Introduction
Humanoid safety is becoming a separate engineering discipline because a human shaped machine combines mobility reach high joint torque batteries cameras networked software and frequent contact with people. A safe deployment therefore cannot rely on one emergency stop or one collision limit. The robot body the task the workspace and the software update process all affect the final risk.
Key facts
- ISO 10218-1 and ISO 10218-2 were revised in 2025 for industrial robot safety.
- ISO is developing a new edition of ISO 13482 for service robots.
- China has active national standard projects for general humanoid safety and industrial humanoid safety.
Why humanoid safety needs its own assessment
A fixed robot arm usually works inside a defined cell. A humanoid may walk through doors climb steps carry objects use tools and work near people. The hazard map changes with each task. Falling stability stored mechanical energy pinch points battery faults perception errors and unexpected contact must be considered together.
The useful unit of analysis is the complete application. A robot that is acceptable while carrying a light tote can present a different risk when it holds a sharp tool or works beside an open machine. Safety limits should follow the real payload speed reach floor condition and human access pattern.
How existing standards fit together
Industrial humanoids can inherit many principles from ISO 10218. Service deployments need the human contact focus found in ISO 13482 and related test guidance. Neither document should be treated as a universal certificate for every humanoid use case. Scope matters because factories homes hospitals and public spaces expose different people to different hazards.
What China is standardizing now
China has moved from broad robot rules toward dedicated humanoid projects. The 2026 program includes general safety and industrial safety with additional work for domestic public service and special application settings. This matters because it creates a framework for hazards that are specific to walking human shaped machines rather than stationary automation.
A practical safety case for deployment
A serious deployment file should document intended tasks foreseeable misuse speed and force limits stop functions safe states battery behavior fall zones remote control permissions software versions maintenance procedures and incident reporting. Validation should include normal operation and faults such as lost perception stale commands network interruption low battery and a failed actuator.
What buyers should ask vendors to prove
Ask for the applicable standards and the exact configuration that was tested. Request evidence for emergency stopping safe motion collision behavior battery protection cybersecurity update control and recovery after faults. A video of a robot walking beside people does not show that these functions were validated across the required operating envelope.
Limitations and missing information
- Several humanoid specific standards remain drafts in 2026.
- A safe robot can become unsafe after a tool change software update or workspace change.
- Certification scope must be checked against the exact deployment.
Conclusion
Safety will become one of the main differences between a research humanoid and a machine that can work every day around employees or customers. The strongest programs will treat safety evidence as a living engineering record rather than a final box checked before launch.
Sources and methodology
This guide separates published standards and official technical documents from engineering practice. Draft standards are described as work in progress. Product capability is not treated as verified unless a source supports it.
Related TechniaHQRobot guides
Share this article
Share the current TechniaHQRobot article page.
Continue reading
Open the latest robotics reporting, Physical AI analysis and hardware notes.