Design and Analysis of Wideband MIMO THz Sensing Antenna Enabled by Machine Learning for Cognitive Radio: Toward 6G

Abstract

In this article, a wideband dual port Multiple Input Multiple Output (MIMO) terahertz (THz) sensing antenna for Cognitive Radio (CR) applications is designed on a polyimide substrate and optimized using machine learning (ML) techniques. The proposed antenna operates in the 2.59 to 8.5 THz range, offering an impedance bandwidth of 106.58%. It achieves a peak gain of 7.6 dBi and a maximum radiation efficiency of 85%. The key MIMO performance parameters include a total active reflection coefficient (TARC) of <–5 dB, diversity gain (DG) of 10 dB, mean effective gain (MEG) of <–3 dB, an envelope correlation coefficient (ECC) of <0.0017, and channel capacity loss (CCL) of <0.5 bits/s/Hz all within acceptable limits. The antenna is optimized using various ML algorithms, including Decision Tree (DT), Random Forest (RF), K-Nearest Neighbors (KNN), Extreme Gradient Boosting (XGB), and Artificial Neural Network (ANN). Among these, the RF algorithm achieved the highest prediction accuracy, exceeding 99.98%, and is used for S-parameter prediction. The proposed MIMO THz antenna is well-suited for sensing operations in future 6G CR applications. © 2025 Elsevier B.V., All rights reserved.