Practical server-side WiFi-based indoor localization: Addressing cardinality & outlier challenges for improved occupancy estimation

Abstract Server-side WiFi-based indoor localization offers a compelling approach for passive occupancy estimation (i.e., without requiring active participation by client devices, such as smartphones carried by visitors), but is known to suffer from median error of 6-8 meters. By analyzing the characteristics of an operationally-deployed, WiFi-based passive indoor location system, based on the classical RADAR algorithm, we identify and tackle 2 practical challenges for accurate individual device localization. The first challenge is the low-cardinality issue, whereby only the associated AP generates sufficiently frequent RSSI reports, causing a client to experience large localization error due to the absence of sufficient measurements from all nearby APs. The second is the outlier resolution issue, whereby clients physically located outside the fingerprinted region but attached to the WiFi network are localized erroneously to a fingerprinted landmark. To tackle the low-cardinality challenge, we identify the stationary-period of devices, and augment the device’s live AP-reported RSSI readings, during such stationary periods, with the useful but ‘stale’ values reported by neighbouring APs. To eliminate extraneous outlier devices, we apply a threshold-based filtering strategy, where the RSSI thresholds for all interior points are derived using a combination of a weighted path-loss propagation model and the Voronoi tessellation of the fingerprinting map. In addition, to overcome intermittent false positives/negatives in localization or subsequent occupancy estimation, we apply two additional techniques: (a) temporal smoothing of location estimates over a time period, and (b) identification and removal of static devices. We experimentally evaluate these combined set of techniques on 3 different indoor work, collaboration & residential spaces, and show how these techniques improve the robustness of location tracking, which subsequently translates into an approx. 80+% reduction in the overall occupancy estimation error.