In December 2025, CATL deployed humanoid robots at scale on a live battery pack production line at its Luoyang, Henan facility. The robot is named Xiaomo. The task: high-precision battery connector insertion. Xiaomo’s verified success rate is 99%. Its daily output is triple that of a human worker, because it operates without breaks across all three shifts.
Stella’s Analysis
Connector insertion is an ideal test case for humanoid robots in manufacturing. It combines three engineering challenges: spatial precision (approximately 1 mm tolerance on connector alignment), force control (enough force to seat the connector, not enough to crack the housing), and cycle time (the production line does not wait).
Xiaomo solves this with vision-guided alignment combined with compliant end effectors. Stereo cameras plus tactile feedback locate the connector, pre-align the approach vector, then apply graduated insertion force while monitoring resistance in real time. If resistance exceeds threshold, the system resets and retries. This is not brute-force automation. It is a closed-loop control system that replicates what an experienced technician does by feel.
The 99% success rate means 297 out of 300 insertions complete correctly. The 3 failures per 300 cycles go to a rejection queue and human quality review. CATL designed the system to handle failure gracefully, not to eliminate it entirely. The alternative, waiting for 100% before deploying, would mean this line never ran. That is a pragmatic engineering decision, and it is the right one.
The triple-output figure comes from uptime. Xiaomo works three shifts continuously. A human technician sustains high-concentration assembly work for 6 to 8 hours at acceptable accuracy. Xiaomo’s accuracy is constant from hour 1 to hour 24.
Industry Impact
CATL’s deployment is significant for three reasons. First, it proves the use case on a high-stakes production line. EV battery connectors are safety-critical. A missed connection can cause a thermal event. This is not a pilot in a low-risk demo cell.
Second, it creates an industry template. CATL holds 38.1% global battery market share and is the only battery supplier worldwide above 30% share. When they deploy a technology at scale on a live line, the industry follows within 18 to 24 months.
Third, it confirms that humanoid robots do not need to do everything to be economically justified. One task, done reliably, at high volume and high uptime, pays for the machine. Additional capability is upside, not requirement.
For context on how aluminum and metal production lines are integrating industrial robotics, see our analysis of aluminum production robotics.
Trends
CATL has announced plans to expand Xiaomo’s deployment across its global battery pack lines. The next development target is multi-task capability: a single robot handling connector insertion, thermal pad placement, and final inspection in sequence on the same station.
The enabling architecture is policy switching: the robot loads a different motion model per task and transitions based on station position. This is how you get from a single-task system to a flexible automation asset without hardware changes.
See our analysis of innovations in China’s inspection landscape for how quality control is evolving alongside this trend. Also relevant: the economics of robot body business in Chinese industry.
