A team of researchers at Digital University Kerala has created an AI-powered system capable of identifying different antenna types and detecting performance issues without needing physical disassembly or conventional testing. This advancement could revolutionize sectors such as communications, defense, and aerospace. It may also pave the way for self-monitoring communication modules in satellites, drones, and Internet of Things (IoT) devices, enabling real-time fault detection and performance assessment. The findings are featured in the IEEE Journal of Microwaves and led by Anitha Gopi, Sruthi Pallathuvalappil, Elizabeth George, and Alex James.
Their study presents a neuro-memristive 3D crossbar framework that interprets antenna radiation patterns similarly to how images are read. The system then classifies antennas—whether dipole, monopole, or patch—using a sophisticated 3D memristive convolutional neural network (3D-CNN).
Traditional antenna evaluation typically requires expensive and lengthy experiments in anechoic chambers, which are specially designed rooms that block sound or electromagnetic reflections.
The new approach leverages pixel-based sampling and AI algorithms to analyze electromagnetic field data, reducing the need for high power, large area setups, and long testing durations. The hardware was implemented via the Skywater 130-nanometer open-source semiconductor process.
Comparisons between the 3D-CNN and other machine learning models, including YOLOv8 and VGG-19, showed that the neuro-memristive design offered superior accuracy and faster processing, even in noisy conditions such as Gaussian or white noise.
“Our system offers a compact, non-invasive way to ensure antennas are working correctly, even in noisy or harsh environments,” Prof. James, the corresponding author, said. This is especially valuable for defence and remote communication applications, he added.
According to Ms. Gopi, this research merges AI with hardware design, offering a smarter, quicker, and more dependable solution for antenna testing.
Click here to learn more about the latest innovations in AI for the RF Industry.
