Dynamic Enhanced Resource Allocation for Beam-Hopping Satellite Networks with Rate-Splitting Multiple Access

Low Earth Orbit (LEO) satellite networks face dual challenges of dynamic channel interference and heterogeneous service quality assurance. This paper proposes a rate-splitting multiple access enhanced cross-layer dynamic resource optimization framework, achieving joint optimization of the physical layer and link layer through deep reinforcement learning. The innovation lies in establishing a closed-loop feedback mechanism for power allocation and queue state, mapping service priorities to interference suppression weights, significantly enhancing dynamic interference management capabilities. Additionally, a beam-power spatiotemporal joint scheduling algorithm based on orbital dynamics is designed to reduce resource fragmentation. A gradient-isolated hierarchical decision network is constructed to decouple the mixed action space, accelerating policy convergence. Simulation results demonstrate that this approach can simultaneously ensure low-latency transmission and high reliability in time-varying scenarios, providing a new paradigm for efficient resource management in integrated satellite-ground networks.