HomeArticle

The Ministry of Industry and Information Technology has issued five industry standards in the field of humanoid robots — Interpretation of intelligent capability classification and training site construction standards

未来一氪2026-11-19 18:59
The Ministry of Industry and Information Technology has issued five standards in the field of humanoid robots, which take effect immediately.

On November 16, 2026, the Ministry of Industry and Information Technology released Announcement No. 30 of 2026, officially issuing and implementing five standards in the field of humanoid robots under the centralized administration of the Humanoid Robot and Embodied Intelligence Standardization Technical Committee of the Ministry of Industry and Information Technology (MIIT/TC8). This batch of standards covers three major directions: intelligent capability evaluation, key components and training of humanoid robots, providing directional guidance for the R&D, testing and application of humanoid robots.

The five standards in this batch include Humanoid Robot - Technical Specification for Training Site Construction (SJ/T 12353—2026), Humanoid Robot - Management Requirements for Training Data (SJ/T 12354—2026), Humanoid Robot - Classification of Intelligent Capability (SJ/T 12355—2026), Humanoid Robot - Technical Specification for Integrated Electric Drive Joint (SJ/T 12356—2026) and Humanoid Robot and Embodied Intelligence - Technical Specification for Dexterous Hand (SJ/T 12357—2026), all of which shall come into force as of the date of issuance.

Among them, the two standards for intelligent capability classification and training site construction are particularly critical: one solves the evaluation problem of "how intelligent the robot is", and the other solves the infrastructure problem of "how to standardize the training site".

I. Humanoid Robot - Classification of Intelligent Capability: A Measuring Ruler for "Intelligence Level"

Humanoid robots are accelerating their march from laboratories to thousands of industries such as industrial manufacturing, home services, medical care and elderly care, public services and special applications. However, for a long time, the industry has lacked a unified evaluation benchmark for intelligent capabilities - different products have different claims on their "intelligence level", users "dare not and are unwilling to use" the products, and it is difficult to make horizontal comparison between products, which restricts large-scale application and industrial collaboration.

Centering on the core capability closed loop of humanoid robots to complete tasks, this standard systematically breaks down the intelligent capability into four core dimensions:

· Perception and Cognition: The ability of robots to "understand" the environment;

· Decision-making and Learning: The ability of robots to "think clearly" and plan tasks independently;

· Execution Performance: The operation ability of robots to "operate precisely";

· Collaboration and Interaction: The collaborative ability of robots to "cooperate seamlessly" with humans.

On the basis of the four dimensions, with "intelligent autonomy" as the main line, the standard divides the intelligent capability of humanoid robots into six levels from L0 to L5, which progresses step by step from "preset execution" to "high autonomy and self-evolution capability", forming a clear capability evolution path.

In terms of evaluation methods, the standard innovatively adopts a three-layer expression mode: using capability profile (radar chart) to visually present the strengths and weaknesses of the four dimensions; using basic level for macro classification and safety bottom line guarantee; using comprehensive level to reflect the overall balance and support horizontal comparison of products. This method turns the abstract "intelligence level" into a measurable and comparable common language, providing a unified reference for product R&D, testing and evaluation, application selection and scenario adaptation.

II. Humanoid Robot - Technical Specification for Training Site Construction: Making Training Site Construction "Follow Clear Standards"

The training site is a special physical environment built to verify, optimize and improve the comprehensive capabilities of humanoid robots, and it is a key infrastructure connecting algorithm R&D and real application. At present, a large number of humanoid robot data collection and training centers have been put into use or are under planning and construction across the country, but they are generally faced with problems such as "fragmented construction, incompatible equipment interfaces, insufficient scenario reproduction capability, and lack of safety system" - repeated construction, incomparable tests and long technology transformation cycle have become common pain points restricting industrial development.

This standard establishes four principles for training site construction - demand orientation, diversity, intelligence and safety, and systematically stipulates the core elements of the training site:

· Functional Zoning: Scenario training area, data center, maintenance area, storage area, office area, and support virtual-real integrated test area, to meet the full-process closed loop of data collection, model training, simulation verification and real machine verification;

· Equipment and Facilities: Configuration requirements for key equipment such as motion capture, teleoperation, data storage and computing platform;

· Safety Protection: Build a three-level protection system of physical safety (collision prevention, emergency braking), data safety (privacy desensitization, encrypted transmission), and ethical safety (prohibition of non-compliant scenarios), to fully reduce the risks of personnel, robots and the environment during the training process;

· Network Facilities: High-performance and low-latency communication support to support large-scale collaborative training of robots.

This is the first dedicated standard for humanoid robot training sites, which systematically covers the full chain of "construction principles - technical requirements - test methods", fills the gap in training site standards, will provide a unified technical basis for the construction of humanoid robot training sites, reduce repeated construction and lower trial-and-error costs.

III. Three Humanoid Robot Standards: Consolidating Key Components and Data Base

Humanoid Robot - Technical Specification for Integrated Electric Drive Joint focuses on the integrated electric drive joint of humanoid robots, unifies interfaces, performance indicators and test methods, establishes comparable and verifiable product evaluation basis for enterprises and testing institutions, and supports the development of integrated joints towards high performance, high precision, high integration and high power density.

Humanoid Robot and Embodied Intelligence - Technical Specification for Dexterous Hand establishes a technical specification system for the design, testing, evaluation and delivery of dexterous hands, improves the problems of non-unified interfaces and low complete machine integration efficiency, and supports the application and promotion of dexterous hands in scenarios such as intelligent manufacturing, home services, medical assistance, education and scientific research.

Humanoid Robot - Management Requirements for Training Data covers the full life cycle management links of training data, including collection, preprocessing, annotation, quality evaluation, storage, transmission, circulation, use, safety compliance, destruction and continuous improvement, promotes the gradual standardization of training data management, annotation and quality evaluation requirements, strengthens data security, privacy protection and process traceability, and provides basic support for the construction of high-quality training data sets and industrial collaboration.

The issuance and implementation of the five standards marks that China's humanoid robot standardization work has further moved from single technology exploration to a systematic layout covering "capability - component - data - application", which helps to improve the collaboration efficiency of the industrial chain, promote the transformation of technological achievements, and drive the more standardized, healthy and orderly development of the humanoid robot industry.

In the next step, the Humanoid Robot and Embodied Intelligence Standardization Technical Committee of the Ministry of Industry and Information Technology will continue to do a good job in standard publicity and implementation, application and effect evaluation, promote the connection between standards and testing certification, training platform, data ecology construction and typical application scenarios, and further improve the standard system in the field of humanoid robots and embodied intelligence. We sincerely invite complete robot manufacturers, core component suppliers, application service providers, universities, scientific research institutes, social organizations and relevant experts and scholars to participate together, and jointly promote the development of the humanoid robot and embodied intelligence industry.