A Topological Map Merging Method Based On Localization Of Multi-robot SLAM

Simultaneous localization and mapping (SLAM) is usually required for robot system to explore an unknown space without acknowledge of global positioning reference. And for SLAM, multi-robot SLAM could achieve exploration tasks more quickly but with more calculation complexity than single-robot. For maps made by SLAM, occupancy grid map is a useful presentation. Environment exploration of multi-robot system is divided into two parts, establishment of local maps and map merging of local maps. Probabilistic information is also accounted for map merging.In the process of local map establishment, this article proposes a method whose base reference attaches to robot based on EKF-SLAM method. In classic EKF-SLAM method, they usually used global reference for base reference. To avoid the error caused by odometry which is not so accurate, this article proposes a method which used the original position of the robot as base reference.During the process of map merging, the generalized Voronoi diagram (GVD) is used to establish the framework of maps. Accounting for the importance of probabilistic information on GVD, this article firstly extracts the GVD skeleton of maps, then adds the probabilistic information into the GVD skeleton. This articles proposes a new probabilistic generalized Voronoi diagram to build topological maps and increases the accuracy. At the same time, accounting for the nonlinear uncertainty during the relative transformation, this article proposes that this problem could be solved by unscented transformation (UT). This method could improve the accuracy by 2%-3% and reduce the calculation complexity by mostly 80%.The method of this article has these advantages than articles before.1. It reduces error during the process of local map establishment;2. It is effective at finding relative transformation;3. Probabilistic information is accounted for during the extraction of skeleton;4. The uncertainty associated with transformation used to fuse the maps is accounted for.