Abstract: The significant advantages of the recently developed Unscented Kalman Filter (UKF) for nonlinear systems are its easly implementation, better accuracy and moderate computational complexity. However, the UKF has as bad robustness as the extended Kalman filter (EKF) in modelling uncertainty, and is sensitive to the initial conditions. To overcome the limitations of the UKF, an improved UKF based on the theory of Strong Tracking Filters (STF) is developed in the paper. The improved filter could adjust a filtering gain matrix on line by introducing a time-varied fading matrix. Its effectiveness is demonstrated using a simulation example of target tracking. The simulation results show that the improved UKF has good robustness and can rapidly converge.

Key words: Unscented Kalman Filter, Strong Tracking Filter, nonlinear systems, target tracking