Multiobjective Particle Swarm Optimization Based on Cosine Distance Mechanism and Game Strategy.

Area of Science:

• Computational Intelligence
• Optimization Algorithms

Background:

• Optimization problems are ubiquitous, with multiobjective optimization presenting unique challenges due to multiple competing objectives.
• Existing multiobjective optimization algorithms often struggle with maintaining both convergence and diversity.

Purpose of the Study:

• To propose a novel competitive multiobjective particle swarm optimizer (MOPSO) that addresses the limitations of existing methods.
• To enhance the performance of MOPSO by integrating a cosine distance measurement mechanism and a novel game strategy.

Main Methods:

• Developed a MOPSO incorporating a cosine distance mechanism for updating the external archive and a candidate set for effective replacement.
• Introduced a global leader selection strategy integrating game theory and the cosine distance mechanism to increase selection pressure.
• Utilized mutation to preserve swarm diversity and prevent premature convergence.

Main Results:

• The proposed algorithm demonstrated superior performance in benchmark comparisons against state-of-the-art MOPSO and multiobjective evolutionary algorithms.
• Experimental results indicated significant improvements in optimization quality, including convergence and diversity.

Conclusions:

• The novel competitive MOPSO effectively balances convergence and diversity in multiobjective optimization.
• The integrated cosine distance and game strategy offers a promising approach for tackling complex optimization tasks.