2998 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL.

22, 2025

Bumpless Transfer Hybrid Non-Fragile Finite-Time

Control for Markovian Jump Systems
and Its Application
Dong Yang , Member, IEEE, Qingchuan Feng , Jing Xie , and Tao Liu , Member, IEEE

Abstract— In this article, the issue of bumpless transfer hybrid tolerate the co-existing additive and multiplicative perturbations
non-fragile finite-time H∞ control for Markovian jump systems with a wider application. This study presents the method for
(MJSs) is investigated, where the transition rates are partially practitioners interested in bumpless transfer controller design.
available. The non-fragile strategy with tolerating both addi-
tive and multiplicative perturbations is designed, which greatly Index Terms— Bumpless transfer, Markovian jump systems,
relaxes the application scope of the traditional controller. The switched systems, non-fragile control, finite-time control.
bumpless transfer control idea is introduced to depict the tran-
sient behavior caused by a jumping controller. First, a bumpless I. I NTRODUCTION
transfer constraint condition is provided to restrict the amplitude
of the hybrid non-fragile jumping controller, for which the
additive and multiplicative perturbations are considered. Then,
a bumpless transfer hybrid non-fragile controller is developed to
A S A special subclass of switched systems, Markovian
jump systems (MJSs) are defined as difference or dif-
ferential equations and a stochastic function. The cooperation
guarantee the solvability of the finite-time H∞ control issue for between different equations is realized through a Markovian
MJSs with partially available transition rates. Finally, an elec- process [1], [2]. MJSs are adopted to depict complex physical
tronic circuit system example is applied to illustrate the usefulness systems subject to sudden variations in structure and param-
of the proposed bumpless transfer hybrid non-fragile control
approach.
eters. The applicability of MJSs is extended to various fields
such as aerospace engineering, robot control, and mechanical
Note to Practitioners—This article is motivated by the manufacturing [3], [4], [5]. The related work of MJSs has
finite-time hybrid non-fragile H∞ control issue of MJSs with aroused heated discussions among scholars, and many excel-
the bump limitation constraint. The bumpless transfer is often lent results have been obtained [6], [7], [8]. For an uncertain
encountered in the Markovian jumping control field because it
is usually impossible to implement unlimited control signals or
nonlinear MJS with Lévy noises, the stability problem is
bumpy control signals and is a main source of instability, and solved by an adaptive sliding mode control method in a mean-
degradation of performance of MJSs. The traditional bumpless square sense [9]. In the conclusions of the above research, the
transfer technique because of the constant control gain cannot full availability of transition rates is an additional prerequisite.
handle effectively the non-fragile bumpless transfer problem for Due to the high cost and difficulty of obtaining transition
MJSs. In this paper, the restriction of the constant control gain
is relaxed. We develop a bumpless transfer control strategy rates, the transition rates are usually unavailable. On account
considering a strong non-fragility of the control gain. In contrast of the unavailability of transition rates, there is a growing
to the existing results, a non-fragile bumpless transfer can interest in studying MJSs with partially available transition
rates [10], [11], [12]. Based on a decoupling technique, the
Manuscript received 23 December 2023; revised 20 March 2024; accepted sufficient conditions ensuring the H∞ performance of MJSs
8 April 2024. Date of publication 16 April 2024; date of current version with multiplicative noise and partially available transition rates
7 February 2025. This article was recommended for publication by Associate are developed [13]. In [14], the condition of mean-square
Editor K. Yu and Editor C. Seatzu upon evaluation of the reviewers’
comments. This work was supported in part by the National Natural Sci- stability is established for singular semi-MJSs with generally
ence Foundation of China under Grant 62273199, Grant 62173287, Grant unavailable transition rates. Nevertheless, when the transition
62273254, and Grant 62173205; in part by the Young Taishan Scholar Project rates are partially available, how to analyze the system dynam-
of Shandong Province of China under Grant tsqn202306181; and in part by the
Natural Science Foundation of Liaoning Province under Grant 2023-MS-219. ics of MJSs within a finite time interval is a serious challenge,
(Corresponding author: Dong Yang.) which partially motivates the present study.
Dong Yang and Qingchuan Feng are with the School of Engineer- In the study of MJSs, the main concern is the traditional
ing, Qufu Normal University, Rizhao, Shandong 276826, China (e-mail:
[email protected]; [email protected]). stability during an infinite time region. In practice, the dynamic
Jing Xie is with the School of Artificial Intelligence, Shenyang behavior of physical systems deserves more attention within a
University of Technology, Shenyang, Liaoning 110870, China (e-mail: finite time interval [15]. Finite-time control is used to deal with
[email protected]).
Tao Liu is with the Department of Electrical and Electronic Engineer- the dynamic behavior problem of systems within finite time
ing, The University of Hong Kong, Hong Kong, SAR, China (e-mail: intervals, which is highly effective in enhancing the transient
[email protected]). performance of MJSs. Under the influence of a designed con-
This article has supplementary downloadable material available at
https://doi.org/10.1109/TASE.2024.3387659, provided by the authors. troller, the system states are constrained to a preset boundary
Digital Object Identifier 10.1109/TASE.2024.3387659 within a specified time if the initial state starts from a given
1558-3783 © 2024 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: Bharathiar University. Downloaded on March 18,2025 at 05:34:26 UTC from IEEE Xplore. Restrictions apply.
YANG et al.: BUMPLESS TRANSFER HYBRID NON-FRAGILE FINITE-TIME CONTROL 2999

threshold, which is called the finite-time control method [16], multiplicative perturbations lead to big variations of
[17]. In recent years, plenty of results about finite-time control control inputs, which directly affect the establishment
of MJSs have been reported [18]. For MJSs with actuator of the bumpless transfer controller. The presented
failures, an event-triggered control method is developed to bumpless transfer control strategy considering a strong
ensure the finite-time boundedness and H∞ performance under non-fragility of the control gain is developed, which not
the time-varying transition probabilities [19]. Based on a dwell only reduces control bumps produced by the jumping
time method, the asynchronous finite-time control problem is controller but also allows for application to more
addressed for MJSs subject to actuator and sensor failures [20]. intricate scenarios.
However, the above results ignore the influence of the jumping (iii) Based on the stochastic multiple Lyapunov function
characteristic of a controller on finite-time stability. approach, the solvability condition achieving the control
In general, the control synthesis problem of MJSs is solved objectives is given to guarantee the bumpless trans-
by designing a jumping controller. The controllers can produce fer hybrid non-fragile finite-time H∞ performance for
different control signals, and result in control bumps during MJSs.
jumping instants [21], [22]. The control bump may destroy the
transient performance and even distort the Lyapunov stability II. P ROBLEM F ORMULATION AND P RELIMINARIES
of systems [23], [24]. The bumpless transfer scheme provides a Consider a type of MJS described as
potential solution to the control bump issue. Under an average
dwell time switching scheme, the bumpless transfer control ẋ(t) = Am(t) x(t) + Bm(t) u(t) + Dm(t) ω(t),
strategy is developed for switched systems to guarantee the z(t) = Jm(t) x(t) + Fm(t) ω(t), (1)
continuity of the control input while ensuring the asymptotic
stability of systems [25]. In [26], a control gain interpolation where x(t) ∈ R , u(t) ∈ R , z(t) ∈ R , and ω(t) ∈ Rq
n g m

technique is proposed to handle the bumpless transfer issue for denote the system state, the control input, the system output,
positive switched systems, and the L 1 gain is maintained. For and the disturbance, respectively. Assume that ω(t) satisfies
switched descriptor systems, by quantifying the state jumping, Z Tb
a switching control strategy that satisfies bumpless transfer ω T (s)ω(s)ds < ϖ, ϖ ≥ 0. (2)
0
performance both input and state is discussed [27].
The above bumpless transfer control strategies are based Am(t) , Bm(t) , Dm(t) , Fm(t) , and Jm(t) are available constant
on the precondition that the controller can be implemented matrices. {m(t), t ≥ 0} is a right continuous Markov process
accurately. The designed control strategy should have a certain taking values in a set of positive integers M = {1, 2, . . . , M}
robustness to system parameters. On account of calculation with the following transition probability
errors, actuator failures, and control channel interference, the Pr {m(t + 1h) = f |m(t) = g}
designed bumpless transfer controller should be able to tolerate (
the uncertainty or parameter perturbation [28], [29]. It is ϑg f 1h + o(1h), g ̸ = f, (3)
=
necessary to consider non-fragile bumpless transfer controllers 1 + ϑgg 1h + o(1h), g = f,
that can withstand uncertainty in gain parameters. As the
in which 1h > 0 satisfies lim o(1h) = 0, and ϑg f ≥ 0,
external environment becomes more complex, the traditional 1h→0 1h
non-fragile controllers that tolerate a single gain perturbation g ̸ = f represents the transition rate from system mode g to
M
no longer meet the practical engineering [30]. Although the ϑg f = −ϑgg for g, f ∈ M. For
P
system mode f with
bumpless transfer constraint for the non-fragile controller is g=1,g̸= f
anticipated, the developed approaches are still difficult to apply more general application scenarios, we consider generalized
due to the non-fragile behavior. This motivates our study. transition rates where ϑg f is partially available. For instance,
In this paper, we present a bumpless transfer hybrid the matrix M may be described as
non-fragile control strategy for MJSs to reduce the large ϑ11 ? · · · ?
 
control bumps while ensuring the finite-time boundedness and  ? ? · · · ϑ2M 
H∞ index. The significant contributions are summarized as M= . . . .. , (4)
 
. .
 . . . .  .
follows:
(i) In the existing work [28], [29], the non-fragile controller ϑ M1 ? · · · ϑ M M
is formed by a single feature which is either additive where ’?’ denotes that the transition rate is unavailable. Define
or multiplicative. Although additive and multiplicative M = Mag ∪ Muag , ∀g ∈ M, and
perturbations are two fundamental cases, it is not explic-
itly stated which one might be preferred since the Mag ≜ { f : ϑg f is available, for f ∈ M},
g ≜ { f : ϑg f is unavailable, for f ∈ M}.
controller is usually affected by the distinct hybrid fea- Mua (5)
tures. We developed a finite-time hybrid non-fragile H∞
controller, which can tolerate the co-existing additive When Mag ̸ = ∅, Mag can be expressed as
and multiplicative perturbations. g g
Mag = {f1 , f2 , . . . , fgp }, 1 ≤ p ≤ M, (6)
(ii) Unlike the traditional bumpless transfer control
g
approaches, the constant control gain without any in which f p denotes the pth transition rate of the set Mag in
perturbation is restrained [3], [31]. The additive and the gth row of the matrix M.

