This paper investigates a class of impulsive pulse-width sampler systems and its steadystate
control in the infinite dimensional spaces. Firstly, some definitions of pulse-width sampler systems
with impulses are introduced. Then applying impulsive evolution operator and fixed point theorem, some
existent results of steady-state of infinite dimensional linear and semilinear pulse-width sampler systems
with impulses are obtained. An example to illustrate the theory is presented in the end.
1. Introduction
In the design of distributed parameter control systems, one of the important problems is to choose controller and actuator. As the dimension of an industrial controller in actual applications is finite, it restricts us to consider the distributed parameter system with a finite dimensional output. In industrial process control systems, on-off actuators have been in engineer's good graces because of the cheap price and the high reliability.
The interest in the pulse-width sampler control systems was aroused as early as 1960s. It was motivated by applications to engineering problems and neural nets modeling. In modern times, the development of neurocomputers promises a rebirth of interest in this field. The theory of pulse-width sampler control systems is treated as a very important branch of engineering and mathematics. Nevertheless, it can supply a technical-minded mathematician with a number of new and interesting problems of mathematical nature. There are some results such as steady-state control, stability analysis, robust control of pulse-width sampler systems [1–7], integral control by variable sampling based on steady-state data, and adaptive sampled-data integral control [8–11].
On the other hand, in order to describe dynamics of population, subject to abrupt changes as well as other phenomena, such as harvesting, diseases and so forth, some authors have used impulsive differential equations to describe the model since the last century. The reader can refer the basic theory of impulsive differential equations in finite dimensional spaces to Lakshmikantham's book [12]. Meanwhile, the impulsive evolution equations and its optimal control problems on infinite dimensional Banach spaces have been investigated by many authors including Ahmed, Liu, Nieto, and us (see for instance [13–25] and references therein).
However, to our knowledge, the pulse-width sampler systems with impulse on infinite dimensional spaces have not been investigated extensively. In this paper, we first study the following steady-state control of infinite dimensional linear system with impulses
ẋ(t)=Ax(t)+f(t)+Cu(t),t≠τk,Δx(τk)=Bkx(τk)+ck,k=1,2,…,z(t)=K1x(t),
where the state variable x(t) takes values in a reflexive Banach space X, A is the infinitesimal generator of a C0-semigroup {T(t),t≥0} on the state space X, f(t)=f·1(t) is T0-periodic step disturbance of the system and f∈X. Control variable u(t)∈ℝq, the input C:ℝq→X is a bounded linear operator. There is only one time sequences {τk∣k∈ℤ0+} satisfing 0<τ1<τ2<⋯<τk⋯ and limk→∞τk=∞, Bk:X→X, 0<τ1<τ2<⋯<τδ<T0, τk+δ=τk+T0, Δx(τk)=x(τk+)-x(τk-), x(τk+)=limh→0+=x(τk+h) and x(τk-)=x(τk)represent, respectively the right and left limits of x(t) at t=τk. K1:X→ℝp is a given bounded linear operator; z(t) is the p dimensional output of the system (1.1).
We, then, study the following steady-state control of infinite dimensional semilinear system with impulses
ẋ(t)=Ax(t)+f(t,x(t))+Cu(t),t≠τk,Δx(τk)=Bkx(τk)+ck,k=1,2,…,z(t)=K1x(t),
where f:[0,∞)×X→X is T0-periodic continuous function.
Suppose that control signal u(t) is the output of the q dimensional pulse-width sampler controller, and v(t) is the input of the q dimensional pulse-width sampler controller, which is the output of some dynamical controller
v̇(t)=Jv(t)+K2z(t),
where J is a q×q matrix, K2 is a q×p matrix, J is determined by the dynamic characteristics of the controller, and K2 is called the feedback matrix which will be chosen in the latter (see Theorem 3.4 and Theorem 3.8). The output signal u(t)=(u1(t),u2(t),…,uq(t))T and the input signal v(t)=(v1(t),v2(t),…,vq(t))T of the pulse-width sampler satisfy the following dynamic relation:
ui(t)={signαni,nT0≤t<(n+|αni|)T0,i=1,2,…,q;0,(n+|αni|)T0≤t<(n+1)T0,n=0,1,…,αni={vi(nT0),|vi(nT0)|≤1,i=1,2,…,q;signvi(nT0),|vi(nT0)|≥1,n=0,1,…,
where T0 is called the sampling period of the pulse-width sampler which is the same as the period of f and τk, k=1,2,….
