Observer-based event-triggered sliding mode control for uncertain descriptor systems with a neural-network event-triggering sampling

scheme Communicated by Prof. Yugang Niu

Journal Pre-proof

Observer-based event-triggered sliding mode control for uncertain

descriptor systems with a neural-network event-triggering sampling
scheme

Yuzhong Wang, Tie Zhang, Junchao Ren, Meng Chen

PII: S0925-2312(19)31775-8
DOI: https://doi.org/10.1016/j.neucom.2019.12.066
Reference: NEUCOM 21695

To appear in: Neurocomputing

Received date: 9 October 2019

Revised date: 1 December 2019
Accepted date: 16 December 2019

Please cite this article as: Yuzhong Wang, Tie Zhang, Junchao Ren, Meng Chen,
Observer-based event-triggered sliding mode control for uncertain descriptor systems
with a neural-network event-triggering sampling scheme, Neurocomputing (2019), doi:
https://doi.org/10.1016/j.neucom.2019.12.066

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition
of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of
record. This version will undergo additional copyediting, typesetting and review before it is published
in its final form, but we are providing this version to give early visibility of the article. Please note that,
during the production process, errors may be discovered which could affect the content, and all legal
disclaimers that apply to the journal pertain.

© 2019 Published by Elsevier B.V.

 Observer-based event-triggered sliding mode control for
uncertain descriptor systems with a neural-network
event-triggering sampling scheme

Yuzhong Wanga , Tie Zhanga,b,∗, Junchao Rena , Meng Chenc

a Department of Mathematics, Northeastern University, Shenyang, Liaoning, 110004, China
b Department of Mathematics and the State Key Laboratory of Synthetical Automation for Process
Industries, Northeastern University, Shenyang 110004, China
c School of Resources and Civil Engineering, Northeastern University, Shenyang 110004, China

Abstract

This paper investigates the problem of the observer-based event-triggered sliding

mode control for uncertain descriptor systems with exogenous disturbance via a new
neural-network event-triggering strategy. Firstly, a new neural-network event-triggering
communication scheme is proposed in order to decide whether sampled data is to be
transmitted for the observer. Secondly, the existence conditions and the stability condi-
tions of the sliding mode dynamics are presented in terms of linear matrix inequalities.
An adaptive event-triggered sliding mode controller is designed to ensure the retainabil-
ity to the sliding surface from beginning almost surely. Thirdly, an effective strategy is
developed to avoid the Zeno behavior without decomposing the descriptor system, and
the positive lower-bound of the inter-execution time intervals is provided. The existing
method can not directly get the positive lower-bound. Thus, the parameters used for
scaling corresponding inequalities are introduced to solve the problem. Finally, simula-
tion results are presented to illustrate the effectiveness of the proposed method and are
superior to the existing event triggering strategies in the literature.
Keywords: Sliding mode observer, descriptor systems, event-triggering, integral-type
switching surface, LMIs

∗ Correspondingauthor
Email addresses: [email protected] (Yuzhong Wang), [email protected] (Tie Zhang )

Preprint submitted to Journal of LATEX Templates December 18, 2019

1 1. Introduction

2 Descriptor systems, are also referred as singular systems, generalized state-space sys-
3 tems, or differential-algebraic systems. Descriptor systems provide a natural representa-
4 tion of dynamical systems, which describe a larger range of systems than normal systems
5 [1]. Therefore, descriptor systems have been widely used in many fields, such as power,
6 economic and circuits systems [2]. The descriptor model consists of ordinary differen-
7 tial equations that describe the dynamical part of the system and algebraic equations
8 that describe the interrelationship of different components in the system. Over the past
9 decades, a lot of research results have been reported on descriptor systems, such as sta-
10 bilization and stability [3], sliding mode observer design [4] etc. The research results for
11 normal system have been extended to descriptor system, such as stability analysis [5, 6],
12 control design [4], estimation and filtering [7].
13 Networked control systems (NCSs) have attracted great attention due to its simple in-
14 stallation and low cost [8, 9]. However, the communication bandwidth is a scarce resource
15 in NCSs. Thus, an issue of widespread concern is how to reduce the utilization of limited
16 communication bandwidth while ensuring the same levels of performance and stability.
17 One of the effective ways to solve this problem is to transmit signal if and only if the
18 signal satisfies the prescribed condition. Moreover, redundant and ineffective information
19 should not be transmitted. It is clear that this policy will use the network communication
20 bandwidth more effectively [10]. Event-triggering sampling (ETS) schemes are one of the
21 most effective results [11, 12, 13]. ETS mechanisms do not transmit all the signal data.
22 When and only when the sampled signal satisfies the triggering condition, sampled data
23 is transmitted to the controller via the network. Thus, only limited data is sent to the
24 controller [14, 15]. Dynamic systems with event-triggering conditions have been widely
25 applied such as descriptor systems [14], switched systems [16, 17], and T-S fuzzy systems
26 [18].
27 As we all know, how to improve the robustness has attracted more attention due
28 to uncertainties and nonlinearity in most practical systems. Sliding mode control is
29 an effective robust control approach from the perspectives of both theory and practice
30 [19, 20]. Currently, the theory system of sliding mode control is quite complete. It has
31 been widely applied in various industrial control systems because of attractive features
2
32 including ease of implementation, insensitivity to matched disturbances, uncertainties
33 and rapid response. Many research results have bee derived for various dynamic sys-
34 tems, such as time-delay systems [21], stochastic systems [21], uncertain systems [22, 23],
35 switchd systems [17] and Markovian jump systems [24]. In [19], the H∞ SMC problem for
36 a class of uncertain discrete-time systems subject to unmatched external disturbances
37 and communication constraints is discussed. The stochastic communication protocol
38 (SCP) is introduced, and both the sliding surface and the sliding mode controller under
39 the SCP scheduling is designed. The H∞ performance is considered for the closed-loop
40 system rather than the sliding mode dynamics. On the other hand, because states are
41 usually not measurable in most practical systems, observers are often used to observe
42 system states. Therein, the sliding mode observer has both the advantages of the SMC
43 and general observers, which is utilized to deal with the fault estimation and robust
44 stabilization etc [4].
45 As a specific ETS, the event-triggered SMC has power potentials in actual applica-
46 tions. However, event-triggering mechanisms keep the state of the system constant in
47 the inter-execution time intervals. The result that the state of the system fails to reach
48 the sliding surface, which will result in no sliding mode dynamics [25]. At the same
49 time, when the state is not measurable, the situation becomes even more complicated.
50 Therefore, how to design event-triggered SMC is the first motivation of this paper. On
51 the other hand, to our best knowledge, event-triggered SMC for descriptor systems has
52 not been adequate studied. Moreover, how to avoid Zeno behavior is also a challenge-
53 able topic, whose the difficulties are from the singular matrix E in descriptor systems.
54 This causes that the existing methods can not directly get the positive lower-bound
55 [13, 26]. The event-triggering strategy increases the difficulty of designing Lyapunov
56 functions, especially for descriptor systems. How to solve above two problems is the sec-
57 ond motivation. It should be noted that the event-triggering communication threshold
58 is a preselected constant in traditional event-triggering schemes [11, 12, 15, 18, 27]. To
59 further reduce the use of the communication bandwidth, constructing a communication
60 threshold which can be dynamically adjusted based on the latest transmitted signals is
61 the third motivation for this paper.
62 In this paper, the event-triggered SMC problem of descriptor systems is studied. The

