Optimizing Wearable Textile Antennas for Accurate Long-Range Localization Using Interpretable Generalized Additive Neural Network

Abstract

The wearable monopole antenna is for long-range localization at the specific frequencies of 915 and 923 MHzwas built on a Rogers Duroid platform. It provides great efficiency, directed radiation, and low specific absorption rate (SAR) values, ensuring safety compliance. Real-world tests revealed a 3-dB gain in signal strength and a 1.5-km localization range. This makes the antenna a promising option for dependable LoRa-based tracking in a variety of environments. In this manuscript, optimizing wearable textile antennas for accurate long-range localization using interpretable generalized additive neural network (OWTA-ALRL-IGANN) is proposed. The IGANN is used to predict the $S_{11}$ response of the antenna. As therefore, antenna performance is increased, design time is decreased, and complex data interactions can be managed more easily. Finally, the performance of OWTA-ALRL-IGANN method attains 19.11%, 17.21%, and 18.24% higher bandwidth; 18.23%, 19.20%, and 17.20% higher SAR; and 16.11%, 17.19%, and 15.21% lower return loss when comparing with existing techniques like machine learning–optimized wearable antenna for LoRa localization (ML-OWA-LL), optimization of a compact wearable LoRa patch antenna for vital sign monitoring in WBAN medical applications utilizing ML (LoRa-VSM-WBAN-ML), and a compact textile monopole antenna for monitoring the healing of bone fractures utilizing un-supervised machine learning algorithm (CTMA-BF-UMLA), respectively.