Received: 10 January 2017 Revised: 30 November 2017 Accepted: 4 December 2017

DOI: 10.1002/acs.2854

RESEARCH ARTICLE

Adaptive sliding mode control for fuzzy singular systems

with time delay and input nonlinearity

Mourad Kchaou1,2 Hamdi Gassara2 Ahmed El-Hajjaji3

1
Department of Electrical Engineering,
College of Engineering, University of Hail, Summary
Hail 2440, Saudi Arabia This paper addresses the sliding mode control problem for a class of uncer-
2
Laboratory of Sciences and Techniques of
tain Takagi-Sugeno fuzzy singular systems with state delay and subject to input
Automatic Control and Computer
Engineering, National Engineering School nonlinearity. Our purpose is focused on designing an adaptive sliding mode
of Sfax, University of Sfax, 3038 Sfax, controller for such a complex system. First, a new fuzzy integral-type sliding
Tunisia
function is designed. Then, an adaptive sliding mode control scheme is estab-
3
Modeling, Information, and Systems
Laboratory, University of Picardie Jules
lished such that the resulting closed-loop system is insensitive to all admissible
Verne, Amiens 80000, France uncertainties and satisfies the reaching condition. Moreover, delay-dependent
sufficient conditions are derived such that the admissibility and the L2 -L∞ per-
Correspondence
Mourad Kchaou, Department of Electrical formance requirement of the sliding mode dynamics can be guaranteed in the
Engineering, College of Engineering, presence of time delays, external disturbances, and sector nonlinearity input.
University of Hail, Hail 2440, Saudi
Finally, the validity and applicability of the proposed theory are illustrated by a
Arabia; or Laboratory of Sciences and
Techniques of Automatic Control and numerical example.
Computer Engineering, National
Engineering School of Sfax, University of K E Y WO R D S
Sfax, 3038 Sfax, Tunisia. adaptive sliding mode control, fuzzy singular systems, LMI, time delay
Email: [email protected]

1 I N T RO DU CT ION
A singular model, also known as a descriptor model, is a mathematical representation that can provide a convenient and
natural description of several physical plants. Singular space-state systems are a generalization of standard space-state
systems, which are composed of ordinary differential equations, describing the dynamical part of the physical process
and algebraic equations to include the interrelationships between the different components in the system.
In studying singular systems, the system regularity and absence of impulses or causality (for discrete systems) must
be guaranteed.1,2 We emphasize that the delays often occur between different components of many practical systems.3,4
Thus, the study of singular systems with time delays becomes an extremely important topic of control engineering. In
view of the generality of singular models and the time-delay phenomenon, many fundamental control problems have
gained much research attention and many relevant results have been reported (see, for example, other works2,5-9 and the
references therein).
It becomes increasingly apparent that fuzzy control provides appealing advantages in several applications. As a pow-
erful method to study nonlinear systems, the Takagi-Sugeno (TS) model–based fuzzy control has become a widespread
approach to deal with complex nonlinear systems.10,11 Based on the TS fuzzy model, there has been considerable research
work appearing to address the control problem of nonlinear singular systems in the presence of time delays.12-17
On a different research front, sliding mode control (SMC) is one of different robust control schemes used to cope with
model uncertainties and nonlinearities by taking advantage of the concepts of sliding mode surface design and equivalent
control.18,19

Int J Adapt Control Signal Process. 2018;1–16. wileyonlinelibrary.com/journal/acs Copyright © 2018 John Wiley & Sons, Ltd. 1
2 KCHAOU ET AL

Sliding mode control, considered as variable structure control, uses a discontinuous control to drive and then constrain
the state trajectories to lie within a neighborhood of a specific switching surface on which the system meets the required
control specifications. The key feature and advantages of the SMC approach include (1) insensitivity to variations of uncer-
tainties, (2) disturbance rejection capability, and (3) tracking ability. During the past decades, various SMC approaches
have been successfully applied for solving many practical control problems.
The integral SMC has been addressed for uncertain systems in the works of Ding et al20 and Chang21 and for stochas-
tic systems in the works of Gao et al22 and Wu and Ho.23 Recently, the second-order integral SMC approach has been
investigated for uncertain systems with control input delay in the work of Zhang et al.24 The problem of SMC of the TS
fuzzy singular system with time-delay was considered in our other work25 and in the work of Wang and Fei26 with the
assumption that the local input matrices are constant, ie, Bi = B.
On the other hand, most of the proposed approaches for SMC are based on the assumption that the systems under
consideration are characterized by linear inputs. However, in practice, due to the physical limitation, the control input
seems to have a nonlinear character such as sectors, saturation, and deadzone. Therefore, the input nonlinearities can
deteriorate the systems' performances and their effects must be taken into account in analyzing and designing any control
scheme. Recently, attention has been paid to input nonlinearity,27-31 but few works have been undertaken for singular
systems.
Enlightened by the aforementioned reasons, it is necessary to develop a new robust control strategy to deal with
nonlinear singular systems where the real factors such as time delay, input nonlinearity, and external disturbance are
unavoidable and often disrupt the desired performances.
In this paper, the TS fuzzy approach is adopted to investigate the SMC problem for a class nonlinear singular systems
in the presence of the aforementioned factors. Precisely, the main contributions can be summarized as follows:
• consideration of a nonrestrictive TS fuzzy singular system where the local input matrices are different and the input is
nonlinear;
• proposition and construction of a new sliding function and establishing sufficient conditions to ensure the robust
admissibility and the L2 -L∞ performance of the corresponding sliding mode dynamics by means of the feasibility of a
convex optimization problem;
• design of a sliding mode controller for the reaching motion such that trajectories of the resulting closed-loop system
can be driven onto a prescribed sliding surface and maintained there for all subsequent times.
The remainder of this paper is organized as follows. In Section 2, system description and preliminaries are presented. The
main results are developed in Section 3. In Section 4, a numerical example is provided to show the effectiveness of the
proposed scheme, and the conclusions are drawn in Section 5.
Notation. The notations in this paper are quite standard except where otherwise stated. The superscript T stands
for matrix transposition; X ∈ Rn denotes the n-dimensional Euclidean space, whereas X ∈ Rn×m refers to the set
of all n × m real matrices; X > 0 (respectively, X ≥ 0) means that the matrix X is real symmetric positive definite
(respectively, positive semidefinite); l2 [0, ∞) is the space of square summable vectors; 𝜆min (A) and 𝜆max (A) denote the
maximum and minimum eigenvalues of a matrix A, respectively; I and 0 represent the identity matrix and a zero
matrix with appropriate dimension, respectively; diag· · · stands for a block-diagonal matrix, sym(X) stands for X + XT ;
||.|| denotes the Euclidean norm of a vector and its induced norm of a matrix. In symmetric block matrices or long
matrix expressions, we use a star ∗ to represent a term that is induced by symmetry. Matrices, if their dimensions
are not explicitly stated, are assumed to be compatible for algebraic operations. To avoid clutter, in the following, 𝜇i
denotes 𝜇 i (𝜃).

