Abordagem para avaliar o comportamento do motorista em tempo real com TinyML

Abstract

The significant increase in the number of vehicular sensors results in a growing volume of data, leveraging the convergence with Internet of Things (IoT) technologies to enable real-time edge analytics through an OBD-II Edge device. In this context, this study aims to develop, embed, and validate a real-time vehicular data processing solution for determining driver behavior. To achieve this, a three-layered approach was devised, utilizing soft sensors and integrating the Typicality and Eccentricity Data Analytics (TEDA) framework and the Adaptive Autocloud algorithm into a low-power hardware, with a focus on TinyML techniques. A case study was conducted in Natal-RN, Brazil, in a real-world scenario involving two participants and incorporating the proposed approach into the Freematics One+. The analyzed results were promising in classifying driver behavior, capturing significant nuances throughout the journey. In conclusion, this study contributes to real-time analysis of driver behavior during driving.