Optimal design of transmitarray antennas via low-cost surrogate modelling

Area of Science:

• Electromagnetics
• Antenna Theory
• Artificial Intelligence

Background:

• Transmitarrays offer high gain and efficient radiation without complex feeds, surpassing reflectarrays by avoiding feed blockage.
• Designing transmitarrays is complex due to the need to control transmission phase and magnitude, requiring full-wave electromagnetic models.
• Traditional surrogate modeling for transmitarrays is costly, demanding thousands of training data points.

Purpose of the Study:

• To propose a novel, cost-efficient methodology for designing transmitarrays.
• To develop artificial intelligence-enabled data-driven surrogates for faster and more reliable transmitarray unit element modeling.
• To reduce the number of training data samples required for surrogate model construction.

Main Methods:

• Development of artificial intelligence-enabled data-driven surrogate models.
• Utilizing a significantly reduced dataset (hundreds of samples) for training surrogate models.
• Validation of the surrogate model's predictive power for reliable transmitarray design.

Main Results:

• Successfully constructed AI-enabled surrogates using only a few hundred training samples.
• Achieved sufficient predictive power for reliable transmitarray design with the proposed method.
• Demonstrated the methodology by designing high-performance transmitarrays across multiple frequency bands (8-14 GHz, 22-28 GHz, 28-36 GHz).

Conclusions:

• The proposed AI-driven surrogate modeling approach enables cost-efficient design of transmitarrays.
• This method significantly reduces the data acquisition cost compared to traditional surrogate modeling.
• The validated methodology facilitates the reliable design of high-performance transmitarrays for various applications.