Abstract
The existence of mild solution and the constrained local controllability of a retarded boundary control system with nonlocal delay condition have been established. The theory of extrapolation spaces is applied to derive the mild solution. Then, the constrained local controllability is established using the generalized open mapping theorem. In the last section, application of the result is shown through examples of control systems represented by hyperbolic partial differential equations.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
During last four decades, engineering mathematics has made a vast global impact on the controllability of abstract semilinear control systems in infinite dimensional spaces, for example, see articles [1–3] and references therein. The boundary control problem arises in a number of physical, chemical, and biological phenomena, some examples can be seen in [4–7]. For these reasons, study of abstract boundary control problems is one of the most exciting areas in applied mathematics. Lasiecka [8] first established the semigroup approach for abstract parabolic boundary problems and then developed theory for boundary control problems jointly with Triggiani [9, 10]. Numerical treatment for the exact and approximate controllability of infinite dimensional control problems is explicated by Glowinski et al. [11]. In 2006, Boulite et al. [12] have proved the existence of solution and the exact controllability of the following semilinear nonautonomous boundary control problem with nonlocal initial condition:
where 0 ≤ t 1 ≤ t 2 ≤ ..... ≤ t p ≤ T, \(f: [0,T] \times X \rightarrow \partial X\) and \(g: [0,T]^{p} \times C([0,T];X) \rightarrow X\).
Engel et al. [13] have developed semigroup approach for classical solutions to abstract linear time invariant Cauchy problem ((BP)0,0). They have also given the maximal reachability space for ((BP)0,0). Kumpf and Nickel [14] have presented the application of the theory of “one-sided coupled” operator matrices, developed by Engel [15], through example of one-dimensional heat equation with dynamic boundary conditions and boundary control.
Nonlocal conditions represent more practical aspect for the physical measurements as compared to the classical initial conditions. The importance of nonlocal conditions was introduced first time by Byszewski [16] and subsequently developed by many authors (few recent references are [17–20]).
In this paper, we consider the following nonautonomous semilinear boundary control system with nonlocal delay condition:
where τ represents the delay factor, x(t) ∈ X is called the state of the system, and X is the state space. The control variable u(t) ∈ U, where U is the control space. The control operators \(B(t): U \rightarrow \partial X\), for all t ∈ [0, T], are bounded linear operators such that \(B(\cdot ) \in L^{\infty }([0,T]; \mathcal {L}(U; \partial X))\) and \(f: [0,T] \times X \times X \times U \rightarrow \partial X\) is nonlinear map, where ∂ X is the boundary space. The nonlocal delay condition is given by h and ϕ. A few expressions for h are given in the last section for application of the theory.
Throughout this paper X, ∂ X, and U represent the Banach spaces with norms ||⋅||, ||⋅|| ∂ X , and ||⋅|| U , respectively. The set \(\mathcal {L}(X;U)\) stands for the Banach space of bounded linear operators from X into U. Other details and standing assumptions on the system operators are given in the next section.
2 Existence of Mild Solution
Let \(\mathcal {C}_{t}=C([-\tau ,\, t];\,X), \tau >0, 0\leq t\leq T<\infty \) be a Banach space of all continuous functions from [−τ, t] into X endowed with the norm \(||\phi ||_{\mathcal {C}_{t}}=\sup \limits _{-\tau \leq \eta \leq t}||\phi (\eta )||_{X}\). This section explains the existence and uniqueness of mild solution of the following system:
under the following assumptions:
-
(A1)
D is a dense linear subspace of X. Moreover, D is equipped with the norm |⋅| which is finer (stronger) than ||⋅|| X such that (D,|⋅|) is complete.
-
(A2)
For all t ∈ [0, T], \(A_{\max }(t): (D, |\cdot |) \rightarrow X\) and \(L(t) : (D, |\cdot |) \rightarrow \partial X\) are bounded linear operators and L(t) is surjective.
-
(A3)
The family of operators, \(A(t) := A_{\max }(t)|_{\ker L(t)}, t \in [0,T]\), generates an exponentially bounded evolution family (S(t, s))(s, t) ∈ ΔT , ΔT:={(a, b):0 ≤ a ≤ b ≤ T}.
-
(A4)
\(h: \mathcal {C}_{0} \rightarrow \mathcal {C}_{0}\) and there exists \(\chi \in \mathcal {C}_{0} \) such that h(χ) = ϕ.
-
(A5)
The map \(b : [0,T] \rightarrow [-\tau ,T]\) is nondecreasing and non-expansive such that it satisfies the delay property, i.e., b(t) ≤ t for t ∈ [0, T].
