While a technician wearing a full-body motion capture suit that features VR goggles, handheld controllers, and sensors strapped to his legs bends over, lifts his legs and turns, a humanoid robot in front of him mirrors every movement in real time.
This is a novel scene at the Yunwan Embodied Intelligence Innovation Center in Guangzhou in south China, displaying the latest invention by the Guangzhou Baiyun Xuanji Technology Industry Investment Co., Ltd.
Last year, the team could only control the upper body using a whole-body control method, and the lower body had to be operated by a remote control,” said Lu Zhicong, deputy general manager of the company. “This year, we are developing a solution to control the full body. It’s still being refined, and once optimized, we expect motion replication accuracy to reach above 90 percent.”
“What we do is to select the right models for specific scenarios and carry out tailored secondary development,” Lu explained.
The company’s technicians are developing a firefighting robot dog. Although robot dogs already exist on the market, the team’s work focuses on secondary development by integrating infrared sensors, spark detection sensors and communication modules, combined with vision recognition and autonomous exploration algorithms, to meet the specific demands of fire rescue.
“Our goal is to get robots into real production and living scenarios to handle the repetitive, high-risk work that humans shouldn’t have to do,” Lu added.
Li-Gong Industrial is pursuing a similar goal. At the 2026 World Artificial Intelligence Conference, multiple robot workstations at Li-Gong’s exhibition booth drew crowds, who got to witness smart retail, intelligent sorting, material handling and warehousing, with picking, restocking and transport forming a complete cycle.
How to make a robot replicate human manipulation? Cao Yuran, the company’s senior marketing manager, pointed to a pair of blue gloves. “This is a data capture kit we launched globally in April.”
He broke down how the system works: a head-mounted binocular camera captures the first-person view and a wrist-mounted micro-camera tracks fine hand movements, while a tactile sensing glove with over a hundred pressure points records fingertip and palm feedback in full detail.
“Tasks like applying sealant or assembly have always depended on the intuitive touch of a master craftsman. Now we capture that feel through vision and touch, convert it into datasets, and train robots to replicate the logic.”
“This doesn’t mean replacing workers,” Cao said. “Manufacturing has a labor shortage, especially young people don’t want repetitive assembly jobs. Robots take over the repetitive work, freeing people for higher-value tasks like machine adjustment, production scheduling and process optimization.”
Robot orders are currently booked through October, he said, adding that overseas markets are growing fast, with clients located in South America, Europe and Southeast Asia.
Most of Li-Gong’s employees are in the research and development department, Cao revealed. “We use self-developed automation equipment to manufacture robots that then produce more robots.”
This is precisely the shift happening in China’s embodied intelligence industry: from “able to walk” to “able to work” and from the showroom to the production line.
Data from the Ministry of Industry and Information Technology show that China-developed quadruped robots account for nearly 70 percent of global sales, while China has more than 400 humanoid robot models, over half of the world’s total.
Li Yan, a researcher at the Development Research Center of the State Council, noted that embodied intelligence is evolving from a production tool that executes single tasks according to fixed programs to an intelligent agent capable of environmental perception, task decision-making and flexible collaboration. “It is expected to gradually find applications in industrial manufacturing, agriculture, special operations and public services.”
The metro system in south China’s Shenzhen offers a vivid example. A fleet of autonomous logistics robots from 1024 Robotics now runs daily on the city’s Metro Line 16. During off-peak hours, they shuttle between stations, transporting supplies and goods. These robots automatically pass through fare gates, take elevators, board and alight from trains, and navigate to their destinations, all without GPS.
He Zhijian, CEO of 1024 Robotics, noted that the company’s growth has relied on the local government’s “scenario forest” approach. The Pingshan district government in Shenzhen has actively connected the company with real-world use cases and customers, helped the company lease 3,400 square meters of integrated R&D and production space, and facilitated autonomous driving permits.
China’s 15th Five-Year Plan (2026-2030) explicitly identifies embodied intelligence as a new economic growth point. An action plan has been launched for humanoid robots and embodied intelligence real-scenario training, aiming to identify more than 100 high-value application scenarios.
On July 30, the Political Bureau of the Communist Party of China Central Committee held a meeting, making it clear that efforts should be made to accelerate the building of a modern industrial system, strengthen long-term and stable support for basic research, further implement the “AI Plus” initiative, foster new forms of the intelligent economy, and improve the AI governance system.
Active moves were also urged to achieve breakthroughs in frontier technologies and advance the development of future industries, with a focus on building emerging pillar industries and continuously promoting the transformation and upgrading of traditional industries, the meeting said.
The China Embodied Intelligence Industry Development Report (2026) revealed that China’s embodied intelligence market has grown from approximately 213.3 billion yuan (about 31.4 billion U.S. dollars) in 2018 to an estimated 1.09 trillion yuan in 2026, with a compound annual growth rate of 22 percent to 23 percent.
China has the world’s richest application scenarios and massive real-world data sources, providing a large-scale testing ground for the deployment, iteration and implementation of embodied intelligence, said Wei Kai, an AI expert with the China Academy of Information and Communications Technology.