Wheeled humanoid robot “Yuanzheng A2-W” developed by Chinese robotics startup Agibot (Yonhap)
SEOUL, Nov. 10 (Korea Bizwire) — China is racing ahead in the development of humanoid robots and physical artificial intelligence (AI), focusing on the collection and production of real-world sensory and motion data that are far more difficult to obtain than large-language model (LLM) datasets, a new report shows.
According to a study released Sunday by the National Information Society Agency (NIA) in Seoul, Chinese companies and government agencies are aggressively building industrial-scale data collection systems to strengthen the country’s dominance in physical AI—technology that allows robots to perceive, move, and interact in three-dimensional environments.
Among the most prominent players is Shanghai-based Agibot, a humanoid robot startup backed by Tencent. The company operates a massive “data factory” that deploys about 100 robots and 200 human workers to generate between 30,000 and 50,000 data samples a day, including motion, gesture, and environmental data. The firm’s open-source dataset, Agibot World, has reportedly supplied around 80 percent of the real-world data used to train NVIDIA’s “Isaac Groot N1” foundation model for humanoid robotics.
The Chinese government is also providing public facilities—such as hospitals, schools, and community centers—as testbeds for collecting high-quality physical AI data across nine sectors, including manufacturing, education, healthcare, and real estate management.
Meanwhile, South Korea remains in the early stages of physical AI data development. The country’s AI Hub platform provides a foundation for robotics and autonomous systems research but lacks diversity, international compatibility, and real-time data, the NIA report said.
To close the gap, the agency urged the government to accelerate the creation of deep datasets encompassing robotic behavior, sensor, and environmental data, and to connect them with digital-twin simulation environments such as ports, airports, hospitals, and smart buildings.
The report also called for a standardized national consortium to oversee safety, data quality, and interoperability across sensor, robotics, and AI platforms, warning that without swift action, Korea risks falling behind in one of the most strategically important AI frontiers.
Kevin Lee (kevinlee@koreabizwire.com)
