Numerical solution of optimal control problems with discrete-valued system parameters (Q1864786)

A new technique is suggested to solve a class of optimal control problems which are governed by ordinary differential equations and involve discrete-valued system parameters. For that, the problem is decomposed into a bilevel optimization problem, where the `upper' problem is a purely discrete optimization problem, which can be solved by any suitable discrete optimization method, and the `lower' problem is a standard optimal control problem with fixed system parameters, for which likewise a variety of solvers is available. For solution of the upper level problem, especially a simulated annealing approach with memory is developed. A small numerical example illustrates the new approach.