Energy-Efficient Activation/Inactivation Strategy for Long-term IoT Network Operation

Abstract

A critical concern for IoT (Internet of Things) networks is how to operate a large number of sensors in an energy-efficient manner. Energy consumption caused by operating these sensors cannot be overlooked, especially for long-term IoT network operation. Inactivating some of the unneeded sensors during the operation is one of the feasible solutions. Recent studies have mainly focused on selecting the unneeded sensors in a duty-cycled manner based on the sensor's geographic location or the sensor's communication status. These studies involve obtaining some of the user's private information, moreover, anomaly detection performance, a significant criterion regarding IoT network operation performance has not been considered. In this paper, we develop a novel activation/inactivation strategy for long-term IoT network operation. In this strategy, a machine learning model is adopted to adaptively select the unneeded sensors to be inactivated during network operation. Moreover, to maintain high sensor data processing performance during long-term operation, by only using the basic sensor data (e.g., temperature and humidity), our proposal selects the unneeded sensor candidates which are periodically updated. Numerical experiments conducted in two IoT network environments over seven-month operation verify the effectiveness of our proposal.