Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Expanding Spatial and Temporal Context for Robotic Imitation Learning With Scene Graphs
Jianing Qian, Qinhe Peng, Emmanuel Panov, Leonor Fermoselle, Dinesh Jayaraman, Bernadette Bucher, Tarik Kelestemur
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| condition diffusion policy on a dynamic task-relevant scene graph instead of RGB only | Diffusion Policy · Simulated mobile manipulation (Throw Trash, Heat-up Tea, Heat-up Tea Long) | task success rate | Improved |
| add memory of object nodes (scene graph) versus instantaneous visible-object representation | Diffusion Policy · Real-world tabletop Feed-Animals | task success rate | Improved |
| use scene graph memory for Feed-Giraffe and Clean-Table | Diffusion Policy · Real-world tabletop Feed-Giraffe, Clean-Table | task success rate | Improved |
| increase history frames to 10 and predict past and future actions (PTP) | Diffusion Policy · Simulated Heat-up Tea tasks | task success rate | Mixed |
| use object-centric representations of visible objects only instead of raw RGB | Diffusion Policy · Real-world Pineapple-Bowl | task success rate | Improved |
| add 3D centroid to scene graph nodes (ablation) | Scene graph policy · Simulated mobile manipulation, later subtasks and final task | task success rate | Improved |
| add 2D and 3D object location information to scene graph nodes (ablation) | Scene graph policy · Simulated mobile manipulation, later subtasks and final task | task success rate | Improved |
| add 3D centroid or object locations to scene graph nodes (ablation) | Scene graph policy · Simulated mobile manipulation, first subtask | first subtask success rate | No clear effect |
| add 3D centroid or object locations to scene graph nodes (ablation) | Scene graph policy · Real-world Pineapple-Bowl, placing in bowl | task success rate | No clear effect |
Closest papers by meaning
- Instant Policy: In-Context Imitation Learning via Graph Diffusion (ICLR 2025 · similarity 0.90)
- Data Scaling Laws in Imitation Learning for Robotic Manipulation (ICLR 2025 · similarity 0.88)
- GravMAD: Grounded Spatial Value Maps Guided Action Diffusion for Generalized 3D Manipulation (ICLR 2025 · similarity 0.88)
- Learning with Language-Guided State Abstractions (ICLR 2024 · similarity 0.87)
- What's the Move? Hybrid Imitation Learning via Salient Points (ICLR 2025 · similarity 0.87)
- SAM-E: Leveraging Visual Foundation Model with Sequence Imitation for Embodied Manipulation (ICML 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
