Nonlinear Uncertainty Propagation in Astrodynamics: Comparing Taylor Differential Algebra with Monte-Carlo on GPUS

In this paper two approaches for nonlinear uncertainty propagation in astrodynamics are compared. The first approach is based on Taylor Differential Algebra and is aimed at the improvement and generalization of standard linear methods. The second approach is aimed at increasing the computational performances of classical Monte-Carlo simulations exploiting their intrinsic parallel structure and taking advantage of the massively parallel architecture of modern GPUs. The two proposed approaches are applied to test cases considering both simple twobody dynamics and full n-body dynamics with JPL ephemeris. The results of the propagations are thoroughly compared with particular emphasis on the computational performances.