Applied Mathematics and Computation 374 (2020) 125041

Contents lists available at ScienceDirect

Applied Mathematics and Computation

journal homepage: www.elsevier.com/locate/amc

New insights on stability of sampled-data systems with

time-delay
Hong-Bing Zeng a, Zheng-Liang Zhai a, Yong He b, Kok-Lay Teo c,d, Wei Wang a,∗
a
College of Electrical and Information Engineering, Hunan University of Technology, Zhuzhou 412007, China
b
School of Automation, China University of Geosciences, Wuhan 430074, China
c
School of Electrical Engineering, Computing and Mathematical Sciences, Curtin University, Perth, WA 6102, Australia
d
Coordinated Innovation Center for Computable Modeling in Management Science, Tianjin University of Finance and Economics, Tianjin
300222, China

a r t i c l e i n f o a b s t r a c t

Article history: This paper analyses the stability of sampled-data systems with time-delay. By employing a
Received 24 July 2019 two-sided looped-functional approach, some improved conditions are derived to guarantee
Revised 6 December 2019
the stability of the system under consideration. Then, based the conditions, some intrinsic
Accepted 5 January 2020
relationships between sampled-data period and time delay are obtained. From the illus-
trative example being solved by the proposed approach, it is observed that the results
Keywords: obtained are significantly better than those obtained by existing methods. More impor-
Stability tantly, from the simulation being carried out, it is discovered that, contrary to the findings
Sampled-data control in previous studies, time-delays in a system may enlarge the interval of sampled-data pe-
Time-delay riods and accelerate the rate of convergence of the system states, rather than deteriorate
Looped-functional the system performance.
© 2020 Elsevier Inc. All rights reserved.

1. Introduction

Since the beginning of the last decade, networked control systems (NCS) have received much attention due to the rapid
development of computer and communication technology and ease in implementation (see [1–5]). In such systems, the
sensors, controllers, actuators and plants are connected through a communication network. It is often assumed that, in this
model, the input to the plant is updated as frequently as it is allowed within the network capacity. However, the sampling
period of the controller may vary if the transient load on the network or the processor is heavy. In addition, the presence
of communication delays are inevitable in a network control system. It is well accepted that either the variation in the
sampling period or the time delays in a system may destabilize the closed-loop system. Therefore, it is of great importance
to develop stability conditions for sampled-data systems which are robust with respect to variations of sampling period and
time delays.
Sampled-data systems have been extensively investigated in the existing literature and many important results have been
obtained [6–11]. The three main methods for the stability analysis of sampled-data systems are: (1) discrete-time methods
[12,13], where a sampled-data system is modeled as a discrete-time system; (2) impulsive system methods [14], where
a sampled-data system is modeled as an impulsive system; and (3) input delay methods [15–18], where a sampled-data
system is treated as a continuous-time system with a delayed input. Recently, a looped-functional approach is proposed in

∗
Corresponding author.
E-mail addresses: [email protected] (H.-B. Zeng), [email protected] (Z.-L. Zhai), [email protected] (Y. He), [email protected] (K.-L. Teo),
[email protected] (W. Wang).

https://doi.org/10.1016/j.amc.2020.125041
0 096-30 03/© 2020 Elsevier Inc. All rights reserved.
2 H.-B. Zeng, Z.-L. Zhai and Y. He et al. / Applied Mathematics and Computation 374 (2020) 125041

[19] to investigate the stability of sampled-data systems. Based on the looped-functional framework, a two-sided looped-
functional that considers the information of intervals x(tk ) to x(t) and x(t) to x(tk+1 ) is proposed in [20]. The conditions
obtained by this approach can produce results which are much less conservative. However, in [20], the sampled-data control
systems are treated as ideal control systems without taking into account the presence of time-delays.
Since the presence of time delays is inevitable in a variety of systems such as industry processes, communication net-
works [21] and biological systems, much effort has been made to the investigation of systems with time delay [22–30].
Lyapunov-based methods have been widely employed to derive stability conditions for time delay systems. However, the
results obtained tend to be rather conservative. Thus, various approaches have been proposed for improvement. See, for
example, model transform approach [31], free-weighting matrix approach [32], integral inequality approach [33–36], and
convex combination approach [37]. For sampled-data systems with time delay, the problem of stability is investigated in
[38], where some stability conditions are reported based on Wirtinger’s inequality. By employing an augmented Lyapunov
functional approach, some improved conditions are obtained in [39,40]. However, in these papers, only the information of
the intervals x(tk ) to x(t) and x(tk − τ ) to x(t − τ ) is used, while the information on the intervals x(t) to x(tk+1 ) and x(t − τ )
to x(tk+1 − τ ) is ignored. Thus, the stability results obtained are still rather conservative. Additionally, in previous studies,
it is without exception that the presence of time delays is regarded as a cause for the destabilization of the system and
deterioration of the system performance.
In this paper, we revisit the problem of stability of sampled-data systems with time delay. A new two-sided looped
functional that takes in account the information of intervals x(tk ) to x(t), x(t) to x(tk+1 ), x(tk − τ ) to x(t − τ ) and x(t − τ )
to x(tk+1 − τ ) is proposed. Based on this two-sided looped functional, some sufficient conditions are derived to guarantee
the stability of the system under consideration. The combined impact of the communication delay and sampling action
is critically investigated and the outcomes are revealed. Through solving a given example by the approach proposed, it is
clearly observed that the results obtained are much superior to those obtained by other methods in the literature. More
importantly, it is discovered that time delays in a system may improve rather than deteriorate the system performance.
Notation: The inverse and transpose of a matrix are denoted by superscripts ‘−1’ and ‘T’, respectively; n-dimensional
Euclidean space is denoted by Rn ; the space of n × m real matrices is denoted by Rn × m ; P > 0 means that the matrix P is
symmetric and positive definite; diag{. . .} represents a block-diagonal matrix; the identity matrix and the zero matrix are
denoted by I and 0, respectively; ‘∗ ’ denotes the symmetric term in a symmetric matrix; and He{X } = X + X T .

