Factory Humanoids

Buying advice

Separate research platforms from production candidates

On a check in October 2026, the entry point was the Unitree H2 at close to $29,900, and the Figure 02 sat highest at close to $130,000, and three of the six had no published price. Purpose varies as much as price. ARMAR-6, for example, comes from a research group at the Karlsruhe Institute of Technology, whose paper describes it as a collaborative robot for industrial maintenance work, built to a payload target of 10 kg per arm. A platform like that can suit a university, a training centre or an automation lab. If you need a robot on a production shift, ask each maker what service cover, spare parts and software support come with it.

Plan for a cell until standards catch up

An A3 article on humanoid safety notes that pilot programmes so far run in walled gardens, where robots and people share a space but not at the same time, and that current standards do not yet cover hazards specific to humanoids. Plan your first station on that basis. Ask your integrator how it will apply ISO 10218-2:2025, which sets safety requirements for industrial robot applications and robot cells, to the station where the humanoid will work. Agree how people enter for replenishment or a jam, what stops the robot and how it is restarted. Budget for guarding or scanners at the start, and treat their removal as a later decision backed by evidence.

Settle programme ownership and acceptance terms

Data recorded on your line can be valuable to a maker, so agree the terms before the robot arrives. Ask who will own the task programmes and any video or sensor data captured at your station, whether that data may leave your site, and whether you could move the programmes to a second robot or to another maker. Then set an acceptance test on your own floor, with a run long enough to cover a shift change and a material replenishment, and define in advance what counts as an intervention. Build in the faults operators clear by hand today, such as parts arriving in the wrong orientation, so the test reflects how the station really runs.

What to look for

• Hand dexterity and grip force for the parts at your station
• Cycle time measured against your takt time, not a demonstration video
• Integration with your MES, PLCs and safety system, and how new tasks are taught

Common questions

Can a humanoid replace a fixed robot cell?

Rarely on speed. Fixed cells remain quicker where volumes are high and stable. Humanoids make sense where tasks vary, the station was designed for a person, or rebuilding the line would cost more than the robot.

Are humanoids working in factories yet?

In pilots and early deployments, yes, with a number of makers testing them on production tasks. Ask for evidence from comparable work when you request a quote, and describe your station in detail.

Sources and further reading

• ARMAR-6: A Collaborative Humanoid Robot for Industrial Environments (IEEE-RAS Humanoids 2018) — Karlsruhe Institute of Technology (H2T).
  Research paper on a humanoid built to help technicians with industrial maintenance, with design targets and test results.
• Safety by Design: How Humanoid Robots Must Evolve to Depart the Walled Garden — Association for Advancing Automation (A3).
  Looks at why humanoid pilots stay segregated from people and which safety advances shared work would need.
• ISO 10218-2:2025 Robotics: safety requirements, Part 2: Industrial robot applications and robot cells — International Organization for Standardization (ISO).
  Safety requirements for integrating robots into applications and cells, from design and commissioning to decommissioning.