-
(A6)
The nonlinear map \(f : [0,T] \times X \times X \rightarrow \partial X \) is uniformly continuous in t and locally Lipschitz in X, i.e., for every t 0 ∈ [0, T] and constant R≥0, there is a constant \(N_{(R, t_{0})}\) such that: \(||f(t,x(t),x(b(t))) - f(t,y(t),y(b(t)))|| \leq N_{(R,t_{0})}\left [||x(t)-y(t)||+||x(b(t))-y(b(t))||\right ]\) for all (x(t), y(t)),(x(b(t)), y(b(t))) ∈ B R (X 2;(χ(0), χ(0))), where
$$B_{R}(X^{2};(\chi(0),\chi(0))) := \{(x_{1}(t),x_{2}(t))\in X^{2} : \sum\limits_{i=1}^{2} ||x_{i}(t)-\chi(0)||_{X} \leq R, t \in [-\tau, t_{0}]\} $$is a ball in X × X.
To prove the main abstract results, we take the setting of Griener [21] and others [12, 13] who have used the semigroup approach to study the well-posedness of systems when operator \(A_{\max }(t)\) is the perturbation of operator A(t) which is a generator of the evolution family and the perturbation is due to the change in its domain.
Under assumptions (A1) − (A3), the proof of the following properties can be seen in [21]. These properties play key role for the well-posedness of the system (2).
Lemma 2.1
[21] For each λ ∈ ρ(A(t)), the resolvent set of A(t), t ∈ [0, T], the following assertions hold:
-
(i)
\(D = D(A(t)) \oplus \ker (\lambda I - A_{\max }(t))\),
-
(ii)
\(L(t)|_{\ker (\lambda I - A_{\max }(t))}\) is an isomorphism from \(\ker (\lambda I - A_{\max }(t))\) onto ∂X and its inverse \(L_{\lambda ,t} := [L(t)|_{\ker (\lambda -A_{\max }(t))}]^{-1} : \partial X \rightarrow \ker (\lambda I - A_{\max }(t))\) is bounded,
-
(iii)
Q λ (t) := L λ,t L(t) is a projection in D onto \(\ker (\lambda I - A_{\max }(t))\) along D(A(t)).
From Lemma 2.1, \(\text {ran}(Q_{\lambda }(t)) = \ker (\lambda I - A_{\max }(t))\) and \(\ker (Q_{\lambda }(t)) = \text {ran}(I - Q_{\lambda }(t)) = D(A(t))\), and for each \(x(t) \in D(A_{\max }(t))\),
Therefore
These observations lead us towards the following result, the detailed proof can be written as the proof of Proposition 2.6 in [13].
Lemma 2.2
[13] Let assumptions (A1) − (A3) be satisfied and let x 0 ∈ X, λ ∈ ρ(A(t)), t ∈ [0, T]. Then, system (2) is equivalent to the following system:
i.e., x(⋅) is the solution of system 2 iff it is the solution of system 3.
In (3), (x(t) − L λ, t f(t, x(t), x(b(t)))) ∈ D(A(t)), \(L_{\lambda ,t}f(t,x(t),x(b(t))) \in \ker (\lambda I - A_{\max }(t))\), and A(t)(x(t) − L λ, t f(t, x(t), x(b(t)))) ∈ X. Using the theory of extrapolation spaces [22], (3) can be written into standard Cauchy problem in a bigger state space X −1 as follows:
where \(A_{-1}(t) : D(A_{-1}(t)) = X \rightarrow X_{-1}\) is the continuous extension of A(t). The bigger state space X −1 is the extrapolated space of X which satisfies the following:
-
(1)
X −1 is a Banach space containing X as a dense subspace.
-
(2)
A −1(t) generates an extrapolated evolution family S −1(t, s) on X −1.
Applying the variation of parameters, the solution of (4) is given by
Since S(t, s) = S −1(t, s)| X , using \(\lim _{\lambda \rightarrow +\infty } \lambda R(\lambda , A_{-1}) x = x\), we obtain
Thus Eq. 5 can be written as follows:
Definition 2.1
Let f ∈ L 1([0, T] × X × X). A function x ∈ C([−τ, T];X) is said to be a mild solution of the system (3) if it satisfies Eq. 5.
Now, we prove the main theorem of this section for the existence of mild solution. The theme of the proof is based on [17, 23].
