1

Strong Prescribed-Time Stabilization of Uncertain Nonlinear

Systems by Periodic Delayed Feedback
Yi Ding, Bin Zhou, Senior Member, IEEE, Kang-Kang Zhang, and Wim Michiels, Senior Member, IEEE

Abstract—In this paper, a novel Lyapunov-based approach for strong multiple variables and function relationships limit the application
prescribed-time stabilization by periodic delayed feedback is established. of this method in practical systems. In view of these problems,
Since the comparison lemma cannot be directly applied to time-delay
prescribed-time stabilization has been proposed, where the settling-
systems, their proofs rely on trajectory analysis. Based on this approach,
a novel control law for strong prescribed-time stabilization of uncertain time can be chosen in advance. Existing Lyapunov-based criteria that
scalar nonlinear systems is obtained, with the appealing properties that i) can achieve prescribed-time stabilization are mainly based on time
singularity problems inherent to time-varying high gain approaches are varying high-gain feedback (see, for example, [5], [22], [27], [30],
avoided, ii) strong prescribed-time stabilization is achieved with control [32], [33] and [34]).
terms exhibiting a linear growth rate in the combined current and delayed
state variable, iii) the achieved fixed-time stability is preserved under However, since an infinite gain is practically impossible to achieve,
classes of additive perturbations, and iv) the setting time of the closed- such kind of controllers are not well-defined at and after the
loop system equals the prescribed value for some admissible uncertainties. prescribed settling time, namely, these approaches suffer from a
Subsequently, using the backstepping procedure, a strongly prescribed- singularity problem. For this problem, the current general approach is
time stabilizing control law for strict feedback uncertain nonlinear
to turn off the time-varying function before the system state reaches
systems is designed. Numerical simulations are shown to verify the
effectiveness of the proposed approaches. the equilibrium, and then replace it with a time-invariant function
Keywords: Prescribed-time stabilization; Strong prescribed-time sta- (see Remark 2 in [32]). This approach does avoid an infinite gain,
bilization; Periodic delayed feedback; High-order nonlinear systems. and thus it avoids the singularity problem. Nonetheless, this approach
will sacrifice the regulation accuracy when implemented.
I. I NTRODUCTION Time delays are inevitable in practical engineering [10], [12]. Time
delays usually have a negative impact on the stability and robustness
Compared with asymptotic stability, finite-time stability has the ad- of the system [9], [26], and time-delay systems are considered to
vantages of faster convergence speed, higher precision, and stronger be far more complex than time-delay-free ones (see [1] and [2]).
robustness [11], [15], [28]. The concept of finite-time stability was However, the impact of time delay on the system is not always
first proposed in the 1960s [25]. The most common existing ap- detrimental. As shown in [24], researchers sometimes artificially add
proaches for achieving finite-time stability are based on terminal time delays to the controllers to improve the performance of the
attractors [31] and homogeneity theory [4]. A rigorous theory of closed-loop system.
finite-time stability was first proposed in [3]. This approach can Recently, particularly relevant for this paper, a periodic delayed
indeed realize finite-time stability of the closed-loop systems, that is, feedback (PDF) scheme for linear systems was proposed in [35] to
the settling-time is a constant less than infinity, which is a stronger achieve prescribed-time stabilization. An appealing feature of the
notion than asymptotic stability (exponential stability). PDF is that the time-varying controller gain is bounded and can
It should be emphasized that the finite-time stability achieved by be chosen continuous, continually differentiable and even smooth
the above criteria has an obvious shortcoming, that is, the settling- [35], [36]. The price to pay is that in the presence of uncertainty on
time of the system state depends on the initial value, and when the the model, fixed-time stability cannot be guaranteed anymore. What
initial value of the state is far away from the equilibrium point, impedes a robustification of the control law, relying on Lyapunov
its settling-time will also be very large, which greatly limits the stability theory and a comparison principle, is that for most time-
application of these approaches [18]. Therefore, aiming at the above delay systems, the comparison lemma does not hold [17]. Therefore,
shortcoming, the concept of fixed-time stability has been proposed, our goal in this paper is to find another way to combine PDF with
which is a novel notion of finite-time stability, where the upper bound Lyapunov stability theorem, and extend the PDF scheme to nonlinear
of the settling-time is a constant that does not depend on the initial systems. Briefly, the major contributions of this paper can be stated
condition. Recently, a large number of theories for fixed-time stability as follows.
have emerged and have been applied to many different systems (see, 1) Novel Lyapunov-based criteria for fixed-time stability and strong
for example, [6], [13], [14], [19], [20] and [37]). fixed-time stability in the presence of delays are established.
It is worth noting that, for fixed-time control, the upper bound Unlike the existing Lyapunov-based approaches for finite-time
of the settling time is a complex function of system parameters. stability mentioned above using the comparison lemma, the
The complex engineering situations, the cumbersome adjustment of proofs rely on trajectory analysis. In addition, compared to exist-
This work was supported in part by the National Natural Science Foundation ing Lyapunov-based approaches for prescribed-time stabilization
of China for Distinguished Young Scholars under Grant 62125303, the Science based on time-varying high-gain feedback, we avoid an infinite
Center Program of National Natural Science Foundation of China under gain by including a time delay term.
Grant 62188101, the Fundamental Research Funds for the Central Universities 2) A novel nonsingular strongly prescribed-time stabilizing control
under Grant HIT.BRET.2021008, the Internal Funds KU Leuven under Grant
C14/22/092 and the Research Foundation-Flanders (FWO - Vlaanderen) under law for uncertain scalar nonlinear systems is proposed by using
Grant G092721N. the proposed Lyapunov-based approach, whose terms exhibit a
(Corresponding author: Bin Zhou) linear growth rate in the combined current and delayed state
Yi Ding, Bin Zhou and Kang-Kang Zhang are with the Center for Control variable, which preserves the original advantage of the PDF.
Theory and Guidance Technology, Harbin Institute of Technology, Harbin,
150001, China. (Email: [email protected]; [email protected].)
Besides, the closed-loop system is still fixed-time stable under
Wim Michiels is with the Department of Computer Science, KU Leuven, classes of additive perturbations, which overcomes the original
B-3001 Leuven, Belgium. (e-mail: [email protected]) shortcoming of the PDF.
 2

