Machine Learning-Based Optimization of Beam Steering in Reconfigurable Mushroom Antenna

This paper presents a machine learning (ML)-based approach to optimize beam steering in a reconfigurable mushroom antenna composed of metamaterial unit cells with PIN diodes. The antenna consists of metamaterial mushroom unit cells, each incorporating a central shorting via and a PIN diode. Beam steering is achieved by dynamically adjusting the states of the PIN diodes. By employing an artificial neural network (ANN) to predict optimal diode configurations, the proposed method achieves beam steering of $\pm 17^{\circ}$ in the E-plane and $\pm 12^{\circ}$ in the Hplane at 5 GHz, validated through full-wave simulations.