Publication update: This work is now published in IEEE Transactions on Aerospace and Electronic Systems, with a publication date of 25 September 2026. The results below reflect the journal version.
Title: Supervised Classification of Synthetic LEO Debris Families Using Multiset Proper Elements Under Atmospheric-Drag-Free Dynamics Authors: Michael Ling and Yang Yang Journal: IEEE Transactions on Aerospace and Electronic Systems, volume 62 (2026) DOI: 10.1109/TAES.2026.3737763
Journal results
The study compares modified equinoctial, Poincaré and quaternion proper elements for classifying whether synthetic debris fragments share a parent. An augmented quaternion representation, QTNp, adds an explicit orbital-size proxy and achieves the highest positive-class F1 (0.1245, compared with 0.1023 for standard QTN), with recall increasing from 0.156 to 0.263.
MEE + QTNp achieves the highest ROC-AUC (0.656). QTNp, MEE + QTNp and MEE + PNC + QTNp share the highest PR-AUC (0.071). More element sets do not consistently improve same-parent retrieval.
The benchmark is synthetic and excludes atmospheric drag. Thresholds for F1 and recall were selected posthoc on the evaluation set; these scores are not deployment-ready estimates or validation on real operational debris.
Congratulations to Michael and the team for this achievement!