Authorized licensed use limited to: Bharathiar University. Downloaded on March 18,2025 at 05:34:26 UTC from IEEE Xplore. Restrictions apply.
3000 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 22, 2025

The hybrid non-fragile controller is developed as Substituting (7)–(9) into MJS (1), the closed-loop system is
obtained by
u(t) = (K m(t) + 1K m(t) )x(t) (7)
ẋ(t) = Am(t) + K m(t) + σm(t) Bm(t) G 1m(t) ℸ1m(t) (t)


× Q 1m(t) + 1 − σm(t) Bm(t) G 2m(t) ℸ2m(t) (t)

with
×Q 2m(t) K m(t) x(t) + Dm(t) ω(t),

(10)
1K m(t) = σm(t) 1K 1m(t) + 1 − σm(t) 1K 2m(t) K m(t) ,


(8)
z(t) = Jm(t) x(t) + Fm(t) ω(t),
where K m(t) is a nominal control gain, which will be designed
For simplicity of subsequent expression, define Am(t) ≜ A g ,
later. 1K m(t) is the hybrid parameter perturbations of the
Bm(t) ≜ Bg , Dm(t) ≜ Dg , Fm(t) ≜ Fg , Jm(t) ≜ Jg , K m(t) ≜ K g ,
controller (7). σm(t) with 0 ≤ σm(t) ≤ 1 is the weight
σm(t) ≜ σg , G 1m(t) ≜ G 1g , G 2m(t) ≜ G 2g , Q 1m(t) ≜ Q 1g ,
coefficient satisfying the transition probability (3) of Markov
Q 2m(t) ≜ Q 2g , ℸ1m(t) (t) ≜ ℸ1g (t), and ℸ2m(t) (t) ≜ ℸ2g (t),
process m(t). 1K 1m(t) and 1K 2m(t) K m(t) are the additive and
when m(t) = g, g ∈ M.
multiplicative perturbations of the form
Due to (K g + 1K g )x(t) ̸ = (K f +1K f )x(t), ∀g, f ∈
1K 1m(t) = G 1m(t) ℸ1m(t) (t)Q 1m(t) , M, hybrid non-fragile controller (7) is discontinuous at the
jumping points. We require that designed hybrid non-fragile
1K 2m(t) = G 2m(t) ℸ2m(t) (t)Q 2m(t) , (9)
controller (7) satisfies the bumpless transfer level
where G 1m(t) , G 2m(t) , Q 1m(t) , and Q 2m(t) are known matrices, ∥u ⋆ (t) − u(t)∥2 < υ∥x(t)∥2 (11)
ℸ1m(t) (t) and ℸ2m(t) (t) are unknown matrix functions satisfy-
ing ∥ℸ1m(t) (t)∥2 ≤ 1 and ∥ℸ2m(t) (t) ∥2 ≤ 1. to make the controller jumping as smooth as possible, in which
Remark 1: Owing to the inherent imprecisions and uncer- the scalar υ > 0 is called a bumpless transfer level, and
tainties, the developed controller should be able to tolerate u ⋆ (t) ≜ K ⋆ x(t) is regarded as a reference control input.
the perturbation and its own parameter uncertainty. It is Remark 4: The bumpless transfer control aims to limit the
worth noting that the perturbation and parameter uncertainty amplitude between any two consecutive jumping controllers,
are mainly categorized as multiplicative case and additive not to find a reference control input u ⋆ (t). The idea of
case. We develop a hybrid non-fragile controller, where the introducing u ⋆ (t) is to adjust the jumping controller (7) such
perturbations that appeared in the controller gains are modeled that the bumpless transfer level (11) is satisfied. The main
as uncertain gains. The proposed hybrid non-fragile control task of u ⋆ (t) is not to stabilize the system but to provide a
strategy is more general, which can describe the more general standard for limiting the distance value between u(t) and u ⋆ (t).
controller gain variation. In contrast to a signal additive or Therefore, we do not care about the actual value of u ⋆ (t).
multiplicative case [28], [29], the designed hybrid non-fragile Moreover, when the real value of u ⋆ (t) is also an important
controller can tolerate the co-existing additive and multiplica- reference, we can take K ⋆ in advance according to the actual
tive perturbations. demand, then solve the control gain K m(t) .
Remark 2: The proposed hybrid non-fragile controller (7) Remark 5: Compared with the existing work [3], [31], the
consists of the additive and multiplicative perturbation terms. bumpless transfer hybrid non-fragile controller is more general
1K 1m(t) is the additive perturbation term which often because the multiple perturbations are considered. The additive
depicts the small perturbations and uncertainties. 1K 2m(t) K m(t) and multiplicative perturbations directly affect the bumpless
is the multiplicative perturbation term that describes the transfer constraint, which brings about technical difficulties for
re-adjustment controller gains. The weight coefficient σm(t) the design of the bumpless transfer controller. The bumpless
is introduced to establish the relationship between the addi- transfer control aims to minimize control bumps between
tive perturbation 1K 1m(t) and multiplicative perturbation (K g + 1K g )x(t) and K ⋆ x(t), ∀g ∈ M.
1K 2m(t) K m(t) . The free weight coefficient can adjust the pro- Definition 1: [34] For given positive scalar Tb ,
portion between the additive and multiplicative perturbations MJS (1) is said to be finite-time bounded with respect to
so that the proposed hybrid non-fragile controller tolerates the (ε1 , ε2 , Sg , Tb , ϖ ), if there exist scalars ε1 > 0, ε2 > 0 with
co-existing additive and multiplicative perturbations. As two ε1 < ε2 , and matrix Sg > 0, g ∈ M such that
special cases, when σm(t) = 1, the controller (7) is simplified E x0T Sg x0 ≤ ε1 ⇒ E x T (t)Sg x(t) < ε2 , ∀t ∈ [0, Tb ].
 
to an additive non-fragile controller. When σm(t) = 0, con-
(12)
troller (7) is reduced to a multiplicative non-fragile controller.
The classical non-fragile controller is shown to be a special Remark 6: The traditional stability reflects the dynamic
case of the proposed hybrid non-fragile controller. behavior of the system during an infinite time region. In prac-
Remark 3: There is a significant difference between the tice, the dynamic behavior deserves more attention within a
availability of transition rates and the measurability of jumping finite time interval [16]. An MJS that is not finite-time stable if
signal [32], [33]. Although the transition rates are unavailable, its state surpasses a physical threshold during the finite-time
the jumping signal m(t) can be detected by the sensors which interval may be stochastic stable. A finite-time stable MJS
is applied to the design of the bumpless transfer hybrid non- could not be stochastic stable.
fragile controller. In most studies of MJSs, the measurability Bumpless transfer hybrid non-fragile finite-time H∞
of the jumping signal has been well recognized as a reasonable control issue: For MJS (1), find a hybrid non-fragile controller
and acceptable assumption [32], [33]. such that

Authorized licensed use limited to: Bharathiar University. Downloaded on March 18,2025 at 05:34:26 UTC from IEEE Xplore. Restrictions apply.
YANG et al.: BUMPLESS TRANSFER HYBRID NON-FRAGILE FINITE-TIME CONTROL 3001

= x T (t) A gT Pg + Pg A g + K gT BgT Pg + Pg Bg K g

(i) Hybrid non-fragile controller (7) satisfies bumpless
transfer level (11); + σg Q 1g
T T
ℸ1g (t)G 1g
T
BgT Pg + σg Pg Bg G 1g ℸ1g (t)Q 1g
(ii) MJS (1) is finite-time bounded with respect to (ε1 , ε2 , M
Sg , Tb , ϖ ); X
+ (1 − σg )K gT Q 2g
T T
ℸ2g (t)G 2g
T
BgT Pg + ϑg f P f
(iii) Under zero initial condition, MJS (1) satisfies
f =1
Z Tb Z Tb
+ (1 − σg )Pg Bg G 2g ℸ2g (t)Q 2g K g x(t)

z T (s)z(s)ds ≤ γ 2 ω T (s)ω(s)ds. (13)
0 0 + ω T (t)DgT Pg x(t) + x T (t)Pg Dg ω(t). (20)

III. M AIN R ESULTS For any symmetric matrix Rg , g ∈ M, the equation

M
ϑg f Rg = 0 always holds. We have
P
In the section, a hybrid non-fragile controller is designed
subject to additive and multiplicative perturbations. Some f =1

feasible conditions are developed. LVg (x(t))

Theorem 1: Given positive scalars Tb , α, ε1 , ε2 , ϖ , σg ,
= x T (t) A gT Pg + Pg A g + K gT BgT Pg + Pg Bg K g

h̄ ℓg , and positive definite matrix Sg , the issue of bumpless X X
transfer hybrid non-fragile finite-time H∞ control for MJS (1) + ϑg f (P f − Rg ) + ϑg f (P f − Rg )
is solvable, if there exist symmetric matrix Rg , matrices K ⋆ , f ∈Mag f ∈Mua
g
K g , positive definite matrices Pg , P̂ g , and positive scalars γ̆ , + K gT Q 2g
T T
ℸ2g (t)G 2g
T
BgT Pg − σg K gT Q 2g
T T
ℸ2g (t)G 2g
T
BgT Pg
h̄ 2g , h̄ 5g , υg for ∀g, f ∈ M, ℓ ∈ {1, 3, 4, 6} such that
+ Pg Bg G 2g ℸ2g (t)Q 2g K g − σg Pg Bg G 2g ℸ2g (t)Q 2g K g
91g Pg Dg JgT Q 1g T T
K gT Q 2g
 