3
63 main contributions are three folds:
64 1) To reduce the use of communication bandwidth, a new neural-network (NN)
65 event-triggering communication scheme is proposed, which can adaptively adjust event-
66 triggering threshold.
67 2) Observer-based event-triggered SMC problem for descriptor systems is solved by
68 constructing event-triggered sliding mode observer. Stability conditions are obtained by
69 constructing a Lyapunov function. An adaptive event-triggered sliding mode controller
70 is synthesized for driving the trajectories of the observer to remain the sliding surface
71 from the initial time.
72 3) An effective strategy is developed to avoid the Zeno behavior without decomposing
73 the descriptor system, and the positive lower-bound of the inter-execution time intervals
74 is provided. The existing method can not directly get the positive lower-bound. Thus,
75 the parameters used for scaling corresponding inequalities are introduced to solve the
76 problem.
77 The paper is organized as follows. In section 2, a NN event-triggering communi-
78 cation scheme and some useful lemmas and definitions are presented. In section 3, an
79 integral-type switching surface, an adaptive event-triggered sliding mode controller and
80 the Zeno behavior are discussed. Two examples are given in section 4. Section 5 presents
81 concluding comments.
82 Notation: N represents the set of positive integers; Rn represents the n-dimensional
83 Euclidean space; The set of all n × m real matrices are denoted by Rn×m ; The function
84 He[A] = A + AT ; P > 0 means P is real symmetric positive definite; The space `2 [0, ∞)
85 represents all square-integrable vectors functions over [0, ∞); matrix diag {· · · } denotes
86 a block-diagonal matrix; λmin (A) represents the minimal eigenvalues of a matrix A.

87 2. NCSs modeling and NN event-triggering communication scheme

88 2.1. Descriptor systems

89 In this paper, an uncertain descriptor system is considered,


 E ẋ(t) = (A + ∆A)x(t) + Bu(t) + Gω(t)
(1)

y(t) = Cx(t)
4
 90 where E is singular matrix. x(t) ∈ Rn and u(t) ∈ Rm are the state and the control
91 input respectively, y(t) ∈ Rp is the system output, and ω(t) ∈ `2 [0, ∞) is the exogenous
92 disturbance, and satisfying kω (t)k ≤ ϑ kx (t)k. ∆A is time-varying uncertain parameter,
93 which satisfying ∆A = M F (t)N . M and N are known real matrices. F (t) represents an
94 unknown matrix function with Lebesgue measurable elements satisfying F T (t)F (t) ≤ I.
The matrix C and N are assumed that rank [E, C] = rank E, rank [E, N ] = rank E.

Z t
u t
x  y t
'HVFULSWRUV\VWHP 6HQVRU
u2EVHUYHU
t 6OGLQJPRGH&RWUROOHU
z t
Ö t
x u t

K
y t
(YHQW
=2+ &ORFNV\QFKURQL]H
WULJJHULQJ

1HWZRUN
y tk y tk
7KHIUDPHZRUNRI1&6

Fig. 1: Framework of NCSs with event-triggering sampling

95

96 The frame of network descriptor system (1) is showed in Fig.1, which is composed of
97 a plant with a sliding mode observer, a sensor, a ZOH, ect.

98 2.2. Definitions, Lemmas and Assumption

99 To simplify the theoretical analysis, some definitions, lemmas and Assumption are
100 given.
E ẋ(t) = Ax(t) (2)

101 Definition 1. [14] Descriptor system (2) is said to

102 1. Regular if det(sE − A) 6≡ 0.
103 2. Impulse free if deg (det ((sE − A))) = rank (E).
104 3. Stable if for any ε > 0, there exists a scalar δ(ε) > 0 such that, for any compatible
105 initial conditions ϕ(t) satisfying kϕ(t)kc < δ(ε), the solution x(t) of descriptor system
106 satisfies kx(t)k < ε for t ≥ 0. Furthermore, limt→∞ x(t) = 0.
107 4. Admissible if it is regular, impulse free and stable.
5
108 Definition 2. [14] Given constants γ > 0, the descriptor system (1) is said to be
109 asymptotically stable with an H∞ disturbance attenuation level γ, if the descriptor sys-
110 tem with ω(t) = 0 is asymptotically stable, for all nonzero ω(t) ∈ `2 [0, ∞) under the zero
111 initial condition, it holds that
Z ∞
 T 
y (t) y (t) − γ 2 ω T (t) ω (t) dt < 0 (3)
0

112 Lemma 1. [14] For any matrices X, Y and a positive scalar ς > 0, The following
113 inequality holds:
2X T Y ≤ ςX T X + ς −1 Y T Y (4)

114 Lemma 2. [14] For symmetric matrix X and singular matrix E = EL E R T with EL ∈
115 Rn×r and ER ∈ Rn×r full column rank such that E L T X E L > 0 and Ψ is nonsingular,
116 Λ and W are full row rank and full column rank respectively and satisfy ΛE = 0 and
117 EW = 0, then, X E + W T ΨΛT is nonsingular and its inverse is expressed as (X E +
118 W T ΨΛT )−1 = X̄ E T + ΛΨ̄W , where X̄ = X̄ T and Ψ̄ is nonsingular matrix such that
119 Ψ̄ = (ΛT Λ)−1 Ψ−1 (W W T )−1 , ER
T
X̄ ER = (ELT X EL )−1 .
+LGGHQ
OD\HU

wk
,QSXW
vk
V (tk 1 )
2XWSXW

D OD\HU OD\HU

Fig. 2: Framework of Neural-Network

120 2.3. An NN event-triggering communication scheme

121 In this section, a neural-network event-triggering mechanism is introduced to save

122 more the limited network bandwidth. The latest transmitted data and the current sam-
123 pled data are denoted as y(tk ) and y(lk ), respectively. The current sampled data y(lk )
124 should be transmitted to the observer based on the following scheme:
n o
tk+1 = inf t > tk eTy (t)Φey (t) > σ (tk ) y T (tk )Φy(tk ) (5)

6
125 where 0 ≤ σ(tk ) < 1, ey (t) = y(tk ) − y(lk ), Φ > 0 is a positive definite matrix. Besides,
126 σ(tk+1 ) satisfies the following rule.

