UCT CORRELATION USING THE BHATTACHARYYA DIVERGENCE

In this paper, we build on recent work to further investigate the use of information theoretic criteria to solve the track-to-track association problem, in which we have multiple uncorrelated tracks (UCTs) to be tested for correlation against a given set of tracks given at a different time instance. Both the tracks and the UCTs are uncertain and are probabilistically described using multivariate normal distributions. This allows for a closed-form solution, based on the unscented transform and on the Bhattacharyya information divergence for Gaussian distributions. We first mathematically describe the approach and make connections to the covariance-based track association (CBTA) technique existing in the literature. We then summarize recent and current results comparing the two approaches in Cartesian and equinoctial element coordinates. We show that while the Bhattacharyya approach is superior to CBTA in Cartesian coordinates, the two approaches are nearly identical in performance when the analysis is done in equinoctial elements with CBTA outperforming the Bhattacharyya metric in certain orbital regimes for the more ambiguous correlation problems. We conclude the paper with a summary of the main outcomes of this paper.