Deep reinforcement learning combines the perception ability of deep learning with the decision-making ability of reinforcement learning, and discovers the optimal strategy for the task through continuous trial and error learning of the agent. This paper studies the deep reinforcement learning method for multi-ship path planning. We combine the characteristics of ship navigation to redesign the reward function and the continuous action space and state space, and improve the Q value calculation method based on the characteristics of ship planning. Finally, we set up a simulation experiment to show that the method proposed in this paper can effectively plan collision-free paths for multiple ships at the same time.
Research on deep reinforcement learning method for multi-ship path planning
International Conference on Mechanisms and Robotics (ICMAR 2022) ; 2022 ; Zhuhai,China
Proc. SPIE ; 12331
10.11.2022
Aufsatz (Konferenz)
Elektronische Ressource
Englisch