Distributed Multi-robot Localization Based on Mutual Path Detection

This paper presents a new algorithm for the problem of multi-robot localization in a known environment. The approach is based on the mutual refinement by robots of their beliefs about the global poses, whenever they detect each other's paths. In contrast to existing approaches in the field the detection of robots (e.g. by cameras) by each other is not required any more. The only requirement is the ability of robots to communicate with each other. In our approach the robots try to detect the paths of other robots by comparing their own perception sensor information to that “seen” by other robots. This way the relative poses of the robots can be determined, which in turn enables the exchange of the beliefs about the global poses. Another advantage of our algorithm is that it is not computationally more complex than the conventional single robot algorithms. The results obtained by simulations show a substantial reduction of the localization time in comparison to single robot approaches.