Vision-guided robot alignment for scalable, flexible assembly automation

This paper introduces a vision-guided alignment method that aims at closing the gap between a manual assembly and a fully automatic assembly. It is intended to be used for assembly of small parts in a highly agile production scenario, which employs both human workers and robots in the same line, with a frequent need for reconfiguration. The proposed method is for precisely aligning assembly features for small part, tight tolerance assemblies which are typically found in electronic (consumer/industrial) products, by utilizing a robot guided by one or more cameras. The conventional methods of vision-guided robotics in assembly automation solutions (e.g. traditional calibration and visual servoing) cannot perform the assemblies with a grasp error and visually inaccessible assembly features. The vision-guided alignment system that utilizes a camera space manipulation control process, on the other hand, relies on the CAD model of the components with a local calibration method to achieve the high accuracy alignment for the final assembly. Experimental verification is performed to test the accuracy of the alignment, the robustness to grasp error, and the automatic initialization function. Under a two camera configuration, the developed 3D vision guided assembly robot system successfully assembles the ideal test parts within sub-millimeter tolerance with no manual calibration.