O uso de visão computacional para o controle de um manipulador robótico

Abstract

Robotic manipulators can support the automation of industrial processes, and due to this, they became broadly used in robotics, which also encourages research in the field. Currently, the majority of the robotic manipulators are limited to sequential movements whose control mesh is directly structured by the engineer. Any changes in its workstation can lead to problems, and a major challenge in this area has been making the manipulator capable of perceiving changes in the real environment and make decisions about which action will be held, based on the information collected in the physical environment. In this context, a simple computational strategy game between the robotic manipulator and a human player was developed, which integrates computer vision for the identification of objects and its location in space and inverse kinematics so that the manipulator can carry out the movements correctly. The adopted computer vision solutions were the color-based segmentation to distinguish the objects of interest. For the moving of the manipulator a fixed root point (anchor point) was adopted in the board, so that all the positions were calculated from that anchoring point.