4102 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS: SYSTEMS, VOL. 53, NO.

7, JULY 2023

Prescribed-Time Control and Its

Latest Developments
Yongduan Song , Fellow, IEEE, Hefu Ye , and Frank L. Lewis , Life Fellow, IEEE

Abstract—Prescribed-time (PT) control for nonlinear systems, history of FT control theory reveals that the related concepts
originated from Song et al., has gained increasing attention can be traced back to the 1960s, when the concept of FT sta-
among the control community. The salient feature of PT con- bility was proposed in [61], [62], and [63], with application
trol lies in its ability to achieve system stability within a finite
settling time user-assignable in advance irrespective of initial to certain simple linear systems. A system is considered to
conditions. It is such a unique feature that has enticed many be FT stable if, given a bounded initial condition, all closed-
follow-up studies on this technically important area, motivating loop signals are bounded and its states converge to zero (or a
numerous research advancements. In this article, we provide a residual set) within a specified FT interval. It is essential to
comprehensive survey on the recent developments in PT control. identify FT stability and asymptotic stability. In fact, a system
Through a concise introduction to the concept of PT control, and
a unique taxonomy covering: 1) from robust PT control to adap- can be FT stable but not asymptotically stable, and vice versa.
tive PT control; 2) from PT control for single-input–single-output Asymptotic stability corresponds to the behavior of a system
(SISO) systems to multi-input–multioutput (MIMO) systems; and within a sufficiently long (in principle, infinite) time interval,
3) from PT control for an isolated system to multiagent systems, while FT stability is a more practical concept that helps to
we present an accessible review of this interesting topic. We high- study the behavior of the closed-loop system over a finite
light key techniques, and fundamental assumptions adopted in
various developments as well as some new design ideas. We also (possibly short) time interval, and hence it finds application
discuss several possible future research directions toward PT whenever it is desired/required that the system states shrink to
control. a certain small threshold (for example, to avoid saturation or
Index Terms—Finite-time (FT) control, prescribed-time (PT) the excitation of nonlinear dynamics) within a short period of
control, state scaling, time scaling, time-varying feedback. time.
The past few decades have witnessed much progress in
FT control of dynamic systems [64], [65], [66], [67], [68],
I. I NTRODUCTION [69], [70], [71], [72], [73], [74], [75], [76]. The homogeneous
approach, terminal sliding mode, higher-order sliding modes,
HE NOTION of prescribed-time (PT) control for nonlin-
T ear systems, pioneered by Song et al. [1], has brought
much vitality to finite-time (FT) control, attracting increasing
and adding a power integrator are suggested sequentially in
an attempt to achieve FT stability for high-order nonlinear
systems [77], [78], [79], [80], [81], [82]. Although conver-
attention from the control community and motivating numer- gence may be pursued in a finite time, estimation of the
ous follow-up studies on this important field during the past settling time relies explicitly on initial conditions. This may
few years (see [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], limit the application scope of those existing results when little
[11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], knowledge of plant initial states are accessible. Later on, a
[22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], notion termed fixed-time (FxT) control [83], [84], [85], [86],
[33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [87], [88], [89], [90], [91], [92] has emerged, which employs
[44], [45], [46], [47], [48], [49], [50], [51], [52], [53], [54], odd-order plus fractional-order feedback to provide various
[55], [56], [57], [58], [59], [60]). Examining the development closed-loop system dynamics. The upper bound of the settling
time can be estimated without using any information on initial
Manuscript received 15 October 2022; accepted 23 January 2023. Date
of publication 15 February 2023; date of current version 16 June 2023. conditions.
This work was supported in part by the National Key Research and Despite the benefits of FxT control in the light of settling
Development Program of China under Grant 2022YFB4701400/4701401; time estimation, no simple and obvious relationship exists
in part by the National Natural Science Foundation of China under
Grant 61991400, Grant 61991403, Grant 62250710167, Grant 61860206008, between the control parameters and the intended upper bound
Grant 61933012, and Grant 62273064; and in part by the China Scholarship of the settling time. The settling time under the FxT con-
Council. This article was recommended by Associate Editor H. Zhang. trol is often overestimated, which may be hundreds or even
(Corresponding author: Yongduan Song.)
Yongduan Song and Hefu Ye are with the Chongqing Key Laboratory thousands of times larger than the true settling time, resulting
of Autonomous Systems, Institute of Artificial Intelligence, School of in an inaccurate description of system performance. On the
Automation, Chongqing University, Chongqing 400044, China, and also with other hand, the settling time is not a directly tunable parame-
the Star Institute of Intelligent Systems, Robotic Systems Lab, Chongqing
400044, China (e-mail: (e-mail: ydsong@cqu.edu.cn; yehefu@cqu.edu.cn). ter for either FT control or FxT control, as it also depends on
Frank L. Lewis is with the UTA Research Institute, University of Texas at other controller design parameters. To alleviate the problem
Arlington, Fort Worth, TX 76118 USA (e-mail: lewis@uta.edu). of overestimation of the settling time while alleviating the
Color versions of one or more figures in this article are available at
https://doi.org/10.1109/TSMC.2023.3240751. dependence of the settling time on design parameters, the
Digital Object Identifier 10.1109/TSMC.2023.3240751 predefined-time (PdT) control approach is exploited in [93],
2168-2216 
c 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: Xian Jiaotong University. Downloaded on June 27,2025 at 10:09:34 UTC from IEEE Xplore. Restrictions apply.
SONG et al.: PT CONTROL AND ITS LATEST DEVELOPMENTS 4103

[94], [95], and [96], where the least upper bound of the set-
tling time can be preset irrespective of initial conditions and
any other design parameter.
Recently, the classical idea that originated in strategic and
tactical missile guidance applications [97], [98], [99] has been
revisited and further applied to high-order nonlinear systems,
namely, PT control, which inherits the advantages of FT con-
trol, FxT control, PdT control, and also allows for presetting
the settling time precisely. This concept is of great importance
in many practical engineering applications where transient
processes must occur within a given time (e.g., missile guid-
ance, multiagent rendezvous, emergency braking, and obstacle Fig. 1. Organization of this article.
avoidance in robotic systems, etc.).
More importantly, the PT control is promising since it is
robust to external disturbances, the control input is always Definition 1 [108, Ch. 4]: For a nonautonomous system as
smooth over the transient process, and there is no need for any ẋ = f (x, t) (1)
information on the upper bound of the nonvanishing pertur-
bations in the control design. The key technical design steps where f : Rn × [0, ∞) → Rn is piecewise continuous in t
for PT control include: converting the original system to a and locally Lipschitz in x. The equilibrium point x = 0 is as
new system by a time-varying transformation (including state follows.
scaling, time scaling, and some other technologies), dealing 1) Stable, if there exists a class of K function β such that
with matching/mismatching uncertainties and unknown con- x(t) ≤ β(x(0)).
trol coefficients to construct appropriate Lyapunov inequalities, 2) Asymptotically stable,1 if it is stable, and x(t) →
and selecting the appropriate control gain k to prove the bound- 0 as t → ∞.
edness of all closed-loop signals, especially the boundedness 3) Exponentially stable, if there exist two positive numbers,
of the control inputs. Furthermore, because all real PT con- λ1 and λ2 , such that for sufficiently small x(0), x(t) ≤
trollers have infinite gain characteristics as time tends to the λ1 x(0)e−λ2 t ∀t ≥ 0.
preset time, they are limited to usage within a finite time The definitions of FT and FxT stability are stated below.
interval. Many infinite time controllers may be integrated with Definition 2 [83, Ch. 4]: For a system as (1), the equilib-
PT algorithms to deliver their infinite time features inside a PT rium point x = 0 is as follow.
window, thereby extending the use of PT control systems. In 1) FT stable, if it is stable and there exists a x(0)-dependent
this article, we perform a complete study on several important settling time function T(x(0)) such that x(t) = 0 for t ≥
theoretical breakthroughs, key technical concerns, and poten- T(x(0)).
tial research problems in PT control, as well as provide a 2) FxT stable, if it is stable and the settling time function
comprehensive literature survey. T(x(0)) is upper bounded on R, i.e., ∃ Tmax > 0, x(t) =
The study will start next in Section II with an overview 0 for t ≥ Tmax .
of some basic propositions of FT/FxT/PdT and PT control. Obviously, the terminal time always attaches itself to x(0)
Section III lists some specific literature on PT control for in FT control, such attachment is however removed in FxT
single-input–single-output (SISO) systems and provides some control. An astonishing scenario in FT/FxT stability is the PT
basic design ideas of prescribed robust and/or adaptive con- stability, where the terminal time has nothing to do with initial
troller design, focused on the introduction of state scaling tech- condition, and thus can be user-set freely in advance.
nology and time scaling technology on PT control. Section IV
lists some state-of-the-art results on multi-input–multi-output B. Propositions on Finite-/Fixed-/Predefined-/Prescribed-
(MIMO) systems and presents a detailed demonstration of PT Time Stability
control for this type of system. It covers square and nonsquare
Achieving FT stability for dynamic systems is of spe-
MIMO systems. Section V lists some interesting studies on PT
cial theoretical and practical interest. The typical approach
distributed control and addresses some basic issues of PT con-
for establishing FT stability is to derive Lyapunov differen-
trol for multiagent systems. The organization of Sections II–V
tial inequalities. Most of these inequalities can be found in
is shown in Fig. 1. Section VI provides some connections
the following works which are summarized as a variety of
between FT and PT control and also discusses some possible
propositions.
open areas of research.
Proposition 1 [78]: For system ẋ = f (x, t), if there exists
a C 1 function V(x) ≥ 0 such that V̇(x) ≤ −kV q (x), where
II. P RELIMINARIES k > 0, 0 < q < 1, then the closed-loop system is FT stable
A. Definitions and the settling time is calculated by
We first consider some basic definitions of infinite-time 1 If x(0) ∈ (−∞, +∞), here we say that the equilibrium point is global
(asymptotic/exponential) stability. asymptotically stable.

Authorized licensed use limited to: Xian Jiaotong University. Downloaded on June 27,2025 at 10:09:34 UTC from IEEE Xplore. Restrictions apply.
4104 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS: SYSTEMS, VOL. 53, NO. 7, JULY 2023

1
T := V 1−q (x(0)). and p, q, m and n are all odd integers, then the closed-loop
k(1 − q) system is FxT stable and the settling time is bounded by
Proposition 2 [69]: For system ẋ = f (x, t), if there exists 1 n 1 q
a C 1 function V(x) ≥ 0 such that V̇(x) ≤ −k1 V q (x) + k2 V(x) T := + .
k1 m − n k2 q − p
where k1 > 0, k2 > 0, and 0 < q < 1, then the closed-
loop system is semi-global FT stable and the settling time is Proposition 9 [93]: For system ẋ = f (x, t), if there exists
calculated by a C 1 function V(x) ≥ 0 such that
  1 V p (x) 1−p
1 k1 1−q V̇(x) ≤ − V (x)
T := ln 1 − V (x(0)) . e (2)
k2 (1 − q) k2 pTp

Proposition 3 [75]: For system ẋ = f (x, t), if there exists where Tp > 0, 0 < p ≤ 1, then the closed-loop system is
a C 1 function V(x) ≥ 0 such that V̇(x) ≤ −kV q (x) + η where weakly PdT stable; if the equal sign in (2) always holds, then
k > 0, 0 < η < ∞ and 0 < q < 1, then the closed-loop the closed-loop system is strongly PdT stable. The settling
system is practical semi-global FT stable and the settling time time is bounded by Tp .
is calculated by Proposition 10 [1]: Consider a time-varying function
⎛ ⎞ μ(t) = T/(T − t), if a C 1 function V : [0, T) → [0, +∞)
  1−q
1 ⎝V 1−q (x(0)) − η q
⎠
satisfies
T :=
kθ (1 − q) k(1 − θ ) μ(t)
V̇(t) ≤ −2kμ(t)V(t) + d(t)2 (3)
4θ
where 0 < θ < 1 is a constant.
for unknown perturbation d(t) and positive numbers k, θ , then
Proposition 4 [76]: For system ẋ = f (x, t), if there exists a
V(t) is bounded for t ∈ [0, T).
C 1 function V(x) ≥ 0 such that V̇(x) ≤ −k1 V q (x)−k2 V(x)+η
Proposition 11: Consider a time-varying function μ(t) =
where k1 > 0, k2 > 0, 0 < η < ∞ and 0 < q < 1, then the
T/(T − t), if a C 1 positive function V : [0, T) → [0, +∞)
closed-loop system is practical FT stable and the settling time
satisfies
is calculated by
⎧ ⎫ V̇(t) ≤ −kμ(t)V(t) + |d(t)| (4)
⎨ ln k2 θV k(x(0))+k1 k2 V 1−q (x(0))+θk1 ⎬
1−q