2 SYSTEM D ESCRIPTION AND PRELIMINARIES

In this study, we consider a nonlinear singular system with state delay described by the following TS fuzzy model:
Ri ∶ If 𝜃1 (t) is F1i and if 𝜃2 (t) is F2i · · · if 𝜃s (t) is Fsi , then
⎧ ̇ = Ai (t)x(t) + Ahi (t)x (t − h(t)) + Bi (𝜙 (u(t)) + 𝑓i (t, x(t))) + Bwi w(t)
Ex(t)
⎪
⎨ z(t) = Ci x(t) + Chi x (t − h(t)) (1)
⎪
⎩ x(t) = 𝜑(t), t ∈ [−hM , 0], i = 1, 2, … , r,
KCHAOU ET AL 3

where x ∈ Rn is the state and u(t) = [u1 , u2 , … , um ]T ∈ Rm is the control input. 𝜙(u) = [𝜙1 (u1 ), 𝜙2 (u2 ), … , 𝜙m (um )]T ,
where 𝜙l (ul ) is a continuous nonlinear function for l = 1, 2, … , m. w(t) ∈ Rw is the external disturbance input, fi (t, x(t))
represents the system nonlinearity and any model uncertainties in the system including external disturbances, z(t) ∈ Rs is
the controlled output, F𝑗i ( j = 1 … s) are fuzzy sets, 𝜃(t) = [𝜃 1 (t), … , 𝜃 s (t)] is the premise variable vector. It is assumed that
the premise variables do not depend on the input variables u(t), which is needed to avoid a complicated defuzzification
process of fuzzy controllers. Delay h(t) is time varying and satisfies
hm ≤ h(t) ≤ hM , ̇ ≤ hd ,
h(t) (2)
where hm and hM are constants representing the bounds of the delay and hd is a positive constant. 𝜑(t) is a compatible
vector-valued initial function in [−hM , 0] representing the initial condition of the system. The system disturbance, w(t),
∞
is assumed to belong to L2 [0, ∞), ie, ∫0 wT (t)w(t)dt < ∞. This implies that the disturbance has finite energy. Matrix
E ∈ Rn×n may be singular with rank (E) = q < n. Ai (t) = Ai + Δ Ai and Ahi = Ahi + Δ Ahi are time-varying system matrices.
Matrices Ai , Ahi , Bi , Bwi , Ci , and Chi are constant with appropriate dimensions, and Bi is of full column rank.
The following assumptions are made in this study.
Assumption 1.
1. Δ Ai (t) and Δ Ahi (t) are real matrices representing norm-bounded parameter uncertainties and satisfy
[ ] [ ]
Δ Ai (t) Δ Ahi (t) = Mi F(t) Ni Nhi , (3)
where Mi , Ni , and Nhi are known real constant matrices and F(t) is unknown time-varying matrix function
satisfying F T (t)F(t) ⩽ I.
2. Nonlinear function fi (t, x(t)), or the so-called matched uncertainty, satisfies
𝑓i (t, x(t)) ≤ 𝜌i ||x(t)||, (4)
where 𝜌i is a positive unknown constant.
̄
3. Exogenous signal w(t) is bounded by upper bound w
̄
||w(t)|| ≤ w. (5)
4. Nonlinear input 𝜙(u) applied to the system satisfies 𝜙(0) = 0 and
uT 𝜙(u) ≥ 𝛼uT u, (6)
where 𝛼 is a positive constant.
5. Matrix pair [Ai , Bi ] is controllable for all i = 1, 2, … , r.

Remark 1. As in our other work25 and in the work of Wang and Fei,26 it is assumed that Bi = B, i = 1, 2, … , r. This
restrictive assumption is not considered in this present study.
The overall fuzzy model is inferred as follows:
⎧ ∑ r
⎪Ex(t)
̇ = 𝜇i (𝜃(t)) {Ai (t)x(t) + Ahi (t)x(t − h(t)) + Bi (𝜙 (u(t)) + 𝑓i (t, x(t))) + Bwi w(t)}
⎪ i=1
⎨ ∑r (7)
⎪z(t) = 𝜇i (𝜃(t)) {Ci x + Chi x (t − h(t))} ,
⎪
⎩ i=1

where 𝜇 i (𝜃(t)) is the normalized membership function defined by

∏
s
( )
F𝑗i 𝜃𝑗 (t)
𝑗=1
𝜇i (𝜃(t)) = , i = 1, 2, … , r
∑∏
s r
( )
F𝑗i 𝜃𝑗 (t)
i=1 𝑗=1

and F𝑗i (𝜃𝑗 (t)) represents the membership degrees of 𝜃 j (t) in fuzzy set F𝑗i . Note that normalized membership 𝜇 i (𝜃(t)) satisfies
∑
r
𝜇i (𝜃(t)) ≥ 0, i = 1, 2, … , r 𝜇i (𝜃(t)) = 1. (8)
i=1
4 KCHAOU ET AL

Before stating the main results, we need to introduce some definitions and some lemmas, which will be used in the
next section.
Consider the following unforced linear singular system with time delay:

Eẋ = Ax(t) + Ah x(t − h(t)), 0 ≤ h(t) ≤ hM x(t) = 𝜑(t), t ∈ [−hM , 0]. (9)

Definition 1. (See the work of Dai1 )

1. System (9) is said to be regular if det(sE − A) ≠ 0.
2. System (9) is said to be impulse free if deg(det(sE − A)) = rank(E).
3. System (9) is said to be admissible if it is regular, impulse free, and stable.
Singular time-delay system (9) may have an impulsive solution. However, the regularity and nonimpulse of (E, A)
guarantee the existence and uniqueness of an impulse-free solution to (9) on [0, ∞).
Lemma 1. (See the work of Gu et al4 )
For any constant matrix M > 0, any scalar hm and hM with 0 < hm < hM , and vector function x(t) ∶ [−hM , −hm ] → Rn
such that the integrals concerned are well defined, then the following holds:
t−hm t−hm t−hm
−hr xT (s)Mx(s)ds ≤ − xT (s)dsM x(s)ds
∫t−hM ∫t−hM ∫t−hM

with hr = hM − hm .

Lemma 2. (See the work of Petersen32 )

Let M, N, Δ be matrices with Δ satisfying ΔT Δ ≤ I and scalar 𝜀 > 0. Then, the following inequality holds:

sym(MΔN) ≤ 𝜀MM T + 𝜀−1 N T N. (10)

3 MAIN RESULT

In this section we aim to design a sliding mode controller for the aforementioned fuzzy singular system (7). The SMC
design involves 2 basic steps. First, we design a fuzzy integral sliding surface as a function of the system state such that
the sliding mode dynamics restricted to the surface is admissible and satisfies the L2 -L∞ performance. The second step is
to synthesize a suitable sliding mode controller to globally drive the system state trajectories to the predefined switching
surface and maintain it there for all subsequent time.