Theorem 2.3
Let assumptions (A1)−(A6) be satisfied and ||λL λ,t || ≤ γ as \(\lambda \rightarrow \infty \), then the boundary value problem (2) has a mild solution x(⋅) on [−τ, T] for some \(0 < T < \infty \) . Moreover, the mapping ϕ ↦ x from \(\mathcal {C}_{0}\) into \(\mathcal {C}_{T}\) is Lipschitz and induces the uniqueness of the mild solution.
Proof
We show that the system (2) has a mild solution on the interval [0, T], where \(0 < T < \infty \). Following the assumption (A4), let us define \(\bar {\chi } \in \mathcal {C}_{T}\) by
From (A3), let M ≥ 1 and ω ≥ 0 be such that ||S(t, s)|| X ≤ M e ω(t − s). Let R > 0 be fixed such that \(\sup _{t \in [0,T]} ||(S(t,0)-I)\chi (0)||_{X} \leq R/2\), and the following hold:
and
where \(N_{f} = \sup \{||f(t, \chi (0), \chi (b(0)))||_{\partial X} : t \in [0,T]\}\) and \(R^{\prime } = 2R+2||\bar {\chi }||_{T}\). Define a map \(F : \mathcal {C}_{T} \rightarrow \mathcal {C}_{T}\) by
From assumption (A4), we know that there exists \(\chi \in \mathcal {C}_{0}\) such that
Therefore, it is sufficient to show that F has a fixed point on [0, T]. In the following, we show that F maps \(B_{R}(\mathcal {C}_{T}; \bar {\chi })\) into itself.
Thus, \(Fx \in B_{R}(\mathcal {C}_{T};\bar {\chi })\) for t ∈ [0, T] implying F maps the ball \(B_{R}(\mathcal {C}_{T}; \bar {\chi })\) into itself. Further, F satisfies local Lipschitz condition in this ball with constant \(N = N_{(R^{\prime },T)}\). Suppose \(x,y \in B_{R}(\mathcal {C}_{T}; \bar {\chi })\), then
Since \(2\gamma Me^{\omega T}T N_{(R^{\prime },T)}<1\), by Banach contraction principle, F has a unique fixed point in \(B_{R}(\mathcal {C}_{T};\bar {\chi })\) for t ∈ [0, T]. This fixed point is the required mild solution of (2) on [0, T].
From the above discussion, we conclude that a mild solution x(⋅) of (2) on the interval [0, T] can be extended to the interval [0,2T] by defining x(t) = y(t) on [T,2T], which is as follows: if 0 < τ < T, then
and if 0 < T < τ, then T − τ < 0 and hence
Here, η(t) is the restriction of χ(t) on the subinterval [T − τ,0] and satisfies assumption (A4). In this way, the interval of solution can be extended up to \(T_{\max } < \infty \) and in such case \(\lim _{t \uparrow T_{\max }} ||x(t)|| = \infty \) for which the proof replicates the explanation of Theorem 6.1.4 in [23].
Let \(x_{1}(\cdot ), x_{2}(\cdot ) \in \mathcal {C}_{T}\) be two mild solutions corresponding to the nonlocal delay functions \(\phi _{1}, \phi _{2} \in \mathcal {C}_{0}\), respectively. From assumption (A4), we get \(\chi _{1}, \chi _{2} \in \mathcal {C}_{0}\) such that h(χ i ) = ϕ i for i = 1, 2. From the above discussion, we have
Therefore, it is sufficient to prove that the mapping χ ↦ x is Lipschitz which renders the uniqueness of mild solution.
For t ∈ [−τ,0], clearly, with the constant 1, the mapping is Lipschitz. Now, we consider the case t ∈ [0, T].
Let t 1 ∈ [0, t] be such that b(s) ≤ 0 for s ∈ [0, t 1]. Then x i (b(s)) = χ i (b(s));i = 1,2; and hence from (10), we get
which by Gronwall’s inequality implies that
which yields the Lipschitz continuity of the map χ ↦ x(⋅) and the uniqueness of the mild solution. □
3 Controllability Result
In this section, we shall discuss the constrained controllability of semilinear boundary control system. We provide sufficient conditions for the constrained exact local controllability on [0, T] assuming that control functions take values in a closed convex cone with vertex at zero. There are several research papers by Klamka [24, 25] on the constrained controllability of abstract control systems with initial condition. Azzouzi et al. [26] have investigated the constrained approximate controllability of boundary control systems for controls in a convex cone. We have extended the work by Chukwu and Lenhart [27] and Klamka [24] to boundary control systems with nonlocal delay condition.
