Impact of Random Number Generators on the performance of particle swarm optimization in antenna design

Evolutionary search algorithms are highly efficient tools to solve problems when it is not possible to use an exhaustive search approach. Particle Swarm Optimization (PSO) is a bio-inspired technique to perform an intelligent search in continuous and multidimensional solution spaces. Random Number Generators play an important role in PSO. They are commonly used within PSO to solve real complex engineering problems. However, only a few studies have been reported on the impact of Random Number Generators (RNG) on the performance of PSO. In this paper, the impact of different RNGs is studied by considering the swarm initialization (both the particles' positions and velocities) and their velocities update. The performance of PSO with different RNGs is tested based on three real-world electromagnetic problems. They are a) a resonant rectangular patch antenna, b) an E-shape microstrip antenna and c) a highly compact and strongly coupled 4-element microstrip array. All the tested RNGs can achieve the design requirements. The Gaussian RNG converges much faster than uniform ones.