东北大学学报（自然科学版）

东北大学学报(自然科学版) ›› 2008, Vol. 29 ›› Issue (1): 17-20.DOI: -

• 论著 • 上一篇    下一篇

基于强化学习方法的ATM网络ABR流量控制

李鑫;井元伟;任涛;张阳;   

  1. 东北大学信息科学与工程学院;东北大学信息科学与工程学院;东北大学信息科学与工程学院;沈阳市东陵农电局 辽宁沈阳110004;辽宁沈阳110004;辽宁沈阳110004;辽宁沈阳110015
  • 收稿日期:2013-06-22 修回日期:2013-06-22 出版日期:2008-01-15 发布日期:2013-06-22
  • 通讯作者: Li, X.
  • 作者简介:-
  • 基金资助:
    国家自然科学基金资助项目(62074009);;

Reinforcement learning approach to ABR traffic control of ATM networks

Li, Xin (1); Jing, Yuan-Wei (1); Ren, Tao (1); Zhang, Yang (2)   

  1. (1) School of Information Science and Engineering, Northeastern University, Shenyang 110004, China; (2) Shenyang Dongling Rural Electric Bureau, Shenyang 110015, China
  • Received:2013-06-22 Revised:2013-06-22 Online:2008-01-15 Published:2013-06-22
  • Contact: Li, X.
  • About author:-
  • Supported by:
    -

RichHTML

1

PDF (PC)

762

摘要： 针对异步传输模式(ATM)网络的拥塞问题,将强化学习方法应用于拥塞控制器的设计之中.该方法不依赖于网络的数学模型和先验知识,而是通过试错和与环境的不断交互获得知识,从而改进行为策略,具有自学习的能力.控制器通过调节可用比特速率(ABR)业务发送数据的速率,使网络中可能发生拥塞的节点的缓冲器队列长度逼近给定值,从而避免拥塞的发生,保证网络的稳定运行.通过一系列仿真实验验证了该方法的有效性.

关键词: ATM网络, ABR业务, 拥塞控制, 流量控制, 强化学习

Abstract: The reinforcement learning approach is applied to the design of controller to solve the congestion problem in ATM (asynchronous transfer mode) networks. This approach does not rely on the mathematic model and priori-knowledge of network, but acquires the knowledge through trial-and-error method and interacts with environmental conditions to improve its behavior strategy. So, it has the self-learning ability and the queue length of buffer at bottleneck node thus approximates to the set value by readjusting the source traffic rate in the ABR (available bit rate) service. The stability of the system is therefore provided and able to avoid possible occurrence of congestion. Simulation results show the effectiveness of the approach proposed.

中图分类号: