Target product grabbing of factory robotic arm based on digital twin technology

Abstract

Introduction – As smart manufacturing upgrades to flexibility and precision, factory robotic arm target grabbing is the core link of material transfer in the production line, and its performance directly determines production efficiency and product quality. However, the current factory robotic arms have insufficient gripping accuracy for target products and have weak anti-interference capabilities. Therefore, the study proposes a factory robotic arm target product grabbing model based on digital twin technology, aiming to break through the accuracy and anti-interference bottlenecks through virtual and real collaborative optimization. Methods – The research first builds a cross-platform communication mechanism between the robot operating system and Unity3D, and encapsulates the underlying protocol through rosSocket middleware. Secondly, a high-fidelity digital twin is built through SolidWorks modeling, 3DMax rendering and Unity3D assembly. The three-dimensional bounding box projection collision detection algorithm is combined to improve planning efficiency. Finally, the Rapidly-exploring Random Tree-Connect algorithm and tool center point calibration technology are integrated to build a composite grabbing model. Results – Experimental results showed that the average coordinate deviation of the center point of the proposed model tool was 2.34 mm. The capture success rate in normal scenarios was 90%, and the success rate in complex scenarios was 80%, which was 10% and 20% higher than that of mainstream deep reinforcement learning methods. The grabbing efficiency was 19.1% higher than that of the point cloud segmentation method, and the energy consumption and fixture wear rate were 25% and 47% lower than those of the traditional method. Discussion – This model achieves collaborative improvements in accuracy, efficiency, and economy through virtual and real collaborative optimization.