Neural adaptive congestion control for broadband ATM networks

The authors investigate the application of neural networks to adaptive congestion control in broadband ATM networks. A neural control scheme is proposed which is a direct application of the backpropagation neural networks with those modifications required to pose the problem in the framework of a general quality-of-service control. The learning algorithms regulating traffic loads to meet performance requirements are described and validated. To illustrate the present scheme´s ability to control, three examples of networks consisting of simple dynamic queuing models are studied through simulations.<>