Network localization and navigation: theoretical framework, efficient operation, and security assurance.
Yuan Shen
2014-01-01
Abstract:Reliable and accurate localization of mobile network nodes is a key enabler for numerous emerging applications in the commercial, public safety, and military sectors. Achieving such location awareness by conventional techniques is challenging in harsh environments with limited infrastructures, e.g., indoors, in urban canyons, and on battlefields. This thesis proposes network localization and navigation (NLN), a new paradigm in which mobile nodes exploit both spatial and temporal cooperation for positional inference. To fully understand the cooperation benefits and associated costs, and then to efficiently harness the benefits with minimum costs in practical networks, we have developed a foundation for NLN with contributions in the following three areas. In the first part, we establish a theoretical framework for NLN and characterize the localization performance by spatiotemporal cooperation, showing that NLN significantly improves the reliability and accuracy of location awareness. We introduce the notion of localization information, and demonstrate by Fisher information analysis that such information can be decomposed into the sum of basic building blocks, each associated with a spatial or temporal cooperation link. We also develop a geometric method to illustrate the evolution and coupling of localization information induced by spatiotemporal cooperation. In the second part, we develop robust resource allocation techniques that guarantee the localization accuracy in the presence of parameter uncertainty with minimum energy consumption. We first discover important functional properties of the localization accuracy metrics, based on which the accuracy constraints are transformed into conic forms. We then design an asymptotically optimal second-order cone program (SOCP)-based algorithm for robust resource allocation with a proven convergence rat, as well as near-optimal but more efficient SOCP-based algorithms using relaxation methods. In the third part, we develop an information-theoretic framework for secret-key generation (SKG) using noisy observations of common sources, and as a case study determine the secret-key rate that can be generated from wideband signals in multipath 3 channels. Since the probability distribution of source parameters may be unavailable in many scenarios, we consider the problem from a non-Bayesian perspective and model the parameters as deterministic but unknown. We then propose a new metric called intrinsic information to characterize the secret-key rate, and derive the intrinsic information for wideband channels as a function of network parameters, such as transmission bandwidth and node mobility. The contributions of this work provide insights into the essence of NLN, yielding theoretical benchmarks and practical guidelines valuable for the design and operation of high-accuracy location-aware networks. Thesis Supervisor: Moe Z. Win Title: Professor