ETD

• ABB
• computer vision
• EGM
• HOG
• robotic cell
• ROS
• RWS
• stereo vision

This thesis project is in line with the concept of Industry 4.0 and consists in the creation of a vision-driven robotic system for the automation of a production process inside a plant of Luxottica Group.
The process consists in picking eyeglass frames from a conveyor belt and hooking them onto a frame using computer vision to dynamically determine the positions of interest, such as pick and place points. In particular, two fixed cameras and a stereo pair were used in the robotic cell and the HOG (Histogram of Oriented Gradients) descriptor was considered to implement the feature detection algorithm.
The cell is mainly composed of an ABB IRB 1200 robotic manipulator, equipped with a custom gripper designed for the application, along with other actuators such as linear guides and pneumatic actuators, allowing the automation of the process. The high-level control of the system is carried out from a PC which runs a ROS (Robot Operating System) application. It is connected via Ethernet to the low-level robot controller on which the EGM (Externally Guided Motion), RWS (Robot Web Services) and StateMachine Add-In functions are used to control and manage robotic cell. Control techniques are based on the visual servoing scheme, in particular of the look and move type. A position control has been implemented for the picking phase, in which the external PC sends vision-based position references to the robot controller for execution. Instead, for the hooking phase, a velocity control based on position error has been developed: using information from the vision system and the position feedback from the robot, a velocity reference is computed and sent to controller for actuation.