On the efficacy of WiFi indoor positioning in a practical setting

We implement two popular WiFi, fingerprinting based indoor tracking mechanisms, namely the k-nearest neighbours and probabilistic positioning methods. Both mechanisms are evaluated in the context of an indoor position-tracking tablet application, following an investigation to determine optimal working parameters. Our results indicate that even after significant optimisation, both fingerprinting algorithms are highly sensitive to the location of the access points and do not produce finely grained location results. Although in this case the results are accurate enough for our purposes, factors such as the effect of natural body obstruction of the user as well as the location of the access points used in fingerprinting must be considered carefully if more accuracy is required.