Corrected Differential Evolution Particle Filter for Nonlinear Filtering

摘要

Particle filter is the most successful nonlinear filter for nonlinear filtering. However its resampling process has the critical problem existing is the particle impoverishment problem. In this letter, we propose a new corrected differential evolution particle filter for solving this problem. In this algorithm, the particles sampling from the importance distribution are regarded as the initial population of the Corrected Differential Evolution (CDE) algorithm, and the corresponding weights as the fitness functions. The optimal particles are obtained by the process of the CDE algorithm. Experiment results indicate that the proposed method relieves the particle degeneracy and impoverishment and improves the estimation precision.