Optimal Sensing for Patient Health Monitoring

In this paper, we construct a framework for optimally sensing a patient's health state with a wireless body area network (WBAN). In such a resource-constrained paradigm, it is often necessary to use lower performance sensing modes, conditionally reducing system performance to increase efficiency and maximize system lifetime. The optimal control architecture trades between shallow and deep sensing modes according to the estimated patient health state, minimizing the expected costs over future states. We construct an implicit formulation for deriving the optimal sensing policy via dynamic programming, an easily implemented myopic sensing policy, and a useful performance bound for evaluating near-optimal approximate sensing policies. We further provide an Monte Carlo experimental evaluation of how such policies depend on key model parameters.