A new approach to dynamic network routing using Omicron Ant Colony algorithm

Abstract

The paper introduces a new approach to network routing using the adaptive learning techniques of Ant Colony Optimization (ACO) framework. The proposed algorithm is based on the two ACO algorithms AntNet and Omicron Ant Colony Optimization (OA). In principle, the algorithm uses the mobile agents (ants) to collect information about the network. The ants exchange this collected data using stigmergic communication. In an attempt to decrease the packet delay and improve throughput new methods and data structures have been introduced. The algorithm adopts OA's approach to initialize and update the pheromone values. In addition, it introduces solution tables to hold a set of good solutions at any given time. Further, to select the next-hops the algorithm includes methods, namely, deterministic dual step method and roulette-wheel selection. The algorithm is simulated on the NSFNET topology using ns-2. On comparing the delay and throughput values with standard AntNet algorithm, a considerable improvement is observed, thereby denoting an enhanced efficiency of routing.