This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON CYBERNETICS 1

Input-to-State Stability for Time-Delay Systems

With Large Delays
Guopin Liu , Changchun Hua , Senior Member, IEEE, Peter Xiaoping Liu , Fellow, IEEE,
and Ju H. Park , Senior Member, IEEE

Abstract—In this article, we consider the input-to-state stability a rough classification, we can classify the stability criteria
(ISS) problem for a class of time-delay systems with intermit- for time-delay systems as either delay-independent or delay-
tent large delays, which may cause the invalidation of traditional dependent ones according to its dependence upon the size of
delay-dependent stability criteria. The topic of this article fea-
tures that it proposes a novel kind of stability criterion for delays. As indicated by its name, the former aims at provid-
time-delay systems, which is delay dependent if the time delay ing stability condition that are suitable for any time-delay
is smaller than a prescribed allowable size. While if the time region, see [10]–[12] and the references therein. Since the
delay is larger than the allowable size, the ISS can be preserved time-delay information is not involved in, the results of delay-
as well provided that the large-delay periods satisfy the kind independent stability criteria are often conservative, especially
of duration condition. Different from existing results on similar
topics, we present the main result based on a unified Lyapunov– when applied to systems with small delays. This deficiency can
Krasovskii function (LKF). In this way, the frequency restriction be addressed by delay-dependent ones; thus, recent research
can be removed and the analysis complexity can be simplified. A works have focused on the later stability issue for time-delay
numerical example is provided to verify the proposed results. systems [13]–[17].
Index Terms—Delay-dependent stability, input-to-state stabil- In the research on delay-dependent stability analysis for
ity (ISS), large delays, time delay. time-delay systems, corresponding criteria are usually obtained
based on the classic Lyapunov stability theory. The research
directions are mainly focused on reducing the conservativeness
that indexed by the permitted delay region, that is, the max-
I. I NTRODUCTION imum delay size in theory, the systems can tolerate without
IME delays exist in numerous kinds of engineering con- losing the stability. In order to achieve the conservativeness-
T trol systems, where examples can be found in networked
teleoperation systems [1], multiagent systems [2], and circuit
reducing purpose, researchers usually devote their efforts to
constructing proper candidate Lyapunov–Krasovskii functions
systems [3], just to name a few. In most of these applica- (LKFs) and estimate corresponding derivate tightly [18], [19].
tions, systems’ stability is frequently degraded by inevitable Up to now, various types of LKFs have been reported. In
time delays, sometimes even end up with instability. Due to regards to the domain of integration, He et al. [5] and
this reason, exploring the influences caused by time delays Li et al. [20] construct LKFs utilizing the entire delay interval,
and finding corresponding stability criteria have attracted con- while Xiao and Zhang [21] and Wu et al. [22] propose
siderable attention over the past decades [4]–[9]. Following LKFs using delay subintervals. Augmented LKFs are proposed
in [23] to obtain the exponential stability criterion for neu-
Manuscript received May 20, 2021; accepted August 13, 2021. This work ral networks with time-varying delay. Based on multiple
was supported in part by the National Key Research and Development
Program of China under Grant 2018YFB1308300; in part by the National integration methods, Yang et al. [24] construct LKFs with
Natural Science Foundation of China under Grant U20A20187 and multiple integral term. In particular, the works in [25]–[27]
Grant 618255304; in part by the Science Fund for Creative Research Groups consider the effect of noise disturbance and provide some
of Hebei Province under Grant F2020203013; in part by the Science and
Technology Development Grant of Hebei Province under Grant 20311803D new ideas and methods. We refer the readers to [19] for
and Grant 19011824Z; and in part by National Defence Fundamental more references on LKFs construction for the delay-dependent
Project under Grant 2020A130. This work of Ju H. Park was supported stability.
by the National Research Foundation of Korea (NRF) Grant funded by the
Korea Government (MSIT) under Grant 2020R1A2B5B02002002. This arti- Note that the aforementioned references propose the stabil-
cle was recommended by Associate Editor Z. Zeng. (Corresponding author: ity criteria based on a common assumption, which requires
Guopin Liu.) the time delays that never exceed the stability interval region.
Guopin Liu and Changchun Hua are with the School of Electrical
Engineering, Yanshan University, Qinhuangdao 066004, China (e-mail: However, in practical control applications, unreliable working
[email protected]; [email protected]). conditions for some time intervals often lead to intermittent
Peter Xiaoping Liu is with the Department of Systems and Computer time-varying delays that are larger than the delay size derived
Engineering, Carleton University, Ottawa, ON K1S 5B6, Canada (e-mail:
[email protected]). from the stability criteria. Examples can be found in networked
Ju H. Park is with the Department of Electrical Engineering, Yeungnam control systems with consecutive packet losses [28], [29], and
University, Gyeongsan 38541, South Korea (e-mail: [email protected]). in control systems with actuator failure [30], [31]. Since
Color versions of one or more figures in this article are available at
https://doi.org/10.1109/TCYB.2021.3106793. the peak of these large delays is generally much larger
Digital Object Identifier 10.1109/TCYB.2021.3106793 than the permitted delay interval, their presence results in
2168-2267 
c 2021 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on September 10,2022 at 06:00:36 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE TRANSACTIONS ON CYBERNETICS

