Combining Gaussian mixture estimates from uncertain sources

This paper examines the exact solution for combining multiple Gaussian mixture (GM) estimates of a state vector, methods for order-reduction of the resulting GM, estimating the likelihood of each source GM, and for classifying sources as good (in agreement) or bad (biased or false/spoofed). Together, these allow combining estimates from GM sources in a way that guarantees some false-rejection probability of good sources in the merged best-estimate and allows generating a set of ellipsoids to approximate the most-probable volume for displaying the result to a user.