+ σg Q 1g
T T
(t)G 1g
T
BgT Pg + σg Pg Bg G 1g ℸ1g (t)Q 1g x(t)

ℸ1g
 ∗ −γ̆ 2 I F T 0 0 
 g
 < 0,

(14) + ω T (t)DgT Pg x(t) + x T (t)Pg Dg ω(t). (21)
 ∗
 ∗ −I 0 0 
 ∗ ∗ ∗ −h̄ 1g I 0  According to (21), we conclude
∗ ∗ ∗ ∗ −h̄ 2g I − h̄ 3g I
LVg (x(t)) + z T (t)z(t) − γ̆ 2 ω T (t)ω(t)
P f − Rg ≤ 0, g ̸ = f, g ,
f ∈ Mua (15)
= x T (t) A gT Pg + Pg A g + K gT BgT Pg + Pg Bg K g

P f − Rg ≥ 0, g = f, g ,
f ∈ Mua (16) X X
ϖ γ̆ 1 − e b
2 −αT

+ ϑg f (P f − Rg ) + ϑg f (P f − Rg )
ρε1 + < ρε2 e−αTb , (17) f ∈Mag f ∈Mua
α
g

92g K ⋆ − K g 0 0
 + K gT T
Q 2g T
ℸ2g (t)G 2g
T
BgT Pg − σg K gT Q 2g
T T
ℸ2g (t)G 2g
T
BgT Pg
 ∗ υg I T
Q 1g T
K gT Q 2g  > 0,
 + Pg Bg G 2g ℸ2g (t)Q 2g K g − σg Pg Bg G 2g ℸ2g (t)Q 2g K g
 (18)
 ∗ ∗ h̄ 4g I 0 
+ σg Q 1g
T T
ℸ1g (t)G 1g
T
BgT Pg + σg Pg Bg G 1g ℸ1g (t)Q 1g x(t)

∗ ∗ ∗ h̄ 5g I + h̄ 6g I
+ x T (t) Pg Dg + JgT Fg ω(t) + ω T (t) DgT Pg


where
+ FgT Jg x(t) + ω T (t) FgT Fg − γ̆ 2 ω(t).



(22)
91g = A gT Pg + Pg A g + K gT BgT Pg + Pg Bg K g − α Pg Denote
X
+ ϑg f (P f − Rg ) + h̄ 1g σg2 Pg Bg G 1g G 1g
T
BgT Pg 
℧1g (t) + JgT Jg Pg Dg + JgT Fg

Ag (t) =
f ∈Mag ∗ −γ̆ 2 I + FgT Fg
+ h̄ 2g + h̄ 3g σg2 Pg Bg G 2g G 2g
T
BgT Pg ,


℧1g (t) Pg Dg
  T
Jg Jg JgT Fg
 
= + , (23)
92g = I − h̄ 4g σg2 G 1g G 1g
T
− h̄ 5g + h̄ 6g σg2 G 2g G 2g
T
, ∗ −γ̆ 2 I ∗ FgT Fg

−1 −1
P̂ g = Sg 2 Pg Sg 2 , where
ρ = max{ρmax ( P̂ g )}, ℧1g (t)
g∈M

ρ = min{ρmin ( P̂ g )}. = A gT Pg + Pg A g + K gT BgT Pg + Pg Bg K g

g∈M X X
+ ϑg f (P f − Rg ) + ϑg f (P f − Rg )
Moreover, the bumpless transfer level with υ = max υg and f ∈Mag f ∈Mua
g
√ g∈M
the H∞ index with γ = eαTb γ̆ are ensured. + σg Q 1g
T T
ℸ1g (t)G 1g
T
BgT Pg + σg Pg Bg G 1g ℸ1g (t)Q 1g
Proof. Consider the stochastic multiple Lyapunov functions + K gT Q 2g
T T
ℸ2g (t)G 2g
T
BgT Pg − σg K gT Q 2g
T T
ℸ2g (t)G 2g
T
BgT Pg
Vm(t) (t) = x T (t)Pm(t) x(t). (19) + Pg Bg G 2g ℸ2g (t)Q 2g K g − σg Pg Bg G 2g ℸ2g (t)Q 2g K g .

Assume m(t) = g ∈ M, then Vm(t) (t) = Vg (t) and Pm(t) = Pg . From (23), one gets
Calculating the weak infinitesimal operator of Vg (t) along the ℧1g (t) Pg Dg JgT
 
trajectory of system (10) obtains Bg (t) =  ∗ −γ̆ 2 I FgT . (24)
LVg (x(t)) ∗ ∗ −I

Authorized licensed use limited to: Bharathiar University. Downloaded on March 18,2025 at 05:34:26 UTC from IEEE Xplore. Restrictions apply.
3002 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 22, 2025

There exist scalars h̄ 1g > 0, h̄ 2g > 0, and h̄ 3g > 0 such that Applying Dynkin’s formula to (30), we derive
σg Pg Bg G 1g ℸ1g (t)Q 1g + σg Q 1g T T
(t)G 1g
T
BgT Pg E eαt Vg (x(t)) − Vg0 (x0 )

ℸ1g
≤ h̄ 1g σg2 Pg Bg G 1g G 1g
T
1g Q 1g Q 1g ,
BgT Pg + h̄ −1 T Z t 
(25)
eαs −z T (s)z(s) + γ̆ 2 ω T (s)ω(s) ds .
 
≤E (37)
Pg Bg G 2g ℸ2g (t)Q 2g K g + K gT Q 2g T T
ℸ2g (t)G 2g
T
BgT Pg 0

2g K g Q 2g Q 2g K g ,
BgT Pg + h̄ −1
T T T Based on (37), we get
≤ h̄ 2g Pg Bg G 2g G 2g (26)
E eαTb Vg (x(t))

and
Z Tb 
− σg Pg Bg G 2g ℸ2g (t)Q 2g K g − σg K gT Q 2g
T T
ℸ2g (t)G 2g
T
BgT Pg ≤E αs
e −z (s)z(s) + γ̆ ω (s)ω(s) ds ,
 T 2 T

(38)
≤ h̄ 3g σg2 Pg Bg G 2g G 2g
T
3g K g Q 2g Q 2g K g .
BgT Pg + h̄ −1 T T
(27) 0

which means
Substituting (25)–(27) into (24) leads to Z Tb Z Tb
℧2g (t) Pg Dg JgT z T (s)z(s)ds ≤ eαTb γ̆ 2 ω T (s)ω(s)ds.
 
(39)
Cg (t) =  ∗ −γ̆ 2 I FgT , (28) 0
√
0

∗ ∗ −I Hence, inequality (13) holds with γ = eαTb γ̆ .

in which It can be derived by condition (18) that
I + ℧3g K ⋆ − K g
 
℧2g (t) = A gT Pg + Pg A g + K gT BgT Pg + Pg Bg K g > 0, (40)
X X ∗ υg I + ℧4g
+ ϑg f (P f − Rg ) + ϑg f (P f − Rg )
f ∈Mag f ∈Mua
where
g

℧3g = −h̄ 4g σg2 G 1g G 1g

T
− h̄ 5g + h̄ 6g σg2 G 2g G 2g
T
,


+ h̄ 1g σg2 Pg Bg G 1g G 1g
T
BgT Pg + h̄ −1 T
1g Q 1g Q 1g
℧4g = −h̄ 4g Q 1g Q 1g − h̄ 5g + h̄ 6g K g Q 2g Q 2g K g .
−1 T −1 −1 T T


+ h̄ 2g + h̄ 3g σg2 Pg Bg G 2g G 2gT
BgT Pg

2g + h̄ 3g K g Q 2g Q 2g K g .
+ h̄ −1 −1 T T


There exist positive scalars h̄ 4g , h̄ 5g , and h̄ 6g such that
 
Applying Schur complement, we know that the inequality ℧3g 0
Cg (t) < 0 is guaranteed by condition (14), which implies that 0 ℧4g
Ag (t) < 0. In view of conditions (14)–(16), we obtain 
0 −σg G 1g ℸ1g (t)Q 1g − (1 − σg )G 2g ℸ2g (t)Q 2g K g

≤ .
LVg (x(t)) ≤ αVg (x(t)) − z T (t)z(t) + γ̆ 2 ω T (t)ω(t). (29) ∗ 0
(41)
Multiplying both sides of (29) by e−αt yields
Therefore
L Vg (x(t))e−αt ≤ −z T (t)z(t) + γ̆ 2 ω T (t)ω(t) e−αt . (30)
   
I K ⋆ − K g − 1K g
 
> 0. (42)
Moreover, we have ∗ υg I
LVg (x(t)) ≤ e−αt γ̆ 2 ω T (t)ω(t). (31) From (42), one has
Letting m(t0 ) = g0 ∈ M yields υg I > [K ⋆ − (K g + 1K g )]T [K ⋆ − (K g + 1K g )], (43)
Z Tb
Vg (x(t)) − Vg0 (x0 ) ≤ γ̆ 2 ϖ e−αs ds.
 −αTb
E e (32) which is equivalent to
0
υg x T (t)x(t)
From (32), we can infer that
"
# > x T (t)[K ⋆ − (K g + 1K g )]T [K ⋆ − (K g + 1K g )]x(t).
αTb ϖ γ̆ 2 1 − e−αTb
E{Vg (x(t))} ≤ e Vg0 (x0 ) + . (33) (44)
α
1 1 When the gth hybrid non-fragile controller is activated, (44)
Defining Pg ≜ Sg2 P̂ g Sg2 , one has guarantees
"
#
αTb ϖ γ̆ 2 1 − e−αTb υg ∥x(t)∥2 > K ⋆ x(t) − (K g + 1K g )x(t) .
2
(45)
E{Vg (x(t))} ≤ e ρε1 + . (34)
α
Obviously, the bumpless transfer level (11) with υ = max υg is
g∈M
Noting that satisfied for hybrid non-fragile controller (7). This completes
x T (t)Pg x(t) ≥ ρE x T (t)Sg x(t) ,

(35) the proof. □
Since the terms K gT BgT Pg , h̄ 1g σg Pg Bg G 1g G 1g
T
BgT Pg , Pg
one obtains
Bg K g , and h̄ 2g (1 − σg )Pg Bg G 2g G 2g
T
BgT Pg appeared in condi-
ϖ γ̆ 2 (1−e−αTb )
h i
eαTb ρε1 + α
tion (14) result in the synthesis condition for hybrid non-fragile
E x T (t)Sg x(t) < < ε2 .