invalidation of almost all these delay-dependent stability crite- switching instant is avoided and the ADT-like assump-
ria. Considering that the large delay only occurs intermittently tion for large-delay frequency is no longer required.
and lasts for a finite period, it is intuitively obvious that the 3) An explicit characterization of the large-delay duration is
stability can be preserved provided that the large-delay dura- provided, under which the ISS of the time-delay system
tion is short enough, despite of its large value. Then, a natural can be preserved. The characterization only depends on
question arises: whether it is possible to provide a explicit the large-delay duration condition, and no restriction is
constrain on the large delays, and then the delay-dependent required on the large-delay frequency.
stability criteria could be extended so as to account for the The remainder of this article is organized as follows.
presence of large delays? Problem formulation and some assumption preliminaries of
Inspired by the literature on switched systems with unsta- the large delay are presented in Section II. Section III dis-
ble subsystems, one can divide the delay into small- and cusses the property of constructed LKF and presents the main
large-delay parts to solve the mentioned question. We can results of this article. In Section IV, a numerical example is
find contributions to the question reported in [32]–[36]. provided to verify the effectiveness of the proposed result.
Sun et al. [32] consider exponential stability for a class of dis- Finally, conclusion is given in Section V.
crete time-varying delay systems with large-delay sequences. Notation: Let R represent the set of reals. Given any α ∈ R,
A switched delay system with one unstable mode is first we denote the set of reals greater than (respectively, no less
introduced to describe the problem caused by large-delay than) α by R>α (respectively, R≥α ). N represents the set of
sequences. Given a certain time interval, by restricting the natural numbers and N0 = N∪{0}. Denote Rn to be a set of all
occurrence number and the duration of large-delay sequences, n-dimensional real column vectors. For any matrix M, M T rep-
explicit delay-dependent conditions are presented to preserve resents its transpose and M denotes the spectral norm of M.
exponential stability. Based on the idea proposed in [32], The smallest and largest eigenvalues of M are represented by
Sun et al. [33] and Yang et al. [36] extend the result to λmin (M) and λmax (M), respectively. Given a measurable time
time-varying linear systems and neural networks, respectively. function g(t) mapping from R≥0 to Rn , denote the L∞ norm
Another relevant work is [35], which introduces the hybrid of g(·) on time interval [0, t) by gt ∞ := sups∈[0,t) g(s).
delay-dependent stability concept to describe the intermittent For any sets A and B, let B\A represent the relative comple-
large delay. Instead of linear matrix inequality techniques, ment of A in B, that is, the set of all elements that belong to
the Lyapunov–Krasovskii functional approach is provided to B, but not to A.
reduce the conservatism.
These listed papers regard the time-delay system as a kind
II. P ROBLEM F ORMULATION
of switched systems, in which field, numerous remarkable
results have been proposed; for instance, in [37], a novel A. System Description and Hybrid Representation
fuzzy control law has been presented for nonlinear stochastic Consider the following time-delay system:
switched systems. From switched system perspective, results
in [32], [35], and [36] construct individual LKF for each sub- ẋ(t) = Ax(t) + Ad x(t − d(t)) + ω(t) (1)
system. While due to that their methods assume the individual
where x ∈ Rn is the state vector, and A and Ad are constant
LKF has a increase at the switching instant, this framework
matrices with compatible dimensions. ω(t) ∈ Rn represents the
has to find a intermediate constant to analyze the succes-
bounded disturbance. d(t) represents the time-varying delay
sive individual Lyapunov functions when switching occurs.
satisfying d(t) ≤ db and |ḋ(t)| ≤ d∗ ≤ 1 for all t ≥ 0, in
Consequently, the switching frequency is seriously restricted,
which d∗ ∈ R≥0 and db ∈ R>0 are constants.
and this is also the reason the average dwell time (ADT)-like
Let there exist a constant h ∈ [0, db ), such that if d(t) ≤
assumption is required in their results. As indicated in [19], a
h, the process dynamics (1) is in a period with small time
more general form of LKF can help to result in a less conserva-
delay. Otherwise, we say that (1) is in a period with large
tive stability criterion. Observing that the dynamics of systems,
time delay. In this article, we consider the case that large-delay
say, system matrices for linear system case, do not evolve in
periods (LDPs) and small-delay periods (SDPs) alternatively
a switching manner, a unified Lyapunov function may be con-
occur in (1). Then, the process dynamics (1) can be rewritten
structed to relax the above-mentioned conservativeness and
as the following system with hybrid time delay:
generalize the existing works.
 
Motivated by the above observation, this article is concerned ẋ(t) = Ax(t) + Ad x t − dσ (t) (t) + ω(t) (2)
with the input-to-state stability (ISS) for time-delay systems
subject to intermittent large delays. The main contributions are where σ (t) : [0, +∞) → P = {1, 2} is the delay switching
summarized as follows. signal, which specifies that the time-varying delay is the small
1) A novel delay-dependent stability criterion is proposed, one or the large one. Without loss of generality, we assume
which can be applied to scenarios with intermittent large that 0 < d1 (t) ≤ h < d2 (t) ≤ db , that is to say, if σ (t) = 1, the
delay in a different way from existing delay-dependent dynamic process is in SDP, while σ (t) = 2 denotes the LDP.
or delay-independent results. The constant h is a permitted delay bound that distinguishes
2) A unified Lyapunov–Krasovskii functional is con- the small and large delays. It can be reviewed as the delay
structed for both small and large-delay modes. Different size for a delay-dependent stability criterion. Note that the
from [32], [35], and [36], the stability analysis at switching property is only used for stability analysis, and it is

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on September 10,2022 at 06:00:36 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

LIU et al.: INPUT-TO-STATE STABILITY FOR TIME-DELAY SYSTEMS WITH LARGE DELAYS 3

unnecessary to know the specific switching instants between The following proposition presents an upper frequency
small and large delays. bound of large delays that satisfies Assumption 1.
The objective of this article is finding a stability condition Proposition 1: Consider the large delays satisfying
under which the ISS can be maintained in the presence of large Assumption 1. Given any time interval [τ, t] and allowable
delays. The following definition will be used for the objective. small time boundary h, it holds that
Definition 1 [38]: Given a delay switching signal σ (t), the κd t−τ
system (2) is said to be input-to-state stable if we can find a N(τ, t) ≤ + (9)
π τd π
KL function β and a K∞ function α and γ such that ∀t ≥ 0,
and the following result holds: where N(τ, t) denote the number of large-delay occur instant
during the interval [τ, t], and π := mini∈N0 {τi }.
α(x(t)) ≤ β(x(0), t) + γ (ωt ∞ ) (3) Proof: According to (4) and (5), one has
for each ω(t) ∈ L∞ (R≥0 ) and initial states x(0) ∈ Rn . N(τ,t)
The following lemma will be utilized in this article. El (τ, t) = τi .
Lemma 1 (Jensen’s Inequality [4]): For any symmetric i=1
matrix R > 0 and a vector function ẋ(s), where s ∈ [a, b],
With π = min{τi }, it holds naturally that N(τ, t)π <
the following result holds: i∈N0
 T    |El (τ, t)|. Then, according to (8), one can obtain (9) imme-
b b b
diately and, thus, complete the proof.
ẋ(s)ds R ẋ(s)ds ≤ (b − a) ẋT (s)Rẋ(s)ds.
a a a Proposition 1 will be used in the stability analysis process.
With that, compared with the existing works [33]–[36], the
B. Time-Constrained LDPs assumption on the frequency of LDPs is removed in our article.
Remark 1 (Large-Delay Duration): This article considers
In the presence of large delay, let {sk }k∈N0 , (s0 ≥ 0), denote
the large delay as the source of instability. Assumption 1
the happening sequence of large delay. That is to say, a large
describes the large-delay duration property that the total dura-
delay will occur at the instant t = sk , k ∈ N0 . Define sequence
tion of large delay is less than a fraction of time interval, which
{tk }k∈N0 , where tk = sk +h represents the instant at which time
is specified by 1/τd . The constant κd is used for regulariza-
delay exceeds the allowable size. Then, we can represent the
tion. The constraint on large delay duration is inspired by [39],
ith period where the time delay exceeds allowable size as
which concerns the stability of networked control systems
Ti := {ti } ∪ [ti , ti + τi ) (4) in the presence of cyber attacks. Under Assumption 1, (9)
holds immediately, which expresses the frequency property
where τi ∈ R≥0 is the interval length of period Ti , over which for LDPs.
time delay exceeds the allowable size. Remark 2 (Unified LKF): In our article, the time interval
Given any time interval [τ, t], where 0 ≤ τ ≤ t, let is classified into disjoint union of SDPs and LDPs. Different
 