3) For strict feedback nonlinear systems with matched uncertain- increase to infinity and are not well defined afterwards, which may
ties, a novel strongly prescribed-time stabilizing control law lead to implementation problems in practical applications.
based on backstepping is proposed, which shows that, with the It is worth noting that the smooth PDF can also achieve prescribed-
proposed approach, the control law can indeed achieve strong time stabilization of linear systems and avoids the singularity problem
prescribed-time stabilization with the exact setting time being mentioned above. To recall the PDF approach, we give some prelim-
a prescribed number for some admissible uncertainties, while inary results.
offsetting the effect of matching uncertainties. Definition 1: [35] Let h > 0 be a given constant and r ≥ 0 be a
Notation 1: For any x ∈ Rn , we denote x = [x1 , x2 , . . . , xn ]T ∈ given integer (extended with infinity). A function Rh (·) : R → R is
R , xi = [x1 , x2 , . . . , xi ]T ∈ Ri and x[a,b] = x(s), s ∈ [a, b] . A
n said be an S(r) (h) function if
continuous function ϕ : [0, α) → [0, ∞) is said to belong to class K 1) Rh (·) is 2h-periodic and Rh (t) = 0, ∀t ∈ [0, h] ;
if it is strictly increasing and ϕ (0) = 0, where α > 0 is a constant 2) Rh (t) ≥ 0, ∀t ∈ [h, 2h] and there is an h∗ ∈ (h, 2h), such that
(can be infinity). Cr is the set of r-time differentiable functions, Rh (h∗ ) > 0;
(i) (i)
where r ≥ 0 is an integer. Chn is the set of continuous functions 3) Rh ∈ Cr , which implies that Rh (h) = Rh (2h) = 0, i =
ψ : [−h, 0] → Rn with h > 0. 0, 1, . . . , r.
Consider the following scalar system
II. M OTIVATION AND P ROBLEM S TATEMENT ẋ(t) = u(t), x(0) = x0 , t ≥ 0, (3)
Consider the following nonlinear system where u(t) is the control input. A PDF with a prescribed degree of
smoothness that can achieve prescribed-time stabilization is given as
ẋ = f (t, x), x(t0 ) = x0 , t ≥ t0 , (1)
follows.
T n
where x = [x1 , x2 , . . . , xn ] ∈ R is the state vector, x0 is the Lemma 1: [35] Consider the following PDF control
initial condition, and f : [t0 , ∞) × Rn → Rn is a continuous u(t) = −ax(t) − K(a,h) (t)x(t − h), (4)
function which satisfies f (t, 0) = 0, ∀t ≥ t0 . Assume that the
nonlinear system (1) has a unique solution for any given initial where a ≥ 0, h > 0 are two constants, x(θ) is an arbitrary bounded
condition x0 . function for θ ∈ [−h, 0),
So far, for the finite-, fixed- and prescribed-time stability 1 of K(a,h) (t) = Rh (t)W e−a(h−2t) , (5)
system (1), a large number of criteria based on the Lyapunov stability
theorem and the comparison lemma have emerged, whose main idea in which W = Wc−1 (a, h) is defined by
is described as follows. Z 2h
Select a positive-definite Lyapunov function V (t, x) : [t0 , ∞) × Wc (a, h) = e2as Rh (s)ds, (6)
h
Rn → R≥0 , which satisfies
with Rh (t) ∈ S(r) (h). Then the state of the closed-loop system
∂V (t, x) ∂V (t, x)
V̇ (t, x)|(1) , + f (t, x) ≤ g(t, V (t, x)), (2)
∂t ∂x ẋ(t) = −ax(t) − K(a,h) (t)x(t − h), t ≥ 0, (7)
where g(t, s) : [t0 , ∞) × R → R is a function. Suppose that the consisting of (3) and (4) satisfies x(t) = 0, ∀t ≥ T = 2h and
scalar system ż = g(t, z) is finite- or fixed- or prescribed-time x(t) 6= 0, ∀t ∈ [0, T ) whenever x(0) 6= 0.
stable. Then, by using (2), the Lyapunov stability theorem and the However, coming back to (2), when g(t, s) is replaced by
comparison lemma, it is obtained that the state of (1) is finite- or g(t, s, s[t−h,t] ), invoking a comparison lemma is not possible due
fixed- or prescribed-time stable. Different choices of the function to the presence of time delay. In fact, for some time-delay systems
g(t, s) commonly used in the literature are collected in Table 1 with special forms, the comparison lemma still holds (for example,
(we only consider t ∈ [t0 , T + t0 ) for prescribed-time stability), see Lemma 2 in [17]). Unfortunately, these forms are not suitable for
where T is the prescribed time and the other parameters satisfy PDF.
0 < α < 1, β > 1, c > 0, d > 0, k > 0, 0 < kv1 < 1, kv2 > This paper is dedicated to generalize the approach shown in Lemma
h
1, µ1 (t) = T +tc0 −t , µ2 (t) = (T +tT −t)h , h > 2. 1 to nonlinear system (1), that is, solve the strong prescribed-time
0
stabilization problem for nonlinear systems by PDF. To this end, we
Table 1: Different choices of the function g(t, s)
will first introduce the definitions for (strong) fixed-time stability of
No. g(t, s) Stability Type References time-delay systems, and then propose novel criteria for (strong) fixed-
1 g1 = −csα finite-time [3] time stability. As an application of these criteria, a novel strongly
2 g2 = −csα − ds fast finite-time [28] prescribed-time stabilizing control law for scalar nonlinear systems
3 g3 = −csα − dsβ fixed-time [29] will be designed firstly. Then, we will generalize it to higher order
4 g4 = −(csv1 + dsv2 )k fixed-time [23] uncertain nonlinear systems.
5 g5 = −µ
 1 (t)s  prescribed-time [33]
6 g6 = −c − 2 µ̇2 (t)
µ2 (t)
s prescribed-time [30] III. F IXED -T IME S TABILITY C ONCEPTS AND P RESCRIBED -T IME
S TABILIZATION P ROBLEMS
Although the control laws designed based on g5 (t, s) and g6 (t, s)
can achieve prescribed-time stabilization, there is a singularity prob- A. Definitions for Fixed-Time Stability and Formulations for
lem, that is, when t → T + t0 , both functions µ1 (t) and µ2 (t) Prescribed-Time Stabilization
Consider the following nonlinear delay system,
1 We point out that the term “prescribed-time stability”, though frequently
used in the literature, is not very precise since the adjective “prescribed” is ẋ = f (t, x[t−h,t] , u), x[t0 −h,t0 ] = ψ, t ≥ t0 , (8)
rather related to control design than to analysis, e.g. one may be able to T
stabilize a system in any prescribe time by a suitable control law. To be
where u ∈ Rm is the control input, x = [x1 , x2 , . . . , xn ] ∈ Rn is
consistent with the literature, we maintain the existing terminology in this the state vector, ψ ∈ Chn is the initial condition, f : [t0 , ∞) × Chn ×
section. Rm → Rn is a function that is piecewise continuous with respect
 3

x (t) x (t)

x[t0 −h,t0 ] = ψ0
d = d0
x[t0 −h,t0 ] = ψ0
x[t0 −h,t0 ] = ψ1
x[t0 −h,t0 ] = ψ1
d = d1

