China is making a significant move in the field of robotics with the launch of its first large-scale humanoid robot training hub. Set to become fully operational by July, the facility in Shanghai is already working with over 100 robot models according to Global times.

Occupying more than 5,000 square meters in the Zhangjiang area, the center is designed to support collaboration among robot developers by providing a shared environment for training and data generation.

According to center director Xu Bin, the facility is addressing longstanding challenges in the industry—such as inconsistent data formats, siloed innovation, and limited coordination between robot manufacturers. By bringing various robot platforms together under one roof, the center enables more efficient development and shared learning.

More than 100 types of humanoid robots from over a dozen companies are already participating in training. These robots are being taught around 45 core skills—simple but essential actions like grasping, placing, and transporting objects. These movements are demonstrated repeatedly by human trainers while the robots collect sensor data and learn to replicate the actions with increasing accuracy.

Yang Zhengye, the center’s director of market systems, explains that even minor variations—such as changes in object orientation—generate valuable data points. This allows the robots to gradually learn how to adapt to dynamic environments and real-world uncertainties.

The facility focuses on ten key application areas, ranging from household support to industrial tasks and tourism services. Complex jobs are broken into manageable subtasks that robots can learn over time.

Currently, the center generates between 20,000 and 30,000 motion data entries daily, with expectations to exceed 50,000 once it reaches full operation. By the end of the year, the goal is to accumulate over 10 million real-world data entries—offering a rich dataset for further development.

In the long term, the center aims to launch a shared data platform, allowing robot developers to access scenario-specific datasets across industries such as healthcare, manufacturing, and home automation. This approach is expected to reduce redundancy and improve interoperability.

The collected data will also support the development of a general-purpose model for embodied intelligence—a kind of shared control system that could allow robots from different companies to work more seamlessly together.

As the Shanghai training hub prepares for full capacity this summer, it marks an important effort in building smarter, more adaptable humanoid robots—built not in isolation, but through shared knowledge and collaboration.