A solution for the Map Fusion Problem in Cooperative Visual SLAM
This paper presents an approach of map fusion in a multirobot SLAM context. Particularly, we have a team of robots exploring and building local maps of the environment. Each robot builds its own map without knowledge of other robots’ positions and observations. The study of the map fusion problem is tackled in two stages: map alignment and map merging. The first one consists in obtaining the transformation between the robots’ reference system. In the second one, we combine the information of the local maps into a single one. Initially, the robots may not share a common area of the environment or may not begin the map building simultaneously. This situation is also analyzed. The robots build 3D landmark-based maps using the FastSLAM algorithm.