A multi-purpose robot perception system enabling closed-loop control for zero defect manufacturing in gluing processes of large parts

Abstract

Significant progress has been made in robot perception facilitating the deployment of advanced automation across a wide range of applications however typically little solutions are presented for large parts manufacturing. This paper presents a versatile robot perception system designed to enhance the flexibility and precision of robotic adhesive dispensing processes. This solution is capable of addressing the unique challenges of implementing automation solutions in large parts manufacturing, such as flexibility to manufacture small lot sizes, perception for complex task sequences, and handling parts with large dimensions that cannot be captured in single camera frames. A solution based on one single vision sensor is presented for part type recognition, part localization, process monitoring, and quality inspection. This includes algorithms for these perception functionalities and a closed-loop control framework aimed at zero-defect manufacturing. The task planning and execution architecture is based on Behavior Trees to allow modular and scalable robot task modeling and execution, whereas a knowledge database updated with process monitoring results via a module named event manager serves to prevent the propagation of defects to the following production steps. The proposed approach was tested and validated in a robotic cell for glue dispensing for a case study inspired by the bus and coach sector. The results indicate that the system can tolerate position uncertainties and random parts feeding, address disruptions in-process or trigger corrective actions post-process, and easily accommodate new variants.