Citation
Abstract
This report presents developments to a genetic algorithm that evolves antenna designs from primitive shapes. The gain patterns of individuals are evaluated using XFdtd electromagnetic simulation software, and fitness is evaluated by comparison to a target gain pattern. This work is an update to a prior report demonstrating evolution of 3D structures to target geometries. The algorithm was updated to evolve antennas by adding feeds, early shorting prevention, and simulation components. Different fitness functions were developed, and differences in convergence to similar gain patterns was explored. Computational efficiency was also improved by converting to an asynchronous steady-state algorithm and incorporating, among other improvements, an elite method, and a diversity forcing feature. The algorithm was able to evolve antennas to match the desired gain pattern using each fitness function. The algorithm was run for a single-frequency (300MHz) design and for a broadband case (200MHz to 800MHz). Later versions of the algorithm will utilize fitness functions connected to science simulation software to generate designs optimized to science outcomes. Future improvements to the algorithm to facilitate more complex designs and to reduce computation time are also discussed.
Details
- Volume
- 42-237
- Published
- May 15, 2024
- Pages
- 1–47
- File Size
- 13.7 MB