Scalable Robot Localization Using Incremental Map-Matching Techniques

Abstract

Map-matching is an effective approach for robot localization using large-size landmark maps. Existing map-matching techniques typically address batch localization applications where the objective maps are given prior to the map-matching task. In contrast to the existing techniques, this paper addresses real-time localization applications where the maps can be incrementally updated even during the map-matching task. We present a solution that combines an incremental RANSAC map-matching with an ANN landmark database. Experiments using real data show promising results.