ATOM: Abstract Transition and Option Modeling - Experience-Informed Hierarchical Navigation
We propose ATOM (Abstract Transition and Option Modeling), a hierarchical navigation framework that combines higher-level graph planning with reusable lower-level option policies for efficient repeated navigation in large-scale 3D environments represented by hierarchical structures such as 3D scene graphs (3DSGs).
ProofPaper ↗
Key points
- ATOM operates on a hierarchical graph representation, implemented using, for instance, the Rooms and Places layers of a 3DSG, where each directed abstract transition is associated with a navigation option.
- Across repeated navigation tasks, ATOM maintains a smoothed empirical distribution over option initiation states and uses it to progressively refine the corresponding abstract transition cost.
- Ablation studies show that ATOM reduces cumulative repeated-planning time by 97.5% relative to a monolithic MDP baseline while maintaining trajectories within approximately 7% of a Global Places A shortest-path reference.
- Experience-informed refinement further reduces abstract transition-cost estimation error by 10.3%.
Sources (1)
- [1]ATOM: Abstract Transition and Option Modeling - Experience-Informed Hierarchical NavigationarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 13, 01:24 AM
We propose ATOM (Abstract Transition and Option Modeling), a hierarchical navigation framework that combines higher-level graph planning with reusable lower-level option policies for efficient repeated navigation in large-scale 3D environments represented by hierarchical structures such as 3D scene graphs (3DSGs).
ATOM operates on a hierarchical graph representation, implemented using, for instance, the Rooms and Places layers of a 3DSG, where each directed abstract transition is associated with a navigation option.
Extractive summary: sentences quoted from the sources.