Visual Exploration of the Effect of Constraint Handling in Multiobjective Optimization.

Constraint handling in multiobjective optimization is more complex than in single-objective optimization, where the values of the objective and constraints are easier to combine. To gain insight into the characteristics of constraint handling techniques (CHTs) for multiobjective optimization, we explore their effect independently from search methods. We regard CHTs as transformations that alter the problem landscape and visualize these modified landscapes. This helps us predict potential strengths and weaknesses for search methods. We then use a simple local search technique to test our predictions. Results of the experiments with six CHTs applied on 12 test problems show specific properties of the studied CHTs that can help us devise better CHTs in the future, as well as find suitable search methods for them.