Poster abstract: MDP framework for sensor network coordination

We consider a body area network application of monitoring a patient continuously [1,2]. In this application, several sensors are used in measuring physiological and metabolic readings of a patient. In such a system, the sampling rates have to be coordinated to maximize the life time of this sensor network system. We formulate the relationship between energy consumption, sensor sampling rates, and utility of coordination as a Markov Decision Process and compute a globally optimal policy for sensor sampling rates. We then present an entropy-based mechanism for communication between nodes to execute this global policy. We show preliminary results on simulated data to demonstrate that this distributed control framework is feasible for a limited number of sensors.