Recently, the Ministry of Industry and Information Technology (MIIT) and the State-owned Assets Supervision and Administration Commission (SASAC) of China jointly issued a circular launching the 2026 Humanoid Robot and Embodied AI Real-Scene Training Special Action.
Driven by practical applications, the initiative targets key scenarios across industrial, special, and service sectors. It highlights an integrated approach to constructing real-scene training spaces, fostering collaborative innovation syndicates, overcoming challenging operational skills, and validating real-world application deployments.
By conducting rigorous training in physical settings, the program aims to continuously optimize embodied AI models and algorithms, accumulate critical high-quality real-machine data, and enhance the physical performance of core components. It also aims to explore and establish comprehensive life-cycle management and security mechanisms for humanoid robots and embodied AI products.
[AgentUpdate Depth Analysis] This national initiative represents a pivotal shift for Embodied AI, transitioning from virtual simulations to physical, real-world validation. While digital AI Agents have made tremendous strides in virtual environments, physical Embodied Agents still face a severe bottleneck: the scarcity of high-quality, real-world sensory-motor data. By systematically constructing "real-scene training spaces," this action addresses the core "data drought" in physical AI. It bridges the gap between hardware manufacturers and foundation model developers. Compared to relying solely on synthetic datasets, training agents in real-world scenarios dramatically improves their generalization capability and safety profiles. This policy-driven push will accelerate the maturity of the Embodied AI ecosystem, laying a robust foundation for next-generation physical agents in complex industrial and domestic environments.
