| |
| In this paper, we investigate the relation between the filtering accuracy and the sampling number drawn by Particle Filtering (PF) based on the confidence interval theory, and a adaptive sampling numbers PF (APF) is proposed. The conventional PF (CPF) algorithm keeps constant sampling numbers during the entirely filtering time, and its filtering precision depends severely on the sampling number, thus CPF bears a larger computational load. Compared with CPF, the proposed APF achieves almost same filtering precision with adaptive variable sampling number, the higher filtering accuracy can be guaranteed by a pre-determined confidence coefficient. Great deals of Monte-Carlo simulations are performed and show that the average sampling number and the computational load of the proposed algorithms are reduced greatly. |
|
| Keywords:Particle Filtering; Adaptive Particle Filtering ; Sampling Number. |
|
