This report presents initial results apropos a genetic algorithm that evolves 3D geometries; this early study is a proof of concept, presenting an algorithm that evolves to a predefined target shape, with the final goal of designing optimized 3D antenna geometries. The algorithm presented in this work builds structures by combining building blocks of different geometric primitives, where the fitness of a design is measured by the similarity to the target shape. Multiple techniques for comparing 3D shapes are explored and compared in the context of a fitness score for evolutionary algorithms. The designed algorithm was capable of evolving to biconical and dipole antenna shapes rapidly using a variety of fitness functions. A more complex log-periodic antenna shape was evolved using a directed fitness function comparing the component shapes. The development of this algorithm preludes incorporating more complex fitness functions that build designs to improve sensitivity to science outcomes. Future improvements to the algorithm and the steps required to achieve designs with improved sensitivity are discussed in the context of antennas but could be applied to other applications.
Julie Rolla, Bryan Reynolds, Jacob Weiler, Amy Connolly, Ryan Debolt, Alex Machtay, Ben Sipe, and Dylan Wells. Design of 3D Antenna Geometries Using Genetic Algorithms. The Interplanetary Network Progress Report, Volume 42-234, pp. 1-26, August 15, 2023. https://ipnpr.jpl.nasa.gov/progress_report/42-234/42-234A.pdf
