Improved Particle Swarm Optimization for Global Optimization with Decaying Adaptive Velocity Limit

Abstract

This paper introduces an innovative enhanced Particle Swarm Optimization (PSO) technique known as IPSO-AVL. It features a time-adaptive velocity limit and a fitness-weighted personal best centroid. To test how effective this new method is, we conducted a comparative analysis against the traditional PSO and an intermediate IPSO approach using five different test functions: the Sphere function, the Rosenbrock function, the Rastrigin function, the Griewank function, and the Ackley function. The experiments were set up with parameters of D=30, N=30, and T=5000 over 30 runs. The findings indicate that the IPSO-AVL method significantly outperforms the traditional models, particularly for the Sphere and Rastrigin functions.