A heuristic for sigma set selection of UKF

In this paper we present a higher order moment-matching algorithm for computing the distribution parameters of nonlinear transformation random variables. The new algorithm has two distinct aspects compared to the standard Unscented Kalman Filter (UKF). First, the sigma points are computed in two steps using the covariance matrix and higher-order moments. Second, the associated weights are positive numbers in the interval [0, 1]. The performance of the new algorithm is illustrated by simulation. Results show improvement in accuracy in comparison to the traditional UKF.