AI News

Beijing releases an action plan for embodied intelligent robots, with a target of 200,000 units per year in complete-machine production capacity by 2028

ZhidongxiUpdated Nov 13 04:07

AI summary

On November 11, the Beijing Municipal Bureau of Economy and Information Technology released the "Beijing Action Plan for High-Quality Development of the Embodied Intelligent Robot Industry (2026-2028)," proposing that by 2028 it will cultivate no fewer than 5 world-class and 10 domestically leading humanoid robot complete-machine enterprises, with the Beijing-Tianjin-Hebei region accounting for 25% of the country's complete-machine production capacity and a supporting rate for key components exceeding 60%. The plan also proposes adding embodied intelligence dedicated computing power clusters with a total scale exceeding 50,000 PFlops, collecting more than 1 million hours of high-quality data in no fewer than 100 real-world scenarios, and forming a complete-machine production capacity of 200,000 units per year in Beijing.

Why read this

AI score 60

Beijing's three-year action plan for embodied intelligence sets specific quantitative targets for production capacity, computing power, and data, which can be used to assess the priorities of local industrial layout.

Topics

Tags

Body · AI translation

Zhidongxi, November 12 news: On November 11, the Beijing Municipal Bureau of Economy and Information Technology released the "Beijing Embodied Intelligent Robot Industry High-Quality Development Action Plan (2026-2028)" (hereinafter referred to as the "Action Plan").

▲ Announcement of the Release of the "Action Plan"

The "Action Plan" proposes that by 2028, Beijing will strive to cultivate no fewer than 5 world-class and 10 domestically leading humanoid robot complete-machine enterprises; the complete-machine production capacity of the Beijing-Tianjin-Hebei region will account for 25% of the country's total, and the supporting rate for key components will exceed 60%; it also proposes goals such as building a security governance system, striving for a world-class level.

The "Action Plan" will focus on implementing six major projects: embodied intelligence leapfrogging, Beijing-Tianjin-Hebei industrial chain coordination, technological innovation source cultivation, large-scale application promotion, industrial ecosystem leadership, and security governance taking the lead, covering more specific goals such as "adding a dedicated embodied intelligence computing cluster of more than 50,000 PFlops," "collecting more than 1 million hours of high-quality data in no fewer than 100 real scenarios," and "appropriately advancing layout to form an annual complete-machine production capacity of 200,000 units in Beijing."

In addition, the "Action Plan" also mentions that Beijing should formulate special support policies to increase funding support for key technology breakthroughs, innovative product research and development, pilot testing capacity building, and typical scenario applications; strengthen efforts to introduce high-level talent and teams, and provide support in areas such as settling in Beijing, children's education, housing, and medical care for urgently needed and scarce talent in the embodied intelligent robot field.

The full text of the "Action Plan" is as follows:

Beijing regards the embodied intelligent robot industry as an important field of high-precision and advanced industries and a key breakthrough point for future industries, accelerates the promotion of high-quality industrial development, and strives to cultivate no fewer than 5 world-class and 10 domestically leading humanoid robot complete-machine enterprises by 2028, with embodied intelligence models, robot bodies, and key technologies all reaching a world-class level; the complete-machine production capacity of the Beijing-Tianjin-Hebei region will account for 25% of the country's total, and the supporting rate for key components will exceed 60%; take the lead in building a scientific and effective security governance system covering the entire life cycle, with multi-department collaborative participation, and build the city into a world-class source of embodied intelligent robot technology and an industrial innovation highland.

I. Implement the Embodied Intelligence Leapfrogging Project

1. Substantially enhance the capabilities of embodied intelligence foundation models. Support the layout of multiple technical routes such as vision-language-action models (VLA), world models (WM), and reinforcement learning, as well as integrated technological development, leading technological innovation across major routes. Support breakthroughs in technologies such as multi-source information fusion, physical space understanding, and human emotion recognition, promoting a leap from "perception" to "cognition," with "brain" cognition reaching the level of understanding constraints of physical laws and proactively recognizing non-verbal instructions such as human actions and expressions; support breakthroughs in technologies such as real-time reasoning, cross-embodiment transfer, and self-learning evolution, substantially enhancing "cerebellum" capabilities in multi-task generalization, highly reliable execution, fast-cycle operations, and ultra-precision manipulation; support breakthroughs in technologies such as swarm collaboration, long-term memory, localized lightweight deployment, and model interpretability, significantly enhancing the real-world application deployment capabilities of embodied intelligence models, with embodied intelligence foundation models as a whole reaching a world-class level.

