One-Way Observation-based Cooperative Robot Mapping

Abstract

This paper addresses the problem of cooperative mapping in multi-robot SLAM (simultaneous localization and mapping) with unknown initial relative poses among robots. If mutual observation measurements between robots are available when the robots encounter each other, the map transformation between robots for cooperative mapping can be easily obtained. However, the mutual observation measurements require too strict assumptions such as a rendezvous and encounters, which is not practical in real multi-robot systems. Moreover, the mutual observation measurements may contain inevitable errors due to the errors in sensors. To relax the assumption and correct the errors, this paper proposes a one-way observation-based technique for cooperative mapping. The proposed technique was tested with datasets obtained by real experiments with two mobile robots. Each robot was equipped with a sensor fusion system to observe other robots. The test results showed that the proposed technique worked well and improved the accuracy of the cooperatively produced map.