基于深度强化学习的多路径调度模型
Multi-path Scheduling Model Based on Deep Reinforcement Learning
赵静1
作者信息
- 1. 兰州职业技术学院信息工程学院,甘肃 兰州 730070
- 折叠
摘要
文章提出一种基于深度强化学习的多路径调度模型,在聚合场景中将深度学习技术应用于流量管理,以解决多路径环境下的数据包调度问题.文章使用了一个多路径快速UDP网络连接协议(MPQUIC)来实现多路径场景中的路径选择,并训练了一个代理人(Agent)来改进最优选择路径的算法,展示了将深度Q网络代理(DQN Agent)应用于数据流量管理问题的优势.实验证明了在实时环境中使用DQN Agent来提高包调度器性能的可行性,以及使用该技术对新的5G网络进行优化的潜力.实验结果表明:基于深度强化学习的多路径调度模型能够自适应地调整路径选择策略,从而提高网络的稳定性和可靠性.改进的模型不仅具有理论价值,还为实际应用提供了有益的参考和借鉴.
Abstract
In this paper,a multi-path scheduling model based on deep reinforcement learning is proposed,and deep learning technology is applied to traffic management in aggregation scenario to solve the problem of packet scheduling in multi-path environment.A multi-path Quick UDP Internet Connection is used to implement path se-lection in multi-path scenarios,and an agent is trained to improve the optimal path selection algorithm,demonstrat-ing the advantages of applying DQN Agent to data traffic management problems.Experiments demonstrate the feasi-bility of using DQN Agent to improve the performance of packet scheduler in real-time environment,and the poten-tial of using this technology to optimize the new 5G networks.The experimental results show that the multi-path scheduling model based on deep reinforcement learning can adaptively adjust the path selection strategy,thereby improving the stability and reliability of the network.The improved model not only has theoretical value,but also provides useful reference for practical application.
关键词
5G网络/多路径/分组调度/深度强化学习/QUIC/MPTCPKey words
5G network/multi-path/packet scheduling/deep reinforcement learning/QUIC/MPTCP引用本文复制引用
基金项目
甘肃省基础研究计划软科学研究计划专项(22JR4ZA108)
出版年
2024
NETL
NSTL
维普
万方数据