Discrete and Continuous Dynamical Systems - Series B

Vol. 27, No. 7, July 2022, pp. 3585–3603

doi:10.3934/dcdsb.2021197

PRACTICAL PARTIAL STABILITY OF TIME-VARYING

SYSTEMS

Abdelfettah Hamzaoui, Nizar Hadj Taieb∗ and Mohamed Ali Hammami

Faculty of Sciences of Sfax, Department of Mathematics

(Communicated by Michael Malisoff)

Abstract. In this paper we investigate the practical asymptotic and exponen-

tial partial stability of time-varying nonlinear systems. We derive some suffi-
cient conditions that guarantee practical partial stability of perturbed systems
using Lyapunov’s theory where a converse theorem is presented. Therefore, we
generalize some works which are already made in the literature. Furthermore,
we present some illustrative examples to verify the effectiveness of the proposed
methods.

1. Introduction. The problem of the stability of motion with respect to some of

the variables, also known as partial stability, arises naturally in applications. The
basic results in this field belong to Rumyantsev ([5], [18], [19]-[22]), the founder of
the theory of partial stability for systems of ordinary differential equations with con-
tinuous right side and had demonstrated the applicability of his results to problems
of stability of more general models of distributed-parameter systems. Subsequently,
a large number of researchers have contributed to the development of theory and
methods for studying partial stability and stabilization and resolved several impor-
tant applied problems.
Practical stability ([12], [13]), being quite different from stability in the sense
of Lyapunov, is a significant performance specification from an engineering point
of view for the following reason: A system might be stable in theory; however, it
is actually unstable in practice because the stable domain or the domain of the
desired attractor is not large enough and on the other hand, sometimes the desired
state of a system may be mathematically unstable and yet the system may oscil-
late sufficiently near this state that its performance is acceptable, thus, stable in
practice. For standard state-space systems, [15], presented a systematic study of
the theory of practical stability and collected most valuable results. In the recent
years, considerable attention has been paid to the practical partial stability. Such
stability ensures the convergence of a part of the solutions towards a ball containing
the origin of the state space as the radius of the ball can be made arbitrarily small.
This notion is motivated by the ISS property introduced by Sontag [17] where the
trajectory exists for large time and it gets arbitrarily close to a certain sphere where

2020 Mathematics Subject Classification. Primary: 34D20, 37B25; Secondary: 37B55.

Key words and phrases. Lyapunov theory, perturbed systems, practical partial stability.
The authors wish to thank the editor and the anonymous reviewers for their valuable and
careful comments.
∗ Corresponding author: Nizar Hadj Taieb.

3585
3586 ABDELFETTAH HAMZAOUI, NIZAR HADJ TAIEB AND MOHAMED ALI HAMMAMI

the radius depends on the control. Using the Lyapunov-like functions, the author
in [3] have studied the practical stability with respect to a part of the variables of
fractional-order nonlinear systems depending on a small parameter. The authors
in [4] have studied the practical stability with respect to a part of the variables of
nonlinear stochastic differential equations.
One of the main result of this paper is to generalize the result given by [8]
related to present the new Lyapunov function based practical stability analysis
approach for time-varying systems. Thus, we study the convergence of a part of the
solutions towards a small ball containing the origin of the state space. Therefore, we
build practical partial stability theorems using Lyapunov functions and then a more
general converse Theorem is presented. Furthermore, we deal with the problem of
global practical uniform asymptotic and exponential partial stability of some classes
of perturbed systems based on practical scalar function. The remainder of this paper
is organized as follows. In section two, we present our basic results. A converse
stability theorem is established in section three. We study a class of perturbed
systems in section four. Moreover, in section five we present an illustrative example
to indicate significant improvements and the application of the results which arise
from an epidemic model.

2. Basic results. We will use the following notations throughout this paper
• R+ = [0, +∞[, R∗+ =]0, +∞[ and Rn denotes the n−dimensional Euclidian
space with appropriate norm k.k.
• PC(R+ , R) is the space of piecewise continuous functions on R+ to R.
• C 0 (R+ , R+ ) is the space of continuous functions on R+ to R+ .
• C 1 (R+ , R+ ) is the space of continuous differentiable functions on R+ to R+ .
• For r ≥ 0, Br = {x ∈ Rn /kxk ≤ r}.
• LP (R+ ) is the space of functions integrable with p − th power on R+ , p ≥ 1,
Z +∞  p1
and kLkp = |Lp (s)|ds is the norm on LP (R+ ).
0
Consider the following nonlinear time-varying system
ẋ(t) = F (t, x(t)) (1)
where t ∈ R+ is the time, x(t) ∈ Rn is the state, F : R+ ×Rn −→ Rn is continuous on
(t, x) and locally Lipschitz on x. Let φ(., t, x), be the unique solution of (1) passing
through (t, x) such that φ(t, t, x) = x. For xT = (x1 , x2 , ...xn ) ∈ Rn , let q be an
integer such that q ≤ n, y T = (x1 , x2 , ..., xq ) ∈ Rq and z T = (xq+1 , xq+2 , ..., xn ) ∈
Rn−q , where T denotes the transposition. According to this partition, the solution
of (1) can be rewritten for all (t, x) ∈ R+ × Rn ,
φ(., t, x) = (y(., t, x), z(., t, x)),
where y(., t, x) and z(., t, x) are respectively the unique solutions of the following
time-varying systems
ẏ(t) = F1 (t, y(t), z(t))
and
ż(t) = F2 (t, y(t), z(t)),
with F (t, x) = (F1 (t, y, z), F2 (t, y, z)), such that z(s, t, x) is assumed to be
bounded for all s ≥ t ≥ 0.
 PRACTICAL PARTIAL STABILITY OF TIME-VARYING SYSTEMS 3587

