Performance of neural networks for call admission control in ATM systems

In this paper, the capability of neural networks to call admission control in asynchronous transfer mode (ATM) networks is investigated. The general problem of call admission control (CAC) and its formulation as a functional mapping are discussed leading to applications of learning algorithms to CAC problems. A modified cascade-correlation network, which combines typical backpropagation and cascade-correlation algorithms together, is used as call admission controller. Its performances are compared with those of typical backup. Simulation results of basic call admission models illustrate the applicability of the proposed controller