I
26666666666664
37777777777775o0; 1pipjpr 50
and
XY40; (51)
hold, where
J1ij
J11ij J11ji ijNT1iN1i N
T1jN1j
J12ij J12jiT ijN
T1iN1i N
T1jN1j
J12ij J12ji ijNT1iN1i NT1jN1j J13ij J13ji ijN
T1iN1i N
T1jN1j
" #;
J11ijYAj AT
jYQ;
J12ijXAj AT
jYFiCj Ci Q;
J13ijAT
jXXAj FiCj CT
jFTi Q;
J2ij ijNT2iN2i N
T2jN2j 2Q;
H1ijYAdi Adj ijN
T1iN2i N
T1jN2j
XAdi Adj FiCdj FjCdi ijNT1iN2i N
T1jN2j
" #;
H2ijYDi Dj
XDi Dj FiEj FjEi
;
H3ijYM1j
XM1j FiM2j
;
H4ijLi Lj
T Yi YjT
Li LjT
" #:
In this case, a desired exponentially stable fuzzy H1 filter is given in the form of (13) and (14) with parameters
as follows:AfiS
1CiY1WT; BfiS
1Fi; Lfi YiY1WT; (52)
where S and W are any nonsingular matrices satisfying
SWT I XY1: (53)
Proof. We first note that (51) implies thatI XY1 is nonsingular; therefore, there always exist nonsingular matrices
Sand Wsuch that (53) holds. Now, set
Y Y1; P1 Y I
WT 0
" #; P2
I X
0 ST
(54)
ARTICLE IN PRESS
S. Xu, J. Lam / Engineering Applications of Artificial Intelligence 17 (2004) 645659 653
-
8/10/2019 Exponential Image Filter Design for Uncertain TakagiSugeno Fuzzy Systems With Time Delay
10/15
and
P P2P11 : (55)
Then, by some calculations, we have
PX S
ST U ;
where
U W1 Y XY Y WT40: (56)
Noting
XSU1ST Y40: (57)
Then, we have that P40:Pre- and post-multiplying the first row and the first column of the LMI in (50) by diag Y; I;
we have that for 1pipjpr;
PT1Oij 2HTQHP1 P
T1Odij P
T1 P
~Dij ~Dji PT1 P
~Mij PT1 P
~Mji PT1
~Lij ~LjiT
OTdijP1 Xij2Q 0 0 0 0
~Dij ~DjiTPP1 0 2g
2I 0 0 0
~MT
ijPP1 0 0 ijI 0 0
~MT
jiPP1 0 0 0 ijI 0
~Lij ~Lji
P1 0 0 0 0 2I
2666666666664
3777777777775o0; (58)
whereO ij; Odijare given in (26)(29) and (22) with Pbeing replaced by P; and Afi; Bfi andLfiin (52). Now, pre- and
post-multiplying (58) by diagPT1 ; I; I; I; I; I and diagP11 ; I; I; I; I; I; respectively, result in
Oij2HTQH Odij P ~Dij ~Dji P ~Mij P ~Mji ~Lij ~Lji
T
OTdij Xij 2Q 0 0 0 0
~Dij ~DjiTP 0 2g2I 0 0 0
~MT
ijP 0 0 ijI 0 0
~MT
jiP 0 0 0 ijI 0
~Lij ~Lji 0 0 0 0 2I
266666666664
377777777775o0; 1pipjpr;
which, by the Schur complement formula, implies that for 1pipjpr;
Oij12~Lij ~Lji
T~Lij ~Lji 2HTQH Odij P ~Dij ~Dji P ~Mij P ~Mji
OTdij Xij 2Q 0 0 0
~Dij ~DjiTP 0 2g2I 0 0
~MTijP 0 0 ijI 0
~MT
jiP 0 0 0 ijI
2666666664
3777777775o0: (59)
By (59) and Theorem 1, the desired result follows immediately. &
Remark 1. Theorem 2 presents a sufficient condition for the solvability of the robust H1 filtering problem for
uncertain time-delay TS fuzzy systems. A desired H1 filter can be constructed by solving the LMIs in (50) and (51),
which can be implemented by resorting to some standard numerical algorithms, and no tuning of parameters will be
involved (Boyd et al., 1994).
