Abstract: Aiming at the harsh environment in the coal mine, a multifunctional integrated mine AGV is designed independently to replace some manual operation procedures and realize inspection and auxiliary rescue functions in emergency situations. In order to overcome the special environment under the mine, the explosion-proof design and the rationality of the mechanical structure are emphatically considered to ensure that the closed part of the vehicle and the external necessary sensors meet the explosion-proof or flameproof standards. In order to further expand the autonomous operation function, a SLAM navigation scheme based on radar-inertial fusion is designed, and an improved loop detection laser mapping algorithm is proposed for test and test. The real-time optimization of the algorithm is carried out by LiDAR with different line numbers. The effect is tested through the indoor ground environment mapping test and the real tunnel mapping test under the mine successively. Among them, the above ground test movement distance is 60m, the maximum sudden fluctuation in the Z-axis movement direction of the vehicle is less than 0.06m, the underground test movement distance is 200m, and the maximum sudden fluctuation in the upper and lower parts of the vehicle is less than 0.25m under the condition of the bumpy ground under the mine. The key part of the Z-axis regularity is only 3.87%, and the X-axis direction is about 0.37%. It is proved that the robot has high stability of motion and excellent robustness of map construction, and can successfully complete inspection and rescue tasks, which is of great significance for the breakthrough of key technologies in the AGV neighborhood in the mine.