Pose-graph based 3D map fusion with distributed robot system

Abstract

Map fusion is an essential and challenging issue in multi-robot simultaneous localization and mapping (SLAM), especially for 3D SLAM applications. In this paper, we propose an efficient pose-graph based approach to fuse several 3D maps built by distributed mobile robotic system. Several robots, each equipped with a Kinect sensor, cooperate to map the operated environment; each of them just scans a part of the environment. A target detection algorithm is presented to help the robot detect other robots during the map building. Key information will be transferred from the detected robot to the detecting robot to build a link between these two robots. And at last, G2O optimization library is applied on the connected pose-graphs to build an optimized map. Experiments are finally performed to demonstrate the effectiveness of the proposed approach.