In the vast battery-producing facility located in Luoyang, a new category of laborer has appeared alongside the assembly line one that never tires, never takes a rest, and is capable of performing tasks that involve hundreds of volts with absolute accuracy. The world’s largest manufacturer of EV batteries, CATL, has initiated the world’s first mass-production line for batteries that utilizes humanoid robots to replace human labor for a number of the most dangerous tasks.
The humanoid robot, also called Xiaomo or “Moz,” depending on the translation, was developed by Spirit AI. It is a robotics startup in Hangzhou that was founded in the year 2024 with initial investments from CATL. Spirit AI has managed to create the robot in less than a year that can perform various complicated tasks in the industry that previously required the dexterity of a human. CATL terms the achievement as “milestone breakthrough in applying embodied intelligence to manufacturing.”
The core task of Xiaomo is related to the test processes of the End of Line (EOL) and Direct Current Internal Resistance (DCR). These were traditional tasks that involved humans connecting high voltage test plugs, which may contain as much as hundreds of volts, into specific locations of a battery pack.
However, this process was prone to high voltage arcs and was difficult to perform with a certain degree of efficiency and quality continuously. But with a sophisticated Vision-Language-Action (VLA) learning model, Xiaomo can analyze the environment and carry out a task based on requirements with a high degree of accuracy despite unforeseen variations.
The functions and abilities of the robot go far beyond simple repetition. It self-corrects for deviations in material position and changes in the point of connection, with real-time dynamic adjustment of its own posture. When engaging with and disconnecting flexible wiring harnesses, it controls the force levels during connection, preventing damage to the equipment yet securing the connection effectively. According to reports from CATL, Xiaomo has an actual connection success rate above 99% and comparable cycle times with human workers during its applications in actual production processes.
Apart from its original purpose, Xiaomo constantly observes the wiring harness connections in order to instantly inform of irregularities in an attempt to decrease the rate of defect production. When there are operational pauses, it automatically changes to an inspecting state, further elevating manufacturing consistency. According to CATL, “Moz has become an indispensable member of our production line.”
The use of Xiaomo is a part of CATL’s ambitious plan to incorporate robotics and AI in industrial settings. The emphasis on embodied AI, or immersing artificial intelligence in physical machines, is part of a greater movement in China’s manufacturing industry. The adoption of intelligent robots among local car firms and suppliers is being pushed ahead in a bid to increase production efficiency and remain competitive in a global market. Other players in this industry are working towards using humanoid robots in car and component production, with Changan Automobile, Xpeng, UBTech Robotics, and AgiBot at the fore.
CATL’s push for automation happens when the company is performing well in the market. During the first ten months of the year, the firm has installed 355.2 GWh of EV batteries, a 36.6 percent increase from last year, and has achieved a 38.1 percent market share globally, the only one above 30 percent. In the Chinese market, the company installed 40.87 GWh of batteries in November, with a market share of 43.71 percent. It has also started mass-producing the fifth-generation LFP batteries with increased energy density and a longer cycle life. Aided by the use of humanoid robots to perform precision-critical tasks in high-voltage processes, CATL not only lowered any risk of industrial accidents but also paved the way for the manufacturing process in the future. The company’s “Xiaomo” robot’s entry into the market signals the beginning of a new era in industrial robotics where adaptability brought about by AI is at par with the needs of a high-throughput production line.