Robot training centre opens

Initially, around 100 humanoid and service robots will learn specific skills and processes in 48 training cells. The large volumes of training data will help make robots safer and more reliable.

In addition to its robot training cells, the center at the LabCampus innovation hub at Munich Airport provides substantial computing power for AI model training and computer simulations. These resources generate large amounts of data to support the development of more intelligent robots. The current focus is on industrial applications and intralogistics.

Unlike generative AI systems such as ChatGPT, which can draw on vast amounts of data from the internet, embodied AI requires training data from real-world physical interactions. Such data is currently scarce. The new training center features robot training cells where robots learn tasks including folding boxes, assembling components and using tools. At the TUM RoboGym, robotic skills can be developed and tested in realistic environments. At the opening, NEURA Robotics also unveils the next generation of its 4NE1 humanoid robot, which will subsequently be available at the TUM RoboGym.

Data generated at the TUM RoboGym flows into a curated cloud data lake, where it is available for training physical AI models. As a central scientific resource, the data lake enables scalable research into generalizable robotic skills, safe human-robot interaction and AI models for robotics.

The project’s core objective is to transfer scientific results into practical applications. “Close, continuous and flexible collaboration with the robotics industry is essential for research in physical AI. It is the only way to systematically and rapidly develop research prototypes into market-ready products,” says Prof. Achim Lilienthal of TUM MIRMI, scientific coordinator of the TUM RoboGym. 

www.tum.de