AR Occlusion & Spatial Anchoring Simulator

AR Scene

Room complexity 5 obstacles

Occlusion-aware AR

Camera orbit speed 0.6

Tracking noise (drift) 15%

Live Readouts

Visible targets

0 / 0

Occluded targets

Tracking confidence

100%

How it works

Real AR SDKs (ARKit / ARCore) scan the room into a depth mesh, then before drawing a virtual object they compare its distance from the camera to the depth-buffer value along the same ray. If real geometry is closer, the virtual pixel is discarded — the object appears to sit correctly behind real furniture instead of floating on top of everything.

for each virtual target T:
  ray = normalize(T.position - camera.position)
  d_target = |T.position - camera.position|
  hit = raycast(camera.position, ray, maxDist = d_target)
  occluded = (hit !== null)   // real mesh is closer than the virtual object

This simulator anchors colored spheres (virtual game targets) at fixed world positions inside a reconstructed room. Toggle Occlusion-aware AR off to see the "naive AR" bug — targets always render on top of the video feed, even through walls. With it on, a THREE.Raycaster tests each target against the wireframe occluder meshes every frame and hides any target whose ray is blocked.

Room complexity — number of reconstructed furniture/wall blocks; more geometry means more chances for a target to be occluded from any given viewpoint.
Camera orbit speed — simulates the player walking around the room; you can also drag to look around manually.
Tracking noise — real AR trackers drift slightly frame to frame from feature-matching error; this jitters the camera position and lowers the reported tracking confidence, which is what causes anchored objects to "swim" in a real headset or phone.

Virtual game target (visible) Reconstructed real-world geometry

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.