To: The industrial and information technology authorities of Beijing, Tianjin, Shanghai, Jiangsu, Zhejiang, Shandong, Hubei, Hunan, Guangdong, and Sichuan; and all relevant central state‑owned enterprises:

In order to thoroughly implement the decisions and arrangements of the Party Central Committee and the State Council, and to put into effect the requirements set forth in the relevant guiding opinions and action plans for the innovative development of the humanoid robot and embodied intelligence industries, we are promoting the normalized deployment and application of humanoid robot and embodied intelligence products in real production and living environments, and accelerating their large‑scale development. To this end, the Ministry of Industry and Information Technology (MIIT) and the State‑owned Assets Supervision and Administration Commission of the State Council (SASAC) have decided to jointly launch the 2026 Special Action for Real‑Scenario Training of Humanoid Robots and Embodied Intelligence. The relevant matters are hereby notified as follows:

I. General Approach and Objectives

Adhering to an application‑driven approach, and targeting key scenarios in the industrial, special‑purpose, and service sectors, we will promote in an integrated manner the construction of real‑scenario training spaces, the formation of innovation application consortia, the development of operational skills, and the verification and deployment of applications. Through real‑scenario training, we will continuously optimize embodied intelligence model algorithms, accumulate high‑quality real‑machine data, improve the performance of key components and parts of the robot body, and explore the establishment of a life-cycle management and support mechanism for humanoid robots and embodied intelligence products. By the end of 2026, humanoid robots and other key products shall have completed application verification and regular deployment in a number of representative scenarios, and shall have entered “operation mode”. We will summarize and form more than 100 high‑value application scenarios, further enrich the embodied intelligence application spectrum, and drive the ability to achieve large‑scale deployment at the level of tens of thousands of units.

II. Main Tasks

(1) Build Real‑Scenario Training Spaces.

Focusing on the application requirements of humanoid robots and quadruped robots in key scenarios such as production and manufacturing, inspection and analysis, maintenance and repair, warehousing and logistics, catering and retail, medical and elderly care, work safety, emergency rescue, and disaster prevention and mitigation in the industrial, service, and special‑purpose sectors, we shall organize relevant enterprises and institutions to select a number of real scenario units with clear target requirements, well‑defined working conditions, high standardization, and economic feasibility (including but not limited to production workstations, service operation points, and emergency response sites) as carriers for real‑scenario training spaces. The entities providing the scenario units (i.e., user entities) shall be determined and supported to carry out adaptation modifications of the operational environment in accordance with the principle of “minimal intervention and reuse of existing facilities”, to define the key technical requirements for scenario training, and to create real‑scenario training spaces that are trainable, testable, and verifiable. All provincial regions and central state‑owned enterprises shall adhere to the principles of demand‑driven, local adaptation, merit‑based selection, and differentiated layout to avoid clustering of scenario types. Each provincial region shall select no fewer than 20 key scenario units, covering at least two of the three sectors (industrial, service, and special‑purpose); each central state‑owned enterprise shall, in light of its industry sector, select no fewer than 10 key scenarios.

(2) Establish Innovation Application Consortia.

Support is to be given to user entities and complete robot manufacturers (or application service providers) as the main bodies, in collaboration with supply chain enterprises (including model algorithm and component suppliers) and research institutes, to establish innovation application consortia for each scenario. User entities shall be responsible for opening up real‑scenario training spaces, quantifying deployment and application goals, refining scenario requirements with respect to perception and decision‑making, operation and execution, interaction and collaboration, environmental adaptability and reliability, and providing necessary operational workflow data and environmental semantic information, as well as cooperating in scenario‑oriented research and effectiveness verification. Complete robot manufacturers (or application service providers) shall make good use of the real‑scenario training spaces, focus on improving capabilities in scenario understanding, task planning, operation and execution, human‑robot collaboration, continuous learning, and reliable operation, and promote the deep adaptation of humanoid robot, quadruped robot, and other complete products to scenario requirements. Supply chain enterprises and research institutes shall leverage their respective strengths to accelerate the iterative upgrading of key components and parts, strengthen research on key common technologies, and provide service support for application verification and large‑scale deployment. Innovation application consortia are encouraged to formulate task statements, designate lead entities, sign cooperation agreements, define research targets, task boundaries, intellectual property ownership, and benefit‑sharing mechanisms, and establish a long‑term operating mechanism.

