Machine Learning-Optimized Compact Wearable Frequency Reconfigurable Antenna for Sub-6 GHz/mm-Wave 5G Integration

Future 5G wireless systems will have substantial challenges in integrating the sub-6 GHz and millimeter-wave (mm-wave) bands due to their massive frequency ratios. This paper proposes a machine learning-optimized compact wearable frequency-reconfigurable antenna for sub-6 GHz/mm-wave 5G integration. Fabricated on a flexible Rogers Duroid substrate (27.8 × 14 × 0.508 mm³), the antenna initially employs a circular structure resonating at 28 GHz. Dual-band operation (3.5 GHz and 28 GHz) is achieved by etching an H-shaped slot into the rectangular patch. A PIN diode is employed to reconfigure the proposed antenna in the ON and OFF states. In the ON state, the antenna operates at 3.5 GHz and 28 GHz, achieving measured bandwidths of 25.4% and 73.2%, gains of 3.63 dBi and 5.25 dBi, and radiation efficiencies of 90.5% and 88%, respectively. In the OFF state, the antenna operates at 28 GHz, achieving a measured bandwidth of 72.9%, gain of 6.2 dBi, and a radiation efficiency of 89%. Bidirectional E-plane and omnidirectional H-plane radiation patterns are maintained across both bands. At 3.5 GHz, the specific absorption rate (SAR) value for 1 g and 10 g of human tissue is 0.438 W/kg and 0.0147 W/kg, while at 28 GHz, the SAR value is 0.801 W/kg and 1.09 W/kg, which comply with the FCC and ICNIRP standards. Bending tests (lap, chest, arm) demonstrate stable on-body performance. The antenna’s S11 was predicted using a supervised ML regression framework. Among tested algorithms, the decision tree achieved state-of-the-art accuracy (R²: 97.80%) with minimal errors (MAE: 0.72, MSE: 0.28, MSLE: 0.56, RMSLE: 0.81, RMSE: 0.66). The proposed antenna system is suitable for future 5G devices.