A hierarchical particle swarm optimizer and its adaptive variant

Area of Science:

• Computational Intelligence
• Optimization Algorithms
• Metaheuristics

Background:

• Particle Swarm Optimization (PSO) is a widely used metaheuristic for complex problem-solving.
• Standard PSO relies on fixed neighborhood structures, which can limit swarm adaptability.
• Dynamic adaptation of particle influence is crucial for improving optimization efficiency.

Purpose of the Study:

• Introduce a hierarchical particle swarm optimization (H-PSO) metaheuristic.
• Investigate the impact of dynamic hierarchy on swarm behavior and performance.
• Explore variants of H-PSO with adaptive hierarchy shapes and differentiated particle behaviors.

Main Methods:

• Developed a hierarchical structure where particles dynamically ascend or descend based on solution quality.
• Implemented H-PSO with a dynamically adapting hierarchy shape.
• Introduced H-PSO variants with level-specific particle behaviors.
• Tested H-PSO and its variants against standard PSO on benchmark optimization functions.

Main Results:

• H-PSO demonstrated enhanced performance compared to standard PSO across various optimization functions.
• The dynamic hierarchy allowed high-quality solutions to exert greater influence on the swarm.
• Adaptive hierarchy variants showed improved convergence and robustness.
• Differentiated particle behaviors tailored to hierarchy levels further boosted optimization effectiveness.

Conclusions:

• The proposed H-PSO metaheuristic offers a significant advancement over traditional PSO.
• Dynamic hierarchical structures and adaptive behaviors are effective strategies for improving swarm intelligence.
• H-PSO and its variants provide a more flexible and powerful tool for complex optimization tasks.