Abstract: Due to the harsh environment in coal mines underground, mapping methods based on single-source odometry information are prone to issues such as drift, occlusion, and missing semantic features. Existing mainstream localization algorithms applied underground in coal mines often encounter localization errors. To address these issues, this paper proposed an underground autonomous mapping and localization method based on multi-source information fusion. The mapping was performed using the multi-source information fusion-based RTAB-Map algorithm, which significantly reduced mapping drift and improved feature capture ability by fusing point cloud and image data. Precise localization was achieved using the Adaptive Monte Carlo Localization (AMCL) algorithm, which combined LiDAR and motion information and employed particle filtering, pose prediction and resampling to achieve adaptive localization, thereby reducing localization inaccuracies and mapping drift. Simulation and experimental results showed that, compared with a single wheel odometry, the absolute value of the relative error of RTAB-Map mapping based on multi-source information fusion was reduced to within 1%, and the map matching accuracy was higher, improving mapping reliability. Particles using the AMCL algorithm converged rapidly within 2 meters, meeting the localization requirements of autonomous auxiliary transport vehicles.