Pattern-preserving optimal control problems with increasing time horizon (Q6870044)

The authors consider two topological spaces \(\mathcal{U}\) and \(\mathcal{X}\), the space of controls and states, respectively, a cost functional \(\mathcal{J }:\mathcal{U}\times \mathcal{X}\rightarrow \mathbb{R}\), a set of admissible control-state pairs \(\mathcal{A}\subset \mathcal{U}\times \mathcal{X}\), and the optimal control problem defined by \(\inf_{\mathcal{U}\times \mathcal{X}} \mathcal{J}(u,x)\) subject to \((u,x)\in \mathcal{A}\), or equivalently \(\inf_{ \mathcal{U}\times \mathcal{X}}(\mathcal{J}+\chi _{\mathcal{A}})\), where \( \chi _{\mathcal{A}}\) is the indicator function of \(\mathcal{A}\). \(u\in \mathcal{U}\) is an optimal control for \(\mathcal{F}=\mathcal{J}+\chi _{ \mathcal{A}}\) if there exists \(x\in \mathcal{X}\) such that \(\mathcal{F} (u,x)=\inf_{\mathcal{U}\times \mathcal{X}}\mathcal{F}\). The purpose of the paper is to describe the behavior of sequences of such optimal control problems, through sequential \(\Gamma \)-limits. A sequence \((\mathcal{F}_{k})\) is sequentially equi-coercive if, for every \(C>0\) and every sequence \( (u_{k},x_{k})\) such that \(\mathcal{F}_{k}(u_{k},x_{k})\leq C\), there exists a subsequence of \((u_{k},x_{k})\) converging to some \((u,x)\) in \(\mathcal{U} \times \mathcal{X}\). A sequence \((u_{k},x_{k})\) in \(\mathcal{U}\times \mathcal{X}\) is minimizing for the sequence \(\mathcal{F}_{k}\) if \( \liminf_{k\rightarrow \infty }\mathcal{F}_{k}(u_{k},x_{k})=\liminf_{k \rightarrow \infty }\inf_{\mathcal{U}\times \mathcal{X}}\mathcal{F}_{k}\). \N\NThe first result considers a sequence \(\mathcal{F}_{k}:\mathcal{U}\times \mathcal{X}\rightarrow \mathbb{R}\) such that \(\Gamma -\mathrm{seq}(\mathcal{U}\times \mathcal{X})\)-converges to \(\mathcal{F}\) and a minimizing sequence \( (u_{k},x_{k})\). The following propositions hold: 1. If \((u_{k},x_{k}) \rightarrow (u,x)\in \mathcal{U}\times \mathcal{X}\), then \(\mathcal{F} (u,x)=\inf_{\mathcal{U}\times \mathcal{X}}\mathcal{F}=\lim_{k\rightarrow \infty }\inf_{\mathcal{U}\times \mathcal{X}}\mathcal{F}_{k}\); actually, \((u,x) \) is an optimal pair for \(\mathcal{F}\). 2. If \(\mathcal{F}\neq \infty \) and \((\mathcal{F}_{k})\) is sequentially equi-coercive, then (a) there exists an optimal pair \((u,x)\) of \(\mathcal{F}\) and a subsequence \( (u_{k_{n}},x_{k_{n}})\rightarrow (u,x)\) in \(\mathcal{U}\times \mathcal{X}\); (b) if, moreover, \(\mathcal{X}\) is metrizable and for every optimal control \( u\in \mathcal{U}\) there exists at most one \(x\in \mathcal{X}\) such that \( F(u,x)<\infty \), then \(u_{k}\rightarrow u\in \mathcal{U}\) implies \( x_{k}\rightarrow x\) and the preceding equalities hold; actually, \(u\) is an optimal control for \(\mathcal{F}\) and \(x\) is the (unique) corresponding optimal state. The authors define the notion of \((T_{k})\)-pattern preserving family, where \((T_{k})\) is a sequence \(0<T_{k}\rightarrow \infty \), and they prove that (sequential) \(\Gamma \)-convergence and equi-coercivity are sufficient conditions for ensuring pattern preservation. They then consider a sequence of problems parametrized by the final time \(T\in (0,+\infty ]\), with cost functionals of the form \(\mathcal{J}_{T}(u,x)=\int_{0}^{T}\ell (t,x(t),u(t))dt\), to be minimized over all admissible control-state pairs \( (u,x)\) belonging to \(\mathcal{A}=\{(u,x)\in \mathcal{U}\times \mathcal{X} :x=x^{u,x_{0}}\}\), where \(x^{u,x_{0}}\in \mathcal{X}=W_{\mathrm{loc}}^{1,1}([0,\infty ),\mathbb{R}^{n})\) denotes the unique global solution to the Cauchy problem \( x^{\prime }(t)=a(t,x(t))+b(t,x(t))u(t)\), \(x(0)=x_{0}\), the functions \(a\) and \(b\) being Borel measurable satisfying bounds and Lipschitz properties. The cost function \(\ell :[0,\infty )\times \mathbb{R}^{n}\times \mathbb{R} ^{m}\rightarrow \lbrack 0,+\infty ]\) is written as \(\ell (t,x,u)=\ell _{1}(t,x)+\ell _{2}(t,x,u)\) and satisfies different hypotheses. \N\NUnder a further coercivity hypothesis, the authors prove that the family of optimal control problems \(F_{T}\) is \((T_{k})\)-pattern preserving. They analyze the state equations for optimal control equations in the preceding context of differential equations. They analyze optimal control problems with increasing time horizons. The paper ends with examples of optimal control problems concerning switched dynamical systems for a large class of physical/engineering phenomena, or epidemics driven by susceptible-infectious-recovered models.