Robust radio localization with FLIP

Abstract

Radio fingerprinting based localization relies on comparing observations to a database of reference radio fingerprint point data. For complex buildings these databases can be very large. As mobile devices have limitations in storage capacity, working memory, processing speed, and power usage, making an on-terminal system that works well even in large installations is challenging. Moreover, the reported Received Signal Strength Indication (RSSI) scale often differ between devices, so that naive approaches for fingerprint similarity easily can fail to produce reliable results. In this paper, FLexible Indoor Position (FLIP) is presented to address these issues. It provides efficient device independent positioning even in complex buildings, while also taking inhomogeneous transmitter power levels and radio map irregularities into consideration. Despite plain accuracy not being the main goal of FLIP, when it was evaluated on the raw UJIIndoorLoc WiFi database it yielded a median positioning error of 4.7 m (and above 93 % floor level/building identification success rate), which is competitive to other significantly more computation intense approaches. In commercial applications with dedicated iBeacon infrastructures, FLIP routinely reaches median errors below 2 m.