Air-Ground Collaborative Mapping Based on Region Matching Under Terrain Constraints

This paper proposes an air-ground collaborative point cloud map fusion and construction framework based on Lidar Odometry And Mapping and Normal Distribution Transform matching, which can still operate normally under low illumination and Global Navigation Satellite System denied conditions. To deal with the localization problem between agents in the absence of initial pose information, this paper designs a matching mechanism based on the front-end and back-end structure. The front-end performs rough matching of the original point cloud, and the back-end achieves fine matching of the map point cloud. A filtering mechanism for spatial overlapping maps is investigated to obtain accurate pose transformation between agents. To demonstrate the feasibility of the design scheme, simulations are conducted in the gazebo environment. The result shows that the error between initial relative pose obtained from mapping and the real setting is under 1.6%.