In this paper, a class of uncertain switching time-delay systems with nonlinear
perturbations is considered. The system parameter uncertainties are time-varying and unknown
with norm-bounded. The delay in the system states is also time-varying. By using an improved
Lyapunov-Krasovskii functional, a state dependent switching rule for robust exponential stability
is designed in terms of solution of Lyapunov-type equations and growth bound of perturbations.
The approach allows for computation of the bounds that characterize the exponential stability rate
of the solution. Numerical examples are given to illustrate the results.

9 trang |

Chia sẻ: thuyduongbt11 | Ngày: 09/06/2022 | Lượt xem: 151 | Lượt tải: 0
Bạn đang xem nội dung tài liệu **Robust exponential stability of uncertain switching time-delay systems with nonlinear perturbations**, để tải tài liệu về máy bạn click vào nút DOWNLOAD ở trên

ROBUST EXPONENTIAL STABILITY
OF UNCERTAIN SWITCHING TIME - DELAY SYSTEMS
WITH NONLINEAR PERTURBATIONS
Le Van Hien
Hanoi National University of Education
Abstract. In this paper, a class of uncertain switching time-delay systems with nonlinear
perturbations is considered. The system parameter uncertainties are time-varying and unknown
with norm-bounded. The delay in the system states is also time-varying. By using an improved
Lyapunov-Krasovskii functional, a state dependent switching rule for robust exponential stability
is designed in terms of solution of Lyapunov-type equations and growth bound of perturbations.
The approach allows for computation of the bounds that characterize the exponential stability rate
of the solution. Numerical examples are given to illustrate the results.
KEY WORDS: Switching system, time-delay system, uncertainty, nonlinear perturba-
tion, exponential stability, Lyapunov equation.
1 Introduction
Switching systems belong to an important class of hybrid systems, which are described
by a family of differential equations together with specified rules to switch between them.
A switching system can be represented by a differential equation of the form
x˙(t) = fσ(t)(t, x), t ≥ 0,
where {fσ(., .) : σ ∈ I} is a family of functions parameterized by some index set I, which
is typically a finite set, and σ(.), which depends on system state at each time, is the
switching rule/signal determining a switching sequence for the given system.
Switching systems arise in many practical processes that cannot be described by ex-
clusively continuous or exclusively discrete models, such as manufacturing, communication
networks, automotive engineering control, chemical processes (e.g. see [2], [4], [9], [10], [11]
and the references therein).
Although many important results have been obtained for linear switched systems,
there are few results concerning the stability of switched linear systems with time delay.
Under the assumption on commutative system matrices, it was shown in [5] that when all
subsystems are asymptotically stable, the switching system is asymptotically stable under
an arbitrary switching rule. The asymptotic stability for switching linear systems with time
delay has been studied in [15], but the result was limited to symmetric systems. In [7], [8],
[13], delay-dependent asymptotic stability conditions are extended to discrete-time linear
switching systems with time delay. There are some other results concerning asymptotic
stability for switching linear systems with time delay, but most of them provide conditions
for asymptotic stability for arbitrary switching signal without focusing on exponential
stability.
1
The exponential stability problem was considered in [16] for switching linear systems
with impulsive effects by using the matrix measure concept, and in [14] for nonholonomic
chained systems with strongly nonlinear input/state driven disturbances and drifts. On
the other hand, it is worth noting that the existing stability conditions for time-delay
systems must be solved upon a grid of the parameter space, which results in testing
a nonlinear Riccati-type equation or a finite number of LMIs. In this case, the results
using finite gridding points are unreliable and the numerical complexity of the tests grows
rapidly. Therefore, finding new conditions for the exponential stability of uncertain linear
switching time-delay systems is of interest.
