Congestion avoidance in large supra-high-speed packet switching networks using neural arbiters

The authors propose artificial neural network (ANN) based arbitration for fast and intelligent processing of transit packets and assigning network access to the local packets in large supra-high-speed packet switched telecommunication networks. Use is made of: (i) S. Grossberg´s (Ed., Neural Networks and Natural Intelligence, MIT press, Cambridge, MA, USA, 1989) adaptive resonance theory based neural network, a two-layer ANN along with a fuzzy logic based preprocessor to maintain distributed cut-through links between the communicating users; and (ii) a buffer stealer (another ANN based arbiter) to throttle local source access to the network during heavy load conditions. With the incorporation of these additional complexities at the nodal structure, it is possible to most efficiently utilize the available network resources under all possible, i.e., both certain and uncertain, operating conditions. The proposed techniques are applied in a 4×4 Manhattan street network. The main objective is to avoid network layer level congestion