Neuron based call admission control method for transport network of 3rd generation mobile systems

Due to the capability to serve a large number and various types of user traffic, ATM (asynchronous transfer mode) networks are going to turn to be strong candidate among the transport networks for third-generation mobile systems (UMTS, IMT200) in the immediate future. To guarantee high-quality communications for a lot of customers and efficient use of network resources the ATM network management has to contain appropriate call admission control (CAC) algorithms. However, the time specifications for CAC decision are very rigorous because of the handoff procedures coming from the terminal mobility. Choosing suitable network and user models for the CAC problem can be traced back to geometrical set separation. We propose a solution based on neural networks for the above problem. Thanks to the parallel operation of neurons and the small number of layers of the suggested network the strictest time requirements for CAC decision can be satisfied