Machine Learning for End-to-End Congestion Control
Machine Learning for End-to-End Congestion Control
复制标题
DOI：
10.1109/mcom.001.1900509
复制
发表时间：
2020-06
影响因子：
11.2
通讯作者：
Ticao Zhang;S. Mao
中科院分区：
文献类型：
--
作者：
Ticao Zhang;S. Mao
端到端拥塞控制作为确保用户之间有效和公平共享网络资源的最重要机制之一，已经被广泛研究了30多年。随着未来网络的日益复杂，传统的基于规则的拥塞控制方法将变得低效甚至无效。受机器学习（ML）在解决大规模和复杂问题方面取得的巨大成功的启发，研究人员开始将注意力从基于规则的方法转移到基于ML的方法。本文对机器学习在端到端拥塞控制领域的最新应用进行了综述。在本调查中，我们首先简要回顾拥塞控制和机器学习之间的关系。然后回顾最近将机器学习应用于拥塞控制的工作。这些工作要么帮助代理做出智能拥塞控制决策，要么实现增强的性能。最后，我们强调了一系列现实挑战，并指出了未来可能的研究方向。
End-to-end congestion control has been extensively studied for over 30 years as one of the most important mechanisms to ensure efficient and fair sharing of network resources among users. As future networks are becoming more and more complex, conventional rule-based congestion control approaches tend to become inefficient and even ineffective. Inspired by the great success that machine learning (ML) has achieved in addressing large-scale and complex problems, researchers have begun to shift their attention from the rule-based method to an ML-based approach. This article presents a selected review of the recent applications of ML to the field of end-to-end congestion control. In this survey, we start with a brief review of the relationship between congestion control and ML. We then review the recent works that apply ML to congestion control. These works either help the agent to make an intelligent congestion control decision or achieve enhanced performance. Finally, we highlight a series of realistic challenges and shed light on potential future research directions.