Particle filter technology based on parallel genetic resampling

Abstract

The invention relates to a particle filter method based on the parallel genetic resampling, which comprises the following steps of: (1) sampling around an initial probability distribution to obtain initial particles, and setting an initial weight; (2) sampling the particles around the state transition probability density by the filter estimation of M particles at the time of k-1 to generate M new particles, wherein M is a natural number; (3) respectively carrying out the weight update on the M particles to obtain the weight of each particle; and (4) optimizing the particle group by using the parallel genetic resampling algorithm. The invention improves the particle filter, inhibits the degradation phenomenon, solves the problem of the particle insufficiency caused by simple random resampling, improves the diversity and the adaptability of the particles, and further improves the performance accuracy for the particle filter.

Description

technical field

[0001] The invention relates to the field of nonlinear filter algorithms, in particular to a particle filter method for resampling by using a parallel genetic algorithm. Background technique

[0002] The particle filter algorithm is based on the sequential importance sampling (SIS: Sequentiai Importance Sampling) filtering idea of ​​Bayesian sampling estimation. Hammersley et al. proposed the basic SIS method in the late 1950s, and it was further developed in the 1960s. However, because the above-mentioned research has not solved the problems of particle scarcity and computational constraints, it has not attracted people's attention. It was not until the end of the 1980s that the further improvement of computer computing power and a new SIS-based bootstrap (Bootstrap) nonlinear filter method was proposed by Gordon et al. in 1993, which really became the widespread research and practical application of particle filter algorithms. Foundation.

[0003] Particl...

Images