Model-based Reinforcement Learning-Optimized Flexible MIMO Antenna for 5G Higher-Band (n257–n260) and Wearable Biotech Applications
摘要
This research presents a novel four-port flexible multiple-input/multiple-output (MIMO) antenna tailored for 5G high-band and wearable biotech applications. The design emphasizes enhanced port isolation, wide bandwidth, and improved gain. A single antenna element is first optimized using a model-based reinforcement learning (MBRL) algorithm, which efficiently explores the design space with reduced computational effort. The optimized element, measuring 6.47 × 5.95 mm2, is scaled into a compact 15 × 11 mm2 flexible millimetre-wave MIMO antenna. Fabricated on a Rogers RT/duroid 5880 substrate (thickness: 0.508 mm) with coaxial probe feeding, the antenna is tested in both on-body and off-body scenarios. The performance remains stable when worn on various body parts, including the hand, leg, and chest. The antenna operates across 23.5–35.5 GHz and 40.5–70 GHz, effectively covering key 5G bands (26, 28, 39, 41, 47, and 60 GHz), with port isolation better than −15 dB. The MIMO performance is validated by evaluation of the envelope correlation coefficient (ECC), diversity gain, total active reflection coefficient (TARC), and channel capacity loss (CCL). For on-body applications, specific absorption rate (SAR) study verifies human safety, with 10-g and 1-g tissue volume-simulated SAR values of 0.0906 and 0.0521 W/kg, respectively, at the resonant frequency. Furthermore, the antenna maintains consistent performance under significant bending, making it ideal for wearable 5G communication systems.