A Method of Multi-USV Reward Design Using Fuzzy Control

This paper investigates the path planning problem of multiple unmanned surface vehicles (USVs). Since the maritime environment is more complex and variable than the land environment, none of the currently known path planning methods can overcome the uncertainties involved with desirable results. Deep reinforcement learning methods offer great promise for solving maritime uncertainty problems. However, the design of reward functions is often very difficult when it comes to solve practical problems. To enhance the path planning capability of multiple USVs in the maritime environment, we analyzed how maritime uncertainty affected path planning and designed a fuzzy logic-based reward function. This function is capable of effectively guiding the training of USVs while possessing good robustness to adapt to environmental uncertainty. In this paper, simulations of the motion model of USVs and the maritime environment are conducted, and the method’s effectiveness is verified through experiments. The results show that the model can demonstrate strong adaptability in the complex maritime environment.