Reinforcement learning control for ship steering based on general fuzzified CMAC

A general fuzzified cerebellar model articulation controller (GFCMAC) is proposed, in which the fuzzy membership functions are utilized as the receptive field functions. The mapping of receptive field functions, the selection law of membership with its parameters and the learning algorithm are presented. Reinforcement learning base on GFCMAC is applied to ship steering control, as provides an efficient way for the improvement of ship steering control performance. It removes the defect that the conventional intelligent algorithm learning must be provided with some sample data. The parameters of controller are on-line learned and adjusted. It can deal with the uncertainty of ship control in a way. Simulation results show that the ship course can be properly controlled in case of the disturbances of wave, wind, current and error in measure apparatus exist. It is demonstrated that the proposed algorithm is a promising alternative to conventional autopilots.