IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 23, NO.

6, DECEMBER 2015 1943

State and Output Feedback Control of Interval

Type-2 Fuzzy Systems With Mismatched
Membership Functions
Hongyi Li, Xingjian Sun, Ligang Wu, Senior Member, IEEE, and H. K. Lam, Senior Member, IEEE

Abstract—This paper is concerned with the problems of state a weighted sum of some simple linear subsystems [8], and its
and output feedback control for interval type-2 (IT2) fuzzy systems weightings are characterized by the type-1 membership func-
with mismatched membership functions. The IT2 fuzzy model and tions. Recently, many stability analysis and controller synthesis
the IT2 state and output feedback controllers do not share the
same membership functions. A novel performance index, which is results for type-1 fuzzy-model-based (FMB) control systems
expressed as an extended dissipativity performance, is introduced have been developed [9]–[11] . The authors in [7] designed
to be a generalization of H∞ , L2 –L∞ , passive, and dissipativity an FMB fault-tolerant controller for nonlinear stochastic sys-
performances indexes. First, the IT2 Takagi–Sugeno fuzzy model tems against simultaneous sensor and actuator faults. It should
and the controllers are constructed by considering the mismatched be mentioned that the above results are based on parallel dis-
membership functions. Second, on the basis of Lyapunov stability
theory, the IT2 fuzzy state and output feedback controllers are tributed compensation (PDC) design concept [12]. Therefore,
designed, respectively, to guarantee that the closed-loop system is the fuzzy model and the fuzzy controllers or the fuzzy filters
asymptotically stable with extended dissipativity performance. The share the same premise membership functions, which assumes
existence conditions of the two kinds of controllers are obtained that the membership functions contain no uncertainties. How-
in terms of convex optimization problems, which can be solved ever, if there are parameter uncertainties in the nonlinear plant,
by standard software. Finally, simulation results are provided to
illustrate the effectiveness of the proposed methods. then the uncertain parameters will be contained in the member-
ship functions of fuzzy model, and the grades of membership
Index Terms—Extended dissipativity, fuzzy control, interval will become uncertain in value. Then, it is natural to bring some
type-2 fuzzy system, Lyapunov stability theory.
conservative stability analysis results if the PDC design concept
is still used.
I. INTRODUCTION Recently, the authors in [13] pointed out that type-2 fuzzy
sets can be very useful to represent and capture the uncertainties
VER the past few decades, the modeling and control prob-
O lems for nonlinear systems have drawn considerable at-
tention. Since type-1 fuzzy set was first proposed in [1], the
effectively. It has been shown that the type-2 fuzzy logic sys-
tems have the potential to provide better performance than the
type-1 one in [14]–[22] . Based on type-2 fuzzy logic theory,
type-1 fuzzy logic control approach has been widely applied to
the problem of the tracking controller design for the dynamic
practical systems to solve the control problem of the complex
of a unicycle mobile robot was considered in [18]. In [23], the
nonlinear systems [2]–[4] . It is well known that Takagi–Sugeno
authors presented a novel reactive control architecture for au-
(T–S) fuzzy model [5] was introduced to carry out stability
tonomous mobile robots that was based on type-2 fuzzy logic
analysis and controller design for nonlinear systems [6], [7].
controller to implement the basic navigation behaviors and the
The T–S fuzzy model can represent the nonlinear systems by
coordination between these behaviors to produce a type-2 hier-
Manuscript received June 2, 2014; revised August 30, 2014; accepted Novem- archical fuzzy logic controller. The authors in [24] proposed an
ber 8, 2014. Date of publication June 8, 2015; date of current version November interval type-2 (IT2) fuzzy logic congestion controller to achieve
25, 2015. This work was supported in part by the National Natural Science Foun- a superior delivered video quality compared with existing tra-
dation of China under Grant 61333012, Grant 61203002, Grant 61174126, and
Grant 61222301, the Program for New Century Excellent Talents in University ditional controllers and the type-1 fuzzy logic congestion con-
(NCET-13-0696), the Program for Liaoning Innovative Research Team in Uni- troller. Because of the advantage of IT2 fuzzy sets over type-1
versity (LT2013023), the Program for Liaoning Excellent Talents in University fuzzy sets, considerable attention has been paid to IT2 fuzzy
(LR2013053), the Heilongjiang Outstanding Youth Science Fund (JC201406),
the Fok Ying Tung Education Foundation (141059), the Key Laboratory of In- systems in [25]–[27]. The authors in [25] and [26] used IT2
tegrated Automation for the Process Industry, and the Fundamental Research membership functions to capture the nonlinear plants and de-
Funds for the Central Universities (HIT.BRETIV.201303). sign state feedback controllers for the IT2 T–S fuzzy systems.
H. Li is with the College of Engineering, Bohai University, Jinzhou 121013,
China (e-mail: [email protected]). When the state variables are not measurable online, the control
X. Sun is with the School of Mathematics and Physics, Bohai University, methods proposed in [25] and [26] are not available. In addi-
Jinzhou 121013, China (e-mail: [email protected]). tion, it should be mentioned that the performances of IT2 fuzzy
L. Wu is with the Space Control and Inertial Technology Research Center,
Harbin Institute of Technology, Harbin 150001, China (e-mail: ligangwu@ systems have not been considered in the literature.
hit.edu.cn). This paper deals with the problems of state and output feed-
H. K. Lam is with the Department of Informatics, King’s College London, back controllers design for IT2 fuzzy systems with mismatched
London WC2R 2LC, U.K. (e-mail: [email protected]).
Color versions of one or more of the figures in this paper are available online membership functions based on a novel performance index.
at http://ieeexplore.ieee.org. The IT2 fuzzy systems and the IT2 state and output feedback
Digital Object Identifier 10.1109/TFUZZ.2014.2387876 controllers do not share the same membership functions. First,
1063-6706 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
Authorized licensed use limited to: Central South University. Downloaded on June 22,2022 at 10:39:34 UTC from IEEE Xplore. Restrictions apply.
1944 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 23, NO. 6, DECEMBER 2015

the state feedback and the output feedback control systems are where θi (x (t)) denotes the lower grades of member-
constructed. A new performance index, referred to extended ship, θi (x (t)) denotes the upper grades of membership,
dissipativity performance, is introduced. The extended dissipa- μW (fs (x (t))) stands for the lower membership functions, and
is
tivity is a generalization of the H∞ performance, the L2 –L∞ μW i s (fs (x (t))) stands for the upper membership functions.
performance, the passivity performance, and dissipativity per- Here, μW i s (fs (x (t))) ≥ μW (fs (x (t))) ≥ 0 and θi (x (t))≥
formance. Second, based on Lyapunov stability theory, the state is
θi (x (t)) ≥ 0 for all i. Then, the overall IT2 T–S fuzzy system
and output feedback controllers are designed, respectively, to
is represented by
guarantee that the closed-loop system is asymptotically stable
with extended dissipativity performance. The existence con- 
r

ditions of the two kinds of controllers are obtained in terms of ẋ (t) = θi (x (t)) [Ai x (t) + Bi u (t) + D1i w (t)]
i=1
convex optimization problems, which can be solved by standard
software. Finally, simulation results are provided to illustrate the 
r

effectiveness of the proposed method. The rest of this paper is z (t) = θi (x (t)) [Ci x (t) + D2i w (t)] (3)
i=1
organized as follows. Section II formulates the problem, and
Section III presents the main results. Section IV uses some sim- 
r

ulation results to illustrate the effectiveness of the proposed IT2 y (t) = θi (x (t)) Cy i x (t)
fuzzy control schemes, and Section V concludes this paper. i=1

Notation: The notation used throughout the paper is fairly where

standard. Rn stands for the n-dimensional Euclidean space,
and Rn ×m stands for the set of all n × m real matrices; [A]s θi (x (t)) = λi (x (t)) θi (x (t))
is used to denote A + AT for simplicity; P > 0 (≥ 0) stands + λi (x (t)) θi (x (t)) ≥ 0 ∀i (4)
for a symmetric and positive definite (semidefinite); L2 –L∞

r
represents the space of square-integrable vector functions over θi (x (t)) = 1 (5)
[0, ∞); diag{...} stands for a block-diagonal matrix; the super- i=1
scripts “T ” and “−1” stand for matrix transposition and inverse,
0 ≤ λi (x (t)) ≤ 1 ∀i (6)
respectively; and In and 0n denote the identity matrix and zero
matrix with n-dimensions, respectively. In symmetric block ma- 0 ≤ λi (x (t)) ≤ 1 ∀i (7)
trices, we use an asterisk () to represent a term that is induced
λi (x (t)) + λi (x (t)) = 1 ∀i (8)
by symmetry.
with λi (x (t)) and λi (x (t)) being nonlinear functions, and
II. PROBLEM FORMULATION θi (x (t)) denote the grades of membership of the embedded
membership functions.
A. Interval Type-2 Takagi–Sugeno Fuzzy Model
Consider the following IT2 fuzzy model with r rules that B. Interval Type-2 Fuzzy State Feedback Control
represents a continuous-time nonlinear system: In this section, we first construct an IT2 fuzzy state feed-
Plant Rule i: IF f1 (x (t)) is Wi1 and · · · and fp (x (t)) is back controller [27] for the following control design. It is worth
Wip , THEN mentioning that the IT2 fuzzy system and the IT2 fuzzy state
feedback controller do not share the same membership func-
ẋ (t) = Ai x (t) + Bi u (t) + D1i w (t) tions. The jth rule of the fuzzy controller is of the following
z (t) = Ci x (t) + D2i w (t) (1) form:
Controller Rule j: IF g1 (x (t)) is Mj 1 and · · · and gp (x (t))
y (t) = Cy i x (t) is Mj p , THEN
where Wis stands for the ith IT2 fuzzy set of the func- u (t) = Kj x (t) (9)
tion fs (x (t)), i = 1, 2, . . . r, s = 1, 2, . . . p; p is the number
of premise variables; x (t) ∈ Rn is the system state vector, where Mj s stands for the jth fuzzy set of the function gs (x (t)),
u (t) ∈ Rm is the input vector, w (t) ∈ Rh denotes the dis- j = 1, 2, . . . r, s = 1, 2, . . . p; p is the number of premise vari-
turbance input which belongs to L2 [0, ∞), z (t) ∈ Rq is the ables; and Kj ∈ Rm ×n is the state feedback gain matrix of rule
control output, and y (t) ∈ Rg is the measure output; and Ai , j. The firing interval of the jth rule is as follows:
Bi , Ci , D1i , D2i , and Cy i are the known matrices with ap-  p 
 
p
propriate dimensions. The firing interval of the ith rule is as η̃j (x (t)) = μM (gs (x (t))) , μM j s (gs (x (t)))
js
follows: s=1 s=1
 p  
  p = η j (x (t)) , η j (x (t)) (10)
θ̃i (x (t)) = μW (fs (x (t))) , μW i s (fs (x (t)))
is
s=1 s=1 where η j (x (t)) denotes the lower grades of membership, η j
 
