PSO-Based Refinement of Map Transformation in Multi-Robot SLAM

초록

This paper addresses the problem of estimating map transformation in multi-robot simultaneous localization and mapping (SLAM) with unknown initial relative poses among multiple robots. While inter-robot observation measurements are not available, the map transformation among multiple robots can be estimated by map matching or loop closure algorithms. However, there may be inevitable errors in the estimated map transformation due to the errors in sensor measurements and the limited algorithm performance. To solve the problem, this paper proposes a particle swarm optimization (PSO)-based refinement method which can improve the accuracy of map transformation initially estimated by map matching or loop closure algorithms. The proposed method does not rely on specific algorithms and is applicable to any map matching or loop closure algorithms once they produce initial guesses of map transformation. The proposed method was tested by experiments with real-world datasets and validated by showing improved accuracy of map transformation.