A Theoretical Framework for Sequential Importance Sampling with Resampling

Abstract

Monte Carlo filters (MCF) can be loosely defined as a set of methods that use Monte Carlo simulation to solve on-line estimation and prediction problems in a dynamic system. Compared with traditional filtering methods, simple, flexible — yet powerful — MCF techniques provide effective means to overcome computational difficulties in dealing with nonlinear dynamic models. One key element of MCF techniques is the recursive use of the importance sampling principle, which leads to the more precise name sequential importance sampling (SIS) for the techniques that are to be the focus of this article.