An improved probabilistic approach for collaborative multi-robot localization

Cooperative multi-robot localization uses sensor measurements to estimate poses (locations, orientations) of robots in a given environment. Existing approach [3] updates a robot's belief of its pose instantly whenever it detects another robot. However, such instant update does not always contribute positively to the localization process, since both robots' pose estimates could be highly uncertain at the time of detection. In this paper, we develop a new information exchange mechanism for collaborative multi-robot localization so that only necessary information exchange which contribute positively to the localization process are allowed. The proposed mechanism has been implemented and tested using real mobile robots as well as simulations. The results demonstrate significant improvements in localization speed and accuracy when compared to the single mobile robot localization approach.