= θi (x (t)) , θi (x (t)) (2) (x (t)) denotes the upper grades of membership, μM (gs (x(t)))
js

Authorized licensed use limited to: Central South University. Downloaded on June 22,2022 at 10:39:34 UTC from IEEE Xplore. Restrictions apply.
LI et al.: STATE AND OUTPUT FEEDBACK CONTROL OF INTERVAL TYPE-2 FUZZY SYSTEMS WITH MISMATCHED MEMBERSHIP FUNCTIONS 1945

stands for the lower membership functions, and μM j s (gs (x(t))) kth rule is the following interval set:
stands for the upper membership functions. Here, μM j s  p 
 
p
(gs (x(t)))≥ μM (gs (x (t))) ≥ 0 and η j (x(t)) ≥ η j (x(t)) ˜ k (x (t)) =
 μN (hs (x (t))) , μN k s (hs (x (t)))
js
≥ 0 for all j. The overall IT2 fuzzy state feedback control law
ks
s=1 s=1
is represented by
= [k (x (t)) , k (x (t))]

r
where k (x (t)) denotes the lower grades of member-
u (t) = ηj (x (t)) Kj x (t) (11)
ship, k (x(t)) denotes the upper grades of membership,
j =1
μN (hs (x(t))) stands for the lower membership functions,
ks
where and μN k s (hs (x(t))) stands for the upper membership functions.
Here, μN k s hs (x(t))) ≥ μN (hs (x(t))) ≥ 0, and k (x(t)) ≥
ks
ν j (x (t)) η j (x (t)) + ν j (x (t)) η j (x (t)) k (x(t)) ≥ 0 for all k. The overall IT2 fuzzy output feedback
ηj (x (t)) = r
control law is represented by
ν l (x (t)) η l (x (t)) + ν l (x (t)) η l (x (t))
l=1 
r

≥ 0 ∀j (12) x̂˙ (t) = k (x (t)) [Ack x̂ (t) + Bck y (t)]

k =1

r
(19)
ηj (x (t)) = 1 (13) 
r

j =1 u (t) = k (x (t)) Cck x̂ (t)

k =1
0 ≤ ν j (x (t)) ≤ 1 ∀j (14)
where
0 ≤ ν j (x (t)) ≤ 1 ∀j (15)
κk (x (t)) k (x (t)) + κk (x (t)) k (x (t))
ν j (x (t)) + ν j (x (t)) = 1 ∀j (16) k (x(t)) = r 
κp (x (t)) p (x (t)) + κp (x (t)) p (x (t))
p=1
with ν j (x (t)) and ν j (x (t)) being predefined functions, and
ηj (x (t)) stands for the grades of membership of the embed- ≥ 0 ∀k (20)
ded membership functions. For a simple description, we use the 
r
following notations: θi (x (t))  θi and ηj (x (t))  ηj , where k (x (t)) = 1 (21)
i, j = 1, 2, . . . r. Applying the IT2 fuzzy controller (11) to sys- k =1
tem (3), the resulting IT2 fuzzy closed-loop system can be ex- 0 ≤ κk (x (t)) ≤ 1 ∀k (22)
pressed as follows:
0 ≤ κk (x (t)) ≤ 1 ∀k (23)

r 
r
κk (x (t)) + κk (x (t)) = 1 ∀k (24)
ẋ (t) = θi ηj [(Ai + Bi Kj ) x (t) + D1i w (t)]
i=1 j =1 in which κk (x (t)) and κk (x (t)) are predefined functions, and
(17)

r 
r k (x (t)) denotes the grades of membership of the embed-
z (t) = θi ηj [Ci x (t) + D2i w (t)] ded membership functions. For a simple description, we define
i=1 j =1 k (x (t))  k , where k = 1, 2, . . . r. Under the property of
r r r r
r r r r θi = k = θi k = 1, it can be seen from (3)
where θi = ηj = θi ηj = 1. i=1 k =1 i=1 k =1
i=1 j =1 i=1 j =1 and (19) that the following closed-loop system is obtained:

r 
r
 
C. Interval Type-2 Fuzzy Output Feedback Control ˙
x̄(t) = θi k Āik x̄ (t) + D̄1i w (t)
i=1 k =1
In this section, we will construct an IT2 fuzzy output feedback (25)
controller in the following form: 
r 
r
 
z (t) = θi k C̄i x̄ (t) + D̄2i w (t)
Controller Rule k: IF h1 (x (t)) is Nk 1 and · · · and hp (x (t))
i=1 k =1
is Nk p , THEN  
x (t)
x̂˙ (t) = Ack x̂ (t) + Bck y (t) where x̄ (t) = is the state vector of the closed-loop
(18) x̂ (t)
u (t) = Cck x̂ (t)  
Ai  T T
Bi Cck
system (25), Āik = , D̄1i = D1i 0 ,
where x̂(t) ∈ R is the state vector of the dynamic output feed-
n
  Bck Cy i Ack
back controller; Nk s stands for the kth fuzzy set of the func- C̄i = Ci 0 , and D̄2i = D2i are the system matrices.
tion hs (x (t)), k = 1, 2, . . . r, s = 1, 2, . . . p; p is the number The main purpose of this paper is to design the IT2 fuzzy
of premise variables; and Ack , Bck , and Cck are control gain state feedback controller (11) and output feedback controller
matrices with appropriate dimensions. The firing strength of the (19) such that the closed-loop system is asymptotically stable

Authorized licensed use limited to: Central South University. Downloaded on June 22,2022 at 10:39:34 UTC from IEEE Xplore. Restrictions apply.
1946 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 23, NO. 6, DECEMBER 2015

with the H∞ , L2 –L∞ , passive, and dissipativity performance in- matrices Φ̃ and Ψ̃1 such that
dexes. In [28], the authors introduced a new performance index,
referred to extended dissipativity performance index, which is Φ = Φ̃T Φ̃, Ψ1 = −Ψ̃T1 Ψ̃1 . (28)
a generalization of H∞ , L2 –L∞ , passive, and dissipativity per-
Remark 1: The first item of Assumption 1 guarantees that
formances indexes. In addition, the authors presented some new
inequality (26) is well defined. The second item enables one to
conditions for the filter design of Markovian jump delay systems
derive a linear matrix inequality (LMI)-based condition for the
based on the new performance index. In the following part, we
investigation of the dissipativity analysis problem. The condi-
introduce the new performance index from [28]. First, the fol-
tions of Assumption 1 similar to 1), 2), and 5) were used in [32],
lowing assumption is given for developing the new performance
[34], and [35]. On the other hand, when considering the L2 –L∞
index.
performance, it is well known that the output of the considered
Assumption 1 (see[28]): Let Φ, Ψ1 ,Ψ2 , and Ψ3 be matrices
system should not include disturbance inputs [36]. Therefore, it
such that the following conditions are satisfied:
should be assumed that D2i = 0 when Φ = 0, which justifies the
1) Φ = ΦT , Ψ1 = ΨT1 and Ψ3 = ΨT3 . need of the third item of Assumption 1. Finally, the fourth item
of Assumption 1 is technically necessary for the development
2) Φ ≥ 0 and Ψ1 ≤ 0. of our analysis and design methods.
3) ||D2 i || · ||Φ|| = 0. In this paper, our objective is to design the state feedback
controller in (11) and output feedback controller in (19) for
4) (||Ψ1 || + ||Ψ2 ||) · ||Φ|| = 0. system (3) such that: 1) the closed-loop system ( 17) [or (25)] is
5) T
D2i T
Ψ1 D2i + D2i Ψ2 + ΨT2 D2i + Ψ3 > 0. asymptotically stable with w(t) = 0; 2) the closed-loop system
(17) [or (25)] guarantees the new performance index (26).
Definition 1 (see[28]): For given matrices Φ, Ψ1 , Ψ2 , and
Ψ3 satisfying Assumption 1, system (17) [or system (25)] is III. MAIN RESULTS
said to be extended dissipative if there exists a scalar ρ such
that the following inequality holds for any t > 0 and all w (t) ∈ This section is concerned with the controllers design problem
L2 [0, ∞): for the IT2 T–S fuzzy system. The existence conditions of the
 t controllers are given in the following theorems. We first present
IT2 fuzzy state feedback controller design results.
J (t) dt − z (t)T Φz (t) ≥ ρ (26)
0 Theorem 1: For given matrices Φ̃, Ψ̃1 , Ψ2 , and Ψ3 satisfying
(28) and Assumption 1, the system in ( 17) is asymptotically
where J(t) = z(t)T Ψ1 z(t) + 2z(t)T Ψ2 w(t) + w (t)T Ψ3 w (t). stable and satisfies the performance index in Definition 1, if
It can be seen from Definition 1 that the following perfor- there exist matrices G = GT > 0, Q = QT > 0, ΛTi = Λi , Mj
mance indexes hold. (i, j = 1, 2, . . . r) with appropriate dimensions, and under the
1) Choosing Φ = 0, Ψ1 = −I, Ψ2 = 0, Ψ3 = γ 2 I, and ρ = condition ηj − σj θj ≥ 0 (0 < σj < 1) for all j = 1, 2, . . . , r,
0, inequality (26) reduces to the H∞ performance [29]. such that the following LMIs are satisfied:
2) Letting Φ = I, Ψ1 = 0, Ψ2 = 0, Ψ3 = γ 2 I, and ρ = 0,
inequality (26) becomes the L2 –L∞ (energy-to-peak) per- Θ1 < 0 (29)
formance [30].
Θ2 < 0 (30)
3) If the dimension of output z (t) is the same as that of
disturbance w (t), then inequality (26) with Φ = 0, Ψ1 = Ωij − Λi < 0 (31)
0, Ψ2 = I, Ψ3 = γI, and ρ = 0 becomes the passivity
σi Ωii − σi Λi + Λi < 0 (32)
performance [31].
4) Letting Φ = 0, Ψ1 = Q, Ψ2 = S, Ψ3 = R − αI, and σj Ωij + σi Ωj i − σj Λi − σi Λj + Λi + Λj ≤ 0, i < j (33)
ρ = 0, inequality (26) reduces to the strict (Q, S, R)-
dissipativity [32]. where
5) When Φ = 0, Ψ1 = − I, Ψ2 = I, Ψ3 = −σI with > 0 ⎡ ⎤
Ω̄11 Ω̄12 Ω̄13  
and σ > 0, inequality (26) becomes the very-strict passiv- ⎢ ⎥ −Q Q
ity performance. In the definition of the very-strict passiv- Ωij = ⎢
⎣  Ω̄22 Ω̄23 ⎥
⎦ , Θ1 =
ity performance, the scalar ρ is not required to be zero. It  G − 2I
  −I
was shown in [33] that ρ should be a nonpositive scalar.
This fact can also be seen from Assumption 1 and Def-  
inition 1. Indeed, when w (t) = 0, from (26), it follows −G C̃iT Φ̃T
that Θ2 = , C̃i = Ci Q, Ω̄13 = C̃iT Ψ̃T1
 t  −I
ρ≤ e (t)T Ψ1 e (t) dt − e (t)T Φe (t) . (27)
0 Ω̄11 = [Ai Q + Bi Mj ]s , Ω̄12 = D1i − C̃iT Ψ2
Note from Assumption 1 that Φ ≥ 0 and Ψ1 ≤ 0. Thus, the  T 
above inequality implies that ρ ≤ 0, and there always exist Ω̄22 = − D2i Ψ2 s − Ψ3 , Ω̄23 = D2i
T
Ψ̃T1 .