Definition 2.1. A continuous function α : [0, +∞[→ [0, +∞[ is said to belong to
class K if it is strictly increasing and α(0) = 0. It is said to belong to class K∞ , if
α(r) → +∞ as r → +∞.
Definition 2.2. A continuous function β : [0, +∞[×[0, +∞[→ [0, +∞[ is said to
belong to class KL if, for each fixed s, the mapping β(r, s) belongs to class K with
respect to r and, for each fixed r, the mapping β(r, s) is decreasing with respect to
s and β(r, s) → 0 as s → +∞.
In the case when the origin in not necessarily an equilibrium point, we will study
the asymptotic behavior of the solutions in the practical sense, it means that the
solution converges to a small ball centered at the origin with respect to some of the
variables.
Definition 2.3. System (1) is said to be globally uniformly practically asymptoti-
cally y−stable, if there exist β ∈ KL and r > 0, such that for all initial state x, we
have
ky(s, t, x)k ≤ β(kxk, s − t) + r, for all s ≥ t ≥ 0. (2)
Definition 2.4. System (1) is is said to be globally uniformly practically exponen-
tially y−stable if there exist γ > 0, k ≥ 0 and r > 0 such that for all initial state x,
we have
ky(s, t, x)k ≤ kkxk exp(−γ(s − t)) + r, for all s ≥ t ≥ 0. (3)
Remark 1. The inequalities (2) and (3) imply that ky(s, t, x)k will be bounded by
a small bound r > 0, that is, ky(s, t, x)k will be small for sufficiently large s, by
taking an initial condition outside the Ball Br . In this situation, a robustness result
can be obtained if we suppose that r depends on a small parameter ε > 0, in this
case if r = r(ε) approaches to zero as ε tends to zero, then, ky(s, t, x)k approaches
the origin exponentially as s goes to infinity.
Now, let’s present the following definition which is given in [8].
Definition 2.5. Let µ, π ∈ PC(R+ , R). The function µ is π−globally uniformly
practically exponentially stable if there exist θ > 0, λ ≥ 0 and ρ > 0, such that for
all s ≥ t, Z s
µ(τ )dτ ≤ −θ(s − t) + λ
t
and Z s
|π(τ )|ψ(s, τ )dτ ≤ ρ,
t
Z s 
where ψ(s, t) = exp µ(τ )dτ .
t

Remark 2. Indeed, the notion of stable scalar functions was firstly introduced in
([23], [24]) after the author in [8] has extended these definitions to the practical
case. It is worth noting that, if the function µ is π−globally uniformly practically
exponentially stable, then the system
ẏ = µ(t)y(t) + π(t)
is globally uniformly practically exponentially stable.
Now, we present the following lemmas to prove our results.
3588 ABDELFETTAH HAMZAOUI, NIZAR HADJ TAIEB AND MOHAMED ALI HAMMAMI

Lemma 2.6. Let α be a class K function. We have for all a, b ≥ 0,

α(a + b) ≤ α(2a) + α(2b).
Lemma 2.7. For all m ≥ 1 and all a, b ≥ 0,
1 1 1
(a + b) m ≤ a m + b m .
In the sequel, we will use the following auxiliary result called the Generalized
Gronwall-Bellman Inequality which is taken from [24].
Lemma 2.8. (Generalized Gronwall-Bellman Inequality) Let λ, ψ ∈ PC(R+ , R),
and ϕ : R+ −→ R+ is a differentiable function, such that for all s ≥ t,
ϕ̇(s) ≤ λ(s)ϕ(s) + ψ(s).
Then, for all s ≥ t, we have
Z s  Z s Z s 
ϕ(s) ≤ ϕ(t) exp λ(τ )dτ + exp λ(τ )dτ ψ(u)du.
t t u

The next theorem spells out conditions under which system (1) is globally uniformly
practically asymptotically y−stable. Notice that in this paper we try to generalize
the work [8] under the fact that the function µ is not required to be nonnegative
for all t.
Theorem 2.9. Assume that there exist V : [0, +∞[×Rn → R+ a continuously
differentiable function, two K∞ − functions αi , i = 1; 2, a constant a ≥ 0 and scalar
functions µ, π ∈ PC(R+ , R), such that, for all t ≥ 0 and x ∈ Rn ,
α1 (kyk) ≤ V (t, x) ≤ α2 (kxk) + a, (4)
∂V ∂V
(t, x) + (t, x)F (t, x) ≤ µ(t)V (t, x) + π(t), (5)
∂t ∂x
then, system (1) is globally uniformly practically asymptotically y-stable if µ is
π−globally uniformly practically exponentially stable.
Proof. Using Lemma 2.8, we have for all s ≥ t,
Z s
V (s, φ(s, t, x)) ≤ V (t, x)ψ(s, t) + π(τ )ψ(s, τ )dτ,
t
Z s 
where ψ(s, t) = exp µ(τ )dτ . Therefore, we have for all s ≥ t,
t
α1 (ky(s, t, x)k) ≤ V (s, φ(s, t, x))
Z s
≤ V (t, x)ψ(s, t) + π(τ )ψ(s, τ )dτ.
t
Since, µ is π−globally uniformly practically exponentially stable, then, for all s ≥ t,
α1 (ky(s, t, x)k) ≤ α2 (kxk)eλ e−θ(s−t) + aeλ + ρ.
It follows that, for all s ≥ t,
ky(s, t, x)k ≤ α1−1 α2 (kxk)eλ e−θ(s−t) + aeλ + ρ .

By Lemma 2.6, we have for all s ≥ t,

ky(s, t, x)k ≤ α1−1 2α2 (kxk)eλ e−θ(s−t) +α1−1 (2aeλ + 2ρ).

Hence, system (1) is globally uniformly practically asymptotically y−stable.

 PRACTICAL PARTIAL STABILITY OF TIME-VARYING SYSTEMS 3589

Remark 3. One can check here that the origin of the considered systems may not
be an equilibrium point. Indeed, the desired system may be unstable and yet the
system may oscillate sufficiently near this state that its performance is acceptable,
thus the notion of practical stability is more suitable in several situations than
traditional stability. One also desires that the state approaches the origin (or some
sufficiently small neighborhood of it) in a sufficiently fast manner. To this end, the
Theorem 2.9 gives sufficient conditions which relates to the Lyapunov function that
not necessary to meet to be definite.

For the exponential case we have the following Corollary.

Corollary 1. Assume that there exist V : [0, +∞[×Rn → R+ a continuously dif-

ferentiable function, some constants c1 > 0, c2 ≥ 0, a ≥ 0, and m ≥ 1 and scalar
functions µ, π ∈ PC(R+ , R), such that, for all t ≥ 0 and x ∈ Rn ,

c1 kykm ≤ V (t, x) ≤ c2 kxkm + a,

∂V ∂V
(t, x) + (t, x)F (t, x) ≤ µ(t)V (t, x) + π(t),
∂t ∂x

then, system (1) is globally uniformly practically exponentially y-stable if µ is

π−globally uniformly practically exponentially stable.

Proof. Using the same arguments as the Theorem 2.9, we have for all t ≥ t0 ,

c2 a ρ
ky(s, t, x)km ≤ kxkm eλ e−θ(s−t) + eλ +
c1 c1 c1
≤ bkxkm eλ e−θ(s−t) + eλ c + σ,

c2 a ρ
where b = ,c= and σ = . Then, by Lemma 2.7, we have for all s ≥ t,
c1 c1 c1

1 λ θ 1
ky(s, t, x)k ≤ b m kxke m e− m (s−t) + (eλ c + σ) m .

