Map Fusion of Visual Landmark-based Maps
M. Ballesta, O. Reinoso, A. Gil, L. Paya and M. Julia
40th International Symposium on Robotics (Barcelona (SPAIN)
11-13 March, 2009)
Ed. AER-ATP ISBN:978-84-920933-8-0 - pp. 75-81
In a multi-robot framework, each robot can build its own map independently. In this situation, the fusion of these local maps into a global one may be necessary. In this paper, we focus on the study of this problem, i.e., the Map Fusion problem. This task is performed
in two diŽerent stages. First, the Map Alignment stage, in which the transformation between pairs of local maps is computed. Then, in the Map Merging stage, the data from each local map is fused into a unique map. Particularly, we use the RANSAC method in order to perform the alignment between maps. Then, in the map merging state, we propose a method that considers the uncertainty in the estimate of the maps. In these experiments, we use visual landmark-based maps, which have been built by the robots using the FastSLAM algorithm.