Rapid approximation of invariant manifolds using machine learning methods

Low-energy mission design in the three-body model leverages invariant manifolds to obtain low-propellant solutions. Optimizing these trajectories requires generating manifolds and searching for the optimal manifold insertion point. Typically, manifolds are generated using numerical methods which can take up to several seconds, thus making the generation of these structures in an optimization framework computationally intractable. In this paper we will explore the application of machine learning algorithms to enable rapid approximation of these structures. The regression models will then be used within an optimization framework. The robustness, accuracy and computational advantages will be benchmarked against Cubic Convolution based approximation methods.