- The heavy-duty embodied AI robot is designed to bring ‘AI workers’ into real industrial settings
- With 50kg dual-arm payload and 8-hour endurance, MOS2 targets demanding factory tasks
Lumos Robotics (鹿明机器人), a Suzhou-based embodied AI startup, launched MOS 2, a heavy-duty wheeled-arm embodied AI robot designed for industrial applications on August 14.
The robot can handle a combined payload of up to 50kg with its two arms and is built for intensive, continuous work in real factories.
Standing about 1.65 meters tall and occupying just 0.5 square meters, MOS 2 requires a theoretical passage width of about 0.8 meters, allowing it to operate in relatively constrained industrial environments.
The robot has 22 degrees of freedom, including a newly added two-degree-of-freedom head and an omni-directional mobile base.

Its dual-arm system uses standard flanges and integrated interfaces, allowing operators to quickly switch end-of-arm tools for different tasks.
Its perception system combines dual multi-line LiDARs on the base, two wrist cameras, dual six-axis force sensors and stereo cameras on the head. The system is designed to combine environmental perception, visual recognition and force feedback.
Adjusting to production-line changeovers
MOS 2 is powered by an Nvidia Orin AGX platform delivering 275 TOPS of computing power, enabling on-device model deployment and local decision-making.
Its wiring has passed five million durability cycles, while a dual-battery hot-swapping system provides more than eight hours of operation.
The robot is aimed at industrial tasks that remain difficult for conventional automation, including frequent production-line changeovers, complex assembly and low-volume jobs where dedicated automation equipment is difficult to justify economically.

Lumos positions MOS 2 as a heavy-duty “AI Worker” combining a robust body capable of around-the-clock operation with environmental perception, task understanding and autonomous manipulation.
Why it matters globally
The significance is broader than its specifications. As embodied AI moves from demonstrations to industrial deployment, the ability to perform reliably on real production lines is becoming a more important competitive test than flashy demonstrations.
According to ITjuzi.com, a market intelligence provider tracking financing activities in China, the country’s embodied AI sector attracted more than 93.5 billion yuan in financing in the first half of 2026.
Capital is increasingly flowing toward robots capable of performing actual industrial work, the firm says.
For the global robotics industry, MOS 2 reflects a shift from “Can robots move?” to “Can robots work?” The companies that can deliver reliable, flexible automation on real factory floors may ultimately determine how quickly embodied AI becomes commercially viable.


