High-Precision Passive Source Localization In Wireless Sensor Array Networks

Wireless sensor array networks (WSANs) have been widely applied in military and national defense due to the advantages including high performance and good concealment etc. The main typical applications involve low altitude, ground and underwater passive target localization and tracking. In real applications, a WSAN has poor computation capability, constrained bandwidth and limited storage of power. Thus, how to achieve high performance for passive target localization using limited resources is one of the most challenging problems in WSANs. Based on overseas and domestic research status about passive source localization in WSANs, this dissertation solves the passive source localization problem in a class of resource constrained WSAN including the following aspects:direction-of-arrival (DOA) estimation, bearing-only target localization, target localization using array covariance matrices and target localization based on coherent processing among arrays etc. The main content of this thesis is summarized as follows:1. A new low computational complexity narrowband DOA estimation method is proposed based on regularized sparse variable projection (SVP) optimization. While the existing sparse signal representation (SSR) methods mostly utilize joint-sparse signal representation (JSSR) to reconstruct signal sources themselves, the new method simplifies the JSSR problem as a single sparse signal representation (SSSR) problem by reconstructing a sparse indicative vector (SIV) instead of estimating all signal sources. The SVP function regularized by l0norm is NP-hard and therefore we use lp<1norm instead. Then we formulate the SVP optimization as an unconstrained optimization problem, which can be solved iteratively. The SVP optimization algorithm is proved to satisfy the local convergence property and the convergence rate is super-linear when0<p<1.2. A new subband information fusion (SIF) method is proposed for wideband DOA estimation and increases the degrees of freedom for array design. The problem of wideband DOA estimation using SIF method is to jointly utilize all the frequency bin information to recover a single sparse indicative vector (SIV). Since the spatial ambiguity is related to frequency bins, we show that the spatial aliasing can be reduced by combining all the frequency components and present a sufficient condition to achieve the maximum number of degree of freedom for array design.3. A new location-penalized maximum likelihood (LPML) estimator is presented for bear-ing only target localization. We develop a new penalized maximum likelihood cost function by transforming the variables of target position and bearings. Then, the LPML estimator is obtained by maximizing the likelihood cost function. To compare the performance of TBML and LPML estimators, we analyze the Cramer-Rao lower bound (CRLB) of two estimators and show that the bound of LPML estimator is lower than TBML estimator for the location estimate. Extensive sim-ulations are performed. The new LPML algorithm demonstrates the superior performance in all simulations and field experiments.4. A new direct multi-target localization method in WSANs is proposed using sparse indica-tive vector (SIV) recovery based on joint sparse representation of array covariance matrices (SIVR-JSRACM). After introducing a binary SIV, the proposed approach begins with finding a joint sparse representation of all the array matrices. After penalizing the sparsity of SIV by an lp≤1-norm, the localization problem is formulated as a unconstrained optimization and can be solved iteratively. The proposed method does not required the prior information of the number of sources and has properties such as super-resolution, no sensitivity to initialization, robustness to noise, no need for synchronizing the arrays etc. Moreover, the computational complexity of the proposed algorithm for each iteration is bounded at O(L2), where L is the number of potential sources.5. A new CRLB is derived for passive source localization based on angles-of-arrival (AOAs), gain ratios of arrival (GROAs) and time differences of arrival (TDOAs) in a wireless sensor array network. We consider the array networks with perfect coherence, partial coherence or zero co-herence among arrays. The AOA-GROA-TDOA (AGT) has lower bound than other well-known methods if the coherence is perfect. The derived CRLB using AOA-GROA-TDOA (AGT) is re-duced to the one using AOA-GROA if no coherence exists across the arrays and lower than the CRLB using AOA-only. When the coherence is considered, the CRLB using AGT measurements is consistently lower than the other known bounds using AOA-only, TDOA-only and AOA-TDOA.