Online Power Scheduling for Distributed Filtering Over an Energy-Limited Sensor Network

In this paper, the problem of power allocation is considered for distributed estimation over a wireless sensor network with limited power. To utilize the power efficiently, an online power scheduling scheme is proposed for consensus-based distributed filtering, where each communication channel is allocated a certain power based on the real-time innovation of the sensor who transmits the data. First, the Gaussian property of the innovations is investigated under the online power scheduling scheme, and an optimal estimator gain is obtained for each sensor by minimizing the state estimation error covariance. Then, a sufficient condition to guarantee the stability of the proposed estimator equipped with the online power allocation scheme is identified, which provides a lower bound of the power needed to guarantee that all the sensors could achieve a given estimation accuracy. Finally, the estimation performance of different power scheduling schemes are compared, and the theoretical results are verified by numerical examples.