Intelligent Spectrum-Energy Co-Optimization in Space-Air-Ground Integrated 6G Networks

Abstract

A Space Air Ground Integrated Network (SAGIN) unifies satellite, aerial, and terrestrial access. We optimize spectrum, power, and routing jointly. Carriers, bandwidths, and backhaul delays differ across segments, creating tight coupling. Under high load and blockage, naive decisions inflate delay and energy cost. Baselines include water filling with heuristic routing. We also test flat Deep Q Network (DQN) and Proximal Policy Optimization (PPO). We use a hierarchical controller that first chooses routing, then allocates spectrum and power. The reward is latency-aware and penalizes energy use and unfairness under Service Level Agreement (SLA) limits. Ground channels come from DeepMIMO (ray-tracing channel generator) at 28 GHz with 256 subcarriers, $N_{t}=64, N_{r}=16$, and $L_{\max }=10$. The setup uses 6 Low Earth Orbit (LEO) nodes at $20 \text{GHz}, 8$ aerial relays at 3.5 GHz, and 12 base stations with 240 users. Over five runs, our policy reaches 10.56 Gbps, 14.9 Mbit/J, and 12.4 ms. The baseline reaches 9.12 Gbps, 11.3 Mbit/J, and 15.2 ms; stress causes a 12.9% throughput drop with 7.3% SLA violations. Hierarchy improves efficiency without sacrificing latency. This supports structured control for practical SAGIN management.