UCPSO: A Uniform Initialized Particle Swarm Optimization Algorithm with Cosine Inertia Weight.

Area of Science:

• Computational Intelligence
• Swarm Intelligence Algorithms

Background:

• Particle Swarm Optimization (PSO) is a popular meta-heuristic algorithm known for its efficiency.
• However, standard PSO can suffer from premature convergence and getting trapped in local optima.
• Enhancements to PSO parameters, initialization, and topology aim to improve its global search capabilities.

Purpose of the Study:

• To propose a novel variant of the Particle Swarm Optimization algorithm, termed UCPSO.
• To enhance the global search ability and overall performance of PSO.
• To address the limitations of premature convergence and local optima in standard PSO.

Main Methods:

• Introduced UCPSO, incorporating three key improvements: a cosine inertia weight, uniform particle initialization, and a rank-based strategy.
• The cosine inertia weight employs a variable-period cosine function for a multi-stage exploration-exploitation balance.
• Uniform initialization prevents particle aggregation, while the rank-based strategy dynamically adjusts inertia weights to improve swarm exploration and exploitation.

Main Results:

• Comparative experiments demonstrated the effectiveness of the proposed UCPSO variant.
• The integrated improvements significantly enhanced the global search ability of the algorithm.
• UCPSO showed superior performance compared to standard PSO in addressing optimization challenges.

Conclusions:

• The UCPSO algorithm effectively overcomes the limitations of traditional PSO, particularly premature convergence.
• The combination of cosine inertia weight, uniform initialization, and rank-based strategy leads to improved optimization performance.
• UCPSO offers a competitive alternative for complex optimization problems requiring robust global search capabilities.