σ (tk+1 ) = max {σ (tk ) , σm } (6)

127 σ (tk ) is approximated by a NN. σm is the known lower bound of σ (tk ).

128 A three layer BP (back propagation) NN architecture is shown in Fig. 2. It is
129 composed of input layer, hidden layer and output layer. The NN model is described by

βk =vk α + θk , βk0 = f1 (βk )

3
!
X
0
σ =f2 wk β (k) + a (7)
k=1
1 2
e =σ − σn , E = e
2
130 The total error criterion function for P training samples is
P
1X 2
EP = e (8)
2
k=1

131 where α is the input, and θk represents the threshold value of the kth node of the hidden
132 layer. vk is the weight between the input layer and the hidden layer, and wk is the weight
133 between the hidden layer and the output layer. f1 (·) and f2 (·) are sigmoid functions. a
134 represents the threshold value of the kth node of the output layer. σn denotes the desired
135 value. The weights correction ∆vk , ∆wk and thresholds correction ∆θk , ∆a are changed
136 as follows by using the gradient descent method:
∂EP ∂EP
∆vk = − η , ∆θk = −η
∂vk ∂θk
(9)
∂EP ∂EP
∆wk = − η , ∆a = −η
∂wk ∂a

137 Remark 1. σ (tk ) is adjusted by online approximation using a NN. We assume

138 α = kx (tk+1 )k − kx (tk )k. If α > 0, we use σ (tk+1 ) < σ (tk ) to impose faster com-
139 munication [28]. Moreover, the greater the α value is, the smaller the σ (tk ) value is; if
140 α < 0, we use σ(tk+1 ) > σ(tk ) to conserve communication bandwidth. The smaller the
141 α value is, the larger the σ (tk ) value is. When σ(tk ) = σ, the event-triggered condition
142 degenerates to the established conditions in [11, 12, 15, 18, 27]. Thus, the proposed

7
143 NN event-triggered condition is more general, which can effectively reduce the waste of
144 communication bandwidth. It will be confirmed by practical example in section 4.
145 The parameters σ(tk+1 ) are selected as follows:
146 Algorithm 1 For given positive scalars σm .
147 Step 1: Set the initial value as σm . Based on Remark 1, the desired communication
148 bandwidth and control performance, we use MATLAB to train the NN.
149 Step 2: According to values of kx (tk+1 )k − kx (tk )k , we use the NN to approximate
150 the values of σ(tk+1 ).
151 Step 3: If σ(tk+1 ) < σm , then, σ(tk+1 ) = σm . Otherwise, σ(tk+1 ) = σ(tk ).

152 3. Main results

153 The aim of the paper is to design the sliding mode observer-based event-triggering
154 strategy to save the limited communication bandwidth and stabilize the system (1). For
155 the descriptor system (1) with unmeasurable state , the system output is sampled and
156 triggered. Firstly, an integral-type sliding surface is designed such that the dynamics of
157 the observer restricted to the sliding surface have the desirable property of asymptotical
158 stability, Secondly, an adaptive event-triggered SMC is derived to drive the dynamics
159 trajectories of the system to remain the sliding surface from the initial time. Finally, the
160 positive lower bound of the inter-execution time T is obtained to avoid Zeno phenomenon.

161 3.1. Observer design

162 In the following, we use the measurement ~y (t) = y(tk ) to design observer.

˙
E x̂(t) = Ax̂(t) + Bu(t) + L(~y (t) − C x̂(t)) (10)

163 where x̂(t) ∈ Rn represents the state estimation of x(t), L ∈ R(n×p) represents the
164 observer gain.
165 The state estimation denote as e(t) = x(t) − x̂(t). The error system can be given as

E ė (t) = Ae (t) + ∆Ax (t) + Dω(t) − L (~y (t) − C x̂ (t)) (11)

8
166 3.2. Observer-based SMC design
167 In this section, an integral-type sliding surface function is defined as:
Z t
s(t) = GE x̂ (t) − GE x̂ (t0 ) − G [(A + BK) x̂ (v)]dv (12)
t0

168 where the matrix G = B T X is determined in the sequel.

169 When s(t) = 0 and ṡ(t) = 0, the equivalent control law can be given by

ueq (t) = K x̂ (t) − (GB)−1 GLCe(t) − (GB)−1 GLey (t) (13)

170 By substituting (13) into (10), the sliding mode dynamics of the observer (10) and
171 the error system can be given as

˙
 E x̂(t) = (A + BK) x̂(t) + (I − B(GB)−1 G)LCe(t) + (I − B(GB)−1 G)Ley (t)
(14)

E ė (t) =(A − LC)e (t) + ∆Ax (t) + Dω(t) − Ley (t)
172 In the next section, the stability conditions of the system (14) are analyzed.

173 3.3. Stability analysis

174 Theorem 1. For positive scalars 0 ≤ σm < 1 γ > 0. The system (14) is admissible
175 with the H∞ performance level γ. If there exist matrix Φ > 0 and nonsingular matrix
176 P such that
ET P = P T E ≥ 0 (15)

177
 
Π11 Π12 Π13 0 0
 
 
 ∗ Π22 σm C T Φ − P T L PTD PTM 
 
 
Π= ∗ ∗ Π33 0 0 <0 (16)
 
 
 ∗ ∗ ∗ −γ 2 I 0 
 
∗ ∗ ∗ ∗ −=I
where
 
Π11 =He P T (A + BK) + σm C T ΦC + C T C + =−1 N T N

Π12 =σm C T ΦC + C T C + P T (I − B(GB)−1 G)LC + =−1 N T N

 
Π22 =He P T (A − LC) + σm C T ΦC + C T C + =−1 N T N

Π13 =σm C T Φ + P T (I − B(GB)−1 G)L

Π33 =σm Φ − Φ
9
178 P = X E + W T ΨΛT with EL ∈ Rn×r and ER ∈ Rn×r full column rank such that
179 E L T X E L > 0 and Ψ is nonsingular, Λ and W are full row rank and full column rank
180 respectively and satisfy ΛE = 0 and EW = 0.
Proof . Firstly, the regularity and no-impulse of the system will be proved. Since
rank (E) = r < n, it exists two nonsingular matrices R and F such that
     
