Adaptive technique for ATM call admission and routing control using traffic prediction by neural networks

This paper discusses a technique for call admission and routing control, based on a global quality function, which is dependent on the allocated bandwidth, the free network capacity and the call rejection rate, and incorporates quality of service functions, predicted by neural networks. The superior capability of this technique to support admission and routing decisions, according to the characteristics of the traffic generated by admitted calls, is demonstrated by simulation results carried out using suitable traffic and network models, which are equally discussed. It is also shown that the proposed technique, being based on several observed traffic parameters, offers better results than methods based only on declared bandwidth parameters.