Authorized licensed use limited to: Central South University. Downloaded on June 22,2022 at 10:39:34 UTC from IEEE Xplore. Restrictions apply.
LI et al.: STATE AND OUTPUT FEEDBACK CONTROL OF INTERVAL TYPE-2 FUZZY SYSTEMS WITH MISMATCHED MEMBERSHIP FUNCTIONS 1947

Then, the IT2 fuzzy state feedback controller gain matrices are where
given as  
ξ (t)T = g (t)T w (t)T

Kj = Mj Q−1 . J (t) = z (t)T Ψ1 z (t) + 2z (t)T Ψ2 w (t) + w (t)T Ψ3 w (t)

⎡ ⎤
Ω̃1ij Ω̃2ij
In this case, the scalar ρ involved in Definition 1 can be chosen as Ω̃ij = ⎣ ⎦
 Ω̃3ij
ρ = −V (x (0)) . (34) Ω̃1ij = [Ai Q + Bi Mj ]s − C̃iT Ψ1 C̃i
Ω̃2ij = D1i − C̃iT Ψ1 D2i − C̃iT Ψ2
Proof: Choose a quadratic Lyapunov function for the stabil-  T 
ity analysis of system (17) as follows: Ω̃3ij = −D2iT
Ψ1 D2i − D2i Ψ2 s − Ψ3 .
r r
T Consider θi (θj − ηj ) Λi = 0, where Λi = ΛTi is an ar-
V (x (t)) = x (t) P x (t) (35) i=1 j =1
bitrary matrix with appropriate dimensions. Then
T
where P = P > 0. Then, the time derivative of V (t) is 
r 
r 
r 
r 
r 
r

given by θi ηj Ωij = θi ηj Ωij + θi (θj − ηj ) Λi

i=1 j =1 i=1 j =1 i=1 j =1

V̇ (x (t)) = 2x (t)T P ẋ (t) 

r 
r
= θi (θj − ηj + σj θj − σj θj ) Λi
r  r   i=1 j =1
= θi ηj x (t)T [P (Ai + Bi Kj )]s x (t)
i=1 j =1 
r 
r
+ θi (ηj + σj θj − σj θj ) Ωij
+ 2x (t)T P D1i w (t) . i=1 j =1


r 
r
= θi θj (σj Ωij − σj Λi + Λi )
Let g (t) = Q−1 x (t), C̃i = Ci Q, and Q = P −1 . Then, it can i=1 j =1
be obtained that

r 
r
+ θi (ηj − σj θj ) (Ωij − Λi )

r 
r   i=1 j =1
V̇ (x (t)) = θi ηj g (t)T [Ai Q + Bi Mj ]s g (t)
i=1 j =1 
r 
r
= θi2 (σi Ωii − σi Λi + Λi )
T i=1 j =1
+ 2g (t) D1i w (t) . (36)
r −1 
 r
+ θi θj × (σj Ωij − σj Λi
From Ψ1 ≤ 0, it can be seen that i=1 j =i+1

⎡ ⎤T + Λi + σi Ωj i − σi Λj + Λj )
r 
r
z (t)T Ψ1 z (t) = ⎣ θi ηj C̃i g (t) + D2i w (t) ⎦ Ψ1 
r 
r
+ θi (ηj −σj θj) (Ωij −Λi) . (38)
i=1 j =1
i=1 j =1
 

r 
r
It can be seen from (31)–(33) that
× θl ηm C̃l g (t) + D2l w (t)
l=1 m =1 
r 
r
θi ηj Ωij < 0.

r 
r  T
≥ θi ηj C̃i g (t) + D2i w (t) Ψ1 i=1 j =1

i=1 j =1 By the Schur complement, one can have


× C̃i g (t) + D2i w (t) . (37) 
r 
r
θi ηj Ω̃ij < 0.
i=1 j =1
Then That is to say
⎛ ⎞

r 
r r 
r
V̇ (x (t)) − J (t) ≤ ξ (t)T θi ηj Ω̃ij ξ (t) V̇ (t) − J (t) < ξ (t)T ⎝ θi ηj Ω̃ij ⎠ ξ (t) < 0.
i=1 j =1 i=1 j =1

Authorized licensed use limited to: Central South University. Downloaded on June 22,2022 at 10:39:34 UTC from IEEE Xplore. Restrictions apply.
1948 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 23, NO. 6, DECEMBER 2015

Therefore, there is always a sufficiently small scalar c > 0 such When w (t) ≡ 0, it follows from (39) that
that Ω̃ij ≤ −cI. This means that

V̇ (x (t)) − J (t) ≤ −c |ξ (t)|2 . (39)

V̇ (t) ≤ z (t)T Ψ1 z (t) − c |ξ (t)|2 . (45)
Thus, J (t) ≥ V̇ (t) holds for any t ≥ 0 , which means
 t
J (s) ds ≥ V (x (t)) − V (x (0)) . (40) Noticing that Ψ1 < 0 under Assumption 1, we obtain that
0
V̇ (t) ≤ −c |ξ (t)|2 . Then, we can show that the closed-loop
It is shown from (G − I)G−1 (G − I) ≥ 0 with G > 0 that system (17) with w (t) = 0 is asymptotically stable. This con-
cludes the proof. 
−G−1 ≤ G − 2I. (41)
In the following part, we will solve the problem of IT2 fuzzy
From (29) and (41), we know that P > G, which means output feedback controller synthesis for the IT2 fuzzy system
(3). By following the same line as the proof of Theorem 1, the
V (x (t)) = x (t)T P x (t) ≥ x (t)T Gx (t) ≥ 0. following theorem is obtained directly.
For inequality (40), it is derived from (34) that Theorem 2: For given matrices Φ̃, Ψ̃1 , Ψ2 , and Ψ3 satisfy-
ing (28) and Assumption 1, the closed-loop system in ( 25)
 t
is asymptotically stable and satisfies the performance index in
J (s) ds ≥ x (t)T Gx (t) + ρ ∀t ≥ 0. (42)
0
Definition 1, if there exist matrices P = P T > 0, G > 0 and
Λ̃Ti = Λ̃i (i = 1, 2, . . . , r) with appropriate dimensions, and un-
According to Definition 1, we need to prove that the following
der the condition k − σ̄k θk ≥ 0 (0 < σ̄k < 1) for all k, such
inequality holds for any matrices Φ, Ψ1 , Ψ2 , and Ψ3 satisfying
that the following LMIs hold:
Assumption 1:
 t
J (t) dt − z (t)T Φz (t) ≥ ρ. (43)
0 G−P < 0 (46)
To this end, we consider the two cases of ||Φ|| = 0 and ||Φ|| = 0, Θ̃2 < 0 (47)
respectively.
Πik − Λ̃i < 0 (48)
First, we consider the case when ||Φ|| = 0. It follows from
(42), for any t ≥ 0, σ̄i Πii − σ̄i Λ̃i + Λ̃i < 0 (49)
 t
σ̄k Πik + σ̄i Πk i − σ̄k Λ̃i − σ̄i Λ̃k + Λ̃i + Λ̃k ≤ 0, i < k (50)
J (s) ds ≥ x (t)T Gx (t) + ρ ≥ ρ. (44)
0

