Enhancing Network Intelligence with LLM-Based IBN and DRL: A Dynamic Approach for SAGIN Resource Management

We propose an innovative Intent-Based Networking (IBN) system that leverages Large Language Models (LLMs) to translate high-level user intents into actionable network policies. By integrating an LLM-based translation engine with a Deep Reinforcement Learning (DRL) agent, our system enables dynamic and intelligent resource allocation within Space-Air-Ground Integrated Networks (SAGINs). The LLM interprets user requests expressed in natural language to generate precise network configurations, which the DRL agent uses to optimize resource utilization while meeting Quality of Service (QoS) requirements. Simulations demonstrate that our approach effectively reduces latency, enhances bandwidth utilization, minimizes QoS violations, and increases Virtual Network Request (VNR) acceptance rates compared to static allocation strategies. These results highlight the potential of combining LLMs with DRL in IBN systems to improve network adaptability, efficiency, and user satisfaction.