We present Visual Navigation from Goals, or ViNG, an autonomous, self-supervised navigation platform for reaching visually indicated goals in open-world environments that can learn navigational skills entirely in the real-world, without access to any spatial information.
"ViNG: Learning Open-World Navigation with Visual Goals"
Dhruv Shah, Benjamin Eysenbach, Gregory Kahn, Nicholas Rhinehart, Sergey Levine
Berkeley AI Research (UC Berkeley) and Carnegie Mellon University