1
ln θk1
T := max , for unknown perturbation d(t) and a positive number k, then
⎩ k2 θ (1 − q) k2 (1 − q) ⎭
V(t) is bounded for t ∈ [0, T) and limt→T V(t) = 0.
where 0 < θ < 1 is a constant. Proof: Solving the differential inequality (4) gives
Proposition 5 [84]: For system ẋ = f (x, t), if there exists a  t
−A(t) −A(t)
C 1 function V(x) ≥ 0 such that V̇(x) ≤ −(αV p (x) + βV q (x))k V(t) ≤ e V(0) + d̄e eA(s) ds (5)
0
where α > 0, β > 0, p > 0, q > 0, k > 0, and pk < t
1, qk > 1, then the closed-loop system is FxT stable and the where A(t) = 0 kμ(s)ds and d̄ = sups∈[0,t] |d(s)|. Since
settling time is bounded by  t = +∞ and Ȧ(t) = kμ(t), then limt→T V(t) =
limt→T A(t)
limt→T ( 0 eA(s) ds/eA(t) )d̄ = limt→T (eA(t) /eA(t) )(d̄/[kμ(t)]) =
1 1
T := + . 0.
α k (1 − pk) β k (qk − 1) Notice that Proposition 10 implies that the unbounded con-
Proposition 6 [83]: For system ẋ = f (x, t), if there exists trol gain ensures that the closed-loop trajectory is bounded
a C 1 function V(x) ≥ 0 such that V̇(x) ≤ −αV p (x) − βV q (x) under unbounded perturbations, which provides a guideline
where α > 0, β > 0, p = 1 − 1/(2γ ), q = 1 + 1/(2γ ), γ > for invoking state-scaling technique to design PT controllers
1, then the closed-loop system is FxT stable and the settling (see [1], [2], [4], [5]), and Proposition 11 implies that the
time is bounded by unbounded control gain ensures that the closed-loop trajec-
πγ tory converges to zero under bounded perturbations, which
T := √ . provides a theoretical basis for the application of L’Hôpital’s
αβ
Rule in stability analysis (see [60]).
Proposition 7 [86]: For system ẋ = f (x, t), if there exists The connections and differences among the aforemen-
a C 1 function V(x) ≥ 0 such that V̇(x) ≤ −αV 2−(p/q) (x) − tioned stability notions are conceptually highlighted in Fig. 2.
βV p/q (x) where α > 0, β > 0, q > p > 0 and both q and Basically, PT stability is most “desirable”, which covers PdT
p are odd integers, then the closed-loop system is FxT stable stability, FxT stability, and of course FT stability, and FxT sta-
and the settling time is bounded by bility implies FT stability, but the reverse does not necessarily
qπ hold.
T := √ . Propositions 1–4 indicate that the terminal time T attaches
2 αβ(q − p)
itself to several design parameters (e.g., k1 , k2 , q, etc.), and
Proposition 8 [87]: For system ẋ = f (x, t), if there exists the initial system state V(x(0)). Propositions 5–8 show that
a C 1 function V(x) ≥ 0 such that V̇(x) ≤ −k1 V m/n (x) − the settling time T is bounded by a computable function,
k2 V p/q (x) where k1 > 0, k2 > 0, q > p > 0, m > n > 0 which is independent of the initial condition V(x(0)), whereas

Authorized licensed use limited to: Xian Jiaotong University. Downloaded on June 27,2025 at 10:09:34 UTC from IEEE Xplore. Restrictions apply.
SONG et al.: PT CONTROL AND ITS LATEST DEVELOPMENTS 4105

TABLE I
R ELATED P ROPOSITIONS ON FT, F X T, P D T, AND PT C ONTROL (T HE M EANING OF THE R ELATED PARAMETERS S EE P ROPOSITIONS 1–11)

Definition 4 [1]: The system ẋ = f (x, t, d) (d represents

nonvanishing perturbations) is PT globally uniformly asymp-
totically stable in time T if there exist class KL functions
β and βf , a class K function γ , and a time-varying function
μ(t) : [0, T) → R+ with μ(t) approaching to ∞ as t → T
such that
   
x(t) ≤ βf β(x(0), t) + γ d[0,t] , μ(t) ∀t ∈ [0, T). (7)

Clearly, Definition 4 is a generalization of Definition 3 in

the presence of nonvanishing perturbations in the system.
Fig. 2. Relationships among FT/FxT/PdT and PT control. Key features: FT To achieve PT convergence, two general PT controller
control—the settling time is related to initial conditions and design parameters;
FxT control—the settling time can be estimated by an upper bounded function design approaches, namely, the state scaling-based approach
independent of initial conditions; PdT control—the upper bound of the settling and the time scaling-based approach, are used in the literature.
time can be user-set freely; and PT control—the (exact) settling time can be 1) State Scaling: Using a monotonically increasing func-
user-assigned arbitrarily.
tion μ(t) that grows to infinity in finite time T to scale
Propositions 9–11 indicate that the settling time T can be pre- the state, thereby constructing a new variable z = μx.
set at users’ will irrespective of initial conditions and any other A control design that keeps z bounded will implicitly
design parameter. To close this section, we summarize the make x go to zero as t → T.
contents of Propositions 1–11 through Table I. 2) Time Scaling: Using a nonlinear temporal transforma-
tion τ = a(t) with a(t) being a function defined such
III. P RESCRIBED -T IME C ONTROL FOR SISO S YSTEMS that a(0) = 0 and limt→T a(t) = ∞. Since this time
scale transformation maps t ∈ [0, T) to τ ∈ [0, ∞). A
In this section, we focus on several fundamental topics
control design that achieves asymptotic convergence in
in PT control, such as robust control and adaptive control
the light of the time variable τ implicitly achieves PT
based on time-varying feedback, as well as some associated
convergence in terms of the time variable t.
technical concerns, including system state convergence, the
In the seminal work for nonlinear system PT stabilizing
boundedness of control input, and the boundedness of param-
design [1], a novel yet simple time-varying scaling of the form
eter estimations. Before this, we review the related concepts
on PT control. T
μ(t) = , t ∈ [0, T) (8)
T −t
A. Preliminaries on Prescribed-Time Control or its extended version is proposed, where T > 0 is the settling
Definition 3 [5]: System (1) is PT globally uniformly time specified by the designer.
asymptotically stable in time T if there exist a function The basis of time scaling-based design, originated
μ : [0, T) → R+ with μ increasing to ∞ as t → T and from [54], is a temporal axis mapping τ = a(t),2 with the
a class KL function β such that properties defined as follows. Let a (t) = (da/dt) and α(τ ) =

x(t) ≤ β(x(0), μ(t)) ∀t ∈ [0, T) (6) 2 A simple time scaling is t = T(1 − e−τ ) ⇔ τ = a(t) = ln T − ln(T − t).
In this case, a (t) = 1/(T − t), which becomes a particular case of the above
where T is a finite number that can be prescribed in the design. relation (8).

Authorized licensed use limited to: Xian Jiaotong University. Downloaded on June 27,2025 at 10:09:34 UTC from IEEE Xplore. Restrictions apply.
4106 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS: SYSTEMS, VOL. 53, NO. 7, JULY 2023

TABLE II
T ECHNICAL D IFFERENCES B ETWEEN D IFFERENT PT C ONTROL L ITERATURE

a (a−1 (τ )), i.e., α(τ ) is the function (da/dt) expressed in the integrator technique [15]; PT estimation and output regulation
light of the new time variable τ. Also, dτ = a (t)dt = α(τ )dt.3 of the linearized schrödinger equation [16]; PT stabilization
1) a(0) > 0 and a(T) = +∞. for stochastic nonlinear systems, where a nonscaling method
2) a(t) is continuously differentiable on t ∈ [0, T). is used [17], [18], [19]; PT control for nonlinear systems
3) a (t) > 0 and grows to infinite as t → T. within a liner decay rate [22]; PT control for normal-form
Here, we provide an overview on some typical works in PT systems, where Faà di Bruno’s formula and Bell polynomi-
control, most of which were initially for stabilization of SISO als are used [23]; frozen-time eigenvalues for PT-stabilized
systems within a preset time. Technical issues to be covered linear time-varying systems [20]; PT control via bounded time-
include controller structure, selection of time-varying func- varying feedback and parametric Lyapunov equation [24];
tions, convergence, robustness and performance, observers, parametric Lyapunov equation-based output feedback PT con-
and output feedback design, etc. trol [25]; bounded time-varying feedback-based PT control for
The well-known proportional navigation law in tactical and normal-form systems and satellite formation flying [26], [27];
strategic missile guidance (see, for instance, [97], [98]) is the PT control for p-normal nonlinear systems [28]; PT sliding
primary motivation for PT control. The early work address- mode control [102], [103], [104]; a general time transforma-
ing optimal PT control of linear time-invariant (LTI) systems tion for PT control [29]; adaptive PT control for strict-feedback
was reported in [99], [100], [101], which however, is dif- systems [30], [31]; PT differentiator and switched feedback-
ficult to be extended to high-order nonlinear systems. The based PT controller [32], [33]; PT stabilization of a perturbed
first work on PT control applicable to nonlinear systems is chain of integrators within the framework of time-varying
documented in [1], that is based on state scaling. Later on, homogeneity [34]; PT control with bounded time-varying con-
various PT control methods and extensions are proposed, trol gain [35], [36]4 ; PT control for affine systems and rigid
among which the most typical ones include: PT control via bodies [37]; PT and prescribed performance tracking con-
time base generator [106], [107]; super-exponential and PT trol for certain nonlinear systems [38]; practical PT control,
precise tracking control for normal-form systems [3]; PT namely, the output state/tracking error converges to a certain
observer [4] and output feedback design for linear time- set within a prescribed time [39], [40], [41].
invariant (LTI) systems based upon the separation princi- The representative results of the PT control for SISO
ple [5]; PT observer for LTI systems with measurement systems via time-varying feedback are summarized in Table II.
delay [6]; predictor-feedback PT stabilization of LTI systems Most of them are based on state feedback. Due to the diffi-
with input delay [7]; arbitrary time stabilization of integra- culties of designing complex uncertain systems, most results
tors via contraction analysis and an special exponential-like assume that the control coefficients (including the control
scaling [8]; PT stabilizing control for a full-scale 4-degrees- direction) of the system model are precisely known without
of-freedom permanent-magnet synchronous motor system via nonvanishing perturbations in the system. In addition, most
a modified exponential-like scaling [9]; PT stabilizing con- results consider only robust control schemes and do not con-
trol for LTI systems via a new modified exponential-like sider adaptive control schemes. Because in adaptive control,
scaling [10]; time scaling-based output feedback design for it is necessary to guarantee the boundedness of parameter
strict-feedback-like systems [11], [12]; PT stabilizing/tracking estimation (it seems to be difficult to do this with the state
control for strict-feedback-like systems via a dynamic gain scaling-based PT control approach) in addition to the bounded-
feedback design [13], [14]; PT stabilization via adding a power ness of the control signal, which usually poses a challenge for
the controller design. The following about robust and adaptive
3 Since both x(t) and x̆(τ ) relate to the value of the same signal at the PT control will be addressed.
same physical time point represented as t in the original time axis and τ in
the converted time axis, we use the notation x̆(τ ) to express a signal x(t) 4 A unified approach is proposed independently from [35] and [35] for FT
as a function of the transformed time variable τ , i.e., x(t) ≡ x̆(τ ). Hence, control and PT control where zero-error regulation is achieved before the
˙ ) = (dx̆/dτ ) = (dt/dτ )(dx/dt) = (1/α(τ ))(dx/dt) = (1/α(τ ))ẋ(t).
x̆(τ prescribed settling time.

