Automated Energy Modeling Framework for Microcontroller-Based Edge Computing Nodes

When IoT-enabled applications utilized edge nodes rather than cloud servers, they aimed to apply diligent energy-efficient mechanisms on edge devices. Accordingly, frameworks and approaches that monitor/model microcontrollers, including Espressif-Processor-based (ESP) edge nodes, have drawn mainstream attention among researchers working in the edge intelligence domain. The traditional approaches to measuring the energy consumption of edge nodes are either not online or prone to complex solutions. This article attempts to develop an Automated Energy Modeling Framework (AEM) for microcontroller-based edge nodes of IoT-enabled applications. The proposed approach baselines the energy consumption values; models energy consumption values of components using a random forest (RF) algorithm; and, automatically suggests the energy consumption of edge nodes in real-time - i.e., during the execution of IoT-enabled applications on edge nodes. Experiments were carried out to validate two applications' automated energy modeling approach using Espressif's ESP devices. The proposed mechanism would benefit energy-conscious IoT-enabled application developers who focus on minimizing the energy consumption of embedded-based edge nodes such as ESPs.