A Self-adaptive Scaling Parameter Selection Algorithm for the Unscented Kalman Filter

In practice, the Unscented Kalman Filter based on the scaled unscented transformation is usually used as a default form with a set of constant scaling parameters. Sometimes this form cannot obtain the ideal performance and is lack of robustness, especially when it is applied to some highly nonlinear models. This paper, therefore, proposes a new method by modifying the main scaling parameter at every step using a self-adaptive algorithm. Simulation results demonstrate that it is more accurate than the default UKF and easier to implement than the augmented UKF.