In this paper, the problem of exponential stability for a class of uncertain hybrid time
varying-delay systems with nonlinear perturbations is addressed. Our objective is to de-
rive delay-dependent conditions for the robust exponential stability by using an improved
Lyapunov-Krasovskii functional. The conditions will be presented in terms of the solution
of Lyapunov-type equations and growth bounds of perturbations. Comparing with the
previous results, a simple geometric design is employed to find the switching rule and our
approach allows to compute simultaneously the two bounds that characterize the exponen-
tial stability rate of the solution. The results obtained can be considered as an extension
of existing results for linear switching time-delay systems.
The paper is organized as follows: Section 2 presents notations, definitions and a
technical lemma required for the proof of the main results. Sufficient conditions for the
robust exponential stability and illustrative examples are presented in Section 3. The paper
ends with a conclusion followed by cited references.
2 Problem formulation
The following notations will be used throughout this paper. R+ denotes the set of all
real non-negative numbers; Rn denotes the n-dimensional space with the scalar product
〈., .〉 and the vector norm ‖.‖; Rn×r denotes the space of all matrices of (n×r)− dimension.
AT denotes the transpose of A; I denotes the identity matrix; λ(A) denotes the set of all
eigenvalues of A; λmax(A) = max{Reλ : λ ∈ λ(A)}; λmin(A) = min{Reλ : λ ∈ λ(A)};
A matrix A is semi-positive definite (A ≥ 0) if 〈Ax, x〉 ≥ 0, for all x ∈ Rn;A is positive
definite (A > 0) if 〈Ax, x〉 > 0 for all x 6= 0; A ≥ B means A−B ≥ 0. Consider a class of
uncertain hybrid time-delay systems of the form{
x˙(t) = Aσ(t)x(t) +Dσ(t)(t)x(t− h(t)) + fσ(t)(t, x(t), x(t − h(t))), t ∈ R
+,
x(t) = φ(t), t ∈ [−h, 0],
(2.1)
where x(t) ∈ Rn is the system state; σ(.) : R+ → I := {1, 2, . . . ,m} is the switching func-
tion, which will be designed. Aσ,Dσ ∈ {[Ai,Di], i = 1, 2, . . . ,m}, Ai,Di are given matrices
and φ(t) ∈ C([−h, 0], Rn) is the initial function with the norm ‖φ‖ = sups∈[−h,0] ‖φ(s)‖;
the parameter perturbations fi(t, y, z) are unknown and assumed to satisfy the following
condition:
‖fi(t, y, z)‖ ≤ βi‖y‖+ γi‖z‖, ∀t ∈ R
+, y, z ∈ Rn,
where βi, γi, i = 1, 2, . . . ,m are given non-negative constants. The time-varying delay func-
tion h(t) is assumed to satisfy the following well-studied condition
0 ≤ h(t) ≤ h, h˙(t) ≤ µ < 1,
2
where h and µ are given non-negative constants. This assumption means that the time
delay may change from time to time but the rate of changing is bounded, i.e. the delay
cannot increase as fast as the time itself.
Definition 2.1. Given α > 0.
The system (2.1) is robustly α−exponentially stable if there exists a switching function
σ(.) and positive number N such that any solution x(t, φ) of the system satisfies
‖x(t, φ)‖ ≤ Ne−αt‖φ‖, ∀t ∈ R+, (2.2)
for all the admissible uncertainties.
Definition 2.2. (see [12]) The system of matrices {Li}, i = 1, 2, . . . ,m, is said to be
strictly complete if for any x ∈ Rn\{0} there is i ∈ {1, 2, . . . ,m} such that
xTLix < 0.
Let us define
Ωi = {x ∈ R
n : xTLix < 0}, i = 1, 2, . . . ,m.
It's easy to show that the system {Li}, i = 1, 2, . . . ,m, is strictly complete if and only if
m⋃
i=1
Ωi = R
n\{0}. (2.3)
Remark 2.3. A sufficient condition for the strict completeness of the system of ma-
trices {Li} is that there exists λi ≥ 0, i = 1, 2, . . . ,m such that
∑m
i=1 λi > 0 and
m∑
i=1
λiLi < 0.
As shown in [12], if m = 2 then the above condition is also necessary for the strict
completeness.
