Multi-robot FastSLAM based on consensus of observed data

This paper is concerned with a fast alorithm for the multi-robot SLAM (Simultaneous Localization and Mapping) in the uncertain envrionment without any host robots or base stations. In the single robot case, the well-known FastSLAM is sucessful in making a proper trade-off between computational efficiency and estimation accuracy by combining the extended Kalman filter and particle filter. We propose a novel fast multi-robot SLAM algorithm by extending the FastSLAM with the aid of a distributed particle filtering technique. The distributed particle filter incorporates a consensus mechanism on the observed data from all robots which enables an efficient and accurate state estimation under the practical situation where the sensor chracteristics of all robots are available as prior information. The effectiveness of the proposed algorithm is verified through a numerical experiment and simulations.