Unscented Kalman filtering for additive noise case: augmented vs. non-augmented

This paper concerns the unscented Kalman filtering (UKF) for the nonlinear dynamic systems with additive process and measurement noises. We find that under some condition, the basic difference between them is that the augmented UKF draws sigma set only once within a filtering recursion while the non-augmented UKF has to redraw a new set of sigma points to incorporate the effect of additive process noise. This difference generally favors the augmented UKF. The analyses are supported by a representative example.