Learning optimal motion planning for car-like vehicles

In this paper we propose a novel and generic approach to obtain the optimal motion of nonholonomic robots, considering kinematic and obstacle constraints. The algorithm uses reinforcement learning to build and update both the vehicle model and the optimal behaviour at the same time. The method overcomes some limitations of reinforcement learning techniques when they are employed in applications with continuous non-linear systems, such as car-like vehicles. In particular, a good approximation to the optimal behaviour is obtained by a look-up table without of using function interpolation. Both simulation and experimental results of learning optimal motion are reported. The results show the satisfactory performance of the method compared with the popular Q-learning algorithm.