Empowering Precision-Guided Automotive Assembly Operations: A Flexible Robot Vision Framework

Abstract

The ever-evolving advancements in machine vision, deep learning, and robotics have prompted the global automotive industry to adopt new automation strategies in order to meet the evolving demands while ensuring enhanced productivity. This research paper presents a comprehensive framework that incorporates two distinct detection modules tailored to address specific requirements of assembly operations. Both modules leverage three-dimensional vision sensors, with the first employing the template matching technique and the second one utilizing state-of-the-art deep convolutional neural network (CNN) for object detection. To assess its efficacy, the proposed framework was thoroughly tested and validated in a real-world automotive operation focusing on the assembly of a vehicle’s motor and gearbox. The results exhibit promising outcomes, as both incorporated modules were successfully demonstrated in a pre-industrial setup using authentic automotive components. The framework’s flexibility enables its potential application in different manufacturing settings, paving the way for improved accuracy, efficiency, and overall productivity as well as establishing a reasonable reference for future research.