Application of Neural Networks in Solving Some Problems in Modern Telecommunications

Modern telecommunication network is characterized by intensive growth and enormous capabilities. A huge number of users, types of services offered, and permanent trend of increasing the transfer rate make very difficult, or even impossible, to model, predict and control such a traffic. The paper describes the possible use of neural networks (NNs) in solving some problems arising in modern telecommunications. The main attention is devoted to applications of NNs in the traffic control: the scheduling in ATM node, the packet switch control, and the routing protocols, particularly for regular structure networks such as mesh topologies. The cellular neural network (CNN), according to its structure, is proposed for arbitration between the packets that are to be deflected or buffered in mesh topology networks. The parallel analog processing embedded in the CNN leads to very fast and optimal routing.