Blended Ranking to Cross Infeasible Regions in ConstrainedMultiobjective Problems
Nam Mi Young · 2006
We present a multiobjective evolutionary algorithm designed to reliably cross infeasible regions of objective space and find the true constrained Pareto front, which may lie across multiple disconnected feasible regions. By blending an individual's rank in objective space with its rank in constraint space, some infeasible solutions may be selected over some feasible solutions, allowing the population to traverse infeasible regions smoothly. Results from artificial benchmark problems qualitatively illustrate this behaviour, in contrast to NSGA-II which must cross infeasible regions in a single generation