Grasped Element Position Recognition and Robot Pose Adjustment during Assembly

Upgrading building envelopes with fully prefabricated 2D modules requires high accuracy during the manufacturing process with tolerances lower than 1 mm. In the research described in this paper, computer designed and accurately manufactured objects have been assembled with robotic arms. However, in previous phases, it was detected that during the assembly process, the placement differed from the planned location due to undesired deviations of the object while being grasped and placed by the robotic tool. The experiments presented in this paper imply correcting this deviation by localizing the grasped object’s position and recalculating the path and final pose of the assembly process. For localizing the deviated grasped object’s location, an intermediate pose was planned. During this pose, the location of the grasped object was estimated by two different means. For the first solution, visual markers have been placed on a known corner of the objects and these have been recognized with a camera. For the second solution, the coordinates of the objects were measured by a digital theodolite. The location of the deviated object was calculated and compared to the planned location so the robot could divert from its original path. The measurable parameter in the experiments was the assembly accuracy. The results in the two experiments have been analyzed and compared. According to the results, the solutions could be implemented not only in the factory, but at on-site processes as well.