Definition
Embodied AI refers to AI agents that perceive, act and learn through a physical body interacting directly with their environment — the idea that intelligence emerges from a perceive-decide-act loop rather than from disembodied computation alone.
How it works
A typical embodied system closes the loop between sensing (vision, touch and other modalities), decision-making (often involving large multimodal models) and execution (robot arms, legged platforms, dexterous hands), trained and deployed in simulation or the real world.
Usage
Industrial and household robotics and warehouse automation are the main application areas; a large body of research trains skills in simulation first and then transfers them to real robots.
Related and confusable terms
Contrast with computer use: computer use gives a model a “body” in digital interfaces, while embodied AI centers on physical embodiment; both share the perceive-decide-act loop.