Research On The Key Techniques Of Passive Target Detection And Localization In Underwater Acoustic Sensor Network

In recent years,with the increasing importance of the ocean on the international stage,the traditional techniques of signal processing with the use of single node can not meet the requirements of exploiting and making full use of the ocean resources and safeguarding the maritime rights.By taking advantage of the information fusion techniques,the underwater acoustic sensor network(UASN)could significantly improve the sensing performance for the long distance target by spatially placing a number of sensors in the interest of area.Based on the two key techniques of the detection fusion and cooperation localization in the UASN,this thesis aims to improve the target sensing performance.The researches of the thesis include distributed detection method using diffusion Kalman filtering,the direction-ofarrival-only(DOA-only)target localization method,the optimal node placement strategy for the DOA-only target localization,and the DOA-only target motion analysis(DOA-TMA)estimate method.The research results of this dissertation have important theoretical significance and practical application value for the target detection and localization in the UASN.The main results and achievements are summarized as follow:1.Distributed detection method of Gauss–Markov signals was proposed in adaptive networks,where nodes made individual decisions by exchanging information with their neighbors and no fusion center was used.Relying on the connection between the log-likelihood ratio(LLR)and the innovations process,a LLR was constructed to perform hypothesis testing by making use of the innovations estimated by the diffusion Kalman filter.The detection performance of the algorithm was analyzed by deriving approximate expressions for the probability of false alarm and detection.The mean and mean-square value of the test statistic,and the probability of detection and receiver operating characteristics of the algorithm were demonstrated by way of Monte Carlo simulations,where a good agreement between theoretical and simulation results was observed.Simulations also confirmed the improved detection performance of the proposed algorithm over non-cooperative detection and showed that the performance of the algorithm converged to that of the centralized scheme.2.To improve the localization performance of least squares(LS)estimation,the Minmax k-means least-square(Minmax k-means LS)method was proposed to figure out and remove the DOA data which were far more beyond their true values.In addition,an improved cross-bearing localization method was proposed based on the Parzen window to reduce the impact of the invalid sensors on the accuracy of target localization in UASN.In this method,the probability of target location was estimated by the Parzen window according to the distribution characteristics of all intersection points,and the target location was selected as the point corresponding to the maximum value of probability.Because of the nonlinear and multi-peak features of the probability distribution,the standard particle swarm optimization method was adopted to solve the problem.Moreover,simulation results indicated that the improved cross-bearing localization method based on the Parzen window avoided effectively the influence of the invalid anchors on the performance of localization,and had better accuracy and robustness compared with LS estimator in the complex underwater environment.3.By resorting to the Cramér-Rao Lower Bound(CRLB)and the Fisher information matrix(FIM),the optimal sensor placement strategies were analysed to further improve the target location performance in UASN.The optimal goal was to maximize the determinant of the FIM so that the area of uncertainty information ellipse(in 2 dimension)or the volume of the uncertainty information ellipsoid(in 3 dimension)was minimized.Starting with the assumption of the constant measurement noise variance,the optimal target-sensor geometry was obtained in 2 dimension.Moreover,for the assumption of the distant-dependent measurement noise,the expression of the FIM was derived for the target localization based on direction-of-arrival(DOA)in 3 dimension.Besides,the optimal sensor placement strategy was analyzed for the surface sensor networks.In addition,considering the optimal average localization performance,the optimal sensor placement was studied in 2 dimension.The simulation results verified the validity of the theoretical analyze using the LS estimator.4.New estimators were proposed for the problem of target motion analysis(TMA)using DOA measurements collected by a single moving sensor whose locations were subject to errors.The existing DOA-TMA algorithms suffered from performance degradation caused by sensor location errors.Thus,those algorithms,namely,the LS,the bias-compensated LS and the weighted instrumental variable(WIV)estimator,were considered by taking into account the presence of sensor location errors.It was analytically shown that the WIV estimator was no longer asymptotically unbiased due to the nonvanishing correlation between the instrumental variable matrix and the LS noise vector as a result of the sensor location noise.To alleviate such a bias problem,a new bias-compensated WIV estimator was then developed and analytically proved to be asymptotically efficient under the small noise assumption.Extensive simulation examples were presented to corroborate the performance advantage of the proposed bias-compensated WIV estimator over the LS,the bias-compensated LS,the conventional and redeveloped WIV estimators.The proposed bias-compensated WIV estimator was shown to exhibit the best performance among the estimators by producing the smallest bias and a root-mean squared error closest to the CRLB.5.Based on the theoretical analysis and computer simulation,lake experiments were conducted to evaluate the target localization performances in the UASN using three localization methods based on the Parzen window,the Minmax k-means LS and LS estimation,respectively.The experiment results showed that the target localization methods resorting to UASN provided the good performance.In addition,the experiment results suggested that the localization methods in UASN can potentially be applied to engineering applications.