Ship Trajectory Tracking Using Improved Simulated Annealing and Reinforcement Learning

It plays a vital role in the efficiency, safety and maneuverability of navigation to control ship's trajectory precisely. Based on reinforcement learning algorithm, compensation is made on the output of traditional PID controller. In order to save the time of adjusting the PID parameters, the paper introduces an improved simulated annealing algorithm. The designed controller can overcome the external interference, such as wind, wave and flow. Simulation results under the MOOS which was written by Paul Newman to support operations with autonomous marine vehicles in the MIT Ocean Engineering and the MIT Sea Grant programs, show that the ship course can be properly controlled even if changeable wave and wind exist. In addition, the algorithm is applicable for several unmanned systems.