GreenEdge-IoT: Carbon-Aware Edge AI Framework for Sustainable IoT-Based Green Computing

Abstract

The rapid growth of IoT-based services has increased edge-processing demand while also raising energy consumption and carbon emissions. This study presents GreenEdgeIoT, a carbon-aware edge AI framework that balances environmental sustainability with latency, energy, task priority and retained AI quality. The framework profiles incoming IoT tasks, estimates edge workload and effective processing capacity, evaluates communication and computation energy, maps energy use to renewable-adjusted carbon cost and selects feasible execution nodes through a weighted decision function. A tunable carbon-sensitivity parameter enables the allocation policy to adapt between performance-oriented and low-carbon operation. Experiments conducted using ESP32 sensing nodes, Raspberry Pi edge gateways and a Jetson Nano device show that GreenEdgeIoT reduces mean carbon cost by 26.14% over Conventional Edge and 41.11% over Cloud-Centric IoT. It also lowers latency by $\mathbf{1 6. 3 5} \boldsymbol{\%}$ and $\mathbf{3 7. 6 6} \boldsymbol{\%}$, respectively while maintaining $\mathbf{9 3. 2 9} \boldsymbol{\%}$ retained AI quality and 99.45% task acceptance. These findings demonstrate that carbon-aware edge intelligence can support reliable, energy-efficient and sustainable computing for smart cities, healthcare monitoring, industrial IoT and green cyberphysical systems.