Optimization methods for computer-aided design of artificial dielectric lens antennas

Abstract

The paper investigates the optimization of an electrically large antenna : a shaped dielectric substrate lens at millimeter wave frequencies (76 – 77 GHz). The aim is to determine the best lens profile that complies with arbitrary desired radiation pattern templates. A feed forward artificial neural network (ANN) was implemented in order to predict the radiation pattern of this antenna. Particle Swarm Optimization (PSO) procedure, combined with neural network modeling, is applied in order to minimize the cost function (the fitness function) and to fit specifications. The optimized results are in a good agreement with 3D MoM simulations.