Hence, system (1) is globally uniformly practically exponentially y−stable.

Example 1. Consider the following nonlinear time-varying system

  1  t|cos(t)|
ẋ1 = − t| cos(t)| x1 + u(t, x1 , x2 )

1+t√ 2 2
+ x1 + x2 1 + x21 + x22 (6)
ẋ2 = −x2 + x2 ,


where t ≥ 0, x = (x1 , x2 ) ∈ R2 and u : R+ ×R2 −→ R is a continuously differentiable

scalar function such that for all t ≥ 0 and (x1 , x2 ) ∈ R2 , t|u(t, x1 , x2 )| ≤ %, with % >
0. Let V (t, x) = x21 + e−t which satisfies the inequalities given in (4) with α1 (|x1 |) =
x21 , α2 (kxk) = x21 + x22 and a = 1. The derivative of V along the trajectories of (6)
3590 ABDELFETTAH HAMZAOUI, NIZAR HADJ TAIEB AND MOHAMED ALI HAMMAMI

satisfies
 1  t|cos(t)|
V̇ (t, x) = 2 2 2 − t| cos(t)| x21 + x1 u(t, x1 , x2 ) − e−t
1 + t + x1 + x2 1 + x21 + x22
 1  2e−t
=2 2 2 − t| cos(t)| V (t, x) − + 2te−t | cos(t)|
1 + t + x1 + x2 1 + t + x21 + x22
t|cos(t)|
+ x1 u(t, x1 , x2 ) − e−t
1 + x21 + x22
 1 
≤2 − t| cos(t)| V (t, x) + 2te−t + t| cos(t)||u(t, x1 , x2 )|
1 + t + x21 + x22
 1 
≤2 2 2 − t| cos(t)| V (t, x) + 2te−t + %| cos(t)|
1 + t + x1 + x2
≤ µ(t)V (t, x) + 2te−t + %| cos(t)|,
2
where µ(t) = − t| cos(t)| and π(t) = 2te−t + %| cos(t)|.
1 + Zt s Z s
Let ψ(s, t) = exp( µ(τ )dτ ). Since, 2τ e−τ ψ(s, τ )dτ −→ 0 as s −→ +∞. There-
t t
fore, there exist
4π  3  hZ s i %eλ
θ= > 0, λ = 2 ln 1 + π +2 > 0 and ρ = sup 2τ e−τ ψ(s, τ )dτ + ,
3 2 s≥t t θ
such that, for all s ≥ t,
Z s
µ(τ )dτ ≤ −θ(s − t) + λ, (see Appendix 3 in [24])
t
and Z s
π(τ )ψ(s, τ )dτ ≤ ρ.
t
It follows that, µ is π−globally uniformly practically exponentially stable. Then,
the system (6) is globally uniformly practically exponentially x1 −stable.
The established result provides the following question: if system (1) is glob-
ally uniformly practically exponentially y−stable, is there a function V (t, x) which
satisfies the hypothesis of the Theorem 2.9? Indeed, Theorem 2.9 gives sufficient
conditions for exponential stability of a small ball of radius r and centered at the
origin. It does not, however, give a prescription for determining the Lyapunov
function V (t, x).
In the next section, we will show that under some assumptions, there is a function
V (t, x) that satisfies conditions similar to those of Theorem 2.9.

3. A converse stability theorem. The question arises whether it is true that

global uniform practical exponential partial stability implies the existence of Lya-
punov functions such as described in the Theorem 2.9. The authors in ([6], [8]) have
established a converse stability theorem for (1), in the classical stability. Here, as
we have already mentioned we will present a converse stability result for (1), in the
practical partial stability case without assuming bounded conditions on the function
F as known in the literature ([2], [11]). Let’s start by the following Lemma.
Lemma 3.1. For all integer p ≥ 1 and all a, b ≥ 0,
(a + b)p ≤ 2p−1 (ap + bp ).
 PRACTICAL PARTIAL STABILITY OF TIME-VARYING SYSTEMS 3591

In order to give the precise assumption imposed on F (., .), we need to introduce the
following subclasses of C 0 (R+ , R+ ), see [6].
Definition 3.2. A function L ∈ C 0 (R+ , R+ ) is said to be of class BN , if there
exists a function M ∈ C 0 (R+ , R+ ) such that
Z t+δ
L(s)ds ≤ M (δ), for all t ≥ 0, δ ≥ 0.
t

Definition 3.3. A function L ∈ C 0 (R+ , R+ ) is said to be of class BN 2 , if there

exists a function N ∈ C 0 (R+ , R+ ) such that
Z t+δ
L2 (s)ds ≤ N (δ), for all t ≥ 0, δ ≥ 0.
t

Also, we need the following definition.

Definition 3.4. We define the Dini derivative or the upper right-hand generalized
derivative of a function V (t, x) along solutions of (1) by:
n1 o
D+ V (t, x) = lim sup [V (t + h, x + hF (t, x)) − V (t, x)] , (7)
h→0+ h
where V (t, x) satisfies the Lipschitz condition with respect the variable x uniformly
in t, i.e.;
|V (t, x) − V (t, y)| ≤ L|x − y|.
Note that, if V (t, x) has a continuous partial derivative with respect to the first
variable, then along the solution of (1) we have
∂V ∂V
D+ V (t, x) = (t, x) + (t, x)F (t, x).
∂t ∂x
Now, in order to prove a converse theorem, we shall suppose the following assump-
tions.

(H1 ) There exist functions R(.) ∈ BN and K(.) ∈ BN 2 , such that

kF1 (t, y, z)k ≤ L(t)kyk + K(t),
F1 is continuously differentiable with respect to y, and
∂F
1
(t, y, z) ≤ L(t),

∂y

where
L(t) = L + R(t),
with L > 0.

(H2 ) The system (1) satisfies (3) for all t ≥ 0, x ∈ Rn and for some non-negative
constants k, γ and r.

Then, one can state the following theorem.

