Design and performance characterization of a Sub-6 GHz 4-Port MIMO antenna for V2X systems using data-driven learning models

This work presents the 4-port Multiple Input Multiple Output (MIMO) antenna, specifically optimized for Vehicle-to-everything (V2X) communication operating in the Sub-6 GHz frequency band. The proposed MIMO antenna exhibits substantial impedance matching bandwidth between the range of 5.12 GHz to 6.72 GHz with a minimum return loss at 5.9 GHz. To address the complexities and computational burdens of traditional design methodologies, a machine learning (ML)-driven inverse-design framework is implemented. Specifically, Gaussian Process, XGBoost, Random Forest, and Multilayer Perceptron surrogate models were trained on extensive electromagnetic simulation data to predict the antenna’s resonant frequency accurately. These models were combined using a Voting Regressor ensemble and optimized through Differential Evolution to identify optimal antenna geometries efficiently. The 4-port configuration provides efficient decoupling and coherence among the antenna elements by achieving isolation that exceeds 18 dBi. The suggested MIMO antenna is suitable for high-performance V2X applications due to its symmetric radiation parameters and isolation above 10 dBi between the co-polarization and cross-polarization patterns. The maximum gain of 4.9 dBi, and optimum value of diversity parameters (ECC<0.002, DG>9.99 dB, and CCL< 0.3 Bits/Sec/Hz) are achieved, which ensures enhanced communication capabilities, making it suitable for MIMO antenna architectures.