Skip to Main Content
The objective of this work is to demonstrate how clustering techniques such as Kohonen self-organizing maps (SOM) can dramatically reduce the amount of time necessary for optimizing the reconfigurable aperture. Herein, a microstrip line-fed rectangular patch antenna is targeted as the reconfigurable aperture of interest. The patch element is divided into smaller elements, which by themselves are too small to radiate efficiently, yet in combination become a viable radiating structure. A 3-D finite-difference time-domain (FDTD) full-wave analysis utilizing perfectly matched layers is performed to determine the broadband return loss of a reconfigured aperture representing the optimization goal of our SOM. Simulations were performed for a large subset of possible state combinations of the reconfigurable aperture in order to cluster in an unsupervised way the states of the reconfigured apertures using SOM subject to broadband return loss. The SOM is shown to be an efficient method for reconfiguring the aperture from one state to the next within its respective cluster for large populations of feasible reconfigurable states.