Theorem 3.5. Under assumptions (H1 ) and (H2 ), there exist a function V : R+ ×
Rn → R, constants c1 > 0, c2 > 0, c3 > 0, a > 0, b > 0, R > 0, an integer p ≥ 2,
and scalar functions µ, π ∈ PC(R+ , R) such that for all t ∈ R+ and x = (y, z) ∈ BR ,
we have
c1 kykp ≤ V (t, x) ≤ c2 kxkp + a,
3592 ABDELFETTAH HAMZAOUI, NIZAR HADJ TAIEB AND MOHAMED ALI HAMMAMI

∂V ∂V
(t, x) + (t, x)F (t, x) ≤ µ(t)V (t, x) + π(t),
∂t ∂x
∂V
k (t, x)k ≤ c3 kxkp−1 + b,
∂y
where µ is π−globally uniformly practically exponentially stable.
Proof. Let
Z t+δ  h e2M (δ) Z s i p2 
V (t, x) = η(s) ky(s, t, x)kp + K 2 (τ )dτ ds,
t 2L t
0 ∗
where η ∈ C (R+ , R+ ) is a decreasing function, with inf (η(t)) > 0, and δ is chosen
t≥0
such that
ln(2p−1 k p )
δ> . (8)
pγ
By assumption (H2 ), we have
Z t+δ h ip h e2M (δ) i p2
V (t, x) ≤ η(t) r + kkxke−γ(s−t) + N (δ) ds
t 2L
Z t+δ h e2M (δ) i p2
≤ η(0) 2p−1 rp + N (δ) +2p−1 k p kxkp e−pγ(s−t) ds
t 2L
2p−1 k p h e2M (δ) i p2

≤ (1 − e−pγδ )η(0)kxkp + 2p−1 rp + N (δ) δη(0).
pγ 2L
Next, we estimate a lower bound of V. We have,
d 
 y T (s, t, x)y(s, t, x) ≤ 2ky(s, t, x)kkF1 (s, y(s, t, x), z(s, t, x))k
 
ds
≤ 2L(s)ky(s, t, x)k2 + 2K(s)ky(s, t, x)k.
Thus,
d T
y (s, t, x)y(s, t, x) ≥ −2L(s)ky(s, t, x)k2 − 2K(s)ky(s, t, x)k. (9)
ds
Letting v(s) = −ky(s, t, x)k and using (9), we deduce (as in [11], Example 3.9, pp.
103 − 104) that
D+ v(s) ≤ −L(s)v(s) + K(s).
Let, Z s
L(τ )dτ
u(s) = v(s)e t .
It follows that Z s
L(τ )dτ
D+ u(s) ≤ K(s)e t .
Taking into account Theorem 9 in [10] on [t, s], yields
Z s Z τ Z s
− L(τ )dτ Z s L(ς)dς  − L(τ )dτ
v(s) ≤ v(t)e t + K(τ )e t dτ e t .
t
On the one hand, we have
Z s Z s
L(τ )dτ = L(s − t) + R(τ )dτ,
t t
≤ L(s − t) + M (δ), for all s ∈ [t, t + δ], t ≥ 0,
 PRACTICAL PARTIAL STABILITY OF TIME-VARYING SYSTEMS 3593

then, Z s
L(τ )dτ
e t ≤ eL(s−t) eM (δ) , for all s ∈ [t, t + δ], t ≥ 0.
On the other hand, we have
Z τ Z s Z s
 s
Z L(ς)dς  − L(τ )dτ Z s  − L(τ )dτ
K(τ )e t dτ e t ≤ eM (δ) K(τ )eL(τ −t) dτ e t .
t t
Using Cauchy-Schwartz inequality, one obtains for all s ∈ [t, t + δ], t ≥ 0,
Z τ Z s
hZ s L(ς)dς i − L(τ )dτ Z s  1 hZ s 1
e2L(τ −t) dτ e−L(s−t)
2 2
K(τ )e t dτ e t ≤ eM (δ) K 2 (τ )dτ
t t t
 e2M (δ) Z s 1
2 2
≤ K (τ )dτ .
2L t

Thus, for all s ∈ [t, t + δ], t ≥ 0,

 e2M (δ) Z s  21
ky(s, t, x)k + K 2 (τ )dτ ≥ kyke−M (δ) e−L(s−t) .
2L t
Now, since
 e2M (δ) Z s s p
h  1 ip
2
hZ e2M (δ) 2 i
2
ky(s, t, x)k + K 2 (τ )dτ ≤ 2p−1 ky(s, t, x)kp + 2p−1 K (τ )dτ ,
2L t t 2L
then, for all s ∈ [t, t + δ], t ≥ 0,
h e2M (δ) Z s i p2 1
ky(s, t, x)kp + K 2 (τ )dτ ≥ p−1 kykp e−pM (δ) e−pL(s−t) .
2L t 2
Therefore,
e−pM (δ) (1 − e−pLδ )
V (t, x) ≥ η(t + δ)kykp
2p−1 pL
e−pM (δ) (1 − e−pLδ )
≥ inf (η(t))kykp .
2p−1 pL t≥0

Thus, V (t, x) satisfies the first inequality of Theorem 3.5 with

e−pM (δ) (1 − e−pLδ )

c1 = inf (η(t)),
2p−1 pL t≥0

2p−1 k p
c2 = (1 − e−pγδ )η(0),
pγ
and
h e2M (δ) i p2

a = 2p−1 rp + N (δ) δη(0).
2L
Next, to prove the existence of the derivative of V along the trajectories of (1), it
suffices to consider
Z s+δ  h e2M (δ) Z u i p2 
V (s, φ(s, t, x)) = η(u) ky(u, s, φ(s, t, x))kp + K 2 (τ )dτ du.
s 2L s
Since the following two solutions of (1),
u 7→ y(u, t, x), and u 7→ y(u, s, φ(s, t, x))
3594 ABDELFETTAH HAMZAOUI, NIZAR HADJ TAIEB AND MOHAMED ALI HAMMAMI