We end this introduction by giving some definitions.
Definition 1.1.
The closed-loop system (1.1), (1.3)–(1.5) is called linear pulse-width sampler control system with impulses. The closed-loop system (1.2), (1.3)–(1.5) is called semilinear pulse-width sampler control system with impulses.
Definition 1.2.
In the closed-loop system (1.1), (1.3)–(1.5) (or system (1.2), (1.3)–(1.5)), the q dimensional vector αn=(αn1,αn2,…,αnq)T is called the duration ratio of the pulse-width sampler in the nth sampling period, n=0,1,….
We defined a closed cube Ω in ℝq as
Ω={α=(α1,α2,…,αq)T∈ℝq∣|αi|≤1,i=1,2,…,q},
then we have αn∈Ω, for n=0,1,….
Definition 1.3.
In the closed-loop system (1.1), (1.3)–(1.5) (or system (1.2), (1.3)–(1.5)), if there exists a q dimensional vector
α=(αn1,αn2,…,αnq)T∈Ω,
and a corresponding periodicity rectangular-wave control signal u(t)=u(t,α) defined by
ui(t)=ui(t,α)={signαi,nT0≤t<(n+|αni|)T0,i=1,2,…,q;0,(n+|αni|)T0≤t<(n+1)T0,n=0,1,….
such that the closed-loop system (1.1), (1.3)–(1.5) (or system (1.2), (1.3)–(1.5)), has a corresponding T0-periodic trajectory x(·)=x(·,α):x(t+T0,α)=x(t,α), t≥0, then the control signal (1.8) is called the steady-state control with respect to the disturbance f. The T0-periodic trajectory x(·) is called steady-state corresponding to steady-state control u(·) and the constant vector α∈Ω of steady-state control (1.8) is called to be a steady-state duration ratio.
Definition 1.4.
In the closed-loop system (1.1), (1.3)–(1.5) (or system (1.2), (1.3)–(1.5)), if there exists some α∈Ω such that
limn→∞αn=α,whereαn=(αn1,αn2,…,αnq)T,α=(α1,α2,…,αq)T,
then system (1.1), (1.3)–(1.5) (or system (1.2), (1.3)–(1.5)), corresponding to the disturbance f is called to be stead-state stable.
Further, system (1.1), (1.3)–(1.5) (or system (1.2), (1.3)–(1.5)), corresponding to the perturbation f is called stead-state stabilizability if we can choose a suitable T0>0 and K2 such that system (1.1), (1.3)–(1.5) (or system (1.2), (1.3)–(1.5)), is stead-state stable.
2. Mathematical Preliminaries
Let L(X,X) denote the space of linear operators from X to X, Lb(X,X) denote the space of bounded linear operators from X to X, Lb(ℝq,X) denote the space of bounded linear operators from ℝq to X, and Lb(X,ℝp) denote the space of bounded linear operators from X to ℝp. It is obvious that Lb(X,X), Lb(ℝq,X), and Lb(X,ℝp) is the Banach space with the usual supremum norm.
Define D̃={τ1,…,τδ}⊂[0,T0], where 0<τ1<τ2<⋯<τδ<T0. We introduce PC([0,T0];X)≡{x:[0,T0]→X|x is continuous at t∈[0,T0]∖D̃, x is continuous from left and has right hand limits at t∈D̃}, and PC1([0,T0];X)≡{x∈PC([0,T0];X)∣ẋ∈PC([0,T0];X)}. Set
∥x∥PC=max{supt∈[0,T0]∥x(t+0)∥,supt∈[0,T0]∥x(t-0)∥},∥x∥PC1=∥x∥PC+∥ẋ∥PC.
It can be seen that endowed with the norm ∥·∥PC(∥·∥PC1), PC([0,T0];X)(PC1([0,T0];X)) is a Banach space.
We introduce the following assumption [H1].
A is the infinitesimal generator of a C0-semigroup {T(t),t≥0} on X with domain D(A).
There exists δ such that τk+δ=τk+T0.
For each k∈ℤ0+, Bk∈Lb(X,X) and Bk+δ=Bk.
We first recall the homogeneous linear impulsive periodic system
x.(t)=Ax(t),t≠τk,Δx(t)=Bkx(t),t=τk,
and the associated Cauchy problem
x.(t)=Ax(t),t∈[0,T0]∖D̃,Δx(τk)=Bkx(τk),k=1,2,…,δ,x(0)=x̅.