This implies that (43) holds by noting that z (t)T Φz (t) ≡ 0. where
Second, we consider the case of ||Φ|| = 0. In this case, it
is required under Assumption 1 that ||Ψ1 || + ||Ψ2 || = 0 and
||D2 i || = 0, which implies that Ψ1 = 0, Ψ2 = 0, and Ψ3 > 0. ⎡  ⎤
P Āik s P D̄1i − C̄iT Ψ2 C̄iT Ψ̃T1
Thus, J (s) = w (s)T Ψ3 w (s) ≥ 0. Then, using the Schur com- ⎢ ⎥
⎢  T  ⎥
plement to the condition (30), it can be obtained that C̃iT ΦC̃i ≤ Πik = ⎢  − D̄2i Ψ2 s − Ψ3 T
D̄2i Ψ̃T1 ⎥
G . For any t ≥ 0, the following inequalities hold: ⎣ ⎦
 t   −I
J (s) ds − z (t)T Φz (t)  
0
−G C̄iT Φ̃T
 Θ̃2 = .
t 
r 
r   −I
≥ J (s) ds − θi ηj (Ci x (t) + D2i w (t))T Φ
0 i=1 j =1

× (Ci x (t) + D2i w (t))] In the following theorem, the control gain matrices Ack , Bck ,
 t  r  r and Cck in (19) will be solved.
= J (s) ds − θi ηj g (t)T C̃iT ΦC̃i g (t) Theorem 3: Considering the IT2 fuzzy system (3), for given
0 i=1 j =1 matrices Φ̃, Ψ̃1 , Ψ2 , and Ψ3 satisfying (28) and Assumption 1,
 t 
r 
r system (25) is asymptotically stable and satisfies the perfor-
≥ J (s) ds − θi ηj x (t)T Gx (t) ≥ ρ. mance index in Definition 1, if there exists matrices Λ̄Ti = Λ̄i ,
 
0 i=1 j =1 G1 G2
i = 1, 2, . . . r, Ḡ = > 0, R >0, S > 0, Ai , Bi ,
Based on the two cases of ||Φ|| = 0 and ||Φ|| = 0, we know that  G3
the closed-loop system (17) is extended dissipative in the sense and Ci with appropriate dimensions, and under the condition
of Definition 1. k − σ̄k θk ≥ 0 (0 < σ̄k < 1) for all k = 1, 2, . . . , r, such that

Authorized licensed use limited to: Central South University. Downloaded on June 22,2022 at 10:39:34 UTC from IEEE Xplore. Restrictions apply.
LI et al.: STATE AND OUTPUT FEEDBACK CONTROL OF INTERVAL TYPE-2 FUZZY SYSTEMS WITH MISMATCHED MEMBERSHIP FUNCTIONS 1949

the following LMIs hold:

T
Ξ̌5ik = D2i Ψ̃T1 , Ξ̌6ik = −Im , Ξ̂1ik = σ̄k Ξ̌1ik + σ̄i Ξ̌1k i
  Ξ̂2ik = σ̄k Ξ̌2ik + σ̄i Ξ̌1k i , Ξ̂3ik = σ̄k Ξ̌3ik + σ̄i Ξ̌3k i
R I
> 0 (51)
I S Ξ̂4ik = σ̄k Ξ̌4ik + σ̄i Ξ̌4k i , Ξ̂5ik = σ̄k Ξ̌5ik + σ̄i Ξ̌5k i
 
R I  
Ḡ − < 0 (52) −In 0
I S Ξ̂6ik = (σ̄k + σ̄i ) Ξ̌6ii , Ξ7ik =
  0 −In
−Ḡ Θ̄2    
< 0 (53) −Im 0 0 R (Ai − Ak )T
 −I Ξ8ik = , χ̌1ik =
0 −Im S 0
Π̌ik − Λ̂i < 0 (54)
 
σ̄i Π̃ii − σ̄i Λ̄i + Λ̄i < 0 (55) 0 R (Ai − Ak )T
χ̂1ik =
Π̂ik − σ̄k Λ̂i − σ̄i Λ̂k + Λ̂i + Λ̂k ≤ 0, i < k (56) (σ̄k − σ̄i ) S 0
 T 
0 R (Cy i − Cy k )
where χ̌2ik =
Bk 0
⎡ ⎤  
Ξ̌1ik Ξ̌2ik Ξ̌3ik χ̌1ik χ̌2ik χ̌3ik 0 CkT
⎢ ⎥ χ̌3ik =
⎢ ⎥ S (Bi − Bk ) 0
⎢  Ξ̌4ik Ξ̌5ik 0 0 0 ⎥
⎢ ⎥  
⎢ ⎥ 0 R (Cy i − Cy k )T
⎢   Ξ̌6ik 0 0 0 ⎥
⎢ ⎥ χ̂2ik =
Π̌ik = ⎢ ⎥
⎢ ⎥ σ̄k Bk − σ̄i Bi 0
⎢    Ξ7ik 0 0 ⎥
⎢ ⎥  
⎢ ⎥ 0 σ̄k CkT − σ̄i CiT
⎢     Ξ8ik 0 ⎥
⎣ ⎦ χ̂3ik = .
S (Bi − Bk ) 0
     Ξ8ik
⎡ ⎤
Ξ̌1ii Ξ̌2ii Ξ̌3ii   Then, the IT2 fuzzy output feedback controller gain matrices
⎢ ⎥ Λ̄i 0
⎢ ⎥ are given as follows:
Π̃ii = ⎢  Ξ̌4ii Ξ̌5ii ⎥ , Λ̂i =
⎣ ⎦ 0 02(n +2m )
  Ξ̌6ii
Cci = Ci M −T (57)
⎡ ⎤
Ξ̂1ik Ξ̂2ik Ξ̂3ik χ̂1ik χ̂2ik χ̂3ik Bci = N −1
Bi (58)
⎢ ⎥
⎢ ⎥ −1 −T
⎢  Ξ̂4ik Ξ̂5ik 0 0 0 ⎥ Aci = N (Ai −S Ai R − B i Cy i R − S Bi Ci) M (59)
⎢ ⎥
⎢ ⎥
⎢   Ξ̂ 0 0 0 ⎥
⎢ 6ik ⎥
Π̂ik = ⎢ ⎥ where M and N are nonsingular matrices satisfying:
⎢ ⎥
⎢    Ξ7ik 0 0 ⎥
⎢ ⎥
⎢ ⎥
⎢     Ξ8ik 0 ⎥ MN T
= I − RS . (60)
⎣ ⎦
     Ξ8ik
⎡ ⎤   Proof: Using the Schur complement, it can be seen from (56)
RCiT Φ̃T D1i − RCiT Ψ2 that
Θ̄2 = ⎣ ⎦ , Ξ̌2ik =
CiT Φ̃T S D1i − CiT Ψ2
⎡ ⎤
  Ξ̂1ik Ξ̂2ik Ξ̂3ik
(Ai R+Bi Ck )s Ai + AkT ⎢ ⎥
Ξ̌1ik = ⎢ ⎥
⎢  Ξ̂4ik Ξ̂5ik ⎥ − σk Λ̄i − σi Λ̄k + Λ̄i + Λ̄k
 (S Ai + Bk Cy i )s ⎣ ⎦
⎡ ⎤   Ξ̂6ik
RCiT Ψ̃T1  T 
Ξ̌3ik = ⎣ ⎦ , Ξ̌4ik = − D2i Ψ2 s − Ψ3 
3
CiT Ψ̃T1 + κ̂j κ̂jT + φ̂j φ̂Tj < 0 ∀i, k
j =1

Authorized licensed use limited to: Central South University. Downloaded on June 22,2022 at 10:39:34 UTC from IEEE Xplore. Restrictions apply.
1950 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 23, NO. 6, DECEMBER 2015

where Considering that P P −1 = I, inequality (60) holds. From (51),

⎡ ⎤ ⎡ ⎤ it is obvious that
0 0  
⎢ ⎥ ⎢ ⎥ −R −I
κ̂1 =⎢ ⎥
⎣ (σ̄k − σ̄i ) S ⎦ , κ̂2 = ⎢ ⎥
⎣ σ̄k Bk − σ̄i Bi ⎦ <0
−I −S
02×1 02×1
    which shows that R − S −1 > 0, this is to say I − RS is non-
T T
R (Ai − Aj ) R (Cy i − Cy k ) singular. This ensures that there are always nonsingular matrices
φ̂1 = , φ̂2 = M and N such that (60) is satisfied. Setting
03×1 03×1    
⎡ ⎤ R I I S
0   X1 = , X2 = (63)
⎢ ⎥ σ̄k CkT − σ̄i CiT
MT 0 0 NT
κ̂3 = ⎣ S (B − B ) ⎦ , φ̂3 = .
i k
03×1 we obtained from (63) that
02×1
P X1 = X2 . (64)
It is easy to see that
It follows that

3 
3  
κ̂j κ̂jT + φ̂j φ̂Tj ≥ κ̂j φ̂Tj + φ̂j κ̂jT R I
j =1 j =1
X1T P X1 = X1T X2 =
I S
which means which means that X1 and X2 are positive definite, and P
⎡ ⎤
Ξ̂1ik Ξ̂2ik Ξ̂3ik can be expressed as P = X2 X1−1 > 0. Consider the following
⎢ ⎥ equations:
⎢ ⎥
⎢  Ξ̂4ik Ξ̂5ik ⎥ − σ̄k Λ̄i − σ̄i Λ̄k + Λ̄i + Λ̄k
⎣ ⎦ S Ai R + B k Cy i R + S Bi Ck + N Ack M T
  Ξ̂6ik = S Ak R + B k Cy k R + S Bk Ck + N Ack M T

3
+ S (Ai − Ak) R +Bk (Cy i − Cy k ) R + S (Bi − Bk ) Ck
+ κ̂j φ̂Tj + φ̂j κ̂jT <0 ∀i, k. (61)
j =1 and

