Summary of "Building a Spatial Experience - Dent Reality @ NSLondon"

Speaker / context

Andrew, CEO of Dent Reality — ex-iOS developer and early ARKit adopter.
Open-sourced a popular ARKit + CoreLocation library.
Talk/demo focused on building practical AR navigation experiences for retail and indoor spaces.

AR platforms
- ARKit / RealityKit vs ARSCNView (SceneKit). Recommendation: use SceneKit/ARSCNView today for richer features, since RealityKit lacks many high-level features.
LiDAR
- Device depth sensors speed up plane detection when available, but design choices can remove the need for ground-plane scanning entirely.
Indoor location
- Sub-meter accuracy is possible but noisy. UX must be built around location limitations (stabilization/smoothing) to avoid jittery AR content.
Spatial coordinate system
- ARKit uses 3D world coordinates. AR content must be placed and oriented in world space, while some elements should behave like screen-space UI.

Two primary interaction modes: map view (2D) and AR overlay — users can switch between them.
Compass view: large on-screen compass that lifts into an AR element when the phone is raised.
Floating directional arrow: big 3D arrow positioned in world space (e.g., ~15 m away) as a primary navigation beacon.
Screen-scale text labels: anchored in world space but scaled for constant on-screen readability.
Updater system: centralized per-frame update manager to keep AR elements synchronized to the phone (e.g., maintain same Y-height).
Fade distance: dynamic alpha fade so objects fade out as the user approaches (e.g., transparent at 1 m, opaque at 2 m) to avoid awkward “walking through.”
Screen-edge indicator (HUD-like): shows direction to the destination when it’s off-screen (inspired by video game HUDs).
Guidance markers / “stars” (dots): floating guidance points oriented between user and destination, positioned to avoid moving under the user when sidestepping.
- Guidance dots tilt/pitch down ~3° and slowly pan up as the user approaches so end labels aren’t obscured.
Minimize in-AR text: keep AR view lightweight; present dense information in screen-space overlays (regular UI) instead of floating AR text.
Debugger / scene inspector: in-app scene view for debugging — view world anchors, arrow positions, camera pose.

Take cues from video games for navigation affordances (edge indicators, fade, stabilization).
Build experiences to match the constraints of the underlying tech — mask and stabilize localization jitter instead of exposing it.
Prefer floating elements (air-based navigation) to avoid mandatory ground plane scanning and improve first-run experience.
Keep AR interactions lightweight and immediately discoverable so nontechnical users can pick them up quickly.

Open-source libraries: Dent Reality built and used open-source components (AR + mapping overlays, path-on-ground libraries). The speaker’s original AR + CoreLocation project became a popular ARKit project.
Path-on-ground library exists, but ground paths can feel wrong when users sidestep; floating guidance often works better.
Updater architecture: use a centralized per-frame updater for consistent element behavior.
Anchoring text to camera: anchor some elements to the camera or use camera-facing constraints to maintain readability and orientation.
Wireless debugging for AR builds remains unreliable; long Lightning cables are still commonly needed for on-device debugging.

Indoor localization accuracy and smoothing: making AR feel stable despite meter-scale noise.
UX trade-offs between world-scale objects and screen-space readability.
Debugging and developer tooling limitations (wireless debugging still problematic).
Balancing polish/animation versus functional clarity — some fancy treatments add little to usability.

Traveling salesperson problem for shopping lists: optimizing route/order for multi-item trips in stores (combinatorial optimization).
Potential UX: animated transitions connecting compass UI to the floating in-world arrow (fly-out / fly-back).
Wearables: interest in wearables, but many AR experiences will likely remain better on handhelds.

Tried delayed-follow behavior for guidance dots; delay added a “catching up” feel and was not helpful.
Animations and visual polish were considered but kept mostly functional based on constraints and user testing.
Wireless debugging remains promised but unreliable.
Biggest development challenge: indoor location technology and integrating it with a stable UX.