World's largest robotics training center planned

Under the leadership of TUM MIRMI professors Lorenzo Masia and Achim Lilienthal, hundreds of robotic systems, including several humanoids, will be further developed and trained with AI support at the research and training center for their use in everyday life. The partners are jointly investing 17 million euros in the TUM RoboGym. Neura Robotics is contributing the lion's share of eleven million euros, primarily to procure robots and ensure the maintenance of the hardware. "In return, Neura Robotics participates in our research," says Prof. Achim Lilienthal. "The combination of high-end robotics technology and top-class academic research in artificial intelligence will give development a huge boost," says the scientific coordinator of the new robotics hub called 'TUM RoboGym (powered by Neura)' and initiator of the new TUM MIRMI cooperation together with Lorenzo Masia. To this end, the partners TUM MIRMI and Neura Robotics have signed a corresponding cooperation agreement.

Humanoid robots are becoming part of everyday life

TUM President Prof. Thomas F. Hofmann emphasizes: "Humanoid robots have long since left the status of science fiction. In the near future, they will be an integral part of our everyday lives and support people in many tasks. Together with Neura Robotics, we at TUM are working on accelerating this development and, above all, ensuring the safety of coexistence between humans and robots in addition to functionality."

The founder and CEO of Neura Robotics, David Reger, says: "The decisive competitive factor in intelligent robotics is no longer mechanics, it lies in the data. Those who have high-quality, realistic training data set the pace. This is precisely where we at Neura Robotics bring our strengths to bear: By setting up robot gyms worldwide and networking training data via our Neuraverse platform, we are creating scalable training infrastructures for physical AI. Together with TUM, we combine excellent research with entrepreneurial implementation power. In this way, we are setting new standards in intelligent robotics and sustainably strengthening Germany's and Europe's technological leadership in one of the key future technologies of our time."

For the Director of TUM RoboGym and TUM MIRMI Executive Director Prof. Lorenzo Masia, the cooperation offers the opportunity to help shape future developments in robotics worldwide: "European sovereignty is extremely important in times of geopolitical competition between East and West. With this research and training center, which is one of the largest in the world, we are creating a unique infrastructure in Europe for our researchers and students, where they can experience, develop and learn new approaches in robotics and AI and thus become a strong core of European experts when they enter the job market."

Gaining valuable data from human movements

In the TUM Convergence Center at Munich Airport, the cooperation partners are jointly researching the next generation of humanoid robots on an area of 2,300 square meters. Among other things, AI-supported training of robotic systems is planned. TUM MIRMI and Neura Robotics will make most of the data generated from this available to the robotics community in an open ecosystem.

Unlike language models such as ChatGPT, it is not possible in embodied AI to access almost any amount of data from the internet and learn with its help. Web videos in which robotic arms are used to perform certain tasks, for example, are hard to find. It is also still unclear how well movements can be read from videos. Simulations are also usually far too imprecise and cannot reliably reproduce frictional forces that exist in the real world.

"These are the reasons why we need a training center where people can teach the robots individual skills, such as folding a box or putting components together," says Prof. Lilienthal. "The robots will learn general skills in the RoboGym, which they can then transfer independently to specific tasks." The prerequisite for this is a large number of data sets, which are now being created in RoboGym by human robot trainers.