Benchmarking deep learning approaches for all-vs-all conjunction screening

Abstract

The all-vs-all problem, for which conjunctions are screened for over all possible sets of catalogued objects, is crucial for space traffic management and space situational awareness, but is a computational challenge owing to the vast and growing number of possible conjunction pairs. In this work, we present the application of deep learning techniques to this problem, framing conjunction screening as a machine learning classification task. We investigate the performance of different input data representations and model architectures on a realistic all-vs-all dataset, generated using the CNES BAS3E space surveillance simulation framework, and consisting of 170 million object pairs over a 7-day screening period. These approaches are benchmarked against operationally used classical filters in both screening capability and computational efficiency, and the ability of deep learning algorithms to cope and aid with the scales required for current and future operational all-vs-all scenarios is demonstrated.