DRL for Energy, Latency, and Throughput Optimization in UAV-Assisted WBANs

Wireless body area networks (WBANs) facilitate real-time physiological data collection, while the Internet of Medical Things (IoMT) leverages this data for enhanced healthcare service delivery. The primary challenge for such a network is ensuring timely and reliable transmission of patients' data to remote monitoring centers without disrupting other activities. As part of the solution, unmanned aerial vehicles (UAVs) are becoming widely deployed in remote monitoring of patients' vital signs in large-scale IoMT due to their flexibility, maneuverability, strong line-of-sight, and cost-effectiveness. However, UAV-assisted WBANs face challenges, including reliable and timely data transmission, limited radio resources, interference, high energy consumption, and optimal positioning for efficient coverage and connectivity. This work tackles the challenge of joint subcarrier allocation, WBANs uplink transmission power, and UAV's flying heights (SUF) in UAVassisted WBAN scenarios. The SUF optimization model is formulated to enhance spectrum, time, and energy efficiency. We leverage non-orthogonal multiple access techniques to achieve spectrum efficiency and consider minimum and maximum physical resource blocks to cater to physiological data criticality. Furthermore, a centralized deep reinforcement learning (DRL) algorithm is employed to develop the SUF strategy for optimal energy, latency, and throughput. Simulation results demonstrate that SUF-optimization approach achieves faster convergence, superior energy consumption, and near optimal total system capacity compared to baseline resource allocation approaches.