一种基于PID控制的自适应随机早期检测算法研究

Adaptive random early detection algorithm based on PID control principle

为了解决随机早期检测(RED)算法中参数调整困难、收敛性差等问题,提出了一种基于比例积分微分(PID)控制的RED改进算法(PID-RED).算法建立了传输控制协议(TCP)/主动队列管理(AQM)负反馈控制模型,利用经典控制理论中的稳定收敛理论求取PID控制系数,根据实际平均队列长度与预期队列长度的偏差值动态调整RED最大丢包率,从而自适应地调整RED参数.仿真结果表明,PID-RED算法具有更快的收敛速度和更小的队列抖动,提高了主动队列管理策略的鲁棒性.

In order to overcome the problems with random early detection （RED） such as difficulty in parameter tuning and poor convergence, a new RED mechanism based proportional-integral-derivative （PID） control principle was proposed to improve the performance of RED. By establishing a negative feedback transfer control protocol/active queue management （TCP/AQM） model, the new PID-RED algorithm can tune the maximum drop rate of RED according to the difference between actual average queue length and expected queue length. The PID control parameters that can tune the parameter of RED were calculated with the convergence theory in the traditional theory of cybernetics. Simulation results verify that the PID- RED algorithm can achieve faster convergence and smaller queue oscillation than RED and so can improve the robustness of AQM.