Particle Swarm Optimization With Interswarm Interactive Learning Strategy.

Area of Science:

• Computational Intelligence
• Swarm Intelligence
• Optimization Algorithms

Background:

• Canonical particle swarm optimization (PSO) suffers from loss of diversity, hindering its ability to escape local optima.
• Maintaining population diversity is critical for effective global search in optimization problems.

Purpose of the Study:

• To propose an improved PSO algorithm, termed IILPSO, that addresses the diversity loss issue in canonical PSO.
• To enhance the global search capability and prevent premature convergence of PSO through an interswarm interactive learning strategy.

Main Methods:

• Particles are divided into two swarms, with interswarm interactive learning (IIL) triggered by stagnation.
• The softmax and roulette methods determine swarm roles (learning/learned) based on fitness values.
• Velocity mutation and global best vibration strategies are incorporated to boost global search.

Main Results:

• IILPSO demonstrated superior performance compared to eight popular PSO variants in numerical experiments.
• The proposed algorithm showed improvements in solution accuracy, convergence speed, and reliability.
• Analysis of population diversity variations explained the effectiveness of the IILPSO approach.

Conclusions:

• The interswarm interactive learning strategy effectively maintains population diversity and enhances PSO performance.
• IILPSO offers a promising alternative for complex optimization problems requiring robust global search capabilities.
• The study provides insights into the mechanisms driving the improved performance of IILPSO.