Abstract: Aiming at the problems of complex structure, short service life, susceptibility to obstruction by obstacles, and poor real-time performance due to large mapping data volume of the mechanical laser radar commonly used in unmanned systems in underground coal mine roadways, a multi-view point cloud fusion mapping method for solid-state LiDAR in underground coal mine roadways based on voxel filtering is proposed, with solid-state LiDAR as the core sensor. Firstly, a distributed multi-solid-state LiDAR fusion perception SLAM system is designed, and a unified coordinate system is adopted to process multi-view laser point cloud data through joint calibration. Based on the voxel filtering method, multi-view point clouds are lightened and fused to reduce redundant points in the overlapping field of view and obtain unified, low-data-volume point cloud data of the surrounding environment. Secondly, based on the fused point cloud, a local map is constructed by combining dynamic key frame selection and line-plane feature extraction, and the global map is spliced using the pose transformation matrix. Then, the global map is optimized using the voxel filtering method to further reduce the map data volume. Subsequently, joint simulation experiments are carried out using the Gazebo and Rviz visualization modules in ROS. The simulation results show that the multi-view point cloud fusion mapping method based on voxel filtering for solid-state LiDAR can effectively reduce the map point cloud data volume while fully restoring the environmental spatial structure. Finally, according to the actual environment of coal mine roadway, the simulation underground roadway was built, and the multi-sensor joint calibration and multi view point cloud fusion mapping test were carried out by using the driverless test vehicle. The test results show that: compared with the traditional mechanical lidar based on LOAM（LiDAR Odometry and Mapping）, LIO-SAM（Lidar-Inertial Odometry via Smoothing and Mapping） and LeGO-LOAM-SC（Light weight and Ground-Optimized Lidar Odometry and Mapping with Scan Context） For the mapping method, the map data volume of the proposed mapping method in this paper is reduced by 89.25%, 10.12% and 58.14% respectively, and the update time of single frame map is reduced by 86.2%, 29.7% and 72.0% respectively, which has higher accuracy of spatial structure restoration, and has the feasibility of practical application in coal mine roadway scene.