Robots enrolled in the first class will begin by learning 45 “basic skills”, such as grasping, lifting, landing and moving objects. These movements seem simple, but tasks such as folding laundry, arranging shelves or cleaning equipment are difficult for humanoid robots.
China has unveiled its first “training school” for humanoid robots, an infrastructure in Shanghai designed to teach machines the practical skills they will need to work in homes, factories, hospitals, service companies and farms.
The center is expected to come into full operation in July and will begin by bringing together more than 100 robots from more than a dozen companies, becoming a test bed for one of the fastest growing technological sectors in the country.
Located in Zhangjiango hub Shanghai high-tech installation, with 5000 m²will host a pilot program aimed at a diverse group of humanoid robots, with different dimensions, designs and mobility capabilities.
The objective is twofoldsays: prepare robots for real environments and collect large volumes of training data that can be shared by the Chinese robotics industry.
The center is operated by the National and Local Co-built Humanoid Robotics Innovation Center, which has been developing the project for several years. Second Xu Bingeneral director of the installation, the objective is to support the “sharing and use of data on a large scale and strengthen the humanoid robotics sector”.
Robots registered in first class will start by learning 45 basic “atomic skills”, including grasp, lift, place and carry objects.
These movements may seem simple, but they are essential for machines expected to work in hotels, factories, healthcare facilities or private homes.
In practice, tasks such as folding clothes, arranging shelvescleaning equipment, or moving objects continue to be difficult for humanoid robotsespecially when they require dexterity, balance and decision-making capacity.
Training will also focus on sequences of actions. According to the center’s director of market systems, Yang Zhengyerobots are expected to go beyond isolated commands and learn to perform more complex taskss based on autonomous decisions, using data collected during training.
One of the biggest challenges will be improving the form how robots interact physically with the world. Grasping objects, e.g.has long been a difficult problem in robotics.
Humans instinctively adjust pressurethe right angle and timing, but robots have to be trained to understand when to hold, move or release objects without damaging them or dropping them.
To generate the data needed for these improvements, human instructors will be able to supervise robots repeat a single movement hundreds of times per day. In some cases, a scientist may be able to guide the same basic action up to 600 times, recording performance data throughout the entire process.
The 2026 group of robots will focus on 10 major task areas linked to likely commercial uses, including domestic work, industrial applications and tourism. The long-term ambition is create a data exchange model that allows robotics companies to draw on shared training resources while also specializing in specific markets such as healthcare or hospitality.
Aggregated data could also contribute to what researchers describe as a general purpose robotic model — a kind of “super brain” that allows robots built by different manufacturers to benefit from the same learning process.
China’s bet comes at a time when humanoid robots are rapidly passing from demonstrations to the first phases of use commercial.
In some industrial contexts, robots already work alongside humans, while consumer models have become cheaper and more capable in recent years.
Even so, technology is far from perfect, and the Shanghai center aims to bridge the gap between performance in the laboratory and usefulness in everyday life.
For China, the project It’s not just about teaching robots to fold shirts or transporting trays. It’s also about building the infrastructure, data and industry standards needed to make viable, on a large scalethe humanoid machines of the future.
