Vision-based multi-robot simultaneous localization and mapping

In this paper we present a vision-based approach for the multi-robot Simultaneous Localization and Mapping (SLAM) problem. We study the case of a team of robots equipped with a single camera and collaborating in the same worksite. We propose to calculate the location of the robots by using a collection of sparse views of the planar surface on which these robots are moving. The camera motions are estimated using inter-image homographies computed from the matching of overhead transformed views. Results of map generated from the estimated robot locations are presented.