Intelligent Traffic Steering for GEO–LEO Satellite Constellation: A Stable Matching Approach

Abstract

To enable global connectivity through 6G, the efficient operation of hierarchical satellite networks that integrate geostationary (GEO) and low-earth orbit (LEO) satellites is paramount. A significant challenge in achieving this operational efficiency lies in the dynamic association between the extensive array of LEO satellites and ground stations (GSs). In LEO satellite constellations, accurately estimating the queuing delay experienced by data along end-to-end (E2E) paths is challenging because of the complex interleaving of routing paths from countless sources and destinations. In particular, the satellite-to-ground links, which possess lower transmission capacity than inter-satellite links, often become critical bottlenecks for delay. Therefore, this study focuses on GS traffic loads and mathematically demonstrates, through convexity verification of queuing delays, that minimizing the maximum load effectively reduces the E2E delay. Building on these findings, we propose a novel GS-LEO association method designed to reduce delay while suppressing the maximum GS load with low computational complexity. Simulation results utilizing real-world parameters, including IXP locations and traffic demand distributions, demonstrate that the proposed method achieves lower E2E delay than existing routing approaches while maintaining a significantly lower computational load compared with strict optimization methods.