The lifetime of a wireless sensor network mainly depends on battery capacity and energy consumption at each node for operations such as, sensing, processing and communication. Popular approaches to save energy have been to intelligently duty cycle and restrict the frequency of these operations, rendering lower quality data at the sink. In this article, we propose Virtual Sensing Framework (VSF), which reduces the frequency of the above mentioned operations at each node while not compromising on the sensing interval, and hence resulting in higher quality data at the sink. VSF creates virtual sensors at the sink to exploit the spatio-temporal correlations among sensed data. Using an adaptive model at every sensing iteration, the virtual sensors can predict multiple consecutive sensor data for dormant physical sensors with the help of only a few active physical sensors. We show that even when the sensed data represents different parameters (e.g., light, temperature), our proposed technique works well. Applying our technique on the real-world data sets, we attain substantial reduction in energy consumption per node while maintaining high accuracy of the sensed data. To achieve higher energy reduction, VSF has to be used in conjunction with various layers and protocols of the communication stack. Thus, it has the potential to open up new research insights to make the best use of statistical properties of collected sensor data in a network.
Mendeley saves you time finding and organizing research
Choose a citation style from the tabs below