The world's first Humanoid Robot Intelligence Grading standard was recently released by the China Institute of Electronics. Led by the Beijing Humanoid Robot Innovation Center located in the Beijing Economic-Technological Development Area (Beijing E-Town), this standard establishes a "four-dimensional, five-level" evaluation framework, provides a unified technical language and evaluation system for the development, testing, management, and application scenarios of humanoid robot intelligence technology and products, as well as aims to help the industry transition from a "function-oriented" approach to "intelligent evolution".

The "four-dimensional, five-level" evaluation framework innovatively combines the grading logic from fields such as autonomous driving and industrial robots, taking into account the uniqueness of humanoid robots. It forms a four-dimensional core of capabilities with "perception and cognition (P), decision-making and learning (D), execution and performance (E), collaboration and interaction (C)", and constructs an L1-L5 five-level intelligence grading system for humanoid robots.

Specifically, perception and cognition refer to the ability of humanoid robots to acquire, process, understand environmental and self-status information, and to engage in reasoning, judgment, knowledge construction, and application.

Decision-making and learning capability entail humanoid robots being able to achieve precise perception, logical reasoning, task planning, and dynamic decision-making in various environments and tasks through methods such as large-scale models, reinforcement learning, and imitation learning.

Execution and performance capability involve supporting precise control of joint motion and end-effector operations, as well as possessing comprehensive abilities for body movement, dynamic balance, positioning navigation, executing complex actions, and tasks.

Collaboration and interaction capability mean that humanoid robots can engage in safe, natural, and efficient communication and collaboration with humans, environments, other robots, or other intelligent entities.

From L1 to L5, the level of intelligence of humanoid robots increases progressively.

Notably, this standard also provides 22 primary indicators, over 100 technical terms, a general safety baseline, and mappings of typical application scenarios, which can serve as intuitive references for enterprises to carry out product design, performance benchmarking, and capability declarations. With the implementation of the grading standard, humanoid robots are expected to break through "display intelligence" and move towards true "general intelligent", achieving large-scale applications in various fields such as special operations, logistics handling, industrial manufacturing, education research, commercial services, and health care for the elderly.

This standard provides the industry with a scientific and unified set of technical language and evaluation tools, which not only helps standardize market order and accurately guide technological research and development directions but also accelerates the application of humanoid robots in various fields, breaking down barriers between technology and scenarios, and promoting the orderly development of the entire industry ecosystem.

In the future, as the standard setter and ecosystem builder in the field of humanoid robots, Beijing Humanoid Robot Innovation Center will continue to promote the deep integration of technology, products, and verification standards with the industry, accelerating the empowerment of humanoid robots across various industries and entering countless households.