Toward automated collision avoidance: Predicting the risk of satellite collisions using machine learning-powered techniques

Active collision avoidance has become an important task in space operations nowadays, and hundreds of alerts corresponding to close encounters of a satellite and other space objects are typically issued for a satellite in Low Earth Orbit every week. Such alerts are provided in the form of conjunction data messages, and only about two actionable alerts per spacecraft and week remain to be resolved after analyzing all cases. Therefore, building fully automated techniques for predicting the collision risk can help make the process of avoiding collisions less costly, as the number of false positives could be substantially reduced. In this paper, we present our machine learning-powered techniques for this task which have been exploited in the Collision Avoidance Challenge organized by the European Space Agency, in which we took the 7^th place (out of 97 registered participants).