Robot With Vision Guidance System for Unscrewing Operation

This work consists in identifying screws along a section of a metal beam, using a camera. Subsequently, the robot moved to the screws with tool (screwdriver) oriented and aligned with the screws, unscrewing them completely automatically. To detect the screws, a dataset of images contextualized with the application was created for personalized training of a neuronal network model. With this solution it is possible to detect and distinguish any type of screw, as the model can be continuously trained and optimized, following the logic of a scalable solution for the following phases of this development.