Comparison of different PSO initialization techniques for high dimensional search space problems: A test with FSS and antenna arrays

Population-based optimization algorithms are widely used in multiple research areas to optimize different kinds of problems. Commonly, they have been successfully applied to low and medium size search spaces in order to tune or adjust several design parameters. Unfortunately, population-based algorithms can require too much CPU time to find a nearly optimal solution as the number of dimensions of the problem to tackle increases significantly. Usually, mathematical or physical-inspired techniques have been applied to improve the success rate as well as the speed of convergence of population-based methods in a high dimensional framework. An alternative choice is to allocate the individuals of the population in an efficient way during the initialization stage of the algorithm considering two possibilities: placing individuals as close as possible to a global optimum or uniformly distributed over the search space. In this work, three different initialization strategies, the orthogonal array initialization, a chaotic technique and the opposition based initialization have been considered and appropriately combined with the heuristic particle swarm optimization (PSO) algorithm. Results comparing the modified PSO algorithm when applied to optimize frequency selective surfaces (FSS) and planar arrays for WiMAX applications are included.