Research on collision avoidance method of intelligent ship navigation based on reinforcement learning

Abstract

To reduce ship collision avoidance accidents caused by human factors and improve the safety of ship navigation, a collision-avoidance model for intelligent ship navigation is proposed, which is based on deep reinforcement learning. The model is based on a reinforcement learning algorithm. Information such as the current position, speed, and phase of the ship is used as the input of the model. The optimal policy is solved by the value function solver model. A specific reward function is designed to combine a Vessel Conflict Ranking Operator (VCRO) with the reward function. In this way, an intelligent ship navigation collision avoidance model is constructed. Firstly, the specific content of this collision sorting factor is introduced. Then, the realization process of the collision avoidance algorithm for intelligent ship navigation is introduced. Finally, the training and simulation tasks of the model are completed through python programming. The results show that the collision avoidance model can effectively avoid obstacles and ensure the safety of ship navigation. The proposed method can provide a reference for the research of intelligent ship collision avoidance technology.