Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 68, NO.

11, NOVEMBER 2023 6975

Event-Triggered Control for Nonlinear Uncertain Systems via a


Prescribed-Time Approach
Pengju Ning , Changchun Hua , Senior Member, IEEE, Kuo Li , Member, IEEE, and Rui Meng

Abstract—The event-triggered control problem is investigated in [2] based on the state-scaling design method, where the nonlinear
for nonlinear uncertain systems via a prescribed-time approach. function or disturbance only matches the control input. The core of
Different from the existing event-triggered control methods that
can only achieve bounded or finite-time stability, this article fo- the state-scaling-based design method is to scale the system state by a
cuses on the prescribed-time stability under the event-based de- time-varying function that grows to infinity toward the specified time. In
sign method. First, a novel prescribed-time stability criterion is addition, the stability criteria in [1] and [2] required that the Lyapunov
presented, which provides explicit theoretical supports and sys- function V satisfies V̇ (t) ≤ −cμ(t)V (t) + μ(t)ε(t), where μ(t) is the
tematic guidelines for addressing the studied problem. Then, we
put forward a new method for co-designing the controller and
time-varying function and |ε(t)| ≤ ε̄ with the positive constant ε̄. Based
event-triggering mechanism by utilizing the idea of backstepping on the state-scaling design idea, many extensions of PT stabilization
design. Based on the proposed stability criterion, it is proved that can be found in [3], [4], [5], [6], and [7]. In [3], the global PT control
the controller constructed with the new method can regulate the problem for the linearized Schrödinger equation was investigated. For
system states to zero in the prescribed-time interval while avoiding the linear systems in observer canonical form and controllable canonical
the Zeno phenomenon, thus improving the previous prescribed-
time stabilization results that are all based on the real-time control form, the PT estimation and output feedback control problems were
signals. Finally, a simulation example is given to show the effec- addressed in [4] and [5], respectively. Gao et al. [6] investigated the
tiveness of proposed control strategy. PT state feedback control problem for switched nonlinear systems.
Index Terms—Event-triggered control (ETC), nonlinear uncer-
The work in [7] focused on the adaptive PT stabilization of nonlinear
tain systems, prescribed-time (PT) stabilization. systems with uncertain parameters. Different from the state-scaling
based design approach, Krishnamurthy et al. [8] developed a temporal
transformation framework on the basic of the dynamic high-gain scaling
I. INTRODUCTION technique to solve the global PT control problem for the nonlinear un-
Recently, prescribed-time (PT) control has attracted more and more certain systems, where the uncertain terms satisfy the known nonlinear
attention from scholars due to its wide applications in the field of missile growth conditions. Then, Krishnamurthy et al. [9] further considered
guidance, aircraft rendezvous, and intelligent robot control. The salient the output feedback PT stabilization of the nonlinear uncertain systems
feature of PT control is that it can regulate system states into the origin in normal form by utilizing the temporal transformation framework.
within a specified time interval [1]. That is, the settling time of the In [10], based on the more generalized time-varying function, a new
system can be preset arbitrarily according to the practical requirements PT stability criterion was proposed, which required the Lyapunov
irrespective of system initial conditions and control parameters [2], function to satisfy V̇ (t) ≤ −cμ(t)V (t) with the positive constant c.
which is fundamentally different from finite-time (FT) control and The works in [11] and [12] used the linear time-varying high-gain
fixed-time control. The scheme for solving PT stabilization of high- feedback technique to study the PT stabilization of linear systems, and
order nonlinear systems was first proposed in [1] and further researched extended the results to strict-feedback nonlinear systems in [13] where
the nonlinearities meet the linear growth condition.
Although the results of [1], [2], [3], [4], [5], [6], [7], [8], [9], [10],
Manuscript received 16 October 2022; revised 1 February 2023; [11], [12], and [13] can achieve PT stability, the computational burden
accepted 3 February 2023. Date of publication 10 February 2023; for derivation of time-varying functions is enormous, especially when
date of current version 26 October 2023. This work was supported the system order is high. Therefore, the work in [14] proposed a PT
in part by the National Natural Science Foundation of China under
Grant 61825304, Grant U20A20187, and Grant 62103221; in part by control strategy for stochastic systems based on the nonscaling design
the Science Fund for Creative Research Groups of Hebei Province framework, which can largely reduce computational burden. Note that
under Grant F2020203013; in part by the Science and Technology the stability analysis of the results in [1], [2], [3], [4], [5], [6], [7], [8],
Development Grant of Hebei Province under Grant 20311803D; in part [9], [10], [11], [12], [13], and [14] was based on the stability criteria
by the Hebei Innovation Capability Improvement Plan Project under
proposed in [1] and [10]. That is, the Lyapunov function needs to
Grant 22567619H; in part by the Hebei Graduate Innovation Funding
Project under Grant CXZZBS2022137; in part by Guangxi Science and appear completely on the right-hand side of its derivative inequality.
Technology Major Special Project Guike under Grant AA22067064; in However, in most cases, only part of the Lyapunov function can appear
part by Liuzhou Science and Technology Plan Project under Grant on the right-hand side of its derivative inequality. Then, an adaptive PT
2022AAB0101; in part by the National Construction of High-Level Uni- stability criterion was put forward in [15], which required the Lyapunov
versity Public Postgraduate Project under Grant 202208130099; and
in part by the Post Doctoral Innovative Talent Support Program under function V to satisfy V̇ (t) ≤ −cμ(t)Ṽ (t) with the positive definite
Grant BX2021157. Recommended by Associate Editor G. Gu. (Corre- function Ṽ (t) ≤ V (t). Moreover, compared to [14], the result in [15]
sponding author: Changchun Hua.) can further reduce the computational burden. Regrettably, the existing
The authors are with the Institute of Electrical Engineering, Yanshan PT stability criteria are independent of each other and have certain
University, Qinhuangdao 066004, China (e-mail: pjn-best@foxmail.com;
limitations. Therefore, it is of great significance to propose a universal
cch@ysu.edu.cn; good-will@foxmail.com; cherry6266@foxmail.com).
Color versions of one or more figures in this article are available at PT stability criterion. Furthermore, the existing PT stabilization results
https://doi.org/10.1109/TAC.2023.3243863. required that the control signals must be continuously transmitted be-
Digital Object Identifier 10.1109/TAC.2023.3243863 tween the controller and actuator, which means that the control signals

