Some remarks on nonconvex optimal control (Q580698)

Some control problems for the equation (1) \(\dot x=f(t,x,u)\), \(x(t_ 0)=x_ 0\), u(t)\(\in C(t)\) for all t in a Banach space X and for the corresponding orientor equation (2) \(\dot x\in N(t,x)=f(t,x,C(t))\), \(x(t_ 0)=x_ 0\) are studied where f is a continuous function and the controls u(\(\cdot)\) are measurable. The sets C(t) and N(t,x) may not be convex ones. Some conditions for a quasitrajectory x(\(\cdot)\) of the system (1) or (2) to be an extremal one (i.e. it lies on the boundary of the corresponding emission zone) are obtained as some geometrical version of the Pontryagin maximum principle. As an example the nonlinear Neumann problem for some second order partial differential equation with boundary controls is considered.