Inteligência artificial em sistemas embarcados

Abstract

The present study focuses on research that includes a theoretical, didactic, and practical analysis of the integration between Artificial Intelligence (AI) and Embedded Systems, with an emphasis on applying the TensorFlow Lite (LiteRT) framework to the ESP32 microcontroller. The motivation arose from the growing demand for intelligent and efficient solutions in an increasingly technological world, where the need to achieve better results in devices with limited resources becomes more significant and relevant, especially in Internet of Things (IoT) and automation applications. The main objective is to provide an easy-to-understand didactic material that can contribute to students, researchers, professionals in the field, and enthusiasts of the subject, addressing fundamental concepts and presenting a practical example of sine wave prediction using artificial neural networks. The methodology consists of adapting an already developed project, including the creation, training, and optimization of a Machine Learning (ML) model in TensorFlow, followed by the conversion of this model to the TensorFlow Lite format and its implementation on the ESP32 microcontroller. The results were promising, indicating high accuracy of the developed model and demonstrating the effectiveness of the proposed solution. It is concluded that TensorFlow Lite on ESP32 not only allows the execution of more complex Machine Learning models but also demonstrates excellent efficiency and proves to be a promising approach for developing solutions in embedded devices.