Remark 2. It is noted that the LMI conditions provided in Theorem 1 are independent of the delay size. Therefore,
they can be applicable to the case when no prior knowledge about the size of the time delay is available.
ARTICLE IN PRESS
S. Xu, J. Lam / Engineering Applications of Artificial Intelligence 17 (2004) 645659654
-
8/10/2019 Exponential Image Filter Design for Uncertain TakagiSugeno Fuzzy Systems With Time Delay
11/15
4. Simulation example
In this section, we provide a simulation example to illustrate theH1filter design approach developed in the previous
section. The uncertain time-delay TS fuzzy system considered in this example is with two rules:
Plant Rule 1: IF x1t is m1 e:g: small THEN
_xt A1
D
A1txt Ad1D
Ad1txt1:
5 D1o
t;
yt C1 DC1txt Cd1 DCd1txt1:5 E1ot;
zt L1xt
and
Plant Rule 2: IF x1t is m2 e:g: big THEN
_xt A2 DA2txt Ad2 DAd2txt1:5 D2ot;
yt C2 DC2txt Cd2 DCd2txt1:5 E2ot;
zt L2xt;
where
A1
1:2 0:8
1 2" #
; Ad1
0:5 0:2
1 0" #
; D1
1
0:2" #
;
C1 1 0; Cd1 0:8 0:6; E1 0:3; L1 1 0:5;
A2 1:6 0:2
0:7 1
; Ad2
0:6 0:2
0:5 0:8
; D2
0:3
0:1
;
C2 0:5 0:6; Cd2 0:2 1; E2 0:6; L2 0:2 0:3:
The membership functions ofm1 and m2 are, respectively, described by
h1x1t
1 for x1o1;12 1
2x1 for jx1jp1;
0 for x141;
8>:
and
h2x1t
0 for x1o1;12 1
2x1 for jx1jp1;
1 for x141:
8>:
The parameter uncertainties DAit; DAdit; DCit and DCdit; i 1; 2; are assumed to satisfy (5) and (6) with
M11 0
0:5
; M21 0:8; N11 0 0:3; N21 0:2 0;
M12
0
0:3
; M22 0:6; N12 0:5 0; N22 0 0:2:
Then, the final outputs of the fuzzy systems are inferred as follows:
_xt X2i1
hix1tfAi DAitxt Adi DAditxtt Diotg; (60)
yt X2i1
hix1tfCi DCitxt Cdi DCditxt t Eiotg; (61)
zt
X2
i1
hix1tLixt: (62)
ARTICLE IN PRESS
S. Xu, J. Lam / Engineering Applications of Artificial Intelligence 17 (2004) 645659 655
-
8/10/2019 Exponential Image Filter Design for Uncertain TakagiSugeno Fuzzy Systems With Time Delay
12/15
The purpose of this example is to design an exponentially stable fuzzy filter such that the filtering error system is
exponentially stable and satisfies a prescribed H1 performance level. In this example, the H1 performance level is
specified to beg 0:8:For this purpose, we resort to the Matlab LMI Control Toolbox to solve the LMIs in (50) and
(51), and obtain the solution as follows:
X 1:8536 0:6848
0:6848 1:3291 ; Y 0:3917 0:0271
0:0271 0:3908 ; Q0:5094 0:1667
0:1667 0:2935 ;Y1 1:0017 0:4944; Y2 0:1968 0:2985;
F1 1:0774
0:9802
; F2
0:8742
0:9654
;C1
3:1581 0:5266
2:2911 2:1262
;
C2 3:0772 1:5642
2:2103 1:6686
; 11 0:5562; 12 2:2393; 11 0:5172:
Therefore, by Theorem 2, it is to see that the robust H1 filtering problem is solvable. To construct such a desired
exponentially stable fuzzy filter, we choose nonsingular matrices Sand Was
S1 0:5
0:8 1 ; W 2:0523 3:1620
0:2021 2:4576 : (63)
It can be verified that the matrices Sand Win (63) satisfy the equality in (53); thus, by Theorem 2, a desired fuzzyH1filter can be constructed as
Filter Rule 1: IF x1t is m1 THEN
_xt Af1xt Bf1yt;
zt Lf1xt
and
Filter Rule 2: IF x1t is m2 THEN
_xt Af2xt Bf2yt;