3.1 Integral sliding mode surface

Choose the following fuzzy switching function:

Ri ∶ If 𝜃1 (t) is F1i and if 𝜃2 (t) is F2i · · · if 𝜃s (t) is Fsi , then

t
si (t) = Gi (E x(t) − E x(0)) − Gi (Ai + Bi Ki ) x (𝜏) + Ahi x (𝜏 − h(𝜏)) d𝜏, (11)
∫0

where Gi ∈ Rm×n is a constant matrix satisfying Gi Bi is nonsingular, and Ki ∈ Rm×n is a real matrix to be designed.
Then, the overall sliding surface can be described as

∑
r
s(t) = 𝜇i si (t). (12)
i=1

According to the SMC theory, when the system trajectories reach onto the switching surface, we have ṡ i (t) = 0. Then, the
ith equivalent control law can be obtained as follows:

𝜙e (u) = −Ki x(t) − 𝑓i (x(t)) − (Gi Bi )−1 Gi (Δ Ai x(t) + Δ Adi (t) x (t − h(t)) + Bwi w(t)) . (13)
KCHAOU ET AL 5

Substituting (13) into (1), we obtain the following ith fuzzy sliding mode dynamics:
Ri ∶ If 𝜃1 (t) is F1i and if 𝜃2 (t) is F2i · · · if 𝜃s (t) is Fsi , then
{
Eẋ = Ai (t)x(t) + Ahi (t)x(t) + Bwi w(t)
(14)
z(t) = Ci x(t) + Chi x (t − h(t)) ,
and the overall sliding mode dynamics can be described by
⎧ ∑ r { }
⎪Ex(t)x
̇ = 𝜇i Ai (t)x(t) + Ahi (t)x(t) + Bwi w(t)
⎪ i=1
⎨ ∑r (15)
⎪z(t) = 𝜇i {Ci x(t) + Chi x (t − h(t))} ,
⎪
⎩ i=1

where
Gi = I − Bi (Gi Bi )−1 Gi , Ai (t) = Ai + Δ Ā i , Ai = Ai + Bi Ki , Ahi = Ahi + Δ Ahi , (16)
[ ] [ ]
Bwi = Gi Bwi , M i = G i Mi , Δ Ai Δ Ahi = M i F(t) Ni Nhi .

3.2 L2 -L∞ sliding mode dynamics analysis

In this subsection, we will develop a sufficient delay-dependent condition that ensures for sliding mode dynamics (15) to
be robustly admissible with L2 -L∞ performance.
Definition 2. Sliding mode dynamics (15) is said to be admissible with L2 -L∞ performance if it is admissible and for
prescribed positive scalar 𝛾, the following norm is satisfied under a zero initial condition:
||z(t)||∞
||Tzw (s)||L2 -L∞ = sup < 𝛾. (17)
0≠w(t)∈L2 ||w(t)||2

Theorem 1. Let hm , hM , and hd be given positive scalars. Sliding mode dynamics of (15) is admissible with L2 -L∞ norm
bound 𝛾, if there exist matrices R, P > 0, Q1 > 0, Q2 > 0, Q3 > 0, Z1 > 0, Z2 > 0, Z3 > 0, and Gj , ( j = 1, … , 5) and
positive scalars 𝛾, 𝜀, and 𝜖 such that the following inequalities hold:
⎡ Φi + sym(Φ1 ) hr X Γ1 𝜖Γ2 ⎤
⎢ ∗ −hr Z3 0 0 ⎥
Ψ1i (E, Ai , Ahi , Bwi , X, G𝑗 ) = ⎢ ∗ ∗ −𝜖I 0 ⎥ < 0 (18)
⎢ ∗ ∗ ∗ −𝜖I ⎥⎦
⎣

⎡ Φi + sym(Φ1 ) hr Y Γ1 𝜖Γ2 ⎤
⎢ −hr Z3 0 0 ⎥
Ψ2i (E, Ai , Ahi , Bwi , Y , G𝑗 ) = ⎢ ∗ ∗ −𝜖I 0 ⎥ < 0 (19)
⎢ ∗ ∗ ∗ −𝜖I ⎥⎦
⎣
⎡ −ET P ∗ CiT ⎤
Ψ3i (E, Ci , Chi ) = ⎢ ∗ −𝜀I ChiT ⎥ < 0, (20)
⎢ ⎥
⎣ ∗ ∗ −𝛾I ⎦
where
⎡ Φ11i Φ12i Φ13i Φ14i Φ15i G1 B̄ wi ⎤
⎢ ∗ Φ22i Φ23i Φ24i Φ25i G2 B̄ wi ⎥
⎢ ∗ ∗ Φ33i 0 −G3 G3 B̄ wi ⎥
Φi = ⎢
∗ ∗ ∗ Φ44i −G4 G4 B̄ wi ⎥
⎢ ⎥
⎢ ∗ ∗ ∗ ∗ Φ55i G5 B̄ wi ⎥
⎣ ∗ ∗ ∗ ∗ ∗ −𝛾I ⎦
[ T ]T [ T T T T T ]T
Γ1 = G 0 M i , G = G1 G2 G3 G4 G5 ,
[ ]T
Γ2 = Ni Nhi 0 0 0 0
6 KCHAOU ET AL

T
Φ11i = Q1 + Q2 + Q3 + sym(G1 Ai ) − ET Z1 E − ET Z2 E, Φ12i = Ai GT1 + G2 Ahi
T T
Φ13i = ET Z1 E + Ai GT3 , Φ14i = ET Z2 E + Ai GT4
T
Φ15i = P + SRT + Ai GT5 − G1 , Φ22i = −(1 − hd )Q2 + sym(G2 Ahi )
Φ23i = AThi GT3 , Φ24i = AThi GT4
Φ25i = −G2 + AThi GT5 , Φ33i = −Q1 − ET Z1 E
Φ44i = −Q3 − ET Z2 E, Φ55i = −sym(G5 ) + h2m Z1 + h2M Z2 + hr Z3 .
[ ]
Φ1 = 0 XE − YE YE −XE 0

Proof. The proof of this theorem is divided into 2 parts. The first one is concerned with the regularity and the
impulse-free characterizations, and the second one treats the stability property of system (15).
First, we consider the nominal case of (15) (ie, Δ Ai (t) = 0, Δ Ahi (t) = 0, and Δ Bwi (t) = 0).
Using the following notation:
∑
r
∑
r
∑
r
∑
r
∑
r
A= 𝜇iĀ i Ah = 𝜇i Ahi Bw = 𝜇i Bwi C= 𝜇i Ci , Ch = 𝜇i Chi , (21)
i=1 i=1 i=1 i=1 i=1

system (15) can be written as

⎧Ex(t)
̇ = A x(t) + Ah x(t − h(t)) + Bw w(t)
⎪
⎨z(t) = C(t)x(t) + Ch x(t − h(t)) (22)
⎪
⎩x(t) = 𝜑(t), t ∈ [−hM , 0].
Since rank(E) = q ≤ n, there always exist 2 nonsingular matrices M and N ∈ Rn × n such that
[ ]
Iq 0
Ê = MEN = . (23)
0 0
[ ]
0 ̂ ∈ R(n−q)×(n−q) is any nonsingular matrix.
Then, R can be characterized as R = MT ̂ , where Φ
Φ
We also define
[ ] [ ]
̂ ̂ ̂
̂A = MAN = A11 A12 , ̂S = NT S = S11 ,
 21  22 Ŝ 21
[ ] [ ]
P̂ P̂ Ẑ Ẑ
P̂ = M−T PM−1 = ̂ 11 ̂ 12 , Ẑ l = M−T Zl M−1 = ̂ 11l ̂ 12l , l = 1, 2. (24)
P21 P22 Z21l Z22l
It follows from (18) that
[ ]
𝜙11i 𝜙12i
∗ 𝜙22i < 0, (25)

where

𝜙11i = sym(G1 Ai ) − ET Z1 E − ET Z2 E
T
𝜙12i = P + SRT + Ai GT5 − G1
𝜙22i = − sym(G5 ).
T
Premultiplying and postmultiplying (25) by [ I Ai ] and its transpose, respectively, we obtain
( )
sym SRT Ai + PAi − ET Z1 E − ET Z2 E < 0. (26)
∑r
Since 𝜇i (𝜃) ≥ 0 and i=1 𝜇i (𝜃) = 1, it yields
( )
sym SRT A + PA − ET Z1 E − ET Z2 E < 0. (27)
KCHAOU ET AL 7

