Bridging Human Dexterity and Physical AI: MANUS Showcases Teleoperation at TUM RoboGym

During the grand opening of the TUM RoboGym—a new robotics facility powered by NEURA Robotics—wearable motion-capture technology took center stage. Live demonstrations illustrated how precision hand tracking transforms natural human movement into actionable training datasets for Physical AI and autonomous robotic manipulation.

Inside the TUM RoboGym Facility

Operator wearing MANUS Metagloves Pro for humanoid robot teleoperation of NEURA 4NE1 Gen 4 at TUM RoboGym

Unveiled on October 22, 2026, at the LabCampus near Munich Airport, the 2,300-square-meter TUM RoboGym serves as a collaborative hub linking academic robotics research with real-world industrial applications. Rather than relying solely on isolated lab tests, the facility blends four core pillars of robot development:

  • Live human demonstrations
  • Real-time teleoperation
  • Simulated environments
  • Structured robot telemetry

By inviting industry partners to supply authentic operational workflows, researchers at the RoboGym can systematically study how machines acquire practical skills and execute complex tasks with higher reliability.

Real-Time Control with MANUS Metagloves Pro

A highlight of the launch event featured a human operator controlling NEURA Robotics’ 4NE1 Gen 4 humanoid robot in real time using MANUS Metagloves Pro. By translating subtle finger and hand articulation directly into robotic commands, the setup demonstrated seamless, low-latency human-in-the-loop control.

Engineered with electromagnetic field (EMF) tracking, Metagloves Pro map a complete 25 degrees of freedom (DoF) anatomical hand model down to millimeter accuracy. Because EMF technology operates without relying on line-of-sight cameras, the system avoids optical occlusion and positional drift during intricate hand movements. Using MANUS Core and the MANUS SDK, operators can pipe this kinematic data straight into active teleoperation loops or machine-learning architectures.

Scaling from Teleoperation to Imitation Learning

Beyond live remote operation, capturing accurate hand kinematics solves one of Physical AI’s biggest bottlenecks: gathering high-quality training data for dexterous manipulation.

When finger trajectories from Metagloves Pro are logged simultaneously alongside a robot’s internal state, environmental perception, and task success metrics, each teleoperation session doubles as a rich training dataset. Robotics teams can feed these synchronized recordings directly into imitation learning models and policy fine-tuning workflows—allowing humanoid robots like the 4NE1 Gen 4 to eventually perform delicate tasks autonomously.