Authorized licensed use limited to: Xian Jiaotong University. Downloaded on June 27,2025 at 10:09:34 UTC from IEEE Xplore. Restrictions apply.
SONG et al.: PT CONTROL AND ITS LATEST DEVELOPMENTS 4107

B. Robust Prescribed-Time Control C. Adaptive Prescribed-Time Control

In this section, we adopt the state scaling method to design a It is interesting yet challenging to develop adaptive control
control u(t) to stabilize a scalar system with unknown control schemes to regulate the system state to zero in a prescribed
coefficient and nonvanishing perturbation in prescribed time time. So far, the related results in this area are very lim-
T. Consider ited. The following sections present three basic frameworks of
adaptive PT control through a scalar system: (Section III-C1)
ẋ = b(x, t)u + f (x, t) (9) adaptive design for systems with time-invariant parameters;
where x and u are the state and the control input, respec- (Section III-C2) adaptive design for systems with time-varying
tively, b(x, t) and f (x, t) are nonlinear time-varying functions parameters; and (Section III-C3) adaptive Nussbaum gain
and satisfy the following assumptions. design for systems with time-varying parameters.
Assumption 1 [1]: The function f (x, t) is smooth and satis- We use the time scaling method to develop our adaptive
fies |f (x, t)| ≤ d(t)ψ(x), where d(t) is a bounded but unknown control design and hence it is necessary to restate some basic
˙ ) = ẋ(t); and 3) α(τ ) > 0.
concepts: 1) x̆(τ ) = x(t); 2) α(τ )x̆(τ
perturbation, and ψ(x) ≥ 0 is a known computable function.
Assumption 2: The time-varying function b(t) is away from 1) Design for Systems With Time-Invariant Parameters:
zero, without losing generality, we assume that b(t) > 0 and Assumption 3 [30]: The nonlinearity f (x, t) can be param-
there exists an unknown b such that 0 < b ≤ |b(x, t)| < ∞ eterized as f (x, t) = θ ψ(x) with ψ(x) being a known smooth
for all x ∈ R, t ∈ [0, +∞). function and ψ(0) = 0, and θ being an unknown constant.
Remark 1: Assumption 2 is more general than the one used Assumption 4 [110]: The function b(x, t), called control
in [1] since the latter requires that b be known. coefficient, satisfies b(x, t) ≡ b with b being an unknown
Theorem 1: Under Assumptions 1 and 2, the closed-loop nonzero constant. The sign of b is available for control design.
system consisting of (9) and the control law (10) is PT stable Furthermore, we assume that there exists a known constant b
in the sense of Definition 4 and all internal signals are bounded satisfies b < |b|.
over [0, T) Since ψ ∈ C 1 and ψ(0) = 0, then by Hadamard’s Lemma,
2
we know that there exists a known smooth mapping ψ̄ such
u = −k(μx) − θ (μx) ψ + |μ̇μ−2 (μx)| (10) that ψ = ψ̄x.
Theorem 2: Under Assumptions 3 and 4, the closed-loop
where k > 0 and θ > 0. system consisting of (9) and the adaptive control law (14) is
Proof: Denote μx by z, and denote ψ + |μ̇μ−2 (μx)| by PT stable in the sense of Definition 3 and all internal signals
(x, t). Choose a Lyapunov function as V = 1/(2b)z2 , then are bounded over [0, T)
b 1  −2  ⎧
V̇ = μzu + μz μ̇μ z + f ⎪ u(t) = ρ̂(t)ū(t)
b b ⎪
⎪ ⎧
  ⎨ ⎪ ū(t) = − ka (t)x(t) + 1 θ̂ 2 x(t) + 1 ψ̄ 2 x(t)
max 1, d[0,t] ⎨ 2 2
b (14)
≤ μzu + μ|z| (x, t). (11) ⎪
⎪ ˙θ̂ (t) = γ x(t)ψ(x), θ̂ (0) ≥ 0
b b ⎪ ⎪
⎩⎩ ˙ θ
ρ̂(t) = −γρ sgn(b)x(t)ū(t)
Let = max{1, sup{|d(t)|}}/b, applying Young’s inequality
with θ > 0, we get where k > 1/(ρ̂(0)b), γθ > 0, γρ > 0 and a (t) = (da/dt)
μ 2 is a time-varying function as defined in footnote 2, the initial
μ|z| (x, t) ≤ θ μz2 2
. + (12) value of ρ is chosen as ρ(0) > 0 for b > 0 (or ρ(0) < 0 for
4θ
b < 0).
Note that b/b ≥ 1, then substituting (10) and (12) into (11), Proof: According to footnote 3, we know that x̆(τ ) = x(t)
we have ˙ ) = ẋ(t). By using Assumption 3, we rewrite (9) as
and α(τ )x̆(τ
μ 2 μ 2
V̇ ≤ −kμz2 + ≤ −2bkμV + . (13) 1 1
4θ 4θ x̆˙ = (bu + θ ψ) = bŭ(τ ) + θ ψ̆(x̆) . (15)
α(τ ) α(τ )
According to Proposition 10, (13) results in V ∈ L∞ [0,T), and
hence z ∈ L∞ [0, T). Furthermore, the state x = μ−1 2bV is Let ŭ = ρ̂(τ )ū(τ ), and choose a Lyapunov function
bounded and converges to zero as t → T. Since μ̇μ−2 = 1/T V(τ ) : [0, +∞) → [0, +∞) as
is bounded, then (x, t) is bounded, establishing the same for  2
1 2 1 2 |b| 1
u(t). Therefore, all signals are bounded, and the closed-loop V(τ ) = x̆ + θ − θ̂ (τ ) + − ρ̂(τ ) . (16)
system is PT stable in the sense of Definition 4. 2 2γθ 2γρ b
Remark 2: It is worth mentioning that, for system (9), as The derivative of V(τ ) along the trajectory of the system (15)
long as b(x, t) satisfies b(x, t) ≥ b > 0 with b being some is
unknown constant, the proposed PT control does not need a
 
priori information on b(x, t), such simple control algorithm
dV(τ ) x̆ θ − θ̂ γθ x̆ψ̆ dθ̂
without involving b can be readily extended to higher-order = ū(τ ) + θ̂(τ )ψ̆ + −
systems in normal form [1]. Other robust PT control results dτ α(τ ) γθ α(τ ) dτ
  
can be found in [13], [17], [29], and [60] and the references |b| 1 γρ sgn(b) dρ̂
− − ρ̂ x̆ū(τ ) + . (17)
therein. γρ b α(τ ) dτ

Authorized licensed use limited to: Xian Jiaotong University. Downloaded on June 27,2025 at 10:09:34 UTC from IEEE Xplore. Restrictions apply.
4108 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS: SYSTEMS, VOL. 53, NO. 7, JULY 2023

Note that the control law and update laws designed in Proof: It is straightforward to prove that
Theorem 2 are equivalent to T k(t)
eA(t) e− 0 T−t dt
1
ū(τ ) = −kα(τ )x̆ − θ̂ 2 (τ )x̆ −
1 ˘2
ψ̄ x̆ lim = lim ≥0
t→T T − t t→T T −t
2 2
dθ̂ (τ ) 1 dρ̂(τ ) 1 and
= γθ x̆ψ̆, =− γρ sgn(b)x̆ū(τ ).
dτ α(τ ) dτ α(τ ) T k(t) T kmin
(18) e− 0 T−t dt e− 0 T−t dt ekmin ln(T−t)
lim ≤ lim = lim
t→T T −t t→T T −t t→T T −t
Since −(1/2)θ̂ 2 (τ )x̆2 − (1/2)ψ̄˘ 2 x̆2 + θ̂ (τ )ψ̆ x̆ ≤ 0, then by = lim (T − t)kmin −1 = 0.
substituting (18) into (17), we get t→T

dV(τ ) According to Squeeze Theorem, we obtain limt→T eA(t) /(T −

= −kx̆2 (τ ) ≤ 0. (19)
dτ t) = 0. Next, we continue to prove limt→T x/(T − t) = 0.
It follows that from (19) that V(τ ) ∈ L∞ , which indicates that Dividing both sides of (22) by (T − t), we have
x̆(τ ) ∈ L2 ∩L∞ . In view of x̆˙ ∈ L∞ , it follows from Barbalat’s  t
x eA(t) f (x, s) eA(t)
Lemma that limτ →∞ x̆(τ ) = 0 (i.e., limt→T x(t) = 0). = ds + x(0). (23)
Furthermore, according to (14) and Assumption 3 we know T −t T − t 0 eA(s) T −t
that there exists a number L such that |dθ̂(τ )/dτ | ≤ Lx̆2 , As t → T, the last term on the right-hand side of (23) con-
and hence dθ̂ (τ )/dτ ∈ L1 . In view of the argument of verges to zero since limt→T eA(t) /(T − t) = 0 and x(0) is
[112, Th. 3.1], it follows that θ̂ (τ ) has a limit as τ → ∞. bounded. Applying L’Hôpital’s Rule to the first term on the
Similarly, one can conclude that ρ̂(τ ) has a limit as τ → ∞. right-hand side of (23), we have
In addition, it follows from (14) that5 ρ̂(t)˙ = γρ sgn(b)xū =
 t  t −A(s)
γρ sgn(b)(ka (t)+θ̂ /2+ψ̄ /2)x ≥ 0 for ∀b > 0 and ρ̂(t)
2 2 2 ˙ ≤0 eA(t) f (x, s) e f (x, s)ds
lim ds = lim 0 −A(t)
for ∀b < 0. Therefore, one can conclude that ρ̂ ≥ 0 if we t→T T − t 0 eA(s) t→T e (T − t)
choose ρ̂(0) > 0 when b > 0, and that ρ̂ ≤ 0 if we choose e−A(t) f (x, t) f (x, t)
ρ̂(0) < 0 when b < 0, which also yields ρ̂(t)b > 0. Such a = lim = lim .
t→T k(t)
e−A(t) (T − t) − e−A(t) t→T k(t) − 1
conclusion is useful for choosing a suitable control gain k, as T−t
seen shortly. (24)
To proceed, we rewrite the closed-loop dynamics as
Since k(t) ≥ kmin > 1 and limt→T f (x, t) = 0,
1 t
ẋ = −kbρ̂a (t)x − bρ̂ θ̂ 2 x + ψ̄ 2 x + θ ψ. (20) then limt→T eA(t) /(T − t) 0 (f (x, s)/eA(s) )ds = 0, implying
2 limt→T x/(T − t) = 0. The proof of Lemma 1 is completed.
Recall footnote 2, we know that a(t) = ln(T/(T − t)), and According to Theorem 1, we know that
1  
a(0) = 1, a(T) = +∞, a (t) = . kx(t) 1 2 1 2
T −t u(t) = −ρ̂(t) + θ̂ x(t) + ψ̄ x(t) .
T −t 2 2
Then, (20) can be simplified as
In terms of Lemma 1, we obtain that the control input u(t)
ẋ = −
k(t)
x + f (x, t) (21) is bounded over [0, T) and limt→T u(t) = 0. Therefore,
T −t the closed-loop system is PT stable in the sense of
where k(t) = kbρ̂(t) is a positive function and f (x, t) = Definition 3.
−(1/2)bρ̂(θ̂ 2 x + ψ̄ 2 x) + θ ψ is a bounded function. Solving Remark 3: Since |ρ̂(t)| is a monotone increasing function
the differential inequality (21) gives and bρ̂ > 0, then kmin = ρ̂(0)b. Therefore, we only need to
 t  t pick k > 1/(ρ̂(0)b) to ensure that kmin > 1. Particularly, when
f (x, s) k(s) the control coefficient b is known, as assumed in [30], there
x(t) = eA(t)
ds + e x(0), A(t) =
A(t)
− ds.
0 e A(s)
0 T −s is no need for using Lemma 1, we just need to choose k > 1.
(22) 2) Design for Systems With Time-Varying Parameters:
Assumption 5: The nonlinearity f (x, t) satisfies that
To show the boundedness of u(t), we state the following f (x, t) = θ (t)ψ(x), where ψ(x) is a known smooth function,
Lemma. ψ(0) = 0, and the time-varying parameter θ (t) takes values
Lemma 1: For (22), if limt→T f (x, t) = 0 and a con- in an unknown compact set, i.e., there exists an unknown
stant kmin = inf{k(t)} satisfies kmin > 1, then the following constant δθ such that |θ (t)| ≤ δθ .
equations hold: Remark 4: Such an Assumption is more general than the
eA(t) x one used in [30], since the latter requires that θ (t) be time-
lim = 0, lim = 0. invariant. It is also more general than the Assumption used
t→T T −t t→T T − t
in [109] because the latter requires that δθ be known.
5 It is important to ensure that xū ≤ 0 since this guarantees the monotonicity Theorem 3: Under Assumptions 4 and 5, the closed-loop
of ρ̂(t) and thus allows to explicitly pick a suitable control gain k. system consisting of (9) and the adaptive control law (25) is

