摘要
为了改进比例积分微分(PID)控制在Ad Hoc网络主动队列管理(AQM)中的动态性能,优化PID控制参数的整定,该文提出了一种基于反向传播神经网络的PID拥塞控制AQM方案。该文将Ad Hoc网络的分组丢弃分为拥塞丢弃和无线丢弃,考虑分组的到达与丢失为流体,推导了拥塞窗口和队列长度的随机微分关系,通过小扰动线性化理论,获得Ad Hoc网络AQM拥塞控制模型。根据该模型,设计了基于反向传播神经网络(BPNN)的PID队列控制器,该算法可以根据网络状况对控制器PID系数进行自适应的调整。MATLAB和网络模拟器(NS)仿真表明,在突发流、链路容量及时延时变的Ad Hoc网络中,新算法在收敛速度和队列抖动上优于PID。
To improve the dynamic performance of the proportional integral derivative（PID） control and the optimize PID control parameter tuning in the Ad Hoc network active queue management（AQM）,this paper proposes an AQM scheme about the PID congestion control based on the back propagation neural network（BPNN）.Packet losses are divided into congestion loss and wireless loss in Ad Hoc network.Taking the arrival and loss packets as fluids,the stochastic differential relation between congestion windows and queue lengths is deduced.An AQM congestion control model in Ad Hoc network is proposed through small perturbations and linearized theory.The PID queue controller based on the BPNN is designed on the basis of the model.The algorithm can make adaptive adjustments to the controller PID coefficients according to the network situation.MATLAB and network simulator simulations indicate that the new algorithm is superior to the PID in rapidity of convergence and queue oscillation under the Ad Hoc network sudden flow,time-varying link capacity and delay condition.
出处
《南京理工大学学报》
EI
CAS
CSCD
北大核心
2010年第5期628-635,共8页
Journal of Nanjing University of Science and Technology
基金
江苏省自然科学基金(BK2007593)
关键词
无线自组网
拥塞控制
反向传播神经网络
比例积分微分
Ad Hoc network
congestion control
back propagation neural network
proportional integral derivative