2. Preliminaries

Consider the following continuous-time system:

x˙ (t ) = Ax(t ) + Bu(t ) (1)

where x(t) ∈ Rn
is the state vector and u(t) ∈ Rm
is the control input. A ∈ and B ∈ Rn × n Rn × m
are known system matrices.
The sampling instant time, tk , of the sampler is assumed such that the following conditions are satisfied.
tk+1 − tk = hk ∈ [h1 , h2 ], k = 0, 1, 2 . . . (2)
where hk is the updating period of two neighbouring instants, tk and tk+1 , h1 and h2 indicate the lower bound and upper
bound of the sampling period.
Let τ be a time delay in a sampled-data control systems, arising due to the transmission through a communication
network. Thus, by taking into consideration of the communication delay τ , the control input u(t) is of the form:

u(t ) = Kx(tk − τ ) (3)

where tk ≤ t < tk+1 and K is a given controller gain. Then, the corresponding closed-loop system is formulated as

x˙ (t ) = Ax(t ) + B̄x(tk − τ ), t ∈ [tk , tk+1 ) (4)

with B̄ = BK. Let ηk = hk + τ , and the upper bound of ηk is denoted by ηM = h2 + τ .
Before presenting our main results, we recall the following lemma, which is needed in the derivation of the main results.

Lemma 1 ([20]). Let x: [α , β ] → Rn be a differentiable function. Then, for a given matrix H > 0 and any matrix W1 , W2 , W3 ,
the following inequalities hold:
 β
− x˙ T (s )H x˙ (s )ds ≤ i , i = 1, 2 (5)
α
where
 1 1
    
i = (β − α )θ T W1 H −1W1T + W2 H −1W2T + W3 H −1W3T θ + He θ T W1 χ1 + W2 χ2 + W3 χ3(i) θi ,
3 5
χ1 = e1 − e2 , χ2 = e1 + e2 − 2e3 ,
χ3(1) = e1 − e2 − 6e3 + 6e4 ,
χ3(2) = e1 − e2 + 6e3 − 6e4 ,
 H.-B. Zeng, Z.-L. Zhai and Y. He et al. / Applied Mathematics and Computation 374 (2020) 125041 3

   β T
1 β T 2 s
θ1 = x T ( β ) x T ( α ) x (s )ds xT (u )duds ,
ν α ν2 α α
   β β T
1 β T 2
θ2 = x T ( β ) x T ( α ) x (s )ds xT (u )duds ,
ν α ν2 α s

ej = 0n×( j−1 )n In 0n×(4− j )n , j = 1, 2, 3, 4,

ν = β − α and θ is any vector with an appropriate dimension with reference to one of Wi , i = 1, 2, 3.

3. Main results

For the simplification of presentation, the following nomenclatures are defined as given below.
 t
1
v1 (t ) = x(s )ds,
τ t−τ
 t  s
2
v2 (t ) = x(u )duds,
τ t−τ t−τ
2

v3 (t ) = x(t ) − x(tk ),
v4 (t ) = x(t ) − x(tk+1 ),
v5 (t ) = x(t − τ ) − x(tk − τ ),
v6 (t ) = x(t − τ ) − x(tk+1 − τ ),
 t
1
v7 (t ) = x(s )ds,
t − tk tk
 t s
2
v8 (t ) = x(u )duds,
(t − tk ) 2
tk tk
 tk+1
1
v9 (t ) = x(s )ds,
tk+1 − t t
 tk+1  tk+1
2
v10 (t ) = x(u )duds,
(tk+1 − t ) t s
2

 T
τ2 T
ζ1 (t ) = x (t ) x (t − τ ) τ v1 (t )
T T
v2 (t ) ,
2
T
ζ2 (t ) = xT (t ) x˙ T (t ) ,
T
ζ3 (t ) = (tk+1 − t )vT3 (t ) (t − tk )vT4 (t ) (tk+1 − t )vT5 (t ) (t − tk )vT6 (t ) ,
T
ζ4 (t ) = vT3 (t ) vT4 (t ) vT5 (t ) vT6 (t ) ,
T
ζ5 (s ) = xT (s ) xT (s − τ ) ,
T
ζ6 (t ) = vT1 (t ) vT2 (t ) vT7 (t ) vT8 (t ) vT9 (t ) vT10 (t ) ,
T
ζ (t ) = ζ5T (t ) ζ5T (tk ) ζ5T (tk+1 ) xT (t − ηk ) x˙ T (t − τ ) ζ6T (t ) ,
T
ζ˜ (t ) = ζ T (t ) x˙ T (t ) ,
ei = 0n×(i−1 )n In 0n×(14−i )n , i = 1, 2, . . . , 14,
e˜ j = 0n×( j−1 )n In 0n×(15− j )n , j = 1, 2, . . . , 15
Now, the stability conditions are presented in the following theorem.