Authorized licensed use limited to: Xian Jiaotong University. Downloaded on June 27,2025 at 10:09:34 UTC from IEEE Xplore. Restrictions apply.
SONG et al.: PT CONTROL AND ITS LATEST DEVELOPMENTS 4109

PT stable in the sense of Definition 3 and all internal signals compact set within preset time T, if the control law and update
are bounded over [0, T) laws are designed as
⎧ ⎧
⎪
⎪ ⎧ = ρ̂(t)ū(t)
u(t) ⎪
⎪
⎪ ⎧ = N (ξ )ū(t)
u(t)
 
⎪
⎪ ⎪
⎪ ⎪ ū(t) = ka (t)x(t) + 1 1 + θ̂ 2 ψ̄ 2 + δ̂ θ 1 + ψ̄ 2 x(t)
⎪
⎪ ⎪
⎪ ū(t) = − ka (t)x(t) + 12 θ̂ 2 (t)x(t) + 12 ψ̄ 2 x(t) + v(t) ⎨⎪
⎪ ⎪
⎪
⎪ ⎪
⎪ ⎪
⎨˙ 
2

⎨⎪
⎪ ⎪   (31)
⎪ δ̂
⎨ v(t) = 2θ x(t) 1 + ψ̄ 2 ⎪ δ̂ θ (t) = γδ x2 (t) 1 + ψ̄ 2 , δ̂ θ (0) ≥ 0
(25) ⎪
⎪ ⎪ ˙
  ⎪ ⎪
⎪ ⎪ δ̂˙ θ (t) = γ2δ x2 (t) 1 + ψ̄ 2 ,
⎪
⎪ δ̂ (0) ≥ 0 ⎪
⎩⎪
⎪ ⎪ θ̂ (t) = γθ x(t)ψ(x),
⎩
θ̂ (0) ≥ 0
⎪
⎪ ⎪ θ
ξ̇ (t) = x(t)ū(t), ξ(0) > 0
⎪⎪
⎪ ⎪ θ̂˙ (t) = γ x(t)ψ(x),
⎪ θ̂ (0) ≥0
⎪
⎪ ⎪ θ
⎩⎪ ⎩˙ where k > 0, γδ > 0, γθ > 0, a (t) = (da/dt) is a
ρ̂(t) = −γρ sgn(b)x(t)ū(t)
time-varying function as defined in Section III-A and N (ξ )
where k > 1/(ρ̂(0)b), γδ > 0, γθ > 0, γρ > 0 and a (t) = is an enhanced type B-L Nussbaum function as defined
(da/dt) is a time-varying function as defined in Section III-A. in [111, Definition 4.2].
The initial value of ρ is chosen as ρ(0) > 0 for b > 0 (or Proof: First, we rewrite (9) as
ρ(0) < 0 for b < 0). 1
Proof: Similar to the proof of Theorem 2, we first x̆˙ = b̆(τ )N̆ (ξ )ŭ(τ ) + θ̂ ψ̆ + θ − θ̂ ψ̆ + θ ψ̆
α(τ )
rewrite (9) as (32)
1 with θ being an unknown constant and θ = θ (τ )−θ . Then,
x̆˙ = bŭ(τ ) + θ̂ ψ̆ + θ − θ̂ ψ̆ + θ ψ̆ (26)
α(τ ) choosing a Lyapunov function V(τ ) candidate as
with θ being some constant and θ = θ (τ ) − θ . We then 1 1 2 1 2
V(τ ) = x̆2 + θ − θ̂ (τ ) + δ θ − δ̂ θ . (33)
choose a Lyapunov function V(τ ) : [0, +∞) → [0, +∞) as 2 2γθ 2γδ
 2 Taking derivative of V(τ ) along the trajectory of (32), we get
1 2 1 2 |b| 1
V(τ ) = x̆ + θ − θ̂ (τ ) + − ρ̂(τ ) dV(τ ) dξ 1 1
2 2γθ 2γρ b = b̆(τ )N̆ (ξ ) + 1 − x̆ū + θ̂ ψ̆
dτ dτ α(τ ) α(τ )
+
1
δ θ − δ̂ θ .
2
(27)  
1 1 1 dθ
2γδ + x̆ θ ψ̆ + θ − θ̂ γθ x̆ψ̆ −
α(τ ) γθ α(τ ) dτ
With (25), it follows that:
1 dδ̂ θ
dV(τ ) 1 + δ θ − δ̂ θ − . (34)
≤ −kx̆2 (τ ) − x̆v(τ ) − x̆ θ ψ̆ γδ dτ
dτ α(τ ) Inserting the control law designed in (31) into (34), yields
1
− δ θ − δ̂ θ x̆2 1 + ψ̄ 2 (τ ) . (28) dV(τ ) dξ
α(τ ) ≤ b̆(τ )N̆ (ξ ) + 1 (35)
dτ dτ
Since where dξ /dτ = x̆ū(τ ) = kα(τ )x̆2 (τ )+(1/2)(1+ θ̂ 2 ψ̄ 2 )x̆2 (τ )+
δ θ δ θ (1/2)(δ̂ θ (1 + ψ̄ 2 ))x̆2 (τ ) ≥ 0 ∀τ ≥ 0. Thus, it follows from
θ ψ̆ = θ ψ̄(τ )x̆ ≤ x̆2 ψ̄ 2 (τ ) +
2
x̆ x̆2 (29)
2 2 Lemma 2 that V(τ ) ∈ L∞ and ξ ∈ L∞ . Note that the
boundedness of N (ξ ) is guaranteed by the boundedness of
then substituting v into (29) yields dV/dτ ≤ −kx̆2 ≤ 0.
ξ . Therefore, it can be concluded that x̆˙ ∈ L∞ , which fur-
Thus, according to an analysis similar to that in the proof of
ther indicates that limτ →∞ x̆(τ ) = 0 via Barbalat’s Lemma. In
Theorem 2, it can be concluded that all signals are bounded
addition, in terms of the analysis similar to that in the proof
and the closed-loop system is PT stable in the sense of
of Theorem 2, the boundedness of all closed-loop signals can
Definition 3.
be guaranteed and hence the closed-loop system is PT stable
3) Adaptive Nussbaum Gain Design:
in the sense of Definition 3.
Assumption 6: The function b(x, t), called control coeffi-
Remark 5: It is noted that the adaptive PT control is
cient, is away from zero and takes values in a compact set.
developed for the system with unknown yet time-varying
However, its magnitude and sign are unknown. There exists a
parameters in both feedback and input channels. These param-
known positive constant b satisfies b ≤ |b(x, t)|.
eters are not slowly time-varying, but rather, are fast time-
Lemma 2 [111]: Consider two C ∞ positive functions
varying or even involve abrupt changes, thereby making the
V(t) : [0, ∞) → R+ and N (t) : [0, ∞) → R+ . Let
controller design quite challenging. Although the control algo-
b(t) : [0, ∞) → [b, b̄] for two constants b and b̄ satisfying
rithm is based on the first-order system, the fundamental idea
bb̄ > 0. If, for ∀t ≥ 0
and the key design steps are worth extending to more gen-
V̇(t) ≤ (b(t)N (ξ ) + 1)ξ̇ (t), ξ̇ (t) ≥ 0 (30) eral systems. Some attempts about PT control for high-order
systems with unknown directions can be found in [42]. In addi-
for an enhanced Nussbaum function N , then ξ(t) and V(t) are tion, since neural network (NN) can be combined with robust
bounded over the whole time interval [0, ∞). adaptive to deal with modeling uncertainties, how to compen-
Theorem 4: Under Assumptions 5 and 6, all internal signals sate the NN reconstruction error to get PT stability represents
are bounded over [0, T) and system state x(t) converges to a an increasing topic for future study.

Authorized licensed use limited to: Xian Jiaotong University. Downloaded on June 27,2025 at 10:09:34 UTC from IEEE Xplore. Restrictions apply.
4110 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS: SYSTEMS, VOL. 53, NO. 7, JULY 2023

IV. P RESCRIBED -T IME C ONTROL FOR MIMO S YSTEMS Proof: Consider V = 1/(2wB )Z  Z with wB being an
PT control for MIMO nonlinear systems is an open area unknown positive constant, then
of research that is both theoretically and practically important 1
V̇ = μZ  BU + F + μ̇μ−2 Z
and urgent, especially with new problems arising from emerg- wB
 
ing applications, such as missile guidance, accurate and timely B + B B − B
weather forecasting, aircraft and spacecraft flight control, and = μZ  − −kZ − θ Z 2
2wB 2wB
obstacle avoidance in robotic systems, all of which require
1
new control technologies for time optimization. There have + μZ  F + μ̇μ−2 Z . (38)
been few research on PT control for MIMO nonlinear systems, wB
particularly when the control gain matrix is unknown, and In light of Assumption 7, there exists some unknown constant
essentially no findings that can provide PT stabilization, reg- wB > 0, such that 0 < wB ≤ (1/2)λmin (B + B ). Therefore
ulation, or tracking. In [107], by using time-based generators, 1 
a PT control algorithm is applied to a 7-DoF robot manipula- Z B + B Z ≥ Z2 . (39)
2wB
tor with a precondition that all information in the control gain
matrix is available. In [24], a parametric Lyapunov function- In addition, (B − B ) is skew symmetric and hence Z  (B −
based PT controller is applied to a spacecraft rendezvous B )Z = 0 ∀Z ∈ Rn . Now, it follows from (38) that
control system, where the mathematical model of such system 1
V̇ ≤ −kμZ2 − θ Z2  2 + μZ  F + μ̇μ−2 Z .
can be viewed as an MIMO linear system. In [21], a PT regula- wB
tion method is developed for the Euler–Lagrange system with (40)
a known inertia matrix. In [60], PT tracking control for MIMO
With Young’s inequality, we get (1/wB )μZ  (F +
systems with unknown control gain matrix and nonvanishing
μ̇μ−2 Z) ≤ (1/wB )μZ(d(t) + μ̇μ−2 Z) ≤
uncertainties are studied. In addition, some other studies con-
μZ ≤ θ μZ2  2 + (μ 2 /4θ ), with θ > 0,
sider the practical PT tracking control (see [113]), whose basic
= (1/wB ) × max{1, sup{|d(t)|}} and = μ̇μ−2 Z + .
idea is to introduce a smooth function that can converge to a
Therefore, we have
given value at the prescribed time, and to convert the original
constrained system into an unconstrained one by using the idea μ 2 μ 2
V̇ ≤ −kμZ2 + = −2wB kμ(t)V + . (41)
of coordinate transformation similar to that in the prescribed 4θ 4θ
performance control theory [114], [115], and finally to obtain It follows from Proposition 10 that V ∈ L∞ [0, T). Using the
the tracking error of the original system that can converge to analysis similar to that below (13), one can conclude that all
a given accuracy at the prescribed time by proving the bound- signals are bounded over [0, T) and X(t) → 0 as t → T.
edness of the converted system. In the following sections, Therefore, (36) is PT stable in the sense of Definition 3.
we introduce a powerful design approach for MIMO system
that applies not only to square systems but also to nonsquare B. Nonsquare System
systems. Now, we consider a nonsquare MIMO system Ẋ = BU + F
satisfying the following Assumption:
A. Square System Assumption 8 [60]: The high-frequency gain matrix
B(X, t) ∈ Rn×m can be characterized as B(X, t) =
Consider an MIMO nonlinear system as follows:
A(X, t)M(X, t), where M ∈ Rm×m is uncertain yet pos-
sibly asymmetric and A ∈ Rn×m is a known matrix with
Ẋ = B(X, t)U + F(X, t) (36)
full row rank. The message usable for synthesis is that
A(M + M )A is symmetric and positive definite.
where U ∈ Rn and X ∈ Rn are the input and the state vec-
Under Assumption 8, we get a new MIMO system as
tor, respectively. F(X, t) = [f1 , . . . , fn ] ∈ Rn denotes the
follows:
modeling uncertainties and external perturbations and each
fi satisfies Assumption 1, i.e., F ≤ d(t) with  = Ẋ = AMU + F(X, t) (42)
[ψ1 , . . . , ψn ] ∈ Rn .
where U ∈ Rm and X ∈ Rn
are the input and the state vector,
Assumption 7 [60]: The matrix B(X, t) ∈ Rn×n is square
respectively.
and unknown. The only information available for control
According to Assumption 7, we known that the positive
design is that (B + B ) is positive definite and symmetric.
definiteness of (B + B ) ensures that λmin (B + B ) is always
Theorem 5: Under Assumptions 1 and 7, the closed-loop
positive and there exists some positive unknown constant wA ,
system consisting of (36) and the control law (37) is PT stable
such that 0 < wA ≤ (1/A)λmin (A(M + M )A ).
in the sense of Definition 4 and all internal signals are bounded
Theorem 6: Under Assumptions 1 and 8, the closed-loop
over the time interval [0, T)
system consisting of (36) and the control law (43) is PT stable
U = −kZ − θ Z 2 (37) in the sense of Definition 4 and all internal signals are bounded
over [0, T)
where k > 0, θ > 0, Z = (TX/(T − t)), and = + A
U=− kZ + θ Z 2 (43)
μ̇μ−2 Z. A