Ir 0 Ã11 Ã12 L11 L12
REV =   , R (A + BK) V =   , R (A − LC) V =  
0 0 Ã21 Ã22 L21 L22
   
P̃ 11 P̃ 12
F̃ F̃
P̃ =R−T P V =   , V T σC T ΦC + C T C V =  11 12

P̃21 P̃22 0 0
   

F 11 F 12
F̄11 F̄12
V T σC T ΦC V =   , V T =−1 N T N V =  
0 0 0 0

181 Since (15), we have P̃12 = 0. From (16), we obtain

 
Π11 Π12
 <0 (17)
∗ Π22
 T  
V 0 V 0
182 Then pre- and post-multiplying (17) by   and  , we have
0 V 0 V
 
# # # #
   
 T 
 # He P̃22 Ã22 # # 
 <0 (18)
 
 # # # # 
   
T
# # # He P̃22 L22

183 This implies that Ã22 and L22 are nonsingular. Thus the pair (E, A + BK) and
184 (E, A − LC) are regular and impulse free.
185 Next, asymptotically stable will be showed by Lyapunov stability theory. The Lya-
186 punov function candidate is given as

V̇ (t) = x̂T (t)E T P x̂(t) + eT (t)E T P e(t) (19)

10
187 Take the time derivative of (19) along (14), we have

V̇ (t) =x̂T (t)P T He (A + BK) x̂(t) + 2x̂T (t)P T (I − B(GB)−1 G)LCe(t)

+2x̂T (t)P T (I − B(GB)−1 G)Ley (t) + eT (t)P T He(A − LC)e (t) + 2eT (t)P T ∆Ax (t)

+2eT (t)P T Dω(t) − 2eT (t)P T Ley (t)

(20)

188 Then, based on event-trigging scheme, it follows that

V̇ (t) ≤ η T (t)Υη(t) (21)

h i
189 where η T (t) = xT (t) eT (t) eTy (t)
 
Π11 Π12 Π13
 
 
Υ= ∗ Π22 σm C T Φ − P T L  (22)
 
∗ ∗ Π33

190 According to Theorem 1, we get V̇ (t) < 0. Thus, the system (14) is asymptotically
191 stable.
192 We now turn to the analysis of the H∞ performance under the zero initial condition.
193 When ω(t) 6= 0, it follows that

R∞ R∞
Jzω = 0
(y T (t)y(t) − γ 2 ω T (t)ω(t) + V̇ (x(t)))dt − 0
V̇ (x(t))dt
R∞
= 0
(y T (t)y(t) − γ 2 ω T (t)ω(t) + V̇ (x(t)))dt − V (∞) + V (0)
R∞ (23)
≤ 0
(y T (t)y(t) − γ 2 ω T (t)ω(t) + V̇ (x(t)))dt
R∞
≤ 0
χT (t)Πχ(t)dt
h i
194 where χT (t) = η T (t) ω T (t) , according to (16), we get Jzω < 0. According to
195 Definition 2, the system (14) has the H∞ control performance. This completes the
196 proof.
197 Remark 2 Note that there exists the nonconvex term P T (I − B(GB)−1 G)LC in
198 (16), which will lead to the matrix inequalities can not be solved directly. Therefore,
199 (16) will be transformed into a LMI in the following.
200 In the following theorem, the controller gain and the observer gain are designed.

11
201 Theorem 2. For positive scalars 0 ≤ σm < 1 γ > 0. The system (14) is admissible
202 with the H∞ performance level γ. If there exist matrix Φ > 0 and nonsingular matrix
203 P such that
 √ 
∆ ∆12 σm C T Φ + Y 2P T B 0 0 0 0
 11 
 
 ∗ ∆22 σm C T Φ − Y 0 PTD CT YT 0 PTM 
 
 
 ∗ ∗ σm Φ − Φ 0 0 0 YT 0 
 
 
 ∗ ∗ ∗ −B T X B 0 0 0 0 
 <0
 
 ∗ ∗ ∗ ∗ −γ 2 I 0 0 0 
 
 
 ∗ ∗ ∗ ∗ ∗ −HeP + X 0 0 
 
 
 ∗ ∗ ∗ ∗ ∗ ∗ −HeP + X 0 
 
∗ ∗ ∗ ∗ ∗ ∗ ∗ −=I
(24)

where
 
∆11 =He P T A + BK + σm C T ΦC + C T C + =−1 N T N
 
∆22 =He P T A − YC + σm C T ΦC + =−1 N T N

∆12 =σm C T ΦC + =−1 N T N + YC

P =X E + W T ΨΛT

204 Since P T BK and P T L are two nonlinear terms, we assume

Y =P T L
(25)
BK =P T BK

205 then, based on Lemma 3, the observer gains L and the controller gains K are constructed
206 by

K =(B T P T B)−1 B T BK
(26)
L =P −T Y

207 Proof .

12
208 The condition (20) can be simplified as

− 2x̂T (t)P T B(GB)−1 GLCe(t) = −2x̂T (t)P T B(B T X B)−1 B T X LCe(t)

≤ x̂T (t)P T B(B T X B)−1 B T P x̂(t) + eT (t)C T LT X LCe(t)

−1
= x̂T (t)P T B(B T X B)−1 B T P x̂(t) + eT (t)C T LT P P T X −1 P P T LCe(t)
(27)
− 2x̂T (t)P T B(GB)−1 GLey (t) = −2x̂T (t)P T B(B T X B)−1 B T X Ley (t)

≤ x̂T (t)P T B(B T X B)−1 B T P x̂(t) + eTy (t)LT X Ley (t)

−1
= x̂T (t)P T B(B T X B)−1 B T P x̂(t) + eTy (t)LT P P T X −1 P P T Ley (t)

209 The condition in (24) is obtained. This completes the proof.

210 Remark 3 Different from the previous works [29][30], the observer gain L and the
211 controller gain K can be solved by Theorem 2.
212 Now we consider the problem of the controller design. Since the system state x(t)
213 and the error e(t) are not completely available and precisely estimated, we can assume
214 that exists some scalars %i > 0 (i = 1, 2) satisfying ke(t)k < %1 ky(tk )k + %2 kŷ(t)k. Then,
215 we can get

kGLCk ke (t)k ≤ κ1 ky(tk )k + κ2 kŷ(t)k (28)

216 Since it is estimated bound that κ1 , κ2 are unknown, an adaptive control method is used
217 to estimate the unknown parameters to facilitate the control design. The estimates of
218 the error are represented as κ̃i = κ̂i − κi (i = 1, 2).

219 3.4. Adaptive event-triggered SMC law synthesis

220 Theorem 3. Consider the uncertain descriptor system (1), the trajectories of the
221 observer (10) can be remained onto the switching surface s(t) = 0 from the initial time
222 by the following SMC law:

−1
u(t) = K x̂ (t) − (GB) [κ̂1 ky (tk )k + κ̂2 kŷ (t)k] − (GB)−1 GLey (t) − (GB)−1 %sign(t)
(29)

223 where the parameter % > 0. The adaptive laws satisfy κ̂˙ 1 = ζ1 ky (tk )k ks(t)k , κ̂˙ 2 =
224 ζ2 kŷ (t)k ks(t)k. ζi > 0 (i = 1, 2) are constantly adaptive gains.

