Chinese startup uses VR teleoperation to train humanoid robots for industrial tasks

In Shenzhen, workers at IO-AI Tech use VR headsets, handheld controllers, and motion-tracking gear to remotely control humanoid robots for tasks such as stocking shelves and picking items from bins. The company’s technology aims to automate industrial and retail work while simultaneously collecting training data to enable future autonomous operation of the robots.

During a demonstration, a reporter controlled 10 robotic hands from different companies using a custom motion-tracking glove, which instantly transferred finger movements to all 50 robotic digits. The system also provided haptic feedback, allowing the reporter to feel objects manipulated by the robotic hands. The company also showcased a system being tested by a Chinese convenience store chain, where workers used a VR headset and grippers to pick up boxes of medication from shelves, simulating real-world automation tasks.

IO-AI Tech’s technology is designed to adapt human movements to a variety of robot forms, addressing the challenge posed by the diversity of humanoid robots and robotic hands available in China’s market. The startup’s algorithms incorporate a blend of human control and autonomy to account for differences in robot shape, size, and weight, ensuring stability and functionality during operation.

The company’s location in Shenzhen, a hub for manufacturing, facilitates collaboration with local manufacturers eager to automate tasks. IO-AI Tech is working with Jack Sewing Machines, a company that produces clothes manufacturing equipment, to train two-armed robots for tasks like ironing shirts. These robots are intended to integrate into existing production lines, automating work currently performed manually.

IO-AI Tech’s cofounder, Si Chin, emphasized the incremental approach to deploying AI-powered automation, comparing it to the development of self-driving cars. She noted that teleoperation data focused on specific tasks is essential for advancing robotic capabilities. The approach is also being adopted in some Chinese vocational schools, reflecting broader interest in robotics training.

The startup’s efforts align with China’s broader manufacturing prowess in producing affordable, high-quality robots like those from Unitree. By combining teleoperation with AI-driven training data, IO-AI Tech is positioning itself to play a key role in advancing the physical capabilities of AI systems.