Production, storage and dispatch – wherever goods are manufactured, sorted or packed, there is also a picking process involved. As part of this, robots often take individual goods out of crates and put them into new groups. Festo has teamed up with partners from Germany and Canada as part of the FLAIROP project to research how to increase the intelligence of picking robots with distributed AI methods. They are investigating how training data from multiple stations, factories or companies can be used without having to release sensitive company data.
"We are investigating how the most versatile training data possible from multiple locations can be used to develop more robust and efficient solutions using artificial intelligence algorithms than with data from just one robot," says Jonathan Auberle from the Institute for Material Handling and Logistics (IFL) at Karlsruhe Institute of Technology (KIT).
As part of this, autonomous robots process items by gripping and moving them at multiple picking stations. The robots are trained using totally different items at the various stations. At the end, they should also be able to grip items from other stations that they have not yet learned about. "By using distributed learning, also called federated learning, we are able to strike the right balance between the amount of data and the security of data in an industrial environment," says Jonathan Auberle.
Until now, federated learning has been used predominantly for analysing images in the medical sector, where the protection of patient data is naturally a particularly high priority. There is therefore no exchange of training data such as images or grasping points for training the artificial neural network. Only parts of the stored knowledge are sent to a central server – the local weights of the neural network that indicate how strongly one neuron is linked to another. The weights from all the stations are collected in the central server and are optimised with the help of various criteria. The improved version is then fed back to the local stations and the process repeats itself.
The aim is to develop new, better-performing algorithms that will enable the robust use of artificial intelligence for Industry and Logistics 4.0 while complying with data protection guidelines.
"In the FLAIROP research project, we are developing new ways for robots to learn from each other, without sharing sensitive data and company secrets. This has two major benefits: we protect our customers’ data and we gain speed because the robots can perform many tasks more quickly. The collaborative robots can thus support production workers with repetitive, difficult and tiring tasks, for example," says Jan Seyler, Head of Advanced Develop. Analytics and Control at Festo.
"DarwinAI is thrilled to be able to make our Explainable (XAI) platform available to the FLAIROP project and pleased to work with such reputable Canadian and German academic organisations and our industry partner, Festo. We hope that our XAI technology will enable high-value human-in-the-loop processes for this exciting project, which represents an important aspect of our offer alongside our novel approach to federated learning. Because of our roots in academic research, we are enthusiastic about this collaboration and the industrial benefits of our new approach for a range of manufacturing customers," says Sheldon Fernandez, CEO of DarwinAI.
"The University of Waterloo is very excited to be working with the Karlsruhe Institute of Technology and a global leader in industrial automation such as Festo to bring the next generation of trustworthy artificial intelligence to manufacturing," says Dr Alexander Wong, Co-Director of the Vision and Image Processing Research Group, University of Waterloo, and Chief Scientist at DarwinAI.
"By harnessing DarwinAI’s Explainable AI (XAI) and federated learning, we can create AI solutions that will help support factory workers in their daily production tasks to maximise efficiency, productivity and safety."