Xiaomi Humanoid Robot Enters EV Factory for Autonomous Assembly Operations

Xiaomi has advanced embodied intelligence by deploying the Xiaomi humanoid robot into a real EV factory environment for complex component assembly. The milestone marks the first time the Xiaomi humanoid robot has participated directly in automotive manufacturing operations within the company’s die-casting workshop. This move reflects the company’s broader strategy to integrate robotics and artificial intelligence into high-precision industrial settings while meeting strict production cycle requirements.

Xiaomi humanoid robot begins autonomous EV factory operations

The Xiaomi humanoid robot officially entered the EV factory’s die-casting workshop to perform self-tapping nut installation tasks. According to the company, the robot successfully completed three consecutive hours of autonomous operation at the workstation, demonstrating stable and repeatable performance under real production conditions.

Assembly accuracy and production cycle targets

Test results showed a 90.2 percent success rate for simultaneous nut installation on both sides of the workstation. The Xiaomi humanoid robot also met the production line’s fastest cycle time requirement of 76 seconds. The task requires precise picking of self-tapping nuts from an automatic feeding device, accurate placement onto positioning fixtures, and coordinated tightening of floor components following integrated die-casting processes.

Overcoming precision challenges in automotive manufacturing

Achieving reliable alignment and engagement of self-tapping nuts presented significant engineering complexity. The spline structure inside the nuts, non-fixed gripping posture, and magnetic interference increased the difficulty of maintaining consistent precision. These constraints required the Xiaomi humanoid robot to adapt dynamically to variations in positioning and force application during assembly.

End-to-end data-driven control architecture

To address these challenges, (Xiaomi) implemented an end-to-end data-driven control system powered by its self-developed 4.7-billion-parameter Vision-Language-Action model, Xiaomi-Robotics-0. Combined with reinforcement learning, this framework reduces reliance on teleoperation data and enables rapid adaptation across diverse operating conditions in the EV factory environment.

Multimodal sensing and real-time motion control

The Xiaomi humanoid robot integrates multimodal inputs, including vision, tactile sensing, and joint proprioception. This reduces state misjudgment in complex assembly scenarios and enhances operational robustness. For full-body motion control, the system combines optimization-based control with reinforcement learning to maintain stability during dynamic movements.

Simulation training and zero-shot transfer

The reinforcement learning controller underwent training through hundreds of millions of simulated random perturbations in virtual environments. This approach enables the Xiaomi humanoid robot to maintain balance under extreme disturbance conditions and achieve zero-shot transfer from simulation to real-world deployment in automotive manufacturing settings.

Scaling humanoid robotics across EV production lines

The self-tapping nut station represents the initial phase of broader industrial deployment. (Xiaomi) is conducting validation across additional workstations, including bin-picking and front badge installation tasks. The company aims to overcome production cycle time and yield rate bottlenecks to enable large-scale deployment of the Xiaomi humanoid robot across its EV factory operations.

Leadership projections indicate that large numbers of humanoid robots could begin working in company factories within the next five years. Meanwhile, global competitors are accelerating robotics investments, underscoring the intensifying race in embodied intelligence and advanced automotive manufacturing technologies. The deployment of the Xiaomi humanoid robot signals a strategic step toward intelligent, automated EV production at scale.