Similarly, for (54), one can see σ̄i S Ak R + B i Cy k R + S Bk Ci + N Aci M T
⎡ ⎤ 
+ σ̄k S Ai R + B k Cy i R + S Bi Ck + N Ack M T
Ξ̌1ik Ξ̌2ik Ξ̌3ik
⎢ ⎥  
⎢ ⎥
3
 = σ̄i S Ai R + B i Cy i R + S Bi Ci + N Aci M T
⎢  Ξ̌4ik Ξ̌5ik ⎥ − Λ̄i + κ̌j φ̌Tj + φ̌j κ̌jT < 0 ∀i, k
⎣ ⎦ 
j =1 + σ̄k S Ak R + B k Cy k R + S Bk Ck + N Ack M T
  Ξ̌6ik
(62) + (σ̄k − σ̄i ) S (Ai − Ak ) R+ (σ̄k Bk − σ̄i Bi )
where
× (Cy i − Cy k ) R + S (Bi − Bk ) (Ck − Ck ) .
⎡ ⎤
0   By performing congruence transformation by diag{X1 -1 ,I,I}
⎢ ⎥ R (Ai − Ak )T
⎢ ⎥
κ̌1 = ⎣ S ⎦ , φ̌1 = to (61) and (62), we know that conditions in (50) and (48)
03×1 hold. On the other hand, we perform congruence transfor-
02×1 mation to (52), (53), and (55 ) by X1−1 , diag{X1−1 , I}, and
⎡ ⎤ diag{X1−1 , I, I}, respectively. We can see that the conditions
0  
R (Cy i − Cy k )T in (46), (47), and (49) hold. Therefore, all the conditions in
⎢ ⎥
⎢ ⎥
κ̌2 = ⎣ Bk ⎦ , φ̌2 = Theorem 2 are satisfied. The proof is complete. 
03×1 Remark 2: The main contributions of this paper can be sum-
02×1 marized as follows: 1) A new performance index, including the
⎡ ⎤ H∞ performance, the L2 –L∞ performance, the passivity per-
0  T 
⎢ ⎥ Ck formance, and dissipativity performance. 2) Based on the new
κ̌3 = ⎣ S (B − B ) ⎦ , φ̌3 = . performance index, a novel IT2 fuzzy state feedback controller
i k
03×1 is designed for IT2 fuzzy systems with mismatched membership
02×1
functions. 3) A new IT2 fuzzy output feedback controller is also
In order to solve the parameters of the IT2 fuzzy output designed for IT2 fuzzy systems with mismatched membership
feedback controller, the matrix P is partitioned and inverted as functions under a unified frame.
    In order to show the advantages of the proposed results over
S N R M
P = , P −1 = . the existing type-1 fuzzy control results, we give the follow-
NT Y MT T ing lemma. In [29], the authors investigated robust H∞ control

Authorized licensed use limited to: Central South University. Downloaded on June 22,2022 at 10:39:34 UTC from IEEE Xplore. Restrictions apply.
LI et al.: STATE AND OUTPUT FEEDBACK CONTROL OF INTERVAL TYPE-2 FUZZY SYSTEMS WITH MISMATCHED MEMBERSHIP FUNCTIONS 1951

problem of T–S fuzzy systems with state and input time delays. TABLE I
MEMBERSHIP FUNCTIONS OF THE PLANT
To compare with our results, we consider the following type-1
fuzzy system with k fuzzy rules (65):
Lower membership functions Upper membership functions

k
0.925 0.925
ẋ(t) = hi (x(t)) [Ai x(t) + Bi u(t) + D1i w(t)] θ 1 (x 1 ) = 0.95 − (x 1 + 4.5)
θ 1 (x 1 ) = 0.95 − (x 1 + 3.5)
i=1 1 + e− 8 1 + e− 8
(65) 0.925 0.925

k θ 2 (x 1 ) = 0.025 + ( x 1 −4 . 5 )
θ 2 (x 1 ) = 0.025 + ( x 1 −3 . 5 )
z(t) = hi (x(t)) (Ci x(t) + D2i w(t)) 1 + e− 8 1 + e− 8

i=1 θ 3 (x 1 ) = 1 − θ 1 (x 1 ) − θ 2 (x 1 ) θ 3 (x 1 ) = 1 − θ 1 (x 1 ) − θ 2 (x 1 )

where hi (x(t)) has been defined in [29]. Based on the PDC

design concept, the following state feedback fuzzy controller
can be obtained:

k
u(t) = hi (x(t))Ki x(t). (66)
i=1

Under (66), the resulting closed-loop system can be represented

as follows:

k 
k
ẋ(t) = hi (x(t))hj (x(t)) [(Ai + Bi Kj )x(t)
i=1 j =1

+ D1i w(t)]

k 
k
z(t) = hi (x(t))hj (x(t)) [Ci x(t)
i=1 j =1 Fig. 1. Membership functions of the IT2 fuzzy system.

+ D2i w(t)] . (67)

Based on the method proposed in [29], the following lemma can employed to illustrate the practicability of the proposed results
be easily obtained. in Example 2, in which the IT2 fuzzy state feedback controller
Lemma 1: For given scalar a2 = 0, system (67) is asymptot- is applied to control the pendulum system.
ically stable with an H∞ norm bound γ > 0, and the feedback Example 1: Consider the following three-rule IT2 fuzzy
gain matrices are given by system:
Ki = Fi X −1 , i = 1, 2, . . . , k Plant Rule i: IF x1 (t) is Wi1 , THEN

if there exist matrices P̄ > 0 and X such that the following ẋ (t) = Ai x (t) + Bi u (t) + D1i w (t)
LMIs hold, for 1 ≤ i, j ≤ k: (70)
z (t) = Ci x (t) + D2i w (t) , i = 1, 2, 3
Πii < 0 (68)
where
Πij + Πj i < 0, i < j (69)    
2.78−5.63 0.2 −3.22
where A1 = , A2 =
⎡ ⎤ 0.01 0.33 0.35 0.12
Φ1 Φ2 D1i XCiT  
⎢ ⎥ −14 −6.63  T
⎢  T ⎥ A3 = , B1 = 2 −1
⎢ Φ3 −a2 D1i D2i ⎥ 0.45 0.15
Πij = ⎢
⎢
⎥ , Φ1 = [Ai X + Bi Fj ]
⎥ s  T  T
⎢   −γ 2 I 0 ⎥ B2 = 8 0 , B3 = −14 −1
⎣ ⎦
   −I  T  T
D11 = 0.1 0.01 , D12 = 0.1 0.01
Φ2 = −X + a2 XATi + a2 FjT BiT + P̄ , Φ3 = −a2 X − a2 X.  T  
D13 = −0.01 0.1 , C1 = 0.01 0.1
   
IV. SIMULATION EXAMPLES C2 = 0.01 0.1 , C3 = −0.02 0.2
To validate the effectiveness and the practicality of the D21 = −0.01, D22 = −0.02, D23 = −0.01.
proposed control design schemes, two simulation examples
are provided in this section. In Example 1, the effectiveness of The lower and upper membership functions are chosen in
both the IT2 fuzzy state feedback and output feedback control Table I. Fig. 1 shows the membership functions of the IT2
schemes is testified. The inverted pendulum application is fuzzy system according to the representation in (70).

Authorized licensed use limited to: Central South University. Downloaded on June 22,2022 at 10:39:34 UTC from IEEE Xplore. Restrictions apply.
1952 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 23, NO. 6, DECEMBER 2015

Fig. 2. State responses of the open-loop system.

TABLE II
MEMBERSHIP FUNCTIONS OF THE CONTROLLER
Fig. 3. Membership functions of the IT2 fuzzy controller.

Lower membership functions Upper membership functions

1 1
η 1 (x 1 ) = 1 − x1+5
η 1 (x 1 ) = 1 − x1+4
− −
1+e 2 1+e 2

1 1
η 2 (x 1 ) = x 1 −5
η 2 (x 1 ) = x 1 −4
1 + e− 2 1 + e− 2

η 3 (x 1 ) = 1 − η 1 (x 1 ) − η 2 (x 1 ) η 3 (x 1 ) = 1 − η 1 (x 1 ) − η 2 (x 1 )

It is assumed that the disturbance w(t) is


0.1sin(5t), 0 ≤ t ≤ 5
w(t) = (71)
0, else.
 T
Under the initial condition x (0) = −10 −5 , Fig. 2 de- Fig. 4. State responses of the closed-loop system under IT2 fuzzy state feed-
picts the state responses of the open-loop system in (2), which back controller.
indicates that the open-loop system (70) is not stable. In this
case, we design the IT2 fuzzy state feedback controller in (11)
to stabilize this unstable system in (70). plotted in Fig. 4. Obviously, the unstable system has been ef-
Next, according to the description in (9) and (10), the lower fectively stabilized by the designed IT2 fuzzy state feedback
membership functions and the upper membership functions in controller. Therefore, the whole simulation in this control pro-
(10) of the IT2 fuzzy controller are defined in Table II . From cedure has demonstrated the effectiveness of the designed IT2
(12), by choosing the constants ν j (x(t)) = 0.5 and ν̄j (x(t)) = fuzzy state feedback control scheme.
0.5 (j = 1, 2, 3), we can obtain the membership functions of the Remark 3: It should be noted that the IT2 membership func-
IT2 fuzzy state feedback controller, which are shown in Fig. 3. tions will generate uncertain grades of membership as presented
In this control scheme, we consider the L2 –L∞ performance in (4). As a result, the existing type-1 stability analysis for the
index for the system in (70). Based on Definition 1, by setting T–S fuzzy system under the PDC concept cannot be applied.
Φ = I , Ψ1 = 0, Ψ2 = 0, and Ψ3 = 0.1I, and according to According to the representation in (70), Fig. 1 shows the mem-
Theorem 1, with the parameters σ̄k (k = 1, 2, 3) chosen as σ̄1 = bership functions of the IT2 fuzzy system. From Tables I and II,
0.1, σ̄2 = 0.9, σ̄3 = 0.1, by solving the conditions (29)–(33), we it is obvious that the membership functions of the plant and
can obtain the L2 –L∞ performance index γ = 1.1364, and the controller are not matched, even though in such a situation, the
controller gain matrices are obtained as follows: plant can also be controlled with desired system performances.
    We continue to consider that the state cannot be measured.
K1 = −0.6620 −0.1275 , K3 = −0.3458 0.0656 , Then, the IT2 fuzzy dynamic output feedback controller is
  designed to control the IT2 fuzzy system in (70). The measured
K2 = −0.7283 −0.1808 .
output
  as y (t) = Cy i x (t) (i = 1, 2, 3),
is given  where Cy 1 =
Thus, under the same initial state condition, we can obtain the 0.78 0.66 , Cy 2 = 0.33 0.75 , Cy 3 = 0.78 0.66 .
state responses of the closed-loop system in (70), which are We consider the same membership functions in Table II for the

