China trains thirty humanoid robots in physical data factory

In Fujian, a new facility is dedicated to collecting physical data to train nearly 30 humanoid robots. Here, robots do not learn from digital information, but by observing and repeating real tasks. Operators wearing virtual reality headsets guide each movement, from wiping tables to sorting fruit, while cameras and sensors record every detail.

industrial robotics training facility, thirty humanoid robots positioned in rows performing repetitive physical tasks, operators wearing VR headsets controlling robot arms via motion capture, robot hand wiping a table surface while fruit sorting station shows robotic grippers handling apples, overhead camera rigs and LiDAR sensors recording movement data, concrete factory floor with cable management systems, holographic data streams displaying joint angles and force feedback, cool blue LED lighting mixed with warm task lights, photorealistic engineering visualization, wide-angle lens capturing depth and scale, mechanical joints with visible actuators and wiring, subtle motion blur on moving robot limbs, dust particles in shaft of light, technical documentation screens on distant wall

How Human Observation Teaches Machines 🤖

The process is meticulous: a human operator, equipped with a virtual reality device, controls the robot in real time. Each joint rotation and gripper pressure is captured by high-precision sensors and cameras. This data, accumulated through repetition, forms a bank of physical information that allows the machine to learn by imitation. It is not about programming, but about showing and recording.

The Dream Job: Being the Boss of a Robot 🎮

Imagine having a robot that wipes tables, sorts fruit, and moves boxes, all while you put on a pair of virtual reality glasses and move your arms. It's like playing video games, but with the responsibility that if you make a mistake, the robot ends up putting an apple in the trash and a box in the fridge. Luckily, they don't complain about overtime.