基于深度确定性策略梯度算法的智能水下机器人局部路径规划  

Intelligent Underwater Vehicle Local Path Planning Based on Deep Deterministic Policy Gradient Algorithm

作　　者：吕茜 党康宁 Lv Xi;Dang Kangning(Science and Technology on Underwater Vehicle Laboratory,Harbin Engineering University,Harbin,China;HangJang-To-WeHe River Velley Water Diversion Project Construction Co.,Ltd.,Xi'an,China)

机构地区：[1]哈尔滨工程大学水下机器人技术重点实验室,黑龙江哈尔滨 [2]陕西省引汉济渭工程建设有限公司,陕西西安

出　　处：《科学技术创新》2023年第20期224-228,共5页Scientific and Technological Innovation

基　　金：陕西省水利科技计划项目:引汉济渭工程输水隧洞检测机器人(项目编号:2020slkj-5)。

摘　　要：路径规划是智能水下机器人技术研究的核心内容之一,是实现其自主航行和作业的关键环节。基于水下机器人的运动学模型,将深度确定性策略梯度(DDPG)算法应用于水下机器人的局部路径规划中,通过构造适当的奖励信号和设置合理的训练评估条件,使算法适用于水下机器人的运动学模型。仿真试验验证了采用DDPG算法训练的水下机器人能够在航道水域环境中安全快速地规划和避开障碍物,实现自主安全航行。Path planning is one of the core contents of intelligent underwater vehicle technology research,and it is the key link to realize its autonomous navigation and operation.Based on the kinematic model of underwater vehicle,this paper applies the depth deterministic policy gradient(DDPG)algorithm to the local path planning of underwater vehicle,and makes the algorithm suitable for the kinematic model of underwater vehicle by constructing appropriate reward signals and setting reasonable training evaluation conditions.Finally,this paper verifies that the underwater reasonable trained by DDPG algorithm can safely and quickly plan and avoid obstacles in the waterway water environment,and realize autonomous and safe navigation.

关 键 词：智能水下机器人 局部路径规划 深度确定性策略梯度(DDPG)算法 自主安全航行 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]