Generative Adversarial Networks and Semantics for Resilient, Flexible Production Robots
Nowadays the industry is rather automated and possesses a large number of industrial robots, that are precise, fast and powerful. Nevertheless, the pandemic has clearly shown a lack in another key area: flexibility. When global supply chains collapsed, industrial production experienced significant difficulties. Companies were unable to make up for missing parts and materials or to quickly switch to new products with available parts and components. An additional critical aspect was the lack of on-site workers, who had to work in home offices due to pandemic regulations.
The goal of the GANResilRob approach is to create a flexible and resilient industry based on the combination of AI technologies such as machine learning, Generative Adversarial Network (GAN), AI-based semantic interpretation and intuitive task programming. These technologies have the potential to support rapid manufacturing of new products or reconfiguration of production lines due to changing supply chains. AI-based assembly methods and technologies can increase the flexibility and efficiency of robotic applications in manufacturing and disassembly by creating intelligent robotic production cells.
The project will also make it possible to integrate general knowledge about the assembly and disassembly process into an adaptive GAN and apply it to a wide range of object types, such as the vehicle disassembly. This is an important contribution, as the automated recycling of electronic or mechatronic components has the potential to drastically reduce the need for new resources, which consequently leads to a reduction in CO2 emissions by saving these resources.
Germany: FZI Forschungszentrum Informatik (lead), ArtiMinds Robotics GmbH
France: Two-I SAS (lead), CNRS IRL2958 GT-CNRS, Secma SA
March 2022 – February 2025
Total costs: € 2,3 million
Funding: € 1,6 milllion