Authorized licensed use limited to: Central South University. Downloaded on June 22,2022 at 10:39:34 UTC from IEEE Xplore. Restrictions apply.
LI et al.: STATE AND OUTPUT FEEDBACK CONTROL OF INTERVAL TYPE-2 FUZZY SYSTEMS WITH MISMATCHED MEMBERSHIP FUNCTIONS 1953

Fig. 5. State responses of the closed-loop system under IT2 fuzzy output
feedback controller. Fig. 6. Inverted pendulum system.

IT2 fuzzy output feedback controller design, i.e., Remark 4: In our control procedure, the Matlab LMI tool-
box was used to solve the LMI-based conditions in Theorems 1
1 (x1 ) = η 1 (x1 ) , 1 (x1 ) = η 1 (x1 ) and 3, respectively. Referring to Figs. 4 and 5, it can be seen
that the controllers can stabilize the system in (70) with desired
2 (x1 ) = η 2 (x1 ) , 2 (x1 ) = η 2 (x1 )
performances. The main difference is that when the system state
3 (x1 ) = η 3 (x1 ) , 3 (x1 ) = η 3 (x1 ) . is unmeasurable, the designed IT2 fuzzy output feedback con-
troller can stabilize the IT2 fuzzy system and satisfy the designed
In this control scheme, we consider the H∞ performance performance index. Particularly, Fig. 5 has shown the effective-
index for the system in (70). From (20), we choose the con- ness of the proposed IT2 output feedback control scheme. It
stants ν j (x(t)) = 0.5 and ν̄j (x(t)) = 0.5 (j = 1, 2, 3). Based should be mentioned that there are few results about the IT2
on Definition 1, by setting Φ = 0, Ψ1 = −I, Ψ2 = 0, and output feedback control for IT2 fuzzy systems in the existing
Ψ3 = 0.1I, and according to Theorem 3, by solving the con- literature.
ditions (51)–(56), with the parameters σ̄k (k = 1, 2, 3) chosen Remark 5: As compared with the existing literature on
as σ̄1 = 0.2, σ̄2 = 0.9, σ̄3 = 0.3, we can obtain the H∞ perfor- type-1 fuzzy systems, the main significant advantage of this
mance index γ = 1.3255, and the controller gain matrices are study is to solve the control problem of the uncertain nonlin-
obtained as follows: ear system under the performances (L2 –L∞ , H∞ , passive, and
  dissipativity performances indexes). In the following part, a
1.2753 −0.0723
Ac1 = practical example will be utilized to show the effectiveness of
−40.0871 1.3775 the proposed results.
  Example 2: In this example, the inverted pendulum, shown
0.6287 0.0341
Ac2 = in Fig. 6, is used to testify the applicability of the proposed
−22.1301 −0.4088 results. In this control procedure, at first, under different initial
  states, the disturbance input is considered in the pendulum sys-
1.0100 0.5520 tem, and then, the disturbance-free case is considered. For the
Ac3 =
−34.4985 −18.1891 limitation of the space, we only provide the simulation results
for the designed IT2 fuzzy state feedback control scheme. The
   T
0.3010 0.6262 dynamic equation for the inverted pendulum [25] is given by
−3
Bc1 = 10 × , Cc1 =
3.2673 0.1546 g sin(θ(t))− amp Lθ̇(t)2 sin(2θ(t))/2−a cos(θ(t))u(t)
θ̈(t) =
   T 4L/3 − amp L cos2 (θ(t))
0.1457 0.0438
Bc2 = 10 × −3
, Cc2 = where θ(t) denotes the angular displacement of the pendu-
  due to gravity g = 9.8 m/s . mp ∈
2
17.3723 0.1381 lum, and the acceleration
   T mp min m  p max = 2 3  kg is the mass  of the pendu-
−0.0096 −0.5557 lum, Mc ∈ Mc min Mc max = 8 16 kg is the mass of
Bc3 = 10−3 × , Cc3 = . 1
the cart, a = m p +M , 2L = 1 m is the length of the pendulum,
−14.0764 −0.6201 c
and u(t) is the control force (in Newtons) applied to the cart.
Thus, under the same initial state condition, we can obtain the The following four-rule T–S fuzzy model is obtained to describe
state responses of the closed-loop system in (70), which are inverted pendulum:
plotted in Fig. 5 . Obviously, the unstable system has been also

4
effectively stabilized by the designed IT2 fuzzy output feedback ẋ(t) = hi (x(t))(Ai x(t) + Bi u(t)) (72)
controller. i=1

Authorized licensed use limited to: Central South University. Downloaded on June 22,2022 at 10:39:34 UTC from IEEE Xplore. Restrictions apply.
1954 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 23, NO. 6, DECEMBER 2015

where
 
0 1
A1 = A2 = ,
f1m in 0
 
0 1
A3 = A4 =
f1m ax 0
 
0
B1 = B3 = ,
f2m in
 
0
B2 = B4 =
f2m ax
   
x1 (t) θ(t)
x(t) = = ,
x2 (t) θ̇(t)
 
x1 (t) ∈ 10.0078 18.4800
  Fig. 7. State responses of the system.
x2 (t) ∈ −0.0261 −0.1765
f1m ax − f1 (x(t)) of membership becomes uncertain due to uncertain mp and
μM̃ 1 1 (f1 (x(t))) = μM̃ 2 1 (f1 (x(t))) = Mc . Obviously, it is infeasible to apply the existing results on
f1m ax − f1m in
type-1 fuzzy system to design the fuzzy controller. However,
f1 (x(t)) − f1m in in order to compare with existing type-1 fuzzy control results,
μM̃ 3 1 (f1 (x(t))) = μM̃ 4 1 (f1 (x(t))) =
f1m ax − f1m in we should consider the certain grade of membership. By set-
f2m ax − f2 (x(t)) ting mp = mp min and Mc = Mc max , applying Lemma 1 with
μM̃ 1 2 (f2 (x(t))) = μM̃ 3 2 (f2 (x(t))) = a2 = 0.11, a type-1 PDC fuzzy state feedback controller could
f2m ax − f2m in
be designed to stabilize the system with the H∞ performance
f2 (x(t)) − f2m ax level γ = 0.0711. The state feedback controller gain matrices
μM̃ 2 2 (f2 (x(t))) = μM̃ 4 2 (f2 (x(t))) =
f2m ax − f2m in are
 
g − amp Lx2 (t)2 cos (x1 (t)) K1 = 380.5986 82.9699
f1 (x(t)) =  
4L/3 − amp L cos2 (x1 (t)) K2 = 746.2101 190.6029
 
sin (x1 (t))  
× K3 = 616.5455 145.6753
x1 (t)  
K4 = 857.2821 227.4987 .
−a cos (x1 (t))
f2 (x(t)) =
4L/3 − amp L cos2 (x1 (t)) In this example, from [25], the four-rule IT2 T–S fuzzy model
is obtained to describe the inverted pendulum system subject to
and the other considered system matrices is given by
parameter uncertainties, which is given in the following format:
 T
D11 = D12 = −0.5 0.1 Plant Rule i: IF x1 (t) is Wi1 and x1 (t) is Wi2 , THEN
 T ẋ(t) = Ai x(t) + Bi u(t) (75)
D13 = D14 = 0.5 −0.1
  where the system matrices are given in (72). The lower mem-
C1 = C2 = −0.01 0.03
  bership functions and upper membership functions of the plant
C3 = C4 = 0.03 −0.01 are defined in Table III.
The lower and upper grades of membership of rule i (i =
D21 = D23 = −0.01, D22 = D24 = −0.02. (73)
1, 2, 3, 4) are defined as follows:
Consider the following disturbance:
 θi (x1 (t)) = μW (f1 (x(t))) × μW (f2 (x(t)))
i1 i2
−1/(2 + t), 0 ≤ t ≤ 5
w(t) = (74) θ̄i (x1 (t)) = μ̄W i 1 (f1 (x(t))) × μ̄W i 2 (f2 (x(t))).
0, else.
Fig. 8 shows the membership functions θi (i = 1, 2, 3, 4) of the
If mp and Mc take the constant values, then the member- plant. The external disturbance input (74) in the pendulum is
ship functions of the T–S fuzzy system (72) are absolutely considered in this control procedure, and the other considered
known. The state responses of the inverted pendulum system system matrices are given in (73).
are depicted in Fig. 7 under an assumed initial state condi- Based on the closed-loop system in (75), similar to the
 5 T
tion x (0) = 12 π 0 , which indicates that the pendulum is procedure in Example 1, according to Theorem 1, assuming
not stable. From system (72), it can be found that the grade that σ1 = σ4 = 0.95, σ2 = σ3 = 0.5, Φ = 0, Ψ1 = −I, and

