Publicación: Entrenamiento de IA para aplicación a robótica industrial: Generación de elementos mediante geometrías para entorno virtual
Archivos
Fecha
2022
Editor/a
Director/a
Tutor/a
Coordinador/a
Prologuista
Revisor/a
Ilustrador/a
Derechos de acceso
info:eu-repo/semantics/openAccess
Título de la revista
ISSN de la revista
Título del volumen
Editor
Universidad Nacional de Educación a Distancia (España), Universidad Politécnica de Madrid. Departamento de Ingeniería Mecánica
Resumen
La introducción de la Inteligencia Artificial y las Redes Neuronales en la industria resulta muy ventajosa, ya que posibilita la automatización de muchas tareas repetitivas que precisan de sentidos físicos. Las Redes Neuronales requieren ser entrenadas con datasets, que consisten en imágenes y datos numéricos de interés. No obstante, el desarrollo de técnicas de generación de datasets para componentes industriales se está realizando de manera independiente y aislada entre empresas, con sus consiguientes dificultades. Este artículo aporta una metodología sencilla y económica para la generación de datasets de objetos industriales que presenten cierta aleatoriedad en su disposición o forma, que ha sido llevada a la práctica en dos componentes distintos de dos empresas reales. La creación de estos datasets es muy versátil, pudiendo utilizarlos posteriormente para entrenar Redes Neuronales con distintos fines, como la generación de trayectorias para brazos robóticos o la Detección de Objetos, entre otros.
The use of Artificial Intelligence and Neural Networks in the industry is very advantageous, due to the fact that it enables the automation of many repetitive tasks that require human senses. Neural Networks are trained employing datasets, which consist of images and some data of interest. Nevertheless, the development of dataset generation techniques for industrial components is being carried out in an independent and isolated way among companies, with its consequent difficulties. This paper provides a simple and economical methodology for the generation of datasets of industrial objects that show some randomness in their layout or shape. This methodology has been put into practice in two different components belonging to two real companies. The creation of these datasets is very versatile, being able to use them to train different Neural Networks, such as the generation of trajectories for robotic arms or Object Detection, among others.
The use of Artificial Intelligence and Neural Networks in the industry is very advantageous, due to the fact that it enables the automation of many repetitive tasks that require human senses. Neural Networks are trained employing datasets, which consist of images and some data of interest. Nevertheless, the development of dataset generation techniques for industrial components is being carried out in an independent and isolated way among companies, with its consequent difficulties. This paper provides a simple and economical methodology for the generation of datasets of industrial objects that show some randomness in their layout or shape. This methodology has been put into practice in two different components belonging to two real companies. The creation of these datasets is very versatile, being able to use them to train different Neural Networks, such as the generation of trajectories for robotic arms or Object Detection, among others.
Descripción
Categorías UNESCO
Palabras clave
Modelos 3D, Dataset, Red Neuronal, Detección de Objetos, Robótica Industrial
Citación
Centro
E.T.S. de Ingenieros Industriales
Departamento
Mecánica