Next, we introduce the following well known lemma, which will be used in the proof
of our results.
Lemma 2.4.(see [6]) For any x, y ∈ Rn and positive definite matrix X one has
2xT y ≤ xTX−1x+ yTXy.
3 Main results
In the sequel, for the sake of brevity, we will denote σ for the switching signal σ(.).
For given non-negative numbers α, h, µ and symmetric positive definite matrices P,R, S
we set
τ = (1− µ)−1, η = (2α+ e2αh), β = max
1≤i≤m
βi, γ = max
1≤i≤m
γi;
Li = A
T
i P + PAi + ηP + τQ+ hR+ S;Q =
m∑
i=1
DTi PDi (3.1)
Theorem 3.1.
The system (2.1) is robustly α-exponentially stable if there exists symmetric positive
definite matrices P,R, S and a positive scalar ε such that the following conditions hold
3
i) The system of matrices {Li}, i = 1, 2, . . . ,m is strictly complete.
ii) β2 + γ2 ≤ λmin
[
εS − ε2(1 + τe2αh)PP
]
.
Moreover, the solution x(t, φ) of the system satisfies
‖x(t, φ)‖ ≤
√
N2
N1
e−αt‖φ‖, t ∈ R+,
where N1 = λmin(P );
N2 = λmax(P ) + h[τ
m∑
i=1
λmax(D
T
i PDi)
+ λmin(S − ε(1 + τe
2αh)PP ) +
1
2
hλmax(R)]
Proof. Consider the following Lyapunov-Krasovskii functional
V (xt) = V1(x(t)) + V2(xt) + V3(xt) + V4(xt)
where xt(s) = x(t+ s), s ∈ [−h, 0] and
V1(x(t)) = x
T (t)Px(t), (3.2)
V2(xt) = τ
t∫
t−h(t)
e2α(s−t)xT (s)Qx(s)ds, (3.3)
V3(xt) = ε
−1γ2
t∫
t−h(t)
e2α(s−t)‖x(s)‖2ds, (3.4)
V4(xt) =
0∫
−h
t∫
t+s
e2α(s−t)xT (θ)Rx(θ)dθds. (3.5)
From (3.2) - (3.5) and ii) of Theorem 3.1, it's easy to verify that
N1‖x(t)‖
2 ≤ V (xt) ≤ N2‖xt‖
2, t ≥ 0. (3.6)
Taking derivative of V1(x(t)) = x
T (t)Px(t) along trajectories of any subsystem ith we have
V˙1(x(t)) = x
T (t)[ATi P + PAi]x(t) + 2x
T (t)PDix(t− h(t))
+ 2xT (t)Pfi(t, x(t), x(t− h(t))).
Applying Lemma 2.4 gives
2xT (t)PDix(t− h(t)) ≤ e
2αhxT (t)Px(t) + e−2αhxT (t− h(t))DTi PDix(t− h(t))
≤ e2αhxT (t)Px(t) + e−2αhxT (t− h(t))Qx(t − h(t)). (3.7)
2xT (t)Pfi(t, x(t),x(t− h(t))) ≤ 2‖x
T (t)P‖(βi‖x(t)‖ + γi‖x(t− h(t))‖)
≤ εxT (t)PPx(t) + ε−1β2i ‖x(t)‖
2
+ ετe2αhxT (t)PPx(t) + ε−1(1− µ)e−2αhγ2i ‖x(t− h(t))‖
2. (3.8)
4
Next, taking derivative of V2(xt) and V3(xt), respectively, along trajectories of subsys-
tem ith yields
V˙2(xt) =− 2αV2(xt) + τx
T (t)Qx(t)− τ(1− h˙(t))xT (t− h(t))e−2βh(t)Qx(t− h(t))
≤− 2αV2(xt) + τx
T (t)Qx(t)− e−2αhxT (t− h(t))Qx(t− h(t)), (3.9)