t0 − h t0 x[t0 −h,t0 ] = ψ2
t0 + T t t0 − h t0 t0 + T t
x[t0 −h,t0 ] = ψ2
d = d2

Fig. 1. Illustration of the exact fixed-time stability

Fig. 2. Illustration of the strong fixed-time stability

to t, locally Lipschitz in ψ and u, and f (t, 0, 0) = 0, ∀t ≥ t0 . We

assume that (8) has a unique solution for any given initial condition such that all possible values of d(t, x[t−h,t] ) belong to this set. For
and piece-wise continuous input. example, we can usually find a known function ρ(t, x[t−h,t] ) such
In this paper we only consider global stability for simplicity. We that d(t, x[t−h,t] ) ≤ ρ(t, x[t−h,t] ) [6].
first give the definition for (exact) fixed-time stability. For (9) it is too strict to require exact fixed-time stability with a
Definition 2: [23] (Fixed-time stability) The time-delay system (8) settling time independent of the realization of the uncertainty. The
with u = 0 is said to be fixed-time stable (FxTS) with the settling following notion of fixed-time stability is more appropriate.
time T < ∞ if Definition 4: (Strong fixed-time stability) The uncertain time-delay
1) It is stable in the usual sense. system (9) with u = 0 is said to be strongly fixed-time stable (SFxTS)
2) It holds that x(t, ψ) = 0 for all t ≥ T + t0 and for all ψ ∈ with the settling time T < ∞ if it is FxTS with settling time T for any
Chn , where x(t, ψ) is the solution of (8) for u = 0 and initial d(t, x[t−h,t] ) ∈ D, and there exists a function d0 (t, x[t−h,t] ) ∈ D,
condition ψ. such that the closed loop system is EFxTS with the settling time T
when d(t, x[t−h,t] ) = d0 (t, x[t−h,t] ).
Definition 3: (Exact fixed-time stability) The time-delay system
A graphical illustration of the SFxTS is shown in Fig. 2.
(8) with u = 0 is said to be exact fixed-time stable (EFxTS) with
Compared with “free-will strong arbitrary time stability” [22],
the settling time T < ∞ if it is FxTS with the settling time T and,
which requires the settling time to be independent of any system
additionally, there exists a bounded initial condition x[t0 −h,t0 ] =
parameters and initial conditions, Definition 4 only requires the
ψ0 ∈ Chn , such that T0 (ψ0 ) = T , where
existence of an initial value and an uncertainty for which the settling
T0 (ψ) = inf{T∗ ≥ 0 : x(t, ψ) = 0, ∀t ≥ T∗ + t0 } time equals the prescribed time T (see Fig. 2). In addition, Definition
4 is also applicable to time-delay systems, and is thus more general.
is the settling time of (8) with initial condition ψ at t = t0 [21].
Finally, making the leap from analysis to synthesis and allowing
Remark 1: Unlike Definition 3, Definition 2 only requires that the
the value of T to be arbitrarily assigned and arbitrarily small, bring
settling time of all solutions is upper bounded by T . In this way,
us to the main problem addressed in the paper.
abstraction is made of its actual value, but the conditions are easier
Problem 2: (Strong prescribed-time stabilization) Find a family of
to check, as for nonlinear system it is often hard to obtain the settling
control laws u(t) = u(t, T, x[t−h,t] ) parametrized by T > 0, such
time and one resorts to deriving upper bounds instead. On the other
that the closed-loop system (9) and u(t) = u(t, T, x[t−h,t] ) is SFxTS
hand, if the prescribed settling time is reduced towards zero, a more
with the settling time T .
aggressive control actions is unavoidable. By ensuring EFxTS, we
For the sake of brevity, we set t0 = 0 hereafter.
aim at reaching convergence in T secs rather then at most T secs,
which could be an indication of more efficiently using the controls.
A graphical illustration of the EFxTS is shown in Fig. 1. B. Fixed-Time Stability Theorems
It is thus natural to consider the following synthesis problem. Considering system (8), we first give the following Lemma.
Problem 1: (Exact prescribed-time stabilization) Find a family Lemma 2: Consider system (8) with u = 0. Suppose that there is
of control laws u(t) = u(t, T, x[t−h,t] ) parametrized by T > 0, a positive definite function V (t, x(t)) such that
such that the closed-loop system consisting of (8) and u(t) =
u(t, T, x[t−h,t] ) is EFxTS with the settling time T . V̇ (t, x(t))|(8) ≤ −aV (t, x(t)) − K(a,h) (t)V (t − h, x(t − h)), (10)
The term “prescribed-time” is in accordance with the requirement
where a ≥ 0 and h > 0 are two constants, and K(a,h) (t) is defined
that T is allowed to be arbitrarily assigned and arbitrarily small.
in (5). Then system (8) with u = 0 is FxTS with the settling time
Problem 1 has been solved in [35] if system (8) is linear, and will
T = 2h.
not be considered in this paper.
Proof. It is noticed that V (t, x(t)) ≥ 0, V̇ (t, x (t)) ≤ 0. Thus by
Now we consider the uncertain nonlinear time-delay system
using the Lyapunov stability theory, it can be simply concluded that
ẋ = f (t, x[t−h,t] , u) + d(t, x[t−h,t] ), x[t0 −h,t0 ] = ψ, t ≥ t0 , (9) system (8) with u = 0 is stable.
T Next we prove T (ψ) ≤ T. Denote V0 = V (0, x(0)). We can write
where u ∈ Rm is the control input, x = [x1 , x2 , . . . , xn ] ∈ Rn is
the state vector, ψ ∈ Chn is the initial condition, d(t, x[t−h,t] ) ∈ D is V̇ (t, x(t)) = −aV (t, x(t)) − K(a,h) (t)V (t − h, x(t − h)) − ϕ(t),
an unknown function, in which D is a known set, f : [t0 , ∞)×Chn × (11)
Rm → Rn is a known function that is piecewise continuous with where ϕ(t) is a function satisfying ϕ(t) ≥ 0.
respect to t, locally Lipschitz in ψ and u, and f (t, 0, 0) = 0, ∀t ≥ t0 . For t ∈ [0, h) , in view of (11) and K(a,h) (t) = 0, ∀t ∈ [0, h), we
Remark 2: Although the uncertainties of the system cannot be have
described precisely, it is usually possible to find a certain set D, V̇ (t, x(t)) = −aV (t, x(t)) − ϕ(t),
 4

