Abstract: Autonomous navigation is the key technological foundation for achieving autonomous driving, and the use of Simultaneous Localization and Mapping (SLAM) technology is currently the mainstream solution. However, in harsh environments such as insufficient lighting, narrow tunnels, and dust interference underground in coal mines, traditional SLAM technology based on a single sensor of vision or LiDAR is prone to failure, resulting in weak adaptability and insufficient reliability. A SLAM method for autonomous navigation of underground unmanned driving based on multi-source mileage fusion is proposed for the harsh environment and complex working conditions of coal mines, and verified through experiments. Firstly, research has been conducted on the core issue of autonomous navigation - environmental reconstruction based on multi-source mileage fusion. The multi-sensor fusion SLAM algorithm RTAB-Map (Real-Time Appearance-Based Mapping) is used to achieve environmental reconstruction, and to improve its reconstruction accuracy in coal mine environments, the laser odometer, visual odometer, wheel odometer, and IMU (Inertial Measurement Unit) are tightly coupled through Kalman filtering, replacing the RTAB-Map in-situ pose estimation information with higher accuracy coupling results; Subsequently, a new multi-point navigation strategy was proposed to address the problems of repetitive and cumbersome single point navigation, as well as reduced autonomy. Based on existing single point navigation methods, multi-point navigation was achieved by constructing intermediate nodes that can store and publish pose information in chronological order; Finally, an unmanned driving test platform was set up in the laboratory and simulated underground tunnel scenes for autonomous navigation experiments, including environmental reconstruction experiments and multi-point autonomous navigation experiments. The results showed that the RTAB-Map algorithm, which integrates multiple sources of mileage, can achieve high-precision environmental reconstruction. The absolute value of the maximum absolute error for the multi measurement size of the simulated scene map is only 11.9 cm; The multi-point autonomous navigation method achieves continuous navigation of at least 3 target points for autonomous vehicles, avoiding repeated single point navigation and effectively improving the efficiency of autonomous navigation operation. The maximum longitudinal error of trajectory fitting is only 0.25 m, and the maximum lateral error is only 0.19 m, with good navigation accuracy.