Position estimation using RSS measurements with unknown measurement model parameters

The availability and performance of satellite-based navigation systems are the weakest in urban areas and indoor spaces, where the user density would be high. In these environments alternative low-cost positioning techniques are needed. This thesis considers positioning using received signal strength (RSS) measurements of terrestrial wireless networks. No prior knowledge of the considered wireless networks is assumed in this thesis, but only a simplified statistical path loss model for signal propagation. The model parameters are estimated for each base station of the network separately using pre-collected learning data. The method is based on Bayesian estimation theory that characterizes the precision of the parameter estimates, which is an essential feature. Three Bayesian position estimation methods are proposed in this thesis. Two versions of each are compared: one uses point estimates for the model parameters and assumes them to be accurate, whereas the other takes the finite parameter precisions into account. Real-data tests are accomplished using cellular networks in outdoor and wireless local area networks (WLAN) in indoor spaces. The tests indicate that taking the finite parameter precisions into account improves positioning accuracy and especially makes error estimation more realistic. Furthermore, RSS-based methods outperform the method that uses only the list of observed base stations and no RSS information. The advantages of parametric methods compared with the k-nearest neighbour method, which can be regarded as the state-of-the-art positioning method, are also shown