Machine Learning Aided Tapered Four-Port MIMO Antenna for V2X Communications with Enhanced Gain and Isolation

In this communication, a 4-port Multiple-Input-Multiple-Output (MIMO) antenna is analyzed and investigated for vehicle-to-everything (V2X) communication centered at 5.9 GHz. The proposed optimized single antenna consists of a tapered radiating antenna with a defective ground structure fed with stepped impedance transmission line feed. The proposed 4-port MIMO antenna has a dimension of $96\times 64\times 0.8$ mm3 printed on the FR4 substrate with a relative permittivity of 4.4 and loss tangent of 0.02. To obtain the proposed single antenna unit, parametric analysis, and evolution stages have been investigated and discussed. The impedance bandwidth of the proposed antenna is 5.66 - 6.00 GHz with a peak gain of 7.85 dB and radiation efficiency of 99%. In addition, machine learning techniques such as XG (Extreme Gradient) Boost, Random Forest, and Deep Neural Networks (DNN) were employed in the optimization process to predict and fine-tune the antenna's design parameters. The stacking ensemble method, combining these models, was used to improve the accuracy of the antenna performance prediction. By leveraging machine learning, the final design was achieved more efficiently, significantly reducing the simulation time and enabling more precise parameter tuning for optimal performance. Further, to validate the MIMO antenna characteristics, different diversity parameters have been calculated such as ECC (Envelope Correlation Coefficient), DG (Diversity Gain), CCL (Channel Capacity Loss), and TARC (Total Active Reflection Coefficient). The fabricated antenna is modeled, and measured findings are found to be in coherence with simulated findings.