A Two-Stage Pose Graph Optimization Method for Multi-Robot Distributed Cooperative SLAM

With the rapid development of robotics research, functional requirements of robots have become more and more complex, and it is difficult for single robot to satisfy the actual working requirements. However, multi-robot collaborative SLAM systems suffer from high computational resource requirements. This is due to the fact that the operation of a multi-robot collaborative SLAM system needs to process the pose information collected by its own sensors and the relative pose information of other robots obtained through inter-robot communication. In this context, the multi-robot distributed cooperative SLAM system is proposed. However, the distributed cooperative technology still faces the problems of difficult to guarantee the consistency of the poses in different coordinates and complicated pose optimization solution when the poses of different robots are inputted into a single robot system for processing. Therefore, how to efficiently complete the pose optimization of multi-robot cooperative SLAM has become a hot topic in recent years. In this paper, we propose a two-stage pose graph optimization method in a multi-robot distributed collaborative SLAM system to reduce the computational effort of the pose graph optimization process in multi-robot collaborative SLAM system. It is shown that our two-stage pose graph optimization method significantly improves the computational speed of pose graph optimization while maintaining similar accuracy as the existing common multi-robot collaborative SLAM pose graph optimization methods.