摘 要
随着互联网规模的持续扩展和新兴应用的不断涌现，通信网络中的拥塞控制问题日益凸显，成为影响用户体验和服务质量的关键因素。传统拥塞控制算法在应对复杂动态网络环境时存在响应滞后、效率低下等问题，难以满足高带宽低延迟场景的需求。为此，本研究旨在优化现有拥塞控制算法，提升其在网络流量管理中的适应性和性能表现。通过结合机器学习与传统传输控制协议（TCP）机制，提出了一种基于深度强化学习的自适应拥塞控制算法，该算法能够实时感知网络状态并动态调整拥塞窗口大小。实验采用模拟网络环境与真实测试床相结合的方式，对所提算法进行验证与评估。
关键词：拥塞控制 深度强化学习 传输控制协议
Abstract
With the continuous expansion of the Internet scale and the continuous emergence of emerging applications, the problem of congestion control in the communication network has become increasingly prominent, which has become a key factor affecting the user experience and service quality. Traditional congestion control algorithm has problems such as response lag and low efficiency in dealing with complex dynamic network environment, which is difficult to meet the requirements of high bandwidth and low latency scenarios. To this end, this study aims to optimize the existing congestion control algorithms to improve their adaptability and performance in network traffic management. By combining machine learning and traditional transmission control protocol (TCP) mechanism, an adaptive congestion control algorithm based on deep reinforcement learning is proposed, which can sense the network state in real time and dynamically adjust the congestion window size. The experiment combines the simulated network environment and the real test bed to verify and evaluate the proposed algorithm.
Keyword:Congestion Control Deep Reinforcement Learning Transmission Control Protocol
目 录
1绪论 1
1.1通信网络拥塞控制的研究背景 1
1.2拥塞控制算法优化的意义与价值 1
1.3国内外研究现状分析 1
1.4本文研究方法与技术路线 2
2拥塞控制算法的基础理论 2
2.1通信网络中的拥塞现象解析 2
2.2常见拥塞控制算法概述 3
2.3拥塞控制算法的性能评价指标 3
2.4现有算法的局限性分析 4
2.5算法优化的基本原则 4
3拥塞控制算法优化的关键技术 4
3.1动态带宽估计方法改进 4
3.2延迟感知的拥塞检测机制 5
3.3流量预测模型的构建与优化 5
3.4自适应调整策略的设计 6
3.5多路径传输中的拥塞控制优化 6
4拥塞控制算法优化的应用与验证 7
4.1实验环境与数据集设计 7
4.2优化算法的实现过程 7
4.3性能测试与结果分析 8
4.4与其他算法的对比研究 8
4.5应用场景的扩展与展望 9
结论 9
参考文献 11
致谢 12
随着互联网规模的持续扩展和新兴应用的不断涌现，通信网络中的拥塞控制问题日益凸显，成为影响用户体验和服务质量的关键因素。传统拥塞控制算法在应对复杂动态网络环境时存在响应滞后、效率低下等问题，难以满足高带宽低延迟场景的需求。为此，本研究旨在优化现有拥塞控制算法，提升其在网络流量管理中的适应性和性能表现。通过结合机器学习与传统传输控制协议（TCP）机制，提出了一种基于深度强化学习的自适应拥塞控制算法，该算法能够实时感知网络状态并动态调整拥塞窗口大小。实验采用模拟网络环境与真实测试床相结合的方式，对所提算法进行验证与评估。
关键词：拥塞控制 深度强化学习 传输控制协议
Abstract
With the continuous expansion of the Internet scale and the continuous emergence of emerging applications, the problem of congestion control in the communication network has become increasingly prominent, which has become a key factor affecting the user experience and service quality. Traditional congestion control algorithm has problems such as response lag and low efficiency in dealing with complex dynamic network environment, which is difficult to meet the requirements of high bandwidth and low latency scenarios. To this end, this study aims to optimize the existing congestion control algorithms to improve their adaptability and performance in network traffic management. By combining machine learning and traditional transmission control protocol (TCP) mechanism, an adaptive congestion control algorithm based on deep reinforcement learning is proposed, which can sense the network state in real time and dynamically adjust the congestion window size. The experiment combines the simulated network environment and the real test bed to verify and evaluate the proposed algorithm.
Keyword:Congestion Control Deep Reinforcement Learning Transmission Control Protocol
目 录
1绪论 1
1.1通信网络拥塞控制的研究背景 1
1.2拥塞控制算法优化的意义与价值 1
1.3国内外研究现状分析 1
1.4本文研究方法与技术路线 2
2拥塞控制算法的基础理论 2
2.1通信网络中的拥塞现象解析 2
2.2常见拥塞控制算法概述 3
2.3拥塞控制算法的性能评价指标 3
2.4现有算法的局限性分析 4
2.5算法优化的基本原则 4
3拥塞控制算法优化的关键技术 4
3.1动态带宽估计方法改进 4
3.2延迟感知的拥塞检测机制 5
3.3流量预测模型的构建与优化 5
3.4自适应调整策略的设计 6
3.5多路径传输中的拥塞控制优化 6
4拥塞控制算法优化的应用与验证 7
4.1实验环境与数据集设计 7
4.2优化算法的实现过程 7
4.3性能测试与结果分析 8
4.4与其他算法的对比研究 8
4.5应用场景的扩展与展望 9
结论 9
参考文献 11
致谢 12