If x̅∈D(A) and D(A) is an invariant subspace of Bk, using [18, Theorem 5.2.2, page 144], step by step, one can verify that the Cauchy problem (2.3) has a unique classical solution x∈PC1([0,T0];X) represented by x(t)=S(t,0)x̅, where
S(·,·):Δ={(t,θ)∈[0,T0]×[0,T0]∣0≤θ≤t≤T0}→Lb(X,X)
given by
S(t,θ)={T(t-θ),τk-1≤θ≤t≤τk,T(t-τk+)(I+Bk)T(τk-θ),τk-1≤θ<τk<t≤τk+1,T(t-τk+)[∐θ<τj<t(I+Bj)T(τj-τj-1+)](I+Bi)T(τi-θ),τi-1≤θ<τi≤⋯<τk<t≤τk+1.
The operator {S(t,θ),(t,θ)∈Δ} is called impulsive evolution operator associated with {Bk;τk}k=1∞.
The properties of the impulsive evolution operator, {S(t,θ),(t,θ)∈Δ} associated with {T(t),t≥0} and {Bk;τk}k=1∞, are collected here.
Lemma 2.1 (see [26, Lemma 2.1] [27]).
Let assumption [H1] hold. The impulsive evolution operator {S(t,θ),(t,θ)∈Δ} has the following properties.
For 0≤θ≤t≤T0, S(t,θ)∈Lb(X,X), there exists a MT0>0 such that sup0≤θ≤t≤T0∥S(t,θ)∥≤MT0.
For 0≤θ<r<t≤T0, r≠τk, S(t,θ) = S(t,r)S(r,θ).
For 0≤θ≤t≤T0, n∈Z+, S(t+nT0,θ+nT0)=S(t,θ).
For 0≤θ≤t≤T0, n∈Z+, S(t+nT0,0)=S(t,0)[S(T0,0)]n.
For 0≤θ<t, there exists an M≥1, ω∈ℝ such that
∥S(t,θ)∥≤Mexp{ω(t-θ)+∑θ≤τn<tln(M∥I+Bn∥)}.
The exponential stability of the impulsive evolution operator {S(t,θ),t≥θ≥0} will be used throughout the paper; we recall them as the following definitions and lemmas.
Definition 2.2.
{S(t,θ), t≥θ≥0} is called exponentially stable if there exist K≥0 and ν>0 such that
∥S(t,θ)∥≤Ke-ν(t-θ),t>θ≥0.
Assumption [H2]: {T(t),t≥0} is exponentially stable, that is, there exist K0>0 and ν0>0 such that
∥T(t)∥≤K0e-ν0t,t>0.
Two important criteria for exponential stability of a C0-semigroup are collected here.
Lemma 2.3 (see [26, Lemma 2.4]).
Assumptions [H1] and [H2] hold. There exists 0<λ<ν0 such that
∏k=1δ(K0∥I+Bk∥)e-λT0<1.
Then {S(t,θ),t≥θ≥0} is exponentially stable.
Lemma 2.4 (see [26, Lemma 2.5]).
Assume that assumption [H1] holds. Suppose
0<μ1=infk=1,2,…,δ(τk-τk-1)≤supk=1,2,…,δ(τk-τk-1)=μ2<∞.
If there exists α>0 such that
ω+1μln(M∥I+Bk∥)≤-γ<0,k=1,2,…,δ,
where
μ={μ1,γ+ω<0,μ2,γ+ω≥0,
then {S(t,θ),t≥θ≥0} is exponentially stable.
Remark 2.5 (see [26, Theorem 3.2]).
If {S(t,θ),t≥θ≥0} is exponentially stable, then [I-S(T0,0)] is inverse and [I-S(T0,0)]-1∈Lb(X,X).
3. Steady-State Control
In this section, we study the steady-state control of pulse-width sampler control system with impulses. First we introduce the following assumptions.
f(t), t≥0, is T0-periodic step perturbation.
Control signal u(t) is T0-periodic, which is defined by the rectangular wave signal u(t,α), α∈Ω given by (1.8).
Similar to the proof of Theorem 3.2 [26], one can obtain the following results immediately.
Lemma 3.1.
