Normality and second-order optimality conditions in state-constrained optimal control problems with bounded minimizers

The author considers the optimal control problem: \(\varphi (p)\rightarrow min \), \(\overset{.}{x}(t)=f(x(t),u(t),t)\) for a.a. \(t\), \(e_{1}(p)\leq 0\), \( e_{2}(p)=0\), \(u(t)\in U(x(t),t)\) for a.a. \(t\), \(g(x(t),t)\leq 0\) for all \(t\) , where \(p=(x_{0},x_{1})\), \(x_{0}=x(0)\), \(x_{1}=x(1)\), \(t\in \lbrack 0,1]\), and \(U(x,t)=\{u\in \mathbb{R}^{m}:r(x,u,t)\leq 0\}\), where the mappings \( \varphi :\mathbb{R}^{2n}\rightarrow \mathbb{R}\), \(e_{i}:\mathbb{R} ^{2n}\rightarrow \mathbb{R}^{k_{i}}\), \(f:\mathbb{R}^{n}\times \mathbb{R} ^{m}\times \mathbb{R}\rightarrow \mathbb{R}^{n}\), \(r:\mathbb{R}^{n}\times \mathbb{R}^{m}\times \mathbb{R}\rightarrow \mathbb{R}^{q}\), and \(g:\mathbb{R} ^{n}\times \mathbb{R}\rightarrow \mathbb{R}\) are such that \(\varphi ,e_{1},e_{2},f,r\) are twice continuously differentiable and \(g\) is thrice continuously differentiable. The control process \((\overline{x},\overline{u}) \) is regular with respect to the constraints, if for all \(t\in \lbrack 0,1]\) and \(u\in \mathcal{U}(t)\), the closure of \(\overline{u}(\cdot )\) with respect to the Lebesgue measure, the system of active gradients \( (r^{j})_{u}^{\prime }(\overline{x}(t),u,t)\), \(j\in J(u,t)\), the set of active indexes for mixed constraints, supplemented, for \(t\in T_{0}=\{t\in \lbrack 0,1]:g(\overline{x}(t),t)=0\}\), the active set for state constraints, with the gradient \(\Gamma _{u}^{\prime }(\overline{x}(t),u,t)\), where \(\Gamma (x,u,t)=\left\langle g_{x}^{\prime }(x,t),f(x,u,t)\right\rangle +g_{t}(x,t)\), is linearly independent. The author proves properties of such regular control processes. The author introduces a notion of normality for this problem, which is based on the linearization of the control problem and the corresponding linearized differential system. The control process \((\overline{x},\overline{u})\) is normal if the rank of some controllability \(k\times k\)-matrix \(Q\) is equal to \(k\). Given the reference control process \((\overline{x},\overline{u})\) and a pair \((\delta x_{0},\delta u)\in \mathcal{X}\), the author introduces appropriate spaces and operators and he proves continuity and surjectivity properties of these operators. The main result of the paper proves second-order necessary optimality conditions.\ If \((\overline{x},\overline{u} )\) is an optimal control process which is regular with respect to the constraints and normal, the state constraints being compatible with the endpoint constraint, the set \(\Lambda (\overline{x},\overline{u})\) of vectors \(\lambda \in (\mathbb{R}^{1+k_{1}+k_{2}})^{\ast }\) for which there exist \(\psi ,\nu ,\mu \) such that the corresponding set of Lagrange multipliers \((\lambda ,\psi ,\nu ,\mu )\) generated by \(\lambda \), satisfies the maximum principle, is non empty. The paper ends with an equivalence result concerning necessary optimality conditions.

Recommendations

Cites work

Cited in

(4)

This page was built for publication: Normality and second-order optimality conditions in state-constrained optimal control problems with bounded minimizers

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6155315)