Premultiplying and postmultiplying (27) by NT and N, respectively, and then using expressions (23) and (24) yield
( )
sym Ŝ 21 Φ
̄ T Â 22 < 0, (28)

and A22 is thus nonsingular. Hence, according to Definition 1, singular time-delay system (22) is regular and impulse
free for any time delay h(t) satisfying (2).
In the following, we will prove that system (22) is asymptotically stable. For this purpose, we construct a candidate
Lyapunov functional as

V(xt ) = V1 (xt ) + V2 (xt ) + V3 (xt )

V1 (xt ) = xt (t)PT Ex(t)
t t xt
V2 (xt ) = xT (s)Q1 x(s)ds + xT (s)Q2 x(s)ds + xT (s)Q3 x(s)ds (29)
∫t−hm ∫t−h(t) ∫t−hM
0 t 0 t
V3 (xt ) = hm ẋ T (s)ET Z1 Ex(s)dsd𝜃
̇ + hM ẋ T (s)ET Z2 Ex(s)dsd𝜃
̇
∫−hm ∫t+𝜃 ∫−hM ∫t+𝜃
−hm t
+ ẋ T (s)ET Z3 Ex(s)dsd𝜃.
̇
∫−hM ∫t+𝜃

Evaluating the derivative of V(xt ) along the solutions of system (22), it yields

V̇ 1 (xt ) = 2xT (t)PT Ex(t)

̇
̇V2 (xt ) ≤ xT (t)(Q1 + Q2 + Q3 )x(t) − xT (t − hm )Q1 x(t − hm ) − (1 − hd )xT (t − h(t))Q2 x(t − h(t))
− xT (t − hM )Q3 x(t − hM )
t
V̇ 3 (xt ) = h2m ẋ T (t)ET Z1 Ex(t)
̇ + h2M ẋ T (t)ET Z2 Ex(t)
̇ + hr ẋ T (t)ET Z3 Ex(t)
̇ − hm ẋ T (s)ET Z1 Ex(s)ds
̇
∫t−hm
t t−hm
− hM ẋ T (s)ET Z2 Ex(s)ds
̇ − ẋ T (s)ET Z3 Ex(s)ds.
̇ (30)
∫t−hM ∫t−hM

From Lemma 1, one can obtain

t [ ]T [ ]
− hm ẋ T (s)ET Z1 Ex(s)ds
̇ ≤ − x(t) − x(t − hm ) ET Z1 E x(t) − x(t − hm )
∫t−hm
t [ ] [ ]
− hm ẋ T (s)ET Z2 Ex(s)ds
̇ ≤ − x(t) − x(t − hM )T ET Z2 E x(t) − x(t − hM )
∫t−hM
(31)

and
t−hm t−h(t) t−hm
− ẋ T (s)ET Z3 Ex(s)ds
̇ =− ẋ T (s)ET Z3 Ex(s)ds
̇ − ẋ T (s)ET Z3 Ex(s)ds.
̇ (32)
∫t−hM ∫t−hM ∫t−h(t)
[ ]T
Defining 𝜉(t) = xT (t) xT (t − h(t)) xT (t − hm ) xT (t − hM ) ẋ T (t)ET , for any appropriately dimensioned matrix X, the
following inequality holds:
t−h(t) [ ]T [ −1 T ][ ]
𝜉(t) XZ3 X X 𝜉(t)
ds ≥ 0. (33)
∫t−hM ̇
Ex(s) ∗ Z3 ̇
Ex(s)
Then, it is easy to verify
t−h(t)
− ẋ T (s)ET Z3 Ex(s)
̇ ds ≤ (hM − h(t)) 𝜉 T (t) XZ3−1 X T 𝜉(t) + 2𝜉 T (t) XE (x(t − h(t)) − x(t − hM )). (34)
∫t−hM

Similarly, for any matrix Y, we obtain

t−hm
− ẋ T (s)ET Z3 Ex(s)ds
̇ ≤ (h(t) − hm ) 𝜉 T (t)YZ3−1 Y T 𝜉(t) + 2𝜉 T (t)YE(x(t − hm ) − x(t − h(t)). (35)
∫t−h(t)
8 KCHAOU ET AL

From (34)-(35), it can be seen that

t−hm {
− ẋ T (s)ET Z3 Ex(s)ds
̇ ≤ 𝜉 T (t) 𝜌(t)hr XZ3−1 X T (1 − 𝜌(t))hr YZ3−1 Y T (36)
∫t−hM
[ ]}
+ 2 0 XE − YE YE −XE 0 𝜉(t),
h −h(t)
where 𝜌(t) = M h .
r
From (22), the following equation holds for any matrices Gj , j = 1, 2, … , 5 with the appropriate dimensions
̇ + A x(t) + Ah x(t − h(t))] = 0.
2𝜉 T (t)G[−Ex(t) (37)
Besides, it is clear that
̇ = 0.
2xT (t)SRT Ex(t) (38)
Considering (30)-(38), we obtain
∑
r
( )
̇ t ) ≤ 𝜉 T (t)
V(x 𝜇i 𝜌(t)Ψ ̄ 2i 𝜉(t),
̄ 1i + (1 − 𝜌(t))Ψ (39)
i=1

where
Ψ1i = Φi + sym(Φ1 ) + hr XZ3−1 X T (40)

Ψ2i = Φi + sym(Φ1 ) + hr YZ3−1 Y T (41)

and
⎡ Φ11i Φ12i Φ13i Φ14i Φ15i ⎤
⎢ ∗ Φ22i Φ23i Φ24i Φ25i ⎥
Φi = ⎢ ∗ ∗ Φ33i 0 −G3 ⎥ (42)
⎢ ∗ ∗ ∗ Φ44i −G4 ⎥
⎢ ∗ Φ55i ⎥⎦
⎣ ∗ ∗ ∗
since 0 ≤ 𝜌(t) ≤ 1, 𝜌(t)Ψ1i + (1 − 𝜌(t))Ψ2i is a convex combination of Ψ1i and Ψ2i . If (18)-(19) are satisfied, then by
applying the Schur complement, we can verify and obtain that 𝜌(t)Ψ1i + (1 − 𝜌(t))Ψ2i < 0. Hence, V(x ̇ t ) ≤ −𝜆||𝜉(t)||2 ,
which implies that nominal singular system (22) with w(t) = 0 is asymptotically stable. Let us now prove that the
system has the L2 -L∞ performance. For this purpose, consider the following performance index:
t
J0 = V(xt ) − 𝛾 wT (s)w(s)ds, (43)
∫0
where V(xt ) is defined as in (29). For any nonzero w(s) ∈ L2 , t > 0 and zero initial state condition 𝜑(t) = 0, t ∈ [−hM , 0],
we have
t t
J0 = V(xt ) − V(0) − 𝛾 wT (s)w(s)ds = ̇ s ) − 𝛾wT (s)w(s)ds.
V(x (44)
∫0 ∫0
[ ]T
Define 𝜁 (t) = 𝜉 T (t) wT (t) . The following null equation holds:
2𝜁 T [GT 0]T × [−Eẋ + A x(t) + Ah x(t − h(t)) + Bw w(t)] = 0. (45)
By following the same procedure as used above, we can verify that
∑
r ( )
̂ 1i + (1 − 𝜌(t))Ψ
̇ t ) − 𝛾wT (t)w(t) ≤ 𝜁 T (t) 𝜇i 𝜌(t)Ψ
V(x ̂ 2i 𝜁 (t), (46)
i=1

where Ψ ̂ 1i = Φi +sym(Φ1 )+hr XZ −1 X T and Ψ

̂ 1i = Φi +sym(Φ1 )+hr YZ −1 Y T . Using the Schur complement equivalence
3 3
of (18) and (19), we can verify that
̇ t ) − 𝛾wT (t)w(t) < 0.
V(x (47)
Thus, J0 < 0, and therefore, we can obtain the following inequality:
t
xT (t)ET Px(t) ≤ V(xt ) < 𝛾 wT (s)w(s)ds. (48)
∫0
Furthermore, using the Schur complement equivalence to (20), it yields
[ T ] [ T ]T [ T ]
Ci Ci E P 0
<𝛾 . (49)
T
Chi T
Chi 0 𝜀I
KCHAOU ET AL 9