Assumptions [H1], [H3], and [H4] hold. Suppose {S(t,θ),t≥θ≥0} is exponentially stable; for every u(t,α), system (1.1) has a unique T0-periodic PC-mild solution
x(t,α)=S(t,0)x0+∫0tS(t,θ)(f(θ)+Cu(θ,α))dθ+∑0≤τk<tS(t,τk+)ck,
where
x0=[I-S(T0,0)]-1∫0T0S(T0,θ)(f(θ)+Cu(θ,α))dθ,[I-S(T0,0)]-1∈Lb(X,X),
which is globally asymptotically stable.
By Lemma 3.1, we have the following results.
Theorem 3.2.
Under the assumptions of Lemma 3.1, if the sampler periodic T0 has the following properties:
iωn∈ρ(J),ωn=2nπT0,n=0,±1,±2,…,
where ρ(J) is the resolvent set of the matrix J, i satisfies i2=-1, then the following open-loop control system
ẋ(t,α)=Ax(t,α)+f(t)+Cu(t,α),t≠τk,Δx(t,α)=Bkx(t,α)+ck,t=τk,z(t)=K1x(t),v̇(t,α)=Jv(t,α)+K2z(t,α)
has a unique T0-periodic PC-mild solution v(t,α) given by
v(t,α)=eJt[(I-eJT0)-1∫0T0eJ(T0-s)K2z(s,α)ds]+∫0teJ(t-s)K2z(s,α)ds,
Proof.
By (3.3), we know that eiωnT0=ei2nπ=1, that is 1∈ρ(eJT0). Thus (I-eJT0)-1 exists and is bounded. It is not difficult to see that
v(t,α)=eJtv0+∫0teJ(t-s)K2z(s,α)ds,
where v0=v(0,α).
Consider
y=(I-eJT0)-1∫0T0eJ(T0-s)K2z(s,α)ds,
which is the unique solution of the following equation:
y=eJty+∫0teJ(t-s)K2z(s,α)ds.
Let
v0=y=(I-eJT0)-1∫0T0eJ(T0-s)K2z(s,α)ds;
it comes from Lemma 3.1 that
z(t+T0,α)=z(t,α),t≥0.
It is easy to verify that
v(t,α)=eJt[(I-eJT0)-1∫0T0eJ(T0-s)K2z(s,α)ds]+∫0teJ(t-s)K2z(s,α)ds
is just the T0-periodic PC-mild solution v(t,α) of open-loop control system (3.4).
In order to discuss the existence of steady-state control of system (1.1), we define a map G:Ω∈ℝq→ℝq given by
G(α)=(I-eJT0)-1∫0T0eJ(T0-s)K2K1x(s,α)ds,α∈Ω,
where x(·,α) is the T0-periodic PC-mild solution of system (1.1) corresponding to α∈Ω. Then we have the following result.
Lemma 3.3.
Under the assumptions of Theorem 3.2, there exists a constant M¯>0 such that
∥G(α)-G(α̅)∥≤M¯∥K2∥∥α-α̅∥,α,α̅∈Ω.
Proof.
Suppose x1(t,α) and x2(t,α̅) are the T0-periodic PC-mild solution of system (1.1) corresponding to α and α̅∈Ω, respectively, then
x1(0)-x2(0)=x1(T0)-x2(T0)=S(T0,0)(x1(0)-x2(0))+∫0T0S(T0,θ)C(u(θ,α)-u(θ,α̅))dθ.
Thus,
∥x1(0)-x2(0)∥≤∥[I-S(T0,0)]-1∥∥S(T0,θ)∥∥C∥Lb(ℝq,X)∫0T0∥u(θ,α)-u(θ,α̅)∥ℝqdθ.
For 0≤θ≤t≤T0, we obtain
∥x1(t,α)-x2(t,α̅)∥≤∥S(t,0)∥∥x1(0)-x2(0)∥+∥S(T0,θ)∥∥C∥Lb(ℝq,X)∫0T0∥u(θ,α)-u(θ,α̅)∥ℝqdθ≤K∥C∥Lb(ℝq,X)(∥Q∥K+1)∫0T0∥u(θ,α)-u(θ,α̅)∥ℝqdθ≤M¯1∫0T0∥u(θ,α)-u(θ,α̅)∥ℝqdθ,
where
M¯1=K∥C∥Lb(ℝq,X)(∥Q∥K+1),Q=[I-S(T0,0)]-1.
