RoboGym opens in Munich
Germany is building RoboGym — a €17 million ‘data farm’ for humanoid robotics at Munich Airport backed by TU Munich and Neurorobotics to close the real‑world training gap. The facility is pitched as the largest robotics learning ground, because simulations alone aren’t supplying enough real data to commercialize dexterous robots (youtube.com).
The training hall will span roughly 2,300 square metres (about 25,000 sq ft) inside TUM’s Convergence Centre, giving researchers room for full-size indoor environments and multi-robot scenarios. (tum.de) NEURA Robotics is supplying the majority of the initial capital with an approximate €11 million contribution, while TUM and other partners supply the remainder of the launch funding. (neura-robotics.com) A large fleet of humanoids is slated to begin regular on-site training from mid‑2026, with the facility described as the staging ground for continuous real-world data collection rather than one-off experiments. (neura-robotics.com) Specific NEURA platforms named for use in the centre include the 4NE‑1 humanoid and the MiPA cognitive assistant, and much of the sensor and interaction data will feed into NEURA’s Neuraverse training ecosystem. (neura-robotics.com) TUM’s Munich Institute of Robotics and Machine Intelligence will run the scientific programme under professors Lorenzo Masia and Achim Lilienthal, with Masia saying the centre is intended as a “counterweight” to US and Chinese competitors in embodied AI development. (tum.de) Project briefs and partner statements indicate the centre expects to train “hundreds” of robot instances using human-in-the-loop instruction and staged everyday tasks to generate the high‑quality, messy real‑world datasets that current simulators lack. (roboticsandautomationnews.com)
