智能系统学报2024，Vol.19Issue(6)：1492-1502,11.DOI:10.11992/tis.202304056
融合专家纠偏策略的移动机器人动态环境避障方法
Collision avoidance approach with heuristic correction policy for mobile robot navigation in dynamic environments
摘要
Abstract
Mapless navigation for mobile robots based on deep reinforcement learning(DRL)has received increasing at-tention from robotics and related research fields.The major challenge in mapless navigation is collision avoidance of dy-namic obstacles in unstructured environments.Therefore,this paper proposes a DRL algorithm that incorporates a heur-istic correction policy for robot autonomous navigation.The algorithm utilizes information from a 24-line laser radar sensor,target location,and robot velocity as inputs for DRL to generate action commands that regulate the robot's mo-tion.Experimental results demonstrate that,compared to other algorithms,the proposed approach can reach the target more efficiently in terms of distance and time while ensuring safety.Moreover,the algorithm is implemented in a real robot to verify and evaluate its performance,providing a technical reference for collision avoidance during its naviga-tion in dynamic environments.关键词
移动机器人/深度强化学习/机器人导航/非结构环境/动态避障/专家纠偏策略/自学习/端到端Key words
mobile robots/deep reinforcement learning/robot navigation/non-structural environment/dynamic colli-sion avoidance/heuristic correction policy/self-learning/end-to-end分类
信息技术与安全科学引用本文复制引用
田顺钰,欧阳勇平,魏长赟..融合专家纠偏策略的移动机器人动态环境避障方法[J].智能系统学报,2024,19(6):1492-1502,11.基金项目
国家自然科学基金项目(52371275). （52371275）
