A Dynamic Multiobjective Evolutionary Algorithm Based on Decision Variable Classification.

Area of Science:

• Computational intelligence
• Optimization algorithms
• Evolutionary computation

Background:

• Dynamic multiobjective optimization problems (DMOPs) are increasingly important.
• Existing dynamic multiobjective evolutionary algorithms (DMOEAs) struggle to balance population diversity and convergence.
• Effective solutions require advanced strategies for handling evolving problem landscapes.

Purpose of the Study:

• To propose a novel DMOEA that enhances performance on DMOPs.
• To address the critical challenge of maintaining diversity and convergence in dynamic environments.
• To introduce a decision variable classification approach for improved optimization.

Main Methods:

• The proposed DMOEA-DVC classifies decision variables into distinct groups for static and dynamic phases.
• During static optimization, two crossover operators are employed for different variable groups to enhance convergence and diversity.
• During change response, three variable groups are managed using maintenance, prediction, and diversity introduction strategies.

Main Results:

• DMOEA-DVC was evaluated against six state-of-the-art DMOEAs on 33 benchmark DMOPs.
• Experimental results indicate superior or comparable performance of DMOEA-DVC.
• The decision variable classification strategy effectively balances convergence and diversity.

Conclusions:

• DMOEA-DVC demonstrates significant potential for solving DMOPs.
• The proposed method offers a robust approach to managing dynamic changes in optimization problems.
• This work contributes to the advancement of evolutionary algorithms for complex dynamic optimization tasks.