Rao-blackwellized particle filter for turn rate estimation

A Rao-Blackwellized particle filter is developed for single target turn rate estimation from noisy position measurements. Missing measurements and false alarms are handled via incorporation of Probabilistic Data Association (PDA) in the particle filter. The performance of the proposed filter is compared to an Extended Kalman Filter (EKF) through Monte Carlo simulations. When assuming perfect data association, it is found that the proposed filter is comparable to the EKF in terms of RMS errors but that it outperforms the EKF in terms of covariance consistency. Simulations including missing measurements and false alarms show that the proposed filter clearly outperforms the EKF with respect to RMSE and track divergence.