 Research
 Open access
 Published:
Stability and solvability for a class of optimal control problems described by noninstantaneous impulsive differential equations
Advances in Difference Equations volume 2020, Article number: 524 (2020)
Abstract
In this paper, we investigate the existence and stability of solutions for a class of optimal control problems with 1mean equicontinuous controls, and the corresponding state equation is described by noninstantaneous impulsive differential equations. The existence theorem is obtained by the method of minimizing sequence, and the stability results are established by using the related conclusions of setvalued mappings in a suitable metric space. An example with the measurable admissible control set, in which the controls are not continuous, is given in the end.
1 Introduction
Impulsive phenomena are results of a sudden change in the state of system due to external interference; they often occur in nature and human activities. According to the duration of the change, the phenomena of this kind are divided into two categories. One is that the duration of this change is relatively short compared with the total duration of the whole process, which is called instantaneous impulse. The other is that they start from any fixed point and remain active for a limited time interval, namely the effects are continuous. We call it noninstantaneous impulse (see [1–8]).
Most of the mathematical models extracted from impulsive phenomena are characterized by impulsive differential equations, which can be classified into two categories in accordance with the type of impulse: instantaneous impulsive differential equations and noninstantaneous ones. Now, a large number of references deal with the impulsive differential equations. By the type of impulse, they include the noninstantaneous case [1–12] and the instantaneous case [13–22].
This paper is devoted to the study of the differential equation with impulse of noninstantaneous type on account of its reality and significance. For instance, the state change process of some elements during intravenous drug injection, periodic fishing, and criterion for pest management is depicted by noninstantaneous impulsive differential equations (see [20]). It is shown that the investigation of noninstantaneous impulsive differential equations is of great importance to nature and human beings themselves.
In the following, we will briefly sketch some existing results about the differential equations with noninstantaneous impulse.
In 2013, Hernández and O’Regan, depending on the background of pharmacokinetics, considered a Cauchy problem for a class of new semilinear evolution equations with noninstantaneous impulse moments (see [1]), which is
where \(t_{i}< s_{i}\), \(i=0, 1,\ldots,m\), \(m\in\mathbb{N}\), \(s_{0}=0\), \(t_{m+1}=T\), \(A:D(A)\subseteq X \rightarrow X\) is an infinitesimal generator of \(C_{0}\) semigroup \(\{T(t),t\geq0\}\) in Banach space X, \(f:[0,T]\times X \rightarrow X\), and \(g_{i}:(t_{i},s_{i}]\times X \rightarrow X\), \(i=1,\ldots,m\).
This paper has initiated the study for differential equations with noninstantaneous impulse. Since then, relevant research work has just been getting started.
Pierri and O’Regan dealt with the existence of solutions for a kind of semilinear abstract differential equations with noninstantaneous impulse in 2013 (see [9]). Then, Fečkan, Wang, and Zhou paid attention to the existence of periodic solutions for the following noninstantaneous impulsive periodic systems [10, 12]:
and
In 2015, Hernandez et al. introduced and studied a new model of abstract impulsive differential equations [4]
where A is the generator of a \(C_{0}\)semigroup of bounded linear operators \((T(t))_{t\geq0}\) defined on a Banach space \((X,\\cdot\)\), \(x_{0}\in X\), \(I_{i}(t)\) is a \(2^{[0,t]}\)set valued function.
Hernandez et al. improved substantially the theory on differential equations with noninstantaneous impulses. And since then, based on their work, many scholars have been devoted to the study of abstract differential equations with noninstantaneous impulses, such as [5–7].
In 2017, Ravi Agarwal et al. considered the initial value problem with noninstantaneous impulses (see [3])
and demonstrated that equation (1.1) was equivalent to the integral one as follows:
In 2018, Yao et al. obtained the sufficient conditions for the existence and uniqueness of extremum solutions for a class of noninstantaneous impulsive boundary value problems (see (1.3)) by means of the method of upper and lower solutions along with monotone iteration technique (see [11]):
Optimal control theory originated in the late 1950s, and the maximum principle founded by the former Soviet mathematician L.C. Pontryagin marked the beginning of a new stage in its process. Kalman put forward the concept of controllability in 1963 (see [23]), which played an important role in the field of mathematical control theory. Today, many researchers are devoted to the controllability of problems, such as those defined on abstract Banach spaces [24–26] (for more details about the study on abstract differential equations, we refer the readers to a series of Hernandez’s articles, such as [27, 28]).
In recent years, with the further development of computer science and mathematics, the optimal control problems have achieved great progress, and the applications in real life are becoming more and more extensive. A number of scholars have been committed to the study on the optimal control problems (see [8, 29–36]).
In [30], Yu investigated the existence and stability of solutions for the problem
where
h, g are continuous and \(x(t)\) satisfies the following differential equation:
In [37], Deng and Wei considered the existence and stability analysis for the above nonlinear optimal control problems with 1mean equicontinuous controls. Different from [30], Deng and Wei got the existence and stability results by weakening the condition of the control. And the control may be not continuous in their results, this is more in line with the general situation.
In [31], Liu et al. studied the optimal control problem for a new class of noninstantaneous impulsive differential equations, and the controllability was proved by constructing a suitable control function, namely finding \(\overline{u}\in U_{ad}\) such that \(J(\overline {x},\overline{u})\leq J(x,u)\) for all \(u\in U_{ad}\), \(x\in PC([0,T],\mathbb{R}^{n_{1}})\), where
\(y_{d}\in PC([0,T],\mathbb{R}^{n_{1}})\) is a given piecewise continuous function and \(x(t)\) satisfies the following differential equation:
In [32], Achim Ilchmann et al. were concerned with the optimal control problem for regular linear differentialalgebraic systems. In their paper, they derived an augmented system as the key to analyzing the optimal control problem with tools well known for the optimal control of ordinary differential equations.