V˙3(xt) =− 2αV3(xt) + ε
−1γ2‖x(t)‖2 − ε−1γ2(1− h˙(t))e−2αh(t)‖x(t− h(t))‖2
≤− 2αV3(xt) + ε
−1γ2‖x(t)‖2 − e−1γ2(1− µ)e−2αh‖x(t− h(t))‖2. (3.10)
Finally, taking derivative of V4(xt) we get
V˙4(xt) =− 2αV4(xt) + hx
T (t)Rx(t)−
0∫
−h
e2αsxT (t+ s)Rx(t+ s)ds
≤− 2αV4(xt) + hx
T (t)Rx(t). (3.11)
From (3.1), (3.6) - (3.11) we get
V˙ (xt) + 2αV (xt) ≤ x
T (t)[ATi P + PAi + ηP + τQ+ hR]x(t)
+ ε(1 + τe2αh)xT (t)PPx(t) +
β2 + γ2
ε
‖x(t)‖2
= xT (t)Lix(t)−
1
ε
xT (t)
[
εS − ε2(1 + τe2αh)PP − (β2 + γ2)I
]
x(t). (3.12)
From condition ii) of the theorem 3.1 we have
xT (t)
[
εS − ε2(1 + τe2αh)PP
]
x(t)− (β2 + γ2)‖x(t)‖2 ≥ 0.
Therefore,
V˙ (xt) + 2αV (xt) ≤ x
T (t)Lix(t). (3.13)
Let us set
Ωi = {x ∈ R
n : xTLix < 0}, i = 1, 2, . . . ,m.
Then by the strict completeness of the system of matrices {Li}, and from (2.3) it follows
that
m⋃
i=1
Ωi = R
n\{0}.
Defining the sets
Ω˜1 = Ω1, Ω˜i = Ωi\
i−1⋃
j=1
Ω˜j, i = 2, 3, . . . ,m.
we see that
m⋃
i=1
Ω˜i = R
n\{0}, Ω˜i ∩ Ω˜j = ∅, i 6= j.
Therefore, for any x(t) ∈ Rn, t ≥ 0, there exists i ∈ {1, 2, . . . ,m} such that x(t) ∈ Ω˜i. By
choosing switching rule as σ(x(t)) = i whenever x(t) ∈ Ω˜i, then from (3.13) we have
V˙ (xt) + 2αV (xt) ≤ x
T (t)Lix(t) ≤ 0, t ≥ 0, i = 1, 2, . . . ,m.
5
This implies that V (xt) ≤ V (φ)e
−2αt, t ≥ 0.
Taking (3.2) into account, we get
N1‖x(t, φ)‖
2 ≤ V (xt) ≤ V (φ)e
−2αt ≤ N2e
−2αt‖φ‖2, t ≥ 0,
and then
‖x(t, φ)‖ ≤
√
N2
N1
e−αt‖φ‖, t ≥ 0,
which concludes the proof of the theorem 3.1.
Remark 3.2. Note that by Remark 2.2, the system {Li} is strictly complete if there
exists λi ≥ 0, i = 1, 2, . . . ,m,
∑m
i=1 λi > 0 such that
m∑
i=1
λiLi < 0. (3.14)
In this case, the switching rule can be chosen as
σ(t) = argmin{xT (t)Lix(t)}, t ≥ 0.
Indeed, as shown in the proof of Theorem 3.1, we have arrived at the estimation
V˙ (xt) + 2αV (xt) ≤ x
T (t)Lix(t), t ≥ 0.
Since λi ≥ 0 and λ =
∑m
i=1 λi > 0, so
min
i=1,2,...,m
xT (t)Lix(t) ≤ λ
−1
m∑
i=1
λix
T (t)Lix(t).
By choosing switching rule as
σ(t) = argmin{xT (t)Lix(t)}, t ≥ 0,
we have
V˙ (xt) + 2αV (xt) ≤ x
T (t)Lix(t) ≤ λ
−1
m∑
i=1
λix
T (t)Lix(t) ≤ 0.
This leads to
‖x(t, φ)‖ ≤
√
N2
N1
e−αt‖φ‖, t ≥ 0,
as desired. The following procedure can be applied to construct the switching rule.
