Proceedings of the International Workshop on Applications of Neural Networks to Telecommunications

A Hybrid Admission Control Scheme for Broadband ATM Traffic

Ernst Nordström

Department of Computer Systems, Uppsala UniversityBox 325, S–751 05 Uppsala, [e-mail]

Abstract

Link admission control (LAC) in broadband ATM networks is based on evaluation of expected traffic performance. The traditional LAC approach relies on approximate analytical performance models, and can lead to an over controlled network. This paper presents a hybrid LAC scheme which uses a multi layer perceptron (MLP) to refine the performance estimate of a traditional analytical approximation. An accurate but complex traffic performance model is used to derive MLP target data. The numerical results show that the MLP can improve the bandwidth efficiency. ...