Theorem 1. For given h2 ≥ h1 ≥ 0 and τ ≥ 0, suppose that there exist matrices P > 0, S > 0, Z1 > 0, Z2 = Z2T , Z3 > 0, Q1 , Q2 ,
G = GT , R1 = RT1 , R2 > 0, R3 = RT3 , R4 > 0, Ni , Mi , Li (i = 1, 2, 3 ), Y j ( j = 1, 2, . . . , 8 ) such that the following LMIs are satisfied
for hk ∈ [h1 , h2 ],
Z2 + Z3 > 0 (6)

R1 + Z3 > 0 (7)

R3 + Z2 > 0 (8)
4 H.-B. Zeng, Z.-L. Zhai and Y. He et al. / Applied Mathematics and Computation 374 (2020) 125041

⎡  √ ⎤
φ1 + h k φ2 hk 1 τ 2
⎢ ⎥
⎣ ∗ − 1 0 ⎦<0 (9)
∗ ∗ −
⎡  √
2
⎤
φ1 + h k φ3 hk 3 τ 2
⎣ ∗ − 3 0 ⎦<0 (10)
∗ ∗ − 2

where
φ1 = He{T1 P2 + T5 (Q1 6 + Q2 7 ) + M1 13 + M2 14 + M3 15 + N1 16 + N2 17 + N3 18 + Y1 19
+ Y2 20 + L1 21 + L2 22 + L3 23 + Y3 24 + Y4 25 − Y5 21 − Y6 16 − Y7 30 − Y8 31 } + T3 S3
−T4 S4 + τ  T Z1  + h2  T Z2  + ηM  T Z3  ,
φ2 = He{T8 (Q1 6 + Q2 7 ) + T9 Q1 10 + Y6 27 + Y8 29 } + T7 G7 +  T R1  + eT8 R3 e8 ,
φ3 = He{T11 (Q1 6 + Q2 7 ) + T12 Q1 10 + Y5 26 + Y7 28 } − T7 G7 +  T R2  + eT8 R4 e8 ,
1 = N1 N2 N3 Y2 Y4 ,

2 = L1 L2 L3 Y3 ,

3 = M1 M2 M3 Y1 ,
1 = diag{R2 , 3R2 , 5R2 , R4 , Z2 + Z3 },
2 = diag{Z1 , 3Z1 , 5Z1 , Z3 },
3 = diag{R1 + Z3 , 3(R1 + Z3 ), 5(R1 + Z3 ), R3 + Z2 },
 = Ae1 + B̄e4 ,
T
1 = eT1 eT2 τ eT9 τ 2 eT ,
2 10
T
2 =  T
eT8 eT1 − eT2 τ (eT9 − eT2 ) ,
T T
3 = eT1  T , 4 = eT2 eT8 ,
T
5 = eT3 − eT1 eT1 − eT5 eT4 − eT2 eT2 − eT6 ,
T
6 = eT1 − eT3 eT1 − eT5 eT2 − eT4 eT2 − eT6 ,
T T
7 = eT5 eT6 , 8 =  T 0 eT8 0 ,
T
9 = eT1 − eT3 0 eT2 − eT4 0 ,
T
10 =  T T eT8 eT8 ,
T
11 = 0  T
0 eT8 ,
T
12 = 0 eT1 − eT5 0 eT2 − eT6 ,
13 = e1 − e3 , 14 = e1 + e3 − 2e11 ,
15 = e1 − e3 − 6e11 + 6e12 , 16 = e5 − e1 ,
17 = e5 + e1 − 2e13 , 18 = e5 − e1 + 6e13 − 6e14 ,
19 = e2 − e4 , 20 = e6 − e2 ,
21 = e1 − e2 , 22 = e1 + e2 − 2e9 ,
23 = e1 − e2 − 6e9 + 6e10 , 24 = e3 − e4 ,
25 = e4 − e7 , 26 = Ae11 + B̄e4 ,
27 = Ae13 + B̄e4 , 28 = Ae12 + B̄e4 ,
29 = Ae14 + B̄e4 , 30 = 2e11 − 2e3 ,
31 = 2e5 − 2e13 .
Then, the closed-loop system (4) under the control input (3) with (2) being satisfied is stable.

Proof. First, choose the following Lyapunov functional for system (4):
W (t ) = V (t ) + V (t ) (11)
 H.-B. Zeng, Z.-L. Zhai and Y. He et al. / Applied Mathematics and Computation 374 (2020) 125041 5

5 6
where V (t ) = j=1 V j (t ) and V (t ) = j=1 V j (t ). Here, their components are given by

V1 (t ) = ζ1T (t )P ζ1 (t ),
 t
V2 (t ) = ζ2T (s )Sζ2 (s )ds,
t−τ
 0  t
V3 (t ) = x˙ T (u )Z1 x˙ (u )dudθ ,
−τ t+θ
 −τ  t
V4 (t ) = x˙ T (u )Z2 x˙ (u )dudθ ,
− ηM t+θ
 0  t
V5 (t ) = x˙ T (u )Z3 x˙ (u )dudθ ,
− ηM t+θ

V1 (t ) = 2ζ3T (t )[Q1 ζ4 (t ) + Q2 ζ5 (tk+1 )],

V2 (t ) = (tk+1 − t )(t − tk )ζ5T (tk+1 )Gζ5 (tk+1 ),
 t
V3 (t ) = (tk+1 − t ) x˙ T (s )R1 x˙ (s )ds,
tk
 tk+1
V4 (t ) = −(t − tk ) x˙ T (s )R2 x˙ (s )ds,
t
 t−τ
V5 (t ) = (tk+1 − t ) x˙ T (s )R3 x˙ (s )ds,
t k −τ
 tk+1 −τ
V6 (t ) = −(t − tk ) x˙ T (s )R4 x˙ (s )ds
t−τ

