基于DRNN的ATM网络拥塞控制及仿真

ATM Network Congestion Control Based on DRNN and its Simulation

提出一种在用户-网络接口处利用对角递归神经网络(DRNN)作为自适应预测器,实现ATM网络自适应拥塞控制的模型。当DRNN预测下一时刻缓冲区中的信元数超过阈值时,控制器产生一个反馈控制信号减小信源进入网络的信元速率以避免拥塞发生。用话音和图象信源所作的仿真表明本文提出的模型较基于常规前馈网络的模型具有系统结构简单、控制效果好、实时性好等优点。

This paper presents an adaptive congestion control model in ATM networks at the user to network interface by using a diagonal recurrent neural network (DRNN) as an predictor. When DRNN predicts that the number of cells in buffer exceeds the threshold limit in the next time cycle, a control signal is generated by the controller to throttle arrival cell rate. Simulations of voice and video sources show that the presented model is simple in system construction and good in performance and real-time.