Bayesian Benchmarking of GBEES Applied to Outer Planet Orbiter Estimation

What is it about?

Novel deep space missions aim to thoroughly explore outer planetary celestial bodies (e.g., Europa, Enceladus, etc.) that may harbor the building blocks of life. To collect data, spacecraft will follow chaotic, quasi-periodic trajectories that maximize proximity to the surface of these bodies for extended periods while minimizing the need for costly maneuvers. At these distances, measurement signals will face time delays on the order of hours. The nonlinearity of the dynamics of the spacecraft combined with the infrequency of correction updates will inherently lead to spacecraft state uncertainty becoming non-Gaussian. Moment filters have been a cornerstone in spacecraft navigation and estimation for decades, but tend to diverge under these conditions. Although ensemble filters offer greater accuracy, they must also demonstrate sufficient robustness and efficiency to meet the considerable demands of practical spacecraft navigation. In this manuscript, we benchmark our novel filter, Grid-based Bayesian Estimation Exploiting Sparsity, against several moment and ensemble filters. We place significant emphasis on accurately approximating the spacecraft's full probability density function, recognizing that first and second central moments alone fail to capture the complete picture when uncertainty is non-Gaussian.

Featured Image