Skip to content
Book Open access

Towards Persistent Dynamic Scene Understanding for Immersive Computing

Oct 2026 · Proceedings of the 4th ACM Workshop on Mobile Immersive Computing, Networking, and Systems · 0 citations · 2 references

Abstract

Future immersive computing systems must understand physical environments not only within the current field of view, but persistently, as those environments change across repeated visits. Recent functional 3D scene graphs produce semantically rich representations of objects, affordances, articulated parts, and functional relationships. However, they remain transient: each is reconstructed from a single observation session and discards object identities, functional relationships, and accumulated knowledge once the session ends. In this work, we present the vision of Functional Temporal 3D Scene Graphs (FT-3DSG), a persistent, multi-layer representation that unifies geometric mapping, semantic object understanding, functional relationships, and temporal memory in a single evolving model. Our FT-3DSG assigns persistent identities to objects and relationships, updates scene knowledge incrementally across sessions, and preserves event-driven snapshots for historical reasoning. We outline a modular architecture and identify four core research challenges: (1) efficient on-device representation, (2) cross-session relocalization and rebinding, (3) persistent cross-layer tracking, and (4) compact queryable temporal snapshot logs.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.