LLM-Enhanced Dynamic Spectrum Management for Integrated Non-Terrestrial and Terrestrial Networks: A Multi-Objective Optimization Approach

Abstract The convergence of satellite and terrestrial networks introduces complex spectrum management challenges in next-generation wireless systems. This work proposes an innovative AI-driven approach for dynamic resource allocation in integrated non-terrestrial and terrestrial networks (NTN-TN). Our framework leverages large language models to intelligently manage spectrum resources across multiple frequency bands, adapting to dynamic network conditions and diverse service requirements. Through comprehensive system evaluation, we demonstrate significant improvements in key performance metrics, including network throughput, interference reduction, and computational efficiency compared to conventional methods. The solution maintains equitable resource distribution while meeting stringent latency requirements for emerging 6G applications. Experimental results validate the effectiveness of our approach under various operational scenarios, showing consistent performance gains across different network scales and configurations.