GA optimization of terminal antennas by the estimation of the population density of probability using dependency trees

This paper presents an application of Bayesian network density of probability estimators to antenna optimisations, that involves relatively big number of parameters, multiple solutions, or local maxima, that increase the likelihood to converge towards not satisfactory suboptimal solutions. Three methods are compared for the case of a triband antenna optimization: dependency tree method (TREE), GA with linkage crossover operator (GLINX) and a simple GA with dual population diversity method (DUAL). Basic theory, the definition of the chosen structure to be optimized, simulation results and convergence curves are presented.