首页|期刊导航|南京航空航天大学学报(英文版)|在干扰中学习生存:人工智能驱动的无线通信物理层抗干扰综述

在干扰中学习生存:人工智能驱动的无线通信物理层抗干扰综述OA

Learning to Survive Jamming:A Survey of AI-Driven Physical-Layer Anti-jamming in Wireless Communications

中文摘要英文摘要

无线信道的广播特性使通信链路易受蓄意干扰.随着频谱密集复用、海量连接以及空天地一体化的发展,这一威胁将在6G网络中进一步加剧.传统物理层(Physical-layer,PHY)抗干扰方法依赖准确的模型和近似平稳的干扰统计特性,其性能在自适应和快速变化的干扰环境下容易下降.人工智能(Artificial intelligence,AI)可直接从信号、频谱和交互数据中学习检测规则与自适应策略,从而降低对上述假设的依赖.本文系统综述人工智能驱动的物理层抗干扰技术,并建立由学习范式和物理层功能构成的统一双轴分类框架.首先,给出统一的干扰模型,并在分布偏移、实现约束和对抗不确定性条件下对相关学习问题进行建模.随后,系统梳理6类学习范式,包括:监督深度学习,深度强化学习,迁移学习、元学习与持续学习,联邦学习与分布式学习,生成式学习与自监督学习,以及端到端学习与深度展开.进一步,从干扰感知、频域自适应、空域抑制、链路自适应和跨域联合优化5类物理层功能对现有研究进行归纳.最后,讨论关键部署约束和验证缺口,并展望面向鲁棒且符合标准的抗干扰技术的未来研究方向.

The broadcast nature of wireless channels exposes communication links to intentional jamming,which is a threat that will be exacerbated in 6G networks owing to dense spectrum reuse,massive connectivity,and space-air-ground integration.Conventional physical-layer(PHY)countermeasures rely on accurate models and approximately stationary jammer statistics,and their effectiveness degrades under adaptive and rapidly varying jamming.Artificial intelligence(AI)relaxes these assumptions by learning detection rules and adaptation policies directly from signal,spectrum,and interaction data.This survey provides a comprehensive review of AI-driven PHY anti-jamming techniques,organized through a unified dual-axis taxonomy of learning paradigms and PHY functions.We first present a unified jamming model and formulate the associated learning problem under distribution shift,implementation constraints,and adversarial uncertainty.Six paradigm groups are then examined:Supervised deep learning;deep reinforcement learning;transfer,meta,and continual learning;federated and distributed learning;generative and self-supervised learning;and end-to-end learning with deep unfolding.Existing work is subsequently organized by five PHY functions:jamming sensing,frequency-domain adaptation,spatial-domain suppression,link adaptation,and cross-domain joint optimization.Finally,we discuss critical deployment constraints and validation gaps,and identify research directions toward robust and standards-compliant anti-jamming solutions.

陈嘉雯;丁旭辉;张园园;李心宇;杨凯;张先超;陆军;安建平

北京理工大学网络空间安全学院,北京 100081,中国北京理工大学网络空间安全学院,北京 100081,中国北京理工大学网络空间安全学院,北京 100081,中国北京理工大学信息与电子学院,北京 100081,中国北京理工大学信息与电子学院,北京 100081,中国嘉兴大学全省多模态感知与智能系统重点实验室,嘉兴 314001,中国中国电子科技集团公司电子科学研究院,北京 100049,中国北京理工大学网络空间安全学院,北京 100081,中国

信息技术与安全科学

物理层抗干扰人工智能深度学习无线通信干扰抑制

physical-layer anti-jammingartificial intelligence(AI)deep learningwireless communicationsjamming mitigation

《南京航空航天大学学报(英文版)》 2026 (4)

475-505,31

This work was supported in part by the National Key Research and Development Program of China(No.2023YFC3305904),in part by the National Natural Science Foundation of China(No.62301049),and in part by the Natural Science Foundation of Shandong Prov-ince of China(No.ZR2024QF028).

10.16356/j.1005-1120.2026.04.001

评论