Advanced Wireless Localisation Methods Dealing with Incomplete Measurements

Positioning techniques have become an essential part of modern engineering, and the improvement in computing devices brings great potential for more advanced and complicated algorithms. This thesis first studies the existing radio signal based positioning techniques and then presents three developed methods in the sense of dealing with incomplete data. Firstly, on the basis of received signal strength (RSS) location fingerprinting techniques, the Kriging interpolation methods are applied to generate complete fingerprint databases of denser reference locations from sparse or incomplete data sets, as a solution of reducing the workload and cost of offline data collection. Secondly, with incomplete knowledge of shadowing correlation, a new approach of Bayesian inference on RSS based multiple target localisation is proposed taking advantage of the inverse Wishart conjugate prior. The MCMC method (Metropolis-within-Gibbs) and the maximum a posterior (MAP) / maximum likelihood (ML) method are then considered to produce target location estimates. Thirdly, a new information fusion approach is developed for the time difference of arrival (TDOF) and frequency difference of arrival (FDOA) based dual-satellite geolocation system, as a solution to the unknown time and frequency offsets. All proposed methods are studied and validated through simulations. Result analyses and future work directions are discussed.