(3) Develop Practical Operational Skills.

Innovation application consortia shall be guided to develop operational skill packages through real‑scenario training in accordance with the actual competency requirements of the corresponding job positions, so as to form implementable and replicable complete‑solution offerings. A number of embodied intelligence foundational models and motion control algorithms with strong scenario adaptability and leading performance indicators shall be developed, continuously improving the anti‑interference and self‑adaptive capabilities of the “brain and cerebellum” model algorithms, and enhancing generalization and fault tolerance under complex or abnormal operating conditions; contributions are encouraged to be open‑sourced in national‑level open‑source communities. A number of high‑quality, high‑fidelity datasets shall be constructed to improve the accuracy and precision of full‑dimensional information recording, including whole‑body motion trajectories, force‑position control curves, operation execution sequences, and temporal logic, while continuously enriching real‑scenario data such as spatial semantics, object properties, anomaly handling, unexpected interference, and boundary operating conditions. Data governance shall be strengthened to ensure the safe and orderly sharing of data, subject to data security, privacy protection, and trade secret safeguards. For different scenario requirements, model compression and inference acceleration technologies shall be optimized, and a variety of deployment models of computing power—such as cloud‑edge‑device collaboration and off‑grid autonomy—shall be distilled. Anti‑fatigue designs—including structural durability, thermal management, and power optimization—shall be reinforced to improve continuous operation capability under long‑duration and high‑load conditions. Safety capabilities such as collision detection, force control limits, emergency braking, and black‑box recording shall be improved to ensure operational safety in human‑robot mixed environments.

(4) Strengthen Real‑Scenario Application Verification and Regular Deployment.

User entities (or third‑party agencies commissioned by them) shall be guided to formulate application verification test procedures and compliance criteria based on scenario characteristics, scientifically evaluate key indicators such as the true success rate, efficiency improvement, safety reliability, and economic feasibility of the complete solutions, and issue application verification reports. For complete solutions that have passed verification, provincial regions and central state‑owned enterprises shall promote their regular deployment at user entities and in similar scenarios or related fields. For mature basic software and hardware products—such as operating systems, simulation platforms, and key components—that have been verified, they shall be promoted for expanded application in relevant complete robot manufacturers, achieving the effect of “verifying one, deploying a batch, and driving a broader scope”. Provincial regions and central state‑owned enterprises shall establish risk early‑warning and dynamic adjustment mechanisms for large‑scale deployment, reserving technology upgrade windows and exit pathways for areas with rapid technological iteration and significant scenario changes. Exploration of “humanoid‑robot‑as‑a‑service” models is encouraged, using innovative commercial approaches such as pay‑for‑performance based on utility and operational leasing to lower the entry threshold for users and accelerate market penetration.

(5) Strengthen Guarantees of Key Factors.

Innovation application consortia are supported to participate in the work of the MIIT Standardization Technical Committee on Humanoid Robots and Embodied Intelligence, and to strengthen the implementation and application of key standards. With reference to relevant standards and specifications for the life-cycle management of humanoid robots, the capacity for “ID” information management of complete robots shall be enhanced, and the human‑robot safety collaboration management mechanism shall be continuously improved. We shall cultivate multidisciplinary talents who are both proficient in core technologies and well‑versed in industry scenario applications, and accelerate the formation of an industrial talent echelon with a reasonable structure and high quality. Comprehensive financial services for enterprises shall be provided through the coordinated use of equity, debt, and insurance instruments. Each province shall, in light of the actual development of its industry, explore local institutional arrangements to promote the development and application of humanoid robot and embodied intelligence industries, and explore policies such as humanoid robot insurance.

(6) Condense Proven Experience.

