VSD-MOEA: A Dominance-Based Multiobjective Evolutionary Algorithm with Explicit Variable Space Diversity Management.

Area of Science:

• Optimization Algorithms
• Computational Intelligence
• Evolutionary Computation

Background:

• State-of-the-art Multiobjective Evolutionary Algorithms (MOEAs) often prioritize objective function space diversity.
• Existing MOEAs frequently overlook the importance of decision variable space diversity.
• This oversight can limit the overall quality and effectiveness of optimization algorithms.

Purpose of the Study:

• To demonstrate the benefits of explicitly managing decision variable space diversity in MOEAs.
• To introduce a novel algorithm, the Variable Space Diversity-based MOEA (vs-d-MOEA), that considers both decision variable and objective function space diversity.
• To improve the quality of MOEAs by balancing exploration and intensification.

Main Methods:

• The proposed vs-d-MOEA explicitly incorporates diversity metrics from both decision variable and objective function spaces.
• A novel density estimator is utilized to manage objective function space diversity.
• The algorithm adaptively balances exploration and intensification based on diversity information, prioritizing variable space diversity initially and objective space diversity later in the process.

Main Results:

• The vs-d-MOEA demonstrated superior performance compared to state-of-the-art MOEAs on benchmark problems with two and three objectives.
• The algorithm exhibited significantly improved results when evaluated using objective space metrics.
• A more stable and robust behavior was observed in the proposed method.

Conclusions:

• Explicitly managing decision variable diversity is crucial for enhancing MOEA performance.
• The novel vs-d-MOEA effectively balances exploration and intensification by considering both variable and objective space diversity.
• The proposed approach offers a more stable and robust optimization solution for multiobjective problems.