Enhanced Diversity based Particle Filter using Local Linear Embedding

Abstract

The advantage of particle filter algorithm is to estimate the state of non-linear and non-Gaussian problems. However, there are still some shortcomings in particle filter methods, such as particle degradation and loss of particle diversity, which make the state estimation error. In this paper, a particle filter algorithm based on local linear embedding is proposed. Firstly, the state of high-dimensional particles is reduced to one-dimensional Euclidean space, and the neighborhood linear relationship between particles is maintained. Secondly, according to the size of particle weights, the particles are divided into two groups, including the reserved group with larger particle weights and the adjusted group with smaller particle weights. The particles with smaller weights are searched for the adjustment scheme and mapped back to the high-dimensional space for particle state adjustment. Then, the high-dimensional particles with larger weights are mapped back to the high-After resampling, the newly generated two groups of particles are merged into the next iteration calculation. Finally, simulation experiments are carried out to verify the effectiveness of the proposed method by comparing the experimental results of fifteen existing methods with those of the proposed method. The improved method proposed in this paper is applicable to all resampling methods, and can effectively solve the problems of particle degradation and loss of diversity by adjusting the particle state.