Building an AIoT system emulator with Zephyr and QEMU

Abstract

Internet of Things (IoT) and Artificial Intelligence (AI) are two of the emerging technologies of the recent years. While each technology is having its own prosperity in the tech ecosystem, innovators are attempting to combine them as one by incorporating AI into IoT systems with the name of AIoT.

The project aims to design a Proof-of-Concept AIoT system with an applicable problem statement in real life. The project specifically experiments testing a variety of Machine Learning (ML) architectures based on an IoT foundation system that simulates collecting real-time data to clarify the relationship between model accuracy, execution time and memory size in a constrained manner.

Zephyr Real-Time Operating System (RTOS) will be leveraged to develop the firmware at the client side and tested inside a QEMU emulated environment, and a novel approach of model structuring using Keras and deployment using C programming onto edge clients will be applied.