Saving energy in wireless sensor networks based on Echo state networks

Abstract

Prolonged lifetime, robustness and scalability were important requirements in Wireless Sensor Networks (WSNs). We investigated the problem of energy-saving in Wireless Sensor Networks using Echo State Networks (ESN). In this research field, one key factor of the problems was how to save energy efficiently in battery-driven sensor nodes. We tried to present an approach addressing these difficulties based on ESN learning information of these sensors' history status when only the data was available. Echo State Networks utilized incremental updates driven by new sensor readings and massive short memory with history inputs, thus varying communication rates can help save energy. We evaluated this method against those traditional approaches to save energy, and observe that the quality of the overall operation was comparable to the approaches. Therefore, the ability of Echo State Networks to prolong lifetime during the sensor network operation made this approach more suitable and applicable. © 2009 Springer Berlin Heidelberg.