The International Unmanned Systems Innovation Competition Ground Unmanned Vehicle Challenge (UGVC) is a historic international event. According to the rules of the race, the team independently design and build a complete ground unmanned platform. This paper mainly focuses on the design and implementation of unmanned platform autonomous navigation system based on lidar. First of all, according to the overall design requirements of the platform, complete the design of the hardware and software system for autonomous navigation; Then, downsample the dense point cloud data collected by the lidar, reducing the density of the point cloud. Then filter the ground point, removing invalid point cloud; Finally, initialize and update the occupied grid map based on the Binary Bayesian filtering method, at the same time, in order to prevent the difference between the continuous data frames of lidar causing the instability of the map, the historical map feedback step is added, which effectively improves the stability of real-time map construction. Simulation experiments and actual tests show that, the system establishes a complete grid map, and the smapping time of point cloud data per frame is only about 0.06s, and the number of unstable pixels is reduced to 45%, abling to effectively complete autonomous navigation tasks in competitions.