Simplified Marginalized Particle Filtering for Tracking Multimodal Posteriors

In this paper we introduce a simplified marginalized particle filtering method for dynamic systems with nonlinear and conditionally linear states with the marginal posteriors of the nonlinear states being multimodal. We propose a particle filter that employs Rao-Blackwellization by only one Kalman filter per mode for marginalizing the unknown linear states of the system. The validity of the method is tested through computer simulations by applying it to a tracking problem with only two static sensors measuring received signal strength. The results show that the new particle filter performs similarly as that based on traditional Rao-Blackwellization while at the same time it requires much less computations.