Increasing consistency of particle filter using the classic method and particle swarm algorithm

Particle filter is one of the main filters to estimate of non-linear/ non- Gaussian systems. However, it is inconsistent over time. Since the choice of the proposal distribution and the resampling method are is very important to improve the accuracy and consistency, in this paper, increasing consistency of particle filter is done using improved sampling and resampling. To optimize the sampling step, particle swarm optimization (PSO) has been merged into the importance sampling. PSO causes the particles to move to the high probability region of the posterior before sampling and therefor the distribution of particles is improved. In order to reduce the impact of resampling on the accuracy and consistency, a new resampling approach is proposed. The new resampling method can maintain diversity among the particles and ensure that the resampled particles asymptotically approximate the samples from the posterior probability density function of the true state. The main advantage of the proposed method is that it reduces computational cost. This is because the proposed resampling method is performed on only part of the particles. The validity of the proposed filter is evaluated using simulation. The results show that the proposed method has better performance than classical filters.