are equals at time u = s, then

y(u, t, x) = y(u, s, φ(s, t, x)), for all u ≥ s ≥ t.
Thus,
Z s+δ  h e2M (δ) Z u i p2 
p
V (s, φ(s, t, x)) = η(u) ky(u, t, x)k + K 2 (τ )dτ du.
s 2L s
This implies that the derivative of V (t, x) along the trajectories of (1) exists and it
is given by
d
V̇ (t, x) = (V (s, φ(s, t, x))/s=t
ds
h e2M (δ) Z t+δ i p2
= η(t + δ)ky(t + δ, t, x)kp − η(t)kykp + η(t + δ) K 2 (τ )dτ
2L t
2M (δ) Z t+δ 2M (δ) Z u p
pδe e 2 −1
h i
2 2
− K (t) η(u) K (τ )dτ du
4L t 2L s
≤ η(t + δ)[kkxke−γδ + r]p − η(t)kxkp + η(t)kxkp − η(t)kykp
h e2M (δ) Z t+δ i p2
+ η(t + δ) K 2 (τ )dτ
2L t
 h e2M (δ) i p2 
≤ (2p−1 k p e−pγδ − 1)η(t)kxkp + η(t) kxkp + 2p−1 rp + N (δ)
2L
 h e2M (δ) i p2 
≤ (2p−1 k p e−pγδ − 1)η(t)kxkp + η(t) Rp + 2p−1 rp + N (δ)
2L
According to the inequality (8), one gets
 h 2M (δ) i p2 
p−1 p e
p−1 p −pγδ
pγ(2 k e − 1) pγδ 2 r + 2LN (δ)
V̇ (t, x) ≤ p−1 p η(t)V (t, x) + η(t)
2 k (1 − e−pγδ ) inf (η(t)) 2p−1 k p (1 − e−pγδ )
t≥0
 h e2M (δ) i p2 
+ η(t) Rp + 2p−1 rp + N (δ) .
2L
Thus, the second inequality of theorem 3.5 is satisfied with
pγ(2p−1 k p e−pγδ − 1)
µ(t) = η(t)
2p−1 k p (1 − e−pγδ ) inf (η(t))
t≥0

and
 h 2M (δ) i p2 
h pγδ 2p−1 rp + 2LNe h e2M (δ) i p2 i
(δ) p p−1 p
π(t) = η(t) + R + 2 r + N (δ) ,
2p−1 k p (1 − e−pγδ ) 2L
where µ is π-globally uniformly practically exponentially stable. Let,
∂
yy (s, t, x) = y(s, t, x).
∂y
Since,
∂F
1
(t, y, z) ≤ L(t),

∂y

then, yy satisfies the following bound
kyy (s, t, x)k ≤ eL(s−t) eM (δ) , for all s ∈ [t, t + δ].
 PRACTICAL PARTIAL STABILITY OF TIME-VARYING SYSTEMS 3595

Hence,
∂V Z t+δ ∂ h  p2 i
(t, x) = η(s) ky(s, t, x)k2 ds

∂y ∂y

t
p Z t+δ h∂ ih i p2 −1
= η(s) ky(s, t, x)k2 ky(s, t, x)k2 ds

2 t ∂y
Z t+δ
= p η(s)y T (s, t, x)yy (s, t, x)ky(s, t, x)kp−2 ds

t
Z t+δ
≤ pη(t) ky(s, t, x)kp−1 kyy (s, t, x)kds
t
Z t+δ
≤ p2p−2 k p−1 kxkp−1 eM (δ) η(t) e(L−(p−1)γ)(s−t) ds
t
Z t+δ
+ p2p−2 rp−1 eM (δ) η(t) eL(s−t) ds
t
Z t+δ
p−2 p−1 p−1 M (δ)
≤ p2 k kxk e η(0) e(L−(p−1)γ)(s−t) ds
t
eLδ − 1
+ p2p−2 rp−1 eM (δ) η(0) .
L
Then, two cases will be treated: γ(p − 1) ≥ L or γ(p − 1) < L.

Case 1. if γ(p − 1) ≥ L, then the last inequality of Theorem 3.5 holds with
p2p−2 rp−1 eM (δ) Lδ
c3 = p2p−2 k p−1 eM (δ) δη(0) and b = (e − 1)η(0).
L
Case 2. if γ(p − 1) < L, then the last inequality of Theorem 3.5 holds with
p2p−2 k p−1 eM (δ) (L−(p−1)γ)δ p2p−2 rp−1 eM (δ) Lδ
c3 = (e −1)η(0) and b = (e −1)η(0).
L − (p − 1)γ L
The Theorem is proved.

Remark 4. One can see that when a = 0 and π(t) = 0, for all t ≥ 0, then, the
Theorem 3.5 ensures that there exist a lyapunov function with indefinite derivative
such that the function µ is globally uniformly exponentially y−stable, see [23].
A major concern in analyzing the stability of dynamical systems is the robustness
of various stability properties to uncertainties in the system’s model. That’s why
the class of perturbed systems is very important from a practical aspect [9].

4. Perturbed systems. Let’s consider the following time varying perturbed sys-
tem
ẋ = f (t, x) + g(t, x), (10)
where f, g : R+ × Rn → Rn are continuous on (t, x) and locally Lipschitz on x. Let
φ(., t, x), be the unique solution of (10) passing through (t, x) such that φ(t, t, x) = x
and φ(., t, x) = (y(., t, x), z(., t, x)).
In the actual literature, the synthesis of stability of equation (10) is based on the
stability of the nominal system
ẋ = f (t, x), (11)
3596 ABDELFETTAH HAMZAOUI, NIZAR HADJ TAIEB AND MOHAMED ALI HAMMAMI

with V (t, x) as a Lyapunov function candidate for the whole system provided that
the size of perturbation is known. However, we cannot usually conclude the behav-
ior of the solutions of the perturbed system (10), by using V (t, x) as a Lyapunov
function candidate ([1], [7]). Let’s consider the following 1−dimensional perturbed
system
e−t
ẋ = −x + + S(t)x, (12)
1 + x2
2 e−t
where S(t) = ∈ BN , with M (δ) = 2 ln(1 + δ), f (t, x) = −x + and
1+t 1 + x2
g(t, x) = S(t)x.
e−t
One can see that the nominal system ẋ = −x + is globally uniformly prac-
1 + x2
tically exponentially stable with V (t, x) = x2 , µ(t) = −1 and π(t) = e−t .
If we calculate the derivative of V along the trajectories of (12), we obtain
V̇ (t, x) = (−1 + S(t))V (t, x) + e−t .
Here µ̃(t) = −1 + S(t) is not π−globally uniformly practically exponentially stable.
In the sequel, our goal is to show that system (10) keeps the same kind of stability
as system (11). Then, we try to construct a function which guarantees the practical
stability of (10).

