CALL ADMISSION CONTROL IN ATM NETWORK: A NEURAL NETWORK APPROACH

Abstract

The CCITT and domestic standard forums have proposed Asynchronous Transfer Mode (ATM) as transport and multiplexing technique across the User Network Interface of Broadband ISDN. By using ATM, Information flow is organized by fixed size packets called cells, cells are transmitted through virtual circuits and are self routed based on their header information. In ATM network, the precise characteristics of the source traffic are not known and the service quality requirements change over time, making it difficult to build an efficient traffic controller. The ATM call admission controller which uses back propagation neural network for learning the relationship between offered traffic and service quality, is considered. The neural network is adaptive and easy to implement, so this controller automatically adapts to new situation. A training data selection method called the "leaky pattern table method" is also considered to learn precise relations. The accuracy and efficiency of call admission control method is also evaluated by computer simulation of simple network model.