0018-9286 © 2023 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: Thapar Institute of Engineering & Technology. Downloaded on January 12,2024 at 07:43:29 UTC from IEEE Xplore. Restrictions apply.
6976 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 68, NO. 11, NOVEMBER 2023

need to be updated in real time. With the continuous development of II. PRELIMINARIES AND PROBLEM FORMULATION
communication technology and network engineering, the information
A. Preliminaries
transmissions between plant and control center are gradually connected
through the network communication channels. Therefore, reducing the Definition 1: Consider the following nonlinear system:
occupation of communication bandwidth and saving communication
˙ = f (t, ξ(t), u(t))
ξ(t) (1)
resources have important application value and theoretical research
significance.
In the past few decades, event-triggered control (ETC) has increas- where ξ ∈ Rn is the system state vector, u ∈ Rm stands for the
ingly become a hot topic in the field of network control, since it can save controller, f (·) : R≥0 × Rn × Rm → Rn is piecewise continuous in
communication bandwidth and reduce the signal transmission burden of t, and locally Lipschitz in ξ with f (t, 0, 0) = 0. For any given positive
channels [16]. Many constructive works of ETC for nonlinear systems constant Tp , if there is a controller
have been proposed in [16], [17], [18], and [19]. However, most of u(t) = υ(t, ξ), υ(t, 0) = 0 (2)
the ETC results can only achieve bounded stability. Then, Zhang and
Yang [20] designed an event-based FT controller, which can regulate such that the system (1) is asymptotically stable, and limt→Tp− ξ
system states to zero in an FT interval. By introducing the sign function (t) = 0, ξ(t) = 0∀t ≥ Tp , then the equilibrium point ξ = 0 of system
in controller design, Hua et al. [21] focused on the FT control of high- (1) is PT stable.
order interconnected nonlinear systems in the presence of constraint Definition 2 ([15]): If a continuous function satisfies
communications. For p-type nonlinear systems and uncertain systems in
the normal form, the event-based FT control problems were investigated μ(t) > 0 ∀t ∈ [0, Tp )
in [22] and [23], respectively. However, the upper bound of the settling
time in these results depends on the system initial conditions and control lim (Tp − t)μ(t) = ρ (3)
t→Tp−
parameters. Therefore, the abovementioned FT control methods cannot
regulate system to the origin at arbitrary specified time. where ρ is a positive constant or +∞, then μ(t) is called a prescribed-
To the best of our knowledge, there are no ETC results that can adjust time adjustment (Tp -PTA) function.
the system to the origin in a PT interval. The difficulties to handle this Lemma 1 ([24]): For a continuous function g(y) ≥ 0 defined on
problem are mainly from the following three aspects. y ∈ [a, b) with the flaw y = b, if it satisfies
1) Lack of clear theoretical support and systematic guidelines; the
existing PT stability criteria cannot be directly used for the stability lim (b − y)g(y) = d (4)
y→b−
analysis of ETC results.
2) How to balance the contradiction between system performance and b
where d is a positive constant or +∞, then a g(y) dy =
saving communication resources; the ETC methods show signifi- s
lims→b− a g(y)dy is divergent. If there is a constant p ∈ (0, 1), the
cant advantages in saving communication resources, but they can following formula holds:
reduce system performance and prolong the settling time compared
to the real-time control methods. lim (b − y)p g(y) = d = ∞ (5)
3) How to ensure that the controller can be updated in time when y→b−

the system returns to the origin; the existing controller and event- b s
triggering mechanism co-design methods are powerless to deal then a g(y)dy = lims→b− a g(y)dy is convergent.
with this situation. Lemma 2 ([28]): For any variables x(t) ∈ R and 0 < π(t) ∈ R,
The main contributions of this article are summarized as follows. the following inequality holds:
1) A new control method for co-designing event-triggering mecha- x2 (t)
nism and controller is proposed via the PT approach. The traditional − < π(t) − |x(t)|. (6)
ETC results in [16], [17], [18], and[19] can only partially eliminate x2 (t) + π 2 (t)
the effect of event-triggered error and achieve bounded stability. Lemma 3 (PT stability criterion): If there are two nonnegative
Different from these results, the ETC method in this article can differentiable functions V (t) and Ṽ (t) satisfying
fully compensate for the event-triggered error and achieve PT
stability, which means that system states converge to the origin Ṽ (t) ≤ V (t), V̇ (t) ≤ −cμ(t)Ṽ (t) + δ(t), t ∈ [0, Tp ) (7)
in a preassigned time interval irrelevant to initial conditions and
control parameters of system. where c is a positive constant, μ(t) is a Tp -PTA function, and δ(t) ≤ δ̄
2) Based on the limit convergence theorem of abnormal integrals and is a continuous function with the constant δ̄, then we can get that V (t)
contradiction method, we propose a novel PT stability criterion, is bounded for all t ∈ [0, Tp ) and limt→Tp− Ṽ (t) = 0.
which provides a powerful tool to address the event-based PT Proof: From the definition of the Tp -PTA function μ(t) and the
control problem. Moreover, the proposed stability criterion is more inequality (7), one can easily get
general and can be simplified to those in [1], [10], and [15].
3) Compared to the existing PT stabilization results in [1], [2], [3], V (t) ≤ V (0) + δ̄Tp , t ∈ [0, Tp ) (8)
[4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], and [15]
that require real-time control signals, our proposed co-design which implies that V (t) is bounded ∀t ∈ [0, Tp ). According to the
method can simultaneously save communication resources and Gronwall’s lemma, the following inequality can be derived:
reduce the computational burden arising from the derivative of the  t
time-varying function. In particular, the proposed event-triggering lim− cμ(t)Ṽ dt ≤ V (0) + δ̄Tp = Δ (9)
t→Tp 0
mechanism can ensure timely updating of the controller when the
system returns to the origin. where Δ is a positive constant.