Note that
{[ ]T [ ][ ]T [ ]}
∑
r
x(t) CiT CiT x(t)
zT (t)z(t) = 𝜇i (𝜃) T T
i=1
x(t − h(t)) Chi Chi x(t − h(t))
≤ 𝛾xT (t)ET Px(t) ≤ 𝛾V(xt ) (50)
t ∞
< 𝛾2 wT (s)w(s)ds ≤ 𝛾 2 wT (s)w(s)ds.
∫0 ∫0
Taking the maximum value of ||z(t)||2∞ , we have ||z(t)||2∞ < 𝛾 2 ||w(t)||22 for any 0 ≠ w(t) ∈ L2 . Consider now the
uncertain case. By following the same procedure as used above, it is easy to verify that
[ ]
Φi + sym(Φ1 ) hr X ( )
+ sym Γ1 F(t)ΓT2 < 0, (51)
∗ −hr Z3
[ ]
Φi + sym(Φ1 ) hr Y ( )
+ sym Γ1 F(t)ΓT2 < 0. (52)
∗ −hr Z3
Then, according to Lemma 2, inequalities (18)-(19) hold using the Schur complement. This completes the proof.

3.3 L2 -L∞ sliding mode dynamics synthesis

In this section, we focus our attention on design gains Ki in (12) such that sliding mode dynamics (15) is robustly
admissible with L2 -L∞ norm bound 𝛾.
Let xs (t) = Ki x(t).The sliding mode dynamics (15) can be written as
⎧ ∑r
{ }
⎪Ex̃̇ = 𝜇i à i x̃ (t) + A
̃ hi x̃ (t − h(t)) + B̃ wi w(t)
⎪ i=1
⎨ (53)
⎪z(t) = ∑𝜇i {C̃ x̃ (t) + C̃ x̃ (t − h(t)) + D
r
}
̃ wi w(t) ,
⎪ i hi
⎩ i=1

[ ]T
where x̃ (t) = xT (t) xsT (t) and
[ ] [ ] [ ] [ ]
[ ]
̃E = E 0 , A ̃ i = Ai Bi , A ̃ hi = Ahi 0 , B̃ wi = Bwi , C̃ i = Ci 0 ,
0 0 Ki −I 0 0 0
[ ] [ ]T [ ] [ ]
C̃ hi = Chi 0 , M ̃ i = M T 0 , Ñ i = Ni 0 , Ñ hi = Nhi 0 .
i (54)

Theorem 2. Let 𝛾 > 0, hm > 0, hM > 0, hd > 0, and 𝜆j , ( j = 0, 1, … , 5) be given scalars. Sliding mode dynamics of (15)
is admissible with L2 -L∞ performance 𝛾, if there exist matrices R, P > 0, Q1 > 0, Q2 > 0, Q3 > 0, Z1 > 0, Z2 > 0, Z3 > 0,
Gj11 , Gj21 , ( j = 1, … , 5), Yi , and Fi , i = 0, 1, … , r such that the following inequalities hold:
( )
̃ A𝑗i , Ah𝑗i , Bw𝑗i , X, G𝑗i < 0
Ψ1i E, (55)

( )
̃ A𝑗i , Ah𝑗i , Bw𝑗i , Y , G𝑗i < 0
Ψ2i E, (56)
( )
̃ C̃ i , C̃ hi < 0,
Ψ3i E, (57)
where
[ ] [ ]
G𝑗,11 Ai + IYi G𝑗,11 Bi − IFi G𝑗,11 Ahi 0
A𝑗i = , Ah𝑗i = ,
G𝑗,21 Ai + 𝜆i Yi G𝑗,21 Bi − 𝜆i Fi G𝑗,21 Ahi 0
[ ] [ ]
G𝑗,11 Bwi G𝑗,11 IFi
Bw𝑗i = , G𝑗i = . (58)
G𝑗,21 Bwi G𝑗,21 𝜆i Fi

Furthermore, Ki = Fi−1 Yi
10 KCHAOU ET AL

Proof. Under the conditions of Theorem 2, a feasible solution satisfies the condition Φ55i = −sym(G𝑗i ) + h2m Z1 +
h2M Z2 + hr Z3 < 0. This implies that Gji is nonsingular, and thus, Fi is also nonsingular. Applying Theorem 1 to system
(53) with the following particular structures of matrices Gji :
[ ] [ ]
G𝑗,11i IFi Im
G𝑗i = , I= 0 , ( 𝑗 = 1, … , 5, i = 1, 2, … , r), (59)
G𝑗,21i 𝜆i Fi (n−m)×m

conditions (55)-(57) hold by setting Yi = Fi Ki . This completes the proof.

Remark 2. In order to obtain the minimum value of L2 -L∞ performance, the minimum allowed 𝛾 satisfying the LMIs
in Theorem 2 can be computed by solving the following optimization problem:

minimize 𝛾 subject to LMIs (55)-(57). (60)

Remark 3. In (53), we construct an augmented dynamic system with a new state matrix showing a decoupling between
control matrices Bi and controller gains Ki . This allows us to avoid the bilinearity problem and to study the general
case of fuzzy singular systems with no common control matrices.

3.4 Adaptive SMC law synthesis

After establishing the appropriate switching surface (12), an adaptive SMC law will be designed to guarantee the
reachability of the specified sliding surface s(t) = 0 even though uncertainties and input nonlinearity are presented.
To achieve the control objective, the following fuzzy dynamic SMC law is employed:

Ri ∶ If 𝜃1 (t) is F1i and if 𝜃2 (t) is F2i · · · if 𝜃s (t) is Fsi , then

s(t)
u(t) = −𝛼(t)(𝜓i
̂ + 𝜒) (61)
||s(t)||
or, equivalently,
s(t)
u(t) = −𝛼(t)(Ω
̂ + 𝜒) , (62)
||s(t)||
where 𝜓 i is designed as

̄
𝜓i = ||Gi || (||Mi ||||Ni ||||x(t)|| + ||Mi ||||Nhi ||||x(t − h(t))|| + ||Bwi ||w(t)) + ||Ki x(t)|| + 𝜌(t)||x(t)||,
̂ (63)
∑
Ω = ri=1 𝜇i 𝜓i , and χ > 0 is a small constant.
The aforementioned adaptive gains are designed as

̂̇ = 𝜅1 𝛼̂ 3 (t)Ω||s(t)||,
𝛼(t) ̂̇ = 𝜅2 ||x(t)||||s(t)||
𝜌(t) (64)

with 𝛼(0)
̂ = 𝛼0 and 𝜌(0)
̂ = 0. Note that 𝛼 0 is a bounded positive initial value of 𝛼(t),
̂ and 𝜅 1 and 𝜅 2 are adjustable positive
constants.
This proposed control scheme will drive the state to reach the sliding surface s(t) = 0. This fact is stated in Theorem 3.
Theorem 3. If the adaptive control input u(t) is designed as (62), with adaptive law (64), then the trajectory of system
(7) converges to the sliding surface s(t) = 0.

