Due to the nonlinear characteristics of the ship's thrusters and the slow response speed, which leads to a mismatch between the thrust allocation output and the controller output, in order to improve the performance of the unmanned ship dynamic positioning system, this paper proposes an unmanned ship dynamic positioning control method based on deep reinforcement learning, which considers the unmanned ship motion controller and the thrust allocation as a whole, and utilizes reinforcement learning algorithms to learn the ship and thruster's nonlinear dynamics of the ship and the thruster, and directly optimize the thruster commands to eliminate the error between the motion controller and thrust allocation. The feasibility and effectiveness of the proposed overall control method are verified through simulation tests.
Research on unmanned ship dynamic positioning system based on deep reinforcement learning
Ninth International Symposium on Advances in Electrical, Electronics, and Computer Engineering (ISAEECE 2024) ; 2024 ; Changchun, China
Proc. SPIE ; 13291
2024-10-16
Conference paper
Electronic Resource
English