Then, by differentiating W(t), we have


5 
6
W˙ (t ) = V˙ j (t ) + V˙ j (t ) (12)
j=1 j=1

where

V˙ 1 (t ) = 2ζ T (t )T1 P 2 ζ (t ),
V˙ 2 (t ) = ζ T (t )(T3 S3 − T4 S4 )ζ (t ),
V˙ 3 (t ) = ζ T (t )τ  T Z1  ζ (t ) + J1 ,
V˙ 4 (t ) = ζ T (t )h2  T Z2  ζ (t ) + J2 ,
V˙ 5 (t ) = ζ T (t )ηM  T Z3  ζ (t ) + J3 ,

V˙ 1 (t ) = 2ζ T (t ){T5 (Q1 6 + Q2 7 ) + (tk+1 − t )[T8 (Q1 6 + Q2 7 ) + T9 Q1 10 ] + (t − tk )

× [T11 (Q1 6 + Q2 7 ) + T12 Q1 10 ]}ζ (t ),
V˙ 2 (t ) = ζ T (t )[(tk+1 − t ) − (t − tk )]T7 G7 ζ (t ),
V˙ 3 (t ) = ζ T (t )[(tk+1 − t ) T R1  ]ζ (t ) + J4 ,
V˙ 4 (t ) = ζ T (t )[(t − tk ) T R2  ]ζ (t ) + J5 ,
V˙ 5 (t ) = ζ T (t )[(tk+1 − t )eT8 R3 e8 ]ζ (t ) + J6 ,
V˙ 6 (t ) = ζ T (t )[(t − tk )eT8 R4 e8 ]ζ (t ) + J7 .
with
 t
J1 = − x˙ T (s )Z1 x˙ (s )ds,
t−τ
 t−τ
J2 = − x˙ T (s )Z2 x˙ (s )ds,
t−ηM
 t
J3 = − x˙ T (s )Z3 x˙ (s )ds,
t−ηM
 t
J4 = − x˙ T (s )R1 x˙ (s )ds,
tk
6 H.-B. Zeng, Z.-L. Zhai and Y. He et al. / Applied Mathematics and Computation 374 (2020) 125041

 tk+1
J5 = − x˙ T (s )R2 x˙ (s )ds,
t
 t−τ
J6 = − x˙ T (s )R3 x˙ (s )ds,
t k −τ
 tk+1 −τ
J7 = − x˙ T (s )R4 x˙ (s )ds.
t−τ

From ηM ≥ ηk and Z2 + Z3 > 0, it follows that

 t−τ  t
J2 + J3 = − x˙ T (s )Z2 x˙ (s )ds − x˙ T (s )Z3 x˙ (s )ds
t−ηk t−ηk
 t−ηk
− x˙ T (s )(Z2 + Z3 )x˙ (s )ds
t−ηM
 t−τ  t
≤− x˙ T (s )Z2 x˙ (s )ds − x˙ T (s )Z3 x˙ (s )ds
t−ηk t−ηk
 t−τ  t
=− x˙ T (s )Z2 x˙ (s )ds − x˙ T (s )Z3 x˙ (s )ds
t k −τ tk
 tk  t k −τ
− x˙ T (s )Z3 x˙ (s )ds − x˙ T (s )(Z2 + Z3 )x˙ (s )ds.
t k −τ t−ηk

Thus, we get
J2 + J3 + J4 + J6 ≤ Jˆ2 + Jˆ3 + Jˆ4 + Jˆ6 (13)
where
 t
Jˆ2 = − x˙ T (s )(R1 + Z3 )x˙ (s )ds,
tk
 t−τ
Jˆ3 = − x˙ T (s )(R3 + Z2 )x˙ (s )ds,
t k −τ
 t k −τ
Jˆ4 = − x˙ T (s )(Z2 + Z3 )x˙ (s ),
t−ηk
 tk
Jˆ6 = − x˙ T (s )Z3 x˙ (s )ds.
t k −τ

Applying Lemma 1, we obtain

 

3
1
J1 ≤ ζ (t )
T
τ Li Z −1 LT + He{L1 21 + L2 22 + L3 23 } ζ (t ), (14)
2i − 1 1 i
i=1
 

3
1
Jˆ2 ≤ ζ (t ) (t − tk )
T
Mi (R1 + Z3 )−1 MiT + He{M1 13 + M2 14 + M3 15 } ζ (t ), (15)
2i − 1
i=1

Jˆ3 ≤ ζ T (t )[(t − tk )Y1 (R3 + Z2 )−1Y1T + He{Y1 19 }]ζ (t ), (16)

Jˆ4 ≤ ζ T (t )[(tk+1 − t )Y4 (Z2 + Z3 )−1Y4T + He{Y4 25 }]ζ (t ), (17)

 

3
1
J5 ≤ ζ (t ) (tk+1 − t )
T
2 Ni + He{N1 16 + N2 17 + N3 18 }
Ni R−1 T
ζ (t ), (18)
2i − 1
i=1

Jˆ6 ≤ ζ T (t )[τ Y3 Z3−1Y3T + He{Y3 24 }]ζ (t ), (19)

J7 ≤ ζ T (t )[(tk+1 − t )Y2 R−1

4 Y2 + He{Y2 20 }]ζ (t ).
T
(20)
Similar to arguments used in [20], we introduce the following equalities
0 = 2ζ T (t )Y5 [(t − tk )26 − 13 ]ζ (t ) (21)

