一种基于模糊神经网络的可靠流量控制模型

A Reliable New Traffic Control Model Based on Fuzzy Neural Network

为控制P2P流量,本文从数据缓冲区使用的实时状态出发,提出了一种基于模糊神经网络的拥塞控制模型,该模型把缓冲区划分为两个队列分别存放P2P和非P2P的数据包,通过模糊神经网络预测评估缓冲区队列的拥塞状况,并建立一个评估函数对各队列的空间分配作出指导,使得能够控制各队列的拥塞状况,并动态的调整缓冲区队列的分配,在缓冲区溢出前主动丢包,避免缓冲区锁定。模拟实验的结果表明,该模型在保证网络资源分配的公平性方面取得了较好的效果,它降低了数据包排队延时和丢包率,提高了路由器处理网络拥塞的能力。

In this paper, we present a kind of congestion control model which based on fuzzy neural network（FNN） from the practical status of data buffer, for the sake of controlling P2P traffic. This model divides data buffer into two queues which store P2P data packets and non-P2P data packets respectively. It forcasts and evaluates conditions of buff- er queues through FNN as well as guides space allocation of each queue through constructing a evaluation function. Thus, this model is able to control congestion condition of each queue and resize allocation of queues in the buffer auto- matically, then it can avoid lock-out of the buffer by actively dropping packets before the buffer is overflow. Results from simulation experiments show that this model has gained better effect in ensuring network resource allocation equi- table, itcan also decreases the delay of packet queuing and the dropping ratio. Thus,it improves the ability of routers in dealing with network congestion.