Aion: Towards Hierarchical 4D Scene Graphs with Temporal Flow Dynamics
Iacopo Catalano, Eduardo Montijano, Julio A. Placed, Javier Civera, Jorge Peña Queralta
AI summary
Problem
Existing 3D scene graphs lack temporal dynamics, while motion prediction models rely on grid-based discretizations that lack semantic awareness and scale poorly to large environments.
Approach
Aion embeds temporal flow dynamics into navigational nodes of a hierarchical scene graph using a sparse graph-based representation and a temporal ownership transfer mechanism to maintain consistency during graph updates.
Key results
- Graph-based sparse Maps of Dynamics for semantic temporal reasoning
- Dynamic topology temporal modeling via position-invariant indexing
- Seamless integration of temporal flow predictions into navigational nodes
- Scalable, memory-efficient spatio-temporal representation for large environments
Why it matters
Enables autonomous robots to anticipate human motion and plan proactively in complex, dynamic environments using interpretable, semantically grounded spatial models.
Abstract
Autonomous navigation in dynamic environments requires spatial representations that capture both semantic structure and temporal evolution. 3D Scene Graphs (3DSGs) provide hierarchical multi-resolution abstractions that encode geometry and semantics, but existing extensions toward dynam- ics largely focus on individual objects or agents. In parallel, † Equal Project Management. This work was partially supported by the Kaute Foundation through the Tutkijat Maailmalle program, by DGA FSE T73 23R and by project UNDERAIBOT (CPP2022-009792) funded by MI- CIU/AEI/10.13039/501100011033 and European Union (NextGenera- tionEU/PRTR). Iacopo Catalano is with the University of Turku, 20014, Turku, Finland. (e-mail: imcata@utu.fi). Jorge Pe ̃na-Queralta is with the Centre for Artificial Intelligence, Z ̈urich University of Applied Sciences, Winterthur, Switzerland. (e-mail: penq@zhaw.ch). Julio A. Placed is with the Instituto Tecnol ́ogico de Arag ́on (ITA) and the University of Zaragoza, Mar ́ıa de Luna 3-7, Zaragoza, Spain (e-mail: jplaced@ita.es). Javier Civera and Eduardo Montijano are with the University of Zaragoza, 50018, Zaragoza, Spain. (e-mail: jcivera@unizar.es, emonti@unizar.es). Maps of Dynamics (MoDs) model typical motion patterns and temporal regularities, yet are usually tied to grid-based discretizations that lack semantic awareness and do not scale well to large environments. In this paper we introduce Aion, a framework that embeds temporal flow dynamics directly within a hierarchical 3DSG, effectively incorporating the tem- poral dimension. Aion employs a graph-based sparse MoD representation to capture motion flows over arbitrary time intervals and attaches them to navigational nodes in the scene graph, yielding more interpretable and scalable predictions that improve planning and interaction in complex dynamic environments.