Residual energy-based adaptive data collection approach for periodic sensor networks

Due to its potential applications and the density of the deployed sensors, distributed wireless sensor networks are one of the highly anticipated key contributors of the big data in the future. Consequently, massive data collected by the sensors beside the limited battery power are the main limitations imposed by such networks. In this paper, we consider a periodic sensor networks (PSNs) where sensors transmit their data to the sink on a periodic basis. We propose an efficient adaptive model of data collection dedicated to PSN, in order to increase the network lifetime and to reduce the huge amount of the collected data. The main idea behind this approach is to allow each sensor node to adapt its sampling rate to the physical changing dynamics. In this way, the oversampling can be minimized and the power efficiency of the overall network system can be further improved. The proposed method is based on the dependence of measurements variance while taking into account the residual energy that varies over time. We study three well known statistical tests based on One-Way Anova model. Then, we propose a multiple levels activity model that uses behavior functions modeled by modified Bezier curves to define application classes and allow for sampling adaptive rate. Experiments on real sensors data show that our approach can be effectively used to minimize the amount of data retrieved by the network and conserve energy of the sensors, without loss of fidelity/accuracy.