Authorized licensed use limited to: Thapar Institute of Engineering & Technology. Downloaded on January 12,2024 at 07:43:29 UTC from IEEE Xplore. Restrictions apply.
IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 68, NO. 11, NOVEMBER 2023 6977

In the following, the contradiction method is applied to prove limited communication resources; unmanned aerial vehicle (UAV)
limt→Tp− Ṽ = 0. First, one assumes that swarm performances, each UAV needs to reach the designated location
within a preassigned time under limited communication bandwidth.
lim Ṽ (t) = o = 0 (10) Therefore, the research on event-based PT control of system (13) is of
t→Tp−
great significance.
where o is a positive constant. On the one hand, since cμ(t)Ṽ ≥ 0,
τ
F (τ ) = 0 cμ(t)Ṽ dt is monotonically increasing on [0, Tp ). Hence, III. MAIN RESULTS
one can easily verify that To reduce communication transmission, save communication re-
 Tp  t sources, and release the limitations of system convergence time on
cμ(t)Ṽ dt = lim− cμ(t)Ṽ dt ≤ Δ (11) initial conditions and control parameters, a new ETC strategy is pro-
0 t→Tp 0
posed via a PT approach in this section.
 Tp
then 0 cμ(t)Ṽ dt is convergent. On the other hand, from the defini-
tion of μ(t), we have A. Virtual Controller Design

lim c(Tp − t)μ(t)Ṽ = cρo > 0 (12) The detailed design procedure of the virtual controller will be pro-
t→Tp− posed by application of the backstepping-like method. For simplicity
T of analysis, the coordinate transformation is described as follows:
one can get that the improper integral 0 p cμ(t)Ṽ dt is divergent
from Lemma 1. Through comparison and analysis, it can be found z1 (t) = ξ1 (t)
that the contents of these two aspects are contradictory. Therefore, the 
assumption in (10) cannot hold, which implies limt→Tp− Ṽ (t) = o = 0. zi (t) = ξi (t) − αi−1 t, ξ̄i−1 (t), θ̂(t) , i = 2, . . . , n (14)
The proof is completed.
Remark 1: Different from the PT stability criteria in [1], [10], where θ̂(t) = θ − θ̃(t) is the estimate of the uncertain parameter vector
and [15], the Lyapunov function V is required to meet V̇ (t) ≤ −cμ(t) θ, θ̃(t) is the estimation error. The virtual controller αi (·) and the
V (t) + μ(t)δ(t), V̇ (t) ≤ −cμ(t)V (t), and V̇ (t) ≤ −cμ(t)Ṽ (t), re- ˙
adaptive law θ̂(·) are designed as
spectively, our proposed stability criterion only requires that the Lya-
punov function satisfies the condition V̇ (t) ≤ −cμ(t)Ṽ (t) + δ(t). In ˙
θ̂(·) = Γτn , α1 (·) = −σ1 μ(t)z1 − f1 − ϕT1 θ̂ (15)
other words, if we set Ṽ (t) = V (t) and/or δ(t) = 0 in (7), our proposed
i−1
stability criterion can be simplified to those in [1], [10] and [15]. In ∂αi−1
addition, from the proof procedure of Lemma 3, it is not hard to deduce αi (·) = − σi μ(t)zi −zi−1 −fi − θ̂T ϕi − ϕj
j=1
∂ξj
that if the continuous function δ(t) in (7) is set as a constant, i.e.,
δ(t) = δ̄, Lemma 3 is also effective. In summary, the stability criterion i−1
∂αi−1 ∂αi−1 ∂αi−1
proposed in this article can be regarded as a generalization of those + (ξj+1 +fj )+ + Γτi
j=1
∂ξj ∂t ∂ θ̂
in [1], [10], and [15].
i−1 i−1
∂αi−1 ∂αj−1
B. Problem Formulation + ϕi − ϕj Γ zj (16)
j=1
∂ξj j=2 ∂ θ̂
The event-triggered PT stabilization problem is investigated for the
following nonlinear uncertain systems: where σi > n − i + 1, i = 1, . . . , n, is a positive design parameter,
∂αi−1
⎧ τ1 = z1 ϕ1 , τi = τi−1 + zi (ϕi − i−1 j=1 ∂ξj ϕj ), i = 2, . . . , n, and
⎨ξ˙i = ξi+1 + fi (ξ¯i ) + ϕTi (ξ¯i )θ
Γ ∈ Rr×r is a positive definite matrix. In addition, to reduce the
i = 1, . . . , n − 1 (13)
⎩˙ computational burden caused by the derivative of the Tp -PTA function
ξn = u + fn (ξ¯n ) + ϕTn (ξ¯n )θ
μ(t), we choose μ(t) as
where ξi ∈ R, θ ∈ Rr , and u ∈ R are system state variable, uncertain
1
parameter vector, and control input, respectively; ξ¯i = [ξ1 , . . . , ξi ]T ∈ μ(t) = , t ∈ [0, Tp ) (17)
Tp − t
Ri ; ξ = [ξ1 , . . . , ξn ]T ; fi (·) : Ri → R is a known smooth function
with fi (0) = 0; ϕi (·) : Ri → Rr is a known smooth function vector μ(t) = ς, t ∈ [Tp , +∞) (18)
with ϕi (0) = 0.
The objective of this article is to construct a new scheme for co- where ς is a positive design parameter. Since most of the existing
designing the event-triggering mechanism and controller, such that all PT control approaches are only valid for [0, Tp ), and cannot extend
state variables of system (13) can be steered to the origin within the PT to [Tp , +∞), we subtly replace the positive design parameter ς with
Tp , while avoiding the Zeno phenomenon. μ(t) after the time Tp , thus extending the time domain to infinity, i.e.,
Remark 2: The strict-feedback nonlinear system (13) in the form [0, +∞).
of a lower-triangular structure can be described as many practical sys- According to the definition of zi in (14), and the form of virtual
tems, such as the single-link robotic manipulator system, mass–spring controller αi in (15)–(16), one can easily get that
mechanical system, teleoperation system, spacecraft attitude control σi zi
system, quad rotor control system, and wheeled mobile robot. For żi = − + Ωi , i = 1, . . . , n − 1 (19)
Tp − t
system (13), the ETC problem is considered in [16], [17], [18], and [19],
i−1 ∂αi−1 T
which can only realize bounded stability. In practice, we often require where Ω1 = z2 + ϕT1 θ̃ and Ωi = zi+1 − zi−1 − j=1 ∂ξj ϕj θ̃ +
mechanical systems to complete a task precisely within a specific time i−1 ∂αi−1 i−1 ∂αj−1 ∂αi−1 ˙
interval, such as spacecraft autonomous rendezvous and docking, the (ϕi − j=1 ∂ξj ϕj )Γ j=2 ∂ θ̂
zj + ∂ θ̂
(Γτi − θ̂) +
spacecraft needs to achieve precise docking within a preset time under ϕTi θ̃, i = 2, . . . , n.

