Research On The Key Technologies For Passive Source Localization Based On Wireless Sensor Networks

In recent years, passive source localization based on wireless sensor networks(PSL-WSNs) has been paid comprehensive attention from domestic and alien scholars.It is widely used both in civil and military realms, and also it has been the hotspot to bestudied and developed in the world. But there are still many problems to be solved,which are related to the “Association Metrics Estimation (AME)” phase based onprocessing the received signals at the sensors and the “Metrics Fusion Localization(MFL)” phase by using association metrics and WSNs sensor positions. Thisdissertation is mainly concerned with the research on several key technologies to solvethese problems. The author’s major contributions are outlined as follows:(1) When WSNs sensors have different received noise intensities or the wirelesstransmission channel has the shadow fading effect, it is studied that the associationmetrics estimation method for Range Ratios of Arrival (RROA) and the passive sourcelocalization algorithm based on RROA. Firstly, the eigenvector decomposition (EVD)approach is used to estimate the RROA association metrics. The received noise intensityof each sensor can be estimated by performing EVD on the received signal covariancematrix. Secondly, by rotating the array reference point to be at each of the array sensors,a number of covariance matrices are constructed and the EVD approach can be used tocancel the shadow fading effect. The RROA association metrics can be estimatedreliably. At last, the weighted-least-squares (WLS) algorithm based on the RROAassociation metrics is proposed. The proposed approach is robust to channel shadowfading effect and different received noise intensities.(2) An efficient iterative algorithm is proposed for passive source localizationbased on TDOA and GROA. It exploits the Broyden-Fletcher-Goldfarb-Shanno (BFGS)Quasi-Newton method to solve nonlinear equations at the source location under theadditive measurement error. The proposed method can overcome the problem that theHessian matrix may be non-positive and the basic Newton method cannot converge tothe global minimum. Compared with two-step WLS method, the proposed approach canachieve the same accuracy and bias with lower computational complexity when SNR ishigh, especially it can achieves better accuracy and smaller bias at lower SNR.Simulation results show that with a good initial guess to begin with, the proposedestimator converges to the true solution and achieves the CRLB accuracy for bothnear-field and far-field sources. It can apply to the actual environment due to itsreal-time property and good robust performance. (3) It is proposed that a hybrid closed-form solution algorithm based on TDOA andGain Ratios of Arrival (GROA) to improve multiple disjoint sources localizationaccuracy with erroneous sensor positions. The algorithm jointly estimates the unknownsources and sensor positions, and then takes the advantage that the TDOA and GROAfrom different sources have the same sensor position displacements to enhance theposition accuracy. It is also derived that the Cramér-Rao lower bound (CRLB) ofmultiple source localization using both TDOA and GROA. Simulations show that theproposed solution is able to reach the CRLB accuracy very well, and the localizationaccuracy improvements contributed by GROA measurements are significant.(4) Two methods are proposed to reduce the bias of the well-known algebraicclosed-form solution for source localization by using both TDOA and GROA. Firstly, itstarts by deriving the bias of the source location estimate from the closed-form solution.And then, two methods called BiasSub and BiasRed are developed to reduce the bias.The BiasSub method directly subtracts the expected bias from the closed-form solution.The BiasRed method augments the equation error formulation and imposes a constraintto improve the source location estimate. Analysis shows that both methods reduce thebias considerably for distant source when the noise is Gaussian and small. The BiasRedmethod is able to lower the bias to the same level as the maximum likelihood estimator.(5) In the presence of sensor position and velocity errors, it is studied that theproblem of simultaneously locating multiple disjoint sources and refining erroneoussensor positions and velocities using TDOA and Frequency Differences of Arrival(FDOA). The proposed method has the existing WLS method based on TDOA andFDOA improved and a new algebraic closed-form solution is given. The new solutiontakes both the source positions and velocities and the sensor positions and velocities asthe targets to be estimated. All of them can achieve the CRLB accuracy very well. Thetheoretical derivation is corroborated by simulations.