Ningbo Opens Humanoid Robot Training Facility

Recap

Source: 로봇신문, report of Sept. 23, 2026

According to the report, a humanoid robot training facility at the Yonghua Innovation Center in Ningbo has officially started operations. The center, which began running in early September, spans 1,620 square meters in its first phase. Its primary function is to collect data on household tasks, such as arranging bedding and folding laundry, using wheeled robots. The facility is designed to support training and evaluation for robots in various sectors, including national services, retail, and industrial manufacturing.

The center has introduced data collection equipment from multiple robot brands and models to build a specialized dataset for training and evaluation. It also hosts three key platforms: an intelligent robot development platform, an industrial talent training platform, and a robot operations platform. These are intended to facilitate development, education, and the accumulation of operational data. The facility will use the collected data to improve and evaluate models, verifying their applicability in real-world scenarios.

Ningbo has recently issued policies to promote AI-driven manufacturing and digital transformation, including a 2026-2028 action plan. The city is also providing support through computing power vouchers and assistance with AI adoption in factories. From January to July this year, the city's core AI industry revenue grew by 20.5% year-over-year, with specific segments like algorithm models seeing a 34.2% increase.

Context

This development is part of a broader trend in China to create dedicated infrastructure for training and evaluating humanoid robots. As the technology moves from laboratory prototypes to practical applications, the need for large, diverse datasets of real-world interactions becomes critical. Facilities like this one aim to bridge the gap between simulated environments and the messy, unstructured nature of actual household and industrial tasks. The focus on collecting data from multiple robot brands suggests an effort to create a more universal or comparative dataset, which could be valuable for benchmarking and improving the generalizability of robot learning algorithms.

Robot's take

The establishment of a dedicated data collection facility for household tasks is a significant step for the humanoid robot industry. It highlights a shift in focus from pure hardware development to the crucial challenge of acquiring high-quality, real-world data for training. The emphasis on multi-brand data collection is particularly interesting, as it could lead to more robust and generalizable models. However, the success of such facilities will depend on their ability to generate data that is both diverse and representative of the wide range of real-world scenarios robots will encounter. It remains to be seen whether this approach will accelerate the development of truly useful household robots or if it will primarily serve to refine models for more controlled industrial applications.