Högskolan i Skövde

Electric cars are becoming increasingly popular worldwide for achieving net-zero carbon emissions. These vehicles have more cables than petrol cars, so new assembly processes have arisen in the automotive industry. Some tasks performed in these procedures are hard and repetitive, which causes injuries in the fingers, hands, and wrists of the operators who carry them out. This thesis seeks a solution to minimise injuries in the assembly of wire harnesses of bumpers. To do that, a GoFa collaborative robot and eye-tracking glasses, Pupil Labs Invisible, are integrated. With them, the robot dynamically adjusts to the intentions of the human by interpreting their gaze directions, focusing this way on Human-Robot Collaboration. To determine the most appropriate way to implement it, two computer vision methods are compared: colour detection using OpenCV and object detection using YOLOv8. Some experiments in controlled environments are conducted to deep-analyse both methods in terms of detection time, confusion in detection, and behaviour under changing lighting conditions. The results show that colour detection is faster and more accurate than object detection, and performs better in different lighting scenarios than YOLOv8. However, the feasibility of introducing the method in an industrial environment is also evaluated; therefore, YOLOv8 is determined as the most appropriate method. The safety of the operator when working near the robot is also considered, and communication protocols to send data from the glasses to the robot are presented. Finally, the problems faced and how they were solved, as well as the sustainable impact of the project, are discussed.