AMCL based map fusion for multi-robot SLAM with heterogenous sensors

This paper proposes an efficient adaptive Monte Carlo Localization (AMCL) based approach to align the occupancy grid maps built by a multi-robot system. Map alignment plays an important role for the map fusion of multi-robot simultaneous localization and mapping (SLAM), especially for the SLAM with heterogenous sensors. Two robots equipped with a laser and Kinect respectively are executing FastSLAM 2.0 in the same environment but at different starting point; the motion and measurement information is recorded with time-stamps. To merge the maps built by different robots, one robot is first relocated in the map built by the other robot by using the recorded motion sequences and measurement information. With the relocation result, the transformation matrix between the two different maps is the calculated; the matrix is further used as the initial relative pose information for ICP process to obtain precise alignment result. Experiments are finally performed to demonstrate the effectiveness of the proposed approach. Index Terms - AMCL; map fusion; multi-robot SLAM; heterogenous sensors.