Extending MOEA/D to Constrained Multi-objective Optimization via Making Constraints an Objective Function

Abstract

Multiobjective Evolutionary Algorithm based on Decomposition (MOEA/D) is effective for solving multi-objective optimization problems. However, in real-world applications, problems with imposed constraints are common. Therefore, research on Constraint Handling Techniques (CHTs) has been done. CHTs focus on improving search performance by utilizing infeasible solutions. Multi-objective-based CHTs are effective in promoting convergence and diversity in solution sets, but existing CHTs for MOEA/D have limitations in terms of flexibility and extensibility (e.g., the scalarization function to be used). To overcome this, this paper proposes a CHT using two sets of weight vectors to make constraints an objective function. The proposed method is flexible and can be used in any MOEA/D variant. It is incorporated into a basic MOEA/D and its effectiveness is demonstrated by comparing it with existing constrained MOEA/D on 2- and 3-objective benchmark problems.