Call admission control in ATM networks based on evolutionary neural networks

In this paper an evolutionary neural network (ENN) based call admission control (CAC) algorithm for asynchronous transfer mode (ATM) networks is proposed. The ENN is used as a controller to decide whether a new call set up request with some quality of service requirement will be accepted or rejected. The control method is applicable to heterogeneous traffic sources. The designed model has several new features that benefit its efficiency and performance. The ENN dimension is independent of the number of traffic classes. The new evolutionary algorithm is proposed not only to overcome the local extreme value problem in conventional neural network training algorithms but also to simplify the neural network structure and enhance the capabilities of generalization. Simulation results show that our ENN based control method is more robust and accurate than other conventional control methods.