0 = 2ζ T (t )Y6 [(tk+1 − t )27 − 16 ]ζ (t ) (22)

 H.-B. Zeng, Z.-L. Zhai and Y. He et al. / Applied Mathematics and Computation 374 (2020) 125041 7

0 = 2ζ T (t )Y7 [(t − tk )28 − 30 ]ζ (t ) (23)

0 = 2ζ (t )Y8 [(tk+1 − t )29 − 31 ]ζ (t )

T
(24)
where Yi (i = 5, 6, 7, 8 ) are any matrices with appropriate dimensions.
Adding the right-hand of (21)–(24) to W˙ (t ), it follows from (14)–(20) that
 
(tk+1 − t ) (t − tk )
W˙ (t )  ζ T (t ) 1 + 2 ζ (t ) (25)
hk hk
where
 1 = φ1 + h k φ2 + φ4 ,
 2 = φ1 + h k φ3 + φ5 ,
with

3
1
φ4 = 2 Ni + τ Li Z1 Li ) + hkY2 R4 Y2 + hkY4 (Z2 + Z3 ) Y4 + τ Y3 Z3 Y3 ,
(h Ni R−1 T −1 T −1 T −1 T −1 T
2i − 1 k
i=1


3
1
φ5 = {h Mi (R1 + Z3 )−1 MiT + τ Li Z1−1 LTi } + hkY1 (R3 + Z2 )−1Y1T + +τ Y3 Z3−1Y3T .
2i − 1 k
i=1

Note that 1 < 0 and 2 < 0 are convex in hk ∈ [h1 , h2 ], it follows that LMIs (9) and (10) are equal to 1 < 0 and
2 < 0, respectively, in the sense of Schur complement. Thus W˙ (t ) < 0, which means that the closed-loop system (4) is
stable. This completes the proof.

Remark 1. The Lyapunov functional (11) consists of two parts, V(t) and V (t ), where V (t ) satisfies the boundary conditions of
looped-functional obtained in [19], i.e., V (tk ) = V (tk+1 ) = 0. Inspired by the two-sided looped-functional approach proposed
in [20], V (t ) contains the information of intervals x(tk ) to x(t), x(t) to x(tk+1 ), x(tk − τ ) to x(t − τ ) and x(t − τ ) to x(tk+1 − τ ).
This information plays an important role for the derivation of results with reduced conservativeness.

Remark 2. The relationships between the sampling period and the time delay are being fully considered through di-
 t−τ  t−τ  t −τ t t
viding t−η x˙ T (s )Z2 x˙ (s )ds into t −τ x˙ T (s )Z2 x˙ (s )ds and t−k η x˙ T (s )Z2 x˙ (s )ds, and t−η x˙ T (s )Z3 x˙ (s )ds into t x˙ T (s )Z3 x˙ (s )ds,
 tk k
 tk −τ T
k k k k

tk −τ x˙ (s )Z3 x˙ (s )ds, and t−ηk x˙ (s )Z3 x˙ (s )ds. Now, we combine integral terms that are with the same integral interval, where
T

R1 , R3 and Z2 are not required to be positive. On this basis, we can derive results with reduced conservativeness.

In the proof of the following theorem, we remove equalities (21)–(24), keep the term x˙ (t ), instead of expressing x˙ (t ) in
terms of the state of the system while taking the derivative of the Lyapunov functional. Then, we add the right-hand side
of the following equality
0 = 2ζ˜ T (t )T [x˙ (t ) − Ax(t ) − B̄u(t )] (26)
to the derivative of the Lyapunov functional.

Theorem 2. For given h2 ≥ h1 ≥ 0 and τ ≥ 0, system (4) under the control input (3) with (2) being satisfied is stable if there
˜ i , L˜i (i = 1, 2, 3 ),
˜ i, M
exist matrices P > 0, S > 0, Z1 > 0, Z2 = Z2T , Z3 > 0, Q1 , Q2 , G = GT , R1 = RT1 , R2 > 0, R3 = RT3 , R4 > 0, N
Y j ( j = 1, 2, 3, 4 ), and T, such that LMIs (6)–(8), and (27) and (28) below are satisfied for hk ∈ [h1 , h2 ],
˜
⎡  √ ⎤
φ˜ 1 + hk φ˜ 2 hk ˜ 1 τ ˜2
⎣ ∗ − 1 0 ⎦<0 (27)
∗ ∗ − 2

⎡  √ ⎤
φ˜ 1 + hk φ˜ 3 hk ˜ 3 τ ˜2
⎣ ∗ − 3 0 ⎦<0 (28)
∗ ∗ − 2

where
φ˜ 1 = He{ ˜ T P
1
˜2+ ˜ T ( Q1 
5
˜ 6 + Q2  ˜ 7) + M ˜ 1
˜ 13 + M ˜ 2
˜ 14 + +M ˜ 3˜ 15 + N ˜ 1˜ 16 + N˜ 2 ˜ 3
˜ 17 + N ˜ 18
+ Y˜1 
˜ 19 + Y˜2  ˜ 20 + L˜1  ˜ 21 + L˜2 ˜ 22 + L˜3 ˜ 23 + Y˜3 ˜ 24 + Y˜4 ˜ 25 + T  ˜ 26 } +  ˜ S
T ˜3− ˜ S
T ˜4
3 4
+ τ e˜15 Z1 e˜15 + h2 e˜15 Z2 e˜15 + ηM e˜15 Z3 e˜15 ,
T T T

φ˜ 2 = He{ ˜ T ( Q1 
8
˜ 6 + Q2  ˜ 7) +  ˜ T Q1 
9
˜ 10 } + ˜ T G
7
˜ 7 + e˜T R1 e˜15 + e˜T R3 e˜8 ,
15 8

φ˜ 3 = He{ ˜ T ( Q1 
11
˜ 6 + Q2  ˜ 7) +  ˜ T Q1 
12
˜ 10 } − ˜ T G
7
˜ 7 + e˜T R2 e˜15 + e˜T R4 e˜8 ,
15 8
8 H.-B. Zeng, Z.-L. Zhai and Y. He et al. / Applied Mathematics and Computation 374 (2020) 125041

Table 1
Maximal value of h̄ for different τ .

