Zhishen Technology Moves Embodied-Intelligence Robots Into Worksites

What changed

At IROS 2026 in Pittsburgh, Singapore’s Nanyang Technological University won the quadruped-robot challenge with China-made Zhishen Tongchui M1, operating fully autonomously. The event ran from September 27 to October 1 and showcased Chinese robotic hands, robot platforms and self-developed joint modules.

Zhishen Technology says it had mass-produced more than 15,000 embodied-intelligence robots by June 2026. Its machines are being used in electricity, petrochemical, firefighting, security and emergency-response work.

Why it matters

The striking part is not simply that a robot won a competition. It is that the winning machine had to keep sensing, planning, navigating and operating a mechanical arm through repeated, demanding tests. A robot that performs once is a demonstration. One that survives an event cycle is closer to equipment.

That distinction is shaping China’s push from manufacturing robots at scale to placing them in working environments. The L2 quadruped platform is reported to offer about 5.5 hours of endurance, centimetre-level positioning and operation from minus 20°C to 55°C. Those are practical specifications for machines expected to move through industrial sites, not just impress on a stage.

China’s broader lead is also visible in supply. Chinese humanoid robots accounted for more than 97% of global shipments in the first half of 2026, while Zhishen says its own embodied-intelligence production has passed 15,000 units. The reported shipment share has not been independently audited, so its meaning depends partly on how the figures were counted.

The bigger test now is repeatability. Zhishen is extending experience from quadrupeds into humanoid robots, using the cycle of deployment, data collection, model updates and product improvements. If that cycle works across more sites, the advantage will belong less to the robot with the flashiest demonstration than to the maker that can keep machines useful after the cameras leave.

A newly opened programme is intended to accelerate that process by offering 1,000 physical robots, development tools, technical support and project opportunities to developers, researchers, universities and companies. The next milestone is therefore mundane but decisive: whether more people can make these machines complete real tasks reliably.