Energy-efficient quickest change detection in sensor networks

We propose a minimax formulation for quickest change detection in sensor networks, where there is a cost associated with taking observations at each sensor. We present two energy efficient algorithms, centralized and distributed, in which on-off observation control is employed and the change is detected under a constraint on the false alarm rate. We show that these algorithms are asymptotically optimal as the false alarm rate goes to zero and have good performance compared to the straightforward approach of fractional sampling in which the observations are selected randomly.