Energy efficient collection of spectrum occupancy data in wireless cognitive sensor networks

We consider the problem of intelligent spectrum sensing based on historical data in the context of Cognitive Wireless Sensor Networks with the goal to minimize the long term energy consumed for spectrum sensing. The problem involves a wireless sensor with cognitive abilities that must decide how often to sense a specific region of the spectrum based on the last known state of its primary user. We model the problem at hand as a pure-exploration multi-armed bandit problem with dependent arms and use two well known arm pulling strategies to tackle it. Next, we identify the existing correlation structure among the candidate solutions and utilize it to significantly speed up the learning process. Numerical results verify the effectiveness of the proposed strategies.