Let \(U_{0} \subset U\) be a closed convex cone with nonempty interior and take the set of admissible controls for the semilinear boundary control systems:
Definition 3.1
The constrained attainable set at time T > 0, denoted by K T (U 0), is defined as:
where x(t, χ(0), u) is a solution of system (1).
The extrapolated semilinear boundary control system is
Following the discussion in Section 2, the mild solution of (11) can be written as
Let us consider the associated linear boundary control system:
which possesses the mild solution
For λ ∈ ρ(A(t)), we define the following operators:
-
(B)
\(\mathcal {B}_{\lambda } : U \rightarrow C([0,T]; X)\) by
$$\mathcal{B}_{\lambda} u = {\int}_{0}^{\cdot} S(\cdot, s) \lambda L_{\lambda,s} B(s) u(s) ds.$$ -
(F)
\(\mathcal {F}_{\lambda } : X \times U = Z \rightarrow C([0,T]; X)\) by
$$\mathcal{F}_{\lambda} (x,u) = {\int}_{0}^{\cdot} S(\cdot, s) \lambda L_{\lambda,s} f(s,x(s),x(b(s)),u(s)) ds.$$
Here, Z = X × U is a product space which is a Banach space under the norm
We impose the following hypotheses:
-
(F1)
the nonlinear map f satisfies the local Lipschitz condition in X × X × U:
$$||f(t,x_{1}(t),x_{1}(b(t)),u_{1}(t)) - f(t,x_{2}(t),x_{2}(b(t)),u_{2}(t))|| $$$$\leq N_{(2R,t_{0})}(||x_{1}(t) - x_{2}(t)|| + ||x_{1}(b(t)) - x_{2}(b(t))|| + ||u_{1}(t) - u_{2}(t)||), $$ -
(F2)
the nonlinear map f is Frechet differentiable in the argument spaces X and U, and satisfies f(t, x(t), x(b(t)), u(t))| u = 0 = 0,
$$D_{x} f(t, x(t), x(b(t)), u(t))|_{u=0} = 0 ~~\text{and} ~~ D_{u} f(t, x(t), x(b(t)), u)|_{u=0} = 0,$$ -
(F3)
\(\mathcal {B}_{\lambda }\) and \(\mathcal {F}_{\lambda }\) are continuously differentiable on U and their derivatives \(\{D_{u} \mathcal {B}_{\lambda }\}\) and \(\{D_{u} \mathcal {F}_{\lambda }\}\) converge uniformly on U and Z, respectively, as \(\lambda \rightarrow \infty \).
Definition 3.2
The system (1) is said to be U 0-exactly locally controllable on [0, T] if the attainable set K T (U 0) contains a neighborhood of x(0) = χ(0) ∈ X in the space X.
Definition 3.3
The system (1) is said to be U 0-exactly globally controllable on [0, T] if K T (U 0) = X.
Definition 3.4
The corresponding linear system (13) is U 0-exactly globally controllable on [0, T] if \(~ran(\mathcal {B}_{\lambda }(U_{0})(T)) = X\).
Before stating the main result, let us recall the generalized open mapping theorem.
Lemma 3.1
[28] Let X, Y be Banach spaces and \(F : B_{r}(x_{0}) \subset X \rightarrow Y\) such that
for some k > 0 and \(T \in \mathcal {L}(X; Y)\) with ran(T) = Y. Then, \(B_{\rho }(Fx_{0}) \subset FB_{r}(x_{0})\) for some ρ > 0 provided that k is sufficiently small.
Now, we shall proceed to the main result.
Theorem 3.2
Let us suppose that (A1)−(A5) and (F1)−(F3) hold and the linear boundary control system (13) is U 0 -exactly globally controllable on [0,T]. Then, the semilinear boundary control system (1) is U 0 -exactly locally controllable on [0,T].
