动态矩阵主动队列管理算法

针对Internet网络这种大延迟,变化复杂,存在严重干扰的系统,基于约束模型预测控制理论,提出了一种新的主动队列管理算法(DMAQM算法).首先根据预测模型和状态估计器,预测瓶颈链路队列的未来动态特性,它是确定丢包率的基础;然后将拥塞控制的控制要求转化为优化目标函数,在线求解优化问题得到丢包率.DMAQM算法通过滚动优化来适应网络环境的变化,提高了算法鲁棒性;同时在求取丢包率时显式地考虑了网络中存在的约束.仿真结果显示DMAQM算法控制性能优于RED算法,在满足约束的前提下,具有较强的抗干扰能力,同时具有较好的稳定性和鲁棒性.

Active queue management algorithm with dynamic matrix control

We propose a novel active queue management algorithm based on the dynamic matrix control(DMAQM) for the Internet system with large time-delay, complex variations and detrimental disturbances. By the predictive model and state estimator, the future queue length in data buffer, which is the basis for determining the packet-loss rate, is predicted. Furthermore, the control requirements are converted to the optimal control objectives, and then, the packet-loss rate is obtained by solving the optimization problem online. The DMAQM algorithm adapts to the varying network environment and improves the robustness by moving the optimization horizon, and handles network constraints in the process for determining the packet-loss rate. Finally, the performances of DMAQM are evaluated through a series of simulations. The simulation results show that the DMAQM algorithm is superior to the random early detection(RED) algorithm in stability, disturbance rejection, and robustness.