On sensor selection in linked information networks

Sensor networks are often redundant by design in order to achieve reliability in information processing. In many cases, the relationships between the different sensors are known a-priori, and can be represented as virtual linkages among the different sensors. These virtual linkages correspond to an information network of sensors, which provides useful external input to the problem of sensor selection. In this paper, we propose the unique approach of using external linkage information in order to improve the efficiency of very large scale sensor selection. We design efficient theoretical models, including a greedy approximation algorithm and an integer programming formulation for sensor selection. Our greedy selection algorithm provides an approximation bound of (e -1)/(2 · e -1), where e is the base of the natural logarithm. We show that our approach is much more effective than baseline sampling strategies. We present experimental results that illustrate the effectiveness and efficiency of our approach.