Research On Distributed Adaptive Target Tracking Algorithms Based On Wireless Sensor Networks

Recently,the distributed signal processing technologies have developed rapidly,which could improve the scalability and flexibility of the network by realizing the localized processing of each sensor in the network and the communication between neighboring sensors.As one of the essential topics in the field of signal processing target tracking has been widely used in many practical engineering fields.Therefore,the distributed adaptive target tracking algorithms based on wireless sensor networks(WSNs)have great development space and high academic research value.Distributed particle filtering algorithms(DPFs)have attracted much attention as one of the most promising methods to solve the problem of target tracking for nonlinear systems in large sensor networks.Recent research results show that the diffusion strategy is easier to implement in a distributed manner,and has been shown more flexibility and robustness in different applications.Therefore,based on the distributed diffusion strategy,this paper mainly studies the distributed target tracking algorithms and diffusion combiners to further improve the performance of distributed target tracking algorithms.The specific research content includes the following three aspects:1)To achieve a tradeoff between tracking performance and computational complexity of the DPFs,a distributed resampling Gaussian particle filtering algorithm(D-ReGPF)for direct tracking based on time delay and Doppler is proposed,which is a DPF algorithm based on heterogeneous networks.By presetting the allowable error parameter,each sensor can adapt its particle number,further,the desired tracking performance can be obtained,with higher flexibility and robustness than the D-GPF.Further,in view of the potential surge of the number of particles in the particle number adaptation algorithm,a variable bin-size scheme is proposed.By limiting the upper and lower limits of the number of particles,and adjusting the bin-size according to the local posterior distribution which changes with time,the surge of the particle number at the early filtering stage is effectively alleviated.2)A Distributed Collaborative Feedback Particle filtering(D-FPF)algorithm is proposed to solve the problem that the traditional particle filtering algorithms based on importance sampling needs to construct proposal distribution and resampling process.The D-FPF algorithm uses the collaborative feedback structure to update each particle of each sensor,and the importance sampling,resampling,or variance calculation which are mandatory in the traditional distributed particle filters are not requisite.The D-FPF could obtain better tracking and particle variance performance than the classical DPF,even with an extremely small number of particles.3)Aiming at the problem that the existing static combiners in the distributed diffusion strategy are not robust to the spatial variation of the signal and noise statistics in the networks,an online(Nonnegative Adaptive Combiners,NAC)optimization with explicitly imposing the nonnegative constraint on the combiners is proposed,and the closed and adaptive solutions of the optimal combiners are given.Simulation results show that compared with the existing static combiner scheme,the NAC method can improve the robustness of the distributed tracking algorithms against the spatial variation of signal and noise statistics in the networks;and compared with the classic adaptive combiner scheme,the NAC method is beneficial to improve the tracking performance of the distributed tracking algorithms,and has lower computational complexity.