Multi-objective worst case optimization by means of evolutionary algorithms

Many real-world optimization problems are subject to uncertainty. A possible goal is then to find a solution which is robust in the sense that it has the best worst-case performance over all possible scenarios. However, if the problem also involves mul- tiple objectives, which scenario is “best” or “worst” depends on the user’s weighting of the different criteria, which is generally difficult to specify before alternatives are known. Evolutionary multi-objective optimization avoids this problem by searching for the whole front of Pareto optimal solutions. This paper extends the concept of Pareto dominance to worst case optimization problems and demonstrates how evolu- tionary algorithms can be used for worst case optimization in a multi-objective setting