One can do better than the unscented Kalman filter for multistatic tracking

The unscented Kalman filter (UKF) is a useful alternative to the extended Kalman filter (EKF) for tracking with nonlinear dynamics models and when the measurements are nonlinear functions of the target state. In this paper, the problem of tracking using monostatic and bistatic measurements is considered. Previous work has demonstrated that the UKF does not always handle measurement nonlinearities in challenging monostatic scenarios better than the EKF, let alone considering more complicated bistatic scenarios. This paper reviews previous work showing that the UKF is one among many numeric integration-based filters. It is demonstrated that a general cubature Kalman filter outperforms the extended Kalman filter for multistatic tracking when cubature points of a sufficiently high order are used. Additionally, cubature-based measurement conversion for track initiation is discussed, and the posterior Cramér-Rao lower bound for basic multistatic tracker assessment is derived.