2. Enhance the supply level of high-quality embodied data. Support innovation in data collection hardware and collection schemes such as bodyless collection and "first-person perspective collection," accelerating cost reduction and efficiency improvement. Support breakthroughs in technologies such as high-fidelity simulation data generation, robot autonomous exploration data collection, human action mapping, and unified representation of multimodal data, and actively explore the formation of a high-quality data supply paradigm that integrates multi-source data and covers the entire process of model training and application deployment. Take the lead in building a national-level embodied intelligent robot data infrastructure covering diversified real scenarios, collecting more than 1 million hours of high-quality data in no fewer than 100 real scenarios.

3. Increase guarantees for dedicated computing power resources. Add and build a dedicated embodied intelligence computing cluster with a total scale of more than 50,000 PFlops, further strengthening the guarantee of high-quality embodied intelligence computing power. Promote key embodied intelligence enterprises and the construction units of key computing power node projects to form computing power supply-demand coordination through a "pairing" approach. Through the "Galaxy Computing Corridor" project, carry out computing power collaboration with cities outside Beijing that gather computing power, and connect embodied intelligence enterprises with high-quality domestic computing power resources.

4. Build an authoritative evaluation system for embodied intelligence. Build the world's first embodied intelligence model testing ground based on real application scenarios and providing diversified bodies, establish and improve an embodied intelligence model evaluation system in which "user units set the questions and testing entities answer them," and rely on professional institutions and relevant central units to build and operate a neutral and authoritative normalized evaluation and certification platform, continuously iterating evaluation standards and indicator systems. Based on the embodied intelligence testing ground, organize real-scenario task challenges for embodied intelligence model R&D teams around the world, and create a globally renowned embodied intelligence model ranking list.

II. Implement the Beijing-Tianjin-Hebei Industrial Chain Coordination Project

5. Create a world-class body platform. Organize collaborative innovation among complete-machine and component enterprises, continuously improve hardware architecture, and systematically enhance comprehensive performance. Strive for the rated load of complete machines to reach more than 40% of their own weight, average mean time between failures to exceed 2,000 hours, the annual failure rate of core moving components and end effectors to be below 10%, the design service life of complete machines to exceed 5 years, and battery endurance under continuous operation conditions to exceed 4 hours. Create 3-5 world-class complete-machine platforms that meet the requirements for large-scale deployment in diversified scenarios, are serialized and modularized, and can be software-defined products.

6. Accelerate the formation of comparative advantages in key components. Focus on high-tech, high-value-added key components, and support research on high-efficiency, high-reliability motors with a torque density of 60 N·m/kg; high-precision reducers with a rated life of more than 15,000 hours; energy-efficient integrated joint modules with a torque density of 100 N·m/kg and transmission efficiency of more than 85%; six-axis force sensors with a full-scale linear error of no more than 0.1% and dynamic response better than 5 kHz; tactile sensors with a force resolution of ≤0.01 N and spatial resolution of ≤1 mm; highly reliable, high-dexterity dexterous hands with a rated load life of 1 million cycles, fingertip force control resolution of 0.1 N, and a unit cost of less than RMB 10,000; high-safety batteries with an energy density of 350 Wh/kg and a cycle life of 1,500 cycles; and highly humanlike bionic faces with an interaction response latency of less than 50 milliseconds and the ability to reproduce more than 200 expressions, among other key components, so that they generally reach the international first-class level.

7. Solidly advance collaborative innovation in R&D, manufacturing, and application. Promote the empowerment of technological achievements in the field of embodied intelligence in Beijing to drive product innovation in industrial robots, service robots, and special robots in the Beijing-Tianjin-Hebei region. Support a combination of contract manufacturing and self-built production lines, with appropriately advanced layout, to form a complete-machine production capacity of 200,000 units per year in Beijing. Build no fewer than 5 embodied intelligent robot manufacturing service platforms, forming pilot-scale verification service capabilities covering software and hardware functions, performance, reliability, safety, environmental adaptability, and other aspects. Relying on the relevant complete-machine pilot production and verification capabilities, realize an innovation loop of coordinated R&D and manufacturing, with a Beijing-Tianjin-Hebei supporting rate of more than 60%, supporting rapid iterative upgrading of body products. Actively introduce cross-industry supply chain enterprises such as intelligent connected vehicles, and support the construction of no fewer than 10 shared processing platforms in the Beijing-Tianjin-Hebei region with rapid response service capabilities in structural part processing, surface treatment and heat treatment, SMT mounting, wiring harness manufacturing, and other links, supporting the shortening of sample delivery time after component design changes to 1-2 days. Encourage Beijing embodied intelligent robot enterprises to collaborate with application scenario units in Tianjin and Hebei to carry out application innovation. Around vehicle-machine integration and electromechanical integration, build a number of R&D and manufacturing parks within the Beijing-Tianjin-Hebei region to promote clustered industrial development.

