State-Space Modeling And Target Detection And Tracking With Distributed Underwater Acoustic Network

Detection and tracking of underwater targets is of great significance in the applica-tions of at-sea defense,ocean exploitation,and marine scientific research,and is one of the important issues in the research of marine information technology.The traditional target detection technology with only one sonar equipment cannot achieve large-scale and high-resolution monitoring.Although some active detection methods can obtain a higher signal-to-noise ratio,it is difficult to achieve long-term and covert monitoring because of high energy transmission.With the development of sensor technology and the theory state-space model filtering,it is possible to deploy distributed acoustic sensors to detect and track targets passively in the covered area of interest.Aiming to solve the problem of large-scale and long-term surveillance,the distributed network topology and communication strategy are studied in this thesis,and the passive detection and tracking of underwater targets is investigated under the Bayesian state-space model framework.Firstly,the state-space modeling of the distributed underwater acoustic sensor network is developed based on the underwater acoustic physics process.With the state-space model,the Bayesian sequential filtering algorithm is studied to provide a theoretical framework for the thesis.The random finite set(RFS),which can model the existence and motion of the target simultaneously,is introduced to replace the vector state in the traditional state-space model,and the RFS Bayesian filter is developed for joint detection and tracking of acoustic targets.Secondly,the network topology and communication strategy are studied.The node deployment strategy in the detection network with data fusion is proposed to improve the detection probability.Taking the statistical model of underwater acoustic signal into accoun-t,the node position selection scheme is derived by minimizing the detection error probability,and simulations in a shallow water environment are carried out to verify the global optimali-ty of the scheme.Facing the problem of limited communication capability in the distributed underwater acoustic sensor network,an ordering transmission method for distributed detec-tion is proposed,which can save communication overhead with optimal Bayesian detection guaranteed.The energy efficiency of the proposed method is shown by simulations in net-works with different topologies.In order to ensure the long-term synchronization of sensor nodes,an improved tiny-sync synchronization algorithm is designed for the underwater en-vironment with long propagation latency.Compared with the previous algorithms,the proposed method has low complexity,low data storage,and takes less time to implement the synchronization procedureThirdly,the joint detection and tracking of targets is studied in the distributed un-derwater acoustic network composed of sensor nodes with single-hydrophone.Although the measurement information that can be obtained by a single-hydrophone node is relatively limited,these nodes have the advantages of small size,low cost,and easy deployment.In this thesis,the Bernoulli filter based on distributed energy measurement is proposed.The spatial sampling of the sound field energy is carried out in the surveillance region.Based on the acoustic propagation model,the statistical characteristics of the energy measurement of the complex sound pressure field are analyzed,and the likelihood functions with the presence of target and white noises are derived correspondingly.With the statistical characteristics of the acoustic energy measurements,the Bernoulli filter is applied to detect and track target.The performance of the proposed method is verified by both simulations and experimental data processing.Lastly,the joint target detection and tracking problem is studied in the distributed underwater acoustic network composed of multiple arrays.The Bernoulli filter is developed to detect and track target in conjunction of beamforming and Matched Field Processing(MFP).The Bernoulli filter based on multi-array bearnforming is first proposed.The di-rection of arrival(DOA)and intensity information in the beampower are used to construct the measurement RFS,and the likelihood functions are derived correspondingly,leading to the DOA-based and beam-based Bernoulli filter for track extraction.A hybrid method combining the above two methods is also proposed.The beam-based method improves the localization accuracy;however,because the main beam becomes broader toward end-fire,it is likely to incorrectly locate the target in the end-fire directions.If there are more than three arrays,such a problem can be eliminated by using the geometric relationship of the arrays.Otherwise,the hybrid method can compensate for this shortcoming caused by the beam-based method.The simulation results wwith three arrays show that the beam-based method and the hybrid method have better tracking performance than the DOA-based method.The processing results of experimental data collected by two arrays in a shallow water environ-ment show that the hybrid method has better tracking performance.The multi-array MFP for target detection and tracking is also studied.The multi-array MFP ambiguity volume functions are derived by maximizing the likelihood function.For passive detection and track-ing of targets,the Bernoulli filter is then applied by choosing multiple peaks of the MFP ambiguity volume to form the measurement set.Simulations and experimental results show that,compared with the conventional MFP localization algorithm,the proposed Bernoulli filter based on MFP can improve the target tracking performance.