AI-native 6G Network based on Large Language Model (LLM)

Abstract

The sixth-generation (6G) mobile communication is evolving towards the AI-native paradigm of integrated communication, sensing, computing and intelligence. The stringent requirements from vertical scenarios such as the Industrial Internet and autonomous driving have increasingly exposed the inherent bottlenecks of traditional solutions represented by convex optimization and shallow machine learning, rendering them unable to underpin the design objectives of AI-native 6G networks. This paper conducts a systematic review themed on the reconstruction of the core technical system of AI-native 6G networks by large language models (LLMs). It defines the core connotation and three core design principles of AI-native 6G, proposes LLM-driven centralized and distributed top-level architectures as well as a three-layer cloud-edge-terminal collaborative deployment scheme, comprehensively sorts out the research progress and enabling mechanisms of LLMs across four core dimensions including spectrum resource allocation and network slice management, systematically summarizes the key technical challenges and corresponding solutions in the integration process, and prospects the priority research directions in the future.