Título: Convolutional Neural Networks for Autonomous Navigation of the CAT793F in a Mining Environment via Digital Model
Autores: Rhuan Lucas Alves Soares, Antonio Rola Neto, Levi Medeiros Magny, Giovani Bernardes Vitor, Denis Silva Loubach & Bruno Messias Gomes da Silva
Resumo: The mining industry is vast and operates on a global scale, demanding efficiency and, above all, safety in its mining operations. In this context, the development of technologies to ensure the protection of workers and the environment has become increasingly essential, given the presence of heavy vehicles, hazardous materials, and unpredictable conditions. In addition, high fuel consumption, vehicle and equipment maintenance, and the costs associated with specialized labor represent significant challenges for the sector. Thus, this work proposes the study of artificial intelligence tools aimed at learning the behavior of an off-road truck operator—specifically, the CAT 793F model—during the vehicle’s driving process. The objective is to develop a convolutional neural network model capable of performing autonomous navigation, trained using a dataset generated from the driving behavior of human operators within a controlled simulation environment. Therefore, a complete methodology was applied for the generation and acquisition of the dataset used to train the model, as well as the adaptation of the intelligent model to consider frames as inputs in order to produce control responses such as steering, braking, and acceleration. As a result, we observed that the trained AI model was able to achieve a Mean Squared Error Loss of around 0.0002, enabling autonomous navigation along a specified trajectory.
Palavras-chave: Convolutional Neural Network; Artificial Intelligence; End-to-end Navigation.
Páginas: 7
Código DOI: 10.21528/CBIC2025-1175696
Artigo em PDF: CBIC_2025_paper1175696.pdf
Arquivo BibTeX:
CBIC_2025_1175696.bib