(36) controller (7) being non-convex. To find the convex solution
ρ
condition, we give the following result.
Clearly, MJS (1) is the finite-time bounded with respect to Theorem 2: Given positive scalars Tb , α, ε1 , ε2 , ϖ , σg ,
(ε1 , ε2 , Sg , Tb , ϖ ). h̄ ℓg , and positive definite matrix Sg , the issue of bumpless

Authorized licensed use limited to: Bharathiar University. Downloaded on March 18,2025 at 05:34:26 UTC from IEEE Xplore. Restrictions apply.
YANG et al.: BUMPLESS TRANSFER HYBRID NON-FRAGILE FINITE-TIME CONTROL 3003

√
transfer hybrid non-fragile finite-time H∞ control for MJS (1) with γ = eαTb γ̆ are achieved. Moreover, controller gain in
is solvable, if there exist positive scalars γ̆ , h̄ 2g , h̄ 5g , ℵ, υg , the hybrid non-fragile controller (7) is given by K g = Yg X g−1 .
matrices K ⋆ , Yg , symmetric matrix Rg , and positive definite Proof. Define Pg ≜ X g−1 and K g ≜ Yg X g−1 . Performing a
matrix X g , for ∀g, f ∈ M, ℓ ∈ {1, 3, 4, 6} such that transformation to condition (14) by diag{X g , I , I , I , I }, one

93g Dg X g JgT X g Q 1g T T
YgT Q 2g 94g
 gets
 ∗ −γ̆ 2 I F T 0 0 0 

81g Dg X g JgT X g Q 1g T T
YgT Q 2g

 g 
 ∗ ∗ −I 0 0 0   ∗ −γ̆ 2 I F T 0 0 
  g
 < 0,
 
 ∗
 ∗ ∗ − h̄ 1g I 0 0 
  ∗
 ∗ −I 0 0  (53)
 ∗ ∗ ∗ ∗ −Wg 0   ∗ ∗ ∗ −h̄ 1g I 0 
∗ ∗ ∗ ∗ ∗ −95g ∗ ∗ ∗ ∗ −h̄ 2g I − h̄ 3g I
< 0, g ∈ Mag , (46) in which
96g Dg X g JgT T
YgT T
97g
 
X g Q 1g Q 2g
 ∗ −γ̆ 2 I 81g = X g A gT + A g X g + YgT BgT + Bg Yg − α X g
 FgT 0 0 0  X  
 ∗
 ∗ −I 0 0 0  + f X g − X g R g X g + h̄ 1g σg Bg
ϑg f X g X −1 2
 ∗
 ∗ ∗ −h̄ 1g I 0 0  f ∈Mag
 ∗ ∗ ∗ ∗ −Wg 0  T
× G 1g G 1g BgT + h̄ 2g + h̄ 3g σg2 Bg G 2g G 2g

T
BgT .
∗ ∗ ∗ ∗ ∗ −98g
< 0,
g ∈ Mua , (47) Let X g Rg X g = Rg . Due to ϑgg < 0, ∀g ∈ M, (53) is dealt
  g with by the two different cases.
−Rg X g Case 1. When g ∈ Mag , (53) is rewritten as
< 0, g ̸ = f, g ,
f ∈ Mua (48)
∗ −X f
ˇ 1g Dg X g JgT X g Q 1g
T T
YgT Q 2g
 
X f − Rg > 0, g = f, f ∈ Mua , (49)
# g  ∗ −γ̆ 2 I F T 0 0
( )

√
" g
ϖ γ̆ 2
1−e −αTb
−ε2 e−αTb + ε1 < 0,  < 0,
 
α (50)
 ∗
 ∗ −I 0 0  (54)
∗ −ℵ  ∗ ∗ ∗ −h̄ 1g I 0 
ℵSg−1 < X g < Sg−1 , (51) ∗ ∗ ∗ ∗ −h̄ 2g I − h̄ 3g I
92g K ⋆ Yg where
 
0 0 0
 ∗ υg I I I Q 1gT
0

ˇ 1g = X g A gT + A g X g + YgT BgT + Bg Yg + (ϑgg − α)X g

 
 ∗ ∗ Xg 0 0 0  > 0,

(52)
X X
ϑg f X g X −1 ϑg f Rg + h̄ 1g σg2
 T
 ∗ ∗ ∗ Xg 0 YgT Q 2g  + f Xg −
 
 ∗ ∗ ∗ ∗ h̄ 4g I 0  f ∈Mag , f ̸=g f ∈Mag
T
BgT + h̄ 2g + h̄ 3g σg2 Bg G 2g G 2g
T
BgT .


∗ ∗ ∗ ∗ ∗ h̄ 5g I + h̄ 6g I × Bg G 1g G 1g
where One can see that condition (46) is guaranteed by (54).
93g = X g A gT + A g X g + YgT BgT + Bg Yg − α X g Case 2. When g ∈ Mua g , (53) is rewritten as
+ h̄ 1g σg2 Bg G 1g G 1g
T
BgT + h̄ 2g + h̄ 3g σg2 Bg G 2g G 2g
T
BgT


ˆ 1g Dg T
X g JgT X g Q 1g T
YgT Q 2g
 

−
X
ϑg f Rg + ϑgg X g ,  ∗ −γ̆ 2 I FgT 0 0 
 < 0,
 
f ∈Mag
 ∗
 ∗ −I 0 0  (55)
hq q q  ∗ ∗ ∗ −h̄ 1g I 0 
94g = ϑgf1g X g ϑgf2g X g · · · ϑgfg−1
g Xg ∗ ∗ ∗ ∗ −h̄ 2g I − h̄ 3g I
q q i
ϑgfg+1
g Xg ··· ϑgfgp X g , where
ˆ 1g = X g A gT + A g X g + YgT BgT + Bg Yg − α X g

n o
95g = diag X f1g , X f2g , . . . , X fg−1
g , X fg+1
g , . . . , X fgp , X X
Wg = h̄ 2g I + h̄ 3g I, + ϑg f X g X −1
f Xg − ϑg f Rg + h̄ 1g σg2
f ∈Mag , f ̸=g f ∈Mag
96g = X g A gT + A g X g + YgT BgT + Bg Yg − α X g T
BgT + h̄ 2g + h̄ 3g σg2 Bg G 2g G 2g
T
BgT .


× Bg G 1g G 1g
+ h̄ 1g σg2 Bg G 1g G 1g
T
BgT + h̄ 2g + h̄ 3g σg2 Bg G 2g G 2g
T
BgT


X One can see that condition (81) is guaranteed by (55). Pre-and
− ϑg f Rg ,
post-multiplying condition (15) by X g , we obtain
f ∈Mag

f X g − Rg < 0,
X g X −1 g ̸ = f, g ,
f ∈ Mua (56)
hq q q i
97g = ϑgf1g X g ϑgf2g X g · · · ϑgfgp X g ,
n o which derives condition (48). Performing a transformation
98g = diag X f1g , X f2g , . . . , X fgp to condition (16) by X g , we know that condition (16) is
g g g g guaranteed by
with f1 , f2 , . . . , f p described as in (6) and fg = g. The
bumpless transfer level with υ = max υg and the H∞ index X f − R f > 0, g = f, g .
f ∈ Mua (57)
g∈M

Authorized licensed use limited to: Bharathiar University. Downloaded on March 18,2025 at 05:34:26 UTC from IEEE Xplore. Restrictions apply.
3004 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 22, 2025

From conditions (50) and (51), we get Condition (67) shows

I K ⋆ − K g − 1K g
 
ϖ γ̆ 1 − e
2 −αTb


ℵ−1 ε1 < ε2 e−αTb + , ≥ 0, (69)
α
(58) ∗ υg I
1 1
ℵI < Sg2 X g Sg2 < I. (59) in which
1 1 1K g = σg G 1g ℸ1g (t)Q 1g + (1 − σg )G 2g ℸ2g (t)Q 2g K g .
Together with Sg = Sg2 X g Sg2 , one has
Bumpless transfer level (11) holds. This completes the
ℵ < ρ(Sg ) < 1, ∀g ∈ M, (60) proof. □
ρ < ℵ−1 , (61) Remark 7: The parameters ε1 , ε2 , υg and g influence the
1 < ρ, (62) conservatism and computational burden for solving condi-
tions (46)–(52). Too small υg may lead to the insolubility
which leads to of condition (52), and too large υg usually brings about big
ϖ γ̆ 2 1 − e−αTb


control bumps. A smaller g can lead to less computational
ρε1 + < ρε2 e−αTb . (63) burden. In practice, we should choose the parameters ε1 , ε2 ,
α
Based on condition (52), we derive υg , and g to realize a trade-off between the actual demand and
allowed computational complexity.
92g K⋆
 
Yg 0
 ∗ υg I − 83g I I 
  IV. E XTENSION
 ∗ ∗ Xg 0 
∗ ∗ ∗ X g − 84g In most existing results, the control problem of MJSs is
widely studied based on a prerequisite of synchronization
I K ⋆ Yg 0 82g 0 0 0
   