13
225 Proof . Choose the following Lyapunov function:

1 T 1 2 1 2
V (t) = s (t)s(t) + κ̃ + κ̃ (30)
2 2ζ1 1 2ζ2 2

226 Taking the time derivative of s(t), we have

1 ˙ 1
V̇ (t) =s (t) ṡ (t) + κ̃1 κ̂1 + κ̃2 κ̂˙ 2
ζ1 ζ2
1 ˙ 1
=s (t) [GBu(t) + GLCe(t) + GLey (t) − GBK x̂(t)] + κ̃1 κ̂1 + κ̃2 κ̂˙ 2 (31)
ζ1 ζ2
≤ − % ks (t)k ≤ 0

227 Thus, when s (t) 6= 0, s (t) ṡ (t) < 0 always holds, the trajectories of the observer (10)
228 will remain on the sliding surface s(t) = 0 since the initial time.

229 3.5. Event-triggering with the inter execution time T

230 In the following, the theorem gives the lower bound of the inter-execution time T and
231 proved to be positive, thereby Zeno phenomenon is avoided.
232 Theorem 4 Consider the uncertain descriptor system (1) and the control law (29).
233 Then, the inter-execution time tk+1 − tk exists positive lower bound as

tk+1 − tk > T
n o (32)
T = min 1 2 3

234 Let
√ √
ΦC Φ
ϕ1 (t) √ √ −1
ς1 = , ς2 = ΦC
, ς5 = Φ
ϕ (t) ϕ (t) ϕ (t)
1  √ −1 √ −1  √ −1
−1
ς3 = √ k(A + BK + ∆A)k + ϑ Φ + α kCk Φ , ς4 = (GB) GL Φ
λmin ( Φ)
ϕ1 (t) √ √ −1
ς6 = Φ Φ , ζ1 = ς1 ς3 + ς5 , ζ2 = ς1 ς4 + ς2 ς3 + ς6 , ζ3 = ς2 ς4
ϕ (t)
(33)

d
235 Proof . There exist time-vary positive scalars ϕ(t) > 0, ϕ1 (t) > 0 such that ϕ (t) dt x (t) ≤
d kv(t)k
√
236
dt kEx (t)k + ϕ1 (t) kx (t)k. Assume ℘ (t) = k$(t)k , where v (t) = Φey (t), $ (t) =
√
237 Φx (t). For any t ∈ (tk , tk+1 ], there exits α > 0 such that

−1 −1
BKe (t) − (GB) κ̂1 ky (tk )k + κ̂2 kŷ (t)k − (GB) %sign(t) ≤ α ky (t)k . (34)
14
238 Then
√ √ √
kv̇ (t)k k$̇ (t)k kv (t)k Φėy (t) Φẋ (t) Φey (t)
℘˙ (t) ≤ + ≤ √ + √ √
k$ (t)k k$ (t)k k$ (t)k Φx (t) Φx (t) Φx (t)
√ √ √ −1 √
1 ΦC kE ẋ (t)k ϕ (t) ΦC Φ Φx (t)
1
≤ √ + √
ϕ (t) Φx (t) ϕ (t) Φx (t)
 √ √ √ −1 √  √
1 Φ kE ẋ (t)k ϕ1 (t) Φ Φ Φx (t) Φey (t)
+ √ + √  √
ϕ (t) Φx (t) ϕ (t) Φx (t) Φx (t)
 √ √ √ 
ΦC ΦΦey (t) ϕ1 (t) √ √ −1
= + √  kE
√
ẋ (t)k
+ ΦC Φ
ϕ (t) ϕ (t) Φx (t) Φx (t) ϕ (t)
√
ϕ1 (t) √ √ −1 Φey (t)
+ Φ Φ √
ϕ (t) Φx (t)
 √ √ 
 
ΦC Φ 1 √ −1 √ −1 
≤ + ℘ (t) √ k(A + BK + ∆A)k + ϑ Φ + α kCk Φ
ϕ (t) ϕ (t) λmin ( Φ)
√ −1  ϕ (t) √ √ −1 ϕ1 (t) √ √ −1
−1 1
+ (GB) GL Φ ℘ (t) + ΦC Φ + Φ Φ ℘ (t)
ϕ (t) ϕ (t)
= (ς1 + ς2 ℘ (t)) (ς3 + ς4 ℘ (t)) + ς5 + ς6 ℘ (t)

= ς1 ς3 + ς5 + (ς1 ς4 + ς2 ς3 + ς6 ) ℘ (t) + ς2 ς4 ℘2 (t)

(35)

239 That is

℘˙ (t) ≤ ζ1 + ζ2 ℘ (t) + ζ3 ℘2 (t) (36)

√
240 When the value of ℘ (t) evolves from 0 to σ, the lower bound on the execution intervals
√
241 can be decided by the solution T = tk+1 − tk of ℘ (T ) = σ, ℘ (tk ) = 0. Solving the
242 above differential inequality with ℘ (tk ) = 0 as initial condition. we get
 2
243 (i) If ζζ23 − 4 ζζ13 < 0, we have
 
1 = (t) − λ
arctan ≤ T1 + C (37)
ζ3 µ u
q
ζ2 ζ1 ζ22
244 where λ = − 2ζ 3
,µ= ζ3 − 4ζ32
,
15
 √
245 When the value of = (t) evolves from 0 to σ. we get
 
1 −λ
C= arctan
ζ3 µ u
√    (38)
1 σ−λ 1 −λ
arctan ≤T+ arctan
ζ3 µ u ζ3 µ u
246 Obviously, if T → 0, the equation is not true. Thus, there exists a constant parameter
247 1 such that T ≥ 1 .
 2
248 (ii) If ζζ23 − 4 ζζ13 ≥ 0, we assume that ϑ1 , ϑ2 are two real roots of ζ1 + ζ2 ℘ (t) +
249 ζ3 ℘2 (t) = 0 and ϑ1 > ϑ2 . Obviously, ϑ1 < 0, ϑ2 < 0, then
1 ℘ (t) − ϑ1
ln ≤ T + C1 (39)
ζ3 (ϑ1 − ϑ2 ) ℘ (t) − ϑ2
√
250 When the value of = (t) evolves from 0 to σ. we get
1 ϑ1
C1 = ln
ζ3 (ϑ1 − ϑ2 ) ϑ2
√ (40)
1 σ − ϑ1 1 ϑ1
ln √ ≤ T1 + ln
ζ3 (ϑ1 − ϑ2 ) σ − ϑ2 ζ3 (ϑ1 − ϑ2 ) ϑ2
251 Obviously, if T → 0, the equation is not true. Thus, there exists a constant parameter
252 2 such that T ≥ 2 .
253 (iii) When ϑ1 = ϑ2 , we have
1
− = T + C2 (41)
ζ3 (℘ (t) − ϑ1 )
√
254 When the value of = (t) evolves from 0 to σ. we get
1
C2 =
ζ3 ϑ 1
(42)
1 1
− √ ≤T+
ζ 3 ( σ − ϑ1 ) ζ 3 ϑ1

