Approximate dynamic programming methods applied to far trajectory planning in optimal control

Abstract

There are several applications that need far trajectory planning within optimal control problems. One use case is the optimal predictive control of plug-in hybrid electric vehicles (PHEV). It is possible to find an optimal control with models of the vehicle and environment and a fast optimization algorithm that is capable to calculate over long distances within seconds so that dynamic information such as traffic can be recognized quickly. In this paper, several methods based on dynamic programming (DP) are combined to generate approximated optimal control trajectories with a reduced computational complexity to achieve close-to-real-time application. The resulting trajectories are transferred as strategic planning trajectory to subordinated vehicle controllers. Close-to-optimal trajectories are achieved with a large reduction in memory.