SMRTR Science & EngineeringOct 7, 2025Interesting Engineering

China develops wearable exoskeleton to train humanoid robots with high accuracy

SMRTR summary

A humanoid robot pushed off course doesn't just stumble away anymore—it walks right back to finish its job. That's the breakthrough from China's National University of Defense Technology, which has created a wearable suit system called HumanoidExo that teaches robots complex tasks by capturing real human movement.

The key innovation lies in solving a fundamental problem: robots trained on videos and simulations often lose their balance because they lack real-world motion data. The new system uses a suit with sensors that track seven arm joints, plus motion detectors on wrists and a back-mounted LiDAR scanner to capture how humans naturally move and maintain balance.

When researchers tested this on a Unitree G1 robot, the results were striking. The robot's success rate on pick-and-place tasks jumped from just 5% to nearly 80% after learning from the exoskeleton data. Even more impressive—the robot learned to walk simply by watching someone walk to a table, despite receiving no specific walking instructions.

This development comes as companies like Tesla, Figure AI, and Google DeepMind race to perfect humanoid robotics for real-world applications.

SMRTR provides this summary for quick context. The original article belongs to Interesting Engineering.

Read the original article
SMRTR Science & Engineering

Get the next batch of curated summaries in your inbox.

This archive is built from SMRTR newsletter summaries. Subscribe for hand-picked stories without the extra noise.