Marginalised iterated unscented Kalman filter
Lubin Chang, Baiqing Hu, Guobin Chang, A. Li · IET Control Theory and Applications · 2012
In this study, the authors investigate the role of iteration in the unscented Kalman filter (UKF) with additive measurement noise and propose a novel filter referred to as the marginalised iterated unscented Kalman filter (MIUKF). In each iteration of the MIUKF, the new sigma points are regenerated and propagated through the same measurement update strategy as the UKF. In order to guarantee the state to be statistically independent with the measurement noise, the state variables are augmented with the measurement noise. Since the measurement noise is additive, the measurement function is conditionally linear of the augmented state, then the marginalised unscented transformation is investigated to reduce the computational burden. Compared with the traditional iterated UKF, the MIUKF is more rigorous in terms of efficiency and accuracy. The simulation results agree well with the theoretical analyses.