255 Obviously, if T → 0, the equation is not true. Thus, there exists a constant parameter
256 3 such that T ≥ 3 . The equality (32) can be obtained.
257 Remark 4. The positive lower bound of the inter execution time is given in Theorem
258 4, thereby Zeno phenomenon is avoided. To our best knowledge, how to avoid Zeno
259 behavior has not been adequate studied for descriptor system. Since the matrix E is
260 singular, it increases the complexity of the system and makes analysis more difficult. In
261 the paper, the parameters used for scaling corresponding inequalities are introduced to
262 solve the problem. This can be seen from the proof of Theorem 4.
16
263 4. Examples

264 In the section, the effectiveness of the proposed controller design method will be
265 illustrated using examples.
266 Case 1. [4] Simplified model of modeling of oil catalytic cracking is given by

ẋ(t) =(A11 + ∆A11 )x1 (t) + (A12 + ∆A12 )x2 (t) + B1 u(t) + D1 ω(t)
(43)
0 =(A21 + ∆A21 )x1 (t) + (A22 + ∆A22 )x2 (t) + B2 u(t) + D2 ω(t)

267 where x1 (t) stands for a measurable vector, and it can be valve position, blower capac-
268 ity and regenerate temperature. x2 (t) represents the vector of policy, administration,
269 business benefits. u(t) reflects regulation value. ω(t) is the disturbance. The parameter
270 values are as follows A11 = 1, A12 = 2, A21 = 0.5, A22 = −1, B1 = B2 = 1, D1 = D2 = 1.
271 Assume y(t) = x1 (t) and it can be measured. The system matrices can be obtaied
       
1 0 1 2 1 1
E= ,A =  ,B =  , D =  ,
0 0 0.5 −1 1 1
  (44)
h i 1 h i
C = 1 0 ,M =  ,N = 1 0
0

272 We assume the exogenous disturbance

ω(t) = 12e−t sin (t) (45)

273 We give σm = 0.01, x0 = [ 8 2 ]T , x̂0 = [ 7.8 1 ]T , and using Theorem 2, the

274 controller gain K, the observer gain L, triggering matrix Φ and the matrix G are
h i
K = −1.8914 −0.5250
h i
L = 18.1672 −1.4080 (46)
h i h i
Φ = 3.7966 , G = 0.2188 2.4617

We get the switching function:

Z t
s (t) = [0.2188 0] x (t) − [0.2188 0] x (0) − [0.0238 0.3227] x (v) dv
0

275

276 Fig. 3 and Fig. 4 plot the observer’s state trajectories and the error system’s state
277 trajectories respectively, which show that the observer’s state trajectories and the error
17
 20 16
x̂1 (t) e1(t)
x̂2 (t)
15 14 e2(t)

12
10

10
5
8
0
6
−5
4

−10
2

−15 0

−20 −2
0 5 10 15 20 0 5 10 15 20

Fig. 3: The state x̂(t) of the observer. Fig. 4: The state e(t) of the error system.

1
7
0.9

0.8 6

0.7
5
Release interval

0.6

0.5 4
κ̂1
0.4 κ̂2
3
0.3
2
0.2

0.1 1

0
0 5 10 15 20
Times(s) 0
0 5 10 15 20

Fig. 5: The release instants and release interval

Fig. 6: Adaptive parameters estimation.
for observer channel.

18
 s(t)
0.5

0.4

0.3

0.2

0.1

−0.1

−0.2

−0.3

−0.4

−0.5

0 5 10 15 20
Times(s)

Fig. 7: The trajectory of sliding function s(t) for the observer.

278 system’s state trajectories are stable respectively. Fig. 6 illustrates the adaptive param-
279 eter estimation. Fig. 7 plots switching functions s(t). Fig. 5 indicates that the average
280 release period is 0.0285 and 700 of the sampled data are transmitted to the observer
281 using the proposed method, which is only 3.5% of the sampled signals.
282 Case 2. In order to further illustrate the advantages of the proposed method over
283 some existing ones, a comparison result with [12] is presented. Consider mass-spring
284 system as follow:

     T  
0 0 1 0 0 1 0.1
       
       
 0 0 0 1   1   0   0 
A=
 k1 +k2
B = 

,C =   ,D = 
 1    
 (47)

k2
 − m − mc1 0   m   0   0 
 1 m1   1     
k2 k2
m2 −m 2
0 − mc2 0 0 0

285 We give E = I, σm = 0.03, x0 = [ −2 0.6 0.5 0.2 ]T , x̂0 = [ 1.3 1 −1 0.2 ]T .

286 A constant sampling period is given as h = 0.03, and using Theorem 2, the controller
287 gain K, the observer gain L, triggering matrix Φ and the matrix G are
h i
K = −0.0130 −1.4626 −1.1644 −0.4835
h i
L = 0.0911 0.1124 0.0582 −0.0217 (48)
h i h i
Φ = 14.2518 , G = −0.7193 5.4467 4.8843 2.8227
19
 1.5 2
x̂1 (t) e1(t)
x̂2 (t)
1.5 e2(t)
x̂3 (t)
1 e3(t)
1

0.5
0.5

0 0

−0.5
−0.5

−1

−1
−1.5

−1.5 −2
0 10 20 30 40 50 60 0 10 20 30 40 50 60

Fig. 8: The state x̂(t) of the observer . Fig. 9: The state e(t) of the error system.

Then, ω(t) is the same as Case 1. We get the switching function:

s (t) = [−0.7193 5.4467 4.8843 2.8227] x̂ (t) − [−0.7193 5.4467 4.8843 2.8227] x̂ (0)
Z t
− [−7.0802 −13.0485 −15.1909 −0.9597] x̂ (v) dv
0

288

2
1
1.8
0.9
1.6
0.8
1.4
0.7
Release interval

1.2
0.6
1
0.5
0.8
0.4
0.6
0.3
0.4
0.2
0.2
0.1
0
0 10 20 30 40 50 60 0
Times(s) 0 10 20 30 40 50 60

Fig. 10: The release instants and release interval for

Fig. 11: The value of σ(tk+1 ).
observer channel.

