Accelerated and more precise sorting processes thanks to augmented reality and AI-based ARRANGE software

The ARRANGE software package has been designed to provide support for sorting processes using augmented reality. Parts that belong together are overlaid in color directly in the field of vision of employees on the production line. In this way, customers in the automotive sector can reduce costs by accelerating processes and minimizing error rates.

Digital support for sorting processes is in high demand in the automotive industry. The ARRANGE software package employs colors to simplify the sorting of sheet metal parts. Using augmented reality glasses, production staff are able to see which customer the punched objects are intended for. For example, the parts belonging to order A are overlaid in red, while those from order B appear in yellow. The tremendous potential of the solution becomes immediately apparent when sheets contain up to a hundred elements.

While existing solutions require actual photographs to train the artificial intelligence, ARRANGE uses 3D models. The software thus recognizes real objects solely on the basis of synthetic data.


Detection of objects and assessment of their position and orientation

Researchers developed the solution as part of a Software Campus project, working in collaboration with the tool manufacturer and automotive supplier Trumpf as an industrial partner. The software is based on methods of object detection and positional assessment.

Users benefit from ARRANGE in a number of ways. The artificial intelligence-based software speeds up the sorting process because production line operatives have the color-coded parts directly in their field of vision and no longer have to compare the sheet metal with images on an external display. Thanks to AR glasses and ARRANGE software, this previous step has been eliminated. “There is no need for abstraction, because the elements are directly overlaid in color in the workers’ field of view,” explains Fabian Rücker, a research assistant and doctoral student at Fraunhofer IGD. “This alone reduces the susceptibility to errors in the sorting process.” If an action is performed incorrectly, the software provides feedback.


Sorting support minimizes costs

Equally relevant is the performance of a concomitant quality assurance task, namely target-actual comparison. If parts do not correspond to the 3D models visualized, for example because they are uneven or have a faulty shape, this also generates an error message. Taken together, the acceleration of the work steps and the minimization of errors help the industrial user to reduce costs.