El (τ, t) := Ti [τ, t] (5) from existing methods [32]–[36], which present the stability
i∈N
criteria based on the multiple piecewise Lyapunov–Krasovskii
  functional for switched systems, we find the system matrices
Es (τ, t) := Si [τ, t] (6) do no switch at all and, thus, provide our result based on
i∈N a unified Lyapunov–Krasovskii functional. By this way, we
denote the sets of LDPs and SDPs, respectively, where can avoid considering the Lyapunov functional behavior at
the switching instant. Hence, the frequency-like assumption
Si := {ti + τi } ∪ ti + τi , ti+1 ). (7)
is no longer required and the analysis framework is simplified
According to (4)–(7), it holds that Es (τ, t) = [τ, t]\El (τ, t). significantly, which can be observed in the following section.
Since the existence of large delays is the source of insta- Remark 3 (LDPs Formulation): In fact, it is more reasonable
bility, in order to guarantee the system stability, it is obvious to denote the beginning of LDP at the instant t = sk + h, k ∈
that one needs to constrain the length of the large time-delay N0 , that is, the instant when delay exceeds allowable size.
period. Let |E(τ, t)| represent the length of LDP over the This is intuitively clear that although a large delay occurs at
interval [τ, t]. Then, the following assumption is established t = sk , the delay-dependent stability result for small delay
to describe the influence caused by the large delay. still holds until the delay exceeds allowable size. Therefore,
Assumption 1 (Duration): There exist constant κd ∈ R≥0 we establish a new LDP by (4). Compared with [32]–[36], our
and τd ∈ R>1 such that characterization for LDP is more accurate and can relax the
t−τ duration constraint for large delay. In the rest of this article,
|El (τ, t)| ≤ κd + . (8) we adopt the notation of LDP by (4).
τd
In fact, Assumption 1 is similar with the ADT condition
III. M AIN R ESULTS
for switched systems. Hence, τd in (8) can be regarded as the
“dwell time” for LDPs. Given any small time delay boundary A. LKF Construction and Analysis
h, with (8), it is easy to obtain that the frequency at which In this part, we aim at proposing a unified LKF to analyze
large delay can occur is finite. the system stability under both LDPs and SDPs.

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on September 10,2022 at 06:00:36 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE TRANSACTIONS ON CYBERNETICS

 
Define V0 (x) = xT Px(t), where P is a positive matrix to be
t P
− c1 h |ẋ(s)|2 ds + + 3c1 h2 ωt 2∞
specified, and then consider the following LKF: t−h δ2
(16)
V(x(t)) = V1 (x(t)) + V2 (x(t)) (10)
where Q1 > 0 is a matrix satisfying
where
 t  t PĀ + ĀT P + (δ1 + δ2 )P + 3c1 h2 ĀT Ā + Q1 = 0. (17)
V1 (x(t)) = V0 (x(t)) + c1 h |ẋ(m)| dmds 2
t−h s Define the parameters W1 and γ1 as
 t
c2 ATd PAd
V2 (x(t)) = V0 (x(s))ds (11) W1 = c1 h − 3c1 h3 ATd Ad −
1 − d∗ t−d2 (t) δ1
h
 t  t
c3 γ1 = P/δ2 + 3c1 h2
+ V0 (x(m))dmds. (12)
1 + db t−d2 (t) s
then (16) can be rewritten as
In the rest of this section, based on (10), the analysis roughly  t
consists of two cases. V̇1 (x(t)) ≤ −x (t)Q1 x(t) − W1
T
|ẋ(s)|2 ds
First, consider the case that t ∈ Es . t−h
Note that + γ1 ωt 2∞ . (18)
 t
ẋ(t) = Āx(t) − Ad ẋ(s)ds + ω(t) (13) Note that
t−d1 (t)  t  t  t  t
|ẋ(m)|2 dmds ≤ |ẋ(m)|2 dmds
where Ā = A + Ad , then we have t−h s
 t t−h
t
t−h
 t
2
|ẋ(t)|2 = Āx(t) − Ad ẋ(s)ds + ω(t) ≤ ds |ẋ(m)|2 dm
t−h t−h
t−d1 (t)  t
 2
t
=h |ẋ(m)|2 dm
≤ 3xT (t)ĀT Āx(t) + 3 ATd Ad ẋ(s)ds
t−d1 (t) t−h
t
+ 3ω(t)2 . (14) =h |ẋ(s)|2 ds (19)
t−h
The time derivative of V1 (x(t)) can be obtained as we have
   t  t
V̇1 (x(t)) = xT (t) PĀ + ĀT P x(t) + 2x(t)T Pω(t) V̇1 (x(t)) ≤ −x(t) Q1 x(t) −TW1
|ẋ(m)|2 dmds
 t h t−h s
− 2x(t) PAd
T
ẋ(s)ds
t−d1 (t) + γ1 ωt ∞ .
2
(20)
 t
+ c1 h (|ẋ(t)|2 − |ẋ(s)|2 )ds According to (12), the time derivative of V2 (x(t)) can be
t−h obtained such that
≤ xT (t)(PĀ + ĀT P + (δ1 + δ2 )P)x(t) V̇2 (t) ≤
c2
V0 (x(t)) − c2 V0 (x(t − d2 (t)))
 t 2 1 − d∗
ATd PAd P  t
+ ẋ(s)ds + ωt 2∞ c3 db c3
δ1 t−d1 (t) δ 2 + V0 (x(t)) − V0 (x(s))ds (21)
 t 1 + db 1 + db t−d2 (t)
+ c1 h |ẋ(t)| − c1 h
2 2
|ẋ(s)| ds.
2
(15) which obviously indicates
t−h

