Improving Particle Filter with Better Proposal Distribution base on selection resampling*

When facing strongly nonlinear filtering problems, the performance of particle filtering is often hindered by impoverishment problem. In extreme cases, it may lead to filter divergence or even state estimation failure. In this paper, an improving particle filtering algorithm is proposed to overcome this issue. The proposed algorithm utilizes selection resampling technique to improve the selection of the proposal distribution, thereby addressing impoverishment problem. After obtaining prior samples and likelihood weights, this method uses the likelihood weights for cyclically selecting from the prior samples, thereby obtaining a set of particles that approximate the posterior distribution. During the recursive process, only these particles need to be transmitted, and there is no need to transmit the weights. Simulation experiments have validated the effectiveness of the proposed algorithm.