A Hybrid Approach to Enhanced SGP4 for Galileo Constellations

An up-to-date catalog of resident space objects orbiting Earth is essential for effective space environment management, which requires the orbital propagation of tens of thousands of objects. Although ephemeris data are publicly accessible through catalogs such as NORAD, other organizations generate independent datasets based on their own observations. Due to the large number of objects, orbit propagators have become indispensable tools that must balance accuracy with computational efficiency. High-fidelity numerical propagators provide accurate results at the expense of computational intensity, whereas analytical models, such as SGP4, offer greater efficiency at the expense of neglecting certain dynamical effects. Hybrid orbit propagators have been proposed as an alternative by combining classical propagators to integrate the deterministic motion with machine learning or statistical time-series techniques to predict unmodeled effects and address intrinsic uncertainties. This work presents a hybrid version of the SGP4 propagator, particularly fitted for Galileo-type orbits. HSGP4 employs machine learning techniques and operates with a Hybrid Two-Line Element set (HTLE), which extends the classical TLE by including model parameters. The propagation process integrates standard SGP4 outputs with a forecast of SGP4 error, yielding more accurate ephemerides while maintaining its computational efficiency.