A Particle Swarm Algorithm Based on a Multi-Stage Search Strategy.

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Swarm Intelligence

Background:

• Particle Swarm Optimization (PSO) suffers from premature convergence to local optima and suboptimal search accuracy.
• Existing methods to enhance PSO diversity and accuracy often fail to balance exploration and exploitation effectively.
• Population-based approaches frequently divide into sub-populations for spatial management.

Purpose of the Study:

• To introduce a novel multi-stage search strategy for Particle Swarm Optimization.
• To address the limitations of easily getting trapped in local optima and low search accuracy in standard PSO.
• To achieve a better balance between exploration and exploitation in PSO algorithms.

Main Methods:

• A multi-stage search strategy is proposed, primarily utilizing mutual repulsion among particles, supplemented by attraction.
• The strategy aims to control population traits by managing particle interactions.
• The algorithm enhances population entropy over iteration time while ensuring convergence.

Main Results:

• The proposed multi-stage search strategy demonstrated improved performance on the CEC2017 test functions.
• The algorithm achieved satisfactory results, outperforming standard PSO and other improved PSO variants.
• A more balanced search process was created, enhancing population diversity and search accuracy.

Conclusions:

• The developed multi-stage search strategy effectively mitigates local optima entrapment in PSO.
• The proposed method offers a more balanced approach to exploration and exploitation, enhancing overall optimization performance.
• This research contributes a valuable improvement to Particle Swarm Optimization for complex problem-solving.