Closing the loop between mission design and navigation analysis

This research presents a contribution to the field of optimisation under uncertainty by proposing a nonlinear navigation analysis approach under both aleatoric and epistemic uncertainty. The navigation analysis is framed as a sequential filtering problem with an alternation of uncertainty propagation arcs and orbit determination instances. The state distribution is modelled as a mixture of kernels where the mixture weights are interval-valued to model the epistemic component. Specific techniques are discussed for Gaussian mixtures in particular. The uncertainty propagation step is solved by using Gaussian Hermite quadrature rules to compute the propagated means and covariances. The observation update is efficiently solved with a combination of variational inference and importance sampling, and a routine is developed to compute the posterior interval-valued weights. Given the distribution representation, lower and upper expectations of a generic quantity of interest are the solutions of linear programming problems and, therefore, are inexpensive to compute. The developed navigation analysis is finally applied to the robust quantification of the probability of impact of Europa Clipper with Jupiter’s moon Europa during one of its close flybys.