Movement model enhanced RSS localisation

This paper introduces a novel method for the localisation of mobile clients using Received Signal Strength (RSS) measurements. Inaccuracies and ambiguities are frequently present in RSS based localisation results from static clients due to inaccuracies in the propagation modelling used to predict the wireless channel between client and sensor/beacon. If the client is mobile, the wireless channel changes from measurement to measurement, providing far more information. The problem then becomes predicting the most likely route taken by the client, rather than a single position. By fusing data from a ray traced propagation model, a map providing data on the physical environment, and a generative path model capturing the dynamic capabilities of the client, we are able to vastly improve the accuracy of RSS based localisation techniques. The proposed technique formulates the geolocation task as an optimisation problem, and applies a particle filter to the task of determining the most likely path followed by the client given a set of RSS measurements and the constraints imposed by the environment and movement model.