Real-World Training! Two Departments Jointly Launch Special Campaign for Humanoid Robots and Embodied Intelligence

Gasgoo Munich- To move humanoid robots and embodied intelligence from lab validation to routine real-world operation, the Ministry of Industry and Information Technology and the State-owned Assets Supervision and Administration Commission have jointly unveiled a special action plan for 2026. The directive prioritizes "training by scenario" and "iterative refinement," outlining six key tasks across industrial manufacturing, public services, and specialized operations. The goal: accelerate a closed-loop cycle of training, iteration, application, and re-optimization.

Image Source: Kepler Robotics

China's humanoid robotics and embodied intelligence sectors are currently at a critical juncture, shifting from demonstration to large-scale deployment. Yet bottlenecks remain in model algorithms, hardware performance, scenario adaptation, and the accumulation of real-world data. The initiative proposes creating centralized, standardized training environments to prevent redundant construction and cut trial-and-error costs. By forcing products to iterate under actual working conditions, the plan aims to bridge the gap from "functional" to "truly useful."

The initiative spans major hubs including Beijing, Tianjin, Shanghai, Jiangsu, Zhejiang, Shandong, Hubei, Hunan, Guangdong, and Sichuan, alongside relevant state-owned enterprises. Leveraging national AI innovation pilot zones, the drive targets three primary scenarios: industrial manufacturing (production, inspection, maintenance, and logistics); public services (catering, retail, and healthcare); and specialized operations (safety production, emergency rescue, and disaster relief). By conducting physical training in authentic environments, the program seeks to refine embodied intelligence algorithms, accumulate high-quality real-world data, and upgrade critical hardware components.

To achieve these goals, the action plan defines six key missions:

Establish real-world training environments that provide authentic operational settings for physical machine training;

Form innovation consortia that pool resources from industry, academia, research, and application around specific scenarios;

Develop practical operational skills, shifting from "stunt mode" to "work mode" by refining skill sets and algorithmic models;

Scale up mature applications to achieve a ripple effect: validate one solution, deploy a batch, and drive adoption across an entire sector;

Reinforce support for key elements by deepening the implementation of standards, talent development, and financial backing;

Consolidate successful training protocols into replicable technical pathways and management mechanisms that can be adopted across different regions and industries.

Regarding support measures, the Ministry of Industry and Information Technology and SASAC will oversee overall coordination. Regions and enterprises demonstrating significant results will receive preferential support in policies, standards, and projects, alongside broader promotion of their achievements. Provincial industrial authorities and participating state-owned enterprises must strengthen organizational guarantees, using special funds and government incentives to open up scenarios and drive technical breakthroughs. A list-based management system with regular tracking and evaluation will be established to ensure tasks are effectively implemented.