Authorized licensed use limited to: Thapar Institute of Engineering & Technology. Downloaded on January 12,2024 at 07:43:29 UTC from IEEE Xplore. Restrictions apply.
6978 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 68, NO. 11, NOVEMBER 2023

Based on the state transformation (14), the Lyapunov function is the effects of d(t) and achieve FT stability, the sign function was intro-
chosen as duced into υ(t), which leads to the discontinuity of υ(t). In this case,
n excluding Zeno behavior would be extremely complicated. Different
1 1 T −1
V = V1 + V2 , V1 = zj2 , V2 = θ̃ Γ θ̃ (20) from these results, the co-design control method of event-triggering
2 j=1
2 mechanism and controller in this article can completely offset the
event-triggered error and ensure the continuity of υ(t). In this way, we
˙
From the state transformation (14), adaptive law θ̂ (15), virtual can not only achieve the PT stability, but also facilitate the elimination
controller (16), and Lyapunov function (20), the derivative of V satisfies of Zeno behavior.
n
V̇ = −μ(t) σj zj2 + zn (u − αn ). (21) C. Stability Analysis
j=1
Theorem 1: For the system (13), by developing the controller (22)
Remark 3: Obviously, if a controller designed without consider- and event-triggering mechanism (23), such that the equilibrium point
ing the introduction of an event-triggering mechanism in the control ξ = 0 of (13) is PT stable and the Zeno behavior is excluded.
scheme, the real-time controller can be directly constructed as u = αn , Proof: According to the description of controller u in (25), the form
then such control strategy is reduced to the one proposed in [15]. of υ in (24) and the Lemma 2, it is not difficult to prove that

zn u ≤ −|zn αn | + 2β(t) (26)


B. Event-Triggered Controller Design
In this section, by introducing a time-varying function, we propose where β(t) = m1 e−m2 μ(t) is a monotonically decreasing function with
m2

a method for co-designing the controller and event-triggering mecha- the upper bounded β̄ = m1 e Tp .
nism, which is shown as follows: Substituting (26) into (21), it is easy to verify that
u(t) = υ(tk ) ∀t ∈ [tk , tk+1 ) (22) cV1 (t)
V̇ (t) ≤ − + 2β(t), ∀t ∈ [0, Tp )
tk+1 = inf{t > tk | |d(t)| > γ|u(t)| + (t)} (23) Tp − t