Substituting (14) into (15), we have c2 c3 db
V̇2 (t) ≤ + V0 (x(t))
1 − d∗ 1 + db
V̇1 (x(t)) ≤ xT (t)(PĀ + ĀT P + (δ1 + δ2 )P)x(t)  t
 t c3
ATd PAd 2 − V0 (x(s))ds. (22)
+ ẋ(s)ds 1 + db t−d2 (t)
δ1 t−d1 (t)
 t Similar to inequality (19), it holds that
P  t  t  t
+ ωt 2∞ − c1 h |ẋ(s)|2 ds
δ2 t−h V0 (x(m))dmds ≤ db V0 (x(s))ds
+ 3c1 h2 xT (t)ĀT Āx(t) + 3c1 h2 ω(t)2 t−d2 (t) s t−d2 (t)
 t 2 then we have
+ 3c1 h2 ATd Ad ẋ(s)ds  t
t−d1 (t)
c2 c3 db
V2 ≤ ∗
+ V0 (x(s))ds (23)
1−d 1 + db t−d2 (t)
which, based on lemma 1 and d1 (t) ≤ h, can be rewritten as
 t which together with (22) yield
AT PAd 
V̇1 (x(t)) ≤ −xT (t)Q1 x(t) + d h |ẋ(s)|2 ds c2 c3 db
δ1 t−h V̇2 (x(t)) ≤ + V0 (x) − W2 V2 (24)
 t 1 − d∗ 1 + db
+ 3c1 h h Ad Ad
2 T
|ẋ(s)|2 ds
t−h where W2 = c3 (1 − d∗ )/(c2 + c2 db + c3 db − c3 db d∗ ).

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on September 10,2022 at 06:00:36 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

LIU et al.: INPUT-TO-STATE STABILITY FOR TIME-DELAY SYSTEMS WITH LARGE DELAYS 5

Using (10)–(12), (20), and (24), we obtain Remark 4: According to above discussion, we can claim
  t that the system is input-to-state stable in the SDPs by (27),
W1 t
V̇(x(t)) ≤ −x (t)Q2 x(t) −
T
|ẋ(m)|2 dmds while, based on the designed LKF, we cannot conclude the
h t−h s system stability in the LDPs by (35). Maybe another suitable
− W2 V2 + γ1 ωt 2∞ (25) LKF can be designed under which the delay-dependent sta-
bility criterion can be obtained for a prescribed large-delay
where Q2 is a matrix satisfying
 size. However, if the delay size exceeds that prescribed size,
c2 c3 the proposed suitable LKF may fail again. Hence, instead of
−Q1 + + db P + Q2 ≤ 0. (26)
1 − d∗ 1 + db proving the stability for both LDPs and SDPs with a delicate
LKF, the interest for this article is providing a condition on
Let ω1 = min{[λmin (Q2 )]/[λmax (P)], [W1 ]/[c1 h2 ], W2 }. the LDPs, under which the system is input-to-state stable. In
Then, we can obtain the next part, it will present an explicit characterization of the
V̇(x(t)) ≤ −ω1 V(x(t)) + γ1 ωt 2∞ . (27) large-delay duration to preserve the ISS purpose.

Next, consider the case that t ∈ El . With σ (t) = 2, the B. Input-to-State Stability With Large Delays
derivative of V1 (x(t)) along (2) satisfies
Now, we are in the position to present the ISS result of our
V̇1 (x(t)) ≤ xT (t)(PA + AT P)x(t) + 2xT (t)Pω(t) article.
+ 2xT (t)PAd x(t − d2 (t)) + c1 h2 |ẋ(t)|2 Theorem 1: Consider the time-delay system (1) with alter-
 t natively small and LDPs, which can be reformulated by (2).
− c1 h |ẋ(m)|2 dm Given any positive symmetric definite matrix P and Qi , i =
t−h
1, 2, 3, 4 satisfying (17), (26), (31), and (34), select the
≤ xT (t)(PA + AT P + (δ3 + δ4 )P)x(t) Lyapunov functions V(x(t)) as in (10)–(12). For any LDP
ATd PAd 2 P sequence satisfying Assumption 1 with arbitrary κd , and with
+ x (t − d2 (t)) + ωt 2
δ3 δ4 τd such that
+ c1 h2 |ẋ(t)|2 . (28) 1 ω1
≤ (36)
Observing that τd ω1 + ω2
where ω1 and ω2 are as defined in Section III-A, the system (1)
|ẋ(t)|2 = |Ax(t) + Ad x(t − d2 (t)) + ω(t)|2 [or equally system (2)] is input-to-state stable.
≤ 3xT (t)AT Ax(t) + 3ATd Ad x2 (t − d2 (t)) Proof: With (4)–(7), for any 0 ≤ τ ≤ t, the interval [τ, t] is
+ 3ω2 (t) (29) the disjoint union of El (τ, t) and Es (τ, t). Given any i ∈ N0 ,
according to (27) and (35), we have
the following result holds by substituting (29) into (28): γ1
V(x(t)) ≤ e−ω1 (t−ti −τi ) V(x(ti + τi )) + ωt 2∞ (37)
V̇1 ≤ xT (t)Q3 x(t) + γ2 x2 (t − d2 (t)) + γ3 ωt 2 (30) ω1
for all t ∈ Si , and
where Q3 is a matrix of the following form:
γ2 ω2 (t−ti )
V(x(t)) ≤ eω2 (t−ti ) V(x(ti )) + e ωt 2∞ (38)
Q3 = PA + A P + (δ3 + δ4 )P + 3c1 h A A
T 2 T
(31) ω2
and γ2 = (ATd PAd /δ3 ) + 3c1 h2 Ad 2 and γ3 = (P/δ4 ) + for all t ∈ Ti , respectively.
3c1 h2 . Let γ := max{(γ1 /ω1 ), (γ2 /ω2 )}, k := sup{i ∈ N0 |ti ≤ t}.
The time derivative of V2 (x(t)) satisfies (21) as in the case Without loss of generality, we suppose t ∈ Es . Then, according
t ∈ Es . Let the parameters c2 and γ2 satisfy to (37) and (38), we can obtain the following estimate:

