Geosynchronous Satellite Pattern-of-Life Characterization Through Machine Learning-based Mode Change Detection and Classification

With the increasing number of satellites in the Geosynchronous Earth Orbit (GEO), there is a need to develop space situational awareness tools that accurately characterize their on-orbit behavior. This paper proposes a two-step machine learning-based approach: first, mode change detection, followed by a multi-class classification model to characterize the patterns-of-life of GEO satellites. The approach utilizes features derived from astrometric data tailored to capture trends in the east-west and north-south behavior of the satellites. The model’s performance demonstrated a significant improvement over the baseline models provided in the Satellite Pattern-of-Life Identification Dataset and development kit for the AI SSA Challenge.