Proof. Consider the following Lyapunov function:

1 T 1 2 1 2
Vs (t) = s (t)s(t) + 𝛼̃ (t) + 𝜌̃ (t), (65)
2 2𝜅1 2𝜅2

̃ = 𝛼̂ −1 (t) − 𝛼, 𝜌(t)
where 𝛼(t) ̃ = 𝜌(t)
̂ − 𝜌,
̄ and 𝜌̄ = maxi 𝜌i .
KCHAOU ET AL 11

FIGURE 1 Rolling disk

According to (12), we get

∑
r
̇ =
s(t) 𝜇i Gi {(Δ Ai (t) − Bi Ki )x(t) + Δ Ahi (t)x(t − h(t))) + Bwi (t)w(t) + Bi (𝜙(u(t)) + 𝑓i (t, x(t)))} . (66)
i=1

Without loss of generality, we can choose Gi = B†i = (BTi Bi )−1 BTi . Hence, Gi Bi is nonsingular. By taking the derivative
of Vs (t), we get
1 ̂̇
𝛼(t) 1
V̇ s (t) = sT (t)s(t)
̇ − 𝛼(t)̃ + 𝜌(t) ̂̇
̃ 𝜌(t)
𝜅1 𝛼̂ 2 (t) 𝜅2
∑r
= sT (t) 𝜇i Gi {ΔAi (t)x(t) + ΔAhi (t)x(t − h(t))) + Bwi (t)w(t)}
i=1
( )
∑
r
+ sT (t) 𝜙(u(t)) + 𝜇i (𝑓i (t, x(t)) − Ki x(t)) − 𝛼(t)
̂ 𝛼(t)Ω||s(t)||
̃ + 𝜌(t)||x(t)||||s(t)||
̃
i=1
∑
r
̄ + ||Ki x(t)||}
≤ ||s(t)|| 𝜇i {||Gi || (||Mi ||||Ni ||||x(t)|| + ||Mi ||||Nhi ||||x(t − h(t))|| + ||Bwi ||w)
i=1

+ 𝜌||x(t)||||s(t)||
̄ + 𝜌||x(t)||||s(t)||
̃ − 𝛼(t)
̂ 𝛼(t)Ω||s(t)||
̃ + sT (t)𝜙(u(t)). (67)
Using (62) and Assumption 1.4, it can be derived that
sT (t)
uT (t)𝜙(u(t)) = −𝛼(t)(Ω
̂ + χ) 𝜙(u(t)) ≥ 𝛼uT (t)u(t) = 𝛼 𝛼̂ 2 (t)(Ω + χ)2 . (68)
||s(t)||
Since (Ω + χ) > 0, we get
sT (t)𝜙(u(t)) ≤ −𝛼 𝛼(t)(Ω
̂ + χ)||s(t)||. (69)
Substituting (69) into (67), we obtain
V̇ s (t) = (Ω − 𝛼(t)
̂ 𝛼(t)Ω
̃ − 𝛼 𝛼(t)(Ω
̂ + χ)) ||s(t)|| < 0, ∀||s(t)|| ≠ 0. (70)
Noting that 𝛼 𝛼(t)
̂ + 𝛼(t)
̂ 𝛼(t)
̃ = 1 and 𝛼(t)
̂ > 0, it is easy to verify that
V̇ s (t) < 0, ∀t > 0. (71)
This means the system trajectories converge to the predefined sliding surface and are restricted to the surface for all
subsequent time, thereby completing the proof.
s(t)
Remark 4. It is well known that discontinuous term ||s(t)||
induces the chattering phenomenon. A solution to reduce
s(t)
this phenomenon is to substitute discontinuous function by some continuous and smooth functions as 𝜖+||s(t)||
, where
𝜀 is a small positive scalar value.

4 A NUMERICAL EXAMPLE

In order to illustrate the efficiency of the approaches proposed in this paper, an example is illustrated in this section.
12 KCHAOU ET AL

Example 1. The following nonlinear singular model, represented by a set of differential and algebraic equations,
describes a disk rolling on a surface without slipping (see Figure 1). The disk is connected to a fixed wall with a
nonlinear spring, a linear damper, and a delayed resonator. K1 and K2 are the spring coefficients, both being positive.
The damping coefficient of the damper is b, which is also positive. The coefficient of the resonator is Cr . The radius
of the disk is r, its inertia is given by J, and the mass of the disk is m. The position of the center of the disk along the
surface is given by x1 , whereas x2 refers to the translational velocity of the same point. The angular velocity of the disk
is denoted by x3 . The control input is denoted by u and is a torque applied at the center of the disk. The contact force
between the disk and the surface is denoted by x4 . Finally, w(t) is the external disturbance.

⎧ẋ 1 = x2
⎪ ( )
K K b 1 C
⎪ẋ 2 = − m1 x1 + m2 x13 − m x2 + m x4 − mr x2 (t − d(t)) + 0.1w(t)
⎨ r 1
(72)
⎪ẋ 3 = − J x4 + J (1 + c0 cos(x2 )) (gu (u(t))u(t) + 𝑓 (t, x(t))
⎪
⎩0 = x2 − rx3 = 0

As in the work of Sjoberg and Glad,33 this model can be written as

⎧ẋ 1 = x2
⎪ ( )
K1 K2 3 b 1 Cr
⎪ẋ 2 = − m x1 + m x1 − m x2 + m x4 − m x2 (t − d(t)) + 0.1w(t)
⎨ (73)
⎪0 = x2 (− rx3 ) ( )
⎪0 = − K1 x + K2 x3 − b x + r2 + 1 x − r (1 + c cos(x )) (g (u(t))u(t) + 𝑓 (t, x(t)) ,
⎩ m 1 m 1 m 2 J m 4 J 0 2 u

√
In this example, it assumed that gu (u(t)) = (0.4 + 0.3sin(5u(t)) and 𝑓 (t, x(t)) = (0.1x12 + 0.2 |x2 |)e−t . We select K1 =
K2 = K ± ΔK = 100 ± 10 Nm−1 , b = 30 kg/s, Cr = 10 kg/s, c0 = 0.1, m = 40 kg, r = 10 cm, and J = 3.2 kgm−2 .
For |x1 (t)| ≤ 𝜓, the following TS singular fuzzy model is obtained using the sector nonlinearity approach:

⎧ ∑
3
⎪Eẋ = 𝜇i (x1 (t)) {(Ai + ΔAi )x(t) + Ahi x(t − h(t)) + Bwi w(t) + Bi (gu (u(t))u(t) + 𝑓 (t, x(t))}
⎪ i=1
⎨ ∑
3
⎪z(t) = 𝜇i (x1 (t)) {Ci x(t) + Chi x(t − h(t))}
⎪
⎩ i=1

