用于队列管理的模糊增益神经元自适应控制器

针对具有参数时变及非线性特性的网络拥塞控制系统，提出了一种模糊增益神经元主动队列管理算法(FNAQM). 采用路由器队列长度及数据流速作为拥塞度量，在检测当前拥塞信息的同时，预测未来拥塞的状况. 结合神经元控制和模糊控制的优点，利用单神经元计算数据包标记概率，采用有监督的Hebb学习规则在线调整加权系数. 设计的模糊控制器可动态调整神经元增益，能获得更好的控制性能. FNAQM具有结构简单、易于实现、自适应能力强等优点. 仿真实验结果表明，FNAQM能快速将队列调整至目标值，并维持较小的队列抖动，对动态数据流和非响应流具有良好的鲁棒性.

Adaptive Neuron Controller with Fuzzy Self-Tuning Gain for Queue Management

In light of the congestion control system with timevarying parameters and nonlinear property, a neuron control algorithm with fuzzy selftuning gain (FNAQM) is proposed for active queue management. Both queue length and traffic rate are employed as congestion indicators which detect both current and incipient congestion states. Combining the advantages of neuron control and fuzzy control strategies, the endtoend mark probability is calculated by the neuron controller, in which the weights are adjusted online by supervisory Hebb learning rule. Additionally, fuzzy logic control is used to tune the gain of the neuron dynamically for improved network performance. The proposed scheme exhibits good adaptability and selflearning ability, being simple in form and easy to implement. Simulation in network simulator2(NS2) demonstrates that FNAQM can quickly stabilize the queue length to the target with small jitter, and shows strong robustness against dynamic traffics and nonresponsive flows. 