Experimental Deployment of Particle Filters in WiFi Networks

Location tracking in wireless networks has many applications, including enhanced network performance. In this work we investigate the experimental use of "particle filter" techniques as applied to the mobile device tracking problem in WiFi networks. Particle filters are well suited to the non-Gaussian and biased error conditions characteristic of signals found in most WiFi networks, as well as allowing for seamless data fusion with intermittent GPS availability and prior map information. Our experimental results show that particle filter tracking can indeed deliver significant performance gains. In some physical regions of our test area, factors of three improvement in location accuracy - relative to optimal unfiltered WiFi positioning - is found. We show how these gains are achieved at a relatively small particle number of 300, meaning real-time implementation of the particle filter can be easily achieved.