 ∗ υg I I I   ∗ 83g 0 0  between the controller and the controlled system. However,
= ∗ ∗ Xg 0  −  ∗ ∗ 0 0 
   due to the packet dropout and stochastic perturbation, such
kind of ideal synchronization may be difficult to be satisfied
∗ ∗ ∗ Xg ∗ ∗ ∗ 84g
in practical applications, which may bring asynchronous phe-
> 0, (64) nomena between the modes of controller and system. Recently,
where the importance of asynchronous control for MJSs has already
begun receiving more attention [35], [36], [37], [38].
82g = h̄ 4g σg2 G 1g G 1g
T
+ h̄ 5g + h̄ 6g σg2 G 2g G 2g
T
,


In order to handle such an asynchronization case, the hidden
83g = h̄ 4g Q 1g Q 1g , 84g = h̄ 5g + h̄ 6g Yg Q 2g Q 2g Yg .
−1 T −1 −1 T T


Markov model framework in [35] is established to characterize
the asynchronous phenomenon between the controller and the
There exist positive scalars h̄ 4g , h̄ 5g , and h̄ 6g such that controlled system. The passivity asynchronous control issue
0 σg G 1g ℸ1g (t)Q 1g 0 −(1 − σg )G 2g ℸ2g (t)Q 2g Yg
 
of delayed singular Markov jump systems is investigated
∗ 0 0 0  in [36]. In [37], when the asynchronous phenomena occur,
 
∗ ∗ 0 0  the dissipative control is investigated for Markovian fuzzy
∗ ∗ ∗ 0 jump systems by designing a kind of special asynchronous

82g 0 0 0
 controller. For 2-D MJSs, the issue of asynchronous fault
 ∗ 83g 0 0  detection is addressed [38]. How to extend the proposed
≤ ∗ ∗ 0 0 . (65) bumpless transfer hybrid non-fragile control strategy to an


∗ ∗ ∗ 84g asynchronous control version for MJSs? It is a significant

problem that deserves further study.
According to (64) and (65), we obtain Motivated by the above discussions, in this section,
I 85g Yg (1 − σg )G 2g ℸ2g (t)Q 2g Yg we extend the synchronization results developed in Section III
 
 ∗ υg I I I  to an asynchronous control version. The hybrid non-fragile
 ≥ 0, (66) asynchronous controller is developed as

 ∗ ∗ Xg 0 
∗ ∗ ∗ Xg u(t) = (K j (t) + 1K j (t) )x(t) (70)
⋆
where 85g = K − σg G 1g ℸ1g (t)Q 1g . From (66), one gets with
I − Yg X g−1 YgT − 86g K ⋆ − Yg X g−1 − 87g
 
1K j (t) = σ j (t) 1K 1 j (t) + 1 − σ j (t) 1K 2 j (t) K j (t) ,


≥ 0, (67) (71)
∗ υg I − 2X g−1
where K j (t) is a nominal control gain, which will be designed
where
later. 1K j (t) is the hybrid parameter perturbations of the
86g = (1 − σg )2 G 2g ℸ2g (t)Q 2g Yg X g−1 YgT Q 2g
T T
ℸ2g (t)G 2g
T
, controller (70). j (t) is a hidden Markov progress, which takes
87g = σg G 1g ℸ1g (t)Q 1g + (1 − σg )G 2g ℸ2g (t)Q 2g Yg X g−1 . values in another positive integer set J = {1, 2, . .. , J}, and
obeys a conditional probability matrix J = ηgl with the
Notice that probability transition as
Yg X g−1 YgT + 86g 0
 
0 2X g−1
≥ 0. (68) Pr { j (t) = l|m(t) = g} = ηgl . (72)

Authorized licensed use limited to: Bharathiar University. Downloaded on March 18,2025 at 05:34:26 UTC from IEEE Xplore. Restrictions apply.
YANG et al.: BUMPLESS TRANSFER HYBRID NON-FRAGILE FINITE-TIME CONTROL 3005

For all g ∈ M, l ∈ J, the conditional probability ηgl belongs to + Pg Bg G 2l ℸ2l (t)Q 2l K l − σl Pg Bg G 2l ℸ2l (t)Q 2l K l
J
+ σl Q 1l
T T
ℸ1l (t)G 1l
T T
Bg Pg + σl Pg Bg G 1l ℸ1l (t)Q 1l x(t)

ηgl = 1. σ j (t) with 0 ≤ σ j (t) ≤ 1 is
P
the interval [0, 1] and
l=1
+ x T (t) Pg Dg + JgT Fg ω(t) + ω T (t) DgT Pg


the weight coefficient satisfying the transition probability (72)
of hidden Markov process j (t). 1K 1 j (t) and 1K 2 j (t) K j (t) is + FgT Jg x(t) + ω T (t) FgT Fg − γ̆ 2 ω(t),



(76)
with the form in (9).
We rewrite the closed-loop system as M
ϑg f Rg = 0 for any symmetric matrix Rg , g ∈
P
with
ẋ(t) = [Am(t) + 1 − σ j (t) Bm(t) G 2 j (t) ℸ2 j (t) (t)Q 2 j (t) K j (t) f =1


M. By using the proof method similar to (23)–(39) as in
+ Bm(t) K j (t) + σ j (t) Bm(t) G 1 j (t) ℸ1 j (t) (t)Q 1 j (t) ]x(t) Theorem 1, we confirm that MJS (1) is the finite-time bounded
+ Dm(t) ω(t), (73) and satisfies the inequality (13).
z(t) = Jm(t) x(t) + Fm(t) ω(t). It can be derived by condition (75) that
I + ℧3l K ⋆ − K l
 
Define K j (t) ≜ K l , σ j (t) ≜ σl , G 1 j (t) ≜ G 1l , G 2 j (t) ≜ G 2l , > 0, (77)
Q 1 j (t) ≜ Q 1l , Q 2 j (t) ≜ Q 2l , ℸ1 j (t) (t) ≜ ℸ1l (t), and ℸ2 j (t) (t) ≜ ∗ υl I + ℧4l
ℸ2l (t), when j (t) = l, l ∈ J. The main corollaries about the where
bumpless transfer hybrid non-fragile finite-time asynchronous
control issue for MJSs are obtained. ℧3l = −h̄ 4l σl2 G 1l G 1l
T
− h̄ 5l + h̄ 6l σl2 G 2l G 2l
T
,


Corollary 1: Given positive scalars Tb , α, ε1 , ε2 , ϖ , h̄ 1gl ,
℧4l = −h̄ 4l Q 1l Q 1l − h̄ 5l + h̄ 6l K l Q 2l Q 2l K l .
−1 T −1 −1 T T


h̄ 2gl , h̄ 3gl , h̄ 4l , h̄ 5l , h̄ 6l , and positive definite matrix Sg , the
issue of bumpless transfer hybrid non-fragile finite-time H∞ There exist positive scalars h̄ 4l , h̄ 5l , and h̄ 6l such that
control for MJS (1) is solvable, if there exist symmetric matrix
Rg , matrices K ⋆ , K l , positive definite matrices Pg , P̂ g , and
 
℧3l 0
positive scalars γ̆ , σl , υl for ∀g ∈ M, ∀l ∈ J such that 0 ℧4l
0 −σl G 1l ℸ1l (t)Q 1l − (1 − σl )G 2l ℸ2l (t)Q 2l K l
 
91gl Pg Dg JgT
 
≤ . (78)
 ∗ −γ̆ 2 I FgT  < 0, (74) ∗ 0
∗ ∗ −I Therefore
92l K ⋆ − K l 0
 
0
I K ⋆ − K l − 1K l
 
 ∗ υl I T
Q 1l K lT Q 2l
T
 > 0,
 > 0. (79)

 ∗ ∗ h̄ 4l I 0  (75) ∗ υl I
∗ ∗ ∗ h̄ 5l I + h̄ 6l I From (79), one has
and (15)–(17) hold, where
υl I > [K ⋆ − (K l + 1K l )]T [K ⋆ − (K l + 1K l )], (80)
J
X
91gl = A gT Pg + Pg A g − α Pg + ηgl K lT BgT Pg + h̄ −1

2gl which is equivalent to
l=1

3gl K l Q 2l Q 2l K l + h̄ 1gl σl Pg Bg G 1l G 1l Bg Pg υl ∥x(t)∥2 > K ⋆ x(t) − (K l + 1K l )x(t) .

2
+ h̄ −1 2

T T T T
(81)
+ h̄ 2gl + h̄ 3gl σl2 Pg Bg G 2l G 2lT T
Bg Pg + h̄ −1 T


1gl Q 1l Q 1l
 X Obviously, the bumpless transfer level (11) with υ = max υl is
+ Pg Bg K l + ϑg f (P f − Rg ), l∈J
satisfied for hybrid non-fragile asynchronous controller (70).
f ∈Mag
This completes the proof. □
92l = I − h̄ 4l σl2 G 1l G 1l
T
− h̄ 5l + h̄ 6l σl2 G 2l G 2l
T
.


