The Study Of Quadrature Particle Filtering And Its Application

At present, the study of Large-Scale passive sensor system and its related key technology has become more and more attention by domestic and foreign scholars. To the Large-Scale passive sensor system, the key and difficult problem of the data processing is that the nonlinear and non-Gaussian observation data processing with shortage of observation information, high data loss rate. Therefore, combined with the existed nonlinear and non-Gaussian filtering algorithm, a novel Gaussian sum quadrature particle filter(GSQPF) based on Gauss-Hermite quadrature is proposed, and puts forward a Auxiliary Gaussian sum quadrature particle filter(AGSQPF), which take characteristics of target into consideration. On this basis, a novel parallel computing algorithm is proposed.To the Large-Scale, nonlinear and non-Gaussian observation data processing, a novel Gaussian sum quadrature particle filter(GSQPF) based on Gauss-Hermite quadrature is proposed. According to the advantage of Gauss-Hermite quadrature points in the nonlinear approximation and the diversity of quadrature points, we introduce a set of quadrature point probability densities to approximate the important density function, the filtering and prediction densities are approximated as finite Gaussian mixtures, which can effectively improve the performance. The simulations show that the presented filter can outperform both Gaussian sum particle filter(GSPF) and quadrature particle filter(QPF).In order to improve the processing of observation data with aperiodic and long interval, based on GSQPF, considering interval of observation information, observation value and target velocity, and take those characteristics of target into the importance density function, this paper puts forward a Auxiliary Gaussian sum quadrature particle filter(AGSQPF), and effectively improve the diversity and accuracy of particles. For the nonlinear and non-Gaussian model, the simulation results show that, the proposed AGSQPF is estimated to be significantly better than the performance of GSQPF, accurately tracking the passive target.For a large amount of observation data processing and high-requirement of communication in the Large-Scale passive sensor system, this paper presents a parallelized Gaussian sum quadrature particle filter. In the algorithm for parallel processing, the particles and weights of GSQPF are update in the subsystem, each subsystem is responsible for one subnetwork and the relevant variables of the neighboring subsystems are communicated, and independence of the subsystems, improve the efficiency of data processing effectively. The experimental results show that the proposed algorithm meets the need of Large-Scale passive sensor system and accurately tracking the target.