Multi-Objective Evolutionary Algorithms’ Performance in a Support Role

This paper presents a diagnostic assessment study, evaluating five leading multi-objective evolutionary algorithms (MOEAs) on their effectiveness, efficiency, reliability, and controllability on four different formulations of the same benchmark conceptual design problem, using the same underlying model. This assessment entails a broad sampling of the parameter space of each MOEA, for each problem formulation, requiring millions of optimization runs and trillions of model evaluations. The results of this assessment show the strengths and limitations of these MOEAs, establishing the Borg MOEA as a leading algorithm.