SLAM (simultaneous localization and mapping) is the computer-vision method that lets a device build a rough map of its surroundings while tracking its own position within it. In AR, SLAM is what keeps digital objects attached to floors, walls and tables as you move.
Also known as: simultaneous localization and mapping, visual-inertial odometry (VIO)

An AR sculpture at Denver Walls, 2024

Feature points in the real scene: grass, signs, trees and the building
How it works
The phone picks out visual feature points across camera frames and combines them with motion-sensor data to estimate movement, often called visual-inertial odometry. The move-your-phone-around step at the start of many AR experiences gives SLAM enough views to build its map.
How it works in illust
illust's World AR runs SLAM-based tracking in the phone's browser, so visitors don't install anything for content to lock onto the ground. Many illust scenes use the 8th Wall engine to do it.
Key facts
- ARCore uses SLAM to understand where the phone is relative to the world around it, detecting visually distinct feature points in the camera image and combining them with inertial measurements from the phone's IMU (Google ARCore documentation).
- Plain surfaces with no texture, like a white wall, can be hard for SLAM to detect (Google ARCore documentation).
FAQ
Why does AR ask me to move my phone first?
It is letting SLAM collect enough views of the scene to find surfaces and estimate distance. A few seconds of slow movement usually does it.
Does SLAM work at night?
Poorly. It needs visible detail, so lit areas work far better than dark ones.
Is SLAM the same as world tracking?
Not quite. SLAM is the method; world tracking is the AR feature built on it, which keeps content fixed in place as you move.