Second order necessary and sufficient optimality conditions for singular solutions of partially-affine control problems (Q1713285)

The control problems involve minimizing a cost functional \(\varphi_0(x(0), x(T))\) among the trajectories of the differential system \[ x'(t) = f_0(x(t), u(t)) + \sum_{j=1}^m v_j(t) f_j(x(t), u(t)) \quad (0 \leq t \leq T) \] where \(v_1(t), \dots, v_m(t)\) are the \textit{affine} controls taking values in a set \(V\) and the components \(u_1(t), \dots, u_p(t)\) of the vector function \(u(t)\) are the \textit{nonlinear} controls taking values in a set \(U.\) The \textit{state variable} \(x(t)\) is a \(n\)-dimensional vector function. In addition, there are a finite number of equality and inequality constraints \[ \varphi_k(x(0), x(T)) = 0 \, , \quad \varphi_\ell(x(0), x(T)) \leq 0 \, . \] The author shows first order necessary optimality conditions based on a generalized version of Lagrange multipliers and two second order necessary conditions; the second one is then strengthened to a sufficient condition. This is used, among other things, to justify convergence of a shooting algorithm, previously proposed and tested by the author.