Methods 0 0.1 0.2 0.4 0.6

[3] 1.68 1.23 1.06 0.78 0.54

[38] 1.3659 1.2232 1.0883 0.8286 0.5770
[39] 1.7208 1.2955 1.1328 0.8682 0.6202
[40] 1.7208 1.4688 1.3122 1.0459 0.7877
[41] 1.7294 1.6266 1.4515 1.1537 0.8617
Theorem 1 1.7294 1.9223 2.0850 1.7843 1.0266
Theorem 2 1.7294 1.9214 2.0644 1.5945 1.0234

˜1 = N
˜1 ˜2
N ˜3
N Y˜2 Y˜4 ,
˜ 2 = L˜1 L˜2 L˜3 Y˜3 ,
˜3 = M
˜1 ˜2
M ˜3
M Y˜1 ,

˜ 26 = e˜15 − Ae˜1 − B̄e˜4 ,

3 are defined in Theorem 1, and i are defined as for i (i = 1, 2, . . . , 25 ) in Theorem 1 with  being replaced by
1, 2,
˜
e˜15 and ej being replaced by e˜ j ( j = 1, 2, . . . , 14 ).

Remark 3. The LMIs of Theorem 2 are affine in A and B̄, hence A and B̄ reside in the uncertain polytope


k 
k
= γi i , γi ≥ 0, γi = 1, i = [Ai , B̄i ] (29)
i=1 i=1

We can examine the robust stability of the system through solving these LMIs simultaneously for all of the k vertices i
using the same decision matrices.

Remark 4. For the problem of exponential stability, we follow the approach proposed in [42], where we set x̄(t ) = eλt x(t ).
Then, by using the input delay method introduced in [17], we rewrite system (4) as
x̄˙ (t ) = (A + λI )x̄(t ) + eλτ̄ (t ) B̄x̄(t − τ̄ (t )) (30)
where τ̄ (t ) = t − tk + τ , eλτ̄ (t ) ∈ [ρ1 , ρ2 ] with ρ1 = eλτ and ρ2 = eληM . Clearly, (30) can be rewritten as in a polytopic form:


2
x̄˙ (t ) = μi [(A + λI )x̄(t ) + ρi B̄x̄(t − τ̄ (t )))] (31)
i=1

ρ −eλτ (t ) eλτ (t ) −ρ
where μ1 = 2ρ −ρ , μ2 = ρ −ρ 1 . Thus, through the analysis of the robust stability of system (31), we can show that
2 1 2 1
system (4) is exponentially stable with decay rate λ > 0.

4. Numerical examples

This section provides a numerical example to demonstrate the improvement and advantage of the criteria obtained in
this paper.

Example 1. Consider system (4) with

   
0 1 0
A= , B̄ = −3.75 −11.5 (32)
0 −0.1 0.1

This example has been used frequently in the literature, see, for example, Liu and Fridman [3,38], Zhang et al. [39,40],
Zeng et al. [41]. Setting h2 = h1 = h̄, the maximal sampling periods calculated by Theorems 1 and 2 are listed in Table 1
along with those obtained in [3,38–41]. From the table, we observe that the maximal sampling periods obtained in previous
studies decrease with the increase of time delay. It is also observed that, for various τ , the result obtained by using the
proposed approach are much less conservative when compared with those obtained using other approaches in the literature.
It is particularly worth noting that, contrary to the findings in previous studies, the maximal sampling periods with τ = 0.1
and τ = 0.2 are larger than that of τ = 0, which means that the time delay in the system results in the enlargement, rather
than reduction, of the upper bound of the sampling periods.
For comparison, setting ηM = 1, the maximal exponential decay rates λ obtained by Theorem 2 are listed in Table 2
along with those obtained in [3,38–40]. It can be seen from the table that decay rates obtained by the proposed approach
are larger than those obtained in the existing literature. To further demonstrate the advantage of the proposed approach,
setting ηM = 1.7, the maximum decay rates λ computed by Theorem 2 for various τ are listed in Table 3. It is clearly
H.-B. Zeng, Z.-L. Zhai and Y. He et al. / Applied Mathematics and Computation 374 (2020) 125041 9

Table 2
Maximal value of λ for different τ with ηM = 1.

Methods 0 0.1 0.2 0.4 0.6

[3] 0.23 0.20 0.15 0.10 0.07

[38] 0.31 0.26 0.23 0.17 0.12
[39] 0.39 0.30 0.25 0.19 0.14
[40] 0.39 0.37 0.33 0.28 0.21
Theorem 2 0.39 0.39 0.38 0.33 0.27

Table 3
Maximal value of λ for different τ with ηM = 1.7.

τ 0 0.1 0.2 0.3 0.4

Theorem 2 0.0109 0.1278 0.1309 0.0871 0.0466

Fig. 1. State response of system (32) in the case of h̄ = 1.7, τ = 0.

