AI News

Beijing Issues Embodied Intelligence Robot Action Plan: Substantially Enhance Embodied Intelligence Foundation Model Capabilities

YicaiUpdated Nov 11 13:50

AI summary

The "Beijing Embodied Intelligence Robot Industry High-Quality Development Action Plan (2026–2028)" proposes to substantially enhance embodied intelligence foundation model capabilities and supports the layout and technological integration of multiple technical routes such as vision-language-action models (VLA), world models (WM), and reinforcement learning. The plan requires breakthroughs in technologies including multi-source information fusion, physical space understanding, human emotion recognition, real-time reasoning, cross-embodiment transfer, self-learning evolution, swarm collaboration, long-term memory, and localized lightweight deployment, so that embodied intelligence foundation models as a whole reach the international first-class level.

Body · AI translation

According to Beijing Economic and Information Technology news, the "Beijing Embodied Intelligent Robot Industry High-Quality Development Action Plan (2026–2028)" has been issued. The plan proposes to substantially enhance the capabilities of embodied intelligence foundation models. It supports the layout and integrated development of multiple technical routes such as vision-language-action models (VLA), world models (WM), and reinforcement learning, leading technological innovation across major routes. It supports tackling key 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; it supports tackling key 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-paced operations, and ultra-precision manipulation; it supports tackling key technologies such as swarm collaboration, long-term memory, localized lightweight deployment, and model interpretability, significantly improving the real-world application deployment capabilities of embodied intelligence models, with embodied intelligence foundation models as a whole reaching the international first-class level.

Source

Yicai