Autonomous Robotic Assembly of a Skateboard Truck

Autodesk Research is exploring flexible and adaptive robotic systems for automated assembly. We combine vision and force sensing with deep-learning based perception and learned behaviors in a systems based approach to assembling complex product.

In this project, we showcase some of our initial work on assembling a skateboard truck requiring coordinated assembly of multiple parts and complex force insertion. We’ve designed a flexible workcell with multiple robots, tool changers, force torque sensing and overhead and in-hand cameras. We use deep-learning based perception to locate some parts while other parts are picked from known-locations in kits. We also use our perception capabilities to re-grasp parts using multiple robots so we can re-orient them for insertion. Our long-term research goal is to automate the CAD to make pipeline, simplifying the workflow of going from designed products to the process of automating their assembly.