通信学报2026，Vol.47Issue(5)：282-292,11.DOI:10.11959/j.issn.1000-436x.TXXB260066
基于元深度强化学习的蜂窝网链路自适应方法
Meta deep reinforcement learning-based link adaptation method for cellular networks
摘要
Abstract
To address the stringent requirements for reliability and data rate in cellular networks,an efficient and highly generalizable link adaptation method was proposed.First,to ensure reliable wireless communication transmission,a con-strained modulation and coding scheme selection strategy was designed to meet the block error rate requirement.Second,to overcome the poor generalization capability of traditional algorithms in unknown transmission environments,a meta-learning mechanism was integrated with deep reinforcement learning.Through offline training followed by online fine-tuning,rapid policy convergence was achieved.Simulation results demonstrate that,while strictly satisfying the block er-ror rate requirement,the proposed method achieves higher data rate performance and stronger generalization capability compared with traditional link adaptation methods.关键词
链路自适应/深度强化学习/元学习/泛化能力Key words
link adaptation/deep reinforcement learning/meta-learning/generalization capability分类
信息技术与安全科学引用本文复制引用
叶小文,林恒羿,吴怡..基于元深度强化学习的蜂窝网链路自适应方法[J].通信学报,2026,47(5):282-292,11.基金项目
国家自然科学基金资助项目(No.62501157,No.U25A20398) （No.62501157,No.U25A20398）
福建省青年科技人员育成基金资助项目(No.2025350410)The National Natural Science Foundation of China(No.62501157,No.U25A20398),The Foundation for Culti-vated Young Talents of Fujian Province(No.2025350410) （No.2025350410）
