Research on key technologies of random stacking components automatically crawl

Abstract

Given the growing demand for intelligence in industrial production, the key technology of robotic arm automatically crawling is being studied and realized using random stacking components as the research object. To begin, the global feature descriptor PPF algorithm is used to extract part features. The visible points that meet the conditions are then combined pairwise in the off-line training stage, and the characteristics of point pairs are calculated to obtain the model describing the object's global information. The components are identified during the online matching stage by employing a voting strategy based on the Hough transform. The RANSAC algorithm is then used for rough pose estimation, and the ICP algorithm is used to fine-tune the pose result to obtain the target's optimal pose estimation. Finally, the transformation matrix after hand-eye calibration based on differential evolution algorithm determines the position and posture of the parts in the real world coordinate system, guiding the manipulator to accurately grab and place the stacked parts. A grab system was built to test it as part of a water pump production line reconstruction project. Using the pump body as an example, the results show that the number of features of the selected point pair is 730,830, the matching accuracy can reach 94.64%, and the matching time consumption is 1.1424 s; the overall average grabbing success rate is 92.3%, which is practical.