Abstract:

For the adjustable parameter selection problem of -κ- in the unscented Kalman filter(UKF), through the study of the impact of the different κ for filtering, the method based on the step prediction information of the measurement, which is an online adjustment of the UKF, is presented. Based on the prediction information of measurement in every filtering time, the filtering parameter is selected, which is optimal and can realize the on-line adjustment. Numerical simulations show that the adjustment UKF based on the step prediction information of the measurement tracks the real state better than the traditional UKF.