c2 P ≥ γ2 I V(x(t)) ≤ e−ω1 (t−tk −τk ) V(x(tk + τk )) + γ ωt 2∞

(32)  
≤ e−ω1 (t−tk −τk ) eω2 τk V(x(tk )) + eω2 τk γ ωt 2∞
then we have
 t + γ ωt 2∞
c3
V̇(x(t)) ≤ x (t)Q4 x(t) −
T
V0 (x(s))ds ..
1 + db t−d2 .
+ γ3 ωt 2∞ (33) ≤ ψ1 (t)V(x(0)) + (1 + 2ψ2 (t))γ ωt 2∞ (39)
where where

c2 c3 db
Q4 = Q3 + + P. (34) ψ1 (t) = e−ω1 Es (0,t)+ω2 El (0,t)
1 − d∗ 1 + db k  
It follows immediately that: ψ2 (t) = e−ω1 Es (ti +τi ,t)+ω2 El (ti ,t) .
i=0
V̇(x(t)) ≤ ω2 V(x(t)) + γ3 ωt 2∞ (35)
To prove the ISS of system (2), we first need to show that
where ω2 = λmax (Q4 ). the sum term of ψ2 (t) is bounded by a constant.

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on September 10,2022 at 06:00:36 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE TRANSACTIONS ON CYBERNETICS

According to Assumption 1, we have hence, from (49), we conclude that there exist a KL function
t − τi β(x(0), t) and K∞ function γ (ωt 2∞ ) such that
El (ti , t) ≤ κd + (40)
τd α(x(t)) ≤ β(x(0), t) + γ (ωt 2∞ ) (51)
for all t ∈ R≥ti , i ∈ N0 .
where
Since Es (ti , t) = [ti , t]\El (ti , t), we have
∗
Es (ti + τi , t) = t − ti − τi − El (ti + τi , t) β(x(0), t) = e(ω1 +ω2 )κd e−α t α(0)
 
  ∗
eα κd τd
= t − ti − El (ti , t). (41) γ ωt ∞ = 1 + 2e
2 κd (ω1 +ω2 )
∗ γ ωt 2∞ .
1 − e−α τd π
Hence, ψ2 (t) satisfies
k
Therefore, the ISS can be concluded according to
−ω1 Es (ti +τi ,t)+ω2 El (ti ,t) Definition 1. This completes the proof.
ψ2 (t) = e
Remark 5 (Tradeoff Between LDP and ISS): The proof
i=0
k     of Theorem is inspired by the stability analysis of switched
t−t t−t
−ω1 t−ti −κd − τ i +ω2 κd + τ i systems with unstable mode. Condition (36) specifies the
≤ e d d

i=0
tradeoff between stability and intensity of large delay. For
k instance, if τd is decreased, the duration ratio of LDPs is then
∗ (t−t )
≤ e(ω1 +ω2 )κd e−α i (42) increased, which means the intensity of LDPs is strong and
i=0
condition (36) may be destroyed. The consequence is clear that
strong intensity of LDPs can degrade the stability performance.
where α ∗ = ω1 − (ω1 + ω2 /τd ).
In order to preserve system stability, it requires to increase ω1
According to Proposition 1, one has
such that (36) can be satisfied. This adjustment is intuitively
t − ti ≥ τd π N(ti , t) − κd τd (43) clear that when the large delay duration increases, it requires
the system states converging rapidly such that the divergence
which indicates that caused by the large delay can be compensated.
k k
∗ (t−t ) ∗κ ∗ τ π N(t ,t) Remark 6 (LDPs Frequency Free): Due to the benefit of the
e−α i ≤ eα d τd e−α d i . (44) unified LKF approach, we do not have to find a intermediate
i=0 i=0 constant to analyze the relationship between multiple LKFs
Note that at the switching instant. Thus, the large-delay frequency is
no longer required in the condition (36), and the analysis
N(ti , t) ≥ k − i (45)
complexity is simplified as well.
then it holds that Remark 7: By solving (26), we can obtain a candidate h and
k k P. It remains to check if proper constants can be selected such
∗ τ π N(t ,t) ∗ τ π(k−i)
e−α d i ≤ e−α d that (32) is satisfied and parameter W1 is positive. To make
i=0 i=0 the procedure simple, in the simulation part of this article, we
k first solve (26) with prescribed h and then verify (32) and W1 .
∗τ π i 1
= e−α d ≤ ∗ . (46) Remark 8: This article aims at analyzing the stability of
1 − e−α τd π systems with large delays via a single unified LKF, and we
i=0
From (42), (44), and (46), we have only utilize very common approaches for time-delay systems.
∗ In the research field on time-delay systems, there are quite
eα κd τd
ψ2 (t) ≤ eκd (ω1 +ω2 ) ∗ (47) many advanced techniques to reduce the conservatism of
1 − e−α τd π stability criteria. Hence, the stability result may be further
which indicates ψ2 (t) is bounded by positive constant ψ̄2 := strengthened if we adopt more advanced analysis approaches,
∗ ∗
eκd (ω1 +ω2 ) [(eα κd τd )/(1 − e−α τd π )]. such as the ones proposed in [40]. This could be an interesting
Similar to the procedure of deriving (42), we know the direction for future work.
function ψ1 (t) in (39) satisfies
∗ IV. N UMERICAL E XAMPLE
ψ1 (t) ≤ e(ω1 +ω2 )κd e−α t . (48)
A numerical example is given in this section to verify the
From (39), (47), and (48), we can obtain that V(x(t)) proposed result of this article. Consider the following system
satisfies with hybrid time delay:
∗  
V(x(t)) ≤ e(ω1 +ω2 )κd e−α t V(x(0)) ẋ(t) = Ax(t) + Ad x(t − dσ (t) (t)) + ω(t)
 ∗

κd (ω1 +ω2 ) eα κd τd
+ 1 + 2e ∗ γ ωt 2∞ . (49) where
1 − e−α τd π    
−2 0 −1 0
A= , Ad = .
Based on the selection of V(x(t)) given by (10)–(12), we 0 −0.9 −1 −1
can find class K∞ functions α(x(t)) and α(x(t)) such that
The delay switching signal σ (t) → {1, 2}. The time delays
α(x(t)) ≤ V(x(t)) ≤ α(x(t)) (50) for each mode are d1 (t) = 0.03sin(t) and d2 (t) = 4.5 +

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on September 10,2022 at 06:00:36 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

LIU et al.: INPUT-TO-STATE STABILITY FOR TIME-DELAY SYSTEMS WITH LARGE DELAYS 7

Fig. 2. Trajectories of states x(t) under time-constrained large delay.

