Human-vs Machine Design of Antennas: Evolution Behavior in Genetic Shape Optimization

Random-based global optimization algorithms have found extensive application in the domain of antenna shape design, especially when conventional solutions relying on human expertise are lacking. In this research contribution, we investigate the performance of random-based global optimization in scenarios where the design problem could otherwise be tackled through conventional human-guided design methods and parameter adjustments driven by simulations. The present case study involves shape optimization of a 2D pixelated domain, performed via binary coding and a Genetic Algorithm (GA). The reference geometry is a square resonant patch-type antenna with optimized probe feeding position. The initial domain is a pin-centered rectangle larger than the patch itself, so that the optimizer is eventually free to indirectly find the best pin position corresponding to the best design of the patch.