An Adaptive Reinforcement Learning-Based Mobility-Aware Routing for Heterogeneous Wireless Body Area Networks

Over the past few years, wireless body area networks (WBANs) have become increasingly popular for monitoring physiological parameters and health-related data of individuals using wearable sensors. Providing timely support to WBAN users requires mobility management to ensure seamless data transmissions. Considering the high demand for heterogeneous WBAN, vehicular ad hoc networks (VANETs) have emerged as promising networks for transporting the healthcare data of passengers to remote healthcare centers. Owing to the high mobility of vehicles and frequent changes in the network topology, VANETs have difficulty in establishing and maintaining end-to-end (E2E) connections. The limitations of conventional position-based routing protocols in VANET make the established routes invalid, which interrupts the communication flows, causing additional delays and overheads, making them unsuitable for routing health data packets. In this study, we introduce an adaptive reinforcement learning (RL)-based mobility-aware routing (ARMR) protocol for heterogeneous WBAN. First, we developed a mobility-aware routing strategy called link lifetime estimation (LLE), which predicts the link lifetime metric for efficient and reliable communication. LLE provides an estimation of how long the nodes remain within the transmission range of one another. Through this method, ARMR can choose nodes that provide a longer connection period, and thus, a higher level of reliability. Furthermore, we proposed an adaptive Q-learning approach that allows adaptive routing decisions in highly dynamic networks. According to the simulation results, our algorithms can make effective routing decisions in highly dynamic wireless networks, resulting in improved performance metrics in terms of the packet delivery ratio (PDR), E2E delay, overhead, network lifetime, and rapid convergence.