基于智能预测控制的网络拥塞主动队列管理算法研究

An active queue management algorithm for network congestion based on intelligent prediction control

路由队列管理是保证网络性能、避免网络拥塞的重要手段,目前采用的主要队列管理方法为被动式队列管理,同时主动式队列管理已经成为近来的主要研究热点.随机早侦测(RED)作为最早提出的主动队列管理方法,更获得了普遍的关注.使用严格的数学模型来描述由端系统和网关组成的系统,并进行队列管理性能分析.提出一种采用快速广义预测控制的RED控制器(FGPC-RED控制器),进行网络拥塞控制的研究.介绍了系统的结构及系统的辨识,并通过仿真证明了FGPC算法在路由队列管理中应用的可行性,可以有效控制队列长度,避免路由拥塞及减小往返延迟.

Queue management is an important means to improve the performance of a network and avoid network congestion. At present, passive queue management dominates the queue management field. Recently active queue management （AQM） has been getting more and more attention, along with the random early detection （RED） algorithm, the first proposed AQM algorithm. In order to analyze a system made up of clients and a gateway, a strict mathematical model was developed to describe the system. A kind of RED controller with fast generalized prediction control was included to analyze congestion control. The system＇s architecture and identification are discussed in detail. A series of simulations were made, showing that fuzzy generalized predictive control （FGPC） is feasible in queue management and effective in controlling queue length, avoiding network congestion, and reducing round trip delay.