Ensuring fast adaptation in an ant-based path management system

Abstract

The Cross-Entropy Ant System (CEAS) is an Ant Colony Optimization (ACO) system for distributed and online path management in telecommunication networks. Previous works on CEAS have focused on reducing the overhead induced by the continuous sampling of paths. In particular, elite selection has been introduced to discard ants that have sampled poor quality paths. This paper focuses on the ability of the system to adapt to changes in dynamic networks. It is shown that not returning ants may cause stagnation as that tends to make stale states persist in the network. To mitigate this undesirable side-e.ect, a novel pheromone trail evaporation strategy, denoted Selective Evaporation on Forward (SEoF), is presented. By allowing ants to decrease pheromone trail values on their way forward, it enforces a local re-opening of the search process in space upon change when elite selection is applied. © Institute for Computer Sciences, Social-Informatics and Telecommunications Engineering 2010.