ENTERPRISE AI ANALYSIS

Graphene based terahertz MIMO antenna with machine learning regression for 6G communications

This study introduces a compact, high-performance multiple-input multiple-output (MIMO) antenna engineered for 6G terahertz (THz) communication systems. The antenna is implemented on a polyimide substrate (dielectric constant er = 3.5, loss tangent tand = 0.0027) with dimensions of 405 × 163.75 μm², providing a miniaturized footprint suitable for integrated wireless devices. The antenna exhibits multi-resonance operation at 3.752, 4.204, 4.652, 5.104, 5.548, 6.000, and 6.460 THz, providing corresponding bandwidths of 0.3348, 0.2634, 0.2389, 0.2289, 0.2158, 0.2063, and 0.2096 THz, ensuring wideband coverage suitable for high-data-rate applications. The antenna achieves a peak gain of 13.353 dB, outstanding isolation of -34.044 dB, and high efficiency of 96.048%, highlighting its suitability for high-data-rate and low-interference 6G communications. Strong diversity performance is demonstrated through an ultra-low envelope correlation coefficient (ECC) of 0.00017856 and a near-ideal diversity gain (DG) of 9.99911, confirming the effectiveness of the proposed design for interference mitigation and channel reliability. CST Microwave Studio (MWS) simulations were employed to generate datasets for supervised regression machine learning to predict antenna gain. Random Forest Regression delivered superior predictive accuracy with MSE = 0.76%, MAE = 5.43%, RMSE = 8.72%, R2 = 93.93%, and variance score = 95.12%, closely matching the simulated results. The integration of high-performance multi-band antenna design with regression-based machine learning demonstrates a reliable framework for rapid performance evaluation. The combination of compact geometry, wide multi-band operation, high gain, strong isolation, and machine learning-based predictive modeling positions the proposed antenna as a promising solution for high-data-rate and interference-resilient 6G THz communication networks.