Particle filtering for stochastic hybrid systems

Abstract

The standard way of applying particle filtering to hybrid systems is to make use of hybrid particles, where each particle consists of two components, one assuming Euclidean values, and the other assuming discrete mode values. This paper develops a novel particle filter for a discrete-time stochastic hybrid system. The novelty lies in the use of the exact Bayesian equations for the conditional mode probabilities given the observations. Therefore particles are needed for the Euclidean valued state component only. The novel particle filter is referred to as the interacting multiple model (IMM) particle filter because it has a switching/interaction step which is of the same form as the switching/interaction step of the IMM algorithm. Through Monte Carlo simulations, it is shown that the IMM particle filter has significant advantage over the standard particle filter, in particular for situations where conditional switching rate or conditional mode probabilities have small values.