Gen-AI for Radio Resource Management in Multi-orbit 5G-NTN Systems

Non-terrestrial networks (NTNs) are poised to play a pivotal role in 5G-Advanced technology, delivering global connectivity and complementing terrestrial infrastructure when needed. Among NTNs, nongeostationary orbit (NGSO) systems, such as low Earth orbit (LEO) constellations, are gaining significant traction due to increasing investments in large-scale deployments. As NTNs evolve towards multi-orbit and dynamic configurations, efficient radio resource allocation emerges as a critical challenge. This paper addresses the resource management problem in 5G NTNs, targeting a fair user demand satisfaction while incorporating multiorbit connectivity, where users can simultaneously connect to multiple satellites belonging to different orbits. In an attempt to overcome the computational complexity of pure optimization-based solutions, we propose two Generative AI (Gen-AI) solutions: a transformer-based model and a generative adversarial network (GAN)-based model. Results demonstrate that the transformer-based model outperforms the alternative optimization-based method as well as the GAN-based one, showcasing its potential for scalable and efficient radio resource management in multi-orbit 5G NTN systems.