III. Implement the Technological Innovation Source Project

8. Make advanced arrangements for frontier technologies. Support research on frontier technologies for embodied intelligence and complete machines, such as the underlying architecture and basic algorithms of embodied intelligence models, and bionic structures and motion control mechanisms. Support research on frontier technologies for key components such as braided motors based on functional fibers and "braided material manufacturing" processes, integrated joints and dexterous hands based on bionic mechanisms, large-area tactile sensing electronic skin, and ultra-low-power intelligent computing chips.

9. Concentrate efforts on tackling key core technologies. Support research on core components such as high-end flexible tactile sensors, micro-precision reducers, and micro coreless motors, whose supply faces the risk of being "controlled by others," as well as key chips such as RISC-V-based system-level intelligent computing chips and high-dynamic-load N-channel MOSFETs, and promote the principle of "deploying all that should be deployed" in tackling key chips and core components. Support complete-machine units in adapting and applying relevant components and chips, and promote the principle of "adapting all that can be adapted" for relevant independently controllable products. Support research on basic software such as simulation and intelligent training platforms, simulation toolchains, robot operating systems and middleware compatible with ROS/ROS2, development toolchains, and resource scheduling platforms, and continuously improve the application ecosystem for independently controllable software.

IV. Implement the Large-Scale Application Promotion Project

10. Continue to polish the robot "two conferences and one competition" brand. Build a humanoid robot competition and training base, continuously optimize the competition events of the games according to the actual needs of application scenarios, and create humanoid robot games and half-marathon races that fully reflect and represent the highest level of development of China's humanoid robot industry. Relying on international exchange and cooperation platforms such as the World Robot Conference, support close collaboration with the Chinese Institute of Electronics, the World Robot Cooperation Organization, and others, organize international humanoid robot academic conferences at a high level, and build a global highland for industrial and academic exchange in the robot field.

11. Deeply advance real-scenario practical training to break through benchmark scenarios. Take the lead in exploring the construction of application innovation platforms in vertical segmented fields to promote real-scenario practical training in related fields. In the industrial field, focus on scenarios such as advanced manufacturing and logistics warehousing, make key breakthroughs in capabilities such as human-machine collaboration, tool use, and precise operation, gradually become competent for tasks such as handling and transfer, inspection and sorting, plug-in assembly, and auxiliary processing, and form no fewer than 20 benchmark application scenarios in the industrial field. In the service field, focus on scenarios such as elderly care services, medical health, public services, culture, tourism, science and education, and supermarket retail, make key breakthroughs in capabilities such as task planning, safe interaction, and flexible operation, gradually become competent for tasks such as consultation and guidance, unmanned retail, security inspection, and caregiving assistance, and form no fewer than 20 benchmark application scenarios in the service field. In the special field, focus on scenarios such as work safety, emergency rescue, and urban security, make key breakthroughs in capabilities such as environmental adaptability and safe disposal, gradually become competent for tasks such as inspection and troubleshooting, security duty, and rescue and disposal, and form no fewer than 10 benchmark application scenarios in the special field.

12. Accelerate the formation of solutions and promote large-scale application. Cultivate embodied intelligent robot solution providers and application innovation platforms, and distill and form no fewer than 15 solutions that can be promoted at scale in fields such as production operations, commercial services, family life, and elderly care services. Innovatively build a comprehensive evaluation system covering multiple dimensions such as task execution efficiency and input-output returns, as well as a feedback mechanism for collecting data under real application conditions, and innovatively use financial and insurance tools such as financial leasing and application insurance; regularly select and publish typical products and typical case scenarios of embodied intelligent robots, organize complete machine manufacturers and solution providers to conduct extensive matchmaking with central state-owned enterprises, universities, and other enterprises and institutions that possess potential large-scale application scenarios, and accelerate the promotion, application, and continuous iterative upgrading of solutions.

13. Explore the construction of regional integrated application scenarios. Support the construction of the Robot Innovation Park in the Olympic Central Area, and create robot integrated application scenarios with market-oriented operation capabilities and scalability for promotion, such as theme restaurants, hotels, and cultural tourism attractions covering robots for food preparation and sales, guided tours and shopping guidance, entertainment performances, and cleaning services. Support the iterative expansion of the functions of the Robot World, and build the Beijing Economic-Technological Development Area into a concentrated demonstration zone for the application of embodied intelligent robots in real scenarios across the entire area.