Proof
Define the control to state operator \(\mathcal {G} : U_{ad} \rightarrow X\) by \(\mathcal {G}(u) = x(T,\chi (0),u)\). Then, from the integral equation (12), we have
Using the assumption (F3) and differentiating (14), we get
where
Let t 1 ∈ [0, T] be such that b(s) ≤ 0 for s ∈ [0, t 1]. Then,
Now, using assumption (F2) in Eq. 15, we get
Since the linear system (13) is U 0-exactly globally controllable, therefore \(D_{u} \mathcal {G}(u)|_{u=0}\), mapping v ↦ y(T, 0, v) is a surjective map with \(D_{u} \mathcal {G}(0)(U_{ad}) = X\). Further, let u 1 and u 2 ∈ U a d for which the corresponding trajectories are x 1(t) = x(t, χ(0), u 1) and x 2(t) = x(t, χ(0), u 2), respectively. Then,
which by Gronwall’s inequality, implies that
for all t ∈ [0, T]. Therefore,
and
Thus,
By Lemma 3.1, the operator \(\mathcal {G}\) transforms a neighborhood of zero in the space U a d onto a neighborhood of \(\mathcal {G}(0) = S(T,0) \chi (0)\) in the space X. Hence, by Definition 3.2, the semilinear system (1) is U 0-exactly locally controllable. □
4 Application
Example 1
Let us consider the semilinear wave equation of a semi-infinite string \((0 \leq x < \infty )\) with boundary control action:
where α, β, ξ,andμ are coefficient functions satisfying the following conditions:
-
(a)
\(\alpha \in C^{1}([0,T]; W^{1,\infty }[0,\infty )), ~ 0 < \alpha _{1} \leq \alpha (t,x) < \infty \),
-
(b)
\(\beta , \xi \in L^{\infty } ([0,T] \times \{0\}): ~ 0 < \nu _{1} \leq \beta (t,0) < \infty \) and \(0 < \nu _{2} \leq \xi (t,0) < \infty \),
-
(c)
\(\mu \in L^{\infty }([0,T]), ~ 0 < \mu _{1} \leq \mu (t) \leq \mu _{2} < \infty \).
The boundary control wave (16) can be reformulated as abstract boundary control problem with nonlocal delay condition. In this connection, let us consider the following:
-
(i)
The state space \(X = L^{1}[0,\infty )\), \(\partial X = \mathbb {R} = U\) and \(D(\mathcal {A}_{\max }(t)) = W^{2,1}[0,\infty )\).
-
(ii)
The linear operator
$$\mathcal{A}_{\max}(t) \psi = \alpha(t,\cdot) \frac{\partial^{2}}{\partial x^{2}}\psi + \left( 1 + \frac{\partial}{\partial x} \alpha(t,\cdot) \right) \frac{\partial}{\partial x}\psi.$$ -
(iii)
The boundary operator
$$\mathcal{L}(t) \psi = \alpha(t,0) \frac{\partial \psi}{\partial x}(0) + \psi(0)$$with domain
$$D(\mathcal{L}(t)) = D(\mathcal{A}_{\max}(t))$$and
$$\ker(\mathcal{L}(t)) = \{\psi \in X : \frac{\partial \psi}{\partial x}(0) = 0 \text{ and } \psi(0) = 0\}.$$ -
(iv)
The control evaluation operator
$$E(t) \psi = \frac{\partial \psi}{\partial t}(t,0),~ \text{ with } ~ \ker(E(t)) = \{\psi \in X : \frac{\partial \psi}{\partial t}(t,0) = 0\}.$$ -
(v)
The control operators
$$B_{1}(t) u(t) = \xi(t,0) u(t), ~~ B_{2}(t) u(t) = \beta(t,0) u(t).$$ -
(vi)
b(t) = t − τ for t ∈ [0, T] satisfies assumption (A6).
-
(vii)
The nonlinear map
$$\begin{array}{@{}rcl@{}} \mathcal{F}(t,\psi(t),\psi(b(t)),u(t)) &=& \mathcal{F}(t,\psi(t,\cdot),\psi(b(t),\cdot),u(t)) \\ &=& {{\int}_{0}^{1}} \mu(t) \left( \psi(t,x) - \psi(t-\tau,x)\right)u^{2}(t) dx. \end{array} $$
The construction of operators to transform into the first order abstract Cauchy system proceeds further. Let us define a Banach space \(Z = W^{1,1}[0,\infty ) \otimes L^{1}[0,\infty )\) equipped with the norm \(\left \| \left (\begin {array}{c} \psi _{1} \\ \psi _{2} \end {array}\right ) \right \|_{Z} = ||\psi _{1}||_{W^{1,1}[0,\infty )} + ||\psi _{2}||_{L^{1}[0,\infty )}\). Now,
-
(vi)
the system operator
$$A_{\max}(t) z(t) = \left( \begin{array}{cc} 0 & I \\ \mathcal{A}_{\max}(t) & 0 \end{array}\right) \left( \begin{array}{c} y(t) \\ \dot{y}(t) \end{array}\right) = \left( \begin{array}{c} \dot{y}(t) \\ \frac{\partial}{\partial x} \left( \alpha(t,\cdot) \frac{\partial y}{\partial x} + y(t,\cdot)\right) \end{array}\right)$$with domain
