Two-Stage Cooperation Multiobjective Evolutionary Algorithm Guided by Constraint-Sensitive Variables.

Area of Science:

• Engineering Optimization
• Computational Intelligence
• Evolutionary Computation

Background:

• Constrained multiobjective optimization problems (CMOPs) are prevalent in engineering.
• Existing constrained multiobjective evolutionary algorithms (CMOEAs) often fail to differentiate variable importance in constraint violation.
• There is a need for algorithms that effectively handle constraint-sensitive variables.

Purpose of the Study:

• To propose a novel two-stage cooperation multiobjective evolutionary algorithm (CV-TCMOEA) guided by constraint-sensitive variables.
• To address the limitations of existing CMOEAs in handling variables with varying degrees of constraint influence.
• To develop an adaptive strategy for updating individuals based on variable sensitivity.

Main Methods:

• A two-stage approach involving an auxiliary problem and a cooperative search.
• Classification of decision variables into constraint-sensitive and constraint-insensitive types.
• A variable-type-guided cooperative individual update strategy employing multistrategies.

Main Results:

• The proposed CV-TCMOEA demonstrated superior performance compared to seven state-of-the-art CMOEAs.
• Effectiveness validated across 28 benchmark functions and 10 engineering problems.
• The algorithm successfully addressed the challenge of constraint-sensitive variables.

Conclusions:

• The CV-TCMOEA offers an effective mechanism for handling constraint-sensitive variables in CMOPs.
• The proposed strategy enhances the performance and robustness of evolutionary algorithms for engineering applications.
• This work advances the field of constrained multiobjective optimization through adaptive variable handling.