For training solutions that are technologically leading, effective, and have mature models, we shall refine full‑process operational guidelines covering scenario adaptation, environmental modification, deployment verification, and routine operation and maintenance, so as to promote cross‑regional and cross‑industry exchanges with a point‑to‑area approach. Based on work practices, each provincial region may recommend no more than 10 excellent training solutions, and each central state‑owned enterprise may recommend no more than 5 excellent training solutions. We shall systematically summarize management norms for human‑robot collaboration management, product safety access, definition of operation and maintenance responsibilities, and emergency response for safety, and consolidate experience in mechanisms such as the establishment of innovation application consortia, data rights confirmation and sharing, and commercial benefit distribution, thereby supporting the rapid replication and promotion of the special action.

III. Work Requirements

(1) Identify Key Scenarios.

The provincial industrial and information technology authorities and relevant central state‑owned enterprises shall strengthen overall work coordination, proactively integrate the special action into their respective provincial (municipal) or group AI industry development plans and existing working mechanisms, accelerate the selection and confirmation of real scenario units and user entities, deepen supply‑demand matching, guide the formation of innovation application consortia, define work plans, and fill in the Work Plan Form for the Real‑Scenario Training of Humanoid Robots and Embodied Intelligence (attached). This form shall be submitted to the MIIT (Department of Science and Technology) and SASAC (Department of Planning) by June 30, 2026. The relevant provincial regions shall leverage the role of the National Pilot Zone for Innovative Application of Artificial Intelligence, deeply explore regional characteristic industrial advantages, proactively plan a number of high‑value benchmark scenarios that are application‑leading, technologically breakthrough, and commercially viable, and take the lead in the formation of innovation application consortia and application verification. Other provinces (municipalities) and enterprises may, in light of their actual work, refer to this action plan and submit work plans for carrying out relevant work.

(2) Strengthen Process Tracking.

The provincial industrial and information technology authorities and relevant central state‑owned enterprises shall implement list‑based management of key scenarios included in the special action, establish work records, and refine implementation measures. We shall adhere to a targeted and differentiated approach, providing ongoing element guarantees and support for scenarios that are already under joint research. For special scenarios involving industry access and safety supervision, communication and consultation with relevant industry authorities shall be strengthened to ensure that application verification and deployment are compliant and feasible. MIIT and SASAC will establish a regular tracking and evaluation mechanism to monitor progress and phased results, organize industry experts to conduct on‑site guidance, and coordinate the resolution of difficulties and bottlenecks encountered during implementation.

(3) Complete Effectiveness Summary.

By November 30, 2026, the provincial industrial and information technology authorities and relevant central state‑owned enterprises shall comprehensively summarize the completion of all tasks, compile statistics and analyze the effectiveness indicators of scenario applications, and submit the summary of the special action’s effectiveness to MIIT (Department of Science and Technology) and SASAC (Department of Planning). MIIT and SASAC will organize an evaluation of the overall implementation, and will promote outstanding achievements through the selection of typical AI application cases.

IV. Safeguard Measures

MIIT and SASAC will coordinate the overall implementation, promote the fulfillment of tasks, and utilize working platforms such as the “Deep Dive” supply‑demand matching events for AI‑empowered new industrialization and the “AI+” Embodied Intelligence Industry Community to publicize and promote the advanced experience generated during the special action, thereby strengthening its leading and demonstration effects. Regions and enterprises with good implementation results will receive preferential support in relevant policies, standards, and projects. The provincial industrial and information technology authorities and relevant central state‑owned enterprises shall strengthen organizational guarantees, fully mobilize the initiative of leading enterprises in key fields and industries, as well as various stakeholders—including complete robot manufacturers, application service providers, supply chain enterprises, and research institutes—in the humanoid robot and embodied intelligence sectors. They shall establish innovation application consortia and increase support for scenario opening, skills research, application promotion, and standard development through special funds, government incentives, and other measures.

Contact Information:

Ministry of Industry and Information Technology:  010 - 68205246

State-owned Assets Supervision and Administration Commission of the State Council:  010 - 63192338

General Office of the Ministry of Industry and Information Technology

General Office of the State-owned Assets Supervision and Administration Commission of the State Council

June 3, 2026

Original URL：https://www.miit.gov.cn/zwgk/zcwj/wjfb/tz/art/2026/art_f291ccd3da4c47ce95741de63cc088e6.html