In-Depth Analysis | MIIT Joint Science And Technology Letter [2026] No. 256: Special Initiative For “Real-World Training” in Humanoid Robots And Embodied Intelligence Officially Launched

The issuance of this document has a clear industrial and temporal context:

1. Key Industry Milestone: China's humanoid robotics and embodied intelligence industries are currently at a critical juncture, transitioning "from the laboratory to real-world scenarios, and from demonstration and validation to routine operations." In the past, market discussions on embodied intelligence focused primarily on robot bodies, joints, motors, and demonstration videos, while the actual implementation of the industry faced core bottlenecks regarding "where to train, what to use for training, and how to validate."

2. Implementation of Higher-Level Directives: The document explicitly aims to thoroughly implement the decisions and directives of the CPC Central Committee and the State Council, and to fulfill the requirements of relevant guidelines and action plans regarding the innovative development of the humanoid robotics and embodied intelligence industries.

3. Inter-departmental Coordination: The joint issuance of this document by the Ministry of Industry and Information Technology (MIIT) and the State-owned Assets Supervision and Administration Commission of the State Council (SASAC) signifies that the policy not only covers local industries but is also deeply embedded in the intelligent transformation process of central state-owned enterprises (SOEs), reflecting the strategic intent to position SOEs as core application scenario platforms.

Overall Approach: Adhering to an "application-driven" approach, we will focus on three major sectors-industrial, specialized, and service-and advance four key tasks in tandem: "construction of real-world training spaces, cultivation of innovation application consortia, breakthroughs in operational skills, and verification of application deployment."

Core Objectives by the End of 2026:

· Key products, such as humanoid robots, will be the first to complete application verification and routine deployment in a number of representative scenarios, marking the launch of "operational mode"

· Identify and refine more than 100 high-value application scenarios

· Further expand the spectrum of embodied intelligence applications

· Drive the development of deployment capabilities on a scale of tens of thousands of units

This set of objectives is highly pragmatic-it emphasizes "validation" and "deployment" rather than mere technological breakthroughs, and prioritizes "operational mode" over demonstration capabilities. The "tens of thousands of units" target also reflects a clear expectation for large-scale implementation.

This is a fundamental task of the special initiative. The core logic is that without real-world scenarios, it is impossible to develop usable embodied intelligence products.

Key points:

· Focus on three major sectors-industry, services, and specialized operations-and target key scenarios such as manufacturing, testing and analysis, maintenance and repair, warehousing and logistics, food and retail, healthcare and wellness, workplace safety, emergency response, and disaster prevention and mitigation

· Select "real-world scenario units" (production workstations, service operation points, and emergency response stations) as practical training platforms

· User organizations will carry out environmental adaptation and renovation in accordance with the principles of "minimal intervention" and "reuse of existing resources"-the key to this principle is to lower the participation threshold and prevent users from being deterred by excessive renovation costs

· Quantitative requirements: Each provincial-level region must select no fewer than 20 key scenario units, covering at least two of the three major sectors; each central state-owned enterprise must select no fewer than 10 key scenarios

The highlight of this mechanism's design lies in breaking the vicious cycle where "users wait for products, and manufacturers wait for demand."

Core Mechanism:

· Centered on user organizations and original equipment manufacturers (or application service providers), in collaboration with model and algorithm developers, component supply chain enterprises, and research institutes

· User organizations are responsible for providing real-world training environments, quantifying deployment goals, and supplying operational workflow data and environmental semantic information

· System integrators are responsible for developing capabilities in scenario understanding, task planning, operation execution, human-machine collaboration, and continuous learning

· The consortium is required to sign a cooperation agreement that clearly defines development metrics, task boundaries, intellectual property ownership, and profit-sharing arrangements, thereby establishing a long-term operational mechanism

Clearly defining intellectual property ownership is crucial-this not only protects the interests of participating parties but also prevents disputes during subsequent commercialization.

This is the core objective of the technical research initiative, emphasizing that robots must be capable of performing tasks effectively in real-world work environments.

Research Focus:

· Task Skill Sets: Develop practical, replicable end-to-end solutions

· Algorithm Level: Enhance the robustness and adaptability of "big brain/small brain" model algorithms, and strengthen generalization and fault-tolerance capabilities under complex or abnormal operating conditions

· Data: Build high-quality, high-fidelity datasets and encourage open sharing while ensuring safety

· Computing Power: Refine deployment models such as cloud-edge-device collaboration and offline autonomy

· Safety: Refine capabilities such as collision detection, force control limits, emergency braking, and black box systems to ensure operational safety in mixed human-vehicle environments

Of particular note is the "black box" requirement-this indicates that policymakers have fully anticipated safety incidents and are seeking to establish a mechanism for reconstructing the facts of accidents.

