The AutoVikki project, funded by the German Federal Ministry of Education and Research under grant 01IS20039D, ran from 1 July 2020 to 31 December 2022. Its overarching aim was to create an autonomous, visually guided industrial robot cell capable of handling highly variable, fine‑grained objects such as forged titanium parts. The consortium was led by Arnold IT (AIT) as the developer and consortium leader, with August Rüggeberg GmbH & Co. KG (ARG) as the project recipient. Other partners included Bock Bio Science GmbH (BBS) for scenario A (autonomous in‑vitro plant multiplication) and Otto Fuchs KG (OFK) together with ARG for scenario B (automatic post‑processing of forged components). The project also involved the Hochschule (HS) and other stakeholders, all contributing to the design, implementation, and validation of the system.

Technically, AutoVikki integrated a suite of 3D cameras and advanced artificial‑intelligence algorithms to enable the robot to perceive, classify, grasp, and manipulate objects with minimal training data. Deep‑learning techniques such as Generative Adversarial Networks, domain randomization, and federated learning were employed to accelerate the learning of surface and structural features, thereby reducing the need for extensive labeled datasets. A digital twin of the processing tool was developed and stored within the system, allowing real‑time monitoring and feedback of tool performance and process parameters. This digital twin facilitated the automatic adjustment of the robot’s motion and force control, improving both efficiency and part quality.

The mechanical redesign of the post‑processing workflow was a key outcome. The original manual band‑sanding process, which required a worker to guide each part along a stationary abrasive belt, suffered from a very short belt life of only 6 to 10 parts before replacement. This limited automation potential. Through extensive screening and testing, the team replaced the band‑sanding with side‑sanding using fiber abrasives mounted on a support spindle, and COMBIDISC abrasive discs on a disc spindle for defect removal. The chosen abrasive was VICTOGRAIN with a COOL coating, a triangular high‑performance grain that offered superior cutting performance on titanium’s hard, low‑thermal‑conductivity surface. The screening trials were conducted on a Mammut Electronic ME 22/240 variable‑speed shaft drive to identify optimal spindle speeds. These changes eliminated the need for manual belt changes, increased throughput, and reduced downtime.

System tests in scenario B demonstrated significant improvements in process control, performance, and robustness. The robot cell achieved a Technology Readiness Level of 5, indicating successful operation in a realistic test environment. The integration of sensor data, real‑time analysis, and closed‑loop control allowed the system to adapt to variations in part geometry and surface finish, maintaining consistent quality across batches. The digital twin’s feedback loop further refined tool wear monitoring and predictive maintenance, extending tool life and reducing unplanned stops.

Overall, the project delivered a validated, semi‑autonomous robot cell that can be rapidly reconfigured for different fine‑grained workpieces. By combining advanced vision, AI, and digital twin technologies with a redesigned mechanical workflow, AutoVikki achieved a robust, efficient, and flexible solution for industrial post‑processing, meeting the needs of both the automotive and biomedical sectors represented by its partners.