On-line optimization of switched-mode systems: Algorithms and convergence properties

This paper concerns the problem of on-line (realtime) computation of solutions to the optimal switching time problem in hybrid systems. The systems under consideration are autonomous, and the performance measure to be optimized has the form of a cost functional defined on the state trajectory. The state variable cannot, however, be measured directly and it has to be estimated by a suitable observer. In this paper, we propose an on-line optimization algorithm based on the state observer, and derive bounds on its convergence rate.