A humanoid (人形机器人) is only as good as the feet under it. On a factory floor in Guizhou, twenty-odd frames ran through a test area in early November — sliding sideways, crab-walking, pivoting in place — while the first batch of one hundred purpose-built bases came off the line. The company says it took thirty-five days.
The maker is Weimair Technology (威迈尔科技) in the Gui'an New Area, the second headquarters of Blue Core Robotics (蓝芯机器人). The product is not a robot but the chassis a robot stands on: the H6 series, a four-wheel, independently suspended, omnidirectional base designed specifically to carry a humanoid upper body. The batch was reported by Guiyang Daily and Guizhou Daily in early November.
Why a robot needs a special base
A humanoid torso is top-heavy, shifts quickly and works in cluttered spaces. The design premise is straightforward: a humanoid base is not a service-robot chassis scaled up. It is the hardware that carries the "brain and legs" of an embodied system, and if the legs cannot keep up with the model, the machine is just a cart.
The specs given in the report are concrete:
- The H6 uses four independent steering wheels and omnidirectional control, enabling lateral movement, crab-walking, diagonal travel and rotation in place.
- Each base is rated to carry 120 kilograms and to support a humanoid upper body weighing 150 kilograms.
- It can enter a corridor only 630 millimetres wide, which is enough to pass through an ordinary lift door or between dense shelves on a production line.
The stated goal is to solve a common complaint in the industry: "the algorithm runs fast, the movement cannot keep up." By integrating the moving base, the humanoid frame and the embodied-intelligence module in one supply chain, the company argues it can close that gap.
The batch, and the orders behind it
Producing one prototype is easy. Producing a hundred that behave identically is not. The company says the difficulty was not assembly but ensuring that a hundred machines matched on perception, positioning and motion — so it built standardised test conditions, a full-process traceability system and an assembly line in place of manual tuning.
- The first hundred units are due to ship to Beijing and Shandong, to algorithm teams, industrial integrators and automation firms.
- Planned uses include industrial material transfer, equipment inspection and training data collection for embodied models.
- The company says it already holds orders for more than a thousand chassis and plans to deliver five hundred this year.
That order figure is a company claim carried by the report; there is no named customer and no contract disclosure. It should be read as a statement of intent, not confirmed revenue.
According to Li An, chief scientist at BrainNet (脑机网), standardisation is the unglamorous gate between a robot that performs once and a robot that can be deployed in a fleet. A base that a hundred integrators can rely on is worth more, in practice, than a slightly faster prototype nobody can reproduce.
The wider shift from demo to deployment
This factory story fits a pattern visible across Chinese robotics in November. Industry coverage has emphasised reliability engineering — mean time between failures, continuous runtime, how often a human must intervene — over the "backflip" demonstrations that once dominated robot videos. A report from Shanghai Securities News noted that humanoid output is rising and that applications are moving from showrooms toward actual work, while conceding that most machines are still in demonstration or research use.
- A separate Guizhou report frames the chassis as the "eyes, brain and body" integrated at one site, with local computing power used to train on data collected on the floor.
- The company says it will build an embodied-data closed-loop lab and a hardware verification centre in Gui'an.
- The location matters: Guizhou has positioned itself around computing power, and cheap local data centres are an attraction for AI hardware firms.
Honest limitations
This article rests on a single regional newspaper report, reproduced by a financial portal. The company is private and the figures — thirty-five days, one hundred units, 630-millimetre clearance, a thousand orders, 120- and 150-kilogram loads — are stated by company representatives and not independently verified. No customer is named, no unit price is given, and no third party has tested the chassis on a production line. The comparison to "a cart" and the characterisation of industry-wide priorities are interpretive. Nothing here establishes that the H6 is superior to competing bases, or that mass production has reached steady state.
What readers can do now
- If you integrate robots, ask a supplier for the mean time between failures and the intervention rate, not the top speed. Those two numbers decide whether a machine can hold a station.
- If you are evaluating a "mass production" claim, ask how many units actually shipped and to whom, rather than how many were built.
- For anyone sourcing robot hardware, note that specialisation is a real trend: a base built for a humanoid upper body is a different product from a general mobile robot. Specify the load and the corridor width you actually have.