$$D = D(A_{\max}(t)) = W^{2,1}[0,\infty) \otimes W^{1,1}[0,\infty),$$ -
(vii)
the boundary operator
$$L(t) z(t) = \left( \begin{array}{cc} 0 & E(t) \\ \mathcal{L}(t) & 0 \end{array}\right) \left( \begin{array}{c} y(t) \\ \dot{y}(t) \end{array}\right) = \left( \begin{array}{c} \dot{y}(t,0) \\ \alpha(t,0) \frac{\partial y}{\partial x}(t,0) + y(t,0) \end{array}\right)$$with domain
$$D(L(t)) = D ~ \text{ and } ~ \ker(L(t)) = \ker(\mathcal{L}(t)) \otimes \ker(E(t)),$$ -
(viii)
the boundary control operator
$$B(t) u(t) = \left( \begin{array}{c} B_{1}(t) \\ B_{2}(t) \end{array}\right) u(t) = \left( \begin{array}{c} \xi(t,0) \\ \beta(t,0) \end{array}\right) u(t),$$ -
(ix)
the nonlinear map
$$f(t,\psi(t),\psi(b(t)),u(t)) = \left( \begin{array}{c} 0 \\ 1 \end{array}\right) \mathcal{F}(t,\psi(t,\cdot),\psi(b(t),\cdot),u(t)),$$ -
(x)
and the nonlocal delay condition is represented by
$$\hbar(z) = \phi, ~~ i.e. ~~ \left( \begin{array}{c} y(\theta,\cdot) \\ \dot{y}(\theta,\cdot) \end{array}\right) = \left( \begin{array}{c} \phi_{1}(\theta,\cdot) \\ \phi_{2}(\theta,\cdot) \end{array}\right), ~~ \theta \in [-\tau,0].$$
We know that \(A_{\max }(t) \left (\begin {array}{c}\psi _{1} \\ \psi _{2} \end {array}\right ) = \left (\begin {array}{c} \psi _{2} \\ \frac {\partial }{\partial x} \left (\alpha (t,x) \frac {\partial \psi _{1}}{\partial x} + \psi _{1} \right ) \end {array}\right ) \in Z\). Then, to verify that D is densely imbedded in Z of assumption (A1), we obtain the following inequality:
This implies that ||⋅|| D is weaker than the graph norm of the system operator \(A_{\max }(t)\). Hence, the domain D equipped with the norm ||⋅|| D is densely imbedded in the Banach space Z. Let us write \(A_{\max }(t) = \mathcal {A}_{1}(t) + \mathcal {A}_{2}(t) + \mathcal {A}_{3}\), where
and
Let us consider the operator \(A_{1} \psi = \frac {\partial ^{2} \psi }{\partial x^{2}}\) with domain
Then, from the semigroup theory, A 1 is the infinitesimal generator of a C 0− semigroup on X. Similarly, \(A_{2} : D(A_{2}) \subset X \rightarrow X\) with domain
is the infinitesimal generator of a C 0−semigroup on X. From (a), we conclude that the associated operators A 1(t) and A 2(t) are infinitesimal generators of C 0−semigroups on X for all t ∈ [0, T]. Hence, the matrix operators \(\mathcal {A}_{1}(t)\) and \(\mathcal {A}_{2}(t)\) for all t ∈ [0, T] are infinitesimal generators of C 0−semigroups on Z. Further, \(\mathcal {A}_{3}\) is bounded linear operator, therefore by Bounded Perturbation Theorem, it implies that \(A(t) = A_{\max }(t)|_{\ker L(t)}\) generates a C 0−semigroup on Z for every t ∈ [0, T]. The null space \(\ker (L(t)), ~ t \in [0,T]\) of L(t) is defined as follows
Further discussion will employ the relative boundedness of two operators. We say that an operator A 2 is A 1-bounded if there exist constants a > 0 and b > 0 such that
For \(\psi _{1} \in \ker (\lambda I - \mathcal {A}_{\max }(t))\), we have
This implies that the conditions of Lemma 2.1 are satisfied. Therefore, the evolution family S(t, s)(s, t) ∈ Δ associated to the boundary control wave Eq. (16) exists and is exponentially stable ||S(t, s)|| ≤ M e ωt. The nonlinear function \(\mathcal {F}\) satisfies assumptions (F1)−(F2) and correspondingly the matrix operator map f. The set of admissible controls is \(U_{ad} = \{u \in L^{\infty }([0,T]; \mathbb {R}) : u(t) \geq 0\}\), i.e., U 0 is the nonnegative real interval \([0,\infty )\). The U 0−exact global controllability of the linear control system associated to the system (16) is well explained in the pioneer work of Klamka [24] and Son [29]. Hence, from Theorem 3.2, the semilinear wave Eq. (16) is U 0−exactly locally controllable.