By elementaly computation,
∥G(α)-G(α̅)∥≤∥(I-eJT0)-1∥∥eJT0∥∥K2∥∥K1∥Lb(X,ℝp)∫0T0∥x1(s,α)-x2(s,α̅)∥ds≤∥(I-eJT0)-1∥∥eJT0∥∥K2∥∥K1∥Lb(X,ℝp)M¯1T0∫0T0∥u(θ,α)-u(θ,α̅)∥ℝqdθ≤M¯2∥K2∥∫0T0∥u(θ,α)-u(θ,α̅)∥ℝqdθ,
where
M¯2=∥(I-eJT0)-1∥∥eJT0∥∥K1∥Lb(X,ℝp)M¯1T0.
For αlα̅l>0. Without loss of generality, we suppose that 0<αl<α̅l, then we have
∫0T0∥u(θ,α)-u(θ,α̅)∥ℝqdθ≤∫αlT0α̅lT0∥u(θ,α)-u(θ,α̅)∥ℝqdθ≤T0∥α-α̅∥.
For αlα̅l<0. For example, αl<0<α̅l, |α̅l|>αl, we have
∫0T0∥u(θ,α)-u(θ,α̅)∥ℝqdθ≤∫αlT0|α̅l|T0∥u(θ,α)-u(θ,α̅)∥ℝqdθ≤2T0∥α-α̅∥.
By (3.18), (3.20) and (3.21), there exists a constant M¯>0 such that
∥G(α)-G(α̅)∥≤M¯∥K2∥∥α-α̅∥,α,α̅∈Ω.
By Lemma 3.3, we have the following result immediately.
Theorem 3.4.
Under the assumptions Theorem 3.2, one can choose a suitable ∥K2∥ such that the systems (1.1), (1.3)–(1.5) have a unique steady-state and the fixed point of G is just the conducting vector.
Proof.
Let x(t,α) be the T0-periodic PC-mild solution of system (1.1) corresponding to α∈Ω, then
x(0)=x(T0)=S(T0,0)x(0)+∫0T0S(T0,θ)(f(θ)+Cu(θ,α))dθ,
that is,
x(0)=[I-S(T0,0)]-1∫0T0S(T0,θ)(f(θ)+Cu(θ,α))dθ.
By virtue of [H3], we can suppose that ∥f(t)∥≤f0, t≥0, then
∥x(0)∥≤∥[I-S(T0,0)]-1∥∥S(T0,θ)∥∫0T0(∥C∥Lb(ℝq,X)q+f0)dθ≤K∥Q∥(∥C∥Lb(ℝq,X)q+f0)T0≡M¯3.
It comes from
G(α)=(I-eJT0)-1∫0T0eJ(T0-s)K2K1S(t,0)x(0)ds+(I-eJT0)-1∫0T0eJ(T0-s)K2K1(∫0tS(t,s)(f(s)+Cu(s,α))ds)ds
that
∥G(α)∥≤M¯4∥K2∥,
where
M¯4=∥(I-eJT0)-1∥∥eJT0∥∥K1∥Lb(X,ℝp)T0M¯3(K+1∥Q∥).
Using Lemma 3.3 and (3.27), it is not difficult to verify that G:Ω→Ω is a contraction map when
0<∥K2∥<1max(M¯,M¯4).
By the application of contraction mapping principle, G has a unique fixed point α*∈Ω. Obviously, the T0-periodic PC-mild solution of system (1.1) corresponding to α* is just the unique steady-state.
Next, we investigate the steady-state control of system (1.2), (1.3)–(1.5). We need to introduce the following assumption [H5].
f:[0,∞)×X→X is measurable for t≥0, and for any x, y∈X, there exists a positive constant Lu>0 such that
∥f(t,x)-f(t,y)∥≤Lu∥x-y∥.
f(t,x) is T0-periodic in t. That is, f(t+T0,x)=f(t,x), t≥0.
Lemma 3.5.
Under the assumptions [H1], [H4] and [H5], the impulsive evolution operator {S(t,θ),t≥θ≥0} is exponentially stable, that is, there exists a constant K>0 and ν>0 such that
∥S(t,θ)∥≤Ke-ν(t-θ),t>θ≥0,
where ν>(LuKT0+lnK)/T0, then system (1.2) has a unique T0-periodic PC-mild solution x(·,α) corresponding to control u(·,α) given by
x(t,α)=S(t,0)x0+∫0tS(t,θ)(f(θ,x(θ))+Cu(θ,α))dθ+∑0≤τk<tS(t,τk+)ck
and is also exponentially stable.
