Robot Path Planning Based On Reinforcement Learning And Artificial Potential Field

Recently, with the request of application strengthening unceasingly, the technique of robot develops continuously. As a high-tech comprehensive Academics, It's development extends many realms of technique research and arouses the development of these realms. Path-planning receives the scholar's attention all the time, as a key-technology in robot system. Now, many new methods of Path-planning have flowed out with the development of robot technique.This paper also presents a new method about robot path planning, combining with artificial potential field and reinforcement learning theory. The method uses a reward function to generate a potential field, and then abstracts some features from the potential field as candidates of sub-goals, and plans path online through some heuristics. The best-known and most often-cited problem in the potential field method is local minima. But our method does not have this limitation because the local minima are used to form sub-goals. The disadvantage of the local minima in the previous approaches of potential field turns out to be an advantage in our method. Simulation results show the feasibility and the validity of the method.