4.1. Practical asymptotic partial stability. Let’s consider the following as-
sumptions:
(H3 ) There exist V : [0, +∞[×Rn → R+ a continuously differentiable function, two
K∞ functions ξi , i = 1; 2, a constant ã ≥ 0 and scalar functions µ̃, π̃ ∈ PC(R+ , R),
such that, for all t ≥ 0 and x = (y, z) ∈ Rq × Rn−q ,
ξ1 (kyk) ≤ V (t, x) ≤ ξ2 (kxk) + ã,
∂V ∂V
(t, x) + (t, x)f (t, x) ≤ µ̃(t)V (t, x) + π̃(t),
∂t ∂x
(H4 ) There exist S ∈ BN , such that, for all t ≥ 0 and x ∈ Rn ,
 ∂V 
(t, x)g(t, x)≤ S(t)V (t, x). (13)
 
∂x


Theorem 4.1. Under assumption (H3 ) and (H4 ), the perturbed system (10) is
globally uniformly practically asymptotically y−stable if µ̃ is π̃−globally uniformly
practically exponentially stable.
Proof. We have µ̃ is π̃−globally uniformly practically exponentially stable, then,
there exist θ > 0 and λ ≥ 0 such that, for all s ≥ t,
Z s
µ̃(τ )dτ ≤ −θ(s − t) + λ.
t

Also, we have S ∈ BN . Thus, there exists a function M ∈ C 0 (R+ , R+ ) such that

Z t+δ
S(s)ds ≤ M (δ), for all t ≥ 0, δ ≥ 0.
t
Let’s consider the following function:
 
Wδ (t, x) = V (t, x) exp ϕδ (t, x) , (14)
 PRACTICAL PARTIAL STABILITY OF TIME-VARYING SYSTEMS 3597

where
Z t+δ Z s
1 ∂V
ϕδ (t, x) = (τ, φ(τ, t, x))g(τ, φ(τ, t, x))dτ ds.
t t δV (τ, φ(τ, t, x)) ∂x
Hence, for well chosen values of δ, we must prove that Wδ satisfies the conditions of
theorem 2.9 to guarantee that (10) is globally uniformly practically asymptotically
stable.
Since,
 ∂V 
(t, x)g(t, x)≤ S(t)V (t, x),
 
∂x

then,
|ϕδ (t, x)| ≤ M (δ).
Therefore, we have for all t ≥ 0, and x ∈ Rn ,
α1 (kyk) ≤ Wδ (t, x) ≤ α2 (kxk).
Then, Wδ satisfies the inequality (4) with
α1 (kyk) = ξ1 (kyk)e−M (δ) , α2 (kxk) = ξ2 (kxk)eM (δ) and a = ãeM (δ) .
Now, we prove that the derivative of ϕδ along the trajectories of (10) exists. We
have for all s ≥ t ≥ 0,
Z s+δ Z u
1 ∂V
ϕδ (s, φ(s, t, x)) = (τ, φ(τ, s, φ(s, t, x)))
s s δV (τ, φ(τ, s, φ(s, t, x))) ∂x
g(τ, φ(τ, s, φ(s, t, x)))dτ du‘
Since the following two solutions of (10) :
τ 7−→ φ(τ, t, x) and τ 7−→ φ(τ, s, φ(s, t, x))
are equals at the time τ = s, then
φ(τ, t, x) = φ(τ, s, φ(s, t, x)) for all s ≥ t ≥ 0.
Thus,
Z s+δ Z u
1 ∂V
ϕδ (s, φ(s, t, x)) = (τ, φ(τ, t, x))g(τ, φ(τ, t, x))dτ du.
s s δV (τ, φ(τ, t, x)) ∂x
This implies that the derivative of ϕδ (t, x) along the trajectories of (10) exists and
it is given by
d
ϕ̇δ (t, x) = (ϕδ (s, φ(s, t, x)))/s=t
ds
Z t+δ
1 ∂V
= (τ, φ(τ, t, x))g(τ, φ(τ, t, x))dτ
t δV (τ, φ(τ, t, x)) ∂x
Z t+δ
1 ∂V
− (t, x)g(t, x)dτ
t δV (t, x) ∂x
Z t+δ
1 ∂V
= (τ, φ(τ, t, x))g(τ, φ(τ, t, x))dτ
t δV (τ, φ(τ, t, x)) ∂x
1 ∂V
− (t, x)g(t, x).
V (t, x) ∂x
3598 ABDELFETTAH HAMZAOUI, NIZAR HADJ TAIEB AND MOHAMED ALI HAMMAMI

Thus, the derivative along the trajectories of system (10) is given by

   
Ẇδ (t, x) = V̇ (t, x) exp ϕδ (t, x) +ϕ˙δ (t, x)V (t, x) exp ϕδ (t, x)
h ∂V ∂V i   ∂V  
= (t, x) + (t, x)f (t, x) exp ϕδ (t, x) + (t, x)g(t, x) exp ϕδ (t, x)
∂t ∂x ∂x
hZ t+δ 1 ∂V
+ (τ, φ(τ, t, x))g(τ, φ(τ, t, x))dτ
t δV (τ, φ(τ, t, x)) ∂x
1 ∂V i  
− (t, x)g(t, x) V (t, x) exp ϕδ (t, x)
V (t, x) ∂x
  M (δ)  
≤ (µ̃(t)V (t, x) + π̃(t)) exp ϕδ (t, x) +V (t, x) exp ϕδ (t, x)
δ
M (δ) M (δ)
≤ (µ̃(t) + )Wδ (t, x) + π̃(t)e .
δ
Therefore, if δ > 0 is chosen such that
M (δ)
< θ, (15)
δ
then, Wδ satisfies the inequality (5) with
M (δ)
µ(t) = µ̃(t) + and π(t) = π̃(t)eM (δ) ,
δ
such that, µ is π−globally uniformly practically exponentially stable which implies
that (10) is globally uniformly practically asymptotically y-stable.
To justify the existence of the inequality (15) which relates to the choice of δ, we
consider the following remark.
Remark 5. Let θ > 0. If S ∈ LP (R+ ), with p ≥ 1, then, by using Holder’s in-
p−1
equality we have S ∈ BN with M (δ) = kSkp δ p . Consequently, Wδ given in (14)
guarantees that the perturbed system is globally uniformly practically asymptoti-
cally y−stable for all
 kSk  p1
p
δ> .
θ
4.2. Practical exponential partial stability. For the exponential case, let’s con-
sider the following assumption.
(H5 ) Assume that there exist V : [0, +∞[×Rn → R+ a continuously differentiable
function, some constants c1 > 0, c2 ≥ 0, a ≥ 0, and m ≥ 1 and scalar functions
µ, π ∈ PC(R+ , R), such that, for all t ≥ 0 and x ∈ Rn ,
c1 kykm ≤ V (t, x) ≤ c2 kxkm + a, (16)
∂V ∂V
(t, x) + (t, x)F (t, x) ≤ µ̃(t)V (t, x) + π̃(t).
∂t ∂x
Theorem 4.2. Under assumption (H4 ) and (H5 ) the perturbed system (10) is glob-
ally uniformly practically exponentially y-stable if µ̃ is π̃−globally uniformly practi-
cally exponentially stable.
Proof. We can use the same arguments as the proof of the Theorem 4.1.
Remark 6. Note that, if we assume in the Theorem 4.1, that
 ∂V 
(t, x)g(t, x)≤ S(t)kykm ,
 