So far, we have found that the research findings on noninstantaneous impulsive differential equations are still few, and the studies on optimal control problems with noninstantaneous impulse are also scarce. In view of the widespread use of optimal control problem in industrial and mining enterprizes, transportation, power industry, and national economic management (see [38, 39]), inspired by [3, 9, 11, 31], we mainly concentrate on the existence and stability of optimal control problem with nonlinear noninstantaneous impulsive differential equations. The problem is as follows.
Problem (P)
Looking for \(u^{*} \in\mathcal{U}[t_{0},T]\) satisfying the equation
where
\(p>0\) is a natural number, \(\mathcal{U}[t_{0},T]=\{uu\in L^{1}([t_{0},T]; \mathbb{R}^{n}), u(t)\in U\}\), \(U\subset\mathbb{R}^{n}\), \(g:\mathbb {R}^{m}\rightarrow\mathbb{R}\), \(h:[t_{0},T]\times\mathbb{R}^{m}\times\mathbb {R}^{n}\rightarrow\mathbb{R}\), and x satisfies the equation
where \(f_{k}:(s_{k},t_{k+1}]\cap[t_{0},T]\times\mathbb{R}^{m}\times\mathbb {R}^{n}\rightarrow\mathbb{R}^{m}\), \(\phi_{k}:(t_{k},s_{k}]\cap[t_{0},T]\times\mathbb {R}^{m}\times\mathbb{R}^{m}\rightarrow\mathbb{R}^{m}\).
In the paper, the existence result is established by the method of minimizing sequence, and the stability analysis is carried out in a suitable metric space by the related conclusions of setvalued mappings.
The layout of the rest is listed as follows. In Sect. 2, we review some standard facts that are necessary for the paper, such as some important definitions and lemmas. In Sect. 3, it is shown that there exists a unique solution for noninstantaneous impulsive differential equation (1.6). Then, in Sect. 4, it is shown that the optimal control problem (P) is solvable in the defined space with the help of minimizing sequence method. Finally, in Sect. 5, we get the results of stability for the optimal control problem by applying related conclusions on setvalued mappings, and in the end, an example is given to illustrate these results.
2 Preliminaries
In order to proceed smoothly, we do some necessary preparations in this section, and the first one is to construct topological spaces.
Let \(\{t_{i}\}_{i=1}^{p+1}\) and \(\{s_{i}\}_{i=0}^{p}\) be numbers such that \(0= s_{0}< t_{i}< s_{i}<t_{i+1}\), \(i=1,2,\ldots p\), \(0=s_{0}< t_{0}< t_{1}\), \(t_{p}< T \le t_{p+1}\) and p is a natural number.
For the sake of convenience, the norms of all function spaces in the following are uniformly written as the symbol \(\\cdot\\) without confusion, and the readers can identify the meaning of the norms from the context easily.
Set
in which the norm is defined by \(\Vert x \Vert =\max_{t_{0}\le t\le T} \Vert x(t) \Vert \), and
in which the norm is defined as \(\Vert x \Vert =\max_{t_{0}\le t\le T} \Vert x(t) \Vert \), where \(x(t)=(x_{1}(t),x_{2}(t),\ldots, x_{m}(t))^{\mathrm{T}}\in PC_{m}[t_{0},T]\) and
It is easy to prove that \((PC_{m}[t_{0},T], \\cdot\)\) is a Banach space.
Suppose that \(\mathcal{U}[t_{0},T]\) meets the following condition (which is set forth in [37]):
 \((H_{u})\)::

U is compact in \(\mathbb{R}^{n}\), and \(\mathcal {U}[t_{0},T]\) is 1mean equicontinuous, that is, \(\forall\epsilon>0\), there exists a constant \(\delta(\epsilon)>0\) such that for \(h<\delta (\epsilon)\) one has
$$\sup_{u\in\mathcal{U}[t_{0},T] } \int_{t_{0}}^{T} \bigl\Vert u(t+h)u(t) \bigr\Vert \,dt< \epsilon , $$where \(u(t)=0\) for \(t\notin[t_{0},T]\).
For U is compact in \(\mathbb{R}^{n}\), U is a nonempty and closed subset of \(\mathbb{R}^{n}\). Then there exists a constant \(M>0\) such that \(\z\\leq M\), \(\forall z \in U\). According to Proposition 2.1 of [37], \(\mathcal{U}[t_{0},T]\) is bounded and closed in \(L^{1}([t_{0},T];\mathbb{R}^{n})\). Besides, in view of Lemma 3.2 of [37], \(\mathcal{U}[t_{0},T]\) is compact in \(L^{1}([t_{0},T];\mathbb{R}^{n})\).
Let
and
Consider the following conditions (\(k=0,1,\ldots,p\)).
 \((H_{w,k})\)::

\(\forall x^{1}, x^{2}\in\mathbb{R}^{m}\), \(\forall u\in B\), \(\forall t\in[s_{k},t_{k+1}]\cap[t_{0},T]\), \(f_{k}:\mathscr{B}_{k}\rightarrow \mathbb{R}^{m}\), then
$$\bigl\Vert f_{k}\bigl(t,x^{1},u\bigr) f_{k}\bigl(t,x^{2},u\bigr) \bigr\Vert \le L_{k} \bigl\Vert x^{1}x^{2} \bigr\Vert , $$and \(\sup_{(t,x,u)\in\mathscr{B}_{k}} \Vert f_{k}(t,x,u) \Vert \le C_{k}\), where \(L_{k}, C_{k}>0\) are constants.
