The Unscented Particle Filter

TLDR

The paper proposes a new particle filter based on sequential importance sampling. The algorithm employs a bank of unscented filters to generate the importance proposal distribution, and incorporates resampling and optional MCMC steps. Experimental results show that the filter efficiently uses the latest information, has heavy‑tailed proposals, and outperforms standard particle filtering and other nonlinear filtering methods, in agreement with its theoretical convergence proof.

Abstract

In this paper, we propose a new particle filter based on sequential importance sampling. The algorithm uses a bank of unscented filters to obtain the importance proposal distribution. This proposal has two very nice properties. Firstly, it makes efficient use of the latest available information and, secondly, it can have heavy tails. As a result, we find that the algorithm outperforms standard particle filtering and other nonlinear filtering methods very substantially. This experimental finding is in agreement with the theoretical convergence proof for the algorithm. The algorithm also includes resampling and (possibly) Markov chain Monte Carlo (MCMC) steps.