∂x

 PRACTICAL PARTIAL STABILITY OF TIME-VARYING SYSTEMS 3599

then, the perturbed system (10) is globally uniformly practically exponentially

y−stable if µ̃ is π̃−globally uniformly practically exponentially stable.
Example 2. Let’s consider the following three-dimensional system.

x1 π(t) 1 x1 ψ(t) 
ẋ1 = µ(t)x1 + + S(t) x1 +


1 + kxk 2 2 2(1 + kxk2 )







x2 π(t) 1 x2 ψ(t)  (17)
ẋ2 = µ(t)x2 + 2
+ S(t) x2 +
1 + kxk 2 2(1 + kxk2 )








ẋ3 = x1 x2 + x3

Z t
where x = (x1 , x2 , x3 ) ∈ R2 , t ∈ R+ , µ, π ∈ PC(R+ , R), and ψ(t) = exp( µ(s)ds).
0
The system has the form of (10) with
 
x1 π(t)
 µ(t)x1 + 1 + kxk2 
 
 
 
 µ(t)x2 + x2 π(t)
f (t, x) =  


 1 + kxk2 

 
x1 x2 + x3
and  
1 x1 ψ(t)
 S(t)( 2 x1 + 2(1 + kxk2 ) ) 
 
 
 
 S(t)( 1 x2 + x2 ψ(t) )
g(t, x) =  .

 2 2(1 + kxk2 ) 
 
 
0
Let
1 2
V (t, x) = (x + x22 ) + ψ(t).
2 1
The derivative along the trajectories of
ẋ = f (t, x), (18)
is given by
V̇ (t, x) = x˙1 (t)x1 (t) + x2 (t)x˙2 (t) + µ(t)ψ(t)
x1 π(t)
x2 π(t)

= µ(t)x1 + 2
x1 + x2 µ(t)x2 + +µ(t)ψ(t)
1 + kxk 1 + kxk2
≤ 2µ(t)V (t, x) + π(t).
Therefore, if µ is π−globally uniformly practically exponentially stable, then, the
nominal system (18) is globally uniformly practically exponentially (x1 , x2 )−stable.
Moreover, we have
 ∂V 
(t, x)g(t, x)≤ S(t)V (t, x),
 
∂x

then, the perturbed system (17) is globally uniformly practically exponentially
(x1 , x2 )−stable if S ∈ BN , such that δ satisfies (15).
3600 ABDELFETTAH HAMZAOUI, NIZAR HADJ TAIEB AND MOHAMED ALI HAMMAMI

5. Illustrative example. The basic SIR epidemic model, see ([14], [16]), with
generalized Logistic death rate and standard incidence rate is given by the following
system:
  νN (t)  S(t)I(t)
Ṡ(t) = bN (t) − d + S(t) − β


k N (t)






˙ =β S(t)I(t)  νN (t) 
 I(t) − d+ I(t) − (γ + α)I(t)

N (t) k (19)
  νN (t) 

 Ṙ(t) = γI(t) − d + R(t)
k



 Ṅ (t) = bN (t) − d + νN (t) N (t) − αI(t)

  

k

where N (t), S(t), I(t) and R(t) denote the densities of host population, the suscep-
tible, the infective, and the recovered of the host population at time t, respectively.
The model are derived with the following assumptions:
• All of the new born are susceptible, and the birth rate is proportional to the
existing host population, with proportionality constant b > 0.
• The death rates of the host population, the susceptible, the infective and the
recovered, equal to

 νN (t)   νN (t)   νN (t)   νN (t) 

d+ N (t), d + S(t), d + I(t) + αI(t), d + R(t),
k k k k

respectively, where d is the natural death rate, ν = b−d > 0 the initial growth
rate constant, carrying capacity k, and α ≥ 0 means the death rate due to the
disease.
S(t)I(t)
• The incidence rate is the standard incidence, i.e., β , where β > 0 is
N (t)
the contact rate constant.
• The conversing rate from infective class to recovered class is proportional to
the size of the existing infective population with proportionality γ > 0.
S(t) I(t) R(t)
Let’s s(t) = , i(t) = and r(t) = , then, system (19) becomes
N (t) N (t) N (t)

ṡ(t) = b(1 − s(t))N (t) − (β − α)s(t)i(t)



 i̇(t) = βs(t)i(t) + αi2 (t) − (α + γ + b)i(t)


ṙ(t) = γi(t) + αi(t)r(t) − br(t) (20)
 Ṅ (t) = N (t) ν 1 − N (t) −αi(t) .

 h   i

k

Here, we will consider a special case which is α = 0. Thus, system (20) becomes

ṡ(t) = b(1 − s(t))N (t) − βs(t)i(t)




 i̇(t) = βs(t)i(t) − (γ + b)i(t)

ṙ(t) = γi(t) − br(t) (21)
 Ṅ (t) = N (t)ν 1 − N (t) .

  

k
 PRACTICAL PARTIAL STABILITY OF TIME-VARYING SYSTEMS 3601

γ
We suppose that + 1 ≤ N (t) ≤ k and s(t) + i(t) ≤ 1. Let’s x = (y, z) with
2b
y = (s, i) and z = (r, N ). The system (21) has the form of (10) with
−bs(t)N (t) + bN (t)
 
 −(γ + b)i(t) 
f (t, x) =  γi(t) − br(t)
 

N (t) 
  
N (t)ν 1 −
k
and  
−βs(t)i(t)
 βs(t)i(t) 
g(t, x) =  .
 0 
0
One can see here that the origin is not an equilibrium point of the system (21).
Then, let’s consider the following Lyapunov function
1 2
(s + i2 ) + e−(γ+b)t
V (t, x) =
2
1
which satisfies the inequality (16) with c1 = c2 = and a = 1. The derivative along
2
the trajectories of
ẋ = f (t, x), (22)
is given by
V̇ (t, x) = ṡ(t)s(t) + i(t)i̇(t) − (γ + b)e−(γ+b)t
= −bN (t)s2 (t) + bs(t)N (t) − (γ + b)i2 (t) − (γ + b)e−(γ+b)t
s2 (t) N 2 (t) γ + b 2
≤ −bN (t)s2 (t) + b +b − i (t) − (γ + b)e−(γ+b)t .
2 2 2

