A Novel Global Relocalization Method Based on Hierarchical Registration of 3D Point Cloud Map for Mobile Robot

摘要

Indoor service mobile robots need to relocate when they are kidnapped, powered off, or lost in long-term work, thus unable to perform daily tasks. Solving this problem is challenging, especially for 3D maps due to the computational complexity. In order to solve this issue, a novel relocalization algorithm based on hierarchical registration is proposed for a known 3D map in this paper. For 3D point cloud maps, the algorithm obtains multi-layer information in the vertical direction through hierarchical registration at the robot's current position. To obtain the best 3D pose for relocalization, we fuse the poses calculated by the multi-layered point cloud into one and use it as the initial pose of the iterative closest point algorithm. The hierarchical registration based algorithm solves the problem of unknown initial value for registration between two large point clouds, improves the recall rate, and ensures the accuracy of algorithm at the same time. The related relocalization experiments are carried out in the indoor environment and the results verify the effectiveness and robustness of the algorithm.