I. Introduction
Long after their first appearance [1], particle filters still represent an active area of research. Because of their generality and simplicity, they have become a topic of constantly growing interest, development, and numerous applications. Particle filters, together with unscented Kalman filters [2] and moving horizon estimators [3], can provide a general framework for state estimation in nonlinear and non-Gaussian dynamical systems. Moreover, provided that the number of particles is high enough, they can easily out perform other estimation methods. Their main drawback is the required computational load. The necessary computational time for an accurate result is, in most cases, prohibitive for real-time applications and this can seriously limit the applicability of particle filters.