Topological mapping and localization for delivery robot

Abstract

The last-mile delivery robot is a hot topic. To enable the robot to move autonomously in the city environment, topological map and navigation are of great significance, since it helps the robot work through obstacle-free spaces with rea- sonable calculation and storage requirements. Thus, it is friendly to indoor and outdoor city environments. This thesis proposes a topological mapping system that builds topological maps suitable for navigation and obstacle avoidance, with the input of point cloud frames collected from 3D Velodyne Lidar. A region-growing algorithm for topo- logical map building is proposed. Two region-growing rules to build convex obstacle-free clusters in a 3D environment are studied and discussed. The sam- pling method to get the initialization seed point for the region growing algo- rithm is studied and discussed. A map projection method to project the 3D topological map to the 2D map is proposed. Keywords: Topological mapping, SLAM.