Proof.
Suppose that x1(t,α) (x2(t,α̅)) is the PC-mild solution of system (1.2) corresponding to initial value x1=x1(0) (x2=x2(0)), respectively, then
∥x1(t)-x2(t)∥≤Ke-vt∥x1-x2∥+LuK∫0te-v(t-θ)∥x1(θ)-x2(θ)∥dθ≤Ke-vt∥x1-x2∥+LuK∫0t∥x1(θ)-x2(θ)∥dθ.
By Gronwall inequality, we can deduce
∥x1(t)-x2(t)∥≤Ke(LuK-ν)t∥x1-x2∥,t∈[0,T0].
Define a map H:X→X given by
H(t)x̅=x(t)=S(t,0)x̅+∫0tS(t,θ)(f(θ,x(θ))+Cu(θ,α))dθ+∑0≤τk<tS(t,τk+)ck.
Then we can verify that
∥H(T0)x1-H(T0)x2∥≤Ke(LuK-ν)T0∥x1-x2∥.
It comes from
ν>LuKT0+lnKT0
that H(T0) is a contraction map on X. Thus, by the application of contraction mapping principle again, H(T0) has a unique fixed point x*∈X satisfying
H(T0)x*=x*.
Using (2), (3), (4) of Lemma 2.1, one can verify that
H(nT0)=[H(T0)]n,n∈ℕ.
By virtue of (3.34), (3.38), and (3.39), we know that x(·)=x(·,x*) is just the T0-periodic PC-mild solution of (1.2) which is exponentially stable.
Similar to the proof of Lemma 3.1, using Lemma 3.5, we can obtain the following result immediately.
Lemma 3.6.
Under the assumptions of Lemma 3.5, if T0 also satisfies (3.3), then the open-loop control system
ẋ(t,α)=Ax(t,α)+f(t,x(t,α))+Cu(t,α),t≠τk,Δx(t,α)=Bkx(t,α)+ck,t=τk,z(t,α)=K1x(t,α),v̇(t,α)=Jv(t,α)+K2z(t,α).
has a unique T0-periodic PC-mild solution v(·,α).
In order to discuss the existence of steady-state control of system (1.2), we define a map G:Ω∈ℝq→ℝq given by
G̃(α)=(I-eJT0)-1∫0T0eJ(T0-s)K2K1x(s,α)ds,α∈Ω,
where x(·,α) is the periodic solution of system (1.2) corresponding to α∈Ω. Then we have the following results.
Lemma 3.7.
Under the assumptions of Lemma 3.6, there exists a constant M̂>0 such that
∥G̃(α)-G̃(α̅)∥≤M̂∥K2∥∥α-α̅∥,α,α̅∈Ω.
Proof.
Suppose that x1(t,α) and x2(t,α̅) are the T0-periodic PC-mild solutions corresponding to α and α̅∈Ω with the initial value x1(0) and x2(0), respectively, then
x1(0)-x2(0)=x1(T0)-x2(T0)=S(T0,0)(x1(0)-x2(0))+∫0T0S(T0,θ)(f(θ,x1(θ))-f(θ,x2(θ)))dθ+∫0T0S(T0,θ)C(u(θ,α)-u(θ,α̅))dθ.
Thus,
x1(0)-x2(0)=[I-S(T0,0)]-1[∫0T0S(T0,θ)(f(θ,x1(θ))-f(θ,x2(θ)))dθ+∫0T0S(T0,θ)C(u(θ,α)-u(θ,α̅))dθ].
Furthermore,
∥x1(0)-x2(0)∥≤∥Q∥MT0[Lu∫0T0∥x1(θ)-x2(θ)∥dθ+∥C∥Lb(Rq,X)∫0T0∥u(θ,α)-u(θ,α̅)∥ℝqdθ].
For 0≤θ≤t≤T0, we have
∥x1(t,α)-x2(t,α̅)∥≤∥Q∥MT02Lu∫0T0∥x1(θ,α)-x2(θ,α̅)∥dθ+MT0∥C∥Lb(ℝq,X)(∥Q∥MT0+1)∫0T0∥u(θ,α)-u(θ,α̅)∥ℝqdθ+MT0Lu∫0t∥x1(θ,α)-x2(θ,α̅)∥dθ.
