Integrando TinyML em Veículos Flex: Novas Perspectivas para Eficiência Energética e Controle de Poluentes

Abstract

In the automotive industry, the increasing demand for energy efficiency and the reduction of CO2 emissions make flex-fuel vehicles a promising alternative, despite the challenges in optimizing their efficiency and minimizing emissions. This study proposed a methodology based on machine learning to estimate wheel efficiency by CO2 emissions, utilizing algorithms such as Decision Tree, Random Forest, and Multilayer Perceptron in a TinyML-oriented vehicle diagnostic system. The Decision Tree stood out for its shortest inference time (4 µs), lowest power consumption (248.02 mW), and a mean absolute error of 0.30, while the Random Forest had the shortest compilation time (46 s) and the lowest RAM usage (23,496 bytes). The MLP Float32, on the other hand, presented the highest accuracy with a MAE of 0.27. These results indicate that, although there are trade-offs between inference time, power consumption, and accuracy, the Decision Tree and Random Forest models are particularly promising for embedded systems where energy efficiency and resource usage are crucial.