Node Density and Quality of Estimation for Infrastructure-based Indoor Geolocation Using Time of Arrival

Infrastructure-based indoor geolocation systems utilizing a regular grid arrangement of sensors are being investigated for many applications in indoor wireless networks. One of the factors affecting the Quality of Estimation (i.e. location estimation accuracy) of these systems is node density. In this dissertation we study the effects of node density on indoor geolocation systems based on time of arrival (TOA). The effects of node density on the performance of various indoor communication networks (e.g. wireless LANs) in the presence of realistic indoor radio propagation models has been analyzed and reported in the literature. However, we have noted the lack of an equivalent analysis on the effects of node density on the performance of infrastructure-based indoor geolocation systems. The goal of this dissertation is to address this knowledge gap. Due to the complicated behavior of the indoor radio channel, the relationship between the node density and Quality of Estimation (QoE) is not straightforward. Specifically, QoE depends on factors such as the bandwidth used to make the TOA-based distance measurements, the existence of undetected direct path (UDP) conditions, and coverage. In this dissertation, we characterize these dependencies. We begin by characterizing the Quality of Estimation for closest-neighbor (CN), least-squares (LS) and weighted LS techniques in the presence of different node densities and a distance measurement error (DME) model based on ray tracing (RT) that was recently proposed in the literature. Then, we propose a new indoor geolocation algorithm, Closest Neighbor with TOA Grid (CN-TOAG), characterize its performance and show that it outperforms the existing techniques. We also propose an extension to this algorithm, known as Coverage Map Search (CMS) that allows it to be used in suboptimal coverage conditions (which we refer to as partial coverage conditions) that may prevent other TOA-based geolocation techniques from being used. We treat the partial coverage case by defining coverage probabilities and relating them to the average radius of coverage and dimensions of the indoor area. Next, we characterize the effects of node density on the performance of the CN-TOAG algorithm using a DME model based on UWB measurements, and show that node density and partial coverage are intimately linked together. Since this second DME model also allows for the effects of UDP conditions (which affect the quality of the link or QoL), we also characterize the effects of varying UDP conditions on the performance. Finally, we conclude the dissertation by presenting an analysis of fundamental performance bounds for infrastructure-based indoor geolocation, specifically focusing on the Cramer-Rao Lower Bound (CRLB)