RANSAC for Exchanging Maps in Multipath Assisted Positioning

Kurzfassung

The classical approach to the multipath problem in positioning algorithms is to try to mitigate the influence of multipath components (sMPCs) on the line-of-sight (LoS) path. Over the last years, a contrary approach has emerged with multipath assisted positioning, where MPCs of terrestrial signals are regarded as LoS signals from virtual transmitters. Thus, the spatial information in MPCs is exploited for localization. The locations of the physical and virtual transmitters can be estimated together with the user position using simultaneous localization and mapping (SLAM). To decrease the possibly long convergence time of SLAM, maps of transmitters can be exchanged among users. The information in a map can then be fused with the user observations. However, in the general case, the users do not know their starting locations and/or headings, and therefore are in their own local coordinate systems. When maps of transmitters are exchanged, the relative rotation and translation between the coordinate systems of a user and a received map need to be estimated. Within this paper, we propose a variant of the random sample consensus (RANSAC) algorithm for this estimation as it is very robust against outliers. We show by simulations in an indoor scenario that RANSAC increases the accuracy significantly.