Authorized licensed use limited to: Xian Jiaotong University. Downloaded on June 27,2025 at 10:09:34 UTC from IEEE Xplore. Restrictions apply.
SONG et al.: PT CONTROL AND ITS LATEST DEVELOPMENTS 4111

where k > 0, θ > 0, Z = μX with μ(t) = (T/(T − t)) and Meanwhile, the Laplace matrix L has only one zero
=  + μ̇μ−2 Z. eigenvalue and all other eigenvalues with positive real
Proof: This proof is omitted as it is straightforward by tak- parts if and only if the corresponding directed graph G
ing the analysis in the proof of Theorem 5. The difference contains a spanning tree [117].
is that we need to replace the inequality Z  (B + B /2)Z ≥ 2) When αij = 0, it corresponds to the discontinuous FT
wB Z2 in (39) with Z  (A(M + M )A /[2A])Z ≥ consensus protocol outlined in [118].
wA Z2 . 3) When 0 < αij < 1, it reduces to the continuous however
Remark 6: The main challenges in designing a PT con- nonsmooth FT consensus protocol established in [119].
troller for a high-order MIMO system are how to cope with It is important to note that with 0 < αij < 1, the finite
the unknown nonlinear perturbations due to the unknown con- settling time T is determined by Proposition 1 as T =
trol matrix and how to relax the assumptions on the control (V 1−α (0)/[c(1 − α)]) with c > 0 being some constant asso-
matrix in order to make more general control algorithms. ciated with the design parameters k, αij , and λ2 (L)6 (which
relies on the structure of G). There are several issues associated
V. L ATEST D EVELOPMENTS IN P RESCRIBED -T IME with the settling time T.
C ONTROL 1) The settling time T is affected by design parameters k
In this section, we aim to present a literature survey of the and αij , the initial state V(0), as well as the topological
foundations of PT decentralized control theory. Knowledge structure.
of graph theory can be found in any of the papers about 2) To produce a lower T, one can increase k or decrease
multiagents, which we have omitted here due to space con- αij (creating a larger c or a smaller α), but the control
straints. effort increases with a smaller αij .
The idea of using time-varying feedback to obtain PT sta- 3) If a settling time T is imposed, it is necessary to try to
bility has already appeared in early distributed control and find the relevant parameters c and α based upon V(0)
has accomplished a large diffusion in recent years. For exam- from Proposition 1, which cannot be explicitly preset
ple, PT consensus on single and double integrator dynamics because αij is implicitly involved in the function and
cases [43], [105]; PT consensus under undirected/directed the initial condition may be unknown.
graph and PT containment under multiple leaders of first-order The following PT consensus protocol circumvents all the
networked multiagent systems [44]; leader-following control aforementioned shortcomings [44]:
 
of high-order multiagent systems without/with mismatched μ̇
uncertainties [45], [46]; PT consensus via time base genera- ui = − k + c ei , i = 1, 2, . . . , n (47)
μ
tor [47]; cluster synchronization of complex networks [48]; lag
consensus of second-order leader-following multiagents [49]; where k > 0, c is a parameter will be designed later, ei =

PT consensus observer for high-order multiagents [50]; PT j∈Ni aij (xi − xj ) is the local neighborhood error, and μ(t) is
bipartite consensus tracking [51], [52], [53]; PT consen- well defined on [0, T) as in (8).
sus over time-varying graph via time scaling [54], and then Theorem 7: Consider system (44) in conjunction with the
generalized in [55], [56], [57], and [58], in which, PT forma- protocol (47). If the graph G is undirected and connected, and
tion tracking, leader-following control, uncertain multiagent the design parameter c is selected as c ≥ 1/λ2 (L), then the
dynamics, multiagent rigid body system, are considered. consensus is attained in PT, namely
1
A. Prescribed-Time Consensus Protocol χ (t) ≤ δ(0)e−kλ2 (L)t ∀t ∈ [0, T) (48)
μ(t)
Consider a multiagent system where the dynamics of each 
subagent is a single integrator where χ = [χ1 , χ2 , . . . , χn ] ∈ Rn and χi = xi −(1/n) ni=1 xi .
Furthermore, the control input remain C 1 smooth and bounded
ẋi = ui , i = 1, 2, . . . , n. (44) over [0, T).
The general consensus protocol is
B. Prescribed-Time Containment Protocol

n
 
ui = −k aij sgn xj − xi |xj − xi |αij , 1 ≤ i ≤ n (45) When the communication topology structure involves
j=1 multiple leaders, the containment control can be naturally
evolved from the consensus control. In this section, the
where 0 ≤ αij ≤ 1 and k > 0. Obviously, protocol (45) covers
achieved consensus result is extended to the scenario of
several common cases.
containment.
1) When αij = 1, it simplifies to the classical asymptotic
Theorem 8: Consider system (44) in conjunction with the
consensus protocol studied in [116]; then the original
protocol (47). If the graph G has a directed spanning tree
system can be abbreviated as ẋ = −Lx, where x =
leaded by the root node xi , and the design parameter c is
[x1 , x2 , . . . , xn ] ∈ Rn and L is the Laplace matrix of
selected as c ≥ 2λmax (P̃)/λ1 (Q̃), then, for ∀t ∈ [0, T), the
the system, and L = [lik ]n×n
 containment is attained in PT, namely
lik = j∈Ni aij , k=i
(46)
−aik , k = i. 6 λ (L) denotes the ith minimum eigenvalue of the Laplace matrix L.
i

Authorized licensed use limited to: Xian Jiaotong University. Downloaded on June 27,2025 at 10:09:34 UTC from IEEE Xplore. Restrictions apply.
4112 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS: SYSTEMS, VOL. 53, NO. 7, JULY 2023



 −kλ1 (Q̃)t Note that if we choose k ≡ 1/(1 − α), the PT controller (50)
1  λmax P̃ 2λmax (P̃)
Z̃(t) ≤  L−1 ⊗ I m Ẽ(0)e becomes the FT controller (51), which means that the FT con-
μ(t) λ 1
min P̃ troller (51) is indeed a special case of the PT controller (50).
In fact, they share the same property that the control gain tends
where L1 ∈ R(n−1)×(n−1) is a nonsingular matrix with all to ∞ as t → T. As a matter of fact, all FT controllers (includ-
eigenvalues satisfying λi (L1 ) > 0, i = 1, . . . , n − 1, whose ing FxT controllers, PT controllers, and PdT controllers) share
specific expression can be obtained according to the Laplace this property. Also, note that the magnitude of the PT control
matrix L, namely input (50) (consisting of a high-gain function −k/(T − t) and
! " a feedback signal x) does not become large when the feedback
0 01×(n−1)
L= signal decays faster than the high-gain function grows.
L2 L1

and Z̃ = [z2 , z3 , . . . , zn ] ∈ Rn−1 with zi = xi − x1 , B. Discussion on Implementation

Ẽ = [e2 , e3 , . . . , en ] ∈ Rn−1 and Q̃ = P̃L1 + L 1 P̃ with In the implementation of FT control algorithms, it is nec-
P̃ = diag{p2 , . . . , pn } and [p2 , . . . , pn ] = (L
1 ) −1 1
n−1 . essary to introduce sign function sgn(x) to avoid singularity
Furthermore, the control ui remains C smooth and bounded
1
when x(t) = 0. For example, the control law u = −x1/3 is pro-
over [0, T). grammed to be replaced by u = − sgn(x)|x|1/3 . Two effective
Proof: The proofs of Theorems 7 and 8 are omitted as they ways of implementing PTC are as follows.
can be found in [44]. 1) Letting T = Ts (scheduled time) + (small constant) so
that the controller works for the scheduled time.
VI. C OMPARISON AND C ONCLUSION 2) Setting an upper bound on the scaling function μ(t)
To close this article, we recap the connection between the before the time variable approaches the desired preset
PT results to FT results and discuss the related numerical time T.
implementation issues. In addition, we compare the differences Anyhow, unbounded control gain will not cause unbounded
between typical FT/FxT/PdT and PT controllers via simulation control input, and many simulation results show that the PT
on a double integrator. Finally, we conclude by giving some regulation is achieved with a suitable control effort, without an
future research challenges. exorbitant price. Both of the above implement methods slightly
sacrifice the control precision in favor of promoting practical
implementation by avoiding unbounded gains. The major con-
A. Controller Structure
cern for time-varying feedback control is its robustness against
Consider the first-order integrator as follows: measurement noise. In [59], a robust approach has been estab-
ẋ = u, x(0) = x0 . (49) lished for second-order systems where bounded disturbances
corrupt the measurements. For the problem of PT stabiliza-
From [24], one can immediately obtain a time-varying tion of high-order systems in the presence of measurement
feedback-based PT controller as disturbances, further research is required.
kx To show the characteristics and the differences of FT, FxT,
uprescribed = − , k ≥ 1. (50) PdT, and PT control schemes, we consider a double integrator
T −t
for numerical simulation.
Also, from [65], we get the classical FT autonomous con-
troller as ẋ1 = x2 , ẋ2 = u. (53)
ufinite = −k sgn(x)|x|α , 0 < α < 1, k > 0 (51) The FT controller [65, Example 1], the FxT controller
[83, Example 5.11], the PdT controller [93, Example 4.2], and
then the solution of (49) with (51) is
# the PT controller [30] are shown below.
  1 $  1/3 5/3
1) ufinite = −x2 − (x1 + (3/5)x2 )1/5 .
x(t) = |x0 |1−α − k(1 − α)t 1−α sgn(x0 ), t ∈ $0, Tf  2) ufixed = −(1 + 3x1 /2) sgn(s) − (s + s3 )[1/2] with s =
2
x(t) = 0, t ∈ Tf , +∞
x2 + (x2[2] + x1 + x13 )[1/2] and v[a] = |v|a sgn(v).
with 3) upredifined = −([∂ 1 (x1 , T1 )]/∂x1 )x2 − 2 (σ, T2 )
1 1 with σ = 1 + x2 and i (v, Ti ) =
Tf = |x0 |1−α . (5/2Ti ) exp(|v|2/5 )|v|3/5 sgn(v).
k (1 − α)
4) uprescribed = −2μ(x2 + 3μx1 ) − 3μ2 x1 − 3μx2 for
Therefore, we have |x0 |1−α = kTf (1 − α), and the control t ∈ [0, T) and uprescribed = 0 for t ∈ [0, ∞) with
law (51) can be rewritten as μ = 1/(T − t).
Two scenarios are considered for simulation: x1 (0) = 0.2 and
ufinite = −k sgn(x)|x|α−1 |x|
% %α−1 0.4 and x2 (0) = −0.2 and 0. Figs. 3–6 illustrate the simula-
% 1
% tion results, from which, it can be seen that the PT controller
= −k%% |x0 |1−α − k(1 − α)t % x
1−α
% achieves FT regulation in T = 1 s, whereas the settling time
x of the FT controller depends on the initial conditions, the FxT
=−  . (52)
(1 − α) Tf − t controller depends on the design parameters and the upper