zt Lf2xt;
where
Af1 2:9870 0:7508
1:8256 1:5045
" #; Af2
1:1822 0:3462
1:1459 1:8463
" #;
Bf1 0:4195
1:3158
" #; Bf2
0:2797
1:1891
" #;
Lf1 0:6002 0:3954; Lf2 0:2737 0:3206:
The simulation results of the state response of the plant and the filter are given in Fig. 1, where the initial conditions are
1 0:8
T
and 0 0
T
; respectively, and the exogenous disturbance input ot is set as
ot 1
2t; tX0;
which belongs to L20;1: Fig. 2 shows the simulation results of the signal zt and zt: Figs. 3 and 4 show the
simulations of the measured output ytand the error response zt zt;respectively. From these simulation results,
it can be seen that the designed fuzzy H1 filter meet the specified requirements.
5. Conclusions
The problem of robust H1 filtering for continuous TS fuzzy systems with time delays and time-varying norm-
bounded parameter uncertainties has been studied. A sufficient condition for the existence of a full-order exponentially
ARTICLE IN PRESS
S. Xu, J. Lam / Engineering Applications of Artificial Intelligence 17 (2004) 645659656
-
8/10/2019 Exponential Image Filter Design for Uncertain TakagiSugeno Fuzzy Systems With Time Delay
13/15
ARTICLE IN PRESS
0 2 4 6 8 10 12 14 16 18 20
1.5
1
0.5
0
0.5
1
Fig. 1. State response ofxt () and xt :
0 2 4 6 8 10 12 14 16 18 20
0
0.5
0.5
1.5
1
Fig. 2. Response of the signalzt () and zt :
0 2 4 6 8 10 12 14 16 18 20
0
0.5
0.5
1
1.5
Fig. 3. Measured outputyt:
S. Xu, J. Lam / Engineering Applications of Artificial Intelligence 17 (2004) 645659 657
-
8/10/2019 Exponential Image Filter Design for Uncertain TakagiSugeno Fuzzy Systems With Time Delay
14/15
stable filter, which ensures that the error system is exponentially stable and the L2-induced gain from the noise signal
to the estimation error is below a prescribed level, has been obtained. An explicit expression of a desired H1filter has
been given. A simulation example has demonstrated the effectiveness of the proposed approach.
References
Boyd, S., El Ghaoui, L., Feron, E., Balakrishnan, V., 1994. Linear Matrix Inequalities in System and Control Theory. SIAM Studies in Applied
Mathematics. SIAM, Philadelphia, PA.
Cao, Y.-Y., Frank, P.M., 2000. Analysis and synthesis of nonlinear time-delay systems via fuzzy control approach. IEEE Transactions on Fuzzy
Systems 8, 200211.
Cao, S.-G., Rees, N.W., Feng, G., 1997. Analysis and design for a class of complex control systems Part II: fuzzy controller design. Automatica 33,
10291039.de Souza, C.E., Xie, L., Wang, Y., 1993. H1 filtering for a class of uncertain nonlinear systems. Systems & Control Letter 20, 419426.
de Souza, C.E., Palhares, R.M., Peres, P.L.D., 2001. Robust H1 filter design for uncertain linear systems with multiple time-varying state delays.
IEEE Transactions on Signal Processing 49, 569576.
Hale, J.K., 1977. Theory of Functional Differential Equations. Springer, New York.
Jeung, E.T., Kim, J.H., Park, H.B., 1998. H1-output feedback controller design for linear systems with time-varying delayed state. IEEE
Transactions on Automatic Control 43, 971974.
