PlaNet learns a world model from image inputs only and successfully leverages it for planning in latent space.


Redirecting to planetrl.github.io, where the article resides.