⎡1 0 0 0⎤ ⎡ K 0 1 0 0 ⎤ ⎡0 0 0 0⎤
⎢0 ⎢ − m (3 + 𝜓 2 ) − mb 0 1
⎥ ⎢0 − mr 0 0 ⎥
C
1 0 0⎥ ⎢ m
⎥ , Ahi = ⎢ ⎥
E = ⎢0 0 0 0 ⎥ , A1 = ⎢ 0 1 −r ( 0 )⎥ ⎢0 0 0 0⎥
⎢0 0 ⎥⎦ ⎢ − K (3 + 𝜓 2 ) 1 ⎥ ⎢0
⎣ 0 0 − mr 0 0 ⎥⎦
2 C
⎣ m − mb 0 rJ +m ⎦ ⎣
⎡ 0 1 0 0 ⎤ ⎡ 0 ⎤ ⎡ 0 ⎤ ⎡ 0 ⎤
⎢−K − b 0 1
⎥ ⎢ 0 ⎥ ⎢ 0 ⎥ ⎢ 0.1 ⎥
=⎢ 0 1 ⎥
m m m
A2,3
⎢ −r ( 0 ) ⎥ , B1,3 = ⎢ 0 ⎥ , B2 = ⎢ 0 ⎥ , Bwi = ⎢ 0 ⎥ ,
⎢−K − b ⎢ r2 ⎥ ⎢ r2 ⎥ ⎢ 0 ⎥
+ m1 ⎥⎦ ⎣ (c − 1) ⎦ ⎣ (−c0 − 1 − 𝜓 2 ) ⎦ ⎣ ⎦
2

⎣ m m 0 rJ J 0 J

where

[ ] [ ]
C11 = C12 = 0.5 0.5 0 0 , Ch1 = Ch2 = 0 0.5 0 0 (74)
x12 (t) 1 + cos(x2 (t)) 𝜓2 − x12 (t) − cos(x2 (t)) + 1
𝜇1 (x1 (t)) = , 𝜇2 (x1 (t)) = , 𝜇3 (x1 (t)) = .
𝜓2 + 2 𝜓2 + 2 𝜓2 + 2
KCHAOU ET AL 13

The time-varying delay is given as h(t) = 0.1 + 0.2e−0.5t . A straightforward calculation gives hm = 0.1, hM = 0.3, and
hd = 0.2. Assume that the uncertain matrices are as follows:

[ ]T [ ]
Mi = 0.25 0 0 0 , Ni = 1 0 1 0 , Nhi = 0 i = 1, 2, 3.

Our aim is to design an SMC in the form of (62) such that the sliding mode dynamics is robustly admissible with a
guaranteed L2 -L∞ noise attenuation level. Set

Gi = B†i , 𝜆1 = 1.5, 𝜆2 = 0.1, 𝜆3 = 0, 𝜆4 = 0, 𝜆5 = 0.05, 𝜀 = 16.3, 𝜀 = 11.

Using the YALMIP toolbox with the SeDuMi solver, problem (60) has a feasible solution with the following parameters:

⎡ 20.0303 2.7591 −36.7289 −0.9480 4.2078 ⎤ ⎡ 24.1147 12.4407 −2.1241 ⎤

⎢ 2.7591 9.3207 −3.7401 0.1988 −0.3779 ⎥ ⎢ −30.4363 −7.6372 1.7454 ⎥
⎢ ⎥ ⎢ ⎥
P = ⎢ −36.7289 −3.7401 75.4059 1.3429 −7.3582 ⎥ , S = ⎢ −24.9101 2.8102 7.2274 ⎥ .
⎢ −0.9480 0.1988 1.3429 8.4966 −0.7397 ⎥ ⎢ −0.4127 −8.9164 0.7416 ⎥
⎢ ⎥ ⎢ 4.1052 −7.2607 ⎥⎦
⎣ 4.2078 −0.3779 −7.3582 −0.7397 8.6754 ⎦ ⎣ 1.4332

The minimum allowed 𝛾 ∗ = 0.054 and the associate controller gains are

[ ] [ ]
K1 = −0.2390 −0.3485 −0.0840 −0.0094 , K2 = −0.2163 −0.3621 −0.0810 −0.0106 ,
[ ]
K3 = −0.2253 −0.3671 −0.0910 −0.0110 . (75)

The existence of a feasible solution shows that there exists a desire sliding surface in (12) such that the resulting sliding
mode dynamics in (15) is admissible with L2 -L∞ performance.
The remaining task is to design a sliding mode controller such that the system trajectories can be driven onto the
predefined sliding surface and maintained there for all subsequent time. For simulation purposes, we take exogenous
input 𝜔(t) = cos(2.5t)e−0.5t and uncertain matrix function F(t) = 0.5 + 0.5sin(2t).
With 𝜒 = 0.75, 𝛼 0 = 2.5, 𝜅 1 = 0.1, and 𝜅 2 = 0.1, the adaptive SMC law can be designed according to (62)-(64).
s(t) s(t)
To prevent the control signal from chattering, we replace ||s(t)|| with 0.1+||s(t)|| .
The simulation results depicted in Figure 2 show that
• for initial condition 𝜑(t) = [0.1, 0, 0, 0]T , t = −0.2, … , 0, Figures 2A to 2E depict respectively, the system state tra-
jectories, the control input, the resulting sliding surface, and the adaptive law when the SMC is applied. We observe
that the system is stable despite the presence of actuator nonlinearity, parameter uncertainties, and external
disturbances.
• Figures 2D and 2E show that the adaptive laws converge to some values depending on initial condition values
𝛼(0)
̂ ̂ and on the adaptation gains 𝜅 1 and 𝜅 2 . However, we can note that 𝛼(t)
and 𝜌(0) ̂ and 𝜌(t)
̂ do not necessarily
converge to nominal values 𝛼 and 𝜌, respectively.
||z(t)||
• From Figure 2F, the ratio of ||w(t)||∞ is less than 0.012 under zero initial condition, which reveals that the L2 -L∞
2
disturbance attenuation level is less than required 𝛾 ∗ = 0.054.
• The proposed scheme can obtain better convergence performance by driving the system trajectories to the specified
sliding surface asymptotically instead of to some neighbor of the surface.
14 KCHAOU ET AL

4 0.3
0.2
3 0.25

0
0.2
2
-0.2 0.15
1 0.1 0.2 0.3 0.4 0.5

0.1
0
0.05
-1
0

-2 -0.05

-3 -0.1
0 5 10 15 20 25 30 0 5 10 15 20 25 30 35 40

(A) (B)
10-3
3 2.55

2.5 2.545

2 2.54

2.535
1.5
2.53
1
2.525
0.5
2.52
0
2.515
-0.5 2.51
-1 2.505

-1.5 2.5
0 10 20 30 40 50 60 70 0 5 10 15 20 25 30 35 40

(C) (D)
10-3
1.2 0.014

0.012
1

0.01
0.8

0.008
0.6
0.006

0.4
0.004

0.2
0.002

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40

(E) (F)

FIGURE 2 Simulation results for rolling disk system. A, State trajectories; B, Input trajectory; C, Surface trajectory; D, Adaptive law;
E, Adaptive law; F, Ratio trajectory [Colour figure can be viewed at wileyonlinelibrary.com]

5 CO N C LUSION S

This paper addresses the problems of SMC for a class of fuzzy singular systems with state delay and nonlinearity input.
A fuzzy integral sliding function is proposed and a delay-dependent sufficient condition is derived to guarantee that the
sliding mode dynamics is robustly admissible with L2 -L∞ disturbance rejection performance. Moreover, an adaptive SMC
KCHAOU ET AL 15

law is designed such that the trajectories of the resulting closed-loop system can be driven onto a prescribed sliding surface
and maintained there for all subsequent time. The existence and the effectiveness of theoretical developments have been
verified by a numerical example. The proposed controller shows that it has the ability to eliminate the model uncertainties
and to reduce the chattering on the sliding surface. It should be emphasized that the computational simplicity of the
suggested method can be an another prominent feature of this work.
As future work, we will further investigate the problem of fault estimation and fault-tolerant control for nonlinear
singular systems via the SMC scheme.

