【作者】 张煜； 杨才霈； 赵盈琪； 刘颉； 魏伟；

【Author】 Zhang Yu;YANG Cai-Pei;Zhao Ying-Qi;Liu Jie;Wei Wei;Wuhan Second Ship Design and Research Institute;School of Artificial Intelligence and Automation,Huazhong University of Science and Technology;School of Civil and Hydraulic Engineering,Huazhong University of Science and Technology;

【机构】 武汉第二船舶设计研究院； 华中科技大学人工智能与自动化学院； 华中科技大学土木与水利工程学院；

【摘要】 面向高随机、多动态障碍物复杂环境下的路径规划任务,提出了一种基于多目标强化学习的水下无人潜航器路径规划方法。首先,在考虑随机目标与多动态障碍物的路径规划主场景基础上,构建多个子目标(动态避障、目标接近)场景;其次,提出基于集成学习与PER-DDQN的路径规划框架;最后,利用人工势场法改进并优化路径规划框架的多目标策略集成方式。实验结果表明,相比于传统路径规划方法,该方法具有更高的路径规划成功率和更短的路径长度。更多还原

【Abstract】 Aiming at the path planning task in the complex environment with high randomness and multiple dynamic obstacles,a path planning method for unmanned underwater vehicle based on multi-objective reinforcement learning is proposed.First,on the basis of the main scenario of path planning considering random targets and multiple dynamic obstacles,multiple sub-target(dynamic obstacle avoidance,target approaching) scenarios are constructed;a path planning framework based on ensemble learning and PER-DDQN is further proposed;finally,using The artificial potential field method improves and optimizes the multi-objective strategy integration method of the path planning framework.The experimental results show that,compared with the traditional path planning method,this method has higher path planning success rate and shorter path length.更多还原