Fig. 1. Trajectories of delay switching signal σ (t).

TABLE I
A LLOWABLE U PPER B OUND h W ITH VARIOUS h1

0.5sin(t), respectively. The simulation parameters are set as

c1 = 5, c2 = 0.75, c3 = 1.08, d∗ = 0.5, db = 5, δ1 =
0.5, δ2 = 0.1, δ3 = 4.2, and δ4 = 0.3, and the disturbances
are ω(t) = 0.8 exp(−0.2t).
In order to make the inequality (26), (32) feasible, and
Fig. 3. Trajectories of states x(t) under large delay.
to guarantee the constant W1 being positive, the small-delay
region is set as h ≤ 0.033. Then, possible matrices P, Q2 , and
Q4 are obtained as follows:
 
0.8338 −0.3072
P=
−0.3072 0.8049
 
0.4169 −0.1536
Q2 =
−0.1536 0.4024
 
2.4832 −1.2287
Q4 = .
−1.2287 4.1181
Then, through basic computation, we can obtain W1 =
0.0696 and W2 = 0.0750, and according to the definition of
ω1 and ω2 , we have ω1 = min{0.2271, 12.7802, 0.0750} =
0.0750 and ω2 = λmax (Q4 ) = 4.7764, which indicates Fig. 4. Trajectories of states x(t) under small delay.
1/τd ≤ [ω1 /(ω1 + ω2 )] = 0.0155. By recalling Assumption 1,
we know the duration of LDPs should less than 1.55% of all
the running time. One eligible delay switching signal that are In fact, the allowable upper bound of delay can be sig-
generated randomly is illustrated in Fig. 1. nificantly improved because we regard the LDPs as unstable
This time-delay system example has been widely investi- modes for the time-delay system. By restricting the working
gated by existing results. Jiang and Han [41], Shao [42], and dwell time for LDPs, the problem is transformed into the sta-
Park et al. [43] derived the stability criteria for system (1) with bility analysis for switched systems with unstable subsystems.
ω(t) = 0 and h1 < d(t) < h2 . The derived allowable upper Therefore, it is meaningless to compare the allowable upper
bounds for time delay d(t) in [41]–[43] are provided in Table I, bound of delay with any other results in this framework.
from which we can find that the delay sizes permitted in all In our article, the ISS can be guaranteed by restricting the
of these mentioned results are less than 2 when h1 = 0. While duration property of LDPs. Without these kinds of restriction,
in our article with h1 = 0, via the conditions restricting the the stability can no longer be preserved, which is shown as in
duration of LDPs, the input to stability can be preserved in Fig. 3.
the presence of time delays with the size that is equal to 5. In If there are no large delays, then the result degrades to the
the presence of large delay, the state trajectories are illustrated delay-dependent one for the small delay. By letting σ (t) = 1
in Fig. 2, which clearly verified the stability of the time-delay for all t ≥ 0, this condition is verified by Fig. 4, in which the
system. states trajectories are convergent.

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on September 10,2022 at 06:00:36 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

