Intelligent Hierarchical Routing Strategy Based on Regional Division for Large-Scale LEO Satellite Constellations

Low Earth Orbit (LEO) satellite networks play a crucial role in the future 6G systems by providing global coverage and low-latency communication. Efficient routing schemes are essential to ensure network performance. Traditional routing methods are struggling to handle the dynamic characteristics and complex requirements of large-scale satellite constellations. Existing machine learning-based solutions mainly focus on static or medium-scale constellations, lacking sufficient scalability and real-time performance to meet the needs of large-scale dynamic constellations. To address this issue, we propose a dual-layer routing strategy based on Medium Earth Orbit (MEO) and LEO satellites, which combines Graph Neural Network (GNN) and Deep Reinforcement Learning (DRL) to optimize routing. In addition, regional partitioning is introduced to alleviate computational burden. The simulation results demonstrate that, compared with the benchmark scheme, our approach significantly reduces end-to-end latency and packet drop rate, while improving network performance and stability. The code and pre-trained models are publicly available at https://github.com/succulents9/routing.