Fingerprint calibrated centroid and scalar product correlation RSSI positioning in large environments

Abstract

The design and result of a fingerprint calibrated weighted centroid localization (FCWC) algorithm is outlined, discussed and evaluated against the UJIIndoorLoc data set. The algorithm is suitable for situations where positions of access points are not known but reference RSSI measurements can be taken in an indoor environment, as it is typically done for fingerprint algorithms. The algorithm comprises a calibration step which is based on the recorded RSSI fingerprints: By performing reverse localization, “Virtual positions” of the access points are calculated. The virtual positions do not necessarily need to closely match the real positions of the access points. The proposed FCWC algorithm uses weighted centroid, which is analogue to center of gravity. The weight function is based on the given RSSI value between the reference point and the WiFi access point, and is manually calibrated by adjusting 2 global parameters. After the virtual positions have been assigned, position estimates can be calculated using RSSI readings of a rover. This is again performed by applying weighted centroid, the position is obtained via the weighted vector sum of the virtual positions of the access points. For comparison, also a scalar product correlation fingerprinting (SPCF) algorithm has been implemented. In the paper, the background of the FCWC and SPCF algorithm, the weight function, and the obtained results are discussed.