Step 1. Define the symmetric positive definite matrices P,Q,R, S (i.e. the solution of
the matrix inequality (3.14)) such that the system {Li} is strictly complete.
Step 2. Construct the sets Ωi, and then Ω˜i.
Step 3. The switching rule is chosen as σ(x(t)) = i, whenever x(t) ∈ Ω˜i.
Example 3.3. Consider the system (2.1), where m = 2, h(t) = 0.5sin2t and
[A1,D1] =
[(
−41 1
−1 1
)
,
(
1 −1
1 −1
)]
,
[A2,D2] =
[(
1 −1
1 −35
)
,
(
1 −1
3 −4
)]
,
6
Note that, both matrices A1 and A2 are unstable. In this case, we have h = 0.5,
µ = 0.5, τ = 2. For α = 0.5, we verify that the symmetric positive definite matrices
P =
(
0.1402 −0.1445
−0.1445 0.2730
)
, R =
(
1.0416 0.0168
0.0168 1.0920
)
, S = I =
(
1 0
0 1
)
be solution of (3.14) and
M = 0.5L1 + 0.5L2 < −0.04I,
where
L1 =
(
−5.6592 0.2671
0.2671 9.3802
)
, L2 =
(
5.5418 −0.3343
−0.3343 −9.6992
)
.
Therefore, the system {L1, L2} is strictly complete. The sets Ω1,Ω2 are defined as
Ω1 = {(x, y) ∈ R
2 : −5.6592x2 + 0.5342xy + 9.3802y2 < 0},
Ω2 = {(x, y) ∈ R
2 : 5.5418x2 − 0.6686xy − 9.6992y2 < 0},
which can be represented by Figure 1.
It can be seen that Ω1 ∪ Ω2 = R
2\{0}. Therefore, the switching regions are given as
Ω˜1 = {(x, y) ∈ R
2 : −5.6592x2 + 0.5342xy + 9.3802y2 < 0};
Ω˜2 = {(x, y) ∈ R
2 : −5.6592x2 + 0.5342xy + 9.3802y2 ≥ 0, (x, y) 6= (0, 0)}.
We have Ω˜1 ∪ Ω˜2 = R
2\{0}, Ω˜1 ∩ Ω˜2 = ∅.
Choose ε = 0.89 then by Theorem 3.1, the system (2.1) is 0.5−exponentially stable
with the switching rule
σ(t) =
{
1 if x(t) ∈ Ω˜1,
2 if x(t) ∈ Ω˜2
and the grown bounds of the perturbations satisfy
β2 + γ2 ≤ 0.4349
Moreover, the solution of the system satisfies
‖x(t, φ)‖ ≤ 11.3e−0.5t‖φ‖, ∀t ∈ R+.
Figure 1.
7
For the case when m = 1 (without switching) and α = 0, β = 0,D = 0, Theorem
3.1 gives a delay-dependent criteria for the robust stability of linear systems with delayed
perturbations considered in [1].
Consider the linear system with delayed perturbations of the form [1]
x˙(t) = Ax(t) + f(t, x(t− h(t))), t ∈ R+, (3.15)
where ‖f(t, x(t− h(t)))‖ ≤ γ‖x(t− h(t))‖.
Note that, the stability conditions in Theorem 1 in [1] is independent of the size of
delay. If R is small, Theorem 1 in [1] can be derived from the following corollary.
Corollary 3.4. The system (3.15) is robust asymptotically stable if there exist positive
definite matrices P,R, S, a positive scalar ε such that the following conditions hold
i) ATP + PA+ hR+ S < 0;
ii) γ2 ≤ λmin
[
εS −
ε2
1− µ
PP
]
.
Proof. Consider the following Lyapunov-Krasovski functional
V (xt) =x
T (t)Px(t) + ε−1γ2
t∫
t−h(t)
‖x(s)‖2ds
+
0∫
−h
t∫
t+s
e2α(s−t)xT (θ)Rx(θ)dθds.
