引用本文:王萍,于鸿昶,刘奇芳,陈虹.主动队列管理的约束H∞状态反馈控制方法[J].控制理论与应用,2013,30(11):1353~1359.[点击复制]
WANG Ping,YU Hong-chang,LIU Qi-fang,CHEN Hong.State-feedback constrained H-infinity control for active queue management[J].Control Theory and Technology,2013,30(11):1353~1359.[点击复制]
主动队列管理的约束H∞状态反馈控制方法
State-feedback constrained H-infinity control for active queue management
摘要点击 2626  全文点击 2804  投稿时间：2013-05-21  修订日期：2013-07-11
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2013.30509
  2013,30(11):1353-1359
中文关键词  主动队列管理  时域硬约束  H∞控制
英文关键词  active queue management  time-domain hard constraints  H-infinity control
基金项目  长江学者和创新团队发展计划资助项目(IRT1017); 吉林大学基本科研业务费资助项目.
作者单位E-mail
王萍 吉林大学 控制科学与工程系 wangping12@jlu.edu.cn 
于鸿昶 吉林大学 控制科学与工程系  
刘奇芳 吉林大学 控制科学与工程系  
陈虹* 吉林大学 控制科学与工程系  
中文摘要
      考虑Internet网络的时滞、物理量变化复杂、物理约束等因素, 将主动队列管理问题(AQM)描述为约束系统的干扰抑制问题, 运用约束H∞控制理论设计AQM鲁棒控制器. 首先将时变且不可准确测量的可用链路容量建模为已知名义常值加上未知时变干扰; 同时考虑网络中存在的物理约束, 利用双椭圆域方法将时域硬约束转化为一组LMI约束; 最后通过求解LMI约束的优化问题得到状态反馈增益, 解决网络拥塞控制系统的干扰抑制问题. 同时与随机早期检测(RED)和PI算法的仿真结果比较显示, 约束H∞状态反馈控制器降低了链路容量的不确定性对系统动态特性的影响, 提高了AQM算法的鲁棒性.
英文摘要
      We propose a robust active queue management (AQM) algorithm based on constrained H-infinity control for the Internet system with time-delay, complex variations and physical constraints. Firstly, the available link bandwidth is modeled as a nominal constant value, which is known to the link, plus a time-variant disturbance, which is unknown. Furthermore, the double-ellipsoid method is used to translate time-domain hard constraints into LMI conditions. Then, the problem of AQM algorithm can be converted into the disturbance attenuation problem of constrained systems, and the controller is designed using constrained H-infinity control theory. By solving an optimization problem with a set of linear matrix inequalities, we obtain the state feedback gain. The results show that the proposed approach is robust against the disturbance caused by the link capacity. The transient response and tracking capability of the proposed approach are superior to that of the RED and PI controller. The effectiveness of the proposed approach is validated via simulations in MATLAB and NS2 (network simulator, version 2).