V. Implement the Industrial Ecosystem Leadership Project

14. Further strengthen innovation entities. Establish a "one-on-one" service mechanism for the listing of embodied intelligent robot enterprises, and cultivate advantageous leading enterprises. In key fields such as complete machines, key components, embodied intelligence models, and data collection, cultivate no fewer than 200 technology-based and innovative small and medium-sized enterprises, high-tech enterprises, specialized, refined, distinctive and innovative small and medium-sized enterprises, specialized, refined, distinctive and innovative "little giant" enterprises, and manufacturing single-champion enterprises. Give full play to the role of industry organizations such as the Beijing Robot Industry Association, strengthen industry self-discipline, oppose "involution-style" competition, and guide the industry to develop in a healthy and orderly manner.

15. Continuously improve the level of open source and openness. Further improve the quality of embodied intelligent robot bodies, embodied intelligence model open source platforms, and open source datasets. Build a globally influential embodied intelligent robot open source community, and explore the establishment of new service platforms such as an embodied intelligent robot application software store. Build a new business format based on general software and hardware platforms and open source datasets, with secondary development for vertical scenarios, and form a domestically leading capability for secondary development of software and hardware. Promote the establishment of 100 robot research teams in universities and research institutes in Beijing, support the relevant teams in cooperating with advantageous enterprises and innovation platforms in the city, and produce high-quality research results based on open source software and hardware platforms for humanoid robots.

16. Actively lead the construction of the standards system. Support the work of organizations such as the Humanoid Robot Subcommittee of the National Robot Standardization Technical Committee (SAC/TC591/SC5), the Humanoid Robot and Embodied Intelligence Standardization Technical Committee (MIIT/TC8), and the Beijing Robot Industry Association. Prepare to establish a municipal-level standardization technical committee related to robots, carry out research on the standards system, and hold standards supply-demand matchmaking activities in the robot field. Encourage relevant units in the city to lead and participate in the drafting of no fewer than 8 international standards, 30 national standards, 70 industry standards, and association standards, with a participation rate of over 90% in national standards. Actively recommend relevant units in the city to carry out national standardization pilot construction related to robots.

17. Support enterprises in going global and participating in global exchange and cooperation. Support qualified key embodied intelligent robot enterprises in the city to "go global," and encourage advantageous foreign embodied intelligent robot enterprises and institutions to establish R&D centers and manufacturing bases. Relying on international exchange and cooperation platforms such as the World Robot Conference, support the Chinese Institute of Electronics in close collaboration with the United Nations Industrial Development Organization, the World Robot Cooperation Organization, and others, to organize high-level international embodied intelligent robot industry exchange activities and academic conferences, and build a highland for international exchange and cooperation in the industry and academic fields.

VI. Implement the Safety Governance Pioneer Project

18. Advance the layout of safety technology innovation. Focusing on the needs of human-machine collaborative production and living scenarios, take the lead in laying out research on intrinsic safety technologies for complete machines such as hardware overload protection, energy and power distribution safety, and active emergency braking; research on embodied intelligence model safety technologies such as decision safety audit and traceability, malicious attack protection, sensitive privacy data handling, and ethical constraints on human-machine interaction; and research on safety governance technologies such as identity recognition, electronic fences, and digital twin traceability.

19. Take the lead in building a safety governance system. Build third-party institutions such as an embodied intelligent robot testing and certification center and an embodied intelligence quality inspection and testing center, and build a testing and certification platform covering full-dimensional testing capabilities including function, performance, safety, reliability, and intelligence level in aspects such as complete machines, components, models, and data. Build a "one machine, one code" whole-process management system for embodied intelligent robots, and take the lead in forming safety governance capabilities such as identity authentication, trusted traceability, and behavior auditing covering the entire cycle of R&D, production, scrapping, and recycling.

VII. Safeguard Measures

Improve the working mechanism for promoting the embodied intelligent robot industry, and strengthen overall coordination among various departments and districts. Formulate special support policies, and increase financial support for key technology breakthroughs, innovative product R&D, pilot testing capacity building, and typical scenario applications. Further give play to the leading and driving role of government guidance funds, and guide market-based funds to support industrial development. Encourage more universities to offer majors related to embodied intelligent robots, deepen the talent cultivation model integrating industry and education, and cultivate a group of professionals urgently needed for industrial development, such as data collection and secondary development, as well as compound talents with refined technical skills and product understanding. Strengthen the introduction of high-level talents and teams, and provide support in areas such as household registration in Beijing, children's education, housing, and medical care for urgently needed and scarce talents in the embodied intelligent robot field.

This plan shall be implemented from the date of issuance and shall remain valid until December 31, 2028.

Source

Zhidongxi