A Robust Approach for Maintenance and Refactoring of Indoor Radio Maps

Abstract

WiFi Fingerprinting techniques are widely used for indoor localization needs, due to better accuracy guarantees. However, the accuracy is limited by the freshness of the radio map, which is used for localization. Over time this radio map might be incompatible due to the changes in signal strength, a consequence of the dynamic nature of the environment. Therefore, repeated radio map calibration becomes a necessity. Currently radio maps are either manually calibrated, use additional infrastructure or complex algorithms to account for the radio map errors. There is therefore a need for a methodology to update the radio map with minimal additional overhead. This paper proposes a crowdsourcing approach which uses data collected from users of the localization system to maintain the freshness of the radio map. This approach leverages inertial sensors present in commonly used handheld devices, like mobile phones and tablets using which, a trajectory of the path of each user is computed. This trajectory is then coupled with the knowledge of the physical map under consideration to get the real time Received Signal Strength Indicator (RSSI) values at the reference points (RPs). A novel cluster based RSSI propagation policy is proposed where the real time RSSI values obtained are propagated within the clusters. Extensive experiments of our localization system implemented in a real indoor environment shows that this approach maintains radio map freshness while keeping the cost of update low and without need for extra infrastructure.

This work has been funded in part by DST(India) grant DyNo. 100/IFD/2764/2012-2013.