RSS-based position estimation in cellular and WLAN networks

Indoor positioning and tracking is gaining more and more interest in the research field, motivated on one hand by the desire to achieve seamless ubiquitous positioning and, on the other hand, by the potential of huge Location Based Services revenues in the future. The received signal strength (RSS)-based positioning has become a key to enter into the field of indoor and urban positioning where Global Positioning System (GPS) often fails to serve. The main objective of this thesis has been to bring contributions in the RSS-based indoor or urban positioning by implementing a generic multi-purpose MATLAB simulator. The motivation of such a work has been partly from literatures and partly from our experimental findings, which shows that RSS-based positioning often requires time consuming measurement campaigns that are site dependent. The built indoor multi-floor MATLAB simulator incorporates adjustable parameters, based on the analysis and statistics from performed measurements from both cellular and WLAN signals. Then accordingly it generates new scenarios and propagation effects and then allows us to output the user positioning accuracy in terms of Root Mean Square Errors. The additional aim is to analyze in more comprehensive way different RSS-based positioning methods. Typically, the current published research work based on RSS-based positioning is limited to fingerprinting methods. Both the fingerprinting and the path-loss method using the Euclidian distance approaches have been implemented in positioning the user in this thesis work. Apart from user estimation the estimation of the AP location and the path-loss parameters are also carried out in this thesis. The thesis results contain two parts: one based on measured data and another based on simulation. The positioning accuracy as estimated via Monte-Carlo simulations with the built simulator is compared with the results based on real measured data. A high correlation between the two has been observed, as expected with the Monte-Carlo accuracy results coming from the simulations being slightly worse than those observed in location with real data, because a larger set of scenarios could be captured with our simulator