Example 2
Consider the following boundary control system:
where p(t, x) is the density of a population of size x at time t; α is the growth rate depending upon the size x and time t; μ is the aging function; β is the birth function. The control function u(t) represents the inflow of zero-size individuals from an external source.
The coefficients satisfy the following conditions:
-
(a)
\(\alpha \in C([0,T]; W^{1,\infty }[0,\infty )): ~ 0 < \alpha _{1} \leq \alpha (t,x) \leq \alpha _{2} < \infty \),
-
(b)
\(\beta \in L^{\infty } ([0,T] \times \mathbb {R}^{+}):~ 0 < \beta _{1} \leq \beta (t,x) \leq \beta _{2} < \infty \),
-
(c)
\(\mu \in L^{\infty } ([0,T] \times \mathbb {R}^{+})\): \(0 < \mu _{1} \leq \mu (t,x) \leq \mu _{2} < \infty \).
-
(d)
\(c \in L^{\infty } ([0,T]; \mathbb {R}):~ c(t) > 0\).
For −τ ≤ t 1 < ... < t n ≤ 0, the Eq. (17c) represents the nonlocal delay condition and p 0 is the initial population of individual of size x.
The population model (17a) resembles the form of abstract boundary control system if we consider the following:
-
(i)
The state space \(X = L^{1}([0,\infty ))\) and \(D = D(A_{\max }(t)) = W^{1,1}([0,\infty ))\), where the operator \(A_{\max }(t),~ t \in [0,T],\) is defined as
$$A_{\max}(t) \psi = -\alpha(t,\cdot) \frac{\partial}{\partial x}\psi - \frac{\partial}{\partial x} \alpha(t,\cdot) \psi - \mu(t,\cdot) \psi.$$ -
(ii)
The boundary \(\partial X = \mathbb {R}\) and the boundary operator L(t), t ∈ [0, T], is given by
$$L(t) \psi = \alpha(t,0) \psi(0).$$ -
(iii)
The control space \(U = \mathbb {R}\) and the control function \(u \in L^{\infty }([0,T]; U)\) and the admissible control set is \(U_{ad} = \{u \in L^{\infty }([0,T]; U) : u(t) \geq 0\}\), i.e. \(U_{0} = \mathbb {R}^{+} = [0, \infty )\). The boundary control operator B(t), t ∈ [0, T], is defined as B(t)u(t) = c(t)u(t) for all t ∈ [0, T].
-
(iv)
The delay function b(t) = t − τ, t ∈ [0, T], which satisfies assumption (A6).
-
(v)
The nonlinear map
$$f(t,\psi(t),\psi(b(t)),u(t)) = {\int}_{0}^{\infty} \beta(t,x) \left( \psi(t,x) + \psi(b(t),x)\right) u^{2}(t)dx.$$ -
(vi)
The nonlocal condition is \(h(p) = {\sum }_{i=1}^{n} k_{i} p(t_{i},\cdot ) = p_{0} = \phi \).
The verification of assumptions (A1)−(A6) are well explained in the work of Boulite et al. [12]. The nonlinear map f satisfies assumptions (F1)−(F2). The assumption (F3) follows from [30]. Further, since c(t) > 0 and u(t) ∈ U 0 for t ∈ [0, T], therefore the polar cone (B(t)U 0)o={0}. The condition in Theorem 2.1 of Son [29] and Corollary 4.1 of Klamka [24] are required for the U 0−exact global controllability of the linear system corresponding to (17a) is satisfied. Hence, by the theorem on controllability of boundary control system, the population model (17a) is constrained controllable.
References
Sukavanam N, Tomar NK. Approximate controllability of semilinear delay control systems. Nonlinear Functional Analysis and Application 2007;12(1):53–59.
Lions JL. Optimal control of systems governed by partial differential equations: Die Grundlehren der Mathematischen Wissenschaften. New York: Springer; 1971. Translated from the French by S. K. Mitter 170 edition.
Troltzsch F. 2010. Optimal control of partial differential equations: theory, methods and applications. American Mathematical Society, Graduate Studies in Mathematics 112 edition.
Fursikov AV, Gunzburger MD, Hou LS. Boundary value problems and optimal boundary control for the Navier-Stokes system: the two-dimensional case. SIAM J Control Optim 1998;36(3):852–894.
