ABSTRACT
Application of neural networks to asynchronous transfer mode (ATM) network traffic control has been proposed. ATM is a key technology for building a broadband integrated services digital network (B-ISDN), which gives users universal high-speed communication channels for all kinds of communications services. One important issue in ATM networks is the design of an efficient traffic control architecture that guarantees quality of service (QoS) for all network users. However, in future networks for multimedia communication, it will be impossible to model user activity and to exhaustively analyze the network traffic dynamics by mathematical calculations and computer simulations. A neural network is an important technology for deriving an unknown nonlinear function for estimating QoS from the network status by real-time training. The advantage of this method is that the QoS can be accurately estimated without detailed user action models or knowledge about the switching system architecture. This section gives an overview of neural network applications in ATM traffic control, and describes ATM connection admission control as a typical example.