where u(t) = υ(tk ) is the controller, d(t) = υ(t) − u(t) stands for the V̇ (t) ≤ −cςV1 (t) + 2β(t), ∀t ∈ [Tp , +∞) (27)
event-triggered error, 0 < γ < 1 is a design parameter, tk , k ∈ N + , is
where c = 2 min{σi , i = 1, . . . , n} > 2. Then, from (27) and
the update time of the controller with t1 = 0 and υ(t) is given as
Lemma 3, it is easy to verify that V (t) is bounded for all t ∈ [0, Tp )
(1 + γ)zn αn 2
zn 2 (t) and limt→Tp− V1 (t) = 0. Therefore, zi , θ̃, and θ̂ are bounded for all
υ(t) = −  − (24)
2 + β 2 (t)
zn2 αn zn2 2 (t) + β 2 (t) t ∈ [0, Tp ) and limt→Tp− zi (t) = 0.
Before continuing the following analysis, an indispensable proposi-
where β(t) = m1 e−m2 μ(t) , mi , i = 1, 2, is a positive design parame- tion is given below. See the Appendix for its proof.
ter, and (t) is a positive monotonic decreasing differentiable function Proposition 1: For the coordinate transformation (14), under
with (t) = 0∀t ≥ Tp . When (23) is triggered, the controller u(t) will the virtual controller with the adaptive law (15)–(16), one has
be updated to υ(tk+1 ), and keep this value until the next trigger. limt→Tp− αi−1 (t) = 0 and limt→Tp− ξi (t) = 0, i = 2, . . . , n.
In the interval [tk , tk+1 ), from the trigger condition described in ˙
(23), we can obtain that |υ(t) − u(t)| ≤ γ|u(t)| + (t). Therefore, From the Proposition 1, the adaptive law θ̂ in (15) and υ(t) in (25),
˙
there exist two continuous time-varying parameters 1 (t) and 2 (t), it is not hard to get that limt→Tp− θ̂(t) = 0 and limt→Tp− υ(t) = 0. With
satisfying i (tk ) = 0, i (tk+1 ) = ±1 and |i (t)| ≤ 1, ∀t ∈ [tk , tk+1 ), the aid of the existence and continuation properties of the solution
i = 1, 2, such that ξ(t), we can get that ξ(Tp ) = 0. According to the event-triggering
mechanism (23), one can get u(Tp ) = 0. Further, by the fact that fi (·)
υ(t) 2 (t)(t) and ϕi (·) vanish at the origin, and μ(t) = ς ∀t ∈ [Tp , +∞), we can
u(t) = − (25)
1 + γ1 (t) 1 + γ1 (t) conclude that ξ(t) = 0 and υ(t) = 0 ∀t ≥ Tp . Combined with the
which is of great significance to the subsequent stability analysis, and event-triggering mechanism (23), we can easily deduce that u(t) = 0
we only pay attention to the existence of i (t) without requiring its ∀t ≥ Tp . That is, system states converge to zero in the PT Tp .
concrete form. Now, we show that the Zeno behavior cannot occur under the
Remark 4: Under the event-triggered controller (22) and event- event-triggering mechanism (23). First, we define t∗k = tk+1 − tk as
triggering mechanism tk+1 = inf{t > tk | |d(t)| > γ|u(t)| + } with the trigger time interval. From (23), one has
the positive constant , Zhang and Yang [20] and Hua et al. [21] can
d(t) = υ(t) − u(t) = υ(t) − υ(tk ) ∀t ∈ [tk , tk+1 ). (28)
render system states to zero in an FT interval. Unfortunately, they cannot
ensure that the controller u(t) is updated to zero in time when the system If t∗k → 0, it can be deduced from (28) and the continuation property
return to the origin. This means that the closed-loop system cannot be of υ(t) that
maintained at the origin. The new ETC design of this article can avoid
this problem and actually achieve the PT stability by introducing the lim |d(tk + t∗k )| = lim |υ(tk + t∗k ) − υ(tk )| = 0. (29)
t∗k →0 ∗ tk →0
continuous time-varying function (t) in the event-triggering mecha-
nism.
Then, from the trigger condition in (23), we can deduce that
Remark 5: The proposed ETC scheme with event-triggering mech-
anism tk+1 = inf{t > tk | |d(t)| ≥ γ|u(t)| + } ( is a constant) and |d(tk+1 )| > γ|u(tk )| + (tk+1 ) ≥ 0. (30)
the signal υ(t) in [16], [17], and [18] can only partially compensate for
the event-triggered error d(t) and achieve bounded stability. Although By comparison, it is easy to find that (29) and (30) are contradictory,
the control methods proposed in [20] and [21] can completely eliminate so the hypothesis t∗k → 0 cannot hold, which implies that t∗k cannot

Authorized licensed use limited to: Thapar Institute of Engineering & Technology. Downloaded on January 12,2024 at 07:43:29 UTC from IEEE Xplore. Restrictions apply.
IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 68, NO. 11, NOVEMBER 2023 6979

approach to zero, i.e., t∗k ≥ t̄ with the positive constant t̄. That is, the
Zeno behavior cannot occur. This proof is finished. 
Remark 6: If the output of (13) is defined as y = ξ1 , by constructing
the following state transformation:
 
