Investigation ofant colonyalgorithm in multiple traffic flow environments

Abstract

The conventional approaches to routing and bandwidth allocation, the two major components of traffic engineering, have proved insufficient to address QoS requirements of flows while optimizing utilization for complex communication networks. In this paper we consider ant colony algorithms to address this problem. Our studies show that the ant-based routing models are sensitive to initial parameters settings. Only careful adjustments of these initial parameters results in an acceptable convergence behavior. The robust behavior of the real ant compared to the routing algorithms derived from it inspires us to investigate the reasons behind the shortcomings of these algorithms. We present results from an in-depth study of ant behavior in a quest for a robust algorithm. In this work we consider a realistic environment in which multiple source-destination flows compete for resources. We study the routing and load balancing behavior that emerges and show how the behavior relates to analytical approaches. We show the results using simulations in OPNET and derive recommendations on the improvement of the ant-like algorithm