The core of this initiative is to establish a scientific evaluation system and promotion mechanism.

Key mechanisms:

· User organizations (or third-party institutions) develop application verification test procedures and compliance criteria

· Evaluation metrics: actual task success rate, efficiency improvement rate, safety and reliability, and economic feasibility

· For solutions that pass validation, promote their routine deployment at user organizations and in similar scenarios

· Encourage exploration of the "humanoid robot-as-a-service" model, lowering barriers to entry for users through pay-for-performance and operating leases

It is significant that "economic feasibility" has been explicitly included as an evaluation criterion-this indicates that the policy encourages not merely technical feasibility, but a commercially viable business model that adds up.

This initiative addresses supporting infrastructure issues:

· Standards: Participate in the work of standardization technical committees to strengthen the management of "identity" information for complete units

· Talent: Cultivate multidisciplinary professionals who possess expertise in both core technologies and practical applications

· Finance: Coordinate equity, debt, and insurance instruments to provide end-to-end financial services

· Institutional Framework: Encourage local governments to explore institutional arrangements and insurance policies that promote industrial development

· Develop comprehensive operational guidelines covering the entire process, including scenario adaptation, environmental customization, deployment verification, and daily operations and maintenance

· Provincial-level regions may recommend up to 10 outstanding training programs, while central state-owned enterprises may recommend up to 5

· Compile best practices regarding mechanisms for establishing innovation consortia, data rights clarification and sharing, and the distribution of commercial benefits

· Plan Submission (by June 30, 2026): Identify key scenarios, complete the work plan form, and submit it to the Ministry of Industry and Information Technology (Department of Science and Technology) and the State-owned Assets Supervision and Administration Commission of the State Council (Planning Bureau)

· Process Monitoring (ongoing): Implement list-based management, establish a work ledger, and undergo regular monitoring and evaluation

· Results Summary (by November 30, 2026): Submit a summary of the special campaign's results and compile performance metrics for scenario applications

The tight timeline (less than one month from the issuance of the document to the first submission) underscores the urgency of implementing this policy.

· Organizational Support: The two ministries will coordinate overall efforts and promote best practices through channels such as the "In-Depth Tours" initiative and the "AI+" Embodied Intelligence Industry Consortium.

· Resource Allocation: Regions and enterprises that demonstrate strong implementation results will receive preferential support in terms of policies, standards, and projects.

· Local Support: Provincial-level competent authorities and centrally administered state-owned enterprises should increase their support through dedicated funds, government incentives, and other means.

Based on public sentiment analysis and policy requirements, attention must be paid to risks across four key areas:

1. Preventing "showpiece" implementations: Some organizations may rush into procurement or duplicate projects to meet deadlines, leading to formalistic practices such as "robots being brought into factories for photo ops," which can easily spark public criticism regarding the "waste of state assets."

2. Mitigating "robot-replacing-human" anxiety: The large-scale application of embodied intelligence will inevitably lead to job restructuring. Companies must proactively communicate with employees and provide retraining to prevent internal anxiety from escalating into public opinion incidents.

3. Strictly prevent technical safety incidents: If a humanoid robot causes injury or equipment damage during practical training, the intensity of public backlash will far exceed that of traditional industrial accidents. The document has explicitly required the improvement of safety mechanisms; enterprises must establish safety assessment and information disclosure mechanisms.

4. Prudently handle property rights and data disputes: As consortia involve the sharing of large amounts of production line data and process parameters, clear boundaries and intellectual property ownership must be established to avoid trust crises such as "the loss of state-owned data."

Policy Significance:

· This is a landmark document marking the transition from "development" to "deployment," signaling a shift in national support for embodied intelligence from the R&D phase to the application phase.

· Through the "real-world training" approach, it places the initiative for industrial upgrading in the hands of end-users, creating a virtuous cycle where demand drives supply.

· The target of 10,000 units, the goal of refining 100 scenarios, and the dual-track approach involving central and local entities all serve as highly actionable, quantifiable drivers.

Implementation Challenges:

· Completing the entire process-from scenario selection, consortium formation, and practical training verification to results summarization-in less than seven months (June to November) presents an extremely tight timeline.

· Issues regarding interest allocation and intellectual property rights within the framework of central-local coordination and multi-stakeholder collaboration require continuous refinement through practical implementation

· Non-technical factors, such as security risks and public sentiment, may pose hidden obstacles to implementation

Overall Assessment:

Document No. 256 [2026] of the Department of Industry and Information Technology (MIIT) is a policy document that combines strategic foresight with practical operability. Its core value lies in establishing a closed-loop industrial mechanism comprising "scenario opening-joint research-practical training and validation-mass deployment." For local industry and information technology departments, central enterprises, and relevant market entities, this document not only provides a policy window but also sets forth clear action requirements and time constraints.