genuy 0.1.3

A genetic algorithm optimization based Uda-Yagi antenna generator,

genuy

A genetic algorithm-based Uda-Yagi antenna optimizer. Uses PyGAD as the GA engine and pymininec for antenna simulation to evolve element lengths and spacings that maximise gain, front-to-back ratio, and VSWR across a specified bandwidth.

Features

• Optimizes reflector, driven element, and an arbitrary number of director elements
• Evaluates fitness across the full bandwidth (low edge, centre, high edge) to avoid narrow-band solutions
• Adaptive mutation and tournament selection for robust multimodal search
• Parallel fitness evaluation using all available CPU cores
• Exports the best solution as an MMANA-GAL .maa file

Requirements

• Python ≥ 3.14
• pymininec ≥ 1.2.0
• pygad ≥ 3.5.0

Installation

pip install genuy

Or from source:

git clone https://github.com/r2axz/genuy
cd genuy
pip install .

Usage

genuy [OPTIONS]

Key options

Option	Default	Description
-n, --num-elements	4	Total number of elements (reflector + driven + directors)
-f, --frequency	145.0 MHz	Target centre frequency
-b, --bandwidth	10.0 MHz	Bandwidth over which to optimise
-r, --element-radius	3.0 mm	Physical radius of all elements
-z, --reference-impedance	50+0j Ω	Reference impedance for VSWR calculation
--num-generations	200	Maximum number of GA generations
--num-solutions	0 (auto: 10 × num genes)	Population size
--percent-mating	10.0	Percentage of population selected as parents
--save-maa	(none)	Save best solution to an MMANA .maa file
--plot-fitness	(off)	Plot fitness vs. generation after the run

Run genuy --help for the full list of options including element length bounds, spacing bounds, mutation rates, VSWR penalty settings, and wire segmentation.

Examples

Optimize a 4-element 145 MHz antenna and save the result:

genuy -n 4 -f 145 -b 10 --save-maa antenna.maa

Optimize a 6-element antenna, plot convergence, and save:

genuy -n 6 -f 145 -b 10 --num-generations 500 --save-maa 6el.maa --plot-fitness

Fitness function

Each candidate solution is evaluated at three frequencies — lower band edge, centre, and upper band edge. The fitness score is:

$$F = \begin{cases} \dfrac{w_\text{vswr}}{\text{VSWR}\text{worst}} + w\text{gain} \cdot G_\text{worst} + w_\text{fb} \cdot \text{FB}\text{worst} & \text{if } \text{VSWR}\text{worst} < \text{threshold} \ p_\text{vswr} + w_\text{gain} \cdot G_\text{worst} + w_\text{fb} \cdot \text{FB}_\text{worst} & \text{otherwise} \end{cases}$$

Default weights: vswr_weight=100, gain_weight=1.0, fb_weight=2.0. The hard VSWR penalty (high_vswr_penalty=-100) discourages solutions that are badly mismatched anywhere in the band.

Output

After the run the best solution is printed to stdout:

Best solution :  [0.512 0.18 0.474 0.15 0.452 0.19 0.438]
Best solution fitness :  47.23
VSWR :  1.34
Impedance :  (42.1+3.2j)
Gain :  10.8
Front-to-Back Ratio :  18.4

License

MIT — see LICENSE.