电子科技2025,Vol.38Issue(11):1-7,7.DOI:10.16180/j.cnki.issn1007-7820.2025.11.001
基于深度强化学习的高铁网络多路径智能拥塞控制算法
MPTCP Congestion Control for High-Speed Railway Networks Based on Deep Reinforcement Learning
摘要
Abstract
The MPTCP(Multipath Transmission Control Protocol)ensures the reliability of communication serv-ices in high-speed railway networks,yet the frequent handovers and wireless losses within these networks negatively impact MPTCP performance.To solve these problems,a HSR-MPCC(High Speed Railway Multipath Congestion Control)based on deep reinforcement learning is proposed.The HSR-MPCC algorithm adds the window factor into the traditional multipath congestion control algorithm,and can intelligently adjust the window factor value according to different network states,thereby adjusting the congestion window when the congestion window calculated by the tra-ditional multipath congestion control algorithm is larger or smaller.On this basis,deep reinforcement learning tech-nology is used to calculate the optimal addition and subtraction window factor in real time,so that the client transmis-sion rate matches the highly dynamic high-speed chain bandwidth.The experimental results show that HSR-MPCC can improve the performance of traditional multi-path congestion control algorithms such as Uncoupled,LIA(Linked Increase Algorithm)and OLIA(Opportunistic Linked Increases Algorithm).The improved multipath congestion algo-rithm can be better adapted to the dynamic high-speed railway network.关键词
高速铁路网络/多路径TCP/拥塞控制/深度强化学习/TD3/系统吞吐量/ns-3/天地一体化网络Key words
high speed railway network/multipath TCP/congestion control/deep reinforcement learning/TD3/system throughput/ns-3/space-ground integrated networks分类
信息技术与安全科学引用本文复制引用
谢周杨,王成群..基于深度强化学习的高铁网络多路径智能拥塞控制算法[J].电子科技,2025,38(11):1-7,7.基金项目
浙江省重点研发计划(2021C01047) Key R&D Program of Zhejiang(2021C01047) (2021C01047)