Marginalized particle filter for dependent Gaussian noise processes

The theory and the applications of the marginalized particle filter (MPF) have attracted much research attention during the last decade. However, the existing MPF framework does not cover dependent process and measurement noises. This dependency is perhaps more common in practice than is acknowledged in the literature. In this article, we propose a general framework for MPF, covering both cases of dependent and independent noises. As a consequence, MPF with independent noises is a special case of this general framework. The treatment of dependency always provides `extra´ information to the state estimation tasks. This beneficial effect is shown through a numerical example.