Pixelated Antenna Design

DC-SADEA addresses a binary geometry problem: preserving useful spatial structure while searching thousands of antenna pixels. XGBoost predicts candidate performance; a bespoke 2-D crossover exchanges rectangular regions rather than flattened vector segments.

Method

01Binary geometryInitial EM evaluations
02Spatial searchPopulation diversity + 2-D crossover
03XGBoost predictionRank new candidate layouts
04EM validationEvaluate the best predicted candidate

↻ Each validated candidate updates the database for the next iteration. Method summary redrawn from the paper.

Prototype and validation

Fabricated UWB antenna, front and back — Fabricated prototype of the 1920-pixel UWB design. Paper Fig. 6 ↗

1920pixels in the UWB example

2.9–13.6 GHzdesign operating band

20 × 30 mmantenna footprint

Simulated and measured UWB reflection coefficient, gain and radiation efficiency — Simulation–measurement comparison. Radiation efficiency was measured only up to 6 GHz because of equipment limitations. Paper Fig. 7 ↗

Outdoor base-station antenna

The second DC-SADEA case uses a hybrid structure: conventional double-oval dipoles and a reflector are retained, while two feeding structures are digitally coded to improve matching and port isolation. The 1496-pixel design covers 3.3–3.8 GHz and 4.8–5.0 GHz with dual-linear polarization.

Fabricated outdoor base-station antenna, front and top views — Fabricated base-station prototype: front and top views. Paper Fig. 10 ↗

DC-SADEA evaluation cases

Case	Scope	Reported evidence
UWB antenna	1920 pixels	Simulated minimum gain 2.64 dBi; minimum efficiency 80.7%
5G base-station feeds	1496 pixels; 14 specifications	Matching and isolation over 3.3–3.8 and 4.8–5.0 GHz
Electrically small antenna	270 pixels	95 MHz simulated and 90 MHz measured −3 dB bandwidth

For the small-antenna case, DC-SADEA reached feasibility in an average of 1623 EM evaluations over five runs. Standard GA used 8910 evaluations in its single run. These budgets describe this benchmark, not a universal speedup.

A separate EuCAP 2026 paper by Qiang Hua, Xinxin Liu, Mobayode O. Akinsolu, Xinrui Wang and Pavlos Lazaridis demonstrates another use of digital coding: keep a square-ring slot antenna fixed and optimize a pixelated parasitic region on either side of its microstrip feed. The paper describes a symmetric 24 × 24 binary pixel field, separated from the feed edge by 0.5 mm.

Binary parasitic pixel field on both sides of a microstrip feed — Pixel-based parasitic region around the feed; the baseline geometry remains fixed. Paper Fig. 3 ↗

Optimized on/off metal pixel pattern beside the microstrip feed — Optimized parasitic metal pattern reported using SADEA-VI. Paper Fig. 4 ↗

The authors report simulated and measured reflection coefficients meeting the −10 dB target across 3.1–10.6 GHz. Reported using SADEA-VI, this related application illustrates how pixel coding can refine a local parasitic region, as well as an entire radiator or a feeding structure.

Qiang Hua et al. A Novel Digitally Coded Ultra-Wideband Antenna Optimized by SADEA-VI. EuCAP 2026. Parasitic-structure figures are cropped from the accepted manuscript (Figs. 3 and 4), attributed under CC BY 4.0.

Explore the submatrix crossover

Interactive illustration

Preserve a spatial motif

A teaching example of rectangular crossover and mutation. It does not simulate antenna performance.

Schematic teaching example. It illustrates the mechanism, not measured or simulated performance from the paper.