Under the above condition, we define the metric space as follows:
with the metric \(\rho_{k}\) defined as
One can demonstrate easily that \((F_{k}, \rho_{k})\) is a complete metric space for each \(k=0,1,\ldots,p\).
Let \(\phi_{k}: ([t_{k},s_{k}]\cap[t_{0},T] )\times\mathbb{R}^{m}\times\mathbb {R}^{m} \rightarrow\mathbb{R}^{m}\), \(k=1,2,\ldots,p\), and \(\phi_{k}\) satisfies condition \((H_{\phi,k})\), \(k=1,2,\ldots,p\).
 \((H_{\phi,k})\)::

\(\phi_{k}\) is continuous, in addition, \(\forall x^{1},x^{2},y^{1},y^{2}\in\mathbb{R}^{m}\),
$$\bigl\Vert \phi_{k}\bigl(t,x^{1},y^{1} \bigr)\phi_{k}\bigl(t,x^{2},y^{2}\bigr) \bigr\Vert \leq Q_{k} \bigl\Vert x^{1}x^{2} \bigr\Vert +\overline {Q}_{k} \bigl\Vert y^{1}y^{2} \bigr\Vert , $$and \(\sup_{(t,x,y)\in([t_{k},s_{k}]\cap[t_{0},T])\times\mathbb{R}^{m}\times \mathbb{R}^{m}} \Vert \phi_{k}(t,x,y) \Vert \leq D_{k}\), where \(Q_{k}, \overline{Q}_{k}, D_{k}>0\) (\(k=1,2,\ldots,p\)) are constants and \(Q_{k}+\overline{Q}_{k}<1\) (\(k=1,2,\ldots,p\)).
Next, one necessary lemma is given.
Lemma 2.1
Assuming that conditions \((H_{u})\), \((H_{\phi,0})\), \((H_{\phi,k})\)and \((H_{w,k})\) (\(k=1,\ldots,p\)) are satisfied, then, \(\forall u\in\mathcal{U}[t_{0},T]\), \(\forall f_{k}\in F_{k}\) (\(k=0,1,\ldots,p\)), the differential equation (1.6) is equivalent to the following integralalgebraic one:
Proof
Similar to the proof in Sect. 2 of the reference [3], it is easy to obtain this result. □
What follows are the concepts and some important conclusions related to setvalued mappings. For more details, readers can refer to [40, 41].
Definition 2.1
([40])
Let U and F be metric spaces, a setvalued mapping \(I:F \rightrightarrows U\) is called

(1)
upper semicontinuous at \(f\in F\) if, for each open set \(G \subset U\) with \(G \supset I(f)\), there exists \(\delta>0\) such that \(G \supset I(f')\) for any \(f'\in F\) with \(\rho(f',f)<\delta\);

(2)
lower semicontinuous at \(f\in F\) if, for each open set \(G \subset U\) with \(G \cap I(f)\neq\emptyset\), there exists \(\delta>0\) such that \(G \cap I(f')\neq\emptyset\) for any \(f'\in F\) with \(\rho (f',f)<\delta\);

(3)
continuous at \(f\in F\) if I is both upper semicontinuous and lower semicontinuous at f.
Definition 2.2
([40])
Let U and F be metric spaces, a setvalued mapping \(I:F \rightrightarrows U\) is called an usco mapping if I is upper semicontinuous and \(I(f)\) is nonempty compact for each \(f\in F\).
Lemma 2.2
([40])
Let U and F be metric spaces, a setvalued mapping \(I:F \rightrightarrows U\)is closed if \(\operatorname{Graph}(I)\)is closed, where \(\operatorname{Graph}(I):= \{(f,u)\in F\times U:u\in I(f) \}\)is the graph of I.
Definition 2.3
([40])
Let U and F be metric spaces, \(I:F \rightrightarrows U\) is a setvalued mapping.

(1)
For each \(f\in F\), \(u\in I(f)\) is called an essential solution if \(\forall\varepsilon>0\), there exists \(\delta>0\) such that, for any \(f'\in F\) with \(\rho(f',f)<\delta\), it has \(\uu'\<\varepsilon\) for some \(u'\in I(f')\).

(2)
The optimal control problem associated with \(f\in F\) is called essential if each \(u\in I(f)\) is essential.
Lemma 2.3
([41])
Let U and F be metric spaces and \(I:F \rightrightarrows U\)be a setvalued mapping. If I is closed and U is compact, then I is upper semicontinuous.
Lemma 2.4
Let U be a metric space, F be a complete metric space, and \(I:F \rightrightarrows U\)be an usco mapping. There exists a dense residual subset E of F such that I is lower semicontinuous at each \(f\in E\).
Definition 2.4
([40])
Let \((X,d)\) be a metric space, \(\mathcal{A}\), \(\mathcal{B}\) are any two nonempty bounded sets in X. We call
the Hausdorff metric between \(\mathcal{A}\) and \(\mathcal{B}\), where
3 Existence and uniqueness of solutions for the differential equation with noninstantaneous impulse
In this section, it is demonstrated that there exists a unique solution for the noninstantaneous impulsive differential equation (1.6).