Corollary 2: Given positive scalars Tb , α, ε1 , ε2 , ϖ , h̄ 1gl ,
h̄ 2gl , h̄ 3gl , h̄ 4l , h̄ 5l , h̄ 6l , and positive definite matrix Sg , the
Moreover, the bumpless transfer level with υ = max υl and
√ l∈J issue of bumpless transfer hybrid non-fragile finite-time H∞
the H∞ index with γ = eαTb γ̆ are ensured. control for MJS (1) is solvable, if there exist positive scalars
Proof. Choose the stochastic multiple Lyapunov functions γ̆ , σl , υl , ℵ, matrices K ⋆ , Yl , K, symmetric matrices Rg , and
J
positive definite matrix X g , for ∀g ∈ M, ∀l ∈ J such that
as (19). Notice that ηgl ∈ [0, 1] and ηgl = 1. We get
P
l=1
−ϒ1 ϒ2gl 0 KT JgT KT
 
0 0 0
LVg (x(t)) + z (t)z(t) − γ̆ ω (t)ω(t)
T 2 T
 ∗ −ϒ3gl Dg 0 0 ϒ4gl ϒ6gl ϒ8gl 
∗ −γ̆ I Fg
  2 T

X X
 ∗
 0 0 0 0  
= x T (t) ϑg f (P f − Rg ) + ϑg f (P f − Rg )  ∗
 ∗ ∗ −I 0 0 0 0  
 ∗ ∗ ∗ ∗ −ξ X g 0 0 0 
 a ua
f ∈Mg f ∈Mg  
J
X
 ∗
 ∗ ∗ ∗ ∗ −ϒ5gl 0 0  
+ A gT Pg + Pg A g + ηgl K lT BgT Pg + Pg Bg K l  ∗ ∗ ∗ ∗ ∗ ∗ −ϒ7gl 0 


l=1 ∗ ∗ ∗ ∗ ∗ ∗ ∗ −ϒ9gl
+ K lT Q 2l
T T
ℸ2l (t)G 2l
T T
Bg Pg − σl K lT Q 2l
T T
ℸ2l (t)G 2l
T T
Bg Pg < 0, g ∈ Mag , (82)

Authorized licensed use limited to: Bharathiar University. Downloaded on March 18,2025 at 05:34:26 UTC from IEEE Xplore. Restrictions apply.
3006 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 22, 2025

−ϒ1 ϒ2gl 0 KT JgT KT

 
0 0 0 Proof. Conditions (82) and (83) are deduced as
 ∗ −ϒ10gl Dg 0 0 53gl ϒ6gl ϒ11gl 
−ϒ1 ϒ2gl 0 KT JgT KT
 
γ̆
 2 T

 ∗ ∗ − I Fg 0 0 0 0 
   ∗ 81gl Dg 0 0 
 ∗ ∗ ∗ −I 0 0 0 0 
γ̆  < 0,
 2 T

 ∗ ∗ − I F 0 (85)
 
 ∗ ∗ ∗ ∗ −ξ X g 0 0 0   g 
 ∗ ∗ ∗ −I 0 
 
 ∗ ∗ ∗ ∗ ∗ −ϒ5gl 0 0 
∗ ∗ ∗ ∗ −ξ X g
 
 ∗ ∗ ∗ ∗ ∗ ∗ −ϒ7gl 0 
∗ ∗ ∗ ∗ ∗ ∗ ∗ −ϒ12gl where
< 0, g ,
g ∈ Mua (83) X  
81gl = ϑg f X g X −1
f X g − Rg − (ξ
−1
+ α)X g
92l K⋆
 
Yl 0 0 0 f ∈Mag

 ∗ υl I I I T
Q 1l 0 
 J
∗ ∗ K 0 0 0
X
ηgl h̄ 2gl + h̄ 3gl σl2 Bg G 2l G 2l
T T


 > 0, + Bg
 
 (84)

 ∗ ∗ ∗ K 0 T
YlT Q 2l 
 l=1
∗ ∗ ∗ ∗ h̄ 4l I 0 T
+ (h̄ −1
2gl + h̄ 3gl )X g K
−1 −1
YlT Q 2lT
Q 2l Yl K−1 X g
 
∗ ∗ ∗ ∗ ∗ h̄ 5l I + h̄ 6l I
1gl X g Q 1l Q 1l X g + h̄ 1gl σl Bg G 1l G 1l Bg .
+ h̄ −1 T 2 T T

and (48)–(51) hold, where
J
X Define Yl ≜ K l K, L̂−1
gl ≜ Lgl , and
ϒ1 = K + K , ϒ2gl = K T T
A gT + ηgl YlT BgT + Xg,  J

ηgl Bg K l
l=1
P
 Ag + I 0 0 0
J l=1
.
 
X X Ugl ≜  0 0 I 0 0 (86)
ϒ3gl = ϑg f Rg − ηgl h̄ 1gl σl2 Bg G 1l G 1l
T

Bg 
f ∈Mag l=1
 Jg 0 0 I 0
I 000 I
+ h̄ 2gl + h̄ 3gl σl2 Bg G 2l G 2l T
Bg + α + ξ −1


T
Pre- and post-multiplying (85) by Ugl and Ugl gets
− ϑgg X g ,


82gl Dg X g JgT
 
√ T √ √
ϒ4gl ηg1 X g Q 11 ηg2 X g Q 12 T
... ηgJ X g Q 1J T
,

=  ∗ −γ̆ 2 I FgT  < 0, (87)
ϒ5gl = diag h̄ 1g1 I, h̄ 1g2 I, . . . , h̄ 1gJ I ,
 ∗ ∗ −I
√ T
T √ in which
ϒ6gl = ηg1 (h̄ 2g1 + h̄ 3g1 )X g K−1 Y1T Q 21 ηg2 (h̄ 2g2
J
T √ X   X
+ h̄ 3g2 )X g K−1 Y2T Q 22 T
... ηgJ (h̄ 2gJ 82gl = ϑg f X g X −1 ηgl X g K lT BgT

f X g − R g +
T f ∈Mag l=1
+ h̄ 3gJ )X g K−1 YJT Q 2J T
,

Bg K l X g + h̄ 2gl + h̄ 3gl σl2 Bg G 2l G 2l T T


+ Bg
ϒ7gl = diag (h̄ 2g1 + h̄ 3g1 )I, (h̄ 2g2 + h̄ 3g2 )I, . . . ,

T
+ (h̄ 2gl + h̄ 3gl )X g K Yl Q 2l Q 2l Yl K X g
−1 −1 −1 T T −1

(h̄ 2gJ + h̄ 3gJ )I ,

1gl X g Q 1l Q 1l X g + h̄ 1gl σl Bg G 1l G 1l Bg
h̄ −1 T 2 T T

hq q +
q
ϒ8g = ϑgf1g X g ϑgf2g X g · · · ϑgfg−1
g Xg + X g A gT + A g X g − α X g .
Let X g = Pg−1 and Rg = X g Rg X g . Performing a transforma-
q q i
ϑgfg+1
g Xg · · · ϑgfgp X g ,
tion to (87) by diag{X g−1 , I , I }, one can see that condition (74)
is guaranteed.
n o
ϒ9g = diag X f1g , X f2g , . . . , X fg−1
g , X fg+1
g , . . . , X fgp ,
Based on condition (84), we derive
K⋆
J
92l
 
X X Yl 0
ϒ10gl = ϑg f Rg − ηgl h̄ 1gl σl2 Bg G 1l G 1l
T

Bg  ∗ υl I − 84l I I 
f ∈Mag l=1
 
 ∗ ∗ K 0 
+ h̄ 2gl + h̄ 3gl σl2 Bg G 2l G 2l T
Bg + α + ξ −1 X g , ∗ ∗ ∗ K − 85l



I K ⋆ Yl 0 83l 0 0 0
   
hq q q i
ϒ11g = ϑgf1g X g ϑgf2g X g · · · ϑgfgp X g ,  ∗ υl I I I   ∗ 84l 0 0 
n o =∗ ∗ K 0 − ∗ ∗ 0 0 
   (88)
ϒ12g = diag X f1g , X f2g , . . . , X fgp ∗ ∗ ∗ K ∗ ∗ ∗ 85l
g g g
with f1 , f2 , . . . , f p described as in (6) and fg = g. The
g > 0,
bumpless transfer level with υ = max υl and the H∞ index where
√ l∈J
with γ = eαTb γ̆ are achieved. Moreover, controller gain in 83l = h̄ 4l σl2 G 1l G 1l
T
+ h̄ 5l + h̄ 6l σl2 G 2l G 2l

T
,
the hybrid non-fragile asynchronous controller (70) is given
84l = h̄ 4l Q 1l Q 1l , 85l = h̄ 5l + h̄ 6l Yl Q 2l Q 2l Yl .
−1 T −1 −1 T T


by K l = Yl K−1 .

Authorized licensed use limited to: Bharathiar University. Downloaded on March 18,2025 at 05:34:26 UTC from IEEE Xplore. Restrictions apply.
YANG et al.: BUMPLESS TRANSFER HYBRID NON-FRAGILE FINITE-TIME CONTROL 3007

There exist positive scalars h̄ 4l , h̄ 5l , and h̄ 6l such that

I 86l (t) Yl (1 − σl )G 2l ℸ2l (t)Q 2l Yl
 
 ∗ υl I I I 
  ≥ 0, (89)
∗ ∗ K 0 
∗ ∗ ∗ K
where 86l (t) = K ⋆ − σl G 1l ℸ1l (t)Q 1l . From (89), one gets
I − Yl K−1 YlT − 87l (t) K ⋆ − Yl K−1 − 88l (t)
 
Fig. 1. Electronic circuit.
≥ 0, (90)
∗ υl I − 2K−1
in which where R = 2 , L 1 = 2 mH, L 2 = 4 mH, L 3 = 5 mH,
87l (t) =(1 − σl ) G 2l ℸ2l (t)Q 2l Yl K
2 −1
YlT
Q 2l ℸ2l (t)G 2l
T T T
, C1 = 20 mF, C2 = 10 mF, and C3 =  40 mF.
UC (t)
88l (t) =σl G 1l ℸ1l (t)Q 1l + (1 − σl )G 2l ℸ2l (t)Q 2l Yl K .
−1 Define the system state x(t) ≜ and the control
i L (t)
Notice that input u(t) ≜ U (t). For g = 1, 2, 3, matrices A g , Bg , Dg , Jg ,
Fg , G 1g , Q 1g , G 2g , and Q 2g are parametrized and expressed
Yl K−1 YlT + 87l (t) 0
 
≥ 0, (91) by
0 2K−1      
−25 50 0 4
which shows A1 = , B1 = , D1 = ,
−500 0 500 2
I K ⋆ − K l − 1K l
 
