Xiaomi made the transition from making smartphones to becoming a serious player in electric vehicles after several years of progress. The company established its automotive operations in 2021 when it started making electric vehicles at its Beijing factory which features advanced automated manufacturing technology. The facility today stands as one of China's most advanced production centers for electric vehicles which operate in its expanding electric vehicle manufacturing field.
The factory design contains automation features which extend through all of its construction elements. The plant operates with approximately 600 to 700 industrial robots which perform welding and assembly and quality inspection work according to Xiaomi founder and CEO Lei Jun. The facility achieves maximum production automation through its extensive robot network which enables it to produce a high output of vehicles.
The latest Xiaomi experiment includes new technology which moves beyond traditional factory robotic systems. The company now uses humanoid robots which can perform human work instead of depending on fixed industrial machines.
Humanoid Robots Enter the Production Line
Xiaomi introduced its humanoid robots which it developed in-house to begin testing operations at its electric vehicle manufacturing facility after the company made its announcement in early 2026. The robots operated under production conditions instead of executing their functions in a lab space which tested vehicle components used for assembly.
The humanoid robot operated at a testing station which dedicated its work to installing self-tapping nuts onto vehicle structures produced in the die-casting workshop. The task requires both systems to work together for precise delivery of parts through automated feeders and at the same time they need to handle both positioning and tightening operations. The work involves repeating a series of technical tasks which require human operators to maintain both alignment and proper torque settings.
The robot executed the assembly task for three hours without needing any human oversight. The machine reached a success rate of 90.2 percent according to test results while it maintained the factory's production speed.
The robot achieved the operational standard which required 76 seconds for each task according to the production line specifications. In modern automotive manufacturing, this production tempo is critical because every workstation must complete its tasks before the next vehicle component arrives.
The Technical Challenges Behind the Task
The task of installing self-tapping nuts appears simple, yet it presents significant challenges for robots who need to perform this task. Each nut needs to be precisely handled by the robot who will extract it from the automatic feeder to match it with the designated mounting point while securing it without making cross-threads or moving it from position. Assembly work suffers from failures when workers make slight errors in either angle or position.
The internal spline structure of the nuts and all possible gripping angles and magnetic interference from nearby metal components between two gripping points make the task extremely hard for engineers to perform according to Xiaomi engineering team observations. The operation works as an ideal benchmark which allows for checking how accurate humanoid robots perform their work and how well they adjust to different industrial environments.
Xiaomi developed control systems which function through various sensor data types to handle the challenges that face their organization. The robots use multiple sensors which enable them to receive visual information and tactile feedback and track the angles of their joints within their body.
AI at the Core of the Robot’s Capabilities
Xiaomi-Robotics-0 artificial intelligence system functions as the core element which drives the robot's operational performance. The system operates using a Vision-Language-Action (VLA) model which contains approximately 4.7 billion parameters.
The robot uses VLA architecture to analyze visual data and comprehend assignment details which it now uses to execute movement actions in the physical space. The robot uses reinforcement learning techniques which enable it to enhance its abilities through training sessions and direct experience with real-world situations.
The robot develops better task performance skills through practical execution instead of following pre-set commands. The research field today views this technique as essential to attaining “embodied AI” which enables machines to comprehend physical spaces while they engage in complex operational activities.
