A Cooperative Evolutionary Framework Based on an Improved Version of Directed Weight Vectors for Constrained...

Area of Science:

• Optimization
• Computational Intelligence
• Multi-objective Optimization

Background:

• Existing constraint violation measures in CMOPs may inaccurately assess solution proximity to feasibility.
• A gap exists in benchmark problems for CMOPs with deceptive constraints.

Purpose of the Study:

• Introduce a new class of problems: Constrained Multiobjective Optimization Problems with Deceptive Constraints (DCMOPs).
• Propose a cooperative framework with improved directed weight vectors to address DCMOPs.

Main Methods:

• A cooperative framework with two switchable phases: exploration and Pareto-optimal solution finding.
• Utilizes two subpopulations for information exchange and an infeasibility utilization strategy.
• Employs an improved version of directed weight vectors for solving DCMOPs.

Main Results:

• The proposed method demonstrates superior performance compared to existing algorithms on most DCMOPs.
• Achieves significant improvements in reliability and stability for finding optimal solutions.
• Effectively handles deceptive constraints in multiobjective optimization.

Conclusions:

• The introduced cooperative framework is effective for solving DCMOPs.
• The method provides a reliable and stable approach for obtaining well-distributed optimal solutions.
• Addresses limitations in current constraint violation measures for specific optimization problems.