Decentralized sparsity-promoting sensor selection in energy harvesting wireless sensor networks

摘要

This paper considers the problem of sensor selection for the estimation of a stochastic source, being the sensor nodes powered by energy harvesting. Therefore, the interest lies in selecting the subset of most informative sensors that transmit their observations to a fusion center (FC). To that end, we propose to minimize the attained distortion at the FC plus a penalization term that promotes sparsity on the power allocation at the sensors. Then, we propose a decentralized algorithm in which the power allocation (and, thus, the selection policy) and distortion minimization problems can be regarded as separated problems. More specifically, the algorithm consists of: (i) a local computation of the power allocation policy, and (ii) a distortion minimization step. Moreover, for the case where sparsity is promoted via the classical ℓ1 norm, we show that the resulting local power allocation policy can be readily computed by means of a waterfilling-like algorithm.