In order to facilitate real-time performance monitoring and issue detection, innovation may result in self-diagnosing communication modules in satellites, drones, and Internet of Things devices.
Without physically inspecting or disassembling the antennas, researchers at Digital University Kerala have created an artificial intelligence (AI)-driven system that can recognize different kinds of antennas and identify performance issues.
The innovation has the potential to revolutionize the aerospace, defense, and communication sectors. In order to facilitate real-time performance monitoring and defect detection, the invention may potentially result in self-diagnosing communication modules in satellites, drones, and Internet of Things (IoT) devices.
Anitha Gopi, Sruthi Pallathuvalappil, Elizabeth George, and Alex James conducted the study, which was published in the IEEE Journal of Microwaves. The work presents a neuro-memristive 3D crossbar system that can interpret patterns of antenna radiation, such as pictures. Additionally, it uses a sophisticated 3D memristive convolutional neural network (3D-CNN) to classify the antenna type (dipole, monopole, or path).
Antenna testing has historically included expensive and time-consuming tests conducted within anechoic chambers, which are spaces intended to prevent sound or electromagnetic wave reflections or echoes.
The researchers claim that the new technique reduces power, area, and testing time by analyzing electromagnetic field data using pixel sampling and AI algorithms. The Skywater 130-nanometer process, an open-source semiconductor platform, was used for hardware implementation.
Prof. James, the corresponding author, said, “Our method provides a small, non-invasive technique to guarantee antennas are operating properly, even in hostile or noisy conditions.” He went on to say that this is particularly useful for situations involving distant communication and defense.
In order to determine if the neuro-memristive technique produced superior accuracy and quicker processing even under signal noise, such as Gaussian and white noise, the research also compared the 3D-CNN with other machine learning models, such as YOLOv8 and VGG-19.
According to Ms. Gopi, the study combines hardware innovation with artificial intelligence to improve the speed, accuracy, and intelligence of antenna testing.