≥ 0, (92) J1 = [ −2 − 7 ], F1 = 0.1, G 11 = 0.35,
∗ υl I Q 11 = [ 2 3 ], G 21 = 0.21, Q 21 = 5,
     
where −50 100 0 1
A2 = , B2 = , D2 = ,
−250 0 250 3
1K l = σl G 1l ℸ1l (t)Q 1l + (1 − σl )G 2l ℸ2l (t)Q 2l K l .
J2 = [ −1 − 5 ], F2 = 0.2, G 12 = 0.42,
Bumpless transfer level (11) holds. This completes the Q 12 = [ 1 1 ], G 22 = 0.52, Q 22 = 3,
proof. □      
−12.5 25 0 2
Remark 8: When the bumpless transfer hybrid non-fragile A3 = , B3 = , D3 = ,
−200 0 200 5
finite-time asynchronous control issue is considered for MJSs,
the existence of the multiplicative perturbation 1K 2 j (t) K j (t) J3 = [ −3 − 2 ], F3 = 0.3, G 13 = 0.33,
prevents the production of the linear matrix inequality solving Q 13 = [ 1 3 ], G 23 = 0.23, Q 23 = 4.
condition from its original version to an asynchronous version.  
1
In the future, the non-convex term may allow by use an effi- Choose the parameters x0 = , ε1 = 2.2, ε2 = 4,
2
cient computational tools of solution. One can solve (82)–(84)  
0.4 0
by resorting to the existing methods [39]. First, take K in Tb = 3 s, S1 = S2 = S3 = , α = 0.01,
advance according to the allowed computational complexity. 0 0.4
X g , g ∈ M can be obtained from (48)–(51). Then, Yl , l ∈ J ω(t) = 0.0001
1+t
, σ1 = 0.0117, σ2 = 0.0297, σ3 = 0.0347,
can be obtained from the conditions in Corollary 2. ℸ11 (t) = ℸ12 (t) = ℸ13 (t) = cos(πt), ℸ21 (t) = ℸ22 (t) =
ℸ23 (t) = sin(π t), h̄ 11 = 6.3, h̄ 12 = 5.3, h̄ 13 = 6.1, h̄ 21 =
0.0028, h̄ 22 = 0.0063, h̄ 23 = 0.0196, h̄ 31 = 2.4, h̄ 32 = 3.4,
V. A N E XAMPLE
h̄ 33 = 1.4, h̄ 41 = 10.9, h̄ 42 = 10.5, h̄ 43 = 10.3, h̄ 51 = 2.4003,
The proposed bumpless transfer finite-time hybrid non- h̄ 52 = 0.6295,  h̄ 53 = 0.8103,
 h̄ 61 = 21, h̄ 62 = 25, h̄ 63 = 23,
fragile H∞ control strategy is applied to an electronic circuit −2 ? ?
model [40] which is shown in Fig. 1. Our control goal is to M = [ϑg f ] =  ? −5 ? .
suppress large power source increments generated at jumping ? 1 ?
instants while ensuring the finite-time H∞ performance of the By adopting two different control strategies, we show a
electronic circuit. The electronic circuit consists of a collection comparative study. The proposed bumpless transfer hybrid
of inductors L g , a collection of capacitors C g , a resistor R, non-fragile finite-time H∞ control strategy is represented as
and a power source increment U (t), g = 1, 2, 3. Different Method (A). The finite-time non-fragile H∞ control strategy
inductors and capacitors are changed by jumping. i R (t) and in [41] is represented as Method (B). By solving Theorem 2,
i L (t) represent the current increment through the resistor R we obtain
and the inductor L g . UC (t) stands for the voltage increment K 1 = [ 0.7423 − 0.1596 ], K 2 = [ 0.5360 − 0.3572 ],
across the capacitors C g . Adopting the Kirchhoff current and
voltage laws gives rise to K 3 = [ 0.7788 − 0.1462 ], K ⋆ = [ 0.7378 − 0.1747 ],
γ = 3.0706, ϱ = 15.1935.
1
C g U̇ C (t) = − UC (t) + i L (t), From [41], we get
R
L g i̇ L (t) = −UC (t) + U (t), (93) K̃ 1 = [ 0.7637 0.1204 ], K̃ 2 = [ 0.5312 − 0.3944 ],

Authorized licensed use limited to: Bharathiar University. Downloaded on March 18,2025 at 05:34:26 UTC from IEEE Xplore. Restrictions apply.
3008 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 22, 2025

Fig. 2. System mode of MJS. Fig. 7. Second component of control gain.

Fig. 3. Response of the voltage increment UC (t). Fig. 8. Power source increment U (t).

variations of control gain, which indicates that the amplitude

of control gain varies less in method (A). Fig. 8 is the variation
of power source increment U (t). The amplitude of the power
source increment U (t) in Method (A) is significantly better
than one in Method (B), although the proposed bumpless
transfer hybrid non-fragile control strategy largely increases
the computational complexity and the control design difficulty.
Fig. 4. Response of the current increment i L (t). Therefore, the proposed bumpless transfer finite-time hybrid
non-fragile H∞ control technique effectively suppresses the
occurrence of large power source increments.

VI. C ONCLUSION
For MJSs, the issue of bumpless transfer hybrid non-fragile
finite-time H∞ control has been dealt with. A non-fragile
strategy has been designed to tolerate the additive and multi-
plicative perturbations. To depict the transient behavior caused
by a jumping controller, a bumpless transfer control idea has
Fig. 5. Curve of x T (t) Sm(t) x(t).
been given. First, a bumpless transfer constraint condition has
been provided to limit the amplitude of the hybrid non-fragile
controller. Second, a bumpless transfer hybrid non-fragile
controller has been developed to guarantee the solvability
of the finite-time H∞ control issue for MJSs with partially
available transition rates. Finally, a practical example has been
given to show the usefulness of the proposed bumpless transfer
hybrid non-fragile control strategy.

Fig. 6. First component of control gain. R EFERENCES

[1] G. Chen, J. Sun, and J. Chen, “Passivity-based robust sampled-data
control for Markovian jump systems,” IEEE Trans. Syst., Man, Cybern.,
K̃ 3 = [ 0.7761 − 0.1727 ], γ̃ = 3.0801. Syst., vol. 50, no. 7, pp. 2671–2684, Jul. 2020.
[2] H. Chen, G. Zong, X. Liu, X. Zhao, B. Niu, and F. Gao, “A
The different jumping signals are given in Fig. 2. Responses sub-domain-awareness adaptive probabilistic event-triggered policy for
attack-compensated output control of Markov jump CPSs with dynami-
of the voltage increment UC (t) and the current increment cally matching modes,” IEEE Trans. Autom. Sci. Eng., pp. 1–13, 2004,
i L (t) are given in Figs. 3 and 4. Fig. 5 depicts the evo- doi: 10.1109/TASE.2023.3296259.
lution of x T (t)Sm(t) x(t). Under the time interval [0, 3], [3] D. Yang, G. Zong, S. K. Nguang, and X. Zhao, “Bumpless transfer H∞
anti-disturbance control of switching Markovian LPV systems under the
x T (t)Sm(t) x(t) ≤ 3 is guaranteed by the bumpless transfer hybrid switching,” IEEE Trans. Cybern., vol. 52, no. 5, pp. 2833–2845,
hybrid non-fragile controller (7). Figs. 6 and 7 give the May 2022.

Authorized licensed use limited to: Bharathiar University. Downloaded on March 18,2025 at 05:34:26 UTC from IEEE Xplore. Restrictions apply.
YANG et al.: BUMPLESS TRANSFER HYBRID NON-FRAGILE FINITE-TIME CONTROL 3009