Fig. 2. State response of system (32) in the case of h̄ = 1.7, τ = 0.2.

10 H.-B. Zeng, Z.-L. Zhai and Y. He et al. / Applied Mathematics and Computation 374 (2020) 125041

Fig. 3. State response of system (32) in the case of h̄ = 2.0, τ = 0.2.

observed in the table that, for this example, a larger exponential rate is obtained with the introduction of time delay, which
means that time delays in a system may, in fact, accelerate the rate of convergence of the system state.
For the verification of time delay that can accelerate the rate of convergence of the system state and enlarge the upper
bound of the sampling periods, simulations are carried out for different h̄ and τ = 0 under the same initial state x(0 ) =
[2, −1.8]. The outcomes are depicted in Figs. 1–3, respectively. It is observed from Figs. 1 and 2 that, for h̄ = 1.7, the state of
the system converges to zero within 30 seconds when τ = 0.2, while it takes 200 seconds to converge to zero when τ = 0.
That is to say, the time delay τ reduce the time of convergence of the system state. In comparison with Figs. 1 and 3, it
is observed that when h̄ = 2 and τ = 0.2, the states of the system converge more rapidly than that of h̄ = 1.7 and τ = 0.
In other words, the time delay τ is able to expedite the convergence of the system states under larger sampling periods.
Therefore, it can be summarized from the simulation results that the introduction of an appropriate time delay in a system
may, instead, enlarge the allowable sampling period and accelerate the convergence rate of the system. In other words, it is
possible to utilize time delays to improve the system performance. It is an interesting future research topic.

5. Conclusion

This paper investigates the problem of stability of aperiodic sampled-data control systems with time-delay. A new
looped-functional that takes into account the information of intervals x(tk ) to x(t), x(t) to x(tk+1 ), x(tk − τ ) to x(t − τ ) and
x(t − τ ) to x(tk+1 − τ ) is proposed. Based on the free-matrix-based integral inequality approach, less conservative conditions
are derived. A numerical example is given to demonstrate that the conditions obtained outperform others in the existing lit-
erature. What’s more, it discovers a surprising phenomenon, namely time delays in a system may result in the improvement
rather than deterioration of the system performance. This is an interesting future research topic.

Acknowledgments

This work was supported by the National Natural Science Foundation of China (61741308, 61703153, 61672225), the Nat-
ural Science Fund of Hunan Province (2018JJ2096), and the Scientific Research Fund of Hunan Provincial Education Depart-
ment (19B149, 19C0582).

References

[1] X.M. Zhang, Q.L. Han, Event-triggered dynamic output feedback control for networked control systems, IET Control Theory Appl. 8 (4) (2014) 226–234.
[2] T.H. Lee, J.H. Park, H. Jung, Network-based H∞ state estimation for neural networks using imperfect measurement, Appl. Math. Comput. 316 (2018)
205–214.
[3] K. Liu, E. Fridman, Networked-based stabilization via discontinuous Lyapunov functionals, Int. J. Robust Nonlinear Control 22 (4) (2012) 420–436.
[4] X.H. Chang, Q. Liu, Y.M. Wang, J. Xiong, Fuzzy peak-to-peak filtering for networked nonlinear systems with multipath data packet dropouts, IEEE Trans.
Fuzzy Syst. 27 (3) (2019) 436–446.
[5] X.H. Chang, C. Yang, J. Xiong, Quantized fuzzy output feedback H∞ control for nonlinear systems with adjustment of dynamic parameters, IEEE Trans.
Syst. Man Cybern. 49 (10) (2018) 2005–2015.
[6] T. Chen, B.A. Francis, Optimal Sampled-Data Control Systems, Springer-Verlag, Tokyo, 1995.
[7] Y. Liu, S.M. Lee, Improved results on sampled-data synchronization of complex dynamical networks with time-varying coupling delay, Nonlinear Dyn.
81 (1-2) (2015) 931–938.
[8] Y. Liu, J.H. Park, B.Z. Guo, Y. Shu, Further results on stabilization of chaotic systems based on fuzzy memory sampled-data control, IEEE Trans. Fuzzy
Syst. 26 (2) (2018) 1040–1045.
[9] C.Y. Kao, An IQC approach to robust stability of aperiodic sampled-data systems, IEEE Trans. Autom. Control 61 (8) (2016) 2219–2225.
 H.-B. Zeng, Z.-L. Zhai and Y. He et al. / Applied Mathematics and Computation 374 (2020) 125041 11

