On Multi-robot Map Fusion by Inter-robot Observations

This paper addresses the problem of aligning and fusing maps built by multiple robots. The proposed method for solving the multi-robot map alignment problem. relies on inter-robot observations to seed the alignment processing and find a transformation between the map reference frames. The method enables one to join maps from robots with or without initial correspondence. However, the poses of each robot during an inter-robot observation need to be synchronized. In this work the method is applied onto Collaborative Smoothing and Mapping (C-SAM), a smoothing approach for merging maps that are created by different robots independently or in teams. The algorithm is proven to work in two different experiments showing the usefulness of the algorithm. The experiments show that alignment can be conducted using only inter-robot observation in both unguided terrain as well as in terrain with many false observations. The key contribution of this work is an algorithm for solving the association problem and eliminating false observations when doing multi-robot map alignment using inter-robot observations during a rendezvous.