移动机器人运动规划中的深度强化学习方法

随着移动机器人作业环境复杂度的提高、随机性的增强、信息量的减少,移动机器人的运动规划能力受到了严峻的挑战.研究移动机器人高效自主的运动规划理论与方法,使其在长期任务中始终保持良好的复杂环境适应能力,对保障工作安全和提升任务效率具有重要意义.对此,从移动机器人运动规划典型应用出发,重点综述了更加适应于机器人动态复杂环境的运动规划方法——深度强化学习方法.分别从基于价值、基于策略和基于行动者-评论家三类强化学习运动规划方法入手,深入分析深度强化学习规划方法的特点和实际应用场景,对比了它们的优势和不足.进而对此类算法的改进和优化方向进行分类归纳,提出了目前深度强化学习运动规划方法所面临的挑战和亟待解决的问题,并展望了未来的发展方向,为机器人智能化的发展提供参考.

Deep reinforcement learning for motion planning of mobile robots

The motion planning ability of mobile robots are facing a severe challenge with complex environment and less prior information. It is important to study the motion planning method and theory for a mobile robot, so that the mobile robot could adapt to complex environment in a long-running and ensure the work security and task efficiency. This paper mainly summarizes the method based on deep reinforcement learning(DRL) which can deal with the dynamic and complicated obstacles better. The DRL methods, which are based on value, policy and actor-critic, are introduced respectively. Then, the typical robot application in simulation environment and complex real world environment are analyzed based on DRL. After comparing the advantages and disadvantages in detail, the improvement and optimization direction for the DRL method are classified, and the challenges faced by motion planning method are put forward respectively. Finally, the prospects in the field of mobile robot motion planning method with DRL are discussed, which will provide reference for the development of intelligent robots.