Improved particle swarm optimization with a collective local unimodal search for continuous optimization problems

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Swarm Intelligence

Background:

• Particle Swarm Optimization (PSO) is a widely used metaheuristic algorithm.
• Premature convergence is a significant challenge in PSO, limiting its effectiveness.
• Existing PSO variants struggle to balance global exploration and local exploitation.

Purpose of the Study:

• To introduce a new local search technique for improving PSO performance.
• To address the issue of premature convergence in particle swarm optimization.
• To enhance the global-local search capabilities of PSO algorithms.

Main Methods:

• A novel local search strategy is proposed, constructing potential particle positions collectively.
• Randomly selected particles contribute values from their personal best positions.
• Local search is performed around the constructed position, compared against the global best.

Main Results:

• The improved PSO algorithms demonstrated superior solution quality.
• Enhanced convergence velocity, precision, stability, and robustness were observed.
• The new technique outperformed four competing PSO variants in benchmark tests.

Conclusions:

• The proposed local search technique effectively mitigates premature convergence in PSO.
• This method significantly improves the overall performance and reliability of PSO.
• The findings suggest a promising direction for advancing swarm intelligence optimization.