which implies Theorem 1: Suppose that there is a positive definite function

Z t V (t, x(t)) such that
V (t, x(t)) = e−at V0 − e−a(t−s) ϕ(s)ds ≤ e−at V0 , (12) a
0 V̇ (t, x(t))|(8) ≤ − V (t, x(t))
1−τ
For t ∈ [h, 2h] , in light of (11) and (12), we know that K(a,h) (t) τ
− V (t, x(t))V 1−τ (t − h, x(t − h)),
Z t 1−τ
V (t, x(t)) =e−a(t−h) V (h, x(h)) − e−a(t−s) ϕ(s)ds (15)
h
Z t where a ≥ 0, h > 0 and 0 < τ < 1 are some constants, and
− e−a(t−s) K(a,h) (s)V (s − h, x(s − h))ds K(a,h) (t) is defined in (5). Then system (8) with u = 0 is FxTS
h
Z t with the settling time T = 2h.
−at
≤e V0 − e−a(t−s) ϕ(s)ds Proof. It is noticed that V (t, x(t)) ≥ 0, V̇ (t, x(t)) ≤ 0. Thus by
h
Z t using the Lyapunov stability theory, it can be simply concluded that
−a(t−s) system (8) with u = 0 is stable.
− e K(a,h) (s)V (s − h, x(s − h))ds
h
 Z t  Then we prove that there exists a constant t1 ∈ [0, T ] , such
=e−at V0 1 − eah K(a,h) (s)ds + Π (t), (13) that x(t1 ) = 0, and we do so by contradiction. Assume it holds
h that x(t) 6= 0, ∀t ∈ [0, T ] . Consider the positive definite func-
where tion W (t, x(t)) = V 1−τ (t, x(t)) ≥ 0, namely, V (t, x(t)) =
1
Z t W 1−τ (t, x(t)). Thus we have
Π (t) , − e−at eas ϕ(s)ds 1 1
0 V̇ (t, x(t)) = W 1−τ −1 (t, x(t))Ẇ (t, x(t))
Z t Z s−h 1−τ
+ e−at eah K(a,h) (s) eaσ ϕ(σ)dσds. a 1
≤− W 1−τ (t, x(t))
h 0 1−τ
Notice that K(a,h) (t) τ
− W 1−τ (t, x(t))W (t − h, x(t − h)).
Z t 1−τ
Π (t) ≤ − e−at eas ϕ(s)ds Since x(t) 6= 0, ∀t ∈ [0, T ], we have
0
Z t Z t
+ e−at eaσ ϕ(σ)dσ
eah K(a,h) (s)ds Ẇ (t, x(t)) ≤ − aW (t, x(t))
0 h
Z t  Z t  − K(a,h) (t)W (t − h, x(t − h)), ∀t ∈ [0, T ] .
= − e−at eas ϕ (s) ds 1 − e2as Rh (s)dsW
0 h By using Lemma 2, we know that x(T ) = 0, which contradicts
≤0. the assumption. Then there exists a constant t1 ∈ [0, T ] such that
x(t1 ) = 0.
Thus Finally we prove that x(t) = 0 always holds when t ≥ t1 . Notice
 Z t  that from x(t1 ) = 0, V (x(t)) ≥ 0 and V̇ (x(t)) ≤ 0, we get x(t) =
V (t, x(t)) =e−at V0 1 − eah K(a,h) (s)ds + Π (t) 0, ∀t ≥ t1 . Therefore x(t) = 0, ∀t ≥ T. The proof is finished.
 Zht  Corollary 1: Consider system (9) with u = 0. If (15) holds and
−at
≤e V0 1 − eah K(a,h) (s)ds , (14) there exists a function d0 (t, x[t−h,t] ) ∈ D, such that
h
a
It is obtained from (14) that V (T, x(T )) ≤ 0. Then according to V̇ (t, x(t))|(9) = − V (t, x(t))
1−τ
V (T, x(T )) ≥ 0, we know that V (T, x(T )) = 0, namely, x(T ) = 0. K(a,h) (t) τ
Then, by using x(T ) = 0, V (t, x(t)) ≥ 0 and V̇ (t, x(t)) ≤ 0, we − V (t, x(t))V 1−τ (t − h, x(t − h)),
1−τ
have x(t) = 0, ∀t ≥ T. The proof is finished.
Remark 3: Unlike other in approaches to ensure finite- and fixed- holds when d(t, x[t−h,t] ) = d0 (t, x[t−h,t] ), then the system (9) with
time stability, Lemma 2 cannot be directly obtained from Lemma 1 u = 0 is SFxTS with the settling time T = 2h.
by invoking a comparison lemma, due to the presence of the delay Proof. According to Theorem 1, system (9) with u = 0 is FxTS with
term. Therefore, we have adopted a trajectory based technique. the settling time T . Next we prove that there exists a constant T > 0
Remark 4: It should be emphasized that, although the proof of and a bounded initial condition x[t0 −h,t0 ] = ψ0 ∈ Chn , such that
Lemma 1 (see [35] for details) is also based on trajectory analysis, T0 (ψ0 ) = T, where T0 (ψ0 ) is the settling time of system (9).
all expressions and conditions are in terms of equalities involving For t ∈ [0, h] , it holds that
the state variable. An implication is that fixed-time stability cannot a
V̇ (t, x(t)) = − V (t, x(t)),
be guaranteed anymore when the system contains uncertainties (see 1−τ
Remark 4 in [35]). Lemma 2 is combined with the Lyapunov stability and thus
a
theory, and the equations in the proof of Lemma 1 are replaced V (t, x(t)) = e− 1−τ t V (0, x(0)). (16)
by inequalities (12) and (14) involving the Lyapunov function. The
resulting conditions for ensuring fixed-time stability are obviously For t ∈ [h, 2h] , we denote
weaker than Lemma 1, which will enable us to maintain fixed-time a
W (t, x(t)) , e 1−τ t V (t, x(t)), (17)
stability in the presence of system uncertainties.
In theory, we can use Lemma 2 to design controllers. However, whose time-derivative can be written as
since the second term in the right-hand side of inequality (10) does a a a
not depend on x(t), the most commonly used Lyapunov function, Ẇ (t, x(t)) =e 1−τ t V̇ (t, x(t)) + e 1−τ t V (t, x(t))
1−τ
that is, quadratic function, is not suitable for Lemma 2. Therefore, K(a,h) (t) a t τ
following [8], we will make some improvements on Lemma 2. =− e 1−τ V (t, x(t))V 1−τ (t − h, x(t − h))
1−τ
 5

K(a,h) (t) ah τ
=− e W (t, x(t))W 1−τ (t − h, x(t − h)). 1
1−τ
According to (16) and (17), for t ∈ [h, 2h] , we have
10-7
K(a,h) (t) ah 1−τ 0.5 8
Ẇ (t, x(t)) = − e V (0, x(0))W τ (t, x(t)). 6
1−τ 4
2
Then it can be obtained that 0
 Z t  3.6 3.8 4
1 0
W (t, x(t)) = 1− e K(a,h) (s)ds V 1−τ (0, x(0)),
ah
0 1 2 3 4 5
1−τ h

