In this dissertation, we address the issue of collaborative information processing for diffusive source parameter estimation using wireless sensor networks (WSNs) capable of sensing in dispersive medium/environment, from signal processing perspective. We begin the dissertation by focusing on the mathematical formulation of a special diffusion phenomenon, i.e., an underwater oil spill, along with statistical algorithms for meaningful analysis of sensor data leading to efficient estimation of desired parameters of interest. The objective is to obtain an analytical solution to the problem, rather than using non-model based sophisticated numerical techniques. We tried to make the physical diffusion model as much appropriate as possible, while maintaining some pragmatic and reasonable assumptions for the simplicity of exposition and analytical derivation. The dissertation studies both source localization and tracking for static and moving diffusive sources respectively. For static diffusive source localization, we investigate two parametric estimation techniques based on the maximum-likelihood (ML) and the best linear unbiased estimator (BLUE) for a special case of our obtained physical dispersion model. We prove the consistency and asymptotic normality of the obtained ML solution when the number of sensor nodes and samples approach infinity, and derive the Cramer-Rao lower bound (CRLB) on its performance. In case of a moving diffusive source, we propose a particle filter (PF) based target tracking scheme for moving diffusive source, and analytically derive the posterior Cramer-Rao lower bound (PCRLB) for the moving source state estimates as a theoretical performance bound. Further, we explore nonparametric, machine learning based estimation technique for diffusive source parameter estimation using Dirichlet process mixture model (DPMM). Since real data are often complicated, no parametric model is suitable. As an alternative, we exploit the rich tools of nonparametric Bayesian methods, in particular the DPMM, which provides us with a flexible and data-driven estimation process. We propose DPMM based static diffusive source localization algorithm and provide analytical proof of convergence. The proposed algorithm is also extended to the scenario when multiple diffusive sources of same kind are present in the diffusive field of interest. Efficient power allocation can play an important role in extending the lifetime of a resource constrained WSN. Resource-constrained WSNs rely on collaborative signal and information processing for efficient handling of large volumes of data collected by the sensor nodes. In this dissertation, the problem of collaborative information processing for sequential parameter estimation in a WSN is formulated in a cooperative game-theoretic framework, which addresses the issue of fair resource allocation for estimation task at the Fusion center (FC). The framework allows addressing either resource allocation or commitment for information processing as solutions of cooperative games with underlying theoretical justifications. Different solution concepts found in cooperative games, namely, the Shapley function and Nash bargaining are used to enforce certain kinds of fairness among the nodes in a WSN.
Statistical signal processing, Wireless sensor network, Detection and Estimation, Source localization, Source tracking, Machine learning, Resource allocation, Cognitive radios, Game theory, Parametric estimation, Non-parametric estimation
Level of Degree
Electrical and Computer Engineering
First Committee Member (Chair)
Second Committee Member
Third Committee Member
Hakim, Kamrul. "Collaborative Information Processing in Wireless Sensor Networks for Diffusive Source Estimation." (2013). http://digitalrepository.unm.edu/ece_etds/110