基于强化学习的AUV对接控制算法研究  

Research on AUV Docking Control Algorithm Based on Reinforcement Learning

作　　者：庄英豪 张天泽 张悦[1] 李沂滨[1] ZHUANG Yinghao;ZHANG Tianze;ZHANG Yue;LI Yibin(Institute of Marine Science and Technology,Shandong University,Qingdao 266000,China;College of Mechanical and Electrical Engineering,China University of Petroleum(East China),Qingdao 266580,China)

机构地区：[1]山东大学海洋研究院,山东青岛266000 [2]中国石油大学(华东)机电工程学院,山东青岛266580

出　　处：《数字海洋与水下攻防》2024年第5期464-470,共7页Digital Ocean & Underwater Warfare

基　　金：国家自然科学基金面上项目“面向多潜艇故障分布式诊断的增量联邦迁移学习”(62273202)。

摘　　要：自主式水下航行器(AUV)是人类探索和利用海洋的重要装备,能否足够智能化地解决路径规划控制问题是AUV完成其它复杂任务的基础。考虑终端姿态约束下的局部路径规划问题,结合AUV的自主对接控制这一实际使用场景,基于改进的深度强化学习算法(DRL)开发了一种对接控制器,使其具备自主对接能力,延长其续航时间。考虑实际工作场景中的复杂海浪干扰因素,使用了非线性扰动观测器(NDO)来估计AUV三维运动中各自由度的外部扰动,并结合可测量的状态量为DRL智能体设计了科学的观测量及奖励函数,使AUV能够在扰动环境中完成三维对接控制任务。仿真结果表明了该方法的有效性和鲁棒性。Autonomous underwater vehicles(AUVs)is an important kind of equipment for human to explore and utilize the ocean.Intelligent solution of path planning and control is the basis for an AUV to accomplish other complex tasks.Considering the local path planning problem under terminal attitude constraint and combining with AUV autonomous docking control,a docking controller is developed based on the improved Deep Reinforcement Learning(DRL)algorithm.It enables the AUV to dock autonomously and can increase AUV endurance.Considering the complex wave disturbance factors in the practical operating scenario,a nonlinear disturbance observer(NDO)is used to estimate the external disturbances of each degree of freedom in AUV three-dimensional motion.In order to ensure that the AUV can accomplish the three-dimensional docking control task in a disturbed environment,scientific observation quantities and reward functions are designed for the DRL agent in combination with measurable state quantities.Simulation results demonstrate the effectiveness and robustness of the proposed method.

关 键 词：自主式水下航行器 路径规划 对接控制 强化学习 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]