By Gronwall inequality again, we obtain
∥x1(t,α)-x2(t,α̅)∥≤eMT0LuT0∥Q∥MT02Lu∫0T0∥x1(θ,α)-x2(θ,α̅)∥dθ+eMT0LuT0MT0∥C∥Lb(ℝq,X)(∥Q∥MT0+1)∫0T0∥u(θ,α)-u(θ,α̅)∥ℝqdθ.
Integrating from 0 to T0, we obtain
∫0T0∥x1(t,α)-x2(t,α̅)∥dt≤M6M5∫0T0∥u(θ,α)-u(θ,α̅)∥ℝqdθ,
where
M5=1-eMT0LuT0∥Q∥MT02Lu>0,M6=eMT0LuT0T0MT0∥C∥Lb(ℝq,X)(∥Q∥MT0+1).
Thus,
∥G̃(α)-G̃(α̅)∥≤∥(I-eJT0)-1∥∥eJT0∥∥K2∥∥K1∥Lb(X,ℝp)∫0T0∥x1(s,α)-x2(s,α̅)∥ds≤∥(I-eJT0)-1∥∥eJT0∥∥K2∥∥K1∥Lb(X,ℝp)M6M5∫0T0∥u(θ,α)-u(θ,α̅)∥ℝqdθ≤2∥(I-eJT0)-1∥∥eJT0∥∥K2∥∥K1∥Lb(X,ℝp)M6M5T0∥α-α̅∥.
Choosing a constant
M̂=2∥(I-eJT0)-1∥∥eJT0∥∥K1∥Lb(X,ℝp)M6M5T0>0,
then,
∥G̃(α)-G̃(α̅)∥≤M̂∥K2∥∥α-α̅∥,α,α̅∈Ω.
Using Lemma 3.7, we have the following result.
Theorem 3.8.
Under the assumptions of Lemma 3.7, there exists a constant Nf>0 such that ∥f(t,x)∥≤Nf, if ∥K2∥ is sufficiently small, then system (1.2), (1.3)–(1.5) has a unique steady-state and the fixed point of G̃ is just the conducting vector.
Proof.
Let x(t,α) be the T0-periodic PC-mild solution of system (1.2) corresponding to α∈Ω, then
x(0)=[I-S(T0,0)]-1∫0T0S(T0,θ)(f(θ,x(θ))+Cu(θ,α))dθ.
Further,
∥x(0)∥≤K∥Q∥(∥C∥Lb(ℝq,X)q+Nf)T0=M7.
Let
M8=∥(I-eJT0)-1∥∥eJT0∥∥K1∥Lb(X,ℝp)KT0[M7+(Nf+∥C∥Lb(ℝq,X)q)T0].
It comes from
G(α)=(I-eJT0)-1∫0T0eJ(T0-s)K2K1S(t,0)x(0)ds+(I-eJT0)-1∫0T0eJ(T0-s)K2K1(∫0tS(t,s)(f(s,x(s,α))+Cu(s,α))ds)ds
that
∥G(α)∥≤M8∥K2∥.
It is not difficult to see that G:Ω→Ω is a contraction map when
0<∥K2∥<1max(M̂,M8).
By application of contraction mapping principle again, G̃ has a unique fixed point α̃*∈Ω. Obviously, the T0-periodic PC-mild solution of system (1.2) corresponding to α̃* is just the unique steady-state.
Finally, an example is given for demonstration. Consider the following system
∂∂tx(t,y)=∂2x(t,y)∂2y+bu(t)+f(y)·1(t),y∈(0,l),2π>t>0,t≠π2,π,3π2,Δx(τi,y)=x(τi+,y)-x(τi-,y)=bkx(τi,y),y∈(0,l),t=π2,π,3π2,x(t,0)=x(t,l)=0,t≥0,z(t)=∫0lk1x(t,y)dy,
and the output v(t) satisfies
dv(t)dt=v(t)+k2z(t),
where b, bk, k1 and k2 are constants.
Then A can generate an exponentially stable C0-semigroup {T(t),t≥0} in L2(0,l) and ∥T(t)∥≤e-(π/l)t, t≥0. We only choose a suitable positive number k2, then all the assumptions are met in Theorem 3.4, our results can be used to system (3.59).
Acknowledgment
The authors acknowledge the support from the National Natural Science Foundation of China (no.10961009), Introducing Talents Foundation for the Doctor of Guizhou University (2009, no.031). Youth Teachers Natural Science Foundation of Guizhou University (2009, no.083).
