Jenk Infang, head of RealMan Robotics’ humanoid robot data training center, introduces its Beijing facility and explains how it collects data to train humanoid robots for household and workplace tasks.
Infang says the center has about 100 humanoid robots operating across 10 real-world training environments. It aims to provide data for large model companies so they can improve their models and develop robots for everyday use.
The center divides training into general skills and vocational scenarios. Infang gives housekeeping and supermarket stocking as examples of jobs robots may learn to perform.
A home setup lets robots practice handling delicate items such as fruit and eggs. In a bedroom and balcony scene, robots practice folding towels, with possible future applications including bed-making and laundry; Infang says clients in California are testing robots in hotels.
An agricultural harvesting setup captures tactile feedback, finger-joint current, and visual inputs. The center collects varied data and selects subsets to train models for particular scenarios.
Infang says each robot produces 100 to 200 data entries per day. Since the platform launched on March 28, the center has received data orders and partnership inquiries and has brought on nearly 10 partners, including Chinese universities and research institutes. It tailors datasets to client needs and universal standards, then supplies them to large model companies.
Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.
Edward Liu of RealMan Robotics explains how RealMan is building hardware, fleet operations, and data infrastructure to move physical AI from demos to daily work at the Humanoids Summit.
Kenneth Wren, director of overseas solutions at RealMan Robotics, discusses robot training and human-robot collaboration in a Bloomberg interview.
Donny Li of RealMan discusses how robot hardware and real-world training data can accelerate everyday robot use at the Humanoids Summit.
