基于自适应模糊神经网络PID控制器的主动队列管理算法

摘要

Active queue management (AQM) is a very important research area in congestion control. But the complexity and dynamic characteristic of the computer network causes the traditional PID control algorithm low adaptability to dynamic environment due to its fixed parameters. In order to overcome such shortcomings, intelligent control theory was introduced to congestion control research, and a new AQM algorithm called FAPIDNN was proposed. Fuzzy controller automatically computers the learning rate -n according to the current network state, and the neural network PID controller calculate the packet dropping probability based on the learning rate provided by the fuzzy controller. Simulation results show that FAPIPNN algorithm is superior to the presented PID controller on the queue stability, convergence speed and time delay.%主动队列管理(AMQ)是一种非常重要的拥塞控制的研究领域.但其自身的复杂性和计算机网络的动态特性,传统的PID控制算法由于其固定参数的方式、导致其对动态环境适应性低.为了克服这些缺点,通过对拥塞智能控制理论的研究,介绍了一个新的AQM算法FAPIDNN.模糊控制器是采用自动学习速率η的计算机,通过神经网络PID控制器根据当前网络状态和对网络数据包下降概率的计算基础上,网络速率自主学习的模糊控制器.仿真结果表明FAPIDNN算法在队列稳定、收敛速度和时间延迟上要优于PID控制器.