Embodiment
Physical AI
The physical body, sensors and controls through which an agent perceives and acts.
Robots can differ in reach, joint layout, grippers, cameras and action conventions. Data from multiple embodiments can support shared learning, but its context needs to remain identifiable. A demonstration from one robot is not automatically an executable action sequence for another.
In practice
Two arms perform the same insertion task, but their joint targets and camera views require different interpretation.