Theorem 3.1
Supposing that conditions \((H_{u})\), \((H_{w,0})\), \((H_{w,k})\), and \((H_{\phi ,k})\) (\(k=1,\ldots,p\)) are all satisfied, then for each \(u\in\mathcal {U}[t_{0},T]\), the differential equation (1.6) has a unique solution \(x\in PC_{m}[t_{0},T]\). Since the unique solution x depends on u, we denote it as \(x_{u}\). Furthermore, the map \(u\rightarrow x_{u}\)is continuous from \(L^{1}([t_{0},T];\mathbb{R}^{n})\)into \(PC_{m}[t_{0},T]\).
Proof
Similar to the proof of Theorem 3.2.1 in [44], by constructing an equivalent norm in \(PC_{m}[t_{0},T]\)
where \(\chi>0\) is a constant and satisfies that
and
we can come to the conclusion according to the contraction mapping principle in a Banach space. By the same discussion of the proof of Theorem 3.2.1 in [44], we can also get the results that the map \(u\rightarrow x_{u}\) is continuous from \(L^{1}([t_{0},T];\mathbb{R}^{n})\) into \(PC_{m}[t_{0},T]\). □
4 Solvability for the optimal control problem
In this section, we will deal with the existence of solutions for the optimal control problem (P) by the method of minimization.
Lemma 4.1
Let \(f_{k}^{q}\in F_{k}\) (\(k=0,\ldots,p\)), \(\{u^{q}\}\subset\mathcal{U}[t_{0},T]\), and \(\phi_{k}^{q}\)be functions satisfying condition \((H_{\phi,k})\) (\(k=1,\ldots,p\)). Supposed that conditions \((H_{u})\), \((H_{w,0})\), and \((H_{w,k})\)all hold, if \(f_{k}^{q} \rightarrow f_{k}\) (\(q\rightarrow+\infty\), \(k=0,\ldots,p\)), \(u^{q}\rightarrow u\) (\(q\rightarrow+\infty\)) in \(L^{1}([t_{0},T];\mathbb{R}^{n})\)and \(\phi_{k}^{q} \rightarrow\phi_{k}\) (\(q\rightarrow+\infty\), \(k=1,\ldots,p\)) in \(C (((t_{k},s_{k}]\cap [t_{0},T])\times\mathbb{R}^{m}\times\mathbb{R}^{m};\mathbb{R}^{m} )\), then \(x^{q}\rightarrow x\) (\(q\rightarrow+\infty\)) in \(PC_{m}[t_{0},T]\), where
and
Proof
By using the Gronwall inequality and Theorem 3.1, similar to the arguments of Theorem 3.2 in [37], we can get this result easily. □
The following corollary is a direct consequence of the above lemma.
Corollary 4.1
Let \(f_{k}\in F_{k}\) (\(k=0,\ldots,p\)), \(\{u^{q}\}\subset\mathcal{U}[t_{0},T]\), and \(\phi_{k}^{q}\)be functions meeting hypothesis \((H_{\phi,k})\) (\(k=1,\ldots ,p\)). Assume that conditions \((H_{u})\), \((H_{w,0})\), and \((H_{w,k})\) (\(k=1,\ldots,p\)) all hold, if \(u^{q}\rightarrow u\) (\(q\rightarrow+\infty\)) in \(L^{1}([t_{0},T];\mathbb{R}^{n})\)and \(\phi_{k}^{q} \rightarrow\phi_{k}\) (\(q\rightarrow+\infty\), \(k=1,\ldots,p\)) in \(C (((t_{k},s_{k}]\cap [t_{0},T])\times\mathbb{R}^{m}\times\mathbb{R}^{m};\mathbb{R}^{m} )\), then
Now, in order to accomplish the objective of our study, we need to do some proper restrictions on the functions g and h in equation (1.5), which are set forth below.
 \((H_{gh})\)::

\(g:\mathbb{R}^{m}\rightarrow\mathbb{R}\) and \(h:[t_{0},T]\times\mathbb{R}^{m}\times\mathbb{R}^{n}\rightarrow\mathbb{R}\) are both continuous functions.
For simplicity, in the latter part of this section, we suppose that conditions \((H_{u})\), \((H_{w,0})\), \((H_{w,k})\), \((H_{\phi,k})\) (\(k=1,\ldots ,p\)), and \((H_{gh})\) are all satisfied.
Lemma 4.2
Let \(f_{k}^{q} \in F_{k}\)and \(\{u^{q}\}\subset\mathcal{U}[t_{0},T]\). If \(f_{k}^{q} \rightarrow f_{k}\) (\(q\rightarrow+\infty\), \(k=0,1,\ldots,p\)) and \({u^{q}\rightarrow u}\) (\(q\rightarrow+\infty\)) in \(L^{1}([t_{0},T];\mathbb {R}^{n})\), then \(J_{f_{0}^{q},f_{1}^{q},\ldots,f_{p}^{q}}(u^{q}) \rightarrow J_{f_{0},f_{1},\ldots ,f_{p}}(u) q\rightarrow+\infty\).
Proof
By the definition of J, we know that
where \(x^{q}\) and x are the solutions of equation (1.6) corresponding to \(f_{k}^{q}\), \(u^{q}\) and f, u, respectively.
From the hypothesis of this lemma and Lemma 4.1, we immediately obtain that \(x^{q}\rightarrow x\).
Since g and h are continuous, it follows that
and
Next, we will show that the solution of equation (1.6) is bounded, which is independent with regard to \(f_{k}\) and u. The demonstration of the boundedness of x is presented in the following three steps.
