A Comparison of Similarity Measures for Localization with Passive RFID Fingerprints

We present a novel approach to self-localization with passive RFID fingerprints using vector space similarity measures and weighted k-nearest neighbors (WKNN). Embedded in a particle filter, our technique provides robust and accurate position estimates. This is shown through various experiments with a mobile robot. The formulation of the observation model guarantees applicability to any type of off-the-shelf RFID reader which counts RFID tag detections. Different similarity measures are investigated, and a new measure is proposed.