Hybrid Particle Swarms Applied to the Synthesis of Planar Array Feeds

Abstract:
A modified particle swarm optimization (PSO) based algorithm is proposed to optimize planar array feeds when considering complex synthesis. In this work, classical PSO schemes have been modified by introducing one of the most effective selection strategies commonly used in genetic algorithms (GA), tournament selection (TS), and the new hybrid schemes have been applied to several planar array synthesis problems in order to analyze the improvements achieved. First of all, a preliminary study has been carried out to test the approach proposed, considering a canonical synthesis problem and up to four different classical PSO schemes. The results achieved demonstrate the usefulness of the hybrid PSO based schemes, which, in some cases, clearly outperform the classical ones. The reason why the hybrid approach cannot be extended to the whole set of PSO schemes deals with the nature of the TS, which increases search pressure over the swarm but, at the same time, can make diversity vanish. Finally, representative synthesis results for a planar array considered as a sectored antenna for local multipoint distribution systems (LMDS) are also included.