Improved Particle Swarm Optimization Based on Velocity Clamping and Particle Penalization

Abstract

The idea of particle swarm optimization falls under the domain of swarm intelligence. Particle swarm optimization technique is widely used for finding the global minima of well-known benchmark functions. The main idea behind this technique is that working in a group improves the performance of a system. A modified particle swarm optimization technique is proposed in this paper and tested on seven standard benchmark functions. The two major modifications are introduced in the standard particle swarm optimization, modify the velocity of a particle such that the particle remains within the confine limits of clamp velocity, and penalize the particle velocity, if the sum of the velocity vector and position vector results in breaching the boundary limits of search space. The results of the modified PSO are compared with the two versions of standard PSO, constant inertial weight with no velocity clamping and linearly decreasing inertial weight with no velocity clamping.