Robotic Insertion of Timber Joints using Visual Detection of Fiducial Markers

Abstract

Timber building industry is facing a major transformation with digitalization and automation being more broadly adopted. Prefabricated timber frame structures can be mass-produced on large robotic assembly lines, increasing the productivity and competitiveness of wood over standard inorganic materials such as steel or concrete. While standardization is likely to limit creativity, new digital tools, on the contrary, give the possibility to design and build complex and unique geometries. Recent research in bespoke digital prefabrication notably led to the development of Integrally Attached Timber Plate Structures (IATPS). This system consists in assembling wooden panels connected only with timber joints inspired by traditional Japanese carpentry. The elements are digitally prefabricated and inserted into one-another to form bespoke architectural structures. In order to propose a fully automated process for IATPS, from design to construction, this paper investigates a method for assembling the panels with a 6-axis robotic arm. Preliminary studies have shown that significant discrepancies can occur between virtual models and physical prototypes due to joint tolerances, hygrometric variations, and self-weight deformations. To address this challenge and to adapt the robot position to the actual location of the elements, a visual feedback loop was developed using fiducial markers. Several tests were performed with structural wood panels to assess the accuracy of the method for different configurations and adapt the geometry of the joints in consequence. Finally, the insertion of a panel with two through-tenon joints was achieved by taking pictures of the target with a camera mounted on the end-effector of the robot.