Högskolan i Skövde

In the current industrial context, collaborative robotics, computer vision and digital twins are becoming increasingly important for developing flexible and safer manufacturing systems. This is especially relevant in battery-cell kitting, where components must be identified, sorted and assembled with the correct colour, position and polarity. This thesis aims to develop a vision-guided robotic kitting system using a UR10e collaborative robot, AI-based computer vision and a RoboDK digital twin. A workstation was designed where the robot can analyse battery kits, sort cells, handle lids and bridge pieces, and coordinate the process between the physical and the digital twin in RoboDK. To validate the system, two vision methods were implemented and compared: an ROI and colour-based approach, and an Artificial Intelligence vision model approach. The results showed that the first method was useful as an initial prototype but presented limitations in colour-dependent detections. The YOLOv8 method provided more reliable results and allowed a more flexible implementation of the kitting process. Finally, the results, limitations and sustainability impact of the system are discussed, showing the potential of combining collaborative robots, computer vision and digital twins for automated battery-cell kitting.