namely,
Fig. 3. The state x of the system in Example 1
a
e− 1−τ t t
 Z 
V (t, x(t)) = 1− eah K(a,h) (s)ds V 1−τ (0, x(0)).
1−τ h
(18) Remark 5: It is obtained from (20) that when τ = 1/2, ρ(t, x) =
For x(0) 6= 0, it holds that V (0, x(0)) 6= 0. Then in light of f (t, x) = 0 and g(t, x) = 1, (20) reduces to (4), which implies
(16) and (18), we know that V (t, x(t)) 6= 0, ∀t ∈ [0, T ). It follows that (4) is actually the limit case for τ = 1/2. If τ ∈ (1/2, 1),
that x(t) 6= 0, ∀t ∈ [0, T ) . Thus for any ψ0 ∈ Chn which satisfies then x(t − h) in the second term of the right-hand side of (4) is
x(0) 6= 0, we have x(t) 6= 0, ∀t ∈ [0, T ) . Then T0 (ψ0 ) = T, which replaced by x2τ −1 (t)x2−2τ (t − h). Note that the powers taken from
implies the system (9) with u = 0 is EFxTS with the settling time x(t) and x(t − h) are both positive and their sum is equal to one.
T . The proof is finished. This ‘distribution of unit power’ over x(t) and x(t − h) is key to
the rejection of additive perturbations, thereby removing the main
shortcoming of the original control law (4).
C. Strong Prescribed-Time Stabilization of Uncertain Scalar Nonlin-
ear Systems Remark 6: Different from [29] and [23], which need higher order
term xβ (t), β > 1 in the control law to drive the state to zero within a
In this subsection, we will design a strongly prescribed-time fixed time, in Theorem 2, the term x2τ −1 (t)x2−2τ (t−h) whose total
stabilizing control law (solution to Problem 2) for uncertain scalar power equal to 1 is sufficient, which implies that the stabilizing terms
nonlinear systems by using Theorem 1. have linear growth rate in the combined argument [x(t); x(t − h)] .
Consider the following uncertain scalar nonlinear system Example 1: Consider the following uncertain nonlinear system
ẋ(t) = f (t, x) + g(t, x)u(t) + d(t, x), x(0) = x0 , t ≥ 0, (19)
ẋ = x3 + (1 + x2 )u + d(t, x),
where u(t) ∈ R is the control input, f (t, x) ∈ R, f (t, 0) = 0
and g(t, x) ∈ R, g(t, x) 6= 0, ∀t ≥ 0, ∀x ∈ R are two known where d(t, x) ≤ 0.1 |x| and the initial condition is x(0) = 1. Let
functions, and d(t, x) is an unknown function satisfying |d(t, x)| ≤ T = 4s. According to Theorem 2, we can design the control law
ρ(t, x) |x| , where ρ(t, x) ∈ R is a known function. The following  
theorem provides a solution to Problem 2. 1 a
u= − − 0.1 x(t)
Theorem 2: Let T > 0 be a prescribed time. Consider the control 1 + x2 2(1 − τ )

law K(a,h) (t) 2τ −1
− x (t)x2−2τ (t − h) − x3 ,

a
 2(1 − τ )
−1
u =g (t, x) − − ρ(t, x) x(t)
2(1 − τ ) where a, τ and h are three constants satisfying τ = 5/7, a =

K(a,h) (t) 2τ −1 0.1, h = 2, and K(a,h) (t) is defined by (5) (we only consider
− x (t)x2−2τ (t − h) − f (t, x) , (20)
2(1 − τ ) t ∈ [0, 2h]) with
where a ≥ 0, h = T /2, K(a,h) (t) is defined in (5), and τ ∈ (1/2, 1) 
0, t ∈ [0, h) ,
is a constant obtained by dividing two odd numbers. Then the closed- Rh (t) =
sin2 πt


h
, t ∈ [h, 2h] .
loop system consisting of (19) and (20) is SFxTS with the settling
time T . The simulation results are shown in Fig. 3 and Fig. 4, from which
Proof. Choose the Lyapunov function V (x(t)) = x2 (t). Then it is we can see that for d(t) = 0.1e−t x, the closed-loop system is FxTS
obtained from (19) and (20) that and the settling-time is around 3.5s. For d(t) = 0.1x, the closed-loop
V̇ (x(t)) =2x(t)(f (t, x) + g(t, x)u(t) + d(t, x)) system is EFxTS with the settling time T = 4s.
1
≤ x2τ (t)(−ax2−2τ (t) − K(a,h) (t)x2−2τ (t − h))
1−τ 0
aV (x(t)) K(a,h) (t) τ
=− − V (x(t))V 1−τ (x(t − h)). -0.2
1−τ 1−τ
(21) -0.4

By using Theorem 1, we know that the closed-loop system consisting -0.6

of (19) and (20) is FxTS with the settling time T . -0.8
If d(t, x) = ρ(t, x)x, then it is obtained from (21) that
-1
aV (x(t)) K(a,h) (t) τ 0 1 2 3 4 5
V̇ (x(t)) = − − V (x(t))V 1−τ (x(t − h)).
1−τ 1−τ
According to Corollary 1, we know the closed-loop system consisting Fig. 4. The control input u for the system in Example 1
of (19) and (20) is SFxTS with the settling time T .
 6

IV. S TRONG P RESCRIBED -T IME S TABILIZATION OF H IGHER where 21 < τ < 1 is a constant obtained by dividing two odd
O RDER N ONLINEAR U NCERTAIN S YSTEMS numbers, and

In this Section, we will generalize Theorem 2 to higher order 0, t ≤ 0,
Kn (t) =
uncertain nonlinear systems. Due to space limitations, we only Kn(a,h) (t), t > 0,
consider the following single input strict feedback nonlinear systems
with matched uncertainties in which Kn(a,h) (t) is defined by (5) with Rh (t) ∈ S(r) , r ≥ 0.

ẋ1 = F1 (t, x1 ) + G1 (t, x1 )x2 ,

 B. Strong Fixed-Time Stability Analysis


 ẋ2 = F2 (t, x2 ) + G2 (t, x2 )x3 ,
Based on the design in Subsection IV-A, we can give the following

.. (22)
 . conclusion.

 ẋn−1 = Fn−1 (t, xn−1 ) + Gn−1 (t, xn−1 )xn ,

 Theorem 3: Let Assumption 1 be satisfied, and let T > 0 be the

ẋn = Fn (t, x) + Gn (t, x)u + φ(t, x), prescribed time. Then the closed-loop system consisting of (22) and
where φ : R≥0 × Rn → R is an unknown nonlinear function (26) is SFxTS with the settling time T .
satisfying φ(t, 0) = 0, ∀t ≥ 0, Fi : R≥0 × Ri → R is some Proof. We first prove that for i = 1, 2, . . . , n, it holds that zi (t) =
known smooth nonlinear functions satisfying Fi (t, 0) = 0, and 0, ∀t ≥ 2(n − i + 1)h. For clarity, we divide the proof into n steps.
Gi : R≥0 × Ri → R is some known smooth nonlinear functions Step 1: Substituting (26) into (25), and using Assumption 1, we
satisfying Gi (t, xi ) 6= 0, ∀xi ∈ Ri , i = 1, 2, . . . , n. get
a Kn (t) 2(1−τ )
V̇ = − zn2 (t) − zn (t − h)zn2τ (t)
A. Design of the Controller 2(1 − τ ) 2(1 − τ )
+ zn φ(t, x) − |zn | ρ(t, x)
In this subsection, we will consider strict feedback nonlinear
systems with matched uncertainties, and design a strongly prescribed- a K(a,h) (t) τ
≤− V (x(t)) − V (x(t))V 1−τ (x(t − h)).
time stabilizing control law u(t) = u(t, T, x[t−T,t] ) (solution to 1−τ 1−τ
(27)
Problem 2) by backstepping.
The following assumption is imposed on system (22). According to Theorem 1, we know that zn (t) = 0, ∀t ≥ 2h.
Assumption 1: [6] There exist two known bounded nonnegative Step i, i = 2, 3, . . . , n: Suppose that at Step i − 1, it holds that
functions ρ(t, x) such that zn−i+2 (t) = 0, ∀t ≥ 2(i − 1)h. It is obtained from (24) that
|φ(t, x)| ≤ ρ(t, x). żn−i+1 (t) = − azn−i+1 (t)
T − Kn−i+1 (t)zn−i+1 (t − h), ∀t ≥ 2(i − 1)h,
Let T > 0 be a prescribed time, and define h = 2n
.
The virtual
controllers are represented as x1d , x2d , . . . , xnd where x1d = 0 and which can be further written as
x2d , x3d . . . , xnd are some functions to be designed later. Then we
define zi = xi − xid , i = 1, 2, . . . , n. σ̇n−i+1 (κn−i+1 ) = − aσn−i+1 (κn−i+1 )
Step i, i = 1, 2, . . . , n − 1: we can get − Kn−i+1(a,h) (κn−i+1 )σn−i+1 (κn−i+1 − h),

