ATM connection admission control using pRAM based artificial neural networks

As they learn from observed data, neural networks have found many applications in communication networks. Here, a neural network for connection admission control (CAC) based on the probabilistic-RAM (pRAM) neuron model is presented. Because pRAM neural networks can be implemented in hardware easily and at a low cost, they make excellent controllers in the ATM environment. The performance of such a controller is analysed through simulations, and the results are compared with the Equivalent Capacity of Connections CAC algorithm. The results show that, the proposed hardware-based controller guarantees the required quality of service (QoS) and at the same time provides an improvement in network utilization.