Visionsstyrd robotruttning av kablar i 3D-miljöer : Ett demonstrationsbaserat angreppssätt för kabelmanipulering med rörelseprimitiver ochvisuell återkoppling

Cable routing remains one of the most time-consuming and labour-intensive tasks in industrial production environments, such as vehicle assembly, where consistency, flexibility, and accuracy are critical. Automating this task can reduce operator workload, improve long-term repeatability, and enable reconfigurable production lines with minimal human supervision. This thesis presents a vision-guided robotic system for semi-autonomous cable routing in three-dimensional workspaces containing discrete fixtures that constrain the cable’s intended path. The system is implemented on a Franka Emika Panda robotic arm and performs cable manipulation through a sequence of predefined motion primitives. These primitives are not learned dynamically but are instead selected based on human demonstrations, which define key poses and associate them with specific actions such as cable pickup or insertion into a fixture. A fixed camera provides point cloud data to detect the initial pickup point on the cable, while a wrist-mounted camera is used for visual servoing during the fixture insertion phase. This visual feedback enables the system to compensate for small inaccuracies in demonstrations or deviations in fixture positions. The system was evaluated in scenarios involving both horizontal and vertical fixture placements, showing that reliable cable routing can be achieved using a single-arm setup. The work serves as a proof-of-concept for flexible automation in industrial settings.