Comparative analysis of Neural Networks approaches for Initial Orbit Determination of Near-Earth Objects

Initial Orbit Determination from Very Short Arcs (VSAs) is a challenging problem in asteroid surveillance and celestial mechanics. Classical orbit determination methods require angular observations spanning a sufficient fraction of the orbit and become ill-conditioned or fail when only a single night's data is available, as is typical for newly discovered Near-Earth Objects (NEOs). This paper presents a comparative analysis of two neural network approaches designed to overcome these limitations.