Authorized licensed use limited to: Xian Jiaotong University. Downloaded on June 27,2025 at 10:09:34 UTC from IEEE Xplore. Restrictions apply.
SONG et al.: PT CONTROL AND ITS LATEST DEVELOPMENTS 4113

topics on PT control for complex dynamic plants include (but

not limited to).
1) Although adaptive PT control is investigated in [30],
the system unknown parameter must remain unchanged,
while the control coefficient must be accessible. In
some modern practical applications (such as autonomous
vehicles, circuit systems, high-speed trains, and high-
Fig. 3. FT control for a double integrator. performance robots [120], [121], [122], [123]), such
assumptions may not be satisfied since we know that
systems with changing structures usually have time-
varying system parameters and that the inertia matrix
(which can be viewed as the control influence gain) of a
robotic system is usually unknown. It is necessary and
challenging to develop more powerful solutions to meet
such scenarios.
2) Output feedback schemes often imply low cost, which
FxT control for a double integrator with the settling time t ≤ Tmax =
Fig. 4. √ is very attractive in practical applications (especially for
π + (π/ 2) ≈ 5.363 s. large-scaled/networked/multiagent systems). However,
the existing results can only achieve output feedback PT
stabilizing control for some special systems (e.g., linear
systems), it is therefore important to explore the output
feedback-based PT control for more general systems.
3) In addition to the PT stabilizing control, the study
of the PT tracking control is more general, however,
when tracking is considered, the desired trajectory to be
tracked would give rise to extra time variation and/or
Fig. 5. PdT control for a double integrator with the settling time t ≤ Tmax = uncertainties and hence brings technical obstacles. How
T1 + T2 = (0.2 + 0.8) s = 1 s. to improve the PT control algorithm so that it completely
rejects these nonvanishing uncertainties (which may be
generated by the desired tracking signal, may come from
some external noise, or may be inherent in the physical
model) is an interesting future research topic.
4) As for PT control for multiagent systems, it is interesting
to generalize the simple framework on first or second-
order integrators to agents having high-order uncertain
nonlinear dynamics and to investigate PT decentral-
Fig. 6. PT control for a double integrator with the settling time T = 1 s. ized control algorithms under complex communication
topologies, as well as to study how to achieve consensus
with as little information interaction between agents as
bound on the settling time is overestimated, and the PdT
possible without losing controllability.
controller requires a larger control input when the initial con-
5) The study of more types of system models, more low-
ditions become slightly larger, and a slighter overestimation
conservative control algorithms, or the pursuing for
of the settling time can be observed. Besides, from Figs. 3–6,
better control performance of closed-loop systems are
it is observed that the PT controller exhibits smoother con-
all interesting future research topics in the field of PT
trol action, avoiding the chattering phenomenon as reflected in
control.
Figs. 3–5. These simulation results show to some extent the
superiority of PT control compared to the other three control
methods. R EFERENCES
[1] Y. Song, Y. Wang, J. Holloway, and M. Krstic, “Time-varying feedback
C. Challenges and Future Opportunities for regulation of normal-form nonlinear systems in prescribed finite
time,” Automatica, vol. 83, pp. 243–251, Sep. 2017.
The core idea in PT control is to design a suitable time- [2] Y. Song, Y. Wang, and M. Krstic, “Time-varying feedback for stabi-
varying gain that goes to ∞ as the system converges to zero, lization in prescribed finite time,” Int. J. Robust Nonlinear Control,
vol. 29, no. 3, pp. 618–633, 2019.
which is derived from proportional navigation in strategic [3] Y. Wang and Y. Song, “A general approach to precise tracking of
and tactical missile guidance applications. In PT control, we nonlinear systems subject to non-vanishing uncertainties,” Automatica,
require more than that, not only to regulate the state precisely vol. 106, pp. 306–314, Aug. 2019.
[4] J. Holloway and M. Krstic, “Prescribed-time observers for linear
to zero but also to ensure that the control signal is bounded, systems in observer canonical form,” IEEE Trans. Autom. Control,
while completely rejecting external disturbance. The potential vol. 64, no. 9, pp. 3905–3912, Sep. 2019.

Authorized licensed use limited to: Xian Jiaotong University. Downloaded on June 27,2025 at 10:09:34 UTC from IEEE Xplore. Restrictions apply.
4114 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS: SYSTEMS, VOL. 53, NO. 7, JULY 2023

[5] J. Holloway and M. Krstic, “Prescribed-time output feedback for lin- [28] K. Zhang, B. Zhou, M. Hou, and G. Duan, “Prescribed-time stabiliza-
ear systems in controllable canonical form,” Automatica, vol. 107, tion of p-normal nonlinear systems by bounded time-varying feedback,”
pp. 77–85, Sep. 2019. Int. J. Robust Nonlinear Control, vol. 32, no. 1, pp. 421–450, 2022.
[6] N. Espitia, D. Steeves, W. Perruquetti, and M. Krstic, “Sensor delay- [29] D. Tran and T. Yucelen, “Finite-time control of perturbed dynamical
compensated prescribed-time observer for LTI systems,” Automatica, systems based on a generalized time transformation approach,” Syst.
vol. 135, Jan. 2022, Art. no. 110005. Control Lett., vol. 136, Feb. 2020, Art. no. 104605.
[7] N. Espitia and W. Perruquetti, “Predictor-feedback prescribed-time [30] C. Hua, P. Ning, and K. Li, “Adaptive prescribed-time control for a
stabilization of LTI systems with input delay,” IEEE Trans. Autom. class of uncertain nonlinear systems,” IEEE Trans. Autom. Control,
Control, vol. 67, no. 6, pp. 2784–2799, Jun. 2022. vol. 67, no. 11, pp. 6159–6166, Nov. 2022.
[8] A. Pal, B. Singh, S. Kamal, S. Nagar, and J. Goyal, “Arbitrary time [31] C. Hua, H. Li, K. Li, and P. Ning, “Adaptive prescribed-time
stabilization of a coupled tank system: A contraction based approach,” control of time-delay nonlinear systems via a double time-varying
in Proc. Int. Conf. Ind. Technol., 2020, pp. 23–28. gain approach,” IEEE Trans. Cybern., early access, Aug. 31, 2022,
[9] N. de la Cruz and M. Basin, “Predefined-time control of full-scale doi: 10.1109/TCYB.2022.3192250.
4D model of permanent-magnet synchronous motor with deterministic [32] R. I. V. Kairuz, Y. Orlov, and L. Aguilar, “Prescribed-time stabilization
disturbances and stochastic noises,” Actuators, vol. 10, no. 11, p. 306, of controllable planar systems using switched state feedback,” IEEE
2021. Control Syst. Lett., vol. 5, no. 6, pp. 2048–2053, Dec. 2021.
[10] H. Ye and Y. Song, “Prescribed-time control for linear systems in [33] Y. Orlov, R. V. Kairuz, and L. Aguilar, “Prescribed-time robust differ-
canonical form via nonlinear feedback,” IEEE Trans. Syst., Man, entiator design using finite varying gains,” IEEE Control Syst. Lett.,
Cybern., Syst., vol. 53, no. 2, pp. 1126–1135, Feb. 2023. vol. 6, pp. 620–625, 2022.
[11] P. Krishnamurthy, F. Khorrami, and M. Krstic, “Robust adaptive [34] Y. Chitour, R. Ushirobira, and H. Bouhemou, “Stabilization for a per-
prescribed-time stabilization via output feedback for uncertain nonlin- turbed chain of integrators in prescribed time,” SIAM J. Control Optim.,
ear strict-feedback-like systems,” Eur. J. Control, vol. 55, pp. 14–23, vol. 58, no. 2, pp. 1022–1048, 2020.
Sep. 2020. [35] Y. Orlov, “Time space deformation approach to prescribed-time stabi-
[12] P. Krishnamurthy, F. Khorrami, and M. Krstic, “Adaptive output- lization: Synergy of time-varying and non-Lipschitz feedback designs,”
feedback stabilization in prescribed time for nonlinear systems with Automatica, vol. 144, Oct. 2022, Art. 110485.
unknown parameters coupled with unmeasured states,” Int. J. Adapt. [36] Y. Song, J. Su, “A unified Lyapunov characterization for finite time
Control Signal Process., vol. 35, no. 2, pp. 184–202, 2021. control and prescribed time control,” Int. J. Robust Nonlinear Control,
[13] P. Krishnamurthy, F. Khorrami, and M. Krstic, “A dynamic high- to be published, doi: 10.1002/rnc.6544.
gain design for prescribed-time regulation of nonlinear systems,” [37] X. Peng, J. Sun, and Z. Geng, “A specified-time control framework for
Automatica, vol. 115, May 2020, Art. no. 108860. control-affine systems and rigid bodies: A time-rescaling approach,”
[14] Y. Wang and Y. Liu, “Prescribed-time exact tracking for a class of Int. J. Robust Nonlinear Control, vol. 29, no. 10, pp. 3163–3182, 2019.
nonlinear systems,” IEEE Control Syst. Lett., vol. 7, pp. 1351–1356, [38] R. Ma, L. Fu, and J. Fu, “Prescribed-time tracking control for non-
2023, doi: 10.1109/LCSYS.2023.3238657. linear systems with guaranteed performance,” Automatica, vol. 146,
Dec. 2022, Art. no. 110573.
[15] F. Gao, Y. Wu, and Z, Zhang, “Global fixed-time stabilization of
switched nonlinear systems: A time-varying scaling transformation [39] J. Wang, Q. Gong, K. Huang, Z. Liu, C. L. P. Chen, and J. Liu,
approach,” IEEE Trans. Circuits Syst. II, Exp. Briefs, vol. 66, no. 11, “Event-triggered prescribed settling time consensus compensation con-
pp. 1890–1894, Nov. 2019. trol for a class of uncertain nonlinear systems with actuator failures,”
IEEE Trans. Neural Netw. Learn. Syst., early access, Dec. 10, 2021,
[16] D. Steeves, M. Krstic, and R. Vazquez, “Prescribed–time estima-
doi: 10.1109/TNNLS.2021.3129816.
tion and output regulation of the linearized Schrödinger equation by
[40] K. Zhang, B. Zhou, H.-Y. Jiang, G.-P. Liu, and G.-R. Duan, “Practical
backstepping,” Eur. J. Control, vol. 55, pp. 3–13, Sep. 2020.
prescribed-time sampled-data control of linear systems with applica-
[17] W. Li and M. Krstic, “Stochastic nonlinear prescribed-time stabilization
tions to the air-bearing testbed,” IEEE Trans. Ind. Electron., vol. 69,
and inverse optimality,” IEEE Trans. Autom. Control, vol. 67, no. 3,
no. 6, pp. 6152–6161, Jun. 2022.
pp. 1179–1193, Mar. 2022.
[41] J. Qiu, T. Wang, K. Sun, I. J. Rudas, and H. Gao, “Disturbance
[18] W. Li and M. Krstic, “Prescribed-time control of stochastic nonlinear observer-based adaptive fuzzy control for strict-feedback nonlinear
systems with reduced control effort,” J. Syst. Sci. Complex., vol. 34, systems with finite-time prescribed performance,” IEEE Trans. Fuzzy
no. 5, pp. 1782–1800, 2021. Syst., vol. 30, no. 4, pp. 1175–1184, Apr. 2022.
[19] W. Li and M. Krstic, “Prescribed-time output-feedback control of [42] J. Su and Y. Song, “Prescribed-time control under unknown control
stochastic nonlinear systems,” IEEE Trans. Autom. Control, early gain and mismatched nonparametric uncertainties,” Syst. Control Lett.,
access, Feb. 14, 2022, doi: 10.1109/TAC.2022.3151587. vol. 171, Jan. 2023, Art. no. 105420.
[20] A. Shakouri, “On the prescribed-time attractivity and frozen-time [43] G. Jing and L. Wang, “Finite-time coordination under state-dependent
eigenvalues of linear time-varying systems,” Automatica, vol. 140, communication graphs with inherent links,” IEEE Trans. Circuits Syst.
Jun. 2022, Art. no. 110173. II, Exp. Briefs, vol. 66, no. 6, pp. 968–972, Jun. 2019.
[21] A. Shakouri and N. Assadian, “Prescribed-time control for perturbed [44] Y. Wang, Y. Song, D. J. Hill, and M. Krstic, “Prescribed-time consen-
Euler–Lagrange systems with obstacle avoidance,” IEEE Trans. Autom. sus and containment control of networked multiagent systems,” IEEE
Control, vol. 67, no. 7, pp. 3754–3761, Jul. 2022. Trans. Cybern., vol. 49, no. 4, pp. 1138–1147, Apr. 2019.
[22] A. Shakouri and N. Assadian, “Prescribed-time control with lin- [45] X. Chen, X. Zhang, and Q. Liu, “Prescribed-time decentralized regula-
ear decay for nonlinear systems,” IEEE Control Syst. Lett., vol. 6, tion of uncertain nonlinear multi-agent systems via output feedback,”
pp. 313–318, 2021. Syst. Control Lett.. vol. 137, Mar. 2020, Art. no. 104640.
[23] A. Shakouri and N. Assadian, “A framework for prescribed-time control [46] Y. Wang, and Y. Song, “Leader-following control of high-order
design via time-scale transformation,” IEEE Control Syst. Lett., vol. 6, multi-agent systems under directed graphs: Pre-specified finite time
pp. 1976–1981, 2022. approach,” Automatica, vol. 87, pp. 113–120, Jan. 2018.
[24] B. Zhou, “Finite-time stability analysis and stabilization by bounded [47] J. Colunga, C. Vázquez, H. Becerra, and D. Gómez-Gutiérrez,
linear time-varying feedback,” Automatica, vol. 121, Nov. 2020, “Predefined-time consensus of nonlinear first-order systems using
Art. no. 109191. a time base generator,” Math. Probl. Eng., vol. 2018, Oct. 2018,
[25] B. Zhou and Y. Shi, “Prescribed-time stabilization of a class of non- Art. no. 1957070.
linear systems by linear time-varying feedback,” IEEE Trans. Autom. [48] X. Liu, D. W. Ho, and C. Xie, “Prespecified-time cluster synchroniza-
Control, vol. 66, no. 12, pp. 6123–6130, Dec. 2021. tion of complex networks via a smooth control approach,” IEEE Trans.
[26] K. Zhang, B. Zhou, H. Jiang, and G. Duan, “Finite-time output regu- Cybern., vol. 50, no. 4, pp. 1771–1775, Apr. 2020.
lation by bounded linear time-varying controls with applications to the [49] Y. Ren, W. Zhou, Z. Li, L. Liu, and Y. Sun, “Prescribed-time clus-
satellite formation flying,” Int J. Robust Nonlinear Control, vol. 32, ter lag consensus control for second-order non-linear leader-following
no. 1, pp. 451–471, 2022. multiagent systems,” ISA Trans., vol. 109, pp. 49–60, Mar. 2021.
[27] K.-K. Zhang, B. Zhou, and G.-R. Duan, “Prescribed-time input-to-state [50] X. Gong, Y. Cui, T. Wang, J. Shen, and T. Huang, “Distributed
stabilization of normal nonlinear systems by bounded time-varying prescribed-time consensus observer for high-order integrator multi-
feedback,” IEEE Trans. Circuits Syst. I, Reg. Papers, vol. 69, no. 9, agent systems on directed graphs,” IEEE Trans. Circuits Syst. II, Exp.
pp. 3715–3725, Sep. 2022. Briefs, vol. 69, no. 4, pp. 2216–2220, Apr. 2022.