żi = Fi (t, xi ) + Gi (t, xi )xi+1d + Gi (t, xi )zi+1 − ẋid . by denoting κn−i+1 = t − 2(i − 1)h and σn−i+1 (κn−i+1 ) =
σn−i+1 (t − 2(i − 1)h) = zn−i+1 (t − h). Then, according to
Now we design Lemma 1, we get σn−i+1 (κn−i+1 ) = 0, ∀κn−i+1 ≥ 2h, namely,
xi+1d = G−1 zn−i+1 (t) = 0, ∀t ≥ 2ih.
i (t, xi )(−azi −Ki (t)zi (t−h)−Fi (t, xi )+ ẋid ), (23)
Next we prove that zi (t) = 0, ∀t ≥ 2(n − i + 1)h, i = 1, 2, . . . , n,
where implies that x(t) = 0, ∀t ≥ T. We divide the proof into n − 1 steps.

0, t ≤ 2(n − i)h, Step i, i = 1, . . . , n − 1: Suppose that
Ki (t) =
Ki(a,h) (t − 2(n − i)h), t > 2(n − i)h, xi = 0, ∀t ≥ T. (28)
(r)
in which Ki(a,h) (t) is defined by (5) with Rh (t) ∈ S , r ≥ n − i. Denote
Then it is obtained that
ξj = −azj−1 − Kj−1 (t)zj−1 (t − h) − Fj−1 (t, xj−1 ) + ẋj−1d ,
żi = −azi − Ki (t)zi (t − h) + Gi (t, xi )zi+1 . (24)
where j = 2, 3, . . . , i. Then xjd = G−1
j−1 (t, xj−1 )ξj , and it follows
Step n: Notice that that
k
żn = Fn (t, x) + Gn (t, x)u + φ(t, x) − ẋnd . dk ξ j dk zj−1 X dk−m Kj−1 (t) dm zj−1 (t − h)
k
=−a k
−
d t d t dk−m t dm t
Consider the positive-definite Lyapunov functions V = 12 zn2 , whose m=0

time-derivative can be written as dk Fj−1 (t, xj−1 ) dk+1 xj−1d

− + ,
dk t dk+1 t
V̇ =zn żn k k−m −1
dk xjd X d Gj−1 (t, xj−1 ) dm ξj
=zn ẋnd + zn (Fn (t, x) + Gn (t, x)u + φ(t, x)). (25) = ,
dk t m=0
dk−m t dm t
Now we design the control input
dk K
j−1 (t)
 in which k = 0, 1, . . . , i − j + 1. In view of dk t
= 0, ∀t ∈
a
u =G−1
n (t, x) − zn (t) + ẋnd − Fn (t, x) [T, T + h] , z j−1 (t) = 0, ∀t ≥ T and x1d = 0, we know that ẋid =
2(1 − τ )
 0, ∀t ≥ T. Therefore, it is obtained from (23) thatxi+1d = 0, ∀t ≥
Kn (t) 2(1−τ )
−ρ(t, x)sign(zn ) − zn (t − h)zn2τ −1 , (26) T. Thus
2(1 − τ ) xi+1 = zi+1 + xi+1d = 0, ∀t ≥ T,
 7