[10] H.B. Zeng, K.L. Teo, Y. He, W. Wang, Sampled-data-based dissipative control of T-S fuzzy systems, Appl. Math. Model. 65 (2019) 415–427.
[11] Z. Wu, P. Shi, H. Su, J. Chu, Stochastic synchronization of Markovian jump neural networks with time-varying delay using sampled data, IEEE Trans.
Cybern. 43 (6) (2013) 1796–1806.
[12] H. Fujioka, A discrete-time approach to stability analysis of systems with aperiodic sample-and-hold devices, IEEE Trans. Autom. Control 54 (9) (2009)
2440–2445.
[13] Y.S. Suh, Stability and stabilization of nonuniform sampling systems, Automatica 44 (12) (2008) 3222–3226.
[14] P. Naghshtabrizi, J.P. Hespanha, A.R. Teel, Exponential stability of impulsive systems with application to uncertain sampled-data systems, Syst. Control
Lett. 57 (5) (2008) 378–385.
[15] L. Mirkin, Some remarks on the use of time-varying delay to model sample-and-hold circuits, IEEE Trans. Autom. Control 52 (6) (2007) 1109–1112.
[16] E. Fridman, A. Seuret, J.P. Richard, Robust sampled-data stabilization of linear systems: An input delay approach, Automatica 40 (8) (2004) 1141–1446.
[17] E. Fridman, A refined input delay approach to sampled-data control, Automatica 46 (2) (2010) 421–427.
[18] T.H. Lee, J.H. Park, Stability analysis of sampled-data systems via free-matrix-based time-dependent discontinuous Lyapunov approach, IEEE Trans.
Autom. Control 62 (7) (2017) 3653–3657.
[19] A. Seuret, A novel stability analysis of linear systems under asynchronous samplings, Automatica 48 (1) (2012) 177–182.
[20] H.B. Zeng, K.L. Teo, Y. He, A new looped-functional for stability analysis of sampled-data systems, Automatica 82 (2017) 328–331.
[21] X.H. Chang, R. Huang, J.H. Park, Robust guaranteed cost control under digital communication channels, IEEE Trans. Ind. Inf. (2019), doi:10.1109/TII.2019.
2916146.
[22] K. Gu, J. Chen, V. Kharitonov, Stability of Time-Delay Systems, Birkhäuser, Boston, 2003.
[23] M.J. Park, O.M. Kwon, J.H. Park, S.M. Lee, A new augmented Lyapunov-Krasovskii functional approach for stability of linear systems with time-varying
delays, Appl. Math. Comput. 217 (17) (2011) 7197–7209.
[24] H.B. Zeng, X.G. Liu, W. Wang, A generalized free-matrix-based integral inequality for stability analysis of time-varying delay systems, Appl. Math.
Comput. 354 (2019) 1–8.
[25] O.M. Kwon, M.J. Park, J.H. Park, S.M. Lee, E.J. Cha, Analysis on robust H∞ performance and stability for linear systems with interval time-varying state
delays via some new augmented Lyapunov-Krasovskii functional, Appl. Math. Comput. 224 (2013) 108–122.
[26] J. Sun, G.P. Liu, J. Chen, D. Rees, Improved delay-range-dependent stability criteria for linear systems with time-varying delays, Automatica 46 (2)
(2010) 466–470.
[27] H.B. Zeng, Y. He, M. Wu, J.H. She, New results on stability analysis for systems with discrete distributed delay, Automatica 60 (2015) 189–192.
[28] H. Bao, J.H. Park, J. Cao, Exponential synchronization of coupled stochastic memristor-based neural networks with time-varying probabilistic delay
coupling and impulsive delay, IEEE Trans. Neural Netw. Learn. Syst. 27 (1) (2016) 190–201.
[29] H. Bao, J. Cao, J. Kurths, A. Alsaedi, B. Ahmad, H∞ state estimation of stochastic memristor-based neural networks with time-varying delays, Neural
Netw. 99 (2018) 79–91.
[30] H. Bao, J. Cao, J. Kurths, State estimation of fractional-order delayed memristive neural networks, Nonlinear Dyn. 94 (2018) 1215–1225.
[31] E. Fridman, U. Shaked, Delay-dependent stability and H∞ control: constant and time-varying delays, Int. J. Control 76 (1) (2003) 48–60.
[32] Y. He, M. Wu, J.H. She, G.P. Liu, Parameter-dependent Lyapunov functional for stability of time-delay systems with polytopic-type uncertainties, IEEE
Trans. Autom. Control 49 (5) (2004) 828–832.
[33] X.M. Zhang, M. Wu, J.H. She, Y. He, Delay-dependent stabilization of linear systems with time-varying state and input delays, Automatica 41 (8) (2005)
1405–1412.
[34] A. Seuret, F. Gouaisbaut, Wirtinger-based integral inequality: application to time-delay systems, Automatica 49 (9) (2013) 2860–2866.
[35] H.B. Zeng, Y. He, M. Wu, J.H. She, Free-matrix-based integral inequality for stability analysis of systems with time-varying delay, IEEE Trans. Autom.
Control 60 (10) (2015) 2768–2772.
[36] A. Seuret, F. Gouaisbaut, Hierarchy of LMI conditions for the stability analysis of time-delay systems, Syst. Control Lett. 81 (2015) 1–7.
[37] P.G. Park, J.W. Ko, C. Jeong, Reciprocally convex approach to stability of systems with time-varying delays, Automatica 47 (1) (2011) 235–238.
[38] K. Liu, E. Fridman, Wirtinger’s inequality and Lyapunov-based sampled-data stabilization, Automatica 48 (1) (2012) 102–108.
[39] C.K. Zhang, L. Jiang, Y. He, H. Wu, Stability analysis for control systems with aperiodically sampled data using an augmented Lyapunov functional
method, IET Control Theory Appl. 7 (9) (2013) 1219–1226.
[40] C.K. Zhang, Y. He, L. Jiang, M. Wu, Q.H. Wu, Stability analysis of sampled-data systems considering time delays and its application to electric power
markets, J. Frankl. Inst. 351 (9) (2014) 4457–4478.
[41] H.B. Zeng, Z.L. Zhai, H.Q. Xiao, W. Wang, Stability analysis of sampled-data control systems with constant communication delays, IEEE Access 7 (2019)
111–116.
[42] A. Seuret, E. Fridman, J.P. Richard, Sampled-data exponential stabilization of neutral systems with input and state delays, in: Proceeding of the IEEE
Mediterranean Conference, Cyprus, 2005.