State estimation of radar tracking system using a robust adaptive unscented Kalman filter

In this work, for a two-dimensional radar tracking system, a new implementation of the robust adaptive unscented Kalman filter is investigated. This robust approach attempts to eliminate the effects of faults associated with measurement models, and varying noise covariances to improve the target tracking performance. An adaptive threshold value is used to identify the need for adapting the noise covariances rather than a fixed threshold value. A forgetting factor and a weighted mix of the most recent and previous estimate data are employed to update the process and measurement noise covariances. By calculating the root mean square error using Monte Carlo simulations under various circumstances, the efficiency of the proposed approach is examined. It has been found that the proposed approach can successfully handles system uncertainties imposed by variable noise covariance and measurement outliers.