依据我国煤矿智能绿色发展战略，深入分析了国内外智能拣矸系统的研究现状，指出研发适用于井下的多机械臂煤矸石智能分拣机器人是破解煤矸分拣难题的重要发展方向，凝练了直接影响和制约我国煤矸石智能分拣高质量发展的“煤矸石准确识别、精准跟踪和可靠抓取、多目标任务多机械臂协同分拣”三大关键共性技术难题，并给出了解决思路和方法。针对煤矿井下煤矸石被煤泥严重包裹识别难，提出了“X射线+视觉”煤矸石识别与匹配方法、基于点云数据的煤矸石抓取特征提取方法，实现目标矸石的快速识别和最优抓取特征提取；针对煤矸石形态各异、动态环境抓取难，提出了基于ORB+BEBLID特征的FLANN动态目标高效匹配方法、基于FDSST的动态目标精准跟踪方法、基于三环PID的机械臂同步跟踪轨迹规划方法，实现机械手对高速传输的动态矸石稳定抓取；针对煤矸石随机分布、障碍多、多机械臂任务分配难，提出了改进匈牙利算法的多机械臂动态空间协同分拣方法，确保系统收益的前提下实现多机械臂在动态空间中高效协同工作。现场工业性试验研究结果表明，针对三大关键共性技术所提出的方法能够有效破解煤矸石高效识别和抓取特征提取、机械臂动态目标同步跟踪稳定抓取、多机械臂高效协同分拣等难题，通过构建完整的多机械臂煤矸石智能分拣机器人系统，提高了煤矸石智能分拣系统的可靠性和分拣效率。

According to the intelligent green development strategy of coal mine in our country, the research status of intelligent coal gangue sorting system at home and abroad is deeply analyzed. It is pointed out that the development of multi-arm coal gangue intelligent sorting robot suitable for underground is an important development direction to solve the problem of coal gangue sorting. The three key common technical problems of “accurate recognition, accurate tracking and reliable grasp of coal gangue, multi-target task and multi-arm collaborative sorting” which directly affect and restrict the high quality development of intelligent sorting of coal gangue in China have been condensed, and the solutions are given. For the problem that coal gangue is badly encapsulated by coal slime in underground coal mine, the method of “X-ray and visual” recognition and matching of coal gangue, the method of extracting feature of coal gangue based on point cloud data are put forward to realize fast recognition and optimal feature extraction of target gangue. An efficient FLANN dynamic target matching method based on the fusion of ORB and BEBLID features, an accurate dynamic target tracking method based on FDSST, and a synchronous tracking trajectory planning method for manipulator based on three-loop PID are proposed to realize the stable grasping of dynamic gangstone with high-speed transmission by manipulator. For the random distribution of coal gangue and difficult task allocation of multi-manipulators with obstacles, a dynamic space cooperative sorting method of multi-manipulators based on improved Hungarian algorithm was proposed, which realized efficient cooperative work of multi-manipulators in dynamic space on the premise of ensuring system revenue. The results of field industrial experiment show that the proposed method for the three key common technologies can effectively solve the problems of efficient recognition and grasping feature extraction of coal gangue, synchronous tracking and stable grasping of dynamic targets of manipulators, efficient cooperative sorting of multi-manipulators, and construct a complete multi-manipulator intelligent sorting robot system for coal gangue. It improves the reliability and efficiency of the gangue intelligent sorting system.