Fuzzy PID Autopilot For Ship Steering Based On BP Neural Network

Fixed quantizing factor of fuzzy controller, which is determined by physical and fuzzy domain of input variables, can’t meet the requirement of control precision if input variables are changing constantly. According to the relationship between physical domain of input variables and quantization, a method of adjusting quantizing factor on line is proposed based on BP neural network.The fluid forces and moments on ship are deposed into fluid forces and moments acting on bare hull, open-water propeller, open-water rudder and fluid forces and moments caused by interference between them. Building ship maneuvering mathematical model group and verifying its accuracy are necessary which is controlled by ship course controller.For embodying the advantages of fuzzy controller, a conventional PID autopilot for ship steering is designed firstly, which parameters are adjusted manually. The control accuracy of autopilot for ship steering will decrease due to uncertainty of controlled model and change of navigational environment. Aiming at the phenomenon, using a fuzzy controller, which inputs are course error and its rate of change, to adjust the parameters of PID can improve the performance of autopilot for ship steering.The quantizing factor of conventional fuzzy controller is set artificially relay on desired course of ship. A BP neural network, which input is desired course and output is quantizing factor, is proposed. Training the BP neural network according to training sample which are desired course and quantization factor determined manually.Simulations show the fuzzy autopilot for ship steering has better performance than conventional PID autopilot for ship steering and the performance of fuzzy autopilot for ship steering based on BP neural network is better than conventional fuzzy autopilot for ship steering...