New approaches to coevolutionary worst-case optimization

Abstract

Many real-world optimization problems involve uncertainty. In this paper, we consider the case of worst-case optimization, i.e., the user is interested in a solution's performance in the worst case only. If the number of possible scenarios is large, it is an optimization problem by itself to determine a solution's worst case performance. In this paper, we apply coevolutionary algorithms to co-evolve the worst case test cases along with the solution candidates. We propose a number of new variants of coevolutionary algorithms, and show that these techniques outperform previously proposed coevolutionary worst-case optimizers on some simple test problems. © 2008 Springer-Verlag Berlin Heidelberg.