LLM-Driven Intelligent Network Slicing for Satellite Direct-to-Device Communications

Seamless global connectivity is becoming a reality with the emergence of 6G non-terrestrial networks (NTNs), enabling direct-to-device (D2D) satellite communications. However, managing dynamic satellite networks efficiently remains a challenge due to unpredictable traffic patterns, limited resources, and high-latency environments. This article introduces a novel approach leveraging Large Language Models (LLMs) to intelligently optimize network slicing, ensuring faster, more efficient resource allocation. By integrating federated learning, the system adapts to real-time network conditions while preserving user privacy. Additionally, blockchain technology enhances security and transparency in model updates. Our results show that this AI-driven solution reduces latency by 40% and improves resource utilization by 60% compared to traditional methods.