Utility Maximizing Node Activation Policies in Networks of Partially Rechargeable Sensors

We address the problem of optimal node activation in a sensor network, where the optimization objective is represented by a time-average generalized coverage metric. We consider a scenario where each sensor is rechargeable, can hold up to K quanta of energy, and can be activated even if it is partially recharged. If the recharge and/or discharge process is random, the problem of optimal sensor activation in this context is a complex stochastic decision question. For the case of identical sensor coverages, we show that there exists a simple threshold policy that achieves near-optimal performance. We then extend this approach to a general sensor network, where coverage areas of different sensors could have full, partial or no overlap with each other. We demonstrate through simulations that a local information based threshold policy, with an appropriately chosen threshold, achieves a performance that is very close to the