基于改进DQN算法的船舶全局路径规划研究  

Ship global path planning based on improved DQN algorithm

作　　者：关巍[1] 曲胜 张显库[1] 胡彤博 GUAN Wei;QU Sheng;ZHANG Xianku;HU Tongbo(Navigation College,Dalian Maritime University,Dalian 116026,China)

机构地区：[1]大连海事大学航海学院,辽宁大连116026

出　　处：《中国舰船研究》2025年第1期107-114,共8页Chinese Journal of Ship Research

基　　金：国家自然科学基金资助项目(51409033,52171342);中央高校基本科研业务费专项资金资助项目(3132023502)。

摘　　要：[目的]为提升实际海域环境下船舶航行路径的经济性与安全性,提出一种改进深度Q网络(DQN)算法的船舶全局路径规划方法。[方法]首先,引入优先经验回放机制赋予重要样本更高的权重,提升学习效率;然后,再通过决斗网络和噪声网络改进DQN的网络结构,使其对特定状态及其动作的价值评估更加准确,并同时具备一定的探索性和泛化性。[结果]实验结果表明,在马尼拉附近海域环境下,相比于A^(*)算法和DQN算法,改进算法在路径长度上分别缩短了1.9%和1.0%,拐点数量上分别减少了62.5%和25%。[结论]实验结果验证了改进DQN算法能够更经济、更合理地规划出有效路径。[Objective]In order to improve the economy and safety of ship navigation path in actual sea environment,this paper proposes a ship global path planning method with an improved Deep Q-Network(DQN)algorithm.[Method]First,a prioritized experience replay(PER)mechanism is introduced to the DQN to give higher weights to important samples and improve learning efficiency.Next,its network structure is improved through a duel network and noise network,enabling it to evaluate the values of specific states and actions more accurately and generalization capabilities.[Result]An experiment is carried out in the marine environment near Manila,and the results show that compared with the A^(*) algorithm and DQN algorithm,the improved algorithm reduces the path length by 1.9%and 1.0%respectively,and the number of turning points by 62.5%and 25%respectively.[Conclusion]It is verified that the improved DQN algorithm can plan the effective path more economically and rationally.

关 键 词：船舶 运动规划 DQN算法 优先经验回放(PER) 

分 类 号：U664.82[交通运输工程—船舶及航道工程] U675.9[交通运输工程—船舶与海洋工程]