AR Object Placement: From Raycast to Stabilization

Placing 3D Objects in AR: From Raycast to Stabilization

We at TrueTech regularly solve this for clients in furniture and retail. Getting a 3D model to sit stably on a detected plane is not a one-evening job. The object must stay level as the camera moves, not "float" under changing lighting, scale appropriately, and not sink into the surface. We offer a turnkey implementation: from model format selection to final app integration. In one project for a furniture retailer, we faced the issue of the object "sinking" into the plane during fast camera movement. The fix was to use ARAnchor with a 2mm dynamic offset along the normal — this eliminated Z-fighting under any lighting.

3D Model Formats and Optimization for AR

The standard for mobile AR is USDZ (iOS/RealityKit) and GLB/GLTF (ARCore/cross-platform). A common mistake: taking an unoptimized model and trying to load it into AR. 500k triangles, 4096×4096 textures without mip-mapping guarantees FPS drops on devices below A13.

Target parameters for mobile AR:

Compression: KTX2 + Basis Universal for GLB, HEIC textures in USDZ. In RealityKit — Reality Composer Pro (Xcode 15 or later) for baking PBR materials directly into .reality format. For Android, use Android Studio with ARCore SDK 1.30+ for GLB conversion.

Why Raycast Over hitTest for Placement?

With ARKit 4+ and ARCore 1.18+, the preferred method for determining placement point is Raycast. It returns ARRaycastResult with target: .existingPlaneGeometry or .estimatedPlane. existingPlaneGeometry uses already detected geometry — accuracy is 2× higher than the deprecated hitTest. Apple recommends raycast as the primary method (see ARKit documentation). Moreover, hitTest was deprecated in iOS 14, and its use may lead to app rejection due to App Store Review Guidelines (section 4.2: minimum functionality).

Example in Swift:

let query = arView.makeRaycastQuery( from: arView.center, allowing: .existingPlaneGeometry, alignment: .horizontal ) let results = arView.session.raycast(query) if let first = results.first { placeEntity(at: first.worldTransform) } 

For ARCore — Session.createRaycastQuery() + Frame.raycast() gives more stable results at plane edges.

Stabilizing the Object After Placement

Once placed, the object should not "wander" as the camera moves. The standard approach is to attach to AnchorEntity via ARAnchor:

let anchor = ARAnchor(transform: worldTransform) arView.session.add(anchor: anchor) let anchorEntity = AnchorEntity(anchor: anchor) anchorEntity.addChild(modelEntity) arView.scene.addAnchor(anchorEntity) 

Explicit ARAnchor locks the position, and the object does not update as the plane refines. Without this, the object shifts for the first 10–15 seconds of the session while ARKit refines geometry. Optionally, apply Kalman filtering to the anchor position if jitter persists.

Shadows and Physical Lighting

An AR object without a shadow looks like it's floating. In RealityKit, a shadow is created by setting castsShadow = true on the entity — contact shadow renders automatically. In SceneKit, you need a directional light with castsShadow = true. Importantly, when using environment texturing (ARKit, A12+), objects receive reflections of the real environment. Metallic surfaces reflect the room — without this, a chrome object looks plastic. Enable environmentTexturing = .automatic in ARWorldTrackingConfiguration.

Supported Gestures for Object Manipulation

After placement, the user should be able to translate, rotate, and scale the object. In RealityKit, gestures are added in one line:

arView.installGestures([.translation, .rotation, .scale], for: entity) 

For custom control, use UILongPressGestureRecognizer with continuous raycast — this gives more control over sensitivity. When implementing rotation, constrain to the Y axis: use simd_quatf(angle:axis:) only around the vertical axis, otherwise the object tilts with imprecise gestures. On Android, similarly use GestureRecognizer from Sceneform or ArGestureRecognizer in ARCore.

Testing AR Placement on Different Devices

It is critical to test on multiple generations of iPhones and Android devices. For example, on iPhone XR (A12) raycast works stably, but environment texturing consumes additional memory — FPS may drop with high-polygon models. We recommend using TestFlight for iOS and internal testing via Google Play for Android — this helps identify depth issues (Z-fighting) on different ARCore versions.

What's Included in the Work

We have completed over 20 successful projects in mobile AR development. Contact us for a project assessment — we deliver turnkey solutions in 2–3 weeks. Get a no-obligation consultation.