I. Introduction
In wireless sensor networks (WSN), energy is a scarce resource. To guarantee prolonged unattended deployment, sensors need to be frugal with respect to energy usage. Sensing, computing and communication need to be minimized to ensure maximum network lifetime. In addition, it may be infeasible to collect and process all the possible sensor data due to bandwidth and computational constraints. Therefore, it is of practical interest to activate only a limited number of sensors at any time instant. The power constraint dictates how many sensors should be operating. For any particular application, the following question arises: which is the optimal subset of sensors to choose in order to maximize a pre-specified performance metric?