The ant colony optimisation class of meta-heuristics simulate the ability of ant colonies to determine shortest paths to food. In this paper, we investigate strategies for achieving a balance between intensification and diversification search phases apart from that implicit in the algorithm itself. This is achieved by continually varying the sensitivity with which ants(agents) perceive pheromone trails. This has the effect of allowing the ants to both refine the search in promising regions and to escape explored areas of the search space. We examine four strategies using the well-known travelling salesman problem. The results indicate that while the explicit intensification/diversification schemes occasionally outperform the standard ant colony optimisation, the algorithm's inherent intensification/diversification characteristics are often sufficient to and good quality solutions.
|Publication status||Published - 1 Dec 2002|