Adaptation of an Established Evolutionary Algorithm (for Neural Net Training) to Graph Discovery for Lunar Track Network Planning(NPO-53723-1)

This algorithm is an adaptation of NEAT (NeuroEvolution of Augmenting Topologies), originally designed for optimizing weights and topologies of neural networks, but repurposed to instead find graphs with 2D spatial meaning. It uses the core ideas behind NEAT for evolving solutions stochastically, and can in principle work with any abstract loss/fitness function, so long as the evaluation is fast enough for the algorithm to make progress.