Multi-Robot Mapping Based on 3D Maps Integration

Abstract

An unknown environment could be mapped more efficiently by a group of robots than a single robot. The time reduction due to parallelization is crucial in complex area mapping. There are two general solutions used in the multi-robot mapping. In the first one, robots exchange raw data from sensors. The second approach assumes that each robot creates a local map independently that is exchanged with other robots and integrated. In this chapter, we present a 3D maps integration algorithm that utilizes overlapping regions in the feature-based alignment process. The algorithm does not need any initial guess about the transformation between local maps. However, for successful integration, maps need to have a common area. We showed that the implemented method is effective in various environments. The approach has been verified in experiments with wheeled mobile robots and using public datasets with octree-based maps.