(1) For \(t\in[t_{0},t_{1}]\cap[t_{0},T]\),
(2) For \(t\in(t_{k},s_{k}]\cap[t_{0},T]\), \(k=1,\ldots,p\), \(\x(t)\=\\phi_{k}(t,x(t),x(t_{k}0))\ \leq D_{k}\);
(3) For \(t\in(s_{k},t_{k+1}]\cap[t_{0},T]\), \(k=1,\ldots,p\),
where \(C_{0}\), \(C_{k}\), \(D_{k}\) (\(k=1,\ldots,p\)) are constants in conditions \((H_{w,0})\), \((H_{w,k})\), and \((H_{\phi,k})\) (\(k=1,\ldots,p\)).
It is followed up by the step (1) to (3) that \(\x\\leq\overline{A}<+\infty\), where \(\overline{A}=\max_{k=1,\ldots,p}\{\x_{0}\+C_{0}T, D_{k}+C_{k}T\}\).
It is noticeable that \(\u\\leq M\) by virtue of condition \((H_{u})\), which implies that \(h(t,x(t),u(t))\) is bounded for \(t\in[t_{0},T]\) in cooperation with the continuity of h. With the aid of the dominated convergence theorem, equation (4.2) leads to the fact that
Combining (4.1) with (4.3), it holds that
Then the proof is finished. □
Then, two corollaries are given in the following as a result of Lemma 4.2.
Corollary 4.2
Let \(f_{k}\in F_{k}\) (\(k=0,1,\ldots,p\)) and \(\{u^{q}\}\subset\mathcal {U}[t_{0},T]\)with \(u^{q}\rightarrow u\) (\(q\rightarrow+\infty\)), then
Corollary 4.3
Let \(f_{k}^{q}\in F_{k}\)with \(f_{k}^{q} \rightarrow f_{k}\) (\(q\rightarrow+\infty\), \(k=0,1,\ldots,p\)) and \(u\in\mathcal{U}[t_{0},T]\), then
Now, let us demonstrate one of the main results in the paper, that is, the existence of solutions for the optimal control problem (P).
Theorem 4.1
Suppose that conditions \((H_{u})\), \((H_{w,0})\), \((H_{w,k})\), \((H_{\phi ,k})\) (\(k=1,\ldots,p\)), and \((H_{gh})\)are all satisfied, problem \(\mathbf {(P)}\)has at least one solution, that is, there exists \(u^{*}\in\mathcal {U}[t_{0},T]\)satisfying the equation \(J_{f_{0},f_{1},\ldots,f_{p}}(u^{*})=\min_{u\in\mathcal {U}[t_{0},T]}J_{f_{0},f_{1},\ldots,f_{p}}(u)\).
Proof
The proof of Lemma 4.2 shows that the solution of equation (1.6) is bounded. Given the continuity of g and h, by condition \((H_{u})\), it implies that
namely \(J_{f_{0},f_{1},\ldots,f_{p}}(u)\) is bounded below.
Take a minimizing sequence \(\{u^{j}\}_{j=1}^{+\infty}\subset\mathcal {U}[t_{0},T]\) such that
In the light of the compactness of \(\mathcal{U}[t_{0},T]\), it has a convergent subsequence \(\{u^{j'}\}_{j=1}^{+\infty}\subset\{u^{j}\}_{j=1}^{+\infty}\subset\mathcal {U}[t_{0},T]\) such that
Therefore,
In accordance with Corollary 4.2, we have that
Combining (4.5) and (4.6), it leads to that
This finishes the proof. □
5 Stability of optimal control problem
In this section, the main task is to discuss the stability of optimal solutions for the optimal control problem (P). We will characterize them by the stability of the set of all solutions for problem (P), which is denoted by \(I(f_{0}, f_{1}, \ldots, f_{p})\)(see below). In other words, if for any \(\varepsilon>0\) there exists \(\delta>0\) such that \(H(I(f'_{0}, f'_{1}, \ldots, f'_{p}), I(f_{0}, f_{1}, \ldots, f_{p}))<\varepsilon\) with \(\widetilde{\rho} (f', f)<\delta\), where H is the Hausdorff metric induced by the metric defined on \(\mathcal{F}\)(see below).
Set a metric space
on which the metric is defined as
where \(f^{1}=(f^{1}_{0},\ldots,f^{1}_{p})\in\mathcal{F}\) and \(f^{2}=(f^{2}_{0},\ldots ,f^{2}_{p})\in\mathcal{F}\).
Obviously, \((\mathcal{F}, \widetilde{\rho})\) is a complete metric space.
Consider the setvalued mapping \(I:\mathcal{F}\rightrightarrows \mathcal{U}[t_{0},T]\),
Whereafter, some meaningful conclusions are approached by similar proofs in [30] in the following.
Theorem 5.1
\(I(f_{0}, f_{1}, \ldots, f_{p})\neq\emptyset\)for each \(f=(f_{0}, f_{1}, \ldots, f_{p})\in\mathcal{F}\).
Proof
The result can be obtained directly by Theorem 4.1. □
Theorem 5.2
\(I:\mathcal{F}\rightrightarrows\mathcal{U}[t_{0},T]\)is a usco mapping.
