基于队列敏感性的无线接入网络拥塞控制算法

摘要

Since the wireless access network is subject to the effects of strong nonlinearity, large delay and random link loss, the classical Active Queue Management (AQM) has the problems of slow convergence rate and long response to queue in the actual control process. By analyzing the characteristics of Random Exponential Marking (REM) algorithm in the wireless access network, this paper proposed a new congestion control method of wireless access network based on sensitive queue, which improved the original price model of REM and overcame the insensitivity of price to the change of queue size. Moreover, a single neuron was utilized to optimize the parameters. Finally, the proposed algorithm was compared with REM, PI on NS2 simulation platform. The simulation results show that the proposed algorithm has fast convergence and strong robustness.%由于无线接入网络存在强非线性、大时延以及随机链路丢包等因素,导致经典主动队列管理(AQM)算法在实际控制时存在队列收敛速度慢、响应时间长等问题.通过分析随机指数标记(REM)算法在无线接入网中的特点,在原先REM价格模型的基础上对其进行了改进,以队列误差的平方项来克服价格对队列变化不敏感的缺陷,从而提出了一种基于队列敏感性的无线接入网络拥塞控制算法,并利用单神经网络对其参数进行了优化.最后,通过NS2仿真平台对所提算法与REM、PI算法进行对比,实验表明所提算法拥有队列收敛快、鲁棒性强的优点.