which implies xi+1 (t) = 0, ∀t ≥ T. |φ0 (t, x)| ≤ ρ(t, x), such that the settling time of the closed-loop
Notice that x1 (t) = z1 (t) = 0, ∀t ≥ T , which satisfies (28) when system satisfies T0 (x[t0 −h,t0 ] ) = T , when x[t0 −h,t0 ] = x0 [t0 −h,t0 ]
i = 1. Thus we have x(t) = 0, ∀t ≥ T. and φ(t, x) = φ0 (t, x).
Next we prove that the closed-loop system consisting of (22) Consider φ(t, x) = 0 and x[t0 −h,t0 ] = x0 [t0 −h,t0 ] which satisfies
and (26) is FxTS with the settling time T . It has been proven that x2d (0) 6= 0 and xid (0) = 0, i = 3, 4, . . . , n + 1, it is obtained that
x(t) = 0, ∀t ≥ T , namely, the settling time of the closed-loop system x1 (0) 6= 0. Notice that
T0 (x[−h,0] ) ≤ T. Thus we just need to prove that the closed-loop
system is stable in the usual sense. żi =0, i = 2, 3, . . . , n,
Since Ki (t) = 0, ∀t ∈ [0, (2(n − i) + 1)h] , i = 1, 2, . . . , n, ẋ1 =ż1 = −ax1 − K1 (t)x1 (t − h), ∀t ≥ 0.
the solution of the closed-loop system depends only
on x(0). In what follows, we denote by χ(t, x(0)) = By using Lemma 1, we know that x1 (t) 6= 0, ∀t ∈ [0, T ) , which
[χ1 (t, x (0)), χ2 (t, x(0)), . . . , χn (t, x(0))]T the solution of the implies T0 (x[t0 −h,t0 ] ) = T. The proof is finished.
closed-loop system. Remark 7: Since system (26) has no mismatched uncertainties, the
Step 1: It is obtained from (27) that design of the virtual control laws x1d , x2d , . . . , xnd doesn’t need to
be based on Lyapunov-like functions. Thus, we only use Theorem 1
|zn (t)| ≤ |zn (0)| = |xn (0) − xnd (0)| in the final step of the backstepping procedure, in order to suppress
which implies |χn (t, x(0))| ≤ Θn (x(0)) + |xnd | , where the matched uncertainties.
Θn (x(0)) , |xn (0) − xnd (0)| , is a continuous function, and
Θn (x(0)) = 0, when kx(0)k = 0.
Step i, i = 2, 3, . . . , n: Suppose that |zn−i+2 (t)| ≤ C. An example
Θn−i+2 (x(0)), where Θn−i+2 (x(0)) is a continuous function, and In this subsection, we present an example to illustrate the effec-
Θn−i+2 (0) = 0. Then, in view of (24), we denote z̃n−i+1 (t) = tiveness of Theorem 3. We consider the trajectory tracking control
eat zn−i+1 (t). Then of a one-link manipulator actuated by a brush dc (BDC) motor [7],
whose dynamics can be written as
z̃˙n−i+1 (t) = − eat Kn−i+1 (t)zn−i+1 (t − h)
+ eat Gn−i+1 (t, xn−i+1 )zn−i+2 , ∀t ≥ 0. M q̈ + B q̇ + N sin(q) = I,
This expression, along with Kn−i+1 = 0, ∀t ∈ [0, 2(i − 1)h], leads LI˙ = Vε − RI − KB q̇ + Vp , (29)
to
where
|z̃n−i+1 (t)|
J mL20 M0 L20 2M0 R02
Z 2(i−1)h M= + + + ,
≤ eas Gn−i+1 (s, χn−i+1 (s, x(0)))Θn−i+2 (x(0))ds Kτ 3Kτ Kτ 5Kτ
0 mL0 G M0 L0 G B0
N= + ,B = ,
+ |z̃n−i+1 (0)| , ∀t ∈ [0, 2(i − 1)h] . 2Kτ Kτ Kτ
A combination with zn−i+1 (t − h) = zn−i+2 (t) = 0, ∀t ∈ in which J is the rotor inertia, m is the link mass, M0 is the load
[2(i − 1)h, ∞) and |zn−i+1 (t)| ≤ |z̃n−i+1 (t)|, ∀t ≥ 0, leads us mass, L0 is the link length, R0 is the radius of the load, G is the
to gravity coefficient, B0 is the coefficient of viscous friction at the joint,
q is the angular motor position, I is the motor armature current, Kτ is
|zn−i+1 (t)| the coefficient which characterizes the electromechanical conversion
Z 2(i−1)h
of armature current to torque, L is is the armature inductance, R is
≤ eas Gn−i+1 (s, χn−i+1 (s, x(0)))Θn−i+2 (x(0))ds
0 the armature resistance, KB is the back-emf coefficient, Vε is the
+ |zn−i+1 (0)| , ∀t ∈ [0, ∞) . input control voltage, and Vp is the perturbation voltage. Select the
2
desired reference signal as qd (t) = (π/2) sin(t)(1 − e−0.1t ) [16].
Thus |zn−i+1 (t)| ≤ Θn−i+1 (x(0)) and |χn−i+1 (t, x(0))| ≤
Denote x = [x1 , x2 , x3 ] = [q − qd (t), q̇ − q̇d (t), I − Id (t)]T and
T
Θ n−i+1 (x(0)) + |xn−i+1d | , where Θn−i+1 (x(0)) , |zn−i+1 (0)| +
R 2(i−1)h u = Vε − Vd (t), where Id (t) = M q̈d (t) + B q̇d (t) + N sin(qd (t))
0
eas Gn−i+1 (s, χn−i+1 (s, x(0)))Θn−i+2 (x(0))ds is a con- and Vd (t) = RId (t) + KB q̇d (t) + LI˙d (t), then (29) can be rewritten
tinuous function. It holds that Θn−i+1 (x(0)) = 0, when kx(0)k = 0. as
At Step n, It is obtained that |χ1 (t, x(0))| = |z1 (t)| ≤ Θ1 (x(0)),
where Θ1 (x(0)) is a continuous function, and Θ1 (0) = 0. Thus ẋ1 =x2 ,
there exists a function Γ1 (kx(0)k) ∈ K, such that |χ1 (t, x(0))| ≤ N N B 1
Γ1 (kx(0)k). Then it is obtained from |z1 (t)| ≤ Γ1 (x(0)) that it ẋ2 = sin(qd (t)) − sin(x1 + qd (t)) − x2 + x3 ,
M M M M
also exists a function Γ2 (kx(0)k) ∈ K, such that |χ2 (t, x(0))| ≤ R KB 1 Vp
ẋ3 = − x3 − x2 + u + , (30)
Θ2 (x(0)) + |x2d (t, x1 (0))| ≤ Θ2 (x(0)) + max |x2d (t, x1 (0))| ≤ L L L L
0≤t≤T
Γ2 (kx(0)k). which in the form of (22) with F1 (t, x) = 0, G1 (t, x) =
The same arguments can be repeated. We known that there exists 1, F2 (t, x) = MN N
sin(qd (t)) − M B
sin(x1 + qd (t)) − M x2 , G2 (t, x) =
a function Γi (kx(0)k) ∈ K, such that |χi (t, x(0))| ≤ Θi (x(0)) + 1 R KB 1 V
, F3 (t, x) = − L x3 − L x2 , G3 (t, x) = L , φ(t, x) = Lp .
M
xid (t, xi−1 (0)) ≤ Θi (x(0)) + max |xid (t, xi−1 (0))| ≤ For simulation, choose J = 1.625 × 10−3 kg · m2 , m =
0≤t≤T
Γi (kx(0)k), i = 3, 4, . . . , n. Then there exists a function Γ (x(0)) ∈ 0.506kg, M0 = 0.434kg, L0 = 0.305m, R0 = 0.023m, B0 =
K, such that kχ(t, x(0))k ≤ Γ (kx(0)k). Thus for any ε > 0, 16.25 × 10−3 N · m · s/rad, L = 25 × 10−3 H, R = 5Ω, Kτ =
there exists a function δ(ε) which satisfies Γ (δ(ε)) = ε such that KB = 0.9N · m/A, Vp = 0.1 sin(50πt)V. The initial conditions
kχ(t, x(0))k < ε, when kx(0)k < δ(ε). are selected as q(0) = −0.2, q̇(0) = −0.1, I(0) = 0.1. According
Finally, we prove that there exists an initial condition x0 [t0 −h,t0 ] to Theorem 3, choose the prescribed time T = 30, namely, h = 5,
which satisfies kx0 (s)k < ∞, ∀s ∈ [t0 − h, t0 ] and a function and select the controller parameters as a = −0.15, τ = 81/83, and
 8

[12] Hu, S., & Zhu, Q. (2003). “Stochastic optimal control and analysis of
stability of networked control systems with long delay,” Automatica,
1
39(11), 1877-1884.
0.5 [13] Hua. C., Ning. P., Li. K. & Guan. X. (2020), “Fixed-time prescribed
0 tracking control for stochastic nonlinear systems with unknown mea-
surement sensitivity,” IEEE Trans. Cybern., early access.
0 5 10 15 20 25 30 35 40
[14] Hua. L., Zhu. H., Zhong. S., Zhang. Y., Shi. K. & Kwon. O. M.
(2022), “Fixed-Time Stability of Nonlinear Impulsive Systems and Its
6 Application to Inertial Neural Networks,” IEEE Transactions on Neural
4
Networks and Learning Systems, doi: 10.1109/TNNLS.2022.3185664.
2
[15] Huang. X., Lin. W. & Yang. B. (2005), “Global finite-time stabilization
0
of a class of uncertain nonlinear systems,” Automatica, 41(5), 881-888.
[16] Li, T. S., Wang, D., Feng, G., & Tong, S. C. (2009), “A DSC approach
-2
0 5 10 15 20 25 30 35 40 to robust adaptive NN tracking control for strict-feedback nonlinear
systems,”IEEE transactions on systems, man, and cybernetics, part b
(cybernetics), vol. 40, no. 3, pp. 915-927.
Fig. 5. The state x and control input u for (30). [17] Liu. B., Lu. W., & Chen. T. (2012), “Stability analysis of some delay
differential inequalities with small time delays and its applications,”
Neural Networks, 33, 1-6.
Ki (a, h)(t), i = 1, 2, 3, which is defined by (5) (we only consider [18] Liu. Y., Li. H., Lu. R., Zuo. Z. & Li. X. (2022), “An overview of
finite/fixed-Time control and its application in engineering systems,”
t ∈ [0, 2h]) with IEEE/CAA Journal of Automatica Sinica, 9(12), 2106-2120.
[19] Meng. Q., Ma. Q. & Shi. Y. (2023), “Adaptive fixed-time stabilization for

