LAUSR

Abstract To reduce computation time while limiting loss in accuracy when propagating an orbit state probability density function, this work seeks to develop an adaptive approach to multi-fidelity uncertainty propagation for applications in astrodynamics. Using the method of stochastic collocation, a set of particles produced via a low-fidelity solver defines a basis used in the surrogate over the space of propagated states. This basis allows for identifying a subset of important samples that are re-propagated using a high-fidelity propagator, which defines a correction for the original basis. This approach reduces computation time for propagating a particle ensemble or a Gaussian mixture model via the unscented transform. This paper demonstrates the efficacy of this method for several Earth-orbit test cases, and provides a means for merging general and special perturbation theories to produce a posterior probability density function more statistically consistent with a precise estimated state.