Simultaneous localization and mapping: swarm robot mutual localization and sonar arc bidirectional carving mapping

Abstract

This work primarily aims to study robot swarm global mapping in a static indoor environment. Due to the prerequisite estimation of the robots' own poses, it is upgraded to a simultaneous localization and mapping (SLAM) problem. Five techniques are proposed to solve the SLAM problem, including the extended Kalman filter (EKF)-based mutual localization, sonar arc bidirectional carving mapping, grid-oriented correlation, working robot group substitution, and termination rule. The EKF mutual localization algorithm updates the pose estimates of not only the current robot, but also the landmark-functioned robots. The arc-carving mapping algorithm is to increase the azimuth resolution of sonar readings by using their freespace regions to shrink the possible regions. It is further improved in both accuracy and efficiency by the creative ideas of bidirectional carving, grid-orientedly correlated-arc carving, working robot group substitution, and termination rule. Software simulation and hardware experiment have verified the feasibility of the proposed SLAM philosophy when implemented in a typical medium-cluttered office by a team of three robots. Besides the combined effect, individual algorithm components have also been investigated.