Abstract: The gangue is wrapped by slurry in underground coal mine, which causes difficult coal-gangue recognition and sorting. The underground working space is narrow, so the equipment layout is difficult, and the diversion of coal-gangue is difficult. Therefore, developing a high-performance, highly reliable intelligent coal-gangue sorting robot is necessary. The paper analyzes the research status of coal-gangue recognition, robot trajectory plan and multi-dynamic-target multi-robot collaborative control technology of intelligent coal-gangue sorting robot. This paper points out that the coal-gangue sorting work environment is complex, and its weight and shape of coal-gangue are irregular and randomly distributed. Therefore, the three key technologies for intelligent coal-gangue sorting robot are recognition and grasping features extraction of coal-gangue in complex environment, stable and reliable grasping of coal-gangue in unstructured environment, and intelligent collaborative sorting of multi-target multi-robot. It is proposed that in order to realize the intelligent sorting of coal-gangue by the robot, further research should be carried out. The research includes the methods of coal-gangue recognition and sorting feature extraction suitable for underground, accurate positioning and synchronous tracking of dynamic targets, online trajectory planning of mechanical arms, and intelligent collaborative control of multiple mechanical arms. By soring out the above three key technologies, it can be concluded as follows. The construction and expansion of coal-gangue data set, recognition and grasping features extraction of coal-gangue are the key technologies to achieve efficient coal-gangue recognition. Precise tracking of dynamic coal-gangue, trajectory planning of synchronous tracking dynamic target of mechanical arm and fast and stable grasping of large quality targets are the key technologies to realize stable coal-gangue grasping by mechanical arms. Multi-task efficient allocation, anti-collision path planning and intelligent collaborative control are the key technologies to achieve efficient intelligent collaborative sorting of multiple mechanical arms. According to the common problems at present, this paper puts forward the solutions. In the aspect of recognition, the method of coal-gangue recognition and grasping feature extraction based on multi-mode deep learning is studied to realize fast coal-gangue recognition suitable for the underground. In the aspect of trajectory planning, the precise positioning and real-time tracking methods of dynamic coal-gangue are studied to realize the adaptive and stable grasping of dynamic coal-gangue by the robot. In the aspect of collaborative sorting, a multi-layer multiple mechanical arms collaborative control model is built to achieve efficient intelligent collaborative sorting of multiple mechanical arms in the complex environment.