Spatial anchors: Difference between revisions - VR & AR Wiki - Virtual Reality & Augmented Reality Wiki

(4 intermediate revisions by the same user not shown)

Line 5:

| used_by = [[ARKit]], [[ARCore]], [[Microsoft HoloLens]], [[Magic Leap]], [[Meta Quest]], [[OpenXR]]

}}

[[File:Spatial anchors1.png|right|300px]]

[[File:Spatial anchors2.png|right|300px]]

[[File:Spatial anchors4.png|right|300px]]

'''Spatial anchors''' are persistent reference points in the real world that [[augmented reality]] (AR) and [[mixed reality]] (MR) systems use to lock virtual objects to a fixed location in physical space.<ref name="MagicLeap">[https://developer-docs.magicleap.cloud/docs/guides/unity/perception/anchors/spatial-anchors-overview/ Magic Leap Developer Docs – Spatial Anchors Overview (2025)]</ref><ref name="MSLearn">[https://learn.microsoft.com/en-us/windows/mixed-reality/design/spatial-anchors Microsoft Learn – Spatial Anchors (2025)]</ref> A spatial anchor establishes a world-locked frame of reference that defines a point in the environment with a unique coordinate frame, capturing a complete '''[[6 degrees of freedom]] (6DOF)''' representation-three translational coordinates (X, Y, Z position) and three rotational coordinates (pitch, yaw, roll orientation).<ref name="OpenXR">[https://registry.khronos.org/OpenXR/specs/1.1/man/html/XrSpatialAnchorMSFT.html Khronos OpenXR – XR_MSFT_spatial_anchor Extension Specification]</ref><ref name="BrownWiki">[https://www.vrwiki.cs.brown.edu/vr-development-software/unity/spatial-anchors VR Software Wiki – Spatial Anchors in Unity]</ref> The AR device continuously tracks this anchor over time, so that any digital content attached to it remains accurately '''[[world-locked]]''' (tied to a real-world position and orientation) rather than floating or drifting as the user moves.<ref name="ARKitAnchor">[https://www.captechconsulting.com/blogs/visualizing-surfaces-detected-by-arkit CapTech Consulting – ARAnchor ARKit Overview (2019)]</ref>

Line 10:

Line 14:

By rendering virtual objects relative to a spatial anchor's coordinate system, those objects appear fixed in the real world with minimal drift or deviation, even as the user changes their viewpoint or returns to the scene later.<ref name="MagicLeap"/><ref name="OpenXR"/> This capability is essential for creating believable and immersive experiences where digital elements appear to be a natural part of the user's surroundings, solving the fundamental AR problem of '''[[drift]]'''-where virtual objects can appear to float away from their intended positions as the system's understanding of the environment updates.<ref name="RecreateFAQ">[https://recreate.nl/faq-items/what-is-a-spatial-anchor/ Recreate – What is a spatial anchor?]</ref>

Spatial anchors enable three critical features in AR/MR applications: '''[[stability]]''' (ensuring virtual content stays precisely fixed in place), '''[[persistence]]''' (allowing virtual content to be saved and reloaded across different sessions), and '''collaboration''' (enabling multiple users and devices to share a common frame of reference for co-located, multi-user experiences).<ref name="MSLearn"/><ref name="MetaDesignAnchors">[https://developers.meta.com/horizon/design/mr-design-spatial-anchors/ Meta for Developers – Spatial Anchors Design]</ref>

Spatial anchors enable three critical features in AR/MR applications: '''[[stability]]''' (ensuring virtual content stays precisely fixed in place), '''[[persistence]]''' (allowing virtual content to be saved and reloaded across different sessions), and '''[[collaboration]]''' (enabling multiple users and devices to share a common frame of reference for co-located, multi-user experiences).<ref name="MSLearn"/><ref name="MetaDesignAnchors">[https://developers.meta.com/horizon/design/mr-design-spatial-anchors/ Meta for Developers – Spatial Anchors Design]</ref>

== Definition and core principles ==

Line 367:

Line 371:

=== AI and spatial computing convergence ===

'''On-device machine learning''' in spatial computing enables Apple's Create ML to train object tracking models directly from command line-converting 3D model files through ML training for visionOS tracking. '''Multimodal AI''' bridges spatial and business data, processing text, image, audio, and spatial data types simultaneously.<ref name="Deloitte_Trends">[https://www2.deloitte.com/us/en/insights/focus/tech-trends/2025/tech-trends-future-of-spatial-computing.html Deloitte Tech Trends 2025 – Future of Spatial Computing]</ref>

'''[[On-device machine learning]]''' in spatial computing enables Apple's Create ML to train object tracking models directly from command line-converting 3D model files through ML training for visionOS tracking. '''[[Multimodal AI]]''' bridges spatial and business data, processing text, image, audio, and spatial data types simultaneously.<ref name="Deloitte_Trends">[https://www2.deloitte.com/us/en/insights/focus/tech-trends/2025/tech-trends-future-of-spatial-computing.html Deloitte Tech Trends 2025 – Future of Spatial Computing]</ref>

'''Spatial AI''' at the intersection of spatial computing and AI/ML includes edge AI processors for local inference, spatial cameras and sensors for AR, hardware-in-the-loop simulation platforms, and geospatial data processing units. Deloitte Tech Trends 2025 predicts spatial computing will converge with '''agentic AI systems''' that are context-aware and capable of executing functions proactively, able to serve the right content at the right time without explicit commands, and multimodal while processing spatial, visual, and business data simultaneously.<ref name="Deloitte_Trends"/>

'''[[Spatial AI]]''' at the intersection of spatial computing and AI/ML includes edge AI processors for local inference, spatial cameras and sensors for AR, hardware-in-the-loop simulation platforms, and geospatial data processing units. Deloitte Tech Trends 2025 predicts spatial computing will converge with '''[[agentic AI]] systems''' that are context-aware and capable of executing functions proactively, able to serve the right content at the right time without explicit commands, and multimodal while processing spatial, visual, and business data simultaneously.<ref name="Deloitte_Trends"/>

=== Future applications ===