289 Using Algorithm 1, set the initial value as σm = 0.03. Based on Remark 1, the desired
290 communication bandwidth and control performance, we can obtain a set of training
291 samples. The values of kx (tk+1 )k−kx (tk )k serves as the input of the NN. The simulation
292 results can be obtained as follows:
20
 0.4
35
0.35

The release instants and release interval

30
0.3

25
κ̂1 0.25
κ̂2
20 0.2

15 0.15

0.1
10

0.05
5

0
0 10 20 30 40 50 60
0 Times(s)
0 10 20 30 40 50 60

Fig. 13: The release instants and release interval

Fig. 12: Adaptive parameters estimation.
by the method of [12].

s(t)
0.5

0.4

0.3

0.2

0.1

−0.1

−0.2

−0.3

−0.4

−0.5

0 10 20 30 40 50 60
Times(s)

Fig. 14: The trajectory of sliding function s(t) for the observer.

293 Fig. 8 and Fig. 9 plot the observer’s state trajectories and the error system’s state
294 trajectories respectively, which show that the observer’s state trajectories and the error
295 system’s state trajectories are stable respectively. Fig. 11 shows the online approxi-
296 mation of σ(tk ). Fig. 12 illustrates the adaptive parameter estimation. Fig. 14 plots
297 switching functions s(t). Fig. 10 indicates that the average release period is 0.2116 and
298 approximately 283 sampled data are transmitted to the observer using our method, which
299 is only 14.15% of the sampled signals. However, we use the same system parameters by
300 the method of [12]. Fig. 13 shows the average release period is 0.1704 and 352 sampled
301 signals are sent to the controller by the method of [12], which is 17.6% of the sampled
21
302 signals. The comparison results show that our proposed method reduces 3.45% of the
303 sample signals transmission. Obviously, the limited communication bandwidth is used
304 more efficiently by the proposed NN event-triggering communication scheme than using
305 the traditional communication scheme.
306 Remark 5. Fig. 10 shows that the communication bandwidth is used more efficiently
307 by the proposed NN event-triggering communication strategy. Fig. 11 indicates that the
308 values of σ(tk ) are relatively small before the system is stable. However, the values of
309 σ(tk ) are relatively large when the system is stable. This can be used to dynamically
310 adjust the event-triggering threshold to sharply reduce the number of transmitted signals.

311 5. Conclusion

312 In this paper, the problem of sliding mode observer-based event-triggering strategy
313 has been studied for uncertain descriptor systems with the exogenous disturbance. A
314 new NN event-triggering communication strategy has been further reduced the utiliza-
315 tion of the communication bandwidth. Then, the existence conditions and the stability
316 conditions of sliding mode dynamics have been presented. An adaptive event-triggered
317 sliding mode controller has been designed such that the resulting system is admissible
318 with the H∞ performance level γ. The Zeno behavior has been discussed for the pro-
319 posed strategy, and we have proved that the continuously sampling and triggering can
320 exclude the Zeno behavior without decomposing the descriptor system. Finally, two
321 examples further indicates that fewer sampled signals need to be transmitted. By com-
322 paring examples, the proposed NN event-triggering communication scheme save more the
323 limited communication bandwidth while at the same time guaranteeing the same control
324 performance.

325 6. Acknowledgements

326 This work was supported by the National Natural Science Foundation of China under
327 Grant 61673100, and Automation for Process Industries Fundamental Research Funds,
328 No. 2013ZCX02.

22
329 References

330 References

331 [1] J. Li, Q. Zhang, Z. Ding, Z. Yi, Sliding mode control for descriptor markovian jump systems with
332 mode-dependent derivative-term coefficient, Nonlinear Dynamics 82 (1-2) (2015) 465–480.
333 [2] Q. Zhang, L. Chao, Z. Xue, Complexity, analysis and control of singular biological systems, Lecture
334 Notes in Control and Information Sciences 421.
335 [3] S. Xu, J. Lam, Y. Zou, J. Li, Robust admissibility of time-varying singular systems with commen-
336 surate time delays, Automatica 45 (11) (2009) 2714–2717.
337 [4] J. Li, Q. Zhang, An integral sliding mode control approach to observer-based stabilization of stochas-
338 tic its descriptor systems, Neurocomputing 173 (2015) 1330–1340.
339 [5] D. Chu, Y. S. Hung, A numerical design of interactors for general descriptor systems, IEEE Trans-
340 actions on Automatic Control 57 (9) 2396–2399.
341 [6] C. Lin, B. Chen, P. Shi, J.-P. Yu, Necessary and sufficient conditions of observer-based stabilization
342 for a class of fractional-order descriptor systems, Systems and Control Letters 112 (2018) 31–35.
343 [7] Y. Tian, Z. Wang, A new integral inequality approach for extended dissipative filters design of
344 singular markovian jump systems with discrete and distributed delays, Circuits Syst. Signal Process.
345 (2019) doi:org/10.1007/s00034-019-01305-9.
346 [8] J. P. Hespanha, P. Naghshtabrizi, Y. Xu, A survey of recent results in networked control systems,
347 Proceedings of the IEEE 95 (2007) 138–162.
348 [9] J. Li, M. J. Er, Y. K. Tan, H. Yu, Z. Peng, Adaptive sampling rate control for networked systems
349 based on statistical characteristics of packet disordering, ISA Transactions 58 (2015) 121–132.
350 [10] C. Peng, Q. L. Han, D. Yue, To transmit or not to transmit: A discrete event-triggered communi-
351 cation scheme for networked takagicsugeno fuzzy systems, IEEE Transactions on Fuzzy Systems 21
352 (2013) 164–170.
353 [11] S. Hu, Y. Zhang, Z. Du, Network-based H∞ tracking control with event-triggering sampling scheme,
354 IET Control Theory and Applications 6 (4) (2012) 533–544.
355 [12] D. Yue, Q. L. Han, J. Lam, Network-based robust H∞ control of systems with uncertainty, Auto-
356 matica 41 (2005) 999–1007.
357 [13] Y. Qi, P. Zeng, W. Bao, Event-triggered and self-triggered H∞ control of uncertain switched linear
358 systems, IEEE Transactions on Systems, Man, and Cybernetics: Systems (2018) 1–13.
359 [14] Y. Wang, T. Zhang, J. Ren, J. Li, Network-based integral sliding mode control for descriptor sys-
360 tems with event-triggered sampling scheme, International Journal of Robust and Nonlinear Control
361 29 (10) (2019) 2757–2776.
362 [15] P. Chen, C. Tai, Event-triggered communication and H∞ control co-design for networked control
363 systems, Automatica 49 (5) (2013) 1326–1332.
364 [16] X. Xiao, Z. Lei, D. W. C. Ho, G. Lu, Event-triggered control of continuous-time switched linear
365 systems, IEEE Transactions on Automatic Control (2018) 1–1.
366 [17] H. Yan, H. Zhang, X. Zhan, Y. Wang, S. Chen, F. Yang, Event-triggered sliding mode control of
23
367 switched neural networks with mode-dependent average dwell time, IEEE Transactions on Systems
368 Man and Cybernetics Systems (2019) 1–11.
369 [18] J. Cheng, H. P. Ju, H. Wang, Event-triggered H∞ control for T-S fuzzy nonlinear systems and its
370 application to truckctrailer system, ISA Trans 65 (2016) 62–71.
371 [19] S. Jun, W. Zidong, N. Yugang, On H∞ sliding mode control under stochastic communication
372 protocol, IEEE Transactions on Automatic Control (2018) 2174 – 2181.
373 [20] C. Edwards, S. K. Spurgeon, Sliding mode control: theory and applications, Taylor and Francis,
374 1998.
375 [21] Y. Niu, D. W. C. Ho, J. Lam, Robust integral sliding mode control for uncertain stochastic systems
376 with time-varying delay, Automatica 41 (2005) 873–880.
377 [22] X. Yan, S. K. Spurgeon, C. Edwards, Sliding mode control for time-varying delayed systems based
378 on a reduced-order observer, Automatica 46 (2010) 1354–1362.
379 [23] C. Edwards, S. K. Spurgeon, Sliding mode stabilization of uncertain systems using only output
380 information, International Journal of Control 62 (1995) 1129–1144.
381 [24] S. Jun, N. Yugang, X. Jing, An event-triggered approach to sliding mode control of markovian jump
382 lur’e systems under hidden mode detections, IEEE Transactions on Systems, Man, and Cybernetics:
383 Systems (2018) 1–12.
384 [25] A. K. Behera, B. Bandyopadhyay, X. Yu, Periodic event-triggered sliding mode control, Automatica
385 96 (2018) 61–72.
386 [26] P. Tabuada, Event-triggered real-time scheduling of stabilizing control tasks, IEEE Transactions on
387 Automatic Control 52 (9) (2007) 1680–1685.
388 [27] D. Zhang, Q. L. Han, X. Jia, Network-based output tracking control for T-S fuzzy systems using
389 an event-triggered communication scheme, Fuzzy Sets and Systems 273 (2015) 26–48.
390 [28] C. Peng, M. Yang, J. Zhang, M. Fei, S. Hu, Network-based H∞ control for T-S fuzzy systems with
391 an adaptive event-triggered communication scheme, Fuzzy Sets and Systems 329 (2017) 61–76.
392 [29] H. He, X. Gao, W. Qi, Distributed event-triggered sliding mode control of switched systems, Journal
393 of the Franklin Institute (2018) doi:10.1016/j.jfranklin.2018.04.020.
394 [30] L. Wu, Y. Gao, J. Liu, H. Li, Event-triggered sliding mode control of stochastic systems via output
395 feedback, Automatica 82 (2017) 79–92.

