Simultaneously merging multi-robot grid maps at different resolutions

Abstract

For an unknown GPS-denied environment, multiple robots can explore various areas of the same environment and build local maps for these areas. Accordingly, it is essential to merge these local maps of different robots into one global map. For grid map merging, this paper proposes an effective approach, which can simultaneously merge multi-robot grid maps at different resolutions. First, we present a scaling pairwise method for merging grid map pairs. Usually, a selected grid map is initialized as the seed map, which is then sequentially updated by performing scaling pairwise merging between itself and other input grid maps. Afterwards, map augmentation and feature fusion strategy is proposed to alternatively integrate two local maps into one initial global map. To address the accumulated error, a coarse-to-fine method is introduced to refine the initial global map and obtain the accurate global map. Experimental results conducted on real robot data sets demonstrate that the proposed approach can merge multiple grid maps simultaneously at different resolutions with good performances.