Research map / 3D vision / 3D generation
Joint Navigation and Manipulation Planning with 3D Interaction Chains
Keming Zhang, Sixian Zhang, Xinhang Song, Hongyu Wang, Yiyao Wang, Yingjie Wang, Shuqiang Jiang
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 |
|---|---|---|---|
| use 3D feature-map aggregated interaction point features instead of point-cloud features (ablation) | 3D-IC · OVMM benchmark | average success rate (%): 39.1 -> 52.7 | Improved |
| add explicit semantic labels to point cloud features (ablation) | PCD Feat. · OVMM benchmark | average success rate (%): 39.1 -> 40.2 | Improved |
| use independent single-stage interaction point decisions instead of heuristic (ablation) | OVMM heuristic baseline · OVMM benchmark | average success rate (%): 34.8 -> 49.7 | Improved |
| jointly plan all four stages as one interaction chain (ablation) | 3D-IC · OVMM benchmark | average success rate (%): 49.7 -> 52.7 | Improved |
| jointly plan only stages 3 and 4 (ablation) | 3D-IC · OVMM benchmark | Place success rate (%): 15.8 -> 20.7 | Improved |
| add token interpretation fine-tuning alone (ablation) | Qwen2.5-VL-7B policy · OVMM benchmark | average success rate (%): 46.0 -> 46.9 | Improved |
| add token preference learning (ablation) | Qwen2.5-VL-7B policy · OVMM benchmark | average success rate (%): 46.0 -> 47.6 | Improved |
| add token interpretation to TP+TS (ablation) | Qwen2.5-VL-7B policy · OVMM benchmark | average success rate (%): 50.6 -> 52.7 | Improved |
| add trajectory selection fine-tuning (ablation) | Qwen2.5-VL-7B policy · OVMM benchmark | average success rate (%): 46.0 -> 49.8 | Improved |
| use 3D-IC on a real Stretch 3 robot instead of heuristic baseline | 3D-IC · Real-world cross-room | average success rate (%): 25.0 -> 40.0 | Improved |
Closest papers by meaning
- OmniManip: Towards General Robotic Manipulation via Object-Centric Interaction Primitives as Spatial Constraints (CVPR 2025 · similarity 0.86)
- Localizing, Structuring, and Rendering: Bridging 3D and 2D Vision-Language-Action Models for Robotic Manipulation (CVPR 2026 · similarity 0.86)
- RobotSmith: Generative Robotic Tool Design for Acquisition of Complex Manipulation Skills (NeurIPS 2025 · similarity 0.86)
- GravMAD: Grounded Spatial Value Maps Guided Action Diffusion for Generalized 3D Manipulation (ICLR 2025 · similarity 0.86)
- AC-DiT: Adaptive Coordination Diffusion Transformer for Mobile Manipulation (NeurIPS 2025 · similarity 0.85)
- OWMM-Agent: Open World Mobile Manipulation With Multi-modal Agentic Data Synthesis (NeurIPS 2025 · similarity 0.85)
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.
