Remote Operators Train Humanoid Robots in Shenzhen's Hardware Factories

Teleoperated Humanoids as a Training-Data Factory

According to Wired AI, IO-AI Tech, a startup located approximately 45 minutes north of Shenzhen, has begun deploying teleoperated humanoid robots across Chinese warehouses and convenience stores. Workers wearing VR headsets, body-tracking sensors, and motion-capture gloves control robots remotely—stocking shelves, picking items from bins, and performing household tasks like folding clothes. The company’s primary objective is dual: enable robots to perform economically useful work today while capturing the motion-control data needed to train autonomous systems tomorrow.

The startup addresses a fragmentation problem endemic to China’s robotics hardware ecosystem. Dozens of incompatible humanoid designs compete in the market, each with different proportions, joint configurations, and control interfaces. IO-AI’s motion-transfer algorithms create a unified teleoperation layer, translating a single human operator’s movements into commands compatible with robots from different manufacturers—including Unitree humanoids demonstrated on-site.

Motion Mapping and Hybrid Autonomy

The core technical challenge is that human morphology does not align with robotic morphology. According to Wired AI, IO-AI’s algorithms blend direct human control with learned autonomy modules. When a human operator wearing a motion-tracking glove manipulates a robot hand, the system transfers finger movements in real-time. However, for larger systems—full-body robots performing tasks like shelf-stocking—pure teleoperation is insufficient. The algorithms grant robots a degree of independent motion, such as balance correction and fine-tuned limb positioning, to prevent collapse when the human input does not perfectly fit the robot’s physical constraints.

This hybrid approach serves both immediate deployment and long-term automation. Workers wearing the company’s VR equipment gather labeled training data—the human operator’s intent paired with the robot’s executed motion and outcome—which IO-AI can later use to train autonomous policies that require no remote operator.

Shenzhen’s Role as a Hardware Testbed

IO-AI Tech cofounder Si Chin emphasized to Wired AI that Shenzhen’s concentration of manufacturers and robotics companies creates an operational advantage. The location enables rapid prototyping across multiple robot morphologies and close collaboration with local industrial partners. The article notes that Jack Sewing Machines, a Shenzhen-based clothes-manufacturing equipment maker, is among the local firms working with IO-AI to automate production tasks.

Why This Matters

Teleoperated humanoid robots reframe the labor economics of warehouse and light-manufacturing automation. Rather than requiring fully autonomous systems before deployment—a threshold that remains years away for many tasks—companies can deploy semi-autonomous robots controlled by remote human operators, paying operator wages to perform work on-site through robots. Simultaneously, every hour of teleoperation becomes a labeled training example for autonomous models, creating a data-flywheel that could accelerate the transition to truly independent systems. For robot manufacturers fragmented across incompatible designs, motion-transfer platforms like IO-AI’s become critical infrastructure, shifting the economic barrier from hardware standardization to software compatibility. If the approach scales beyond Shenzhen’s manufacturing ecosystem, it could reshape job categories in logistics and retail—replacing on-site workers with remote operators, while accelerating the obsolescence of those roles once autonomous systems mature.