1 λ1 y(t) λ2 y(t)
z1 = + (31)
2 λ1 (t) − y λ2 (t) + y
our control scheme can be directly used to solve the event-based PT sta-
bilization problem with output constraints −λ2 (t) < y < −λ1 (t),
where λi , i = 1, 2, is a positive design parameter, (t) is a performance Fig. 1. Spring-damping mechanical system schematic diagram.
function, and can be chosen as (t) = 0 e−kt + ∞ with the design
parameters 0 > 0, k > 0, and ∞ ≥ 0. Different from the performance
functions given in [25], [26], [27], [28], [29], [30], [31], [32], and [33],
where the performance parameter ∞ is required to satisfy ∞ > 0, our
proposed performance function can set ∞ = 0, which means that the
steady-state requirement is limt→+∞ y(t) = 0 instead of converging to
a specific range, since the system output can be rendered to zero in a
preassigned time interval. In addition, the similar transformations can
be found in [25], [26], [27], [28], [29], [30], and [31], but they are unable
to deal with the asymmetric constrained and unconstrained problems
simultaneously without changing the control structure. Although [32]
and [33] have solved this problem, they can only realize that system Fig. 2. Responses of ξ1 under different initial conditions. (a) General
states coverage to the origin within a finite/fixed time, which is related scale. (b) Logarithmic scale.
to initial conditions and control parameters of system.
Remark 7: The systems considered by [1], [2], [3], [4], and [5] were
integrator chains with uncertainties and/or disturbances matched with
control input. Although [6] and [8] considered general cascade systems
that contained state-dependent uncertainties in entire system dynamics,
where the unknown nonlinear terms need to satisfy the known linear
growth conditions. In this article, the strict-feedback nonlinear systems
with a lower-triangular structure are considered, and each differential
term contains a state-dependent nonlinearity with uncertain parameters.
Remark 8: For the system (13), the stabilization and tracking control
problems have been studied by many works, such as [34], [35], [36],
[37], [38], [39], [40], [41], and [42]. However, most of the research Fig. 3. Responses of ξ2 and u under different initial conditions. (a) ξ2 .
results can only achieve bounded or FT stability. Different from these (b) u.
results, the control algorithm proposed in this article can steer the
system states converge to the origin within a specified time, which
can be preset by the designer according to actual requirements. By defining ξ1 = M s, ξ2 = M ṡ, u = F , the system (32) can be
Remark 9: By introducing the time-varying function μ(t) = rewritten as
Tpm+n
with the positive integers m and n to scale all system
(Tp −t)m+n ξ˙1 = ξ2 , ξ˙2 = u + f2 (ξ1 , ξ2 ) (33)
states, the control methods proposed in [1], [2], [3], [4], [5], [6], and
ξ1
[7] can achieve the PT stability, but the computational burden for the where f2 (ξ1 , ξ2 ) = − M1
(kξ1 e− M + θξ2 ) is a smooth function.
derivative of μ(t) is largely increased. Then, a new nonscaling design In this simulation, the system parameters are taken as M = 1, k =
framework was proposed in [14] to reduce the computational burden, 1, θ = 0.5, and the design parameters are selected as σ1 = 3, σ2 =
T
which introduces μ(t)δi with μ(t) = ( Tpp−t )m and δi = 3 · 5i−2 into 2, m1 = 5, m2 = 3, Tp = 3, and γ = 0.5. The positive monotonic
the virtual controller. Compared to [14], our work can further reduce decreasing differentiable function (t) is selected as
computational burden and save control effort, because the power of the 
time-varying function μ(t) can equal to one, i.e., μ(t) = (Tp 1−t)k with 0.002(Tp − t), t ∈ [0, Tp )
(t) =
k = 1. 0, t ≥ Tp
and the system initial conditions are given by three different values
IV. SIMULATION EXAMPLES (ξ1 (0), ξ2 (0)) = (1.5, 3.5), (2.0, 3.0), (2.5, 3.0). The initial value of
Consider the networked-based spring-damping mechanical system adaptive parameter is chosen as θ̂(0) = 0. The simulation results are
shown in Fig. 1. According to the Newton’s law, the kinematic model shown in Figs. 2–4.
is characterized as follows: The responses of ξ1 with general scale and logarithmic scale under
different initial conditions are plotted in Fig. 2, from which we can find
M s̈ = F − kse−s − θṡ (32)
that ξ1 converges to zero within the PT Tp . The responses of ξ2 and u
where s, ṡ, s̈, and M denote position, velocity, acceleration, and mass of under different initial conditions are plotted in Fig. 3, from which we
the trolley, respectively; k stands for the elastic coefficient of the spring; can see that ξ2 and u converge to zero within the PT Tp . The triggering
θ is the uncertain viscosity coefficient of the damper; F represents the instants and the responses of θ̂ are shown in Fig. 4, it can be seen that
external force. the number of trigger events is limited and θ̂ is bounded.

Authorized licensed use limited to: Thapar Institute of Engineering & Technology. Downloaded on January 12,2024 at 07:43:29 UTC from IEEE Xplore. Restrictions apply.
6980 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 68, NO. 11, NOVEMBER 2023

