The robot proved its effectiveness by independently assembling car components without the help of an operator.
Xiaomi’s humanoid robot began work in a real automobile production, taking part in the complex assembly of components. As part of tests in the Xiaomi EV injection molding shop, the robot successfully worked autonomously for three hours in a row at the self-tapping nut installation site. This is reported by CNEV Post, Dengi.ua reports. The PromPolitInform portal informs.
Tests confirmed that the humanoid robot is able to work in the rhythm of a modern automobile factory. It achieved a success rate of 90.2% when simultaneously installing parts on both sides of the workstation. At the same time, the robot fully adhered to the tight production cycle: it takes only 76 seconds to complete the task.
The complexity of the operation lies in the fact that the robot must grab self-tapping nuts from an automatic feeder and place them on the fixtures with extreme precision. The process is complicated by the spline structure of the nuts, interference from magnetic forces, and the need to coordinate actions with a moving conveyor belt to fix the cast components of the car floor.
To control the “iron worker,” Xiaomi used its own VLA (Vision-Language-Action) model called Xiaomi-Robotics-0. The system has 4.7 billion parameters and uses an end-to-end data-driven control method.
Key technological features include:
Multimodality: The robot navigates not only through vision, but also through tactile feedback and sensors in the joints.
Hybrid architecture: The optimization controller makes decisions in less than 1 millisecond, providing instant response.
Virtual learning: The AI went through hundreds of millions of simulations in a digital environment, which allowed the robot to maintain balance even under extreme external obstacles.
Thanks to reinforcement learning, the robot quickly adapts to real-world conditions without the need for manual remote control. This makes it a fully autonomous link in the Xiaomi EV production chain.
As we wrote earlier, researchers from the US Agricultural Research Service (ARS) in collaboration with Michigan State University presented a new model of a harvesting robot equipped with two “hands”.
We have also already written that the hasty introduction of AI and the mass replacement of workers with algorithms can deprive businesses of competitive advantages, because such a model can be easily copied by other market players.
