"We are investigating how the most versatile training data possible from multiple locations can be used to develop more robust and efficient solutions with the help of artificial intelligence algorithms than with data from just one robot," says Jonathan Auberle from the Institute of Materials Handling and Logistics Systems (IFL) at KIT. Autonomous robots process items at several picking stations by gripping and moving them. The robots are trained with different articles at the various stations. At the end, they should also be able to pick items from other stations that they have not yet learned to pick. "The distributed learning approach, also known as federated learning, allows us to strike a balance between data diversity and data security in an industrial environment," explains Auberle.

The FLAIROP (Federated Learning for Robot Picking) project is a partnership between Canadian and German organizations. The Canadian project partners focus on object recognition through deep learning, explainable AI and optimization, while the German partners contribute their expertise in robotics, autonomous grasping through deep learning and data security. FLAIROP is funded by the Canadian National Research Council (NRC) and the German Federal Ministry for Economic Affairs and Energy (BMWi).

To date, federated learning has mainly been used in the medical sector for image analysis, where the protection of patient data is particularly important, explains Auberle. Consequently, there is no exchange of training data such as images or grasping points for training the artificial neural network; instead, only the local weights of the neural network, i.e. parts of stored knowledge, are transferred to a central server. There, the weights from all stations are collected and optimized using various criteria. The improved version is then transferred back to the local stations and the process is repeated. The aim is to develop new, more powerful algorithms for the robust use of artificial intelligence for Industry and Logistics 4.0 while complying with data protection guidelines.

During the project, a total of four autonomous picking stations will be set up to train the robots: two at the Institute of Materials Handling and Logistics Systems at KIT and two at Festo in Esslingen. Other partners are the Institute for Applied Informatics and Formal Description Methods (AIFB) at KIT, Darwin AI and the University of Waterloo.

"In the FLAIROP research project, we are developing new ways for robots to learn from each other without sharing sensitive data and trade secrets. This has two major benefits: We protect our customers' data and we gain speed because the robots can take on many tasks faster this way. For example, the collaborative robots can support production workers with repetitive, heavy and tiring tasks," says Jan Seyler, Head of Advanced Develop. Analytics and Control at Festo.