Festo in German-Canadian collaboration to develop AI technology for picking robots

In the research project FLAIROP, the robots are trained with different articles in separate locations. The weights from all stations are collected and optimized using various criteria. Then the improved version is played back to the local stations and at the end, they should be able to grasp articles from other stations that they have not yet learned about.

Robotic picking is an important application in production, warehousing, and shipping. Festo is teaming up with researchers from the Karlsruhe Institute of Technology (KIT) and two Canadian partners – the University of Waterloo and software company DarwinAI of Waterloo, Ont. – in the Federated Learning for Robot Picking (FLAIROP) project, which seeks to enhance picking robots through the use of distributed Artificial Intelligence (AI) methods. The partners are researching how to use training data for robots acquired from multiple stations, plants, or even companies without needing participants to share confidential company information.

Four autonomous picking stations

The FLAIROP project is investigating how this training data can help develop more robust and efficient solutions using AI algorithms, so that picking robots are trained with varying objects and should eventually be able to grasp unfamiliar items from other stations. These potent algorithms will allow for the robust use of AI for industry and Logistics 4.0, while remaining compliant with data-protection guidelines. In other words, Festo and its partners are developing new ways for picking robots to learn from each other without exchanging sensitive or confidential info. Work flows will accelerate as these robots handle diverse tasks more quickly. This is particularly useful with collaborative robots working side-by-side with workers to relieve them of repetitive, heavy, and/or tiring aspects of their job.

As part of FLAIROP, four autonomous picking stations will be set up for robot training in Germany, two at the Festo campus in Esslingen am Neckar near Stuttgart, two at the KIT Institute for Material Handling and Logistics in Karlsruhe.

Up to now, federated learning has been mostly limited to the medical industry for image analysis, in which patient confidentiality is extremely important. It dispenses with the need for exchanging confidential data, like images or grasp points, in training artificial neural networks; rather, pieces of stored data are transferred to a central server to develop AI algorithms. The FLAIROP project opens the way to advance AI for robotics. With Festo’s help, picking is about to become better, faster, and easier.

For more information, contact Festo.

• Tweet