Definition
Plain language
Extra camera information telling the robot how far away each point in the scene is.
As stated in the literature
Per-pixel range data (RGB-D or stereo-derived), used to lift 2D detections into 3D object poses for manipulation.
Also called: depth sensing, depth data
Why it matters: Grasping requires knowing exactly where something is in three dimensions, and a plain color image alone cannot say how far away anything is.
For example, a flat photo shows a plate and a fork overlapping, but depth tells the robot the fork is five centimeters closer, so it knows which one to grab first.
Heard on the show
“At k going to infinity — unlimited attempts — you're asking the cleanest possible question: can this model ever solve this problem at depth T?”Episode 011 — When RL Actually Teaches Agents Something New, And When It Doesn't