Sampling-based Optimal Motion Planning for Non-holonomic Dynamical Systems

Sampling-based motion planning algorithms, such as the Probabilistic RoadMap (PRM) and the Rapidly-exploring Random Tree (RRT), have received a large and growing amount of attention during the past decade. Most recently, sampling-based algorithms, such as the PRM* and RRT*, that guarantee asymptotic optimality, i.e., almost-sure convergence towards optimal solutions, have been proposed. Despite the experimental success of asymptotically-optimal sampling-based algorithms, their extensions to handle complex non-holonomic dynamical systems remains largely an open problem. In this paper, with the help of results from differential geometry, we extend the RRT* algorithm to handle a large class of non-holonomic dynamical systems. We demonstrate the performance of the algorithm in computational experiments involving the Dubins' car dynamics.