Multi-objective optimization (MOO) algorithms often use external archives to keep track of the Pareto- optimal solutions. Vector evaluated particle swarm optimization (VEPSO) is one such algorithm. In contrast to other MOO algorithms, VEPSO does not clearly define how to implement the archive. In this paper, the performance of various archive implementations, as found throughout the literature, are evaluated using the well-known Inverted Generational Distance (IGD) measure. A new archive implementation based on the hypersurface contribution is proposed and evaluated. The results show that overall the well-known crowding distance archive outperformed all other archive implementations. The hypersurface contribution archive also showed promise. Finally, it is shown that the distance metric and nearest neighbor archives perform worse than even the random archive.