24
396 Yuzhong Wang received the B.Sc. degree in Mathematics from North-
397 eastern University, ShenYang, China, in 2016, the M.Sc degree in Operations research
398 and cybernetics from Northeastern University, ShenYang, China, in 2018. He is now pur-
399 suing the ph.D. degree in School of computer science and engineering at Northeastern
400 University, Shenyang, China. His research interests include switched systems, descriptor
401 systems, sliding mode control , event-triggered, adaptive control.

402 Tie Zhang, Professor, Department of Mathematics, Northeastern

403 University, Shenyang, China. Research interests: Numerical Methods for Partial Differ-
404 ential Equations and Integro-Differential Equations, Scientific Computation.

405 Junchao Ren received the B.Sc. degree in applied mathematics

406 from Northeastern University, Shenyang, China, in 2000 and the M.Sc. degree in control
407 theory and control engineering from Guangdong University of Technology, Guangzhou, in
408 2003. Dr. Ren is currently an associate professor with college of Sciences, Northeastern
409 University, Shenyang, China. His research interests include robust control, sliding control
410 and singular systems.

25
411 Chen Meng received the ph.D. in Civil Engineering from Wuhan Uni-
412 versity of Technology, Wuhan, China, in 2011. He is currently a lecturer with College of
413 Resources and Civil engineering, Northeastern University, Shenyang, China. His research
414 interests include dynamic properties of high performance fibre reinforced concrete and
415 Intelligent construction of Civil Engineering.

26
Conflict of Interest Form

Conflict of Interest Form

We wish to draw the attention of the Editor to the following facts which may be considered as
potential conflicts of interest and to significant financial contributions to this work. We wish to
confirm that there are no known conflicts of interest associated with this publication and there
416

has been no significant financial support for this work that could have influenced its outcome.

We confirm that the manuscript has been read and approved by all named authors and that
there are no other persons who satisfied the criteria for authorship but are not listed. We further
confirm that the order of authors listed in the manuscript has been approved by all of us.
We confirm that we have given due consideration to the protection of intellectual property
associated with this work and that there are no impediments to publication, including the timing
of publication, with respect to intellectual property. In so doing we confirm that we have
followed the regulations of our institutions concerning intellectual property.

We understand that the Corresponding Author is the sole contact for the Editorial process
(including Editorial Manager and direct communications with the office). He/she is responsible
for communicating with the other authors about progress, submissions of revisions and final
approval of proofs. We confirm that we have provided a current, correct email address which is
accessible by the Corresponding Author and which has been configured to accept email from
[email protected].

Signed by all authors as follows:

[email protected]
[email protected]
[email protected]
Declaration of Interest Statement

Declaration of Interest Statement

We wish to draw the attention of the Editor to the following facts which may be considered as
417
potential conflicts of interest and to significant financial contributions to this work. We wish to
confirm that there are no known conflicts of interest associated with this publication and there
has been no significant financial support for this work that could have influenced its outcome.
Author Contributions Section

The author contributions

418
1) To reduce the use of communication bandwidth, a new neural-network (NN) event-triggering
communication scheme is proposed, which can adaptively adjust event-triggering threshold.

2) Observer-based event-triggered SMC problem for descriptor systems is solved by constructing

event-triggered sliding mode observer. Stability conditions are obtained by constructing a
Lyapunov function. An adaptive event-triggered sliding mode controller is synthesized for driving
the trajectories of the observer to remain the sliding surface from the initial time.

3) An effective strategy is developed to avoid the Zeno behavior without decomposing the
descriptor system, and the positive lower-bound of the inter-execution time intervals is provided.
The existing method can not directly get the positive lower-bound. Thus, the parameters used for
scaling corresponding inequalities are introduced to solve the problem.