Predicting satellite close approaches using statistical parameters in the context of artificial intelligence

In order to ensure a sustainable use of low earth orbit in particular and near Earth space in general, reliable and effective close approach prediction between space objects is key. Only this allows for efficient and timely collision avoidance. Space Situational Awareness (SSA) for commercial and government missions will be facing the rapidly growing amount of small and potentially less agile satellites as well as debris in the near earth realm, such as the increase in CubeSat launches and upcoming large constellations. At the same time, space object detection capabilities are expected to increase significantly, allowing for the reliable detection of smaller objects, e.g. when the Air Force Space Fence radar becomes operational. In combination, the space object catalog is expected to increase tremendously in size. In this paper, we introduce an investigative approach based on the latest capabilities in artificial intelligence in fostering the potential for fast and accurate close approach predictions. We consider the study of statistical and information theory parameters in contrast and complementary to the classical probability of collision computation alone, in order to determine the feasibility of reliably predicting close approaches.