ORCID

Mourad Kchaou http://orcid.org/0000-0002-6849-1745

Ahmed El-Hajjaji http://orcid.org/0000-0002-9380-9890

REFERENCES
1. Dai L. Singular Control Systems. Lecture Notes in Control and Information Sciences. vol. 118. New York, NY: Springer; 1989.
2. Duan GR. Analysis and Design of Descriptor Linear Systems. New York, NY: Springer Verlag; 2010.
3. Mahmoud MS. Switched Time-Delay Systems. Boston, MA: Springer-Verlag; 2010.
4. Gu K, Kharitonov VL, Chen J. Stability of Time-Delay Systems. Berlin, Germany: Springer-Verlag; 2003.
5. Napp D, Tadeo F, Chaabane M. State feedback positive stabilization of discrete descriptor systems. Int J Innov Comput, Inf Control.
2014;10(5):1853-1859.
6. Kchaou M, Gassara H, El-Hajjaji A. Robust observer-based control design for uncertain singular systems with time-delay. Int J Adapt
Control Signal Process. 2014;28(2):169-183.
7. Cui W, Fang J, Shen Y, Zhang W. Dissipativity analysis of singular systems with Markovian jump parameters and mode-dependent mixed
time-delays. Neurocomputing. 2013;110:121-127.
8. Feng Z, Lam J, Gao H. 𝛼-Dissipativity analysis of singular time-delay systems. Automatica. 2011;47(11):2548-2552.
9. Kchaou M, Tadeo F, Chaabane M. A partitioning approach for H∞ control of singular time-delay systems. Optim Control Appl Methods.
2013;34(4):472-486.
10. Takagi T, Sugeno M. Fuzzy identification of systems and its application to modelling and control. Trans Syst Man Cybern.
1985;15(1):116-132.
11. Benzaouia A, Oubah R. Stability and stabilization by output feedback control of positive Takagi–Sugeno fuzzy discrete-time systems with
multiple delays. Nonlinear Anal: Hybrid Systems. 2014;11:154-170.
12. Li F, Shi P, Wu L, Zhang X. Fuzzy-model-based D-stability and nonfragile control for discrete-time descriptor systems with multiple delays.
IEEE Trans Fuzzy Syst. 2014;22(4):1019-1025.
13. Manivannan A, Muralisankar S. Robust stability analysis of Takagi-Sugeno fuzzy nonlinear singular systems with time-varying delays
using delay decomposition approach. Circuits Syst Signal Process. 2016;35(3):791-809.
14. Gassara H, Hajjaji AE, Kchaou M, Chaabane M. Observer based (Q, V, R)-𝛼-dissipative control for TS fuzzy descriptor systems with time
delay. J Franklin Inst. 2013;351(1):187-206.
15. Kchaou M, Souissi M, Toumi A. Delay-dependent stability and robust L2 -L∞ control for a class of fuzzy descriptor systems with
time-varying delay. Int J Robust Nonlinear Control. 2011;23(3):284-304.
16. Rakkiyappan R, Chandrasekar A, Lakshmanan S. Stochastic sampled data robust stabilisation of T-S fuzzy neutral systems with randomly
occurring uncertainties and time-varying delays. Int J Syst Sci. 2016;47(10):2247-2263.
17. Li X, Rakkiyappan R. Stability results for Takagi–Sugeno fuzzy uncertain BAM neural networks with time delays in the leakage term.
Neural Comput & Applic. 2013;22(1):203-219.
18. Wu L, Zheng WX, Gao H. Dissipativity-based sliding mode control of switched stochastic systems. IEEE Trans Autom Control.
2013;58(3):785-791.
19. Liu J, Vazquez S, Wu L, Marquez A, Gao H, Franquelo LG. Extended state observer-based sliding-mode control for three-phase power
converters. IEEE Trans Ind Electron. 2017;64(1):22-31.
20. Ding Y, Zhu H, Zhong S. Exponential stabilization using sliding mode control for singular systems with time-varying delays and nonlinear
perturbations. Commun Nonlinear Sci Numer Simulat. 2011;16(10):4099-4107.
21. Chang J. Dynamic output feedback integral sliding mode control design for uncertain systems. Int J Robust Nonlinear Control.
2012;22(8):841-857.
22. Gao Q, Feng G, Liu L, Qiu J, Wang Y. An ISMC approach to robust stabilization of uncertain stochastic time-delay systems. IEEE Trans
Autom Control. 2014;61(12):6986-6994.
23. Wu L, Ho DWC. Sliding mode control of singular stochastic hybrid systems. Automatica. 2010;46(4):779-783.
24. Zhang X, Su H, Lug R. Second-order integral sliding mode control for uncertain systems with control input time delay based on singular
perturbation approach. IEEE Trans Autom Control. 2015;60(11):3095-3100.
16 KCHAOU ET AL

25. Kchaou M, Gassara H, El-Hajjaji A, Toumi A. Dissipativity-based integral sliding-mode control for a class of Takagi–Sugeno fuzzy singular
systems with time-varying delay. IET Control Theory Appl IET. 2014;8(17):2045-2054.
26. Wang Y, Fei J. Adaptive sliding mode control for PMSM position regulation system. Int J Innov Comput, Inf Control. 2015;11(3):881-891.
27. Yu F-M, Chung H-Y, Chen S-Y. Fuzzy sliding mode controller design for uncertain time-delayed systems with nonlinear input. Fuzzy Set
Syst. 2003;140(2):359-374.
28. Liu Y, Niu Y, Zou Y. Sliding mode control for uncertain switched systems subject to actuator nonlinearity. Int J Control Autom Syst.
2014;12(1):57-62.
29. Gao L, Wang D, Wu Y. Non-fragile observer-based sliding mode control for Markovian jump systems with mixed mode-dependent time
delays and input nonlinearity. Appl Math Comput. 2014;229:374-395.
30. Nguyen T-B-T, Liao T-L, Yan J-J. Improved adaptive sliding mode control for a class of uncertain nonlinear systems subjected to input
nonlinearity via fuzzy neural networks. Math Probl Eng. 2015;2015:1-13.
31. Liu L, Pu J, Song X, Fu Z, Wang X. Adaptive sliding mode control of uncertain chaotic systems with input nonlinearity. Nonlinear Dyn.
2014;76(4):1857-1865.
32. Petersen IR. A stabilization algorithm for a class of uncertain linear systems. Syst Control Lett. 1987;8(4):351-357.
33. Sjoberg J, Glad T. Computing the controllability function for nonlinear descriptor systems. Paper presented at: Proceedings of the 2006
American Control Conference; 2006; Minneapolis, MN.

How to cite this article: Kchaou M, Gassara H, El-Hajjaji A. Adaptive sliding mode control for fuzzy
singular systems with time delay and input nonlinearity. Int J Adapt Control Signal Process. 2018;1–16.
https://doi.org/10.1002/acs.2854