Estimation of the blocking probabilities in an ATM network node using Artificial Neural Networks for Connection Admission Control

摘要

The aim of this paper is an application of neural networks to Connection Admission Control (CAC) in an ATM network. We propose a method to estimate if, and for how long, an ATM node is in a state of congestion. We show that estimations with neural networks are accurate, need a small observation of the traffic and can therefore react quickly to traffic changes.