大语言模型驱动的通信制式识别的挑战、机遇与应用前景
CSTR:
作者:
作者单位:

南京邮电大学通信与信息工程学院,南京 210003

作者简介:

通讯作者:

桂冠,男,教授,博士生导师,E-mail: guiguan@njupt.edu.cn。

中图分类号:

TN911.71

基金项目:


Challenges, Opportunities, and Future Directions of Large Language Model Empowered Wireless Technology Recognition
Author:
Affiliation:

School of Communications and Information Engineering, Nanjing University of Posts and Telecommunications, Nanjing 210003, China

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    随着无线通信技术的快速演进,5G、Wi-Fi和窄带物联网(Narrowband internet of things, NB-IoT)等多种通信制式并存,网络结构与信道环境的复杂性显著提升,使通信制式识别在频谱管理、干扰抑制与安全监测中的作用愈加重要。传统依赖人工特征与规则的方法在动态复杂场景及未知协议下适应性有限,而深度学习通过端到端建模与自动特征提取显著提升了识别精度与鲁棒性,但其在跨制式泛化、数据稀缺与计算开销方面仍面临瓶颈。以大语言模型与多模态模型为代表的大规模预训练模型凭借强泛化、跨任务迁移和少样本学习能力,展现出在复杂通信制式识别中的巨大潜力。本文系统梳理了通信制式识别技术的发展脉络,重点探讨了大语言模型驱动方法的最新进展,剖析了其在泛化能力、可解释性与高效部署等方面所面临的挑战,挖掘了其在智能频谱管理与安全监测中的应用机遇,并展望了其在5G/6G智能网络管理中的发展前景。

    Abstract:

    With the rapid evolution of wireless communication technologies, multiple standards such as 5G, Wi-Fi, and narrowband internet of things(NB-IoT) coexist, which lead to significantly increased complexity in network architectures and channel environments. This makes wireless technology recognition increasingly vital for spectrum management, interference mitigation, and security monitoring. Traditional methods relying on handcrafted features and rule-based approaches exhibit limited adaptability in dynamic and complex scenarios, particularly when facing unknown protocols. Deep learning has substantially improved recognition accuracy and robustness through end-to-end modeling and automatic feature extraction. However, it still encounters bottlenecks in cross-standard generalization, data scarcity, and computational overhead. In recent years, large-scale pre-trained models, represented by large language models and multimodal models, have demonstrated remarkable potential in complex wireless technology recognition owing to their strong generalization, cross-task transfer, and few-shot learning capabilities. This paper provides a systematic review of the development of wireless technology recognition, with a focus on the latest advances driven by large language models. We analyze the challenges related to generalization ability, interpretability, and efficient deployment, explore the opportunities in intelligent spectrum management and security monitoring, and finally present a forward-looking perspective on their potential applications in intelligent network management for 5G/6G.

    参考文献
    相似文献
    引证文献
引用本文

陈思,胡力凡,章伟杰,唐甜甜,桂冠.大语言模型驱动的通信制式识别的挑战、机遇与应用前景[J].南京航空航天大学学报,2025,57(5):822-830

复制
分享
相关视频

文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2025-06-25
  • 最后修改日期:2025-08-26
  • 录用日期:
  • 在线发布日期: 2025-10-27
  • 出版日期:
文章二维码
您是第位访问者
网站版权 © 南京航空航天大学学报
技术支持:北京勤云科技发展有限公司