PED: Path-Encoding-based Derivative-free Method

With the development of technology and the growing demands, hybrid systems are becoming more and more complicated and larger in scale. Due to the notorious state explosion problem, existing methods for optimal hybrid control synthesis problems can not scale well to compositional real-time hybrid systems. In this paper, we propose a path-encoding-based derivative-free method (PED) to solve such problems. We encode the search of the potential path and other control parameters as a combined optimization problem, which can then be solved by existing solvers efficiently.

Due to the paper space limitation, we upload the detail of all the configurations and results of the experiments on this website. The experiments show our method can handle such complex cases efficiently with satisfactory success rates.