A new particle swarm algorithm and its globally convergent modifications.

Area of Science:

• Computational intelligence
• Optimization algorithms
• Swarm intelligence

Background:

• Particle Swarm Optimization (PSO) is a population-based technique widely used for optimization.
• Standard PSO can exhibit limitations in global search ability during later iteration stages.
• Understanding particle trajectory is key to improving PSO performance.

Purpose of the Study:

• To investigate particle behavior in standard PSO using Monte Carlo methods.
• To develop a novel PSO variant (MRPSO) with enhanced exploration capabilities.
• To introduce a hybrid MRPSO (HMRPSO) with a mutation strategy for improved global optimum finding and balanced search.

Main Methods:

• Monte Carlo simulations to analyze PSO particle trajectories.
• Development of a Moderate-Random-Search strategy for PSO (MRPSO).
• Implementation of a hybrid MRPSO (HMRPSO) incorporating a new mutation strategy.
• Performance evaluation using thirteen benchmark functions.

Main Results:

• Identified reasons for PSO's diminished global search ability in later stages.
• MRPSO demonstrated improved exploration and convergence rates.
• HMRPSO significantly outperformed standard PSO and other PSO variants on multimodal and unimodal functions.
• HMRPSO showed promising search performance compared to recent evolutionary algorithms.

Conclusions:

• The proposed HMRPSO algorithm effectively enhances global search and convergence.
• HMRPSO achieves a better balance between exploration and exploitation in solution spaces.
• HMRPSO represents a significant advancement over existing PSO and evolutionary algorithms for complex optimization tasks.