Mobile robot control in dynamic environments based on hybrid intelligent system

Abstract

A hybrid intelligent control system is proposed for robot navigation. Fuzzy modelization and planning is combined with reinforcement learning to perform robust and optimal control. Fuzzy modelization and planning is efficient, but it lacks self-adaptability. Learning control can obtain the optimal control policy by on-line learning, but it takes long time for the strategy to converge. The hybrid method obtains the advantages of both and has better control performance than either of them. In the computer simulation experiment, the hybrid intelligent method can perform efficient and optimal control for robot navigation in dynamic environments.