Wireless Body Area Network Control Policies for Energy-Efficient Health Monitoring

Wireless body area networks (WBANs) have strong potential in the field of health monitoring. However, energy consumption required for accurate monitoring determines time between battery charges wearable sensors, which is a key performance factor (and can be critical case implantable devices). In this paper, we study inherent trade-off power sensors and probability misclassifying patient’s state. We formulate as dynamic problem, at each step, choose to activate subset that provide noisy measurements assume (unknown) state follows Markov chain, so our problem formulated partially observable decision (POMDP). show all past summarized belief on true patient, allows tackling POMDP an MDP Then, empirically greedy one-step look-ahead policy compared optimal obtained by solving program. For purpose, use open-source Continuous Glucose Monitoring (CGM) dataset 232 patients over six months extract transition matrix sensor accuracies from data. find saves ≈50% costs while reducing misclassification less than 2% most possible always activates sensors. Our sensitivity analysis reveals remains nearly across different cost parameters varying number The results also practical importance, because too complicated, easily implemented WBAN systems.