Efficient and Accurate WLAN Positioning with Weighted Graphs

This paper concerns indoor location determination by using existing WLAN infrastructures and WLAN enabled mobile devices. The location fingerprinting technique performs localization by first constructing a radio map of signal strengths from nearby access points. The radio map is subsequently searched using a classification algorithm to determine a location estimate. This paper addresses two distinct challenges of location fingerprinting incurred by positioning moving users. Firstly, movement affects the positioning accuracy negatively due to increased signal strength fluctuations. Secondly, tracking moving users requires a low-latency overhead which translates into efficient computations to be done on a mobile device with limited capabilities. We present a technique to simultaneously improve the positioning accuracy and computational efficiency. The technique utilizes a weighted graph model of the indoor environment to improve positioning accuracy and computational efficiency by only considering the subset of locations in the radio map that are feasible to reach from a previously estimated position. The technique is general and can be used on top of any existing location system. Our results indicate that we are able to achieve similar dynamic localization accuracy to static localization. Effectively, we are able to counter the adverse effects of added signal fluctuations caused by movement. However, as some of our experiments testify, any location system is fundamentally constrained by the underlying environment. We give pointers to research which allows such problems to be detected early and thereby avoided before deploying a system.