An Analysis of Algorithmic Components for Multiobjective Ant Colony Optimization: A Case Study on the Biobjective TSP

In many practical problems, several conflicting criteria exist for evaluating solutions. In recent years, strong research efforts have been made to develop efficient algorithmic techniques for tackling such multiobjective optimization problems. Many of these algorithms are extensions of well-known metaheuristics. In particular, over the last few years, several extensions of ant colony optimization (ACO) algorithms have been proposed for solving multi-objective problems. These extensions often propose multiple answers to algorithmic design questions arising in a multi-objective ACO approach. However, the benefits of each one of these answers are rarely examined against alternative approaches. This article reports results of an empirical research effort aimed at analyzing the components of ACO algorithms for tackling multi-objective combinatorial problems. We use the bi-objective travelling salesman problem as a case study of the effect of algorithmic components and their possible interactions on performance. Examples of design choices are the use of local search, the use of one versus several pheromone matrices, and the use of one or several ant colonies.