Energy Prediction Models for Wireless Sensor Nodes
Scientists of Tallinn University of Technology have introduced a novel energy prediction model based on sampling operators.
Recent research and development in IoT (Internet of Things), remarkably changed the entire scenario of computer networks or internet and evolves the infinite smart devices with embedded systems. Moreover, IoT is the combination of various devices. Generally, IoT is based on the Wireless Sensor Nodes. They can sense, then gather the data, afterwards send to the network. WSNs node is used to provide real time accurate monitoring, especially in human inaccessible locations.
Wireless sensor nodes operating without energy storage are desirable for the implementation of applications for which electrochemical batteries or supercapacitors are unsuitable due to concerns encompassing physical constraints, maintenance, safety, environmental regulations, etc. This calls for miniaturized energy harvesting technologies, which typically exhibit an intermittent behavior. It has been shown that applications that can tolerate such intermittent behavior can build upon transient computing techniques that allows saving and restoring the state and/or data of a node depending on the available energy. Independently, it has also been shown that energy prediction can be used to control e.g. task management in wireless sensor nodes, and thus better manage the energy/QoS trade-off.
A novel energy prediction model was introduced based on sampling operators. The main idea of the method is to approximate the smooth trend of the harvested energy profile via sampling series with high order of approximation. Numerical experiments showed that the proposed method has prediction accuracy, comparable or better than state of the art models.
A test device was installed at the Thomas Johann Seebeck Department of Electronics, Tallinn University of Technology, which implements a wireless peer-to-peer application where the above-mentioned approaches are uniquely combined. The two main elements used to achieve this were the selected energy prediction model and the proposed light-weight task management mechanism where transient computing and energy prediction are actually combined.
The implementation is based on FRAM-based wireless nodes that operate without energy storage. They are powered by a solar panel used in real-life conditions. The results show that the proposed implementation enhances the communication quality between the nodes, and consequently improve the application’s reliability. The results also highlight that careful tuning of various parameters, such as communication threshold, sampling period and prediction time, is needed to achieve the desired trade-off between adaptability and robustness.
Yannick Le Moullec, professor, head of the Research Laboratory for Cognitronics, Thomas Johann Seebeck Department of Electronics, Tallinn University of Technology, e-mail: firstname.lastname@example.org,
Gert Tamberg, senior research scientist, Division of Mathematics, Department of Cybernetics, Tallinn University of Technology, e-mail: email@example.com