Nondegenerate necessary conditions for nonconvex optimal control problems with state constraints (Q1289040): Difference between revisions

The optimal control problem \[ \min g(x(1)) \] is considered, with state equation on \([0,1]\), \[ \begin{cases} x'(t)= f(t,x(t),u(t))\\ x(0)= x_0\end{cases} \] and control/state constraints on \([0,1]\) \[ \begin{cases} u(t)\in \Omega(t)\\ h(t, x(t))\leq 0,\quad u(1)\in C.\end{cases} \] If a pair \((\overline x,\overline u)\) is a strong local minimizer, that is realizes the minimum among all pairs \((x,u)\) such that \(\|x-\overline x\|\leq \delta\), under suitable regularity assumptions the following necessary conditions for optimality can be derived. There exist \(p(t)\) absolutely continuous, \(\mu\) a nonnegative measure, \(\lambda\in [0,+\infty[\) such that \[ \begin{aligned} & \mu([0,1])+ \|p\|_{L^\infty}+ \lambda>0,\\ & -p'(t)= \Biggl(p(t)+ \int_{[0,t[} h_x(s, \overline x(s)) d\mu(s)\Biggr)\cdot f_x(t,\overline x(t),\overline u(t)),\\ & -\Biggl(p(1)+ \int_{[0,1]} h_x(s,\overline x(s)) d\mu(s)\Biggr)\in N_C(\overline x(1))+\lambda g_x(\overline x(1)),\\ & \text{spt }\mu\subset \{t\in [0,1]: h(t,\overline x(t))= 0\},\\ & \overline u(t)\text{ maximizes }\Biggl(p(t)+ \int_{[0,t[} h_x(s,\overline x(s)) d\mu(s)\Biggr)\cdot f_x(t,\overline x(t),\cdot)\text{ over }\Omega(t),\end{aligned} \] where \(N_C\) is the normal cone to \(C\). The degenerate case \(h(0,x_0)= 0\) makes the conditions above trivially satisfied, and additional constraint qualifications are needed to derive nondegenerate necessary conditions and optimality. This is the goal of the paper, in the framework of nonsmooth analysis.