Authorized licensed use limited to: Xian Jiaotong University. Downloaded on June 27,2025 at 10:09:34 UTC from IEEE Xplore. Restrictions apply.
SONG et al.: PT CONTROL AND ITS LATEST DEVELOPMENTS 4115

[51] B. Ning, Q.-L. Han, and Z. Zuo, “Bipartite consensus tracking [76] J. Yu, P. Shi, and L. Zhao, “Finite-time command filtered backstep-
for second-order multiagent systems: A time-varying function-based ping control for a class of nonlinear systems,” Automatica, vol. 92,
preset-time approach,” IEEE Trans. Autom. Control, vol. 66, no. 6, pp. 173–180, Jun. 2018.
pp. 2739–2745, Jun. 2020. [77] A. Levant, “Homogeneity approach to high-order sliding mode design,”
[52] X. Gong, Y. Cui, J. Shen, Z. Shu, and T. Huang, “Distributed Automatica, vol. 41, no. 5, pp. 823–830, May 2005.
prescribed-time interval bipartite consensus of multi-agent systems on [78] S. P. Bhat and D. S. Bernstein, “Geometric homogeneity with applica-
directed graphs: Theory and experiment,” IEEE Trans. Netw. Sci. Eng., tions to finite-time stability,” Math. Control Signal Syst., vol. 17, no. 2,
vol. 8, no. 1, pp. 613–624, Jan.–Mar. 2021. pp. 101–127, May 2005.
[53] A. Pal, S. Kamal, X. Yu, S. Nagar, and X. Xiong, “Free-will arbitrary [79] M. Chen, Q. Wu, and R. Cui, “Terminal sliding mode tracking control
time consensus for multiagent systems,” IEEE Trans. Cybern., vol. 52, for a class of SISO uncertain nonlinear systems,” ISA Trans., vol. 52,
no. 6, pp. 4636–4646, Jun. 2022. no. 2, pp. 198–206, Nov. 2013.
[54] Z. Kan, T. Yucelen, E. Doucette, and E. Pasiliao, “A finite-time [80] W. Lin and C. Qian, “Adding a power integrator: A tool for global sta-
consensus framework over time-varying graph topologies with tem- bilization of high-order lower-triangular systems,” Syst. Control Lett.,
poral constraints,” J. Dyn. Syst. Meas. Control, vol. 139, no. 7, 2017, vol. 39, no. 5, pp. 339–351, Apr. 2000.
Art. 071012. [81] Z. Sun, M. Yun, and T. Li, “A new approach to fast global finite-
[55] T. Yucelen, Z. Kan, and E. Pasiliao, “Finite-time cooperative engage- time stabilization of high-order nonlinear system,” Automatica, vol. 81,
ment,” IEEE Trans. Autom. Control, vol. 64, no. 8, pp. 3521–3526, pp. 455–463, Jul. 2017.
Aug. 2019. [82] C. Zhang, J. Yang, Y. Yan, L. Fridman, and S. Li, “Semiglobal finite-
[56] E. Arabi, T. Yucelen, and J. Singler, “Finite-time distributed control time trajectory tracking realization for disturbed nonlinear systems via
with time transformation,” Int. J. Robust Nonlinear Control, vol. 31, higher-order sliding modes,” IEEE Trans. Autom. Control, vol. 65,
no. 1, pp. 107–130, 2021. no. 5, pp. 2185–2191, May 2020.
[57] D. Tran, T. Yucelen, and S. Sarsilmaz, “Finite-time control of [83] A. Polyakov, Generalized Homogeneity in Systems and Control. Berlin,
multiagent networks as systems with time transformation and sep- Germany: Springer, 2020.
aration principle,” Control Eng. Pract., vol. 108, Mar. 2021, [84] A. Polyakov, “Nonlinear feedback design for fixed-time stabilization
Art. no. 104717. of linear control systems,” IEEE Trans. Autom. Control, vol. 57, no. 8,
[58] D. Kurtoglu and T. Yucelen, “A time transformation approach to finite- pp. 2106–2110, Aug. 2012.
time distributed control with reduced information exchange,” IEEE [85] A. Polyakov, D. Efimov, and W. Perruquetti, “Finite-time and fixed-
Control Syst. Lett., vol. 6, pp. 485–563, 2022. time stabilization: Implicit Lyapunov function approach,” Automatica,
[59] D. Steeves and M. Krstic, “Prescribed-time stabilization robust to vol. 51, pp. 332–340, Jan. 2015.
measurement disturbances,” in Proc. Amer. Control Conf., 2022, [86] Z. Zuo and L. Tie, “A new class of finite-time nonlinear consensus
pp. 3873–3878. protocols for multi-agent systems,” Int. J. Control, vol. 87, no. 2,
[60] H. Ye and Y. Song, “Prescribed-time tracking control of MIMO non- pp. 363–370, 2014.
linear systems with non-vanishing uncertainties,” IEEE Trans. Autom. [87] Z. Zuo and L. Tie, “Distributed robust finite-time nonlinear consensus
Control, early access, Jul. 26, 2022, doi: 10.1109/TAC.2022.3194100. protocols for multi-agent systems,” Int. J. Syst. Sci., vol. 47, no. 6,
[61] P. Dorato, Short-Time Stability in Linear Time-Varying Systems, pp. 1366–1375, 2016.
Polytechn. Inst. Brooklyn, New York, NY, USA, 1961. [88] Z. Zuo, Q.-L. Han, B. Ning, X. Ge, and X.-M. Zhang, “An overview
of recent advances in fixed-time cooperative control of multiagent
[62] L. Weiss and E. Infante, “Finite time stability under perturbing forces
systems,” IEEE Trans. Ind. Informat., vol. 14, no. 6, pp. 2322–2334,
and on product spaces,” IEEE Trans. Autom. Control, vol. AC-12, no. 1,
Jun. 2018.
pp. 54–59, Feb. 1967.
[89] C. Chen and Z. Sun, “Fixed-time stabilisation for a class of high-
[63] H. D’Angelo, Linear Time-Varying Systems: Analysis and Synthesis.
order non-linear systems,” IET Control Theory Appl. vol. 12, no. 18,
Boston, MA, USA: Allyn and Bacon, 1970.
pp. 2578–2587, Dec. 2018.
[64] V. Haimo, “Finite time controllers,” SIAM J. Control Optim., vol. 24,
[90] B. Ning, Q.-L. Han, Z. Zuo, L. Ding, Q. Lu, and X. Ge, “Fixed-
no. 4, pp. 760–770, 1986.
time and prescribed-time consensus control of multi-agent systems and
[65] S. Bhat and D. S. Bernstein, “Continuous finite-time stabilization of its applications: A survey of recent trends and methodologies,” IEEE
the translational and rotational double integrators,” IEEE Trans. Autom. Trans. Ind. Inf., vol. 19, no. 2, pp. 1121–1135, Feb. 2023.
Control, vol. 43, no. 5, pp. 678–682, May 1998. [91] X. Li, C. Wen, and J. Wang, “Lyapunov-based fixed-time stabilization
[66] X. Huang, W. Lin, and B. Yang, “Global finite-time stabilization of control of quantum systems,” J. Autom. Intell., vol. 1, no. 1, Dec. 2022,
a class of uncertain nonlinear systems,” Automatica, vol. 41, no. 5, Art. no. 100005.
pp. 881–888, 2005. [92] Z. Wu, M. Ma, X. Xu, B. Liu, and Z. Yu, “Predefined-time parameter
[67] F. Amato, M. Ariola, and P. Dorato, “Finite-time control of lin- estimation via modified dynamic regressor extension and mixing,” J.
ear systems subject to parametric uncertainties and disturbances,” Franklin Inst., vol. 358, no. 13, pp. 6897–6921, 2021.
Automatica, vol. 37, no. 9, pp. 1459–1463, 2001. [93] J. Sánchez-Torres, D. Gómez-Gutiérrez, E. López, and A. Loukianov,
[68] A. Levant, “Higher-order sliding modes, differentiation and output- “A class of predefined-time stable dynamical systems,” IMA J. Math.
feedback control,” Int. J. Control, vol. 76, no. 9, pp. 924–941, 2003. Control Inf., vol. 35, no. 1, pp. 1–29, 2018.
[69] Y. Shen and X. Xia, “Semi-global finite-time observers for nonlinear [94] A. Muoz-Vázquez, J. Sánchez-Torres, M. Defoort, and S. Boulaaras,
systems,” Automatica, vol. 44, no. 12, pp. 3152–3156, 2008. “Predefined-time convergence in fractional-order systems,” Chaos
[70] Y. Hong and Z.-P. Jiang, “Finite-time stabilization of nonlinear systems Solitons Fractals, vol. 143, Feb. 2021, Art. no. 110571.
with parametric and dynamic uncertainties,” IEEE Trans. Autom. [95] R. Aldana-López, D. Gómez-Gutiérrez, E. Jiménez-Rodríguez,
Control, vol. 51, pp. 1950–1956, Dec. 2006. J. D. Sánchez-Torres, and M. Defoort, “Enhancing the settling time esti-
[71] F. Amato, M. Ariola, M. Carbone, and C. Cosentino, “Finite-time con- mation of a class of fixed-time stable systems,” Int. J. Robust Nonlinear
trol of linear systems: A survey,” in Current Trends Nonlinear System Control, vol. 29, no. 12, pp. 4135–4148, 2019.
Control, L. Menini, L. Zaccarian, and C. T. Abdallah, Eds. Boston, [96] J. Sánchez-Torres, A. Muñoz-Vázquez, M. Defoort,
MA, USA: Birkhauser, 2006, pp. 195–213. E. Jiménez-Rodríguez, and A. Loukianov, “A class of predefined-time
[72] S. Ding and S. Li, “A survey for finite-time control problems,” Control controllers for uncertain second-order systems,” Eur. J. Control,
Decis., vol. 26, no. 2, pp. 161–169, 2011. vol. 53, pp. 52–58, May 2020.
[73] Y. Liu, Y. Jin, X. Liu, and X. Li, “Survey on finite-time control for [97] P. Zarchan, Tactical and Strategic Missile Guidance, 6th ed. Reston,
nonlinear systems,” Control Theory Appl., vol. 37, no. 1, pp. 1–12, VA, USA: AIAA, 2012.
Jan. 2020. [98] G. Slater and W. Wells, “Optimal evasive tactics against a proportional
[74] M. Wang, H. Sun, and S. Qin, “An adaptive memristive navigation missile with time delay,” J. Spacecr. Rockets, vol. 10, no. 5,
dynamical system to nonsmooth optimization problems,” pp. 309–313, 1973.
Nonlinear Dyn., vol. 111, pp. 4451–4468, Mar. 2023, doi: [99] Y. Ho, A. Bryson, and S. Baron, “Differential games and optimal
10.1007/s11071-022-08075-1. pursuit-evasion strategies,” IEEE Trans. Autom. Control, vol. AC-10,
[75] H. Wang, B. Chen, C. Lin, Y. Sun, and F. Wang, “Adaptive finite-time no. 4, pp. 385–389, Oct. 1965.
control for a class of uncertain high-order non-linear systems based [100] Z. Rekasius, “An alternate approach to the fixed terminal point regulator
on fuzzy approximation,” IET Control Theory Appl., vol. 11, no. 5, problem,” IEEE Trans. Autom. Control, vol. AC-9, no. 3, pp. 290–292,
pp. 677–684, 2017. Jul. 1964.

