Learning autonomous behaviours for non-holonomic vehicles

In this paper we propose a generic approach to acquire navigation skills for nonholonomic vehicles in unknown environments. The algorithm uses reinforcement learning to update both the vehicle model and the optimal behaviour at the same time. After the training phase, the vehicle is able to explore the environment through a wall-following behaviour. The vehicle can also reach any goal position by the virtual wall concept. The method does not require function interpolation to obtain a good approximation to the optimal behaviour. The learning time was only a few minutes to acquire the wall-following behaviour. Both simulation and experimental results are reported to show the satisfactory performance of the method.