Authorized licensed use limited to: Central South University. Downloaded on June 22,2022 at 10:39:34 UTC from IEEE Xplore. Restrictions apply.
LI et al.: STATE AND OUTPUT FEEDBACK CONTROL OF INTERVAL TYPE-2 FUZZY SYSTEMS WITH MISMATCHED MEMBERSHIP FUNCTIONS 1955

TABLE III TABLE IV

LOWER AND UPPER MEMBERSHIP FUNCTIONS LOWER AND UPPER MEMBERSHIP FUNCTIONS OF THE CONTROLLER

Lower membership functions Upper membership functions Lower membership functions Upper membership functions

μW (f 1 (x(t))) = μ W (f 1 (x(t))) μ̄ W 1 1 (f 1 (x(t))) = μ̄ W 2 1 (f 1 (x(t))) x2

1
11 21 μM (x 1 ) = 0.3e ( − 0 . 3 5 ) μ M 1 1 (x 1 ) = μ M (x 1 )
11 11
−f 1 ( x ( t ) ) + f 1 m a x −f 1 ( x ( t ) ) + f 1 m a x
= f 1 m a x −f 1 m i n , μW (f 1 (x(t))) = f 1 m a x −f 1 m i n , μ̄ W 3 1 (f 1 (x(t)))
31 μM (x 1 ) = 0.3 − μ M (x 1 ) μ M 1 2 (x 1 ) = μ M (x 1 )
12 11 12
f 1 ( x ( t ) ) −f 1 m i n f 1 ( x ( t ) ) −f 1 m i n
= μW (f 1 (x(t))) = f 1 m a x −f 1 m i n = μ̄ W 4 1 (f 1 (x(t))) = f 1 m a x −f 1 m i n x2
41 1 )
μM (x 1 ) = 0.7e ( − 0 . 3 5 μ M 1 3 (x 1 ) = μ M (x 1 )
13 13
with x 2 (t) = 0, m p = m p m a x and with x 2 (t) = x 2 m a x , m p = m p m a x
μM (x 1 ) = 0.7 − μ M (x 1 ) μ M 1 4 (x 1 ) = μ M (x 1 )
14 13 14
M c = M c m in and M c = M c m i n

μW (f 2 (x(t))) = μ W (f 2 (x(t))) μ̄ W 1 2 (f 2 (x(t))) = μ̄ W 2 1 (f 2 (x(t)))

12 32

−f 2 ( x ( t ) ) + f 2 m a x −f 2 ( x ( t ) ) + f 2 m a x
= f 2 m a x −f 2 m i n , μW (f 2 (x(t))) = f 2 m a x −f 2 m i n , μ̄ W 3 1 (f 2 (x(t)))
31

f 2 ( x ( t ) ) −f 2 m i n f 2 ( x ( t ) ) −f 2 m i n
= μW (f 2 (x(t))) = f 2 m a x −f 2 m i n = μ̄ W 4 2 (f 2 (x(t))) = f 2 m a x −f 2 m i n
42

with m p = m p m a x and with m p = m p m i n and M c = M c m i n

Mc = Mcm ax

Fig. 9. Membership functions of the IT2 fuzzy controller.

Fig. 8. Membership functions of the IT2 fuzzy systems.

Ψ2 = 0. Then, we can obtain that the H∞ performance level

γm in = 0.0711 and the controller gains:
 
K1 = 103 × 1.0603 0.2413
 
K2 = 103 × 2.0408 0.4578 Fig. 10. Trajectories of x1 .
 
K3 = 103 × 2.0122 0.4567 IT2 fuzzy model can deal with the uncertainties in membership
 
K4 = 103 × 2.3213 0.5188 . functions well.
These graphs indicate that the designed IT2 fuzzy state feed-
The lower and upper membership functions of the controller are back controller can stabilize the inverted pendulum system in
defined in Table IV . Fig. 9 depicts the membership functions spite of under different initial states.
ηj (j = 1, 2, 3, 4) of the controller. Figs. 10 and 11 plot the On the other hand, when w(t) = 0, (t ≥ 0), under the con-
trajectories of the state responses of the closed-loop system trollers designed in this example, the trajectories of the state re-
with disturbance input under various initial conditions shown sponses of the closed-loop system are shown in Figs. 12 and 13
 5 T
in the graphs. From Figs. 10 and 11, we can obtain that the under the initial condition 12 π 0 . From Figs. 12 and 13,
IT2 fuzzy state feedback controller can stabilize the inverted it can be seen that the disturbance affects the stability of the
pendulum system better than the type-1 fuzzy controller, and the system, and the speed to be stable for the closed-loop system

Authorized licensed use limited to: Central South University. Downloaded on June 22,2022 at 10:39:34 UTC from IEEE Xplore. Restrictions apply.
1956 IEEE TRANSACTIONS ON FUZZY SYSTEMS, VOL. 23, NO. 6, DECEMBER 2015

V. CONCLUSION
In this paper, the problems of state and output feedback con-
trollers design have been solved for the IT2 T-S fuzzy system
with mismatched membership functions. Under a unified frame-
work, the IT2 fuzzy controllers have been designed for IT2
fuzzy systems based on a new performance index. In this new
performance index, H∞ , L2 –L∞ , passive, and dissipativity per-
formances are included. By using Lyapunov stability theory and
the convex optimization technique, the existence conditions of
the state and output feedback controllers have been expressed.
Two numerical examples have illustrated the effectiveness of the
proposed designed method. In future work, the actuator delay
and fault will be considered in the IT2 fuzzy systems, and the
Fig. 11. Trajectories of x2 . fault-tolerant controller will be designed for the systems with
actuator delay and fault.

REFERENCES
[1] L. Zadeh, “Fuzzy sets,” Inf. Control, vol. 8, no. 3, pp. 338–353, 1965.
[2] C.-T. Lin and C. S. G. Lee, “Neural-network-based fuzzy logic control and
decision system,” IEEE Trans. Comput., vol. 40, no. 12, pp. 1320–1336,
Dec. 1991.
[3] C.-T. Lin and C. G. Lee, “Reinforcement structure/parameter learning for
neural-network-based fuzzy logic control systems,” IEEE Trans. Fuzzy
Syst., vol. 2, no. 1, pp. 46–63, Feb. 1994.
[4] S. Tong, X.-L. He, and H.-G. Zhang, “A combined backstepping and
small-gain approach to robust adaptive fuzzy output feedback control,”
IEEE Trans. Fuzzy Syst., vol. 17, no. 5, pp. 1059–1069, Oct. 2009.
[5] T. Takagi and M. Sugeno, “Fuzzy identification of systems and its ap-
plications to modeling and control,” IEEE Trans. Syst., Man Cybern.,
vol. SMC-15, no. 1, pp. 116–132, Jan./Feb. 1985.
[6] J. Qiu, G. Feng, and H. Gao, “Asynchronous output-feedback control of
networked nonlinear systems with multiple packet dropouts: T-S fuzzy
Fig. 12. Trajectory of x1 . affine model-based approach,” IEEE Trans. Fuzzy Syst., vol. 19, no. 6,
pp. 1014–1030, Dec. 2011.
[7] M. Liu, X. Cao, and P. Shi, “Fuzzy-model-based fault-tolerant design
for nonlinear stochastic systems against simultaneous sensor and actuator
faults,” IEEE Trans. Fuzzy Syst., vol. 21, no. 5, pp. 789–799, Oct. 2013.
[8] G. Feng, “A survey on analysis and design of model-based fuzzy control
systems,” IEEE Trans. Fuzzy Syst., vol. 14, no. 5, pp. 676–697, Oct. 2006.
[9] L. Wu, X. Su, P. Shi, and J. Qiu, “A new approach to stability analysis
and stabilization of discrete-time T–S fuzzy time-varying delay systems,”
IEEE Trans. Syst., Man, Cybern. B, Cybern., vol. 41, no. 1, pp. 273–286,
Feb. 2011.
[10] Y.-H. Chang, C.-W. Chang, C.-L. Chen, and C.-W. Tao, “Fuzzy sliding-
mode formation control for multirobot systems: design and implemen-
tation,” IEEE Trans. Syst., Man, Cybern. B, Cybern., vol. 42, no. 2,
pp. 444–457, Apr. 2012.
[11] Y.-H. Chang, W.-S. Chan, and C.-W. Chang, “T-S fuzzy model-based
adaptive dynamic surface control for ball and beam system,” IEEE Trans.
Ind. Electron., vol. 60, no. 6, pp. 2251–2263, Jun. 2013.
[12] H. O. Wang, K. Tanaka, and M. F. Griffin, “An approach to fuzzy control
of nonlinear systems: Stability and design issues,” IEEE Trans. Fuzzy
Fig. 13. Trajectory of x2 . Syst., vol. 4, no. 1, pp. 14–23, Feb. 1996.
[13] J. M. Mendel, R. I. John, and F. Liu, “Interval type-2 fuzzy logic systems
made simple,” IEEE Trans. Fuzzy Syst., vol. 14, no. 6, pp. 808–821, Dec.
without disturbance is faster. Overall, these simulation results 2006.
[14] P. Melin and O. Castillo, “A new method for adaptive control of non-linear
show that our proposed control schemes are effective to control plants using type-2 fuzzy logic and neural networks,” Int. J. Gen. Syst.,
the uncertain nonlinear systems with desired performances. vol. 33, nos. 2/3, pp. 289–304, 2004.
Remark 6: Considering the simulation results in [37], and [15] D. Wu and W. W. Tan, “A type-2 fuzzy logic controller for the liquid-level
process,” in Proc. IEEE Int. Conf. Fuzzy Syst., 2004, vol. 2, pp. 953–958.
Figs. 7, 10, and 11, it can be seen that the IT2 fuzzy controller [16] O. Castillo and P. Melin, Type-2 Fuzzy Logic. New York, NY, USA:
designed from Theorem 1 can always stabilize the inverted pen- Springer, 2008.
dulum system for different initial status. In this example, the [17] R. Sepúlveda, O. Castillo, P. Melin, A. Rodrı́guez-Dı́az, and O. Montiel,
“Experimental study of intelligent controllers under uncertainty using
IT2 fuzzy system and fuzzy controllers do not share the same type-1 and type-2 fuzzy logic,” Inf. Sci., vol. 177, no. 10, pp. 2023–2048,
premise membership functions, that is to say, the controller can- 2007.
not be obtained based on the PDC concept. Thus, the stability [18] R. Martı́nez, O. Castillo, and L. T. Aguilar, “Optimization of interval type-
2 fuzzy logic controllers for a perturbed autonomous wheeled mobile robot
conditions proposed in [37] cannot be availed in this example. using genetic algorithms,” Inf. Sci., vol. 179, no. 13, pp. 2158–2174, 2009.