Jin, S.H., Park, J.B., 2001. Robust H1 filtering for polytopic uncertain systems via convex optimisation. IEE ProceedingsControl Theory
Applications 148, 5559.
Johansson, M., Rantzer, A., A rzen, K.-E., 1999. Piecewise quadratic stability of fuzzy systems. IEEE Transactions on Fuzzy Systems 7, 713721.
Kim, E., 2002. A fuzzy disturbance observer and its application to control. IEEE Transactions on Fuzzy Systems 10, 7784.
Lee, K.R., Kim, J.H., Jeung, E.T., Park, H.B., 2000. Output feedback robust H1 control of uncertain fuzzy dynamic systems with time-varying
delay. IEEE Transactions on Fuzzy Systems 6, 657664.
Lee, K.R., Jeung, E.T., Park, H.B., 2001. Robust fuzzy H1control for uncertain nonlinear systems via state feedback: an LMI approach. Fuzzy Sets
and Systems 120, 123134.
Nagpal, K.M., Khargonekar, P.P., 1991. Filtering and smoothing in an H1 setting. IEEE Transactions on Automatic Control 36, 152166.Reif, K., Unbehauen, R., 1999. The extended Kalman filter as an exponential observer for nonlinear systems. IEEE Transactions on Signal
Processing 47, 23242328.
Takagi, T., Sugeno, M., 1985. Fuzzy identification of systems and its applications to modelling and control. IEEE Transactions on Systems Man and
Cybernetics 15, 116132.
Tanaka, K., Sugeno, M., 1992. Stability analysis and design of fuzzy control systems. Fuzzy Sets and Systems 45, 135156.
Tanaka, K., Ikeda, T., Wang, H.O., 1996. Robust stabilization of a class of uncertain nonlinear systems via fuzzy control: quadratic stabilizability,
H1 control theory, and linear matrix inequalities. IEEE Transactions on Fuzzy Systems 4, 113.
Tanaka, K., Ikeda, T., Wang, H.O., 1998. Fuzzy regulators and fuzzy observers: relaxed stability conditions and LMI-based designs. IEEE
Transactions on Fuzzy Systems 6, 250264.
Wang, Y., Xie, L., de Souza, C.E., 1992. Robust control of a class of uncertain nonlinear systems. Systems & Control Letters 19, 139149.
Xie, L., de Souza, C.E., 1992. RobustH1control for linear systems with norm-bounded time-varying uncertainty. IEEE Transactions on Automatic
Control 37, 11881191.
Xie, L., de Souza, C.E., Wang, Y., 1996. Robust filtering for a class of discrete-time uncertain nonlinear systems: an H1 approach. International
Journal of Robust and Nonlinear Control 6, 297312.
ARTICLE IN PRESS
0 2 4 6 8 10 12 14 16 18 201
0.8
0.6
0.4
0.2
0
0.2
Fig. 4. Error response ofzt zt:
S. Xu, J. Lam / Engineering Applications of Artificial Intelligence 17 (2004) 645659658
-
8/10/2019 Exponential Image Filter Design for Uncertain TakagiSugeno Fuzzy Systems With Time Delay
15/15
Xu, S., 2002. RobustH1 filtering for a class of discrete-time uncertain nonlinear systems with state delay. IEEE Transactions on Circuits Systems I
49, 18531859.
Xu, S., Chen, T., 2003. Robust H1 filtering for uncertain impulsive stochastic systems under sampled measurements. Automatica 39, 509516.
Xu, S., Van Dooren, P., 2002. RobustH1filtering for a class of non-linear systems with state delay and parameter uncertainty. International Journal
of Control 75, 766774.
Xu, S., Lam, J., Yang, C., 2001a. H1 and positive real control for linear neutral delay systems. IEEE Transactions on Automatic Control 46,
13211326.
Xu, S., Lam, J., Yang, C., 2001b. Quadratic stability and stabilization of uncertain linear discrete-time systems with state delay. Systems & ControlLetters 43, 7784.
ARTICLE IN PRESS
S. Xu, J. Lam / Engineering Applications of Artificial Intelligence 17 (2004) 645659 659