A novel quadrature particle filtering based on fuzzy c-means clustering

作者:

Highlights:

摘要

In this paper, a novel particle filter (PF) which we refer to as the quadrature particle filter (QPF) based on fuzzy c-means clustering is proposed. In the proposed algorithm, a set of quadrature point probability densities are designed to approximate the predicted and posterior probability density functions (pdf) of the quadrature particle filter as a Gaussian. It is different from the Gaussian particle filter that uses the prior distribution as the proposal distribution, the proposal distribution of the QPF is approximated by a set of modified quadrature point probability densities, which can effectively enhance the diversity of samples and improve the performance of the QPF. Moreover, the fuzzy membership degrees provided by a modified version of fuzzy c-means clustering algorithm are used to substitute the weights of the particles, and the quadrature point weights are adaptively estimated based on the weighting exponent and the particle weights. Finally, experiment results show the proposed algorithms have advantages over the conventional methods, namely, the unscented Kalman filter(UKF), quadrature Kalman filter(QKF), particle filter(PF), unscented particle filter(UPF) and Gaussian particle filter(GPF), to solve nonlinear non-Gaussian filtering problems. Especially, to the target tracking in Aperiodic Sparseness Sampling Environment, the performance of the quadrature particle filter is much better than those of other nonlinear filtering approaches.

论文关键词:Sequential importance sampling,Quadrature particle filter,Gauss–Hermite quadrature,Fuzzy c-means clustering,Aperiodic sparseness sampling

论文评审过程:Received 18 December 2015, Revised 14 May 2016, Accepted 17 May 2016, Available online 19 May 2016, Version of Record 18 June 2016.

论文官网地址:https://doi.org/10.1016/j.knosys.2016.05.034