Cooperative simultaneous localisation and mapping using independent Rao–Blackwellised filters

摘要

This study focuses on an approach to the multi-robot simultaneous localisation and mapping (SLAM) problem. The authors consider that a team of robots start the map building process from known initial poses and move through the environment observing a set of landmarks. In particular, the authors consider that the robots are equipped with vision sensors and are capable to compute the relative distance to natural visual landmarks. In the proposed approach, each robot computes an own global map and integrates its own observations in it. In addition, each robot is also capable of integrating the measurements obtained by other robots in his own map, thus cooperating to compute a global map using the information provided by all the robots in the team. In this way, each robot in the team builds a different global map. The approach presented here differs from others in the sense that each robot maintains an independent SLAM filter, and still each robot can introduce the observations of other robots in his own map. The authors present a set of experimental results obtained in a simulated environment that validate the proposed approach.