Micu S, Zuazua E. Boundary controllability of a linear hybrid system arising in the control of noise. SIAM J Control Optim 1997;35(5):1614–1637.
Zhang B. Exact boundary controllability of the Korteweg-de vries equation. SIAM J Control Optim 1999;37(2):543–565.
Greiner G, Kuhn KG. Linear and semilinear boundary conditions: the analytic case. Semigroup theory and evolution equations, volume 135 of lecture notes in pure and applied mathematics. New York: Marcel Dekker; 1991. p. 193–211.
Lasiecka I. Unified theory for abstract parabolic boundary problems: a semigroup approach. Appl Math Optim 1980;6:287–333.
Lasiecka I, Triggiani R. Dirichlet boundary control problem for parabolic equations with quadratic cost: analyticity and Riccati’s feedback synthesis. SIAM J Control Optim 1983;21(1):41–67.
Lasiecka I, Triggiani R. Abstract parabolic systems. Control theory for partial differential equations: Continuous and approximation theories. i. Encyclopedia of mathematics and its applications. 71st ed. Cambridge: Cambridge University Press; 2000.
Glowinski R, Lions JL, He J. 2008. Exact and approximate controllability for distributed parameter systems: a numerical approach. Cambridge University Press, Encyclopedia of Mathematics and its Applications 117 edition.
Boulite S, Idrissi A, Maniar L. Controllability of semilinear boundary problems with nonlocal initial conditions. J Math Anal Appl 2006;316:566–578.
Engel KJ, Fijavz MK, Kloss B, Nagel R, Sikolya E. Maximal controllability for boundary control problems. Appl Math Optim 2010;62:205–227.
Kumpf M, Nickel G. Dynamic boundary conditions and boundary control for the one-dimensional heat equation. J Dyn Control Syst 2004;10(2):213–225.
Engel KJ. Spectral theory and generator property for one-sided coupled operator matrices. Semigroup forum. Springer; 1999. p. 267–295.
Byszewski L. Theorems about the existence and uniqueness of solutions of a semilinear evolution nonlocal Cauchy problem. J Math Anal Appl 1991;162:494–505.
Dubey SA, Bahuguna D. Existence and regularity of solutions to nonlocal retarded differential equations. Appl Math Comput 2009;215:2413–2424.
Jeong JM, Kim HG. Controllability of semilinear functional integrodifferential equations. Bulletin of the Korean Mathematical Society 2009;46(3):463–475.
Jeong JM, Kim JR, Roh HH. Controllability for semilinear retarded control systems in Hilbert spaces. J Dyn Control Syst 2007;13(4):577–591.
Tomar NK, Kumar S. Approximate controllability of nonlocal semilinear time-varying delay control systems. Nonlinear Dynamics and Systems Theory 2012;12(3): 303–310.
Greiner G. Perturbing the boundary conditions of a generator. Houst J Math 1987; 13(2):213–229.
Engel KJ, Nagel R. One-parameter semigroups for linear evolution equations. Berlin: Springer; 1995.
Pazy A. Semigroups of linear operators and applications to partial differential equations. New York: Springer; 1983.
Klamka J. Constrained controllability of nonlinear systems. J Math Anal Appl 1996;201(2):365–374.
Klamka J. Constrained controllability of semilinear systems with delays. Nonlinear Dyn 2009;56(1-2):169–177.
Azzouzi ME, Bouslous H, Maniar L, Boulite S. Constrained approximate controllability of boundary control systems. IMA J Math Control Inf 2015;33(3):669–683.
Chukwu EN, Lenhart SM. Controllability questions for nonlinear systems in abstract spaces. J Optim Theory Appl 1991;68(3):437–462.
Deimling K. Nonlinear functional analysis. Berlin: Springer;1985.
Son NK. A unified approach to constrained approximate controllability for the heat equations and the retarded equations. J Math Anal Appl 1990;150:1–19.
Mortad MH. Products and sums of bounded and unbounded normal operators: Fuglede-putnam versus Embry. Rev Roumaine Math Pures Appl 2011;56(3):195–205.
Acknowledgments
The authors would like to thank the anonymous reviewers for their insightful comments that significantly improved the quality of this paper. The first author would like to acknowledge the University Grants Commission, New Delhi, for providing financial support under grant No. F. 2-10/2011(SA-I).
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Kumar, S., Tomar, N.K. Mild Solution and Constrained Local Controllability of Semilinear Boundary Control Systems. J Dyn Control Syst 23, 735–751 (2017). https://doi.org/10.1007/s10883-016-9355-2
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10883-016-9355-2