Proof
In view of Lemma 2.3 and the compactness of \(\mathcal {U}[t_{0},T]\), it is sufficient to show that \(\operatorname{Graph}(I)\) is closed, where
Let \(\{f^{q}\}\subset\mathcal{F}\) with \(f^{q}\rightarrow f=(f_{0}, f_{1}, \ldots, f_{p})\in\mathcal{F}\) and \(\{u^{q}\}\subset I(f_{0}, f_{1}, \ldots, f_{p})\) with \(u^{q}\rightarrow u^{*}\in\mathcal{U}[t_{0},T]\). In order to prove the closeness of \(\operatorname{Graph}(I)\), we shall just conclude that \(u^{*}\in I(f_{0}, f_{1}, \ldots, f_{p})\).
In fact, in consideration of \(u^{q}\in I(f^{q})\) for each \(q\in\mathbb {N}\), we have
where \(f^{q}=(f^{q}_{0},\ldots,f^{q}_{p})\in\mathcal{F}\).
Since \(f^{q}\rightarrow f\) (\(q\rightarrow+\infty\)) and \(u^{q}\rightarrow u^{*}\) (\(q\rightarrow+\infty\)), according to Lemma 4.2 and Corollary 4.3, one can get that
and
respectively. Thus, it has
which indicates that \(u^{*}\in I(f_{0}, f_{1}, \ldots, f_{p})\). Then the proof is completed. □
Theorem 5.3
\(I:\mathcal{F}\rightrightarrows U\)is lower semicontinuous at \(f=(f_{0}, f_{1}, \ldots, f_{p})\in\mathcal{F}\)if and only if problem \(\mathbf{(P)}\)associated with f is essential.
Proof
First of all, we show that the lower semicontinuity of \(I:\mathcal {F}\rightrightarrows\mathcal{U}[t_{0},T]\) at \(f=(f_{0}, f_{1}, \ldots, f_{p})\in\mathcal{F}\) results in the fact that problem (P) associated with f is essential.
Let \(u\in I(f_{0}, f_{1}, \ldots, f_{p})\), \(\forall\varepsilon>0\), then \(V(u,\varepsilon)\cap I(f_{0}, f_{1}, \ldots, f_{p})\neq\emptyset\), where \(V(u,\varepsilon)\) is the open neighborhood of u. For the setvalued mapping I is lower semicontinuous at f, there exists \(\delta>0\) such that \(V(u,\varepsilon)\cap I(f'_{0}, f'_{1}, \ldots, f'_{p})\neq\emptyset\) for any \(f'\in\mathcal{F}\) subject to \(\widetilde {\rho}(f',f)<\delta\). Take \(u'\in V(u,\varepsilon)\cap I(f'_{0}, f'_{1}, \ldots, f'_{p})\), then \(u'\in I(f'_{0}, f'_{1}, \ldots, f'_{p})\) and \(\uu'\ <\varepsilon\). Hence, u is essential, which implies that problem (P) associated with f is essential.
Conversely, if problem (P) associated with f is essential, any element of \(I(f_{0}, f_{1}, \ldots, f_{p})\) is essential in accordance with Definition 2.3. For any open set G with \(G\cap I(f_{0}, f_{1}, \ldots, f_{p})\neq\emptyset\), there exists \(\varepsilon >0\) such that the open neighborhood \(V(u,\varepsilon)\) of u satisfies \(V(u,\varepsilon)\subset G\) for each \(u\in G\cap I(f_{0}, f_{1}, \ldots, f_{p})\). Because u is an essential solution, there exists \(\delta>0\) such that, for any \(f'\in\mathcal{F}\) satisfying \(\widetilde{\rho }(f',f)<\delta\), we have that \(\uu'\<\varepsilon\) for some \(u'\in I(f'_{0}, f'_{1}, \ldots, f'_{p})\). It shows that \(u'\in(V(u,\varepsilon )\cap I(f'_{0}, f'_{1}, \ldots, f'_{p}))\). Consequently, \(G \cap I(f'_{0}, f'_{1}, \ldots, f'_{p})\neq\emptyset\) for any \(f'\in\mathcal{F}\) with \(\widetilde {\rho}(f',f)<\delta\), that is, I is lower semicontinuous at f. Then the proof is ended. □
Remark 5.1
Theorem 5.2 and Theorem 5.3 indicate that if problem (P) associated with \(f=(f_{0}, f_{1}, \ldots, f_{p})\in \mathcal{F}\) is essential, then the setvalued mapping \(I:\mathcal {F}\rightrightarrows\mathcal{U}[t_{0},T]\) is continuous at f. What is more, since \(I:\mathcal{F}\rightrightarrows\mathcal{U}[t_{0},T]\) is compactvalued, then by Theorem 17.15 of [41], the solution set \(I(f_{0}, f_{1}, \ldots, f_{p})\) is stable.
Remark 5.2
Considering Lemma 2.4, there exists a dense residual subset \(\mathcal{E}\) of \(\mathcal{F}\) such that problem (P) associated with \(f=(f_{0}, f_{1}, \ldots, f_{p})\in\mathcal{E}\) is essential, the solution set \(I(f_{0}, f_{1}, \ldots, f_{p})\) is stable for most \(f=(f_{0}, f_{1}, \ldots, f_{p})\in\mathcal{F}\). Moreover, it suggests that each optimal control problem associated with \(f=(f_{0}, f_{1}, \ldots, f_{p})\in\mathcal{F}\) can be closely approximated arbitrarily by an essential optimal control problem.
In the end, an example is presented for illustrations.
Example 5.1
Let
Obviously, \(\mathcal{U}[1,6]\) satisfies condition \((H_{u})\). In fact, \(\forall u\in\mathcal{U}[1,6]\), \(\forall\epsilon>0\), there exists \(\delta=\min\{1,\frac{\epsilon}{5}\}\) such that when \(h<\delta\),
So, it is a nonempty compact subset of \(L^{1}([1,6])\).