[4] L. Zhang, H. Liang, Y. Sun, and C. Ahn, “Adaptive event-triggered [25] Y. Shi and X.-M. Sun, “Bumpless transfer control for switched linear
fault detection scheme for semi-Markovian jump systems with output systems and its application to aero-engines,” IEEE Trans. Circuits Syst.
quantization,” IEEE Trans. Syst., Man, Cybern., Syst., vol. 51, no. 4, I, Reg. Papers, vol. 68, no. 5, pp. 2171–2182, May 2021.
pp. 2370–2381, Apr. 2021. [26] Y. Zhao, J. Zhao, and J. Fu, “Bumpless transfer control for switched
[5] D. Tong, C. Xu, Q. Chen, W. Zhou, and Y. Xu, “Sliding mode control positive linear systems with L1 gain property,” Nonlinear Anal. Hybrid
for nonlinear stochastic systems with Markovian jumping parameters Syst., vol. 33, pp. 249–264, Aug. 2019.
and mode-dependent time-varying delays,” Nonlinear Dyn., vol. 100, [27] J. Shi and J. Zhao, “State bumpless transfer control for a class of
no. 2, pp. 1343–1358, Apr. 2020. switched descriptor systems,” IEEE Trans. Circuits Syst. I, Reg. Papers,
[6] J. Cheng, L. Xie, D. Zhang, and H. Yan, “Novel event-triggered protocol vol. 68, no. 9, pp. 3846–3856, Sep. 2021.
to sliding mode control for singular semi-Markov jump systems,” [28] S. Fan, H. Yan, H. Zhang, H. Shen, and K. Shi, “Dynamic event-based
Automatica, vol. 151, May 2023, Art. no. 110906. non-fragile dissipative state estimation for quantized complex networks
[7] T. Wu, Y. Zhu, L. Zhang, and P. Shi, “Bumpless transfer model predictive with fading measurements and its application,” IEEE Trans. Circuits
control for Markov jump linear systems,” IEEE Trans. Autom. Control, Syst. I, Regul. Pap., vol. 68, no. 2, pp. 856–867, Feb. 2021.
vol. 69, no. 2, pp. 1348–1355, Feb. 2024. [29] H. Shen, X. Hu, J. Wang, J. Cao, and W. Qian, “Non-fragile H∞
[8] N. Zhang, S. Huang, L. Ning, and W. Li, “Semi-global sampling control synchronization for Markov jump singularly perturbed coupled neural
for semi-Markov jump systems with distributed delay,” IEEE Trans. networks subject to double-layer switching regulation,” IEEE Trans.
Autom. Sci. Eng., pp. 1–12, 2004, doi: 10.1109/TASE.2023.3282053. Neural Netw. Learn. Syst., vol. 34, no. 5, pp. 2682–2692, May 2023.
[9] C. Xu, D. Tong, Q. Chen, W. Zhou, and P. Shi, “Exponential stability of [30] B. Lyu, C. Liu, and X. Yue, “Hybrid nonfragile intermediate observer-
Markovian jumping systems via adaptive sliding mode control,” IEEE based T-S fuzzy attitude control for flexible spacecraft with input
Trans. Syst., Man, Cybern., Syst., vol. 51, no. 2, pp. 954–964, Feb. 2021. saturation,” Aerosp. Sci. Technol., vol. 128, Sep. 2022, Art. no. 107753.
[10] G. Zong, Y. Li, and H. Sun, “Composite anti-disturbance resilient control [31] D. Yang, G. Zong, and S. Nguang, “H∞ bumpless transfer reliable con-
for Markovian jump nonlinear systems with general uncertain transition trol of Markovian switching LPV systems subject to actuator failures,”
rate,” Sci. China Inf. Sci., vol. 62, no. 2, Feb. 2019. Inf. Sci., vol. 512, pp. 431–445, Feb. 2020.
[11] H. Yan, Y. Tian, H. Li, H. Zhang, and Z. Li, “Input–output finite-time [32] Y. Zhang, Y. He, M. Wu, and J. Zhang, “Stabilization for Markovian
mean square stabilization of nonlinear semi-Markovian jump systems,” jump systems with partial information on transition probability based
Automatica, vol. 104, pp. 82–89, Jun. 2019. on free-connection weighting matrices,” Automatica, vol. 47, no. 1,
pp. 79–84, Jan. 2011.
[12] H. Liang, L. Zhang, Y. Sun, and T. Huang, “Containment control of
[33] L. Zhang and E.-K. Boukas, “Stability and stabilization of Markovian
semi-Markovian multiagent systems with switching topologies,” IEEE
jump linear systems with partly unknown transition probabilities,” Auto-
Trans. Syst., Man, Cybern., Syst., vol. 51, no. 6, pp. 3889–3899,
matica, vol. 45, no. 2, pp. 463–468, Feb. 2009.
Jun. 2021.
[34] X. Luan, F. Liu, and P. Shi, “Observer-based finite-time stabilization
[13] X. Liu, J. Zhuang, and Y. Li, “H∞ filtering for Markovian jump linear
for extended Markov jump systems,” Asian J. Control, vol. 13, no. 6,
systems with uncertain transition probabilities,” Int. J. Control, Autom.
pp. 925–935, Nov. 2011.
Syst., vol. 19, no. 7, pp. 2500–2510, 2021.
[35] Z.-G. Wu, P. Shi, Z. Shu, H. Su, and R. Lu, “Passivity-based asyn-
[14] B. Jiang, Y. Kao, H. R. Karimi, and C. Gao, “Stability and stabilization
chronous control for Markov jump systems,” IEEE Trans. Autom.
for singular switching semi-Markovian jump systems with generally
Control, vol. 62, no. 4, pp. 2020–2025, Apr. 2017.
uncertain transition rates,” IEEE Trans. Autom. Control, vol. 63, no. 11,
pp. 3919–3926, Nov. 2018. [36] F. Li, C. Du, C. Yang, and W. Gui, “Passivity-based asynchronous
sliding mode control for delayed singular Markovian jump systems,”
[15] B. Niu, B. Yan, X. Zhao, B. Zhang, T. Zhao, and X. Liu, IEEE Trans. Autom. Control, vol. 63, no. 8, pp. 2715–2721, Aug. 2018.
“Event-triggered adaptive command filtered bipartite finite-time track-
[37] Z.-G. Wu, S. Dong, H. Su, and C. Li, “Asynchronous dissipative control
ing control of nonlinear coopetition MASs with time-varying dis-
for fuzzy Markov jump systems,” IEEE Trans. Cybern., vol. 48, no. 8,
turbances,” IEEE Trans. Autom. Sci. Eng., pp. 1–13, 2004, doi:
pp. 2426–2436, Aug. 2018.
10.1109/TASE.2023.3297253.
[38] P. Cheng et al., “Asynchronous fault detection observer for 2-D Markov
[16] F. Amato and M. Ariola, “Finite-time control of discrete-time linear
jump systems,” IEEE Trans. Cybern., vol. 52, no. 12, pp. 13623–13634,
systems,” IEEE Trans. Autom. Control, vol. 50, no. 5, pp. 724–729,
Dec. 2022.
May 2005.
[39] L. I. Allerhand and U. Shaked, “Robust state-dependent switching of
[17] T. Zhang, X. Li, and S. Song, “Finite-time stabilization of switched linear systems with dwell time,” IEEE Trans. Autom. Control, vol. 58,
systems under mode-dependent event-triggered impulsive control,” IEEE no. 4, pp. 994–1001, Apr. 2013.
Trans. Syst., Man, Cybern., Syst., vol. 52, no. 9, pp. 5434–5442,
[40] D. Yang, Q. Feng, and G. Zong, “Asynchronous bumpless transfer finite-
Nov. 2022.
time H∞ control for Markovian jump systems with application to circuit
[18] Z. Cao, Y. Niu, and J. Song, “Finite-time sliding-mode control of Marko- system,” IEEE Trans. Circuits Syst. II, Exp. Briefs, vol. 69, no. 12,
vian jump cyber-physical systems against randomly occurring injection pp. 4929–4933, Dec. 2022.
attacks,” IEEE Trans. Autom. Control, vol. 65, no. 3, pp. 1264–1271,
[41] R. Kavikumar, R. Sakthivel, O. M. Kwon, and B. Kaviarasan, “Reliable
Mar. 2020.
non-fragile memory state feedback controller design for fuzzy Markov
[19] X. Gao, F. Deng, X. Li, and Q. Zhou, “Event-triggered finite-time jump systems,” Nonlinear Analysis: Hybrid Syst., vol. 35, Feb. 2020,
reliable control for nonhomogeneous Markovian jump systems,” Int. J. Art. no. 100828.
Robust Nonlinear Control, vol. 30, no. 5, pp. 1831–1849, Mar. 2020.
[20] F. Li, C. Du, C. Yang, L. Wu, and W. Gui, “Finite-time asynchronous
sliding mode control for Markovian jump systems,” Automatica,
vol. 109, Nov. 2019, Art. no. 108503.
[21] P. Cheng, H. Chen, S. He, and W. Zhang, “Asynchronous decon- Dong Yang (Member, IEEE) received the M.S.
volution filtering for 2-D Markov jump systems with packet loss degree in mathematics from the College of Engineer-
compensation,” IEEE Trans. Autom. Sci. Eng., pp. 1–12, 2004, doi: ing, Qufu Normal University, Qufu, China, in 2010,
10.1109/TASE.2023.3292891. and the Ph.D. degree in control theory and con-
[22] K. Ding and Q. Zhu, “Extended dissipative anti-disturbance control for trol engineering from the College of Information
delayed switched singular semi-Markovian jump systems with multi- Science and Engineering, Northeastern University,
disturbance via disturbance observer,” Automatica, vol. 128, Jun. 2021, Shenyang, China, in 2017.
Art. no. 109556. In 2020, he was a Senior Research Assistant
[23] T. Yang and J. Dong, “Predefined-time adaptive fault-tolerant control with the Department of Electrical and Electronic
for switched odd-rational-power multi-agent systems,” IEEE Trans. Engineering, The University of Hong Kong, SAR,
Autom. Sci. Eng., vol. 20, no. 4, pp. 2423–2434, Oct. 2023, doi: China. In 2021, he was a Post-Doctoral Fellow with
10.1109/TASE.2022.3208029. the Department of Electrical and Electronic Engineering, The University of
[24] Y. Jiang, B. Niu, G. Zong, X. Zhao, and P. Zhao, “Distributed adaptive Hong Kong. He is currently an Associate Professor with the College of
secure consensus tracking control for asynchronous switching nonlinear Engineering, Qufu Normal University, Rizhao, China. His research interests
MASs under sensor attacks and actuator faults,” IEEE Trans. Autom. include Markovian jump systems, switched LPV systems, switched systems,
Sci. Eng., pp. 1–11, 2004, doi: 10.1109/TASE.2023.3277488. bumpless transfer, H∞ control, and their applications.

Authorized licensed use limited to: Bharathiar University. Downloaded on March 18,2025 at 05:34:26 UTC from IEEE Xplore. Restrictions apply.
3010 IEEE TRANSACTIONS ON AUTOMATION SCIENCE AND ENGINEERING, VOL. 22, 2025

Qingchuan Feng received the B.S. degree in Tao Liu (Member, IEEE) received the B.E. degree
automation and the M.S. degree in electrical engi- from the College of Information Science and Engi-
neering from the College of Engineering, Qufu neering, Northeastern University, Shenyang, China,
Normal University, Qufu, China, in 2020 and 2023, in 2003, and the Ph.D. degree from the Research
respectively. His research interests include Marko- School of Engineering, The Australian National Uni-
vian jump systems, bumpless transfer, H∞ control, versity, Melbourne, Australia, in 2011.
and electronic circuit. In 2012, he was a Visiting Scholar with the Cen-
tre for Future Energy Networks, The University of
Sydney. From 2012 to 2015, he was a Post-Doctoral
Fellow with The Australian National University, the
University of Groningen, and The University of
Hong Kong. In 2015, he became a Research Assistant Professor. He is
currently an Assistant Professor with the Department of Electrical and
Electronic Engineering, The University of Hong Kong. His research interests
Jing Xie received the M.S. degree in mathematics include power system analysis and control, complex dynamical networks,
from the College of Sciences, Northeastern Univer- distributed control, and event-triggered control.
sity, Shenyang, China, in 2009, and the Ph.D. degree
in control theory and control engineering from the
College of Information Science and Engineering,
Northeastern University, in 2018.
In 2018, she was a Lecturer with the School
of Artificial Intelligence, Shenyang University of
Technology, Shenyang. She is currently an Associate
Professor with the School of Artificial Intelligence,
Shenyang University of Technology. Her research
interests include switched systems, bumpless transfer, adaptive control, and
their applications.

Authorized licensed use limited to: Bharathiar University. Downloaded on March 18,2025 at 05:34:26 UTC from IEEE Xplore. Restrictions apply.