Comparison of heuristic methods when applied to the design of reflectarrays

This paper presents and compares different configurations of several heuristic optimization methods with the aim of seeking the one with the best performance when applied to phase synthesis of reflectarray antennas. This phase synthesis is a hard task and improving the synthesis method can result in significant reductions in the associated computational cost. The performance of algorithms such as real-valued particle swarm optimization (PSO), simulated annealing (SA), micro-genetic algorithms (μGA) and free search-differential evolution (FSDE) are compared. The phase synthesis of reflectarrays with several sizes under different radiation pattern restrictions using a large number of initialization seeds are considered to obtain reliable conclusions. Furthermore, the effect of chaotic, opposite and orthogonal initialization techniques on PSO method is also investigated. Finally, representative results including radiation patterns obtained in the synthesis process as well as simulations of the effect introduced by the manufacturing tolerances are presented.