8 IEEE TRANSACTIONS ON CYBERNETICS

V. C ONCLUSION [19] C. Zhang, Y. He, L. Jiang, and M. Wu, “Stability analysis for delayed
neural networks considering both conservativeness and complexity,”
This article has analyzed the ISS for a class of time-delay IEEE Trans. Neural Netw. Learn. Syst., vol. 27, no. 7, pp. 1486–1500,
systems subject to intermittent large delays. Different from Jul. 2016.
existing results, this article derives a novel stability criterion [20] T. Li, L. Guo, C. Sun, and C. Lin, “Further results on delay-dependent
stability criteria of neural networks with time-varying delays,” IEEE
for time-delay systems, which is delay dependent if the time Trans. Neural Netw., vol. 19, no. 4, pp. 726–730, Apr. 2008.
delay is within a given interval. For the case of large-delay [21] S.-P. Xiao and X.-M. Zhang, “New globally asymptotic stability criteria
size, the ISS can also be preserved if the large-delay duration for delayed cellular neural networks,” IEEE Trans. Circuits Syst. II, Exp.
Briefs, vol. 56, no. 8, pp. 659–663, Aug. 2009.
satisfies certain conditions. Besides, the stability analysis is [22] Z.-G. Wu, J. Lam, H. Su, and J. Chu, “Stability and dissipativity analysis
finalized based on a unified Lyapunov–Krasovskii functional, of static neural networks with time delay,” IEEE Trans. Neural Netw.
which relaxes the frequency restriction of large delays and Learn. Syst., vol. 23, no. 2, pp. 199–210, Feb. 2012.
[23] C. Hua, X. Yang, J. Yan, and X. Guan, “New exponential stability criteria
simplifies the analysis complexity. A numerical example is for neural networks with time-varying delay,” IEEE Trans. Circuits Syst.
provided to verify the proposed results. II, Exp. Briefs, vol. 58, no. 12, pp. 931–935, Dec. 2011.
[24] Q. Yang, Q. Ren, and X. Xie, “New delay dependent stability criteria for
recurrent neural networks with interval time-varying delay,” ISA Trans.,
vol. 53, no. 4, pp. 994–999, 2014.
R EFERENCES [25] Q. Zhu and T. Huang, “Stability analysis for a class of stochastic delay
nonlinear systems driven by G-Brownian motion,” Syst. Control Lett.,
[1] D. Sun, F. Naghdy, and H. Du, “Neural network-based passivity con- vol. 140, Jun. 2020, Art. no. p104699.
trol of teleoperation system under time-varying delays,” IEEE Trans. [26] Q. Zhu, “Stabilization of stochastic nonlinear delay systems with exoge-
Cybern., vol. 47, no. 7, pp. 1666–1680, Jul. 2017. nous disturbances and the event-triggered feedback control,” IEEE Trans.
[2] X. Wu, Y. Tang, J. Cao, and W. Zhang, “Distributed consensus of Autom. Control, vol. 64, no. 9, pp. 3764–3771, Sep. 2019.
stochastic delayed multi-agent systems under asynchronous switching,” [27] W. Hu, Q. Zhu, and H. Karimi, “Some improved Razumikhin stability
IEEE Trans. Cybern., vol. 46, no. 8, pp. 1817–1827, Aug. 2016. criteria for impulsive stochastic delay differential systems,” IEEE Trans.
[3] F. Chen, X. Zhang, and J. Peng, “Time-delayed chaotic circuit design Autom. Control, vol. 64, no. 12, pp. 5207–5213, Dec. 2019.
using all-pass filter,” IEEE Trans. Circuits Syst. I, Regular Papers, [28] X. Sun, G. Liu, W. Wang, and D. Rees, “Stability analysis for networked
vol. 61, no. 10, pp. 2897–2903, Oct. 2014. control systems based on average dwell time,” Int. J. Robust Nonlinear
[4] K. Gu, V. Kharitonov, and J. Chen, Stability of Time-Delay Systems. Control, vol. 20, no. 15, pp. 1774–1784, 2010.
Boston, MA, USA: Birkhäuser, 2003. [29] X. Huang and Y. Tian, “Asynchronous distributed localization in
[5] Y. He, G. Liu, and D. Rees, “New delay-dependent stability criteria for networks with communication delays and packet losses,” Automatica,
neutral networks with time-varying delay,” IEEE Trans. Neural Netw., vol. 96, pp. 134–140, Oct. 2018.
vol. 18, no. 1, pp. 310–314, Jan. 2007. [30] X. Sun, G. Liu, D. Rees, and W. Wang, “Stability of systems with
[6] Y. Liu, S. M. Lee, O. M. Kwon, and J. H. Park, “New approach to sta- controller failure and time-varying delay,” IEEE Trans. Autom. Control,
bility criteria for generalized neural networks with interval time-varying vol. 53, no. 10, pp. 2391–2396, Nov. 2008.
delays,” Neurocomputing, vol. 149, pp. 1544–1551, Feb. 2015. [31] J. Wang, Z. Liu, C. Chen, and Y. Zhang, “Fuzzy adaptive compensation
[7] J. Chen, S. Meng, and J. Sun, “Stability analysis of networked control control of uncertain stochastic nonlinear systems with actuator failures
systems with aperiodic sampling and time-varying delay,” IEEE Trans. and input hysteresis,” IEEE Trans. Cybern., vol. 49, no. 1, pp. 2–13,
Cybern., vol. 47, no. 8, pp. 2312–2320, Aug. 2017. Jan. 2019.
[8] J. Chen, C. Lin, B. Chen, and Q. Wang, “Regularization and stabilization [32] X. Sun, G. Liu, D. Rees, and W. Wang, “Delay-dependent stability
for rectangular T–S fuzzy discrete-time systems with time delay,” IEEE for discrete systems with large delay sequence based on switching
Trans. Syst., Man, Cybern., Syst, vol. 49, no. 4, pp. 833–842, Apr. 2019. techniques,” Automatica, vol. 44, no. 11, pp. 2902–2908, 2008.
[9] R. Baranitha, R. Mohajerpoor, and R. Rakkiyappan, “Bilateral teleop- [33] X. Sun, G. Liu, W. Wang, and D. Rees, “Stability analysis for systems
eration of single-master multislave systems with semi-Markovian jump with large delay period: A switching method,” Int. J. Innov. Comput.
stochastic interval time-varying delayed communication channels,” IEEE Inf. Control, vol. 8, no. 6, pp. 4235–4247, 2012.
Trans. Cybern., vol. 51, no. 1, pp. 247–257, Jan. 2021. [34] X.-M. Sun, S.-L. Du, P. Shi, W. Wang, and L.-D. Wang, “Input-to-
[10] A. A. Zevin and M. A. Pinsky, “Delay-independent stability condi- state stability for nonlinear systems with large delay periods based on
tions for time-varying nonlinear uncertain systems,” IEEE Trans. Autom. switching techniques,” IEEE Trans. Circuits Syst. I, Regular Papers,
Control, vol. 51, no. 9, pp. 1482–1486, Sep. 2006. vol. 61, no. 6, pp. 1789–1800, Jun. 2014.
[35] X.-M. Sun, X.-F. Wang, and F. Mazenc, “Delay-hybrid-dependent sta-
[11] W. Chen and W. Zheng, “Delay-independent minimum dwell time for
bility for systems with large delays,” IEEE Trans. Syst., Man, Cybern.,
exponential stability of uncertain switched delay systems,” IEEE Trans.
Syst., vol. 50, no. 7, pp. 2752–2759, Jul. 2020.
Autom. Control, vol. 55, no. 10, pp. 2406–2414, Oct. 2010.
[36] B. Yang, M. Hao, M. Han, X. Zhao, and G. Zong, “Exponential
[12] V. S. Bokharaie and O. Mason, “On delay-independent stability of a class
stability of discrete-time neural networks with large delay,”
of nonlinear positive time-delay systems,” IEEE Trans. Autom. Control,
IEEE Trans. Cybern., vol. 51, no. 5, pp. 2824–2834, May 2021,
vol. 59, no. 7, pp. 1974–1977, Jul. 2014.
doi: 10.1109/TCYB.2019.2923244.
[13] Q. Ma, G. Feng, and S. Xu, “Delay-dependent stability criteria for [37] J. Cheng, W. Huang, H. Lam, J. Cao, and Y. Zhang, “Fuzzy-model-based
reaction–diffusion neural networks with time-varying delays,” IEEE control for singularly perturbed systems with nonhomogeneous Markov
Trans. Cybern., vol. 43, no. 6, pp. 1913–1920, Dec. 2013. switching: A dropout compensation strategy,” IEEE Trans. Fuzzy Syst.,
[14] C. Tan, H. Zhang, and W. S. Wong, “Delay-dependent algebraic riccati early access, Dec. 1, 2020, doi: 10.1109/TFUZZ.2020.3041588.
equation to stabilization of networked control systems: Continuous-time [38] L. Vu, D. Chatterjee, and D. Liberzon, “Input-to-state stability of
case,” IEEE Trans. Cybern., vol. 48, no. 10, pp. 2783–2794, Oct. 2018. switched systems and switching adaptive control,” Automatica, vol. 43,
[15] X. Li and X. Yang, “Lyapunov stability analysis for nonlinear no. 4, pp. 639–646, 2007.
systems with state-dependent state delay,” Automatica, vol. 112, pp. 1–6, [39] C. D. Persis and P. Tesi, “Input-to-state stabilizing control under denial-
Feb. 2020. of-service,” IEEE Trans. Autom. Control, vol. 60, no. 11, pp. 2930–2944,
[16] P. Pepe, “Discrete-time systems with constrained time delays and delay- 2015.
dependent Lyapunov functions,” IEEE Trans. Autom. Control, vol. 65, [40] R. Mohajerpoor, L. Shanmugam, H. Abdi, R. Rakkiyappan,
no. 4, pp. 1724–1730, Apr. 2020. S. Nahavandi, and P. Shi, “New delay range–dependent stability
[17] X. Liu and K. Zhang, “Input-to-state stability of time-delay systems with criteria for interval time-varying delay systems via Wirtinger-based
delay-dependent impulses,” IEEE Trans. Autom. Control, vol. 65, no. 4, inequalities,” Int. J. Robust Nonlinear Control, vol. 28, no. 2,
pp. 1676–1682, Apr. 2020. pp. 661–677, 2018.
[18] Z. Wu, H. Su, J. Chu, and W. Zhou, “Improved delay-dependent [41] X. Jiang and Q.-L. Han, “On H∞ control for linear systems with interval
stability condition of discrete recurrent neural networks with time- time-varying delay,” Automatica, vol. 41, no. 12, pp. 2099–2106, 2005.
varying delays,” IEEE Trans. Neural Netw., vol. 21, no. 4, pp. 692–697, [42] H. Shao, “New delay-dependent stability criteria for systems with
Apr. 2010. interval delay,” Automatica, vol. 45, no. 3, pp. 744–749, 2009.

