cross-embodiment

Cross-embodiment refers to using or transferring a single model across robots whose form and joint layout differ, such as arms, wheeled robots and legged robots. The aim is to train on data from many robots so that adapting to a new robot requires less data.

Software AI models run the same on different computers, but robot control models often fail when the body or sensors change, which makes cross-embodiment a major research problem in robot AI.

This entry is based on AIPOST articles and widely known facts. If something is wrong, please send us a correction request.