A Multiobjective Evolutionary Algorithm Based on Objective-Space Localization Selection.

Area of Science:

• Optimization Algorithms
• Computational Intelligence
• Multi-Objective Optimization

Background:

• Many-objective optimization problems often feature irregular Pareto fronts, posing challenges for existing evolutionary algorithms.
• Traditional methods require predefined weight vectors or complex indicator calculations, limiting their adaptability.
• Developing robust algorithms for irregular Pareto fronts is crucial for advancing optimization research.

Purpose of the Study:

• To propose a simple yet effective multiobjective evolutionary algorithm (EA) capable of handling irregular Pareto fronts.
• To design an EA that avoids the need for predefining weight vectors and calculating indicators during the search.
• To enhance the selection of promising search directions using local crowdedness information.

Main Methods:

• The proposed algorithm adaptively selects search directions based on crowdedness in local objective spaces.
• It employs a dynamic environmental selection process, removing poor-quality individuals.
• The selection mechanism probabilistically removes the worst among a few crowded individuals, preserving representative solutions.

Main Results:

• The algorithm was validated against seven state-of-the-art algorithms on complex problems with 2 to 15 objectives.
• Empirical results show strong competitiveness in performance and algorithm compactness.
• The proposed EA effectively handles diverse irregular Pareto fronts and varying numbers of objectives.

Conclusions:

• The developed EA offers a competitive and compact solution for multiobjective optimization problems with irregular Pareto fronts.
• Its adaptive strategy simplifies the optimization process by eliminating the need for weight vectors and complex indicators.
• The algorithm demonstrates robust performance across a wide range of objective numbers and problem complexities.