TSP optimisation using multi tour ants

Abstract

Ant colony optimisation has proved useful for solving problems that can be cast in a path length minimisation form, particularly the travelling sales person (TSP) problem. Finding good, if not optimal, solutions in a reasonable time requires a balance to be struck between exploring new solutions and exploiting known information about possible solutions already examined. A new algorithm in which individual ants each live long enough to explore multiple solutions is introduced. Results are presented that show that enabling ants to learn from their own prior experience in addition to the collective wisdom of the colony improves performance on two standard test TSP data sets and suggests that the algorithm may well be useful for the whole class of TSP problems.