Machine Learning-based Prediction of Starlink Satellite Lifespan for Sustainable Space Operations

Mega-constellation growth at Starlink requires accurate prediction of their lifespan for mission planning and proper space debris management. The study explores using- Decision Tree Regressor, XGBoost, and Random Forest - to forecast Starlink satellite operating lifespan. With the use of publicly available operational data and orbital parameters, we train and validate these models with comparison of prediction accuracy. We examine the impact of essential characteristics, including launch altitude, inclination, and operational experience, on satellite lifespan. The research will reveal insights into the determinants of satellite lifespan and advance the creation of durable predictive models for the sustainable use of near-Earth space. The comparison of these three algorithms provides useful insights into choosing the best method for predicting lifespan in the context of large satellite constellations. This work ultimately aims to enhance our knowledge of satellite degradation and provide insights into how to mitigate the growing problem of space debris.