World-Task Factorization

Robot learning framework separates world and task for better generalization

Researchers have introduced a new framework for robot learning that separates the 'world' from the 'task' to improve generalization. This approach formalizes the asymmetry between environmental properties and task logic…