γ
Since + 1 ≤ N (t) ≤ k, then,
2b
bk 2
V̇ (t, x) ≤ −(γ + b)V (t, x) + .
2
bk 2
Let µ(t) = −γ − b and π(t) = . It is easy to see that µ is π−globally uniformly
2
practically exponentially stable. Consequently, the nominal system (22) is globally
uniformly practically exponentially y−stable. Moreover, since s + i ≤ 1, then,
 ∂V   
(t, x)g(t, x) = −βs2 i + βsi2 
   
∂x

≤ 2βV (t, x).
It is clear to see that 2β ∈ BN , with M (δ) = 2βδ. Thus, the assumption (H4 ) is
satisfied. Then, by the theorem 4.1, the system (21) is globally uniformly practically
exponentially y−stable for all δ > 0, such that θ = γ + b > 2β.
Now, inversely, we only need to consider the following subsystem of system (21)

ṡ(t) = −bs(t)N (t) + bN (t) − βs(t)i(t)
(23)
i̇(t) = βs(t)i(t) − (γ + b)i(t).
3602 ABDELFETTAH HAMZAOUI, NIZAR HADJ TAIEB AND MOHAMED ALI HAMMAMI

Let’s take 
b(1 − s(t))N (t) − βs(t)i(t)
F1 (t, y, z) =
βs(t)i(t) − (γ + b)i(t).
One can see that the assumption (H1 ) is satisfied with
L(t) = b2 k 2 + 4β 2 + 2bβk + (γ + b)2 and K(t) = b2 k 2 .

6. Conclusion. We have studied in this paper the global uniform practical asymp-
totic and exponential partial stability of time varying systems. A converse theorem
is established to guarantee the practical exponential partial stability of a class of
time-varying systems using the Lyapunov theory. Moreover, the asymptotic behav-
iors of solutions of perturbed systems are studied and some examples are given that
show the applicability of the results.

Acknowledgments. The authors wish to thank the editor and the anonymous
reviewers for their valuable and careful comments.

REFERENCES

[1] A. Benabdallah, M. Dlala and M. A. Hammami, A new Lyapunov function for stability of
time-varying nonlinear perturbed systems, Systems and Control Letters, 56 (2007), 179–187.
[2] A. Benabdallah, I. Ellouze and M. A. Hammami, Practical stability of nonlinear time-varying
cascade systems, Journal of Dynamical and Control Systems, 15 (2009), 45–62.
[3] A. Ben Makhlouf, Partial practical stability for fractional-order nonlinear systems, Mathe-
matical Methods in the Applied Sciences, (2020).
[4] T. Caraballo, F. Ezzine, M. A. Hammami and L. Mchiri, Practical stability with respect to a
part of variables of stochastic differential equations, Stochastics, 93 (2021), 647–664.
[5] M. Yu. Filimonov, Global asymptotic stability with respect to part of the variable for solutions
of systems of ordinary differential equations, Differential Equations, 56 (2020), 710–720.
[6] B. Ghanmi, N. Hadj Taieb and M. A. Hammami, Growth conditions for exponential stability
of time-varying perturbed systems, International Journal of Contol, 86 (2013), 1086–1097.
[7] N. Hadj Taieb and M. A. Hammami, Some new results on the global uniform asymptotic
stability of time-varying dynamical systems, IMA Journal of Mathematical Control and In-
formation, 35 (2018), 901–922.
[8] N. Hadj Taieb, Stability analysis for time-varying nonlinear systems, International Journal
of control, (2020), https://doi.org/10.1080/00207179.2020.1861332.
[9] N. Hadj Taieb, Indefinite derivative for stability of time-varying nonlinear systems, IMA
Journal of Mathematical Control and Information, 38 (2021), 534–551.
[10] J. W. Hagood and S. T. Brian, Recovering a function from a Dini derivative, Amer. Math.
Monthly, 113 (2006), 34–46.
[11] H. Khalil, Nonlinear Systems, Third ed. Prentice-Hall Englewood Cliffs, NJ, 2002.
[12] V. Lakshmikantham, S. Leela and A. A. Martynyuk, Practical Stability of Nonlinear Systems,
World Scientific Publishing Co., Inc., Teaneck, NJ, 1990.
[13] J. Lasalle and S. Lefschetz, Stability by lyapunov direct method, with application, Mathe-
matics in Science and Engineering, Academic Press, New York-London, 4 (1961).
[14] Q. G. Linda, M. L. Jaime and W. H. Herbert, Four SEI endemic models with periodicity and
separatrices, Math. Biosci., 128 (1995), 157–184.
[15] A. Martynyuk and Z. Sun, Practical stability and its applications, Beijing: Science Press,
(2003).
[16] S. Ruiqing, J. Xiaowu and C. Lansun, The effect of impulsive vaccination on an SIR epidemic
model, Applied Mathematics and Computation, 212 (2009), 305–311.
[17] E. D. Sontag, Smoimoth stabilization implies copre factorization, IEEE Trans. Automat.
Control, 34 (1989), 435–443.
[18] V. V. Rumyantsev, Partial stability of motion, Mosk. Gos. Univ., Mat. Mekh. Fiz. Astronom.
Khim., 4 (1957), 9–16.
[19] V. V. Rumyantsev, Stability of equilibrium of a body with a liquid-filled hollow, Dokl. Akad.
Nauk SSSR, 124 (1959), 291–294.
 PRACTICAL PARTIAL STABILITY OF TIME-VARYING SYSTEMS 3603

[20] V. V. Rumyantsev, Stability of rotational motion of a liquid-filled solid, Prikl. Mat. Mekh.,
23 (1959), 1057–1065.
[21] V. V. Rumyantsev, Stability of motion of a gyrostat, Prikl. Mat. Mekh., 25 (1961), 9–19.
[22] V. V. Rumyantsev, Stability of motion of solids with liquid-filled hollows, Tr. II Vses. sâezda
Po Teor. Prikl. Mekh. Proc. 2 All-Union Congress: Theoretical and Applied Mechanics,
Moscow: Nauka, 1 (1965), 57–71.
[23] B. Zhou, On asymtotic stability of linear time-varying systems, Automatica, 68 (2016), 266–
276.
[24] B. Zhou, Stability analysis of nonlinear time-varying systems by lyapunov functions with
indefinite derivatives, IET Control Theory and Applications, 11 (2017), 1434–1442.

Received January 2021; revised April 2021. Early access August 2021.
E-mail address: [email protected]
E-mail address: [email protected]
E-mail address: [email protected]