Global localization of multirobot formations using ceiling vision SLAM strategy

Global localization is an important matter in multirobot formations, but the issue has not been sufficiently studied yet. In this paper, we successfully extend the single robot ceiling vision SLAM to multirobot formations for addressing global localization problem. Each robot is equipped with a monocular camera that looks upward to the ceiling. The monocular camera system used for ceiling observation appears to be more convenient than other active sensors such as laser and panoramic camera. A public global map shared by every robot is developed for positioning update. Two global localization strategies are proposed. The first strategy is to globally localize one robot only and then localize the others based on the relative poses amongst the robots. The second strategy is to globally localize all the robots simultaneously. The former requires less computational resource, and the later exhibits better localization performance. A feature-based matching approach is utilized to calculate the relative poses amongst the robots. Simulation experiments are finally performed to demonstrate the effectiveness of the proposed approach.