0, t ∈ [0, h) ,
Rh (t) = a class of uncertain nonlinear systems,” IEEE Transactions on Automatic
(t − h)5 (2h − t)5 , t ∈ [h, 2h] . Control, doi: 10.1109/TAC.2023.3244151.
The simulation results are shown in Fig. 5, from which we can see [20] Min. H., Shi. S., Xu. S., Guo. J. & Zhang. Z. (2022), “Fixed-
time Lyapunov criteria of stochastic nonlinear systems and its
that the state of the closed-loop system converges to zero at the
generalization,” IEEE Transactions on Automatic Control, doi:
prescribed time T = 30s. 10.1109/TAC.2022.3218592.
[21] Moulay. E., Dambrine. M. & Yeganefar. N. (2008). “Finite-time stability
V. C ONCLUSIONS and stabilization of time-delay systems,” Systems & Control Letters,
57(7), 561-566.
A novel Lyapunov-based approach for strong prescribed-time sta- [22] Pal, A. K., Kamal, S., Nagar, S. K., Bandyopadhyay, B., & Fridman,
bilization based on PDF has been established. Compared to existing L. (2020), “Design of controllers with arbitrary convergence time,”
Lyapunov-based approaches, the proposed approaches are not based Automatica, 112, 108710.
on the comparison lemma. A novel control law for uncertain scalar [23] Polyakov. A. (2012), “Nonlinear feedback design for fixed-time stabi-
lization of linear control systems,” IEEE Trans. Autom. Control, 57(8),
nonlinear systems has been proposed, whose terms exhibit a linear 2106–2110.
growth rate in the combined current and delayed state variable. In [24] R. Sipahi, S. Niculescu, C. Abdallah, W. Michiels, & K. Gu. (2011),
addition, a strongly prescribed-time stabilizing law for strict feedback “Stability and stabilization of systems with time delay,” IEEE Control
nonlinear systems has been designed. Unlike existing Lyapunov- Systems Magazine, 31(1), 38–65.
[25] Rang E. (1963), “Isochrone families for second-order systems,” IEEE
based approaches for prescribed-time stabilization, the control laws Transactions on Automatic Control, 8(1), 64-65.
designed in this paper do not have singularity problems. Numerical [26] Richard. J. P. (2003), “Time-delay systems: an overview of some recent
simulations have shown the effectiveness of the proposed approaches. advances and open problems,” automatica, 39(10), 1667-1694.
[27] Song. Y., Wang. Y., Holloway. J. & Krstic. M. (2017), “Time-varying
R EFERENCES feedback for regulation of normal-form nonlinear systems in prescribed
finite time,” Automatica, 83, 243–251.
[1] Artstein. Z. (1982), “Linear systems with delayed controls: A reduction,” [28] Sun. Z. Y., Shao. Y. & Chen. C. C. (2019), “Fast finite-time stability
IEEE Trans. Autom. Control, 27(4), 869–879. and its application in adaptive control of high-order nonlinear system,”
[2] Besancon. G., Georges. D., & Benayache. Z. (2007), “Asymptotic Automatica, 106, 339–348.
state prediction for continuous-time systems with delayed input and [29] Sun. Z. Y., Yun. M. M. & Li. T. (2017), “A new approach to fast global
application to control,” Proc. Eur. Control Conf., 1786–1791. finitetime stabilization of high-order nonlinear system,” Automatica, 81,
[3] Bhat. S. P. & Bernstein. D. S. (2000), “Finite-time stability of continuous 455–463.
autonomous systems,” SIAM J. Control Optim., 38(3), 751–766. [30] Wang. Y., Song. Y., Hill. D. J. & Krstic. M. (2019), “Prescribed-time
[4] Bhat. S. P. & Bernstein. D. S. (2005), “Geometric homogeneity with consensus and containment control of networked multiagent systems,”
applications to finite-time stability,” Mathematics of Control, Signals and IEEE Transactions on Cybernetics, 49(4), 1138-1147.
Systems, 17, 101-127. [31] Zak. M. (1989), “Terminal attractors in neural networks,” Neural net-
[5] Chitour. Y., Ushirobira. R., & Bouhemou. H. (2020), “Stabilization for works, 2(4), 259-274.
a perturbed chain of integrators in prescribed time,” SIAM J. Control [32] Zhang. K. K., Zhou. B. & Duan. G. -R. (2022), “Prescribed-time input-
Optim., 58(2), 1022–1048. to-state stabilization of normal nonlinear systems by bounded time-
[6] Corradini. M. & Cristofaro. A. (2018), “Nonsingular terminal sliding- varying feedback,” IEEE Transactions on Circuits and Systems I, 69(9),
mode control of nonlinear planar systems with global fixed-time stability 3715-3725.
guarantees,” Automatica, 95, 561-565. [33] Zhou. B. (2020), “Finite-time stability analysis and stabilization by
[7] Dawson. D. M., Carroll. J. J., & Schneider, M. (1994), “Integrator bounded linear time-varying feedback,” Automatica, 121, 109191.
backstepping control of a brush DC motor turning a robotic load,”IEEE [34] Zhou. B. & Zhang. K. -K. (2023), “A linear time-varying inequality ap-
Transactions on Control Systems Technology, vol. 2, no. 3, pp. 233-244. proach for prescribed time stability and stabilization,” IEEE Transactions
[8] Efimov, D., & Aleksandrov, A. (2020), “On estimation of rates of con- on Cybernetics, 53(3), 1880-1889.
vergence in Lyapunov-Razumikhin approach,” Automatica, 116, 108928. [35] Zhou. B., Michiels. W. & Chen. J. (2022), “Fixed-time stabilization
[9] Fridman, E. (2014). Introduction to time-delay systems: Analysis and of linear delay systems by smooth periodic delayed feedback,” IEEE
control. Springer. Transactions on Automatic Control, 67(2), 557-573.
[10] Garone, E., Gasparri, A., & Lamonaca, F. (2015), “Clock synchroniza- [36] Zhou. B., Ding. Y., & Zhang. K. K., “Prescribed time control based on
tion protocol for wireless sensor networks with bounded communication the smooth periodic delayed sliding mode surface without singularities,”
delays,”Automatica, 59, 60-72. Submitted for publication.
[11] Hong. Y., Wang, J. & Cheng. D. (2006), “Adaptive finite-time control [37] Zuo. Z., Song. J., Tian. B. & Basin. M. (2022), “Robust fixed-time stabi-
of nonlinear systems with parametric uncertainty,” IEEE Trans. Autom. lization control of generic linear systems with mismatched disturbances,”
Control, 51(5), 858–862. IEEE Trans. Syst., Man, Cybern., Syst., 52(2), 759–768.