Map-merging using maximal empty rectangles in a multi-robot SLAM process

Abstract

A map-merging algorithm is proposed where reduced element maps are applied instead of grid maps and the maximal empty rectangles are applied as their features. Simultaneous localization and mapping (SLAM) refer to the process where a robot provides the environment map without any knowledge about its own position. Due to error accumulation, required time, saving lives and reasons alike, applying a single robot in the SLAM process is not justified. In such applications, many robots are to be applied in the SLAM process in a parallel sense. The map-merging process is one of the challenging topics in a multi-robot simultaneous localization and mapping process in producing a global map of the environment. In this study, a centralized algorithm is introduced for map-merging based on maximal empty rectangles as the features of local maps without any knowledge about robots’ initial or relative positions. Three examples and one experiment are applied in validating the performance of this newly proposed algorithm. The obtained results indicate that this algorithm can merge local maps with small overlapping areas in relation to the whole map, subject to multiple sources of error due to the difference in scales, diversity of agents applied and measurement noise.