A Novel Graph-Based Energy Efficient Sensor Selection Scheme in Edge Computing

Abstract

IoT sensors are usually used for data collection and monitoring in various environments over the well-known edge computing architecture. However, the sensors' lifetime is a significant challenge when the sensors are battery-powered. To reduce energy consumption, this paper presents a novel graphbased sensor selection scheme to select working and sleeping sensors that maximizes the number of sleeping nodes while keeping the accuracy comparable with the state-of-the-art methods using a hierarchical prediction scheme. Moreover, an entropy-aware publishing method is proposed to reduce the edgeto-cloud transmissions to avoid transmitting predictable measurements to the cloud by utilizing the temporal correlation in a sensor measurement. The experimental results showed that the proposed method could surpass the previous approaches by turning off 14% more sensors on average under equal accuracy. Moreover, the entropy-aware publishing method could reach 44% average saving in data transmission.