Humanoids among people: the safety impact of moving general-purpose robots into shared human spaces
Humanoid robots are leaving controlled environments and entering warehouses, hospitals, and homes, where they share space with people who never trained to work beside a machine. This work characterizes the hazard classes a deployed humanoid introduces and asks whether today's robot safety standards were built for a body like this.
Most robot safety frameworks assume a machine that is bolted down, fenced off, or both. The humanoid breaks that assumption on purpose. It is untethered, general-purpose, and meant to operate in the same unstructured space as people, which is precisely what makes it useful and precisely what makes it hard to govern. As pilots move from factory floors into warehouses, hospitals, and eventually homes, the safety question shifts from "can it do the task" to "what happens to the people around it when it does the task wrong."
The hazards that matter are not the ones a fixed industrial arm presents. A deployed humanoid introduces distinct classes of risk:
- Uncaged operation: no physical barrier separates the machine from people, removing the primary control most industrial robot safety depends on
- Falls and toppling: a tall, top-heavy body that loses balance becomes a heavy, high-energy object near people, a failure mode bolted-down arms do not have
- General-purpose, learning-driven behavior: one body runs many tasks under learned policies, so its actions are harder to enumerate, test, and bound than a fixed automation routine
- Proximity to vulnerable people: homes, hospitals, and elder care place humanoids beside children, patients, and the elderly rather than trained workers
- Unstructured, shared environments: cluttered and changing human spaces defeat the assumption of a known, controlled workcell
This work surveys the real and anticipated deployment contexts for humanoids, characterizes these hazard classes, and tests them against the standards a deployer would actually reach for, including the industrial and collaborative robot standards (ISO 10218, ISO/TS 15066) and the personal-care robot standard (ISO 13482). The central question is coverage: whether frameworks written for fixed arms, fenced cells, and narrow collaborative tasks can carry the weight of a mobile, general-purpose, bipedal machine operating among ordinary people, and where the gaps leave deployers without a defensible safety case.
What you're reading is the abstract.