What is Open X-Embodiment / RT-X?

Robot learning is a research field where robots acquire new skills, or adapt to their surroundings, through machine learning. A core problem is that robots are usually specialists: a model is typically trained separately for each task, robot and environment, and changing one thing can mean starting over. Open X-Embodiment is a shared collection of data from many different kinds of robots, gathered in one common format. RT-X is the name given to robot-control models trained on this data, such as RT-1-X, which its makers say shows skills transfer across many robot types. Google DeepMind and partners from 33 academic labs pooled data from 22 robot types for the dataset.

Written from primary sources and reference pages, and checked against them by an independent AI fact-checker. Drafted with AI assistance.