No-gap second-order optimality conditions for optimal control problems with a single state constraint and control (Q5900530): Difference between revisions

This paper studies the optimal control problem \[ \min_{u,y} \int_0^T \ell(u(t),y(t))\,dt+\varphi(y(T)) \tag{P} \] subject to \(\dot y(t)=f(u(t),y(t))\) a.e. on \([0,T]\), \(y(0)=y_0\), \(g(y(t))\leq0\) for all \(t\in[0,T]\), where the data are \(\ell:{\mathbb R}\times{\mathbb R}^n\to{\mathbb R}\), \(\varphi:{\mathbb R}^n\to{\mathbb R}\), \(f:{\mathbb R}\times{\mathbb R}^n\to{\mathbb R}^n\), \(g:{\mathbb R}^n\to{\mathbb R}\), \(T>0\), \(y_0\in{\mathbb R}^n\). The control \(u(\cdot)\) is scalar-valued. It is assumed that the corresponding Hamiltonian is strongly convex in \(u\) (so that the control \(u(\cdot)\) is continuous on \([0,T]\)), and that the trajectory \((u,y_u)\) has a finite set of junction times (split into regular entry, exit, and touch points). The authors develop second order necessary and sufficient optimality conditions, which they call ``no gap'' conditions. This means that the change between the necessary and sufficient optimality conditions is between the strict and nonstrict inequality. A special attention is devoted to weak local solutions of (P) satisfying a certain quadratic growth condition. Main tools are abstract optimization (applied to the optimal control setting) and junction conditions analysis. At the end of the paper the authors discuss the infinite number of junction points as an open problem.