移动通信2025，Vol.49Issue(7)：21-30,10.DOI:10.3969/j.issn.1006-1010.20250520-0002
大语言模型驱动的无线通信物理层任务智能处理技术综述
A Survey of LLM-Driven Intelligent Processing for Physical-Layer Tasks in Wireless Communication
摘要
Abstract
6G aims to build intrinsically intelligent networks,with the physical layer facing multiple challenges such as high-frequency channel modeling,ultra-large-scale antenna array optimization,and dynamic task adaptation.Traditional approaches,which rely on mathematical formulations or dedicated models,often fall short in addressing the complexity of emerging scenarios.As a core foundation model,large language models(LLMs)offer promising capabilities for physical-layer intelligence,leveraging their strong generalization ability and multimodal fusion potential.This paper provides a systematic survey of recent advances in applying LLMs to physical-layer tasks,including channel prediction,beam management,and resource allocation.It also reviews key enabling techniques such as architectural innovations,parameter-efficient fine-tuning,and multimodal information integration.By analyzing current model capabilities and limitations,the paper identifies future research directions such as constructing dynamic LLM-based channel knowledge bases and enhancing semantic communication through knowledge reinforcement.Finally,the integration of trustworthy AI,model-data fusion,and edge intelligence is discussed to support the evolution of intelligent physical-layer processing in 6G.关键词
大语言模型/物理层智能化/6G通信/信道建模/多模态学习/参数高效微调/基于LLM的信道知识库Key words
large language models(LLMs)/intelligent physical layer/6G communication/channel modeling/multimodal learning/parameter-efficient fine-tuning/LLM-based channel knowledge base分类
信息技术与安全科学引用本文复制引用
王文,孙亚萍,许晓东,何业军,陈昊,马楠,崔曙光..大语言模型驱动的无线通信物理层任务智能处理技术综述[J].移动通信,2025,49(7):21-30,10.基金项目
国家自然科学基金"语义知识库驱动的零样本多层级语义编码与特征传输研究"(62301471) （62301471）
国家自然科学基金"语义知识库构建方法与智能演进机理"(62293482) （62293482）
移动信息网络国家科技重大专项(2024ZD1300700) （2024ZD1300700）
