中文摘要：

针对Internet系统，通过对流体流模型的分析，提出了一个新的预测模型。该模型形式简单，参数的计算相对容易，并且能根据当前的网络情况有效的预测拥塞窗口的变化。结合动态矩阵控制（Dynamic Matrix Control ，DMC ）理论，提出了一种新的主动队列管理算法---DMCAQM算法，给出了DMCAQM 的详细设计过程，稳定性分析和参数选取原则。大量不同网络环境的仿真实验表明DMCAQM 算法是有效的。与PI、RaQ和REM等算法相比较，DMCAQM 有收敛速度快、队列抖动小的优点。同时，由于DMCAQM 的采样间隔相对较大，而算法实现简单，所以计算量小，占用的路由器资源更少。

英文摘要：

Through the analysis of the fluid-based model ,a new predictive model is proposed for Internet system .This model is simple ,and the way of selecting parameters is relatively easy .It′s able to predict the change of congestion window according to the network environment .Based on the theory of Dynamic Matrix Control （DMC） ,a new Active Queue Management （AQM） algo-rithm is proposed ,called DMCAQM .The design details ,the stability analysis and the principles of choosing parameters are also pre-sented .The results of simulations under different network environments demonstrate that DMCAQM is effective in controlling the queue length to the desired value .In comparison to other AQM algorithms ,such as PI ,RaQ and REM ,DMCAQM have faster re-sponse and lower queue fluctuations .In addition ,DMCAQM has larger interval while with simple implementation ,so it use less re-source on the routers .