Light-Weight ML Aided Autonomous IoT Networks

Abstract

A common challenge in Internet-of-Things (IoT) networks is managing and connecting a large number of field deployed IoT nodes. With the massive growth of IoT applications, any human intervention is even more impractical. Hence, node- and network-level coordination for low-cost sensing, energy sustainability, self-organized fault remediation capability, auto-calibration of field nodes are some of the desirable features. Aiming at autonomous IoT, this article first presents the recent progress on light-weight machine learning aided strategies for context-aware IoT applications, wherein, depending on the application context and adaptation requirement, the data-driven intelligence operates at the field nodes or at the nearby edge node or at the cloud storage. A few motivating results towards operational autonomy and network scalability on the chosen use cases are presented. Next, the requirements towards fully autonomous and self healing IoT networks are presented, highlighting several future research directions and challenges.