Applications of Unscented Kalman Filtering to State Estimation

For nonlinear system filtering problem,Kalman filter cannot be used.Though the extended Kalman filter can be used in nonlinear system,but the outcome of the filter is a partial estimation and the system is not very robust.To get better state estimation,a filtering method,named Unscented Kalman filter,is introduced.The idea of the filter is: the UKF generates better estimates of mean and the covariance of the state,leading to higher accuracy based on the unscented transformation.Compared with the filtering estimate results given from a time series state estimation example based on Mackey-Glass Model,the simulation shows that UKF not only leads to more accurate results than EKF,but also can avoid computing Jacobi matrix.