Track-to-track fusion with missing information: Empirical study: Tracking a variety of target types

Copious research has been done comparing various track-to-track methods via their calculated (theoretical) covariance. What has not been studied is how these various track-to-track methods actually perform (i.e., their empirical variance) against a variety of target types. This paper will compare best of breed, centralized measurement fusion, track to track (TTT) without memory (with and without cross correlation), information matrix fusion, and TTT with memory to see which method empirically has the lowest covariance against a variety of targets. These targets are a discrete white noise acceleration, constant velocity, and constant acceleration This study will further investigate how a track-to-track fusion center handles missing information, particularly estimating the process noise values (σ_α) for the sensor trackers and what happens to the relative quality of the estimates as the variance assumed at the fusion center is mismatched with those from the sensor trackers.