When is the maximum principle for state constrained problems nondegenerate? (Q1340561): Difference between revisions

In the paper the following optimal control problem is considered: \[ \text{minimize } g(x(b)) \] subject to the state equation \[ x'= f(t, x, u) \] and to constraints \[ x(a)= x_ 0,\quad u(t)\in \Omega,\quad h(t, x(t))\leq 0,\quad d(x(b))\leq 0. \] On the other hand a pair \((x, u)\) of functions solving the minimization problem above is said to satisfy the maximum principle with multipliers \(p\in W^{1,1}(a, b; \mathbb{R}^ n)\), \(\mu\) a nonnegative measure, \(\alpha\geq 0\), \(\lambda\geq 0\), if \hskip10mm \(- p'(t)=\left(p(t)+ \int_{[a, t[} h_ x(s, x(s)) d\mu(s)\right) f_ x(t, x(t), u(t))\), \hskip10mm \(u(t)\in \text{argmax}_{\xi\in \Omega} \left(p(t)+ \int_{[a,t[} h_ x(s, x(s)) d\mu(s)\right) f_ x(t, x(t),\xi)\), \hskip10mm \(\text{spt } \mu\subset \{t\in [a, b]: h(t, x(t))= 0\}\), \hskip10mm \(\alpha_ i= 0\text{ whenever } d_ i(x(b))< 0\), \hskip10mm \(p(b)+ \int_{[a, b]} h_ x(s, x(s)) d\mu(s)= -(\alpha\cdot d_ x(x(b))+ \lambda g_ x(x(b)))\). The authors investigate the assumptions which lead a minimizing pair \((x, u)\) to satisfy the maximum principle with nontrivial multipliers, that is \(\mu\), \(\alpha\), \(\lambda\) not all equal to zero.