Predictive modelling and high-performance enhancement smart thz antennas for 6 g applications using regression
Md Ashraful Haque1, Md Mostafa Arafat2, Isha Das3
1Department of Electrical and Electronic Engineering, Daffodil International University, Dhaka, 1341, Bangladesh. limon.ashraf@gmail.com.
Scientific Reports
|October 5, 2025
Summary
A novel graphene MIMO antenna for 6G terahertz applications offers broad bandwidths and high gain. Machine learning models, particularly Extra Trees Regression, significantly enhance design prediction accuracy and optimization.
Area of Science:
- Antenna Engineering
- Terahertz Technology
- Materials Science
Background:
- Sixth generation (6G) wireless systems require advanced antenna solutions for terahertz (THz) frequencies.
- Graphene's unique electromagnetic properties make it a promising material for high-performance antennas.
- Existing antenna designs often face challenges in achieving wide bandwidth, high gain, and efficient isolation at THz frequencies.
Purpose of the Study:
- To introduce a novel graphene-based multiple-input multiple-output (MIMO) antenna design for 6G THz applications.
- To evaluate the antenna's performance in terms of bandwidth, gain, isolation, and efficiency.
- To integrate machine learning (ML) for enhanced predictive modeling and optimization of the antenna design process.
Main Methods:
- Utilized CST Studio Suite for electromagnetic simulation and design.
- Validated the antenna design using an RLC equivalent circuit model in ADS.
- Employed five supervised regression ML models (Extra Trees, Random Forest, Decision Tree, Ridge Regression, Gaussian Process Regression) for predictive analysis.
- Trained ML models on a dataset derived from parametric variations in antenna geometry.
Main Results:
- The proposed graphene MIMO antenna exhibits multi-resonant behavior across three broad bandwidths: 2.479 THz, 0.516 THz, and 1.694 THz.
- Achieved a peak gain of 13.41 dB, isolation of -34.2 dB, and up to 90% efficiency.
- The Extra Trees Regression model demonstrated superior prediction accuracy with an R² score of 98.91%, MAE of 2.51%, and MSE of 0.44%.
- Demonstrated excellent diversity gain (9.9993) and a low envelope correlation coefficient (ECC) of 0.00013.
Conclusions:
- The novel graphene MIMO antenna design meets the demanding requirements for 6G THz applications.
- The integration of ML and RLC circuit modeling significantly reduces simulation time and improves design optimization.
- The developed framework offers a robust approach for future advancements in THz wireless communication and biomedical applications.