Authorized licensed use limited to: Central South University. Downloaded on June 22,2022 at 10:39:34 UTC from IEEE Xplore. Restrictions apply.
LI et al.: STATE AND OUTPUT FEEDBACK CONTROL OF INTERVAL TYPE-2 FUZZY SYSTEMS WITH MISMATCHED MEMBERSHIP FUNCTIONS 1957

[19] H. Hagras, “Type-2 FLCs: A new generation of fuzzy controllers,” IEEE Xingjian Sun received the B.S. degree in mathemat-
Comput. Intell. Maga., vol. 2, no. 1, pp. 30–43, Feb. 2007. ics from Bohai University, Jinzhou, China, in 2012,
[20] S. Barkat, A. Tlemçani, and H. Nouri, “Noninteracting adaptive control where he is currently working toward the M.S. degree
of PMSM using interval type-2 fuzzy logic systems,” IEEE Trans. Fuzzy in control theory with the School of Mathematics and
Syst., vol. 19, no. 5, pp. 925–936, Oct. 2011. Physics, Automation Research Institute.
[21] C.-F. Juang and C.-H. Hsu, “Reinforcement interval type-2 fuzzy con- His research interests include fuzzy control, robust
troller design by online rule generation and Q-value-aided ant colony control, and their applications.
optimization,” IEEE Trans. Syst., Man, Cybern. B, Cybern., vol. 39, no. 6,
pp. 1528–1542, Dec. 2009.
[22] C.-F. Juang and C.-Y. Chen, “An interval type-2 neural fuzzy chip with on-
chip incremental learning ability for time-varying data sequence prediction
and system control,” IEEE Trans. Neural Netwo. Learning Syst., vol. 25,
no. 1, pp. 216–228, Jan. 2014.
[23] H. A. Hagras, “A hierarchical type-2 fuzzy logic control architecture
for autonomous mobile robots,” IEEE Trans. Fuzzy Syst., vol. 12, no. 4,
pp. 524–539, Aug. 2004.
[24] E. A. Jammeh, M. Fleury, C. Wagner, H. Hagras, and M. Ghanbari, Ligang Wu (M’10–SM’12) received the B.S. degree
“Interval type-2 fuzzy logic congestion control for video streaming across in automation from the Harbin University of Science
IP networks,” IEEE Trans. Fuzzy Syst., vol. 17, no. 5, pp. 1123–1142, Oct. and Technology, Harbin, China, in 2001, the M.E.
2009. degree in navigation guidance and control from the
[25] H. K. Lam and L. D. Seneviratne, “Stability analysis of interval type-2 Harbin Institute of Technology, Harbin, in 2003, and
fuzzy-model-based control systems,” IEEE Trans. Syst., Man, Cybern. B, the Ph.D. degree in control theory and control engi-
Cybern., vol. 38, no. 3, pp. 617–628, Jun. 2008. neering from the Harbin Institute of Technology, in
[26] H. K. Lam, H. Li, C. Deters, H. Wuerdemann, E. Secco, and K. Althoefer, 2006.
“Control design for interval type-2 fuzzy systems under imperfect premise From January 2006 to April 2007, he was a Re-
matching,” IEEE Trans. Ind. Electron., vol. 61, no. 2, pp. 956–968, Feb. search Associate with the Department of Mechan-
2014. ical Engineering, The University of Hong Kong,
[27] J. Cao, P. Li, and H. Liu, “An interval fuzzy controller for vehicle active Hong Kong. From September 2007 to June 2008, he was a Senior Research
suspension systems,” IEEE Trans. Intell. Transp. Syst., vol. 11, no. 4, Associate with the Department of Mathematics, City University of Hong Kong,
pp. 885–895, Dec. 2010. Hong Kong. From December 2012 to December 2013, he was a Research As-
[28] B. Zhang, W. X. Zheng, and S. Xu, “Filtering of Markovian jump delay sociate with the Department of Electrical and Electronic Engineering, Imperial
systems based on a new performance index,” IEEE Trans. Circuits Syst. I, College London, London, U.K. In 2008, he joined the Harbin Institute of Tech-
Reg. Papers, vol. 60, no. 5, pp. 1250–1263, May 2013. nology as an Associate Professor, and was then promoted to a Professor in
[29] B. Chen, X. Liu, C. Lin, and K. Liu, “Robust H ∞ control of Takagi– 2012. He has published more than 100 research papers in international referred
Sugeno fuzzy systems with state and input time delays,” Fuzzy Sets Syst., journals. He is the author of the monographs Sliding Mode Control of Uncertain
vol. 160, no. 4, pp. 403–422, 2009. Parameter-Switching Hybrid Systems (New York, NY, USA: Wiley, 2014), and
[30] Y. Chen, A. Xue, and S. Zhou, “New delay-dependent L 2 –L ∞ filter Fuzzy Control Systems with Time-Delay and Stochastic Perturbation: Analysis
design for stochastic time-delay systems,” Signal Process., vol. 89, no. 6, and Synthesis (New York, NY, USA: Springer, 2015). His current research in-
pp. 974–980, 2009. terests include switched hybrid systems, computational and intelligent systems,
[31] S. Xu, W. X. Zheng, and Y. Zou, “Passivity analysis of neural networks sliding mode control, optimal filtering, and model reduction.
with time-varying delays,” IEEE Trans. Circuits Syst. II, Exp. Briefs, Dr. Wu currently serves as an Associate Editor for a number of journals, in-
vol. 56, no. 4, pp. 325–329, Apr. 2009. cluding IEEE TRANSACTIONS ON AUTOMATIC CONTROL, Information Sciences,
[32] Z. Li, J. Wang, and H. Shao, “Delay-dependent dissipative control for Signal Processing, and IET Control Theory and Applications. He is also an
linear time-delay systems,” J. Franklin Inst., vol. 339, no. 6, pp. 529–542, Associate Editor for the Conference Editorial Board, IEEE Control Systems
2002. Society.
[33] R. Lozano, B. Brogliato, O. Egeland, and B. Maschke, Dissipative Systems
Analysis and Control: Theory and Applications. London, U.K.: Springer,
2000.
[34] Z. Feng, J. Lam, and H. Gao, “α-dissipativity analysis of singular time-
delay systems,” Automatica, vol. 47, no. 11, pp. 2548–2552, 2011.
[35] Z. Feng and J. Lam, “Reliable dissipative control for singular Markovian
systems,” Asian J. Control, vol. 15, no. 3, pp. 901–910, 2013.
[36] K. M. Grigoriadis and J. T. Watson, Jr,“Reduced-order H ∞ and L 2 -L ∞ H. K. Lam (M’98–SM’10) received the B.Eng.
filtering via linear matrix inequalities,” IEEE Trans. Aerosp. Electron. (Hons.) and Ph.D. degrees from the Department of
Syst., vol. 33, no. 4, pp. 1326–1338, Oct. 1997. Electronic and Information Engineering, The Hong
[37] H. K. Lam and F. H. F. Leung, “Stability analysis of discrete-time fuzzy- Kong Polytechnic University, Hong Kong, in 1995
model-based control systems with time delay: Time delay-independent and 2000, respectively.
approach,” Fuzzy Sets Syst., vol. 159, pp. 990–1000, 2008. From 2000 to 2005, he was a Postdoctoral Fellow
and a Research Fellow with the Department of Elec-
tronic and Information Engineering, The Hong Kong
Hongyi Li received the B.S. and M.S. degrees in Polytechnic University. In 2005, he joined King’s
mathematics from Bohai University, Jinzhou, China, College London, London, U.K., as a Lecturer, where
in 2006 and 2009, respectively, and the Ph.D. de- he is currently a Reader. His current research inter-
gree in intelligent control from the University of ests include intelligent control systems and computational intelligence. He has
Portsmouth, Portsmouth, U.K., in 2012. served as a program committee member and international advisory board mem-
He is currently a Professor with the College of ber for various international conferences and a reviewer for various books,
Engineering, Bohai University. He was a Research international journals, and international conferences.
Associate with the Department of Mechanical En- Dr. Lam is an Associate Editor for the IEEE TRANSACTIONS ON FUZZY
gineering, University of Hong Kong and Hong Kong SYSTEMS and International Journal of Fuzzy Systems, and a Guest Editor for
Polytechnic University, from June 2010 to September a number of international journals. He is in the editorial boards of a number
2010 and from September 2012 to December 2012, of journals including IET Control Theory and Applications. He is the coedi-
respectively. His research interests include fuzzy control, robust control, and tor for two edited volumes: Control of Chaotic Nonlinear Circuits (Singapore:
their applications. World Scientific, 2009) and Computational Intelligence and Its Applications
Dr. Li is an Associate Editor/Editorial Board member for several interna- (Singapore: World Scientific, 2012), and the coauthor of the book Stability Anal-
tional journals, including IEEE Transactions on Neural Networks and Learning ysis of Fuzzy-Model-Based Control Systems (New York, NY, USA: Springer,
Systems, Neurocomputing, Circuits, Systems, and Signal Processing, Shock and 2011).
Vibration, etc.

Authorized licensed use limited to: Central South University. Downloaded on June 22,2022 at 10:39:34 UTC from IEEE Xplore. Restrictions apply.