Automatic Correction System for Learning Activities in Remote-Access Laboratories in the Mechatronics Area

Abstract

In recent years, the educational field has evolved rapidly owing to the integration of several technologies, especially experiments in remote laboratories in the engineering area. Therefore, this article addresses the development of an innovation system for automatically correcting experiments in remote laboratories in mechatronics using digital twins, convolutional neural networks (CNNs), and generative artificial intelligence technologies. This system was designed to overcome the limitations of physical laboratories and teacher’s availability and assist in learning, enabling automatic acquisitions at any time. The digital twin captures data from the teacher’s and student’s experiments, allowing accurate comparisons to identify successes and errors. The application of CNNs serves to validate the results of the experiments through image analysis, whereas generative AI helps to identify patterns. The system was evaluated in a didactic plant, effectively correcting experiments with digital inputs and outputs. In addition, it provides students with detailed feedback on their performance, including specific errors and suggestions for improvement. With a three-layer architecture, i.e., experiments, didactics, and management, the system efficiently processes data from teachers and students, contributing to correcting experiments and optimizing teaching in remote environments.