Improved error estimation in cases of occasional full covariance

Correct track-to-track correlation is a key component of multi-sensor tracking and fusion. Some tracking systems provide full covariance only upon request, making the already difficult problem of multi-sensor track correlation even more difficult. To improve the correctness of correlation results involving track updates from such systems that contain state data without covariance, an estimated covariance can be formed (sometimes from other information contained within the track update). In this paper, an approach is proposed that uses the occasional transmitted full covariance to derive sensor position and compute model parameters. Once determined, the information will be used to create an estimated covariance for subsequent state vector updates that do not have full covariance.