Gaussian sum unscented Kalman filter with adaptive scaling parameters

The paper deals with state estimation of nonlinear non-Gaussian discrete dynamic systems by a bank of unscented Kalman filters. The stress is laid on an adaptive choice of a scaling parameter of the unscented Kalman filters to increase estimate quality over the standard Gaussian sum unscented Kalman filter. Several optimization criteria for adapting the scaling parameter are proposed and discussed and to apply the scaling parameter adaptation within the Gaussian sum framework, three adaptation procedures are proposed. Performance of the proposed estimation methods is analyzed through the root mean square error and non-credibility index in a numerical example.