Energy-Efficient Edge Intelligence for Batteryless IoT Sensors Powered by RF Harvesting
TL;DR
The authors propose an Energy-Efficient Edge Intelligence Framework to be utilised by existing batteryless RF-harvesting IoT sensors, which integrates adaptive energy management, ultralight-weight TinyML inference and duty-cycled computation to maximise sensing/processing performance under dynamic harvested energy availability.
Abstract
The tremendous growth of the Internet of Things (IoT) has led to several billion devices that are connected and desire steady, maintenance-free power. Conventional battery-enabled IoT nodes have limited lifetime, as well as replacement cost and environmental pollution challenges. In this respect, Radio Frequency (RF) energy harvesting appears to be a forward-looking solution as it allows batteryless operation by transforming surrounding electromagnetic energy into accessible electrical power. Nevertheless, incorporating edge intelligence on batteryless IoT devices is still challenging because of stringent energy constraints and intermittent power supplies. To that end, in this paper, the authors propose an Energy-Efficient Edge Intelligence Framework (E ${ }^{\mathbf{3}} \mathbf{I F}$) to be utilised by existing batteryless RF-harvesting IoT sensors. The proposed hybrid framework integrates adaptive energy management, ultralight-weight TinyML inference and duty-cycled computation to maximise sensing/processing performance under dynamic harvested energy availability. Simulation results show that the proposed framework is 32% more energy-efficient, reduces inference latency by 24%, and increases operational uptime by 41%, as compared to the radio-frequency (RF)-powered sensor architecture currently known. The proposed approach offers a feasible path towards sustainable intelligent IoT deployments in smart cities, industrial monitoring and healthcare strategies.