Optimality for periodic control of functional differential systems (Q1820424)

The problem considered is to minimize \(\int^{t_ 0}_{t_ 1}g(x(t),u(t))dt\) for the functional differential system of equations \(\dot x(t)=f(x_ t,u(t))\) under the constraint that the final state is equal to the initial state \(x_{t_ 0}=x_{t_ 1}\). The control u(t) is in \(U(t)\subset R^ m\) a.e., with U(t) closed convex and measurable in t, \(x_ t(s)=x(t+s)\) for \(s\in [-r,0]\), and x(t) is in \(R^ n\). Assuming some conditions weaker than approximate controllability necessary first order optimality conditions are derived. Next, second order necessary conditions are derived which allow to derive a suitable generalization of the \(\pi\)-criterion known for systems without delays (useful to check whether there exists a periodic control better than the constant one).