Battery Health Estimation for IoT Devices using V-Edge Dynamics

Deployments of battery-powered IoT devices have become ubiquitous, monitoring everything from environmental conditions in smart cities to wildlife movements in remote areas. How to manage the life-cycle of sensors in such large-scale deployments is currently an open issue. Indeed, most deployments let sensors operate until they fail and fix or replace the sensors post-hoc. In this paper, we contribute by developing a new approach for facilitating the life-cycle management of large-scale sensor deployments through online estimation of battery health. Our approach relies on so-called V-edge dynamics which capture and characterize instantaneous voltage drops. Experiments carried out on a dataset of battery discharge measurements demonstrate that our approach is capable of estimating battery health with up to $80%$ accuracy, depending on the characteristics of the devices and the processing load they undergo. Our method is particularly well-suited for the sensor devices, operating dedicated tasks, that they have constant discharge during their operation.