Safety Certification Still Limits Industrial Humanoids

Single-source brief: A new arXiv paper maps a key safety gap but does not claim certified deployment.

What Changed

A paper posted on arXiv studies safety certification for industrial humanoid robots.

Its central claim is direct. Safety certification, not movement skill, may be the larger barrier.

The authors call this problem the “fail-passive gap.”

Most industrial safety systems assume power removal creates a safe state. That works for many fixed machines.

It does not work the same way for a walking humanoid. Removing power can cause an uncontrolled fall.

What the Feasibility Study Found

The study used a certified external safety chain. It included a light curtain, emergency stop, fail-safe input, PLC, and wireless PROFIsafe.

The authors say these external parts can be measured with established safety methods.

The remaining issue sits inside the robot-side reaction chain.

They tested the approach in a Unitree G1 EDU pick-and-place cell. The workspace measured 3 meters by 1.5 meters.

The paper also examined stop limits, fall risk, separation distance, and timing.

Deployment Reality

This is a feasibility study, not a certified industrial system.

The authors explicitly do not claim end-to-end PL e or SIL 3 certification.

They placed an industrial software-defined automation controller on the robot. That moved safety-message receipt onto a standardized IEC 61131-3 interface.

Yet the G1 onboard computer is not safety-rated hardware, according to the paper.

The unresolved boundary is the link between that controller and the robot’s balancing policy.