An Improved Particle Swarm Optimization Algorithm
Meerah Karunanithi, Hajar Mouchrik, Arslan A. Rizvi, Talha Ali Khan, Rand Kouatly, Iftikhar Ahmed
Abstract
Particle Swarm Optimization (PSO) is a sophisticated optimization technique deeply rooted in artificial intelligence. This meta-heuristic algorithm addresses complex, nonlinear, and multidimensional problems, often delivering commendable solutions with minimal parameter tuning. PSO iteratively enhances candidate solutions at its core, progressively converging toward near-optimal outcomes, drawing inspiration from natural phenomena like birds and fish flocking behaviour. The PSO process commences by initializing a pool of random solutions and diligently pursuing the best solution through successive iterations. Since its inception in 1995, PSO has undergone numerous refinements to optimize its performance and outcomes through parameter fine-tuning. This paper advances the conventional PSO framework by introducing an improved PSO variant characterized by a linearly diminishing inertia weight and a novel velocity equation. Subsequently, the proposed PSO variant undergoes rigorous testing against a benchmark function set to assess its performance. It is compared to alternative PSO variants to gauge its competitiveness. Furthermore, its efficacy in practical applications is validated by scrutinizing an engineering design optimization problem and engaging in comparative experiments with other algorithmic iterations.