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on September 10,2022 at 06:00:36 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

LIU et al.: INPUT-TO-STATE STABILITY FOR TIME-DELAY SYSTEMS WITH LARGE DELAYS 9

[43] P. Park, J. W. Ko, and C. Jeong, “Reciprocally convex approach to sta- Peter Xiaoping Liu (Fellow, IEEE) received the
bility of systems with time-varying delays,” Automatica, vol. 47, no. 1, B.Sc. and M.Sc. degrees from Northern Jiaotong
pp. 235–238, 2011. University, Beijing, China, in 1992 and 1995, respec-
tively, and the Ph.D. degree from the University of
Alberta, Edmonton, AB, Canada, in 2002.
He has been with the Department of Systems
and Computer Engineering, Carleton University,
Ottawa, ON, Canada, since 2002, where he is cur-
rently a Professor. He is also with the Faculty of
Mechanical Engineering and Automation, Zhejiang
Sci-Tech University, Hangzhou, China, as an
Adjunct Professor. His research interests include interactive networked
Guopin Liu received the B.Sc. and Ph.D. degrees systems and teleoperation, haptics, surgical simulation, robotics, control, and
from Yanshan University, Qinhuangdao, China, in intelligent systems.
June 2014 and January 2020, respectively. Prof. Liu has served as an Associate Editor for several jour-
From 2018 to 2019, he was a Visiting nals, including the IEEE/CAA J OURNAL AUTOMATICA S INICA, the
Researcher with the Faculty of Science and IEEE T RANSACTIONS ON C YBERNETICS, the IEEE T RANSACTIONS ON
Engineering, University of Groningen, Groningen, AUTOMATION S CIENCE AND E NGINEERING, the IEEE T RANSACTIONS
The Netherlands. From 2020 to 2021, he worked ON I NSTRUMENTATION AND M EASUREMENT , and the IEEE/ASME
as a Postdoctoral Fellow with Yeungnam University, T RANSACTIONS ON M ECHATRONICS. He is a Licensed Member of the
Gyeongsan, South Korea. His research interests Professional Engineers of Ontario and a fellow of the Engineering Institute
include switched systems and networked control of Canada.
systems.

Ju H. Park (Senior Member, IEEE) received the

Ph.D. degree in electronics and electrical engineer-
ing from POSTECH, Pohang, Republic of Korea, in
1997.
From May 1997 to February 2000, he was a
Research Associate with ERC-ARC, POSTECH. He
Changchun Hua (Senior Member, IEEE) received joined Yeungnam University, Gyeongsan, Republic
the Ph.D. degree in electrical engineering from of Korea, in March 2000, where he is currently
Yanshan University, Qinhuangdao, China, in 2005. the Chuma Chair Professor. From 2006 to 2007,
He was a Research Fellow with the National he was a Visiting Professor with the Department
University of Singapore, Singapore, from 2006 to of Mechanical Engineering, Georgia Institute of
2007. From 2007 to 2009, he was with Carleton Technology. His research interests include robust control and filtering,
University, Ottawa, ON, Canada, funded by the neural/complex networks, fuzzy systems, multiagent systems, and chaotic
Province of Ontario Ministry of Research and systems. He has published a number of papers in these areas.
Innovation Program. From 2009 to 2010, he was Prof. Park is a recipient of Highly Cited Researcher Award by Clarivate
with the University of Duisburg–Essen, Essen, Analytics (formerly, Thomson Reuters) in 2015, and listed in three fields,
Germany, funded by Alexander von Humboldt engineering, computer sciences, and mathematics in 2019. He serves as an
Foundation. He is currently a Full Professor with Yanshan University. He Editor for the International Journal of Control, Automation, and Systems.
has authored or coauthored over 80 papers in mathematical, technical jour- He is also a Subject Editor/Advisory Editor/Associate Editor/Editorial Board
nals, and conferences. He has been involved in over ten projects supported Member for several international journals, including IET Control Theory and
by the National Natural Science Foundation of China, the National Education Applications, Applied Mathematics and Computation, Journal of The Franklin
Committee Foundation of China, and other important foundations. His current Institute, Nonlinear Dynamics, Engineering Reports, Cogent Engineering, and
research interests include nonlinear control systems, control systems design IEEE T RANSACTIONS ON F UZZY S YSTEMS. He is a fellow of the Korean
over network, teleoperation systems, and intelligent control. Academy of Science and Technology.

Authorized licensed use limited to: Motilal Nehru National Institute of Technology. Downloaded on September 10,2022 at 06:00:36 UTC from IEEE Xplore. Restrictions apply.