Date of Graduation

2016

Document Type

Dissertation

Degree Type

PhD

College

Statler College of Engineering and Mineral Resources

Department

Lane Department of Computer Science and Electrical Engineering

Committee Chair

Matthew C Valenti

Committee Co-Chair

Mark V Culp

Committee Member

Yaser P Fallah

Committee Member

Vinod K Kulathumani

Committee Member

Natalia A Schmid

Abstract

Wireless sensor networks (WSNs) are typically formed by a large number of densely deployed, spatially distributed sensors with limited sensing, computing, and communication capabilities that cooperate with each other to achieve a common goal. In this dissertation, we investigate the problem of distributed detection, classification, estimation, and localization in WSNs. In this context, the sensors observe the conditions of their surrounding environment, locally process their noisy observations, and send the processed data to a central entity, known as the fusion center (FC), through parallel communication channels corrupted by fading and additive noise. The FC will then combine the received information from the sensors to make a global inference about the underlying phenomenon, which can be either the detection or classification of a discrete variable or the estimation of a continuous one.;In the domain of distributed detection and classification, we propose a novel scheme that enables the FC to make a multi-hypothesis classification of an underlying hypothesis using only binary detections of spatially distributed sensors. This goal is achieved by exploiting the relationship between the influence fields characterizing different hypotheses and the accumulated noisy versions of local binary decisions as received by the FC, where the influence field of a hypothesis is defined as the spatial region in its surrounding in which it can be sensed using some sensing modality. In the realm of distributed estimation and localization, we make four main contributions: (a) We first formulate a general framework that estimates a vector of parameters associated with a deterministic function using spatially distributed noisy samples of the function for both analog and digital local processing schemes. ( b) We consider the estimation of a scalar, random signal at the FC and derive an optimal power-allocation scheme that assigns the optimal local amplification gains to the sensors performing analog local processing. The objective of this optimized power allocation is to minimize the L 2-norm of the vector of local transmission powers, given a maximum estimation distortion at the FC. We also propose a variant of this scheme that uses a limited-feedback strategy to eliminate the requirement of perfect feedback of the instantaneous channel fading coefficients from the FC to local sensors through infinite-rate, error-free links. ( c) We propose a linear spatial collaboration scheme in which sensors collaborate with each other by sharing their local noisy observations. We derive the optimal set of coefficients used to form linear combinations of the shared noisy observations at local sensors to minimize the total estimation distortion at the FC, given a constraint on the maximum average cumulative transmission power in the entire network. (d) Using a novel performance measure called the estimation outage, we analyze the effects of the spatial randomness of the location of the sensors on the quality and performance of localization algorithms by considering an energy-based source-localization scheme under the assumption that the sensors are positioned according to a uniform clustering process.

Download

Share

COinS