A novel adaptive unscented Kalman filter

Abstract

For solving the problem that the conventional unscented Kalman filter (UKF) declines in accuracy and further diverges when the system's noise statistics are unknown and time-varying, an adaptive UKF is proposed based on moving window and random weighting methods. The moving window estimation defined in linear system is generalized to the nonlinear filter — UKF. The noise statistics are calculated by applying the moving window estimation and then the weights on each window are adjusted by utilizing the random weighting method. The proposed algorithm has the ability to estimate and adjust the noise statistics online, making the best of the moving window and the random weighting methods. Simulation and comparison analysis demonstrate that the proposed adaptive UKF performs much better than the standard UKF under the condition that system's noise statistics are unknown and time-varying.