基于速率的ATM流量反馈控制--一种基于神经网络预测的方法

在传统的流量控制方法中,基于速率的ATM流量反馈控制往往是在拥塞发生后采取控制措施.这样造成的结果是必然造成拥塞部分业务的服务质量(QoS)下降.本文利用神经网络能近似计算多变量的非线性函数,及自学习、自适应和大规模并行处理的特点,采用基于预测的原理上对信源反馈控制,实验表明这种方法对于降低信元丢失率(CLR),提高业务服务质量有明显效果.

The Rate-based Traffic Feedback Control in ATM Networks--A Neural Network Approach

In the conventional traffic control, the rate-based feedback control always takes the measures after the congestion has taken place in ATM networks. This will result in the reduction of QoS in the part of congestion during the up-load periods . The paper utilizes the neural networks features, such as calculating the non-liner functions, self-study, self-adaptation and parallel-processing on large scale. And it presents the neural network prediction based method for rate-based feedback control in ATM networks. In all examples presented, the feedback control which based on the proposed neural network prediction significantly outperforms the conventional method in lowering down CLR and improve QoS promoting.