In fingerprint-based indoor localization systems, it is necessary to keep the fingerprint database up-to-date since the differences between actual RSS signatures in real world and the fingerprints in the database may lead to severe performance degradation in terms of localization accuracy. The goal of this paper is two folds, by updating the fingerprint database in a crowdsourced manner, (1) increasing the localization accuracy and (2) minimizing the cost for site surveying. However, updating fingerprint database remains challenging since correct localizations are prerequisites to correct updates. To address this problem, the proposed update technique includes a novel path detection algorithm, which finds a set of fingerprints representing user’s path through a whole sequence of RSS vectors. By updating initial database with the signatures, any environmental changes are applied to the fingerprint database to keep the database up-to-date. Our extensive sets of experiments report up to 10% more accurate localization estimations comparing to clustering-based localization techniques and furthermore, the location error rate decreases as being updated.

Keywords: indoor localization, WiFi fingerprint database update, AP relocations