Germany Builds "Robot Gym", a Facility Where Humanoids Are Trained for Everyday Tasks
A robot gym centre is set to open soon. This facility will serve as a place where humans train robots to perform everyday tasks, rather than a fitness centre as gyms are typically known.
The facility, named TUM RoboGym, is being built by Germany’s Technical University of Munich (TUM) in partnership with Neura Robotics, a robotics company based in Metzingen.
TUM RoboGym spans an area of 2,300 square metres and is located at the TUM Convergence Centre near Munich Airport.
The robot training facility, which includes humanoid robots, is being created with an investment of $19.8 million (approximately Rp 333 billion) from partners. Of this, $12.8 million (approximately Rp 215 billion) comes from Neura Robotics.
“Interaction between advanced robotics technology and cutting-edge academic research in AI will provide a major boost to the development of this technology,” said Achim Lilienthal, who is also leading the facility.
Lilienthal will be joined by Lorenzo Masia PhD, a professor of intelligent biorobotic systems and executive director of the Munich Institute of Robotics and Machine Intelligence (MIRMI).
The duo aims to accelerate progress in humanoid robots and the concept of embodied artificial intelligence, which involves AI integrated with a robot’s body that interacts directly with the physical environment.
In the future, Lilienthal and his team will develop AI-supported methods at TUM RoboGym to enable robots to acquire general skills. The goal is for robots to be adopted in the real world for specific scenarios.
Some of these skills include folding boxes, assembling components, and manipulating objects. For now, these skills remain a significant challenge for many robot systems.
Neura Robotics founder and CEO David Reger stated that the key factor in modern robotics competition is no longer mechanics, but data.
“The determining competitive factor in intelligent robotics is no longer mechanics, but data. Those who have high-quality, realistic training data will lead the development,” said Reger.
Therefore, through TUM RoboGym, Lilienthal hopes the facility will address the shortage of high-quality training data, which is one of the biggest challenges in developing embodied AI.
This is because robots require physical data from the real world to understand movements and interactions with objects. Such data will be collected through direct training by humans at RoboGym.
“For that reason, we need a training centre where humans can teach robots individual skills, such as folding boxes or assembling components,” said Lilienthal.
TUM RoboGym is expected to house hundreds of robots, as reported by KompasTekno from Interesting Engineering. In addition to research, RoboGym will also serve as a training ground for students and engineers.