By the same argument used in the proof of Theorem 3.1 we get
V˙ (xt) ≤x
T (t)[ATP + PA+ hR]x(t) +
ε
1− µ
xT (t)PPx(t) + ε−1γ2‖x(t)‖2
≤xT (t)[ATP + PA+ hR+ S]x(t)−
1
ε
xT (t)
[
εS −
ε2
1− µ
PP − γ2I
]
x(t)
≤xT (t)[ATP + PA+ hR+ S]x(t) < 0.
Therefore, the system (3.15) is robust asymptotically stable [3, Theorem 2.1, p. 132], which
conclude the proof.
4 Conclusion
This paper has proposed a switching design for robust exponential stability of un-
certain linear switching time-delay systems with nonlinear perturbations. The stability
conditions have been derived in terms of the solution of Lyapunov-type equations and
grown bounds of the uncertain parameters. The approach has allowed for the use of ef-
ficient techniques for computation of the two bounds that characterize the exponential
stability rate of the solutions. Numerical illustrates the result given.
8
REFERENCES
[1] Alpaslan Parlakci M.N., 2004. Robust stability of linear systems with delayed per-
turbations. Proceeding on Circuits, Signals, and Systems. Clearwater Beach, FL, USA.
[2] DeCarlo R.A., Branicky M.S., Pettersson S., Lennartson B, 2000. Perspectives and
results on the stability and stabilizability of hybrid systems. Proceedings of IEEE, 87, pp.
1069-1082.
[3] Jack K. Hale and S.M. Verduyn Lunel, 1993. Introduction to Functional Differential
Equations. Springer Verlag, New York.
[4] Liberzon, D., 2003. Switching in Systems and Control. Birkhauser, Boston.
[5] Narendra K.S. and Balakrishnan V., 1994. A common Lyapunov function for stable
LTI systems with commuting matrices. IEEE Trans. Aut. Contr, 39, pp. 2469 - 2471.
[6] Park P., 1999. A delay-dependent stability criterion for systems with uncertain
time-invariant delays. IEEE Trans. Aut. Contr, 44, pp. 876-877.
[7] Phat V.N., 2005. Robust stability and stabilizability of uncertain linear hybrid sys-
tem with state delays. IEEE Trans. on CAS II, 52, pp. 94-98.
[8] Phat V.N. and S. Pairote, 2006. Global stabilization of linear periodically time-
varying switched systems via matrix inequalities. J. Contr. Theory Appl., 1, pp. 24-29.
[9] Savkin A.V. and Evans R.J., 2002. Hybrid Dynamical Systems: Controller and
Sensor Switching Problems, Boston, Birkhauser.
[10] Sun Z. and Ge S.S., 2005. Switched Linear Systems: Control and Design, Springer-
London.
[11] Sun Zh. Ge S.S., 2005. Analysis and synthesis of switched linear control systems.
Automatica, 41, pp. 181-195.
[12] Uhlig F., 1979. A recuring theorem about pairs of quadratic forms and extensions.
Linear Algebra and Application, 25, pp. 219-237.
[13] Xie G. and Wang L., December, 2004. Quadratic stability and stabilization of
discrete-time switched systems with state delay. Proc. IEEE Conf. on Decision and Control.,
Atlantics, pp. 3235-3240.
[14] Zairong Xi, Feng G., Jiang Z.P. and Cheng D., 2003. A switching algorithm for
global exponential stabilization of uncertain chained systems. IEEE Trans. Aut. Contr., 48,
pp. 1793-1798.
[15] Zhai G., Lin H., Antsaklis P.L., 2003. Quadratic stabilizability of switched linear
systems with polytopic uncertainties. Int.J. of Control., 76, pp. 2469 - 2471.
[16] Zong G.D. and Wu Y.Q., December, 2003. Exponential stability of a class of
switched and hybrid systems. Proc. IEEE on Contr. Aut. Robotics and Vision, Kuming,
China, pp. 2244-2249.
9