[2] Y. Song, Y. Wang, and M. Krstic, “Time-varying feedback for stabilization


in prescribed finite time,” Int. J. Robust Nonlinear Control, vol. 29, no. 3,
pp. 618–633, 2019.
[3] D. Steeves, M. Krstic, and R. Vazquez, “Prescribed-time estimation and
output regulation of the linearized Schrödinger equation by backstepping,”
Eur. J. Control, vol. 55, pp. 3–13, 2020.
[4] J. Holloway and M. Krstic, “Prescribed-time observers for linear systems
in observer canonical form,” IEEE Trans. Autom. Control, vol. 64, no. 9,
pp. 3905–3912, Sep. 2019.
[5] J. Holloway and M. Krstic, “Prescribed-time output feedback for linear
systems in controllable canonical form,” Automatica. vol. 107, pp. 77–85,
2019.
Fig. 4. Responses of tk and θ̂ under different initial conditions. (a) tk . [6] F. Gao, Y. Wu, and Z. Zhang, “Global fixed-time stabilization of
(b) θ̂. switched nonlinear systems: A time-varying scaling transformation ap-
proach,” IEEE Trans. Circuits Syst. II, Express Briefs, vol. 66, no. 11,
V. CONCLUSION pp. 1890–1894, Nov. 2019.
[7] P. Ning, C. Hua, K. Li, and R. Meng, “Adaptive fixed-time control for
In this article, we study the ETC problem for a class of nonlinear uncertain nonlinear cascade systems by dynamic feedback,” IEEE Trans.
uncertain systems via a PT approach. First, by utilizing the contradiction Syst., Man, Cybern. Syst., doi: 10.1109/TSMC.2022.3218599.
[8] P. Krishnamurthy, F. Khorrami, and M. Krstic, “A dynamic
method and the limit convergence of abnormal integral lemma, we
high-gain design for prescribed-time regulation of nonlinear
propose a new PT stability criterion, which is more general than the systems,” Automatica, vol. 115, May 2020, Art. no. 108860,
existing stability criteria. Then, a new approach for co-designing the doi: 10.1016/j.automatica.2020.108860.
event-triggering mechanism and controller is put forward to completely [9] P. Krishnamurthy, F. Khorrami, and M. Krstic, “Robust adaptive
compensate the event-triggered error. Finally, an event-triggered con- prescribed-time stabilization via output feedback for uncertain nonlinear
strict-feedback-like systems,” Eur. J. Control, vol. 55, pp. 14–23, 2020.
troller is obtained in the form of time-varying feedback, which can [10] Y. Orlov, Nonsmooth Lyapunov Analysis in Finite and Infinite Dimensions.
render all system states to zero in the preset time interval. In future, we Berlin, Germany: Springer, 2020.
will focus on the PT stabilization problem for stochastic systems. [11] B. Zhou, “Finite-time stabilization of linear systems by bounded linear
time-varying feedback,” Automatica, vol. 113, Mar. 2020, Art. no. 108760,
doi: 10.1016/j.automatica. 2019.108760.
APPENDIX [12] B. Zhou, “Finite-time stability analysis and stabilization by bounded linear
The proof of Proposition 1 is given below. time-varying feedback,” Automatica, vol. 121, Nov. 2020, Art. no. 109191,
doi 10.1016/j.automatica.2020.109191.
Proof: From the definition of V in (20), and the inequality (27), we [13] B. Zhou and Y. Shi, “Prescribed-time stabilization of a class of
can derive that V̇ ≤ − TcV
p −t
+ TcV 2
p −t
+ 2β(t). According to the compar- nonlinear systems by linear time-varying feedback,” IEEE Trans.
cV2 Autom. Control, vol. 66, no. 12, pp. 6123–6130, Dec. 2021,
ison lemma, if Proposition 1 holds for V̇ = − TcV
p −t
+ Tp −t
+ 2β(t),
doi: 10.1109/TAC.2021.3061645.
then Proposition 1 also holds for V̇ < − TcV
p −t
+ TcV 2
p −t
+ 2β(t). There- [14] W. Li and M. Krstic, “Stochastic nonlinear prescribed-time stabilization
cV2 and inverse optimality,” IEEE Trans. Autom. Control, vol. 67, no. 3,
fore, we only prove Proposition 1 for V̇ = − TcV
p −t
+ Tp −t
+ 2β(t). By pp. 1179–1193, Mar. 2022, doi: 10.1109/TAC.2021.3061646.
utilizing the Gronwall’s lemma, some calculations show that [15] C. Hua, P. Ning, and K. Li, “Adaptive prescribed-time control for a class of
 c  t uncertain nonlinear systems,” IEEE Trans. Autom. Control, vol. 67, no. 11,
t c V̇2 (s) pp. 6159–6166, Nov. 2022.
V = 1− V1 (0) − (Tp − t) ds
Tp 0 (T p − s)
c [16] L. Xing, C. Wen, Z. Liu, H. Su, and J. Cai, “Event-triggered adaptive
control for a class of uncertain nonlinear systems,” IEEE Trans. Autom.
 t
2β(s) Control, vol. 62, no. 4, pp. 2071–2076, Apr. 2017.
+ V2 (t) + (Tp − t)c ds (34) [17] L. Xing, C. Wen, Z. Liu, H. Su, and J. Cai, “Event-triggered output
0 (T p − s)
c
feedback control for a class of uncertain nonlinear systems,” IEEE Trans.
Autom. Control, vol. 64, no. 1, pp. 290–297, Jan. 2019.
and we can directly get that [18] J. Huang, W. Wang, C. Wen, and G. Li, “Adaptive event-triggered
 c  t control of nonlinear systems with controller and parameter estimator
t V̇2 (s)
V1 = 1 − V1 (0) − (Tp − t)c ds triggering,” IEEE Trans. Autom. Control, vol. 65, no. 1, pp. 318–324,
0 (Tp − s)
Tp c Jan. 2020.
 t [19] C. Wang, C. Wen, and Q. Hu, “Event-triggered adaptive control for
2β(s) a class of nonlinear systems with unknown control direction and sen-
+ (Tp − t)c ds (35)
0 (Tp − s)
c sor faults,” IEEE Trans. Autom. Control, vol. 65, no. 2, pp. 763–770,
Feb. 2020.
m2 [20] C. Zhang and G. Yang, “Event-triggered global finite-time control for a
− T −t
where β(t) = m1 e p . Due to limt→Tp− (Tp − t)p (Tβ(t)
−t)c
= 0 for class of uncertain nonlinear systems,” IEEE Trans. Autom. Control, vol. 65,
 Tp
any p ∈ (0, 1), it can be deduced from Lemma 1 that 0 p (T2β(s) p −s)
c ds
no. 3, pp. 1340–1347, Mar. 2020.
[21] C. Hua, Q. Li, and K. Li, “Event-based finite-time control for high-order
is convergent and limt→Tp− V1 (t) = limt→Tp− (Tp − t)V̇2 = 0. In other interconnected nonlinear systems with asymmetric output constraints,”
˙ IEEE Trans. Autom. Control, vol. 67, no. 11, pp. 6135–6142, Nov. 2022,
words, limt→Tp− (Tp − t)θ̂ = 0. Then, similar to the proof of Theo- doi: 10.1109/TAC.2021.3128471.
rem 2 in our previous work [15], we can derive that limt→Tp− αi−1 = 0 [22] M. Li, J. Guo, and Z. Xiang, “Global adaptive finite-time stabilization for
a class of p-normal nonlinear systems via an event-triggered strategy,” Int.
and limt→Tp− ξi = 0, i = 2, . . . , n. This proof is completed.  J. Robust Nonlinear Control, vol. 30, no. 10, pp. 4059–4074, 2020.
[23] W. Sun, J. Zhao, W. Sun, J. Xia, and Z. Sun, “Adaptive event-triggered
REFERENCES global fast finite-time control for a class of uncertain nonlinear systems,”
Int. J. Robust Nonlinear Control, vol. 30, no. 9, pp. 3773–3785, 2020.
[1] Y. Song, Y. Wang, J. Holloway, and M. Krstic, “Time-varying feedback [24] Department of Mathematics, Tongji University, Advanced Mathemat-
for regulation of normal-form nonlinear systems in prescribed finite time,” ics, 6th ed. (in Chinese). Beijing, China: Higher Educ. Press, 2007,
Automatica, vol. 83, pp. 243–251, 2017. pp. 264–265.

