Hybrid Optimization Framework for MIMO Antenna Design in Wearable IoT Applications Using Deep Learning and Bayesian Method

The growing adoption of wearable Internet of Things (IoT) devices requires efficient wireless communication systems for applications like healthcare and fitness. Multiple-input multiple-output (MIMO) technology improves signal quality by using multiple antennas but presents challenges in compactness and low power consumption and offers minimal mutual coupling while interacting with the human body. Ensuring compliance with specific absorption rate (SAR) limits is also crucial. In this paper, we present a hybrid optimization framework for designing MIMO antennas for such wearable devices. We integrate deep learning with Bayesian optimization. We use artificial neural networks (ANNs) to model antenna performance and Bayesian optimization to explore the design space efficiently. The final optimized MIMO antenna has an overall size of 122.44 mm \(\times \) $\times$ 58.32 mm. The proposed MIMO antenna achieved minimized mutual coupling in free space, on hand, and leg with values of < \(-\) $-$ 75.77 dB, < \(-\) $-$ 46 dB, and < \(-\) $-$ 37 dB, respectively. In every scenario, the antenna has a stable resonance at a 2.45 GHz frequency and maintains a SAR within the 1.6 W/kg safety limit. Additionally, our optimization reduced computational effort by about 30% compared to traditional methods. These results show the effectiveness of combining ANNs and Bayesian optimization in designing high-performance MIMO antennas, advancing wireless communication for wearable IoT devices.