AI for autonomous CAM execution

This paper combines a previously developed Intelligent Classification Systems (ICS) for collision risk prediction witha simple Collision Avoidance Manoeuvre (CAM) allocation procedure. The Intelligent Classification System is basedon a combination of Evidence Theory for collision risk assessment and a Machine Learning model that classifiesconjunction events given the encounter geometry, the uncertainty in the probability of collision and the time at whichthe conjunction event occurs.We introduce a quick method to compute a Collision Avoidance Manoeuvre when the Intelligent Classification Systemsuggests that a CAM is needed. The method presented in this paper accounts for epistemic uncertainty in the collisionprediction. The inclusion of the epistemic uncertainty requires solving a min-max problem to find the optimal impulsefor the worst-case scenario. Finally, the paper introduces the basis for a future ML-based system able to predict theoptimal CAM under epistemic uncertainty.