Authorized licensed use limited to: Thapar Institute of Engineering & Technology. Downloaded on January 12,2024 at 07:43:29 UTC from IEEE Xplore. Restrictions apply.
IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 68, NO. 11, NOVEMBER 2023 6981

[25] Y. Liu, X. Liu, Y. Jing, and Z. Zhang, “A novel finite-time adaptive fuzzy [34] M. Krstic, I. Kanellakopoulos, and P. Kokotovic, “Adaptive nonlinear
tracking control scheme for nonstrict feedback systems,” IEEE Trans. control without overparametrization,” Syst. Control Lett., vol. 19, no. 3,
Fuzzy Syst., vol. 27, no. 4, pp. 646–658, Apr. 2019. pp. 177–185, 1992.
[26] C. Hua, P. Ning, K. Li, and X. Guan, “Fixed-time prescribed tracking [35] I. Kanellakopoulos, P. Kokotovic, and A. Morse, “Systematic design
control for stochastic nonlinear systems with unknown measurement sen- of adaptive controllers for feedback linearizable systems,” IEEE Trans.
sitivity,” IEEE Trans. Cybern.. vol. 52, no. 5, pp. 3722–3732, May 2022, Autom. Control, vol. 36, no. 11, pp. 1241–1253, Nov. 1991.
doi: 10.1109/TCYB.2020.3012560. [36] D. Seto, A. Annaswamy, and J. Baillieul, “Adaptive control of nonlinear
[27] T. Berger, H. Le, and T. Reis, “Funnel control for nonlinear systems with systems with a triangular structure,” IEEE Trans. Autom. Control, vol. 39,
known strict relative degree,” Automatica, vol. 87, pp. 345–357, 2018. no. 7, pp. 1411–1428, Jul. 1994.
[28] C. Hua, K. Li, and X. Guan, “Event-based dynamic output feedback [37] J. Zhou, C. Wen, and G. Yang, “Adaptive backstepping stabi-
adaptive fuzzy control for stochastic nonlinear systems,” IEEE Trans. lization of nonlinear uncertain systems with quantized input sig-
Fuzzy Syst., vol. 26, no. 5, pp. 3004–3015, Oct. 2018. nal,” IEEE Trans. Autom. Control, vol. 59, np. 2, pp. 460–464,
[29] J. X. Xu and X. Jin, “State-constrained iterative learning control for a Feb. 2014.
class of MIMO systems,” IEEE Trans. Autom. Control, vol. 58, no. 5, [38] X. Yu and Y. Lin, “Adaptive backstepping quantized control for a class
pp. 1322–1327, May 2013. of nonlinear systems,” IEEE Trans. Autom. Control, vol. 62, no. 2,
[30] Y. Song, Y. Wang, and C. Wen, “Adaptive fault-tolerant PI tracking control pp. 981–985, Feb. 2017.
with guaranteed transient and steady-state performance,” IEEE Trans. [39] J. Zhou, C. Wen, and W. Wang, “Adaptive control of uncertain nonlinear
Autom. Control, vol. 62, no. 1, pp. 481–487, Jan. 2017. systems with quantized input signal,” Automatica, vol. 95, pp. 152–162,
[31] Y. J. Liu and S. Tong, “Barrier Lyapunov functions for Nussbaum gain 2018.
adaptive control of full state constrained nonlinear systems,” Automatica, [40] Y. Li, “Finite time command filtered adaptive fault tolerant control for a
vol. 76, pp. 143–152, 2017. class of uncertain nonlinear systems,” Automatica, vol. 106, pp. 117–123,
[32] Z. Sun, Y. Peng, C. Wen, and C. Chen, “Fast finite-time adaptive sta- 2019.
bilization of high-order uncertain nonlinear system with an asymmetric [41] Y. Li and G. Yang, “Adaptive asymptotic tracking control of uncertain
output constraint,” Automatica, vol. 121, Nov. 2020, Art. no. 109170, nonlinear systems with input quantization and actuator faults,” Automatica,
doi: 10.1016/j.automatica.2020.109170. vol. 72, pp. 177–185, 2016.
[33] X. Jin, “Adaptive fixed-time control for MIMO nonlinear systems with [42] G. Lai, Z. Liu, Y. Zhang, C. Chen, and S. Xie, “Adaptive backstepping-
asymmetric output constraints using universal barrier functions,” IEEE based tracking control of a class of uncertain switched nonlinear systems,”
Trans. Autom. Control, vol. 64, no. 7, pp. 3046–3053, Jul. 2019. Automatica, vol. 91, pp. 301–310, 2018.

Authorized licensed use limited to: Thapar Institute of Engineering & Technology. Downloaded on January 12,2024 at 07:43:29 UTC from IEEE Xplore. Restrictions apply.

You might also like