煤矿安全2025，Vol.56Issue(3)：233-241,9.DOI:10.13347/j.cnki.mkaq.20241415
融合快速遍历随机树和Q强化学习的煤矿轮式机器人路径规划关键技术
Key technologies for path planning of coal mine wheeled robots integrating fast traversal random trees and Q-reinforcement learning
摘要
Abstract
There are a lot of dangerous environment and large volume transportation tasks in the coal mining process.Improving the working quality of coal mine equipment is the key to the intelligent construction of coal mine.In order to improve the efficiency of coal mine production and the safety of transportation tasks,a path planning method of coal mine wheeled robot based on reinforce-ment learning is proposed.This method uses greedy strategy to guide the direction of random tree expansion,uses Markov decision to reduce the invalid nodes generated during expansion,smooths the path trajectory through third-order Bessel curve,and adds ex-pert experience playback pool to improve the computational efficiency.The experimental results show that the global path length generated by the research method can be reduced by at least 10.71%compared with other algorithms in the planned path length test.In the multi-obstacle scenario planning time test,the planning time of the research method is only 0.452 s.In the analysis of obstacle avoidance effect,the research method can effectively avoid static obstacle and dynamic obstacle.The research method has faster path planning efficiency and can generate safer robot running paths.关键词
煤矿轮式机器人/路径规划/快速遍历随机树/Q强化学习/IDDPG-GAIL算法Key words
coal mine wheeled robot/path planning/fast traversal of random trees/Q reinforcement learning/IDDPG-GAIL al-gorithm分类
矿业与冶金引用本文复制引用
温天飞,高宇,王全,杨闯..融合快速遍历随机树和Q强化学习的煤矿轮式机器人路径规划关键技术[J].煤矿安全,2025,56(3):233-241,9.基金项目
国家重点研发计划资助项目(2022YFB4703600) （2022YFB4703600）
2023年辽宁省人工智能创新发展计划重大专项资助项目(2023JH26/10100006) （2023JH26/10100006）
中煤科工集团科技创新基金资助项目(2023-QN003) （2023-QN003）
