Self-learning Adaptive DRNN For Ship Heading Hybrid Control System

Course control is one of the key technologies of ship motion.It is very important for ship navigation safety,route planning and economy.It has been the focus and research hotspot in the field of shipping.Unmanned ship and intelligent shipping pose new challenges to ship navigation control.It requires intelligent ship,higher redundancy,safety,reliability and economy.It needs multi-source propulsion to realize the autonomy and flexibility of the ship.The research on ship course control is mostly in the field of autopilot,and the research heat has been reduced.The research on ship dynamic positioning has been paid attention to,but the autopilot control is indispensable for ocean navigation,and the cooperative control of autopilot and bow thruster is lacking.In addition,the ship’s characteristics are significantly affected by the interference factors such as loading,speed,draft,wind,wave and current.The ship’s course control has the characteristics of slow speed change,large maneuvering and frequent adjustment.Its hysteresis,overshoot,oscillation and navigation fuel saving are always difficult problems.In order to optimize the ship navigation control objectives and improve the navigation economy,an online learning adaptive heading compound control strategy based on extended Kalman filter(EKF),adaptive observer,dynamic recurrent neural network(DRNN)and flexible PID is proposed.The ship course compound control system is designed and the approximate model of ship motion is established.The ship course control is divided into EKF estimation,course observation,course keeping feedback control and ship steering feed forward control.The innovations are as follows(1)The coordinated control of bow thruster and autopilot optimizes the economy of ship navigation;(2)The ship’s heading direction and the direction of wind,wave and current interference force are combined to form the ship’s sailing direction;(3)A learning adaptive composite algorithm based on EKF,adaptive observer,DRNN and flexible PID is proposed.A semi-physical simulation system for ship course control is developed.The MATLAB simulation and semi physical simulation test of the above algorithm are carried out,the results are as follows:(1)Based on EKF online update DRNN network to speed up the algorithm;(2)Taking the minimum criterion of the sum of squares of heading error performance function as the evaluation function,the steepest descent method of gradient chain rule and the Jacobian sensitivity information optimization DRNN as the optimization methods,the adaptive learning process is realized;(3)Bow thruster and autopilot coordinate to adjust the ship’s heading has higher tracking accuracy,stability and better fuel saving performance than autopilot alone.Therefore,the course compound control based on EKF,adaptive observer,DRNN and flexible PID learning adaptive,and the cooperative control of bow thruster and autopilot have practical significance and promotion value.