Authorized licensed use limited to: Xian Jiaotong University. Downloaded on June 27,2025 at 10:09:34 UTC from IEEE Xplore. Restrictions apply.
4116 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS: SYSTEMS, VOL. 53, NO. 7, JULY 2023

[101] M. Sidar, “On closed-loop optimal control,” IEEE Trans. Autom. Yongduan Song (Fellow, IEEE) received the Ph.D.
Control, vol. AC-9, no. 3, pp. 292–293, Jul. 1964. degree in electrical and computer engineering from
[102] Z. Chen, X. Ju, Z. Wang, and Q. Li, “The prescribed time sliding mode Tennessee Technological University, Cookeville,
control for attitude tracking of spacecraft,” Asian J. Control, vol. 24, TN, USA, in 1992.
no. 4, pp. 1650–1662, 2022. He held a Tenured Full Professor with North
[103] N. Harl and S. N. Balakrishnan, “Impact time and angle guidance with Carolina A&T State University, Greensboro, NC,
sliding mode control,” IEEE Trans. Control Syst. Technol., vol. 20, USA, from 1993 to 2008, and a Langley
no. 6, pp. 1436–1449, Nov. 2011. Distinguished Professor with the National Institute
[104] N. Harl and S. Balakrishnan, “Reentry terminal guidance through slid- of Aerospace, Hampton, VA, USA, from 2005 to
ing mode control,” J. Guid. Control Dyn., vol. 33, no. 1, pp. 186–199, 2008. He is currently the Dean with the Institute
2010. of Artificial Intelligence, Chongqing University,
[105] Y. Cai, G. Xie, and H. Liu, “Reaching consensus at a preset time: Chongqing, China. He was one of the six Langley Distinguished Professors
Double-integrator dynamics case,” in Proc. 31st Chin. Control Conf., with the National Institute of Aerospace (NIA), Hampton, VA, USA, and the
2012, pp. 6309–6314. Founding Director of Cooperative Systems with NIA. His current research
[106] T. Tsuji, P. Morasso, and M. Kaneko, “Feedback control of nonholo- interests include intelligent systems, guidance navigation and control, and
nomic mobile robots using time base generator,” in Proc. IEEE Int. bio-inspired adaptive and cooperative systems.
Conf. Robot. Autom., vol. 2, 1995, pp. 1385–1390. Prof. Song was a recipient of several competitive research awards from
[107] H. Becerra, C. Vázquez, G. Arechavaleta, and J. Delfin, “Predefined- the National Science Foundation, the National Aeronautics and Space
time convergence control for high-order integrator systems using time Administration, the U.S. Air Force Office, the U.S. Army Research Office,
base generators,” IEEE Trans. Control Syst. Technol., vol. 26, no. 5, and the U.S. Naval Research Office. He has served/been serving as an
pp. 1866–1873, Sep. 2018. Associate Editor for several prestigious international journals, including the
[108] J. Slotine and W. Li, Applied Nonlinear Control. Englewood Cliffs, IEEE T RANSACTIONS ON AUTOMATIC C ONTROL, IEEE T RANSACTIONS
NJ, USA: Prentice-Hall, 1991. ON N EURAL N ETWORKS AND L EARNING S YSTEMS , IEEE T RANSACTIONS
[109] K. Chen, and A. Astolfi, “Adaptive control for systems with time- ON I NTELLIGENT T RANSPORTATION S YSTEMS , and IEEE T RANSACTIONS
varying parameters,” IEEE Trans. Autom. Control, vol. 66, no. 5, ON S YSTEMS , M AN , AND C YBERNETICS : S YSTEMS . He is currently the
pp. 1986–2001, 2020. Editor-in-Chief for the IEEE T RANSACTIONS ON N EURAL N ETWORKS AND
[110] J. Zhou and C. Wen, Adaptive Backstepping Control of Uncertain L EARNING S YSTEMS.
Systems: Nonsmooth Nonlinearities, Interactions or Time-Variations.
New York, NY, USA: Springer-Verlag, 2008.
[111] Z. Chen, “Nussbaum functions in adaptive control with time-varying Hefu Ye received the B.Eng. degree in measure-
unknown control coefficients,” Automatica, vol. 102, pp. 72–79, ment and control technology and instrument from
Apr. 2019. Harbin Institute of Technology, Harbin, China, in
[112] M. Krstic, “Invariant manifolds and asymptotic properties of adap- 2019. He is currently pursuing the Ph.D. degree with
tive nonlinear stabilizers,” IEEE Trans. Autom. Control, vol. 41, no. 6, the School of Automation, Chongqing University,
pp. 817–829, Jun. 1996. Chongqing, China.
[113] Y. Cao, J. Cao, and Y. Song, “Practical prescribed time control of His research interests include robotic systems,
Euler–Lagrange systems with partial/full state constraints: A settling robust adaptive control, prescribed performance con-
time regulator-based approach,” IEEE Trans. Cybern., vol. 52, no. 12, trol, and prescribed-time control.
pp. 13096–13105, Dec. 2022. Dr. Ye is an Active Reviewer for many interna-
[114] C. Bechlioulis and G. Rovithakis, “Robust adaptive control of feedback tional journals, including the IEEE T RANSACTIONS
linearizable MIMO nonlinear systems with prescribed performance,” ON AUTOMATIC C ONTROL , IEEE T RANSACTIONS ON S YSTEMS , M AN ,
IEEE Trans. Autom. Control, vol. 53, no. 9, pp. 2090–2099, AND C YBERNETICS : S YSTEMS , IEEE T RANSACTIONS ON N EURAL
Oct. 2008. N ETWORKS AND L EARNING S YSTEMS, and IEEE T RANSACTIONS ON
[115] Y. Cao, C. Wen, and Y. Song, “A unified event-triggered control I NTELLIGENT T RANSPORTATION S YSTEMS.
approach for uncertain pure-feedback systems with or without state
constraints,” IEEE Trans. Cybern., vol. 51, no. 3, pp. 1262–1271,
Mar. 2021. Frank L. Lewis (Life Fellow IEEE) received the
[116] R. Olfati-Saber, and R. Murray, “Consensus problems in networks of Ph.D. degree in electrical engineering from the
agents with switching topology and time-delays,” IEEE Trans. Autom. Georgia Institute of Technology, Atlanta, GA, USA,
Control, vol. 49, no. 9, pp. 1520–1533, Sep. 2004. in 1981.
[117] H. Liang, G. Liu, H. Zhang, and T. Huang, “Neural-network-based He is currently a Distinguished Scholar Professor
event-triggered adaptive control of nonaffine nonlinear multiagent and a Distinguished Teaching Professor with the
systems with dynamic uncertainties,” IEEE Trans. Neural Netw. Learn. University of Texas at Arlington, Fort Worth, TX,
Syst., vol. 32, no. 5, pp. 2239–2250, May 2021. USA, and the Moncrief-O’Donnell Chair with the
[118] G. Chen, F. Lewis, and L. Xie, “Finite-time distributed consensus via University of Texas at Arlington Research Institute,
binary control protocols,” Automatica, vol. 47, no. 9, pp. 1962–1968, Fort Worth. He is involved in feedback control,
2011. intelligent systems, cooperative control systems, and
[119] F. Xiao, L. Wang, J. Chen, and Y. Gao, “Finite-time formation control nonlinear systems. He has authored six U.S. patents, numerous journal spe-
for multi-agent systems,” Automatica, vol. 45, no. 11, pp. 2605–2611, cial issues, journal papers, and 20 books, including optimal control, aircraft
2009. control, optimal estimation, and robot manipulator control.
[120] L. Hu, H. Hu, W. Naeem, and Z. Wang, “A review on COLREGs- Dr. Lewis was a recipient of the Fulbright Research Award, the NSF
compliant navigation of autonomous surface vehicles: From tradi- Research Initiation Grant, the Terman Award from the American Society
tional to learning-based approaches,” J. Autom. Intell., vol. 1, no. 1, for Engineering Education, the Gabor Award from the International Neural
Dec. 2022, Art. no. 100003. Network Society, the Honeywell Field Engineering Medal from the Institute
[121] S. Luo and Y. Song, “Chaos analysis-based adaptive backstepping con- of Measurement and Control, U.K., the Neural Networks Pioneer Award from
trol of the microelectromechanical resonators with constrained output IEEE Computational Intelligence Society, the Outstanding Service Award
and uncertain time delay,” IEEE Trans. Ind. Electron., vol. 63, no. 10, from the Dallas IEEE Section, and the Texas Regents Outstanding Teaching
pp. 6217–6225, Oct. 2016. Award in 2013. He was elected as an Engineer of the year by the Fort Worth
[122] W. Cai, D. Li, and Y. Song, “A novel approach for active adhesion IEEE Section. He was listed in the Fort Worth Business Press Top 200 Leaders
control of high-speed trains under antiskid constraints,” IEEE Trans. in Manufacturing. He is a member of the National Academy of Inventors, a
Intell. Transp. Syst., vol. 16, no. 6, pp. 3213–3222, Dec. 2015. Fellow of the International Federation of Automatic Control and the Institute
[123] K. Kim, P. Spieler, E. Lupu, A. Ramezani, and S. Chung, “A bipedal of Measurement and Control, U.K., a Texas Board of Professional Engineer, a
walking robot that can fly, slackline, and skateboard,” Sci. Robot., vol. 6 U.K. Chartered Engineer, and a Founding Member of the Board of Governors
no. 59, 2021, Art. no. eabf8136. of the Mediterranean Control Association.

Authorized licensed use limited to: Xian Jiaotong University. Downloaded on June 27,2025 at 10:09:34 UTC from IEEE Xplore. Restrictions apply.