Optimising the Wireless Body Sensors Network (WBSN) of Elderly Health Monitoring System Through Actor-Critic Mechanism

Elderly people represent a vulnerable population requiring continuous, reliable and timely health monitoring to maintain quality of life and reduce medical risks. Although existing wireless body sensor network (WBSN)-based systems primarily focus on energy efficiency, limited attention has been given to optimising time efficiency and real-time decision-making performance. This study proposes an elderly health monitoring system based on WBSN using an actor-critic deep reinforcement learning (ACDRL) framework to address this gap. The system utilises physiological state parameters, including heart rate, body temperature and oxygen saturation, to dynamically optimise monitoring and data transmission strategies. Device validation experiments demonstrated high sensing accuracy with a mean absolute percentage error (MAPE) of 0.68%. Model optimisation results indicate that a discount factor of 0.4 yields the best performance, achieving a mean absolute error (MAE) of 0.0401. Comparative evaluations against deep Q-network (DQN)-based, deep deterministic policy gradient (DDPG) and twin delayed DDPG (TD3) algorithms show that the proposed ACDRL model consistently outperforms existing approaches across mean absolute error (MAE), integral absolute error (IAE) and integral squared error (ISE) metrics. Furthermore, real-world WBSN implementation involving multiple sensor nodes confirms that the proposed method significantly reduces time consumption, recording 1104 ms with 10 nodes lower than all benchmark models. These results demonstrate that the proposed ACDRL-based WBSN framework provides a scientifically validated, time-efficient and scalable solution for real-time elderly health monitoring, contributing to the advancement of intelligent healthcare systems.