Necessary conditions for infinite horizon optimal control problems with state constraints (Q2001539)

The authors first consider the minimization problem \(\mathcal{M}(g^{\tau },\tau )\) on \([0,\tau ]\) written as minimize \(g^{\tau }(x(\tau ))\) over all trajectories of the following differential inclusion with state constraints \(x' (t)\in F(t,x(t))\), \(x(t)\in \Omega \), for \(t\in \lbrack t_{0},\tau ]\), \(x\in W^{1,1}(t_{0},\tau ;\mathbb{R}^{n})\), with the initial datum \((t_{0},x_{0})\in \lbrack 0,\tau ]\times \Omega \), where \( F:[0,\infty )\times \mathbb{R}^{n}\rightrightarrows \mathbb{R}^{n}\) is a multifunction and \(\Omega \) is a non-empty closed subset of \(\mathbb{R}^{n}\). The infimum of the minimization problem \(\mathcal{M}(g^{\tau },\tau )\) with the initial datum \((t_{0},x_{0})\) is denoted \(V^{\tau }(t_{0},x_{0})\). The main result concerning this problem proves that if \(g\) is a locally Lipschitz continuous function from \(\mathbb{R}^{n}\) to \(\mathbb{R}\) then \( V^{\tau }(.,.)\) is locally Lipschitz continuous on \([0,\tau ]\times \Omega \). The authors here assume different hypotheses on the data of the problem. They also prove properties of the minimizer \(\overline{x}(.)\) of \(\mathcal{M} (g,\tau )\). They then consider the infinite horizon optimal control problem \( \mathcal{B}_{\infty }:\) minimize \(\int_{t_{0}}^{\infty }L(t,x(t),u(t))dt\) over all trajectory-control pairs subject to the state constrained control system \(x' (t)=f(t,x(t),u(t))\), \(u(t)\in U(t)\), \(x(t)\in A\) for a.e. \(t\in \lbrack t_{0},\infty )\) and the initial condition \((t_{0},x_{0})\in \lbrack 0,\infty )\times A\). Here \(f:[0,\infty )\times \mathbb{R}^{n}\times \mathbb{R}^{m}\rightarrow \mathbb{R}^{n}\) and \(L:[0,\infty )\times \mathbb{R} ^{n}\times \mathbb{R}^{m}\rightarrow \mathbb{R}\) are given, \(A\) is a non-empty closed subset of \(\mathbb{R}^{n}\), \(U:[0,\infty )\rightrightarrows \mathbb{R}^{n}\) is a Lebesgue measurable set valued map with closed non-empty images. The authors introduce the Hamiltonian on \([0,\infty )\times \mathbb{R}^{n}\times \mathbb{R}^{n}\) as \(\mathcal{H}(t,x,p)=\sup \{\left\langle p,f(t,x,u)\right\rangle -L(t,x,u):u\in U(t)\}\). They impose hypotheses on the data of the problem which ensure the existence and uniqueness of a solution to the above differential equation on \(x\) for every datum \(x_{0}\) and every control. The purpose of the paper is to prove partial and full sensitivity relations for this problem. The main result here proves that if \(V(i,.)\) is locally Lipschitz continuous on \(A\) for all \( i\in \mathbb{N}\) large enough, then \(V\) is locally Lipschitz continuous on \( [0,\infty )\times A\). The authors also prove properties of the optimal solution \((\overline{x},\overline{u})\) to \(\mathcal{B}_{\infty }\) at \( (t_{0},x_{0})\in \lbrack 0,\infty )\times int\Omega \). For the proof, the authors essentially use tools from nonsmooth analysis. Finally the authors consider the class \(L(t,x,u)=e^{-\lambda t}l(x,u)\). Assuming that \(f\) is independent of \(t\), they prove that \(v(.)=V(0,.)\) is Lipschitz continuous on \(A\) for \(\lambda \geq 0\) large enough. The value function \(V(t,x)\) of \( \mathcal{B}_{\infty }\) which is equal to \(e^{-\lambda t}v(x)\) is Lipschitz continuous on \(A\) uniformly in \(t\geq 0\) for \(\lambda >0\) large enough.