ABR流量控制-一种神经网络方法

本文提出ATM网络中ABR业务流量控制的一种基于神经网络方法。利用神经网络对缓冲队列的长度及其变化和相应可能的信元丢失率进行预测，通过自适应地调整业务源端的发送速率，从而更准确地确定RM信元中相应域的显式速率值。同传统静态反馈方法相比，本方法不仅提高了网络资源利用率，而且改善了信元丢失率。另外，对在具有CBR背景业务情况下本方法的性能也进行了讨论，仿真结果证明了神经网络方法的有效性。

ABR Flow Control : A Neural Network Approach

A neural network approach to ABR flow control in ATM networks is proposed. By this approach, the queue length with its variation and the possible cell loss ratio are predicted, accordingly the source rate is regulated in an adaptive way. Therefore the explicit rate in the corresponding field of the RM cells can be determined more appropriately. This approach performs better than traditional static feedback control either in resource utilization or cell loss ratio. The efficiency of this algorithm under CBR traffic condition is also discussed. Simulation results show that the neural network is effective.