Advancements in information technology and communication systems enabled the development of a wide variety of location based applications and services. The global navigation satellite systems are among the fundamental localization solutions. In harsh environments (e.g., urban canyons and indoor areas), these solutions do not provide a good accuracy or even become unavailable. In order to offer accurate and ubiquitous localization solutions, wireless communication systems have been considered, where several location dependent parameters of the transmitted signals can be measured and exploited (e.g., the time-of-arrival (ToA), the received signal strength (RSS)). In this work, the topic of wireless localization is explored from a statistical signal processing perspective with a focus on two axes. The first axis is cooperative localization applied to ad-hoc networks, where the nodes perform pair-wise ranging measurements (i.e., ToA or RSS) between each other in order to simultaneously estimate their positions. The unique solvability conditions are studied based on the two approaches of graph rigidity and semidefinite programming, and the identifiability conditions are derived. Location estimation solutions are considered with a focus on probabilistic estimation and its application in Markov random fields using the nonparametric belief propagation (NBP) algorithm. The second axis is mobile terminals tracking based on RSS measurements. These measurements are affected by a shadowing phenomenon. The improvement brought by the knowledge of the shadowing maps to the position estimation accuracy is studied. The classical solution for obtaining these maps is fingerprinting, which can be costly in time and effort. Solutions are developed to overcome these difficulties. Several solutions are proposed and investigated via Monte Carlo simulations in different deployment and application scenarios, and several theoretical and practical results are derived.
