An improved implicit equal-weights particle filter

Abstract

The implicit equal-weights particle filter (IEWPF) algorithm is developed to solve the problem of curse of dimensionality. Based on the implicit particle filter, the algorithm could estimate the high-dimensional state with the disturbance of the Gaussian white noise. In this paper, a new algorithm is proposed for extending the application of the IEWPF algorithm. New algorithm could switch the IEWPF algorithm and implicit particle filter depending on the ensemble size and dimension. Besides, the colored noise problem is converted into a Gaussian white noise problem in the new method. The improved algorithm is tested in the linear model, the Lorenz96 model and the target movement model based on the MSC coordinate system, compared with the primary IEWPF algorithm. The results is presented that the improved algorithm has better performance than the original IEWPF algorithm.