Real-time antenna array synthesis using machine learning

With potential applications in 6G wireless communications, on smart and flexible satellite payloads, and in advanced radar systems, flat-panel array antennas hold great promise for contemporary and future technology. TICRA is at the forefront of developing advanced tools for analysis and design of large-scale antenna arrays and machine learning approaches to antenna design.

The innovative machine learning-based framework allows us to address the array synthesis problem in a completely new way, offering unprecedented speed while maintaining high accuracy

In this context, TICRA recently completed the activity Fast Design Algorithms for Antenna Arrays Using Machine Learning that was supported by the European Space Agency (ESA). Here, the array synthesis problem of determining the element excitation amplitudes and phases to meet specific radiation pattern requirements was addressed.

Niels Skovgaard Jensen and Lasse Hjuler Christiansen, two of TICRA’s machine learning engineers and leading contributors to the recently completed ESA project.

As the key outcome of the activity, TICRA has developed a machine learning framework that facilitates the design process for phased arrays. The framework is based on specialised neural networks that, in milliseconds, can accurately compute excitation coefficients in response to changing conditions and radiation pattern requirements, thus paving the way for real-time beamforming, sidelobe suppression, interference nulling, and more.

Real-time synthesis of a 36 x 36 element antenna array tracking two satellites in low-earth orbit.

The ability to perform real-time synthesis with complex constraints opens new and exciting avenues in fields like Earth observation, climate science, navigation, and telecommunications

Erio Gandini, ESA’s Technical Officer on the activity, says “The innovative machine learning-based framework allows us to address the array synthesis problem in a completely new way, offering unprecedented speed while maintaining high accuracy.” On potential applications, he continues: “The ability to perform real-time synthesis with complex constraints opens new and exciting avenues in fields like Earth observation, climate science, navigation, and telecommunications.”

Follow us on LinkedIn or sign up for our newsletter below to get more insights and updates from TICRA.