洋流干扰下低速欠驱动AUV的三维路径规划  

3D path planning for low-speed underdriven AUV under ocean current disturbance

作　　者：陈世同[1] 鲁子瑜 CHEN Shitong;LU Ziyu(College of Intelligent Systems Science and Engineering,Harbin Engineering University,Harbin 150001,China)

机构地区：[1]哈尔滨工程大学智能科学与工程学院,黑龙江哈尔滨150001

出　　处：《智能系统学报》2025年第2期425-434,共10页CAAI Transactions on Intelligent Systems

摘　　要：海洋洋流对低速欠驱动水下运载器(automatic underwater vehicle,AUV)的航行影响巨大,会增加航行时间、增大能源消耗以及改变航行轨迹,故在洋流干扰的情况下规划出一条最优航行路线有着重要的意义。本文主要分析了洋流对AUV的影响机理,由于传统的强化Q网络(deep Q-network,DQN)路径规划算法在复杂环境下存在过估计的问题,提出了基于优先经验回放方法的改进DQN路径规划算法,同时对动作设计和奖励函数进行优化。在基于S57海图数据建立的三维海洋环境下,利用地球与空间研究机构(earth&space research,ESR)提供的洋流数据进行路径规划仿真。实验结果表明,在充分考虑洋流干扰的情况下,改进后的DQN算法能够规划出较优的全局路径规划,提供一条时间最短且能耗最低的航行路线,为AUV水下航行提供参考。Ocean currents,which have a substantial impact on the navigation of low-speed,underdriven AUVs,can increase navigation time,raise energy consumption,and change the navigation trajectory.Therefore,planning an optimal navigation route that accounts for the disturbance of ocean currents is of considerable importance.This study mainly analyzes the mechanism by which ocean currents influence AUVs and proposes an improved DQN path planning algorithm based on the prioritized experience replay method.This modification addresses the problem of overestimation,which is a common issue when using a traditional DQN path planning algorithm.Additionally,the action design and reward functions are optimized.Path planning simulations are conducted in a 3D ocean environment,which is established based on S57 chart data and ocean current data provided by Earth&Space Research.Experimental results show that the improved DQN algorithm generates a more effective global path planning,offering a navigation route that minimizes time and energy consumption.This work provides a valuable reference for underwater AUV navigation,fully considering the impact of ocean current disturbances.

关 键 词：自主水下运载器 强化学习 洋流干扰 路径规划 三维海洋环境 强化Q网络 S57海图 奖励函数 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]