Anchor of big objects in open area

ARReferenceImages need to exhibit a certain amount of visual features and texture in order to be recognized reliably. How do your reference images look like? If they show for example just a portion of the ground, they would likely not detect very well.

When you say that content moves after initial detection, this could either be due to the reference image being out of sight as you move further away, or due to degraded world tracking in general.

To address the first issue, you could try to just detect and not track the image (set maximumNumberOfTrackedImages to 0). For the second, watch out for a limited tracking state in ARCamera.TrackingState of your ARFrames.

I used a minimum of code. All the logic was put into the .pсproject file. Now I think it works on a small scale (not for placing large objects on the street). Now I'm trying to implement this using ARWorldMap and SceneKit. Am I on the right track?

I would recommend using ArcGIS runtime sdk for iOS. You can create a full-scale AR map to overlay over the camera (but it can be invisible). Then you can use real world coordinates to pin objects to the 1-to-1 scale map. The way it works is that the user's real world location is transferred to the user's digital world location, then the runtime moves the map and everything pinned to it around so that the user's real world movement is matched in the digital world's map. IMO, This is the best way around not being in the ARGeoTracking areas.