AI to support decision making in collision risk assessment

This paper introduces an Artificial Intelligence(AI)-based system to support operators to manage the risk of collisions. The system implements an Artificial Neural Network (ANN) based technique to predict the risk of collision between two space objects, where one of the two is an operational satellite and the other one is a piece of space debris. The ANN-based technique provides a prediction of the Probability of Collision (PC), MOID and B-parameter between a primary satellite and a set of space debris objects during an interval of time starting from a database including the initial states synthetic space objects. Results of the system over a database of synthetic objects not previously seen during training show the good accuracy predicting the evolution of the aforementioned variables. The achieved results suggest that the proposed system can be able to implement in collision assessment as a method to identify - quickly, accurately and automatically - possible conjunctions between space objects in an interval of time with no use of dynamic models or orbit propagators. A revised calculation of the PC is also proposed to mitigate the Dilution of Probability that affects the usual definition of this quantity. This phenomenon gives the counterintuitive idea that the lower the quality of the data (or amount of information available to the operators), the smaller the probability of collision, which can lead to a false confidence in the likelihood of a collision or forces operators to accept very large margins. The method presented here will account for epistemic uncertainty under the assumption of Dempster-Shaffer’s Theory of Evidence which leads to the definition of confidence intervals on the probability of a collision. Confidence intervals incorporate the dependency of the probability of collision on the amount and quality of the available information, using the concepts of Belief and Plausibility introduced in Theory of Evidence. The result of this revised calculation of the PC is a more informed decision. At the same time, a lack of information can lead to a higher uncertainty on the decision to be made. Thus the paper will propose a possible approach to make optimal decisions under epistemic uncertainty where the cost of the decision is the risk associated with the decision are concurrently taken into account.