Now, consider the following optimal control problems: Looking for \(u^{*}\in\mathcal{U}[1,6]\) satisfying
where
and
In addition, \(x(t)\) is the solution of the noninstantaneous impulsive differential equation
where \(1=s_{0}< t_{1}=2< s_{1}=3< t_{2}=4< s_{2}=5< t_{3}=6\).
Through simple calculation, \(x(t)\) can be written as
where
Then
It shows that \(J_{f_{0},f_{1},f_{2}}(u)\) reaches the minimum when \(\alpha= 18\), i.e., \(I(f_{0},f_{1},f_{2})=\{u^{18}\}\).
Take
therefore, the solution \(x^{m}(t)\) of the above differential equation with noninstantaneous impulse corresponding to \(f^{m}=(f_{0}^{m},f_{1}^{m},f_{2}^{m})\) can be represented as
where
Then
Similarly, \(J_{f_{0},f_{1},f_{2}}(u)\) attains the minimum when \(\alpha= 18\), i.e., \(I(f_{0}^{m},f_{1}^{m},f_{2}^{m})=\{u^{18}\}\).
From the previous discussion, we can see that
and for any \(m>0\),
Consequently, \(u^{18}\) is essential. What is more, \(H(I(f_{0}^{m},f_{1}^{m},f_{2}^{m}),I(f_{0}, f_{1}, \ldots, f_{p}))=H(u^{18},u^{18})= 0\), so the optimal controller \(u^{18}\) is stable.
6 Conclusions
In this paper, we obtain some important conclusions about the existence and stability of solutions for the optimal control problem described by nonlinear noninstantaneous impulsive ordinary differential equations. It is concluded that most of the optimal control problems are stable in the complete metric space \(\mathcal{F}\), that is, the optimal solution set \(I(f_{0}, f_{1}, \ldots, f_{p})\) will not perturb largely when some disturbances occur in the function \(f=(f_{0}, f_{1}, \ldots, f_{p})\in \mathcal{F}\).
In the future, we will deal with the optimal control problem with nonlinear differential equations with noninstantaneous impulses of fractional order.
References
Hernández, E., O’Regan, D.: On a new class of abstract impulsive differential equations. Proc. Am. Math. Soc. 141, 1641–1649 (2013)
Agarwal, R., Hristova, S., O’Regan, D.: Noninstantaneous Impulses in Differential Equations. Springer, Berlin (2017)
Agarwal, R., O’Regan, D., Hristova, S.: Monotone iterative technique for the initial value problem for differential equations with noninstantaneous impulses. Appl. Math. Comput. 298, 45–56 (2017)
Hernandez, E., Pierri, M., O’Regan, D.: On abstract differential equations with non instantaneous impulses. Topol. Methods Nonlinear Anal. 46(2), 1067–1088 (2015)
Pierri, M., Henriquez, H.R., Prokopczyk, A.: Solutions for abstract differential equations with noninstantaneous impulses. Mediterr. J. Math. 13(4), 1685–1708 (2016)
Chen, P., Zhang, X., Li, Y.: Iterative method for a new class of evolution equations with noninstantaneous impulses. Taiwan. J. Math. 21(4), 913–942 (2017)
Saadati, R., Pourhadi, E., Samet, B.: On the PCmild solutions of abstract fractional evolution equations with noninstantaneous impulses via the measure of noncompactness. Bound. Value Probl. 2019, Article ID 19 (2019)
Wang, J., Feckan, M.: Time optimal control of a system governed by noninstantaneous impulsive differential equations. J. Optim. Theory Appl. 182, 573–587 (2019)
Pierri, M., O’Regan, D., Rolnik, V.: Existence of solutions for semilinear abstract differential equations with not instantaneous impulses. Appl. Math. Comput. 219, 6743–6749 (2013)
Fečkan, M., Wang, J., Zhou, Y.: Periodic solutions for nonlinear evolution equations with noninstantaneous impulses. Nonauton. Dyn. Syst. 1, 93–101 (2014)
Yao, M., Hu, J.: Boundary value problems for first order noninstantaneous impulsive differential equations. J. Shanxi Univ. (Nat. Sci. Ed.) 42, 78–82 (2019) (in Chinese)
Wang, J., Fečkan, M., Zhou, Y.: Random noninstantaneous impulsive models for studying periodic evolution processes in pharmacotherapy. In: Mathematical Modeling and Applications in Nolinear Dynamics. Nonlinear Systems and Complexity, vol. 14, pp. 87–107 (2016)
Hernandez, E., Henriquez, H.R., McKibben, M.A.: Existence results for abstract impulsive secondorder neutral functional differential equations. Nonlinear Anal., Theory Methods Appl. 70(7), 2736–2751 (2009)
Benchohra, M., Henderson, J., Ntouyas, S.K.: Impulsive Differential Equations and Inclusions. Hindawi Publishing Corporation, New York (2006)
Hernandez, E., Tanaka Aki, S.M.: Global solutions for abstract impulsive differential equations. Nonlinear Anal., Theory Methods Appl. 72(3), 1280–1290 (2010)
Hernandez, E., O’Regan, D.: Existence results for a class of abstract impulsive differential equations. Bull. Aust. Math. Soc. 87, 366–385 (2013)
Wang, J., Feckan, M., Zhou, Y.: A survey on impulsive fractional differential equations. Fract. Calc. Appl. Anal. 19, 806–831 (2016)
Hernandez, E.: Global solutions for abstract impulsive neutral differential equations. Math. Comput. Model. 53(2), 196–204 (2011)
Hernandez, E., Azevedo, K.A.G., Gadotti, M.C.: Existence and uniqueness of solution for abstract differential equations with statedependent delayed impulses. J. Fixed Point Theory Appl. 21, Article ID 36 (2019)
