Learning 3D Affordances for Blade Insertion in Cluttered Stowing
Quick Answer
VulcanVoxel introduces a novel approach for learning 3D affordances in robotic stowing, achieving a top-5 coverage of 0.89 against 0.71 from traditional pose-based methods.
Quick Take
VulcanVoxel introduces a novel approach for learning 3D affordances in robotic stowing, achieving a top-5 coverage of 0.89 against 0.71 from traditional pose-based methods. Trained on 10,000 warehouse stow episodes, it utilizes a masked autoencoder for efficient voxel-based inference, reducing processing time from 1.4 seconds to 30 ms. This advancement enhances blade insertion capabilities in cluttered environments.
Key Points
- VulcanVoxel uses a masked autoencoder for 3D occupancy field reconstruction.
- Achieves top-5 coverage of 0.89, outperforming the best pose-based baseline.
- Processes blade occupancy locally at each voxel for enhanced feasibility assessment.
- Trained on 10,000 real warehouse stow episodes without human annotation.
- RGB-to-voxel inference is completed in 30 ms, significantly faster than voxel-to-voxel.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Many manipulation tasks require reasoning about free-space affordances: discovering volumes where an extended rigid tool can safely navigate, complementary to surface contact affordances for grasping. Robotic stowing is a canonical instance, where a blade must sweep items aside inside cluttered fabric bins to create insertion space. Production stow systems generate millions of such episodes, but standard approaches with unimodal data infer affordances as SE(3) pose distributions, a geometric question asked in the wrong domain. VulcanVoxel keeps inference spatial: a masked autoencoder over 3D occupancy fields reconstructs blade occupancy conditioned on scene geometry, computing feasibility locally at each voxel and recovering multi-modal predictions from unimodal data. Blade affordances are spatial objects, subsets of 3D space defined by geometric feasibility. Pose parameters carry no structure for reasoning whether unobserved placements are feasible, and standard generative objectives including flow matching faithfully learn the unimodal distribution produced by execution policies and cannot recover geometric alternatives. Trained on 10,000 real warehouse stow episodes without human annotation, VulcanVoxel achieves top-5 coverage of 0.89 versus 0.71 for the best pose-based baseline, with a distilled student providing RGB-to-voxel inference in 30 ms. vs. 1.4 s. for voxel-to-voxel. We have released a dataset of real blade insertion cycles with RGB-D observations and pose trajectories at this https URL. html.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Robotics (cs.RO) |
| Cite as: | arXiv:2607.02549 [cs.CV] |
| (or arXiv:2607.02549v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2607.02549 arXiv-issued DOI via DataCite |
Submission history
From: Harpreet Sawhney [view email]
[v1]
Thu, 25 Jun 2026 22:09:56 UTC (1,853 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →LLM-Guided ANN Index Optimization for Human-Object Interaction Retrieval
A phase-aware LLM agent optimizes human-object interaction retrieval, outperforming Optuna TPE by 33.3% and VDTuner by 34.2% on the HICO-DET benchmark. This method enhances throughput by 15.3x over UniIR and demonstrates strong transferability across vector database management systems.