Mobile robot localization using multiple geomagnetic field sensors

Abstract

This paper proposes a novel approach to substantially improve the performance of the conventional vector field SLAM (simultaneous localization and mapping) by using multiple geomagnetic field sensors. The main problem of the conventional vector field SLAM is the assumption of known data association. If a robot has a high uncertainty of the pose estimate, the probability of data association failure increases when the robot's pose is located in a wrong cell. To deal with this problem, we propose a multi-sensor approach utilizing multiple geomagnetic field sensors. As the multi-sensor approach updates nodes of one or more cells simultaneously, the probability of data association failure significantly decreases. The proposed multi-sensor-based localization is implemented based on a Rao-Blackwellized particle filter (RBPF) with geomagnetic field sensors. Simulation results demonstrate that the proposed approach greatly improves the performance of the vector field SLAM compared to the conventional approach with a single sensor.