Air-ground Cooperation For Environment Perception And Loop-closure Detection

Recently, Air-ground mobile robots have attracted much attention in hazardous tasks, such as disaster rescue, military scouting. Due to self-limitation, Unmanned Ground Vehicles (UGVs) have the disadvantages of slow speed, small monitoring range and are adversely impacted by terrain and geological conditions. Compared to UGVs, Unmanned Aerial Vehicles (UAVs) can observe large-scale environment from a high perspective. However, their load capacity and battery life is limited, and they are also very sensitive to airflow disturbance. This paper conducts a study on environment perception and loop-closure detection with a custom-built air-ground multi-robot system.To meet the demand of online loop-closure detection using sequential 2-D laser data, a robust ISW-NDT (Incremental Sliding Window-based NDT) approach is proposed which compares the appearance similarity between two scans by sliding a window with fixed size. In order to ensure the efficiency of ISW-NDT concurrently, we calculate in an incremental way and rotate the NDT grids directly rather than the point clouds. Due to the inherent errors in UAV's pose estimation, a 3D reconstruction algorithm is also adopted to perform 3D map building before UAV's loop-closure detection. Experimental results conducted on both UAV and UGV show the validity and robustness of ISW-NDT method in large-scale outdoor environments. Compared with 3D-NDT approach, the proposed ISW-NDT algorithm is capable of providing superior performance in large scale outdoor environments, which achieves higher recall rate at 100% precision and well ensures the online implementation.In addition, some research is done for solving air-ground cooperative loop-closure detection, which achieves the consistency in data acquisition, model establishment and semantic generation. By using the scene registration approach, a large-scale map can be generated with laser data obtained by UAV and UGV. The differences in viewing angle and field of view are also discussed in this paper to accomplish air-ground loop-closure detection. The performance is tested with different flying height of UAV and different sliding increment, which shows the effectiveness of air-ground loop-closure detection.