NEURAL NETWORK MODELS FOR OPTIMAL ROUTING IN COMPUTER NETWORKS

Abstract

The routing of packets from source to destination is an important issue in the design of packet-switched computer networks, where the goal is to minimize the network wide average time delay. The routing algorithms rely heavily on the shortest path computations that have to be carried out in realtime. This dissertation addresses the application of neural networks to the optimal routing problem. Three neural network models are compared. Their performance in giving optimal routes is analysed through simulation results by selecting three different communication network topologies. The neural network models compared are Lee-Chang model, Zhang-Thomopoulos model and Mustafa-Faouzi model, all based on Hopfield neural networks. All-through Lee-change model gives multiple optimal, suboptimal routes simultaneously it is not fool proof in giving all optimal routes. But Mustafa-Faouzi model is found to be giving all optimal routes. The performance of these models in finding the multiple optimal routes simultaneously and the conditions there in are analysed through simulation results. Other factors like divergence problems, computational power requirement have also been examined.