Tian, Y.: Existence and stability of solutions for initial value problems of a class of noninstantaneous impulsive differential equations. Guizhou University (2017) (in Chinese)
Hernandez, E., Balachandran, K., Annapoorani, N.: Existence results for a damped second order abstract functional differential equation with impulses. Math. Comput. Model. 50(11), 1583–1594 (2009)
Hernandez, E., Pierri, M., Goncalves, G.: Existence results for an impulsive abstract partial differential equation with statedependent delay. Comput. Math. Appl. 52, 411–420 (2006)
Kalman, R.E.: Mathematical description of linear dynamical systems. J. Soc. Ind. Appl. Math., A, on Control 1, 152–192 (1963)
Hernandez, E., O’Regan, D.: Controllability of Volterra–Fredholm type systems in Banach spaces. J. Franklin Inst. 346, 95–101 (2009)
Rozanova, A.V.: Controllability for a nonlinear abstract evolution equation. Math. Notes 76, 511–524 (2004)
Balachandran, K., Anandhi, E.R., Dauer, J.P.: Boundary controllability of Sobolevtype abstract nonlinear integrodifferential systems. J. Math. Anal. Appl. 277, 446–464 (2003)
Hernandez, E., O’Regan, D.: Global solutions for a new class of abstract neutral differential equations. Proc. Am. Math. Soc. 143(8), 3561–3571 (2015)
Hernandez, E., dos Santos, J.S.D., Azevedo, K.A.G.: Existence of solutions for a class of abstract differential equations with nonlocal conditions. Nonlinear Anal., Theory Methods Appl. 74(7), 2624–2634 (2011)
Liu, Z., Li, X., Sun, J.: Controllability of nonlinear fractional impulsive evolution systems. J. Integral Equ. Appl. 25, 395–406 (2013)
Yu, J., Liu, Z.: Existence and stability analysis of optimal control. Optim. Control Appl. Methods 35, 721–729 (2014)
Liu, S., Wang, J., Zhou, Y.: Optimal control of noninstantaneous impulsive differential equations. J. Franklin Inst. 354, 7668–7698 (2017)
Ilchmann, A., Leben, L., Witschel, J., Worthmann, K.: Optimal control of differentialalgebraic equations from an ordinary differential equation perspective. Optim. Control Appl. Methods 40, 351–366 (2019)
Balasubramaniam, P., Samath, J.A., Kumaresan, N.: Optimal control for nonlinear singular systems with quadratic performance using neural networks. Appl. Math. Comput. 187(2), 1535–1543 (2007)
Kumar, A.V.A., Balasubramaniam, P.: Optimal control for linear singular system using genetic programming. Appl. Math. Comput. 192(1), 78–89 (2007)
Kumaresan, N., Balasubramaniam, P.: Optimal control for stochastic nonlinear singular system using neural networks. Comput. Math. Appl. 56(9), 2145–2154 (2008)
Kumaresan, N., Balasubramaniam, P.: Optimal control for stochastic linear quadratic singular system using neural networks. J. Process Control 19(3), 482–488 (2009)
Deng, H., Wei, W.: Existence and stability analysis for nonlinear optimal control problems with 1mean equicontinuous controls. J. Ind. Manag. Optim. 11(4), 1409–1422 (2015)
Hafdallah, A., Ayadi, A.: Optimal control of electromagnetic wave displacement with an unknown velocity of propagation. Int. J. Control 92, 2693–2700 (2019)
Huang, Z., Matthew, B., James, K.: An investigation of a highspeed ground manoeuvre under optimal control. Proc. Inst. Mech. Eng. 233, 4363–4379 (2019)
Yu, J.: Limited Rationality and the Stability of the Set of Equilibrium Points in Game Theory. The Science Publishing Company, Beijing (2017) (in Chinese)
Aubin, J., Frankowska, H.: SetValued Analysis. Birhäuser, Boston (1990)
Fort, M.K.: Points of continuity of semicontinuous functions. Publ. Math. (Debr.) 2, 100–102 (1951)
Tan, K., Yu, J., Yuan, X.: The stability of Ky Fan’s points. Proc. Am. Math. Soc. 123, 1511–1519 (1995)
Bressan, A., Piccoli, B.: Introduction to the Mathematical Theory of Control. American Institute of Mathematical Sciences Press, New York (2007)
Acknowledgements
The authors are grateful to the referees for their careful reading of the manuscript. The authors thank the referees and the editors for their valuable comments. The authors are grateful to Dr. Zhanmei Lv for her guidance.
Availability of data and materials
Data sharing not applicable to this article as no data sets were generated or analysed during the current study.
Funding
This research is supported by the Fundamental Research Funds for the Central Universities (NO. 2019QNA32).
Author information
Authors and Affiliations
Contributions
All of the authors contributed equally in writing this paper. They both read and approved the final manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare that they have no competing interests.
Additional information
Abbreviations
Not applicable.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Chen, Y., Meng, K. Stability and solvability for a class of optimal control problems described by noninstantaneous impulsive differential equations. Adv Differ Equ 2020, 524 (2020). https://doi.org/10.1186/s1366202002919z
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s1366202002919z