Multi-Robot Map Fusion Framework using Heterogeneous Sensors

Fusing local 3D maps generated by individual robots to a globally consistent 3D map is a fundamental challenge in multi-robot missions. With the emergence of diverse sensors, different robots could carry heterogeneous sensors(laser scanners and vision-based sensor). However, there have been few works on real time 3D map fusion with different map data type, especially merging the sparse map with the dense map. In this paper, a general probabilistic framework is proposed to address the integrated map fusion problem, which is independent of sensor types and SLAM algorithms. The multiple data association method between those different types of map provides good insights into merging maps with the different physical and geometrical properties. This paper also provides a time- sequential map merging framework that makes fusing maps from distributed multi-robot system efficiently. The proposed approach is evaluated using mapping data collected from both indoor and mixed indoor-outdoor environments with heterogeneous sensors, which shows its robustness and generality in 3D map fusion for multi-robot mapping missions.