Comparative Analysis of Collision Avoidance Decision-Making Across Organizations

Abstract

The growing density of human-made objects in orbit is resulting in an increase in conjunction events. Collision avoidance has become an essential component of satellite operations, with many organizations relying on on call satellite operators to initiate avoidance maneuvers to mitigate collision risks. Given the expected proliferation of conjunctions in the coming decade, it will be essential to automate this process, thus reducing operator workload and enhancing scalability. This study compares the decision-making processes of satellite operators across five organizations, including both space agencies and private companies. Six analysts are given a set of 30 critical conjunction events and tasked with making go/no-go decisions. Their decision criteria and thresholds are analyzed to identify variations in their approaches. Based on these insights, deterministic rule-based classifiers are developed to replicate each analyst’s decision-making process. Additionally, a Long Short-Term Memory (LSTM) model, trained on a synthetic dataset of 8000 critical conjunctions, is tested on the same 30 events assessed by the analysts, and its classifications are compared to theirs. The performance of both the rule-based classifiers and the LSTM model is further evaluated on the synthetic dataset, with the LSTM achieving an F1-score of 88%, demonstrating its strong potential for automating or supplementing decision-making processes.