Resolution of track association ambiguities using expectation-maximization algorithm

Space object observations cannot always be unambiguously associated to catalogued objects. Multiple objects, typically orbiting in close proximity to each other, can pass an assignment criterion and can therefore become likely candidates for the association. If the state uncertainty of the catalogued objects is large, this ambiguity can only be resolved when new observations arrive again. In order to screen for conjunctions between space objects, the most likely associations have to be identified at all time steps.
This works studies an approach where multiple objects with unresolved associations are grouped and then their orbits are determined simultaneously. The measurements are loosely associated first with a weighting factor depending on the statistical distance to the current reference trajectory. The factors are iteratively updated using an expectation-maximization approach and repeated orbit determination and association runs. The method is presented along with practicality considerations/adjustments for an operational implementation.