TemanR.1988New York, NY, USASpringerBernsteinD. S.HollotC. V.Robust stability for sampled-data control systems1989133217226MR101723510.1016/0167-6911(89)90067-4ZBL0684.93052ZhouH. X.XuQ. L.A PMW temperature control system with a distributed parameter1998325364LiQ. G.ZhaoY.Steady-state control of a class of parabolic pulse-width sampler systems2002194516520MR1923132ZBL1109.49311HouL.MichelA. N.Stability and analysis of pulse-width-modulated feedback systems200137913351349MR211033710.1016/S0005-1098(01)00100-5ZBL0995.93058Al-SunniF. M.Al-AmerA. H.Robust control of sampled data systems199814522362402-s2.0-002473632310.1049/ip-cta:19981848HuL. S.ShiP.HuangB.Stochastic stability and robust control for sampled-data systems with Markovian jump parameters2006313250451710.1016/j.jmaa.2005.08.019MR2182514IlchmannA.TownleyS.Adaptive sampling control of high-gainstabilizable systems1999441019611966MR171616710.1109/9.793786ZBL0956.93056ÖzdemirN.2000Exeter, UKUniversity of ExeterÖzdemirN.TownleyS.Integral control by variable sampling based on steady-state data2003391135140MR213656110.1016/S0005-1098(02)00182-6ZBL1016.93044KeZ.LogemannH.TownleyS.Adaptive sampled-data integral control of stable infinite-dimensional linear systems2009584233240MR251063610.1016/j.sysconle.2008.10.015ZBL1159.93336LakshmikanthamV.BaĭnovD. D.SimeonovP. S.19896Teaneck, NJ, USAWorld Scientific Publishingxii+273Series in Modern Applied MathematicsMR1082551AhmedN. U.Optimal impulse control for impulsive systems in Banach spaces2000113752MR1734517ZBL0959.49023AhmedN. U.Some remarks on the dynamics of impulsive systems in Banach spaces200182261274MR1824808ZBL0995.34050AhmedN. U.TeoK. L.HouS. H.Nonlinear impulsive systems on infinite dimensional spaces2003545907925MR199251110.1016/S0362-546X(03)00117-2ZBL1030.34056AhmedN. U.Existence of optimal controls for a general class of impulsive systems on Banach spaces200342266968510.1137/S0363012901391299MR1982287ZBL1037.49036AhmedN. U.Optimal feedback control for impulsive systems on the space of finitely additive measures2007703-4371393MR2310657ZBL1164.34026LiuJ. H.Nonlinear impulsive evolution equations1999617785MR1679758ZBL0932.34067NietoJ. J.Rodriguez-LopezR.Boundary value problems for a class of impulsive functional equations2008551227152731MR240142410.1016/j.camwa.2007.10.019ZBL1142.34362NietoJ. J.O'ReganD.Variational approach to impulsive differential equations200910268069010.1016/j.nonrwa.2007.10.022MR2474254ZBL1167.34318WeiW.XiangX.PengY.Nonlinear impulsive integro-differential equations of mixed type and optimal controls2006551-2141156MR222172910.1080/02331930500530401ZBL1101.45002XiangX.WeiW.Mild solution for a class of nonlinear impulsive evolution inclusions on Banach space2006302367376MR2245948ZBL1121.34063XiangX.WeiW.JiangY.Strongly nonlinear impulsive system and necessary conditions of optimality2005126811824MR2178682ZBL1081.49027YuX.XiangX.WeiW.Solution bundle for a class of impulsive differential inclusions on Banach spaces20073271220232MR227740610.1016/j.jmaa.2006.03.075ZBL1113.34046WangJ. R.XiangX.WeiW.Optimal periodic controls for linear impulsive periodic system4Proceedings of International Conference on Computer Science and Software Engineering (CSSE '08)December 2008Wuhan, China1381412-s2.0-33646451517WangJ. R.XiangX.WeiW.Global behaviors and optimal harvesting of a class of impulsive periodic logistic single-species system with continuous periodic control strategy20082008211923532-s2.0-5084911342510.1155/2008/192353ZBL1157.34349WangJ. R.XiangX.PengY.Periodic solutions of semilinear impulsive periodic system on Banach space20097112e1344e13532-s2.0-3884913059710.1016/j.na.2009.01.139