FoldBack: Self-Correcting Masked Generative Policy for Long-Horizon Garment Folding
FoldBack lets a folding robot undo and retry failed grasps without retraining.
FoldBack is a self-correcting masked generative policy for long-horizon garment folding that recovers when a grasp is missed or slips. At inference it decides when to verify, how to roll back to a retryable pre-grasp pose while keeping successful grasps, and how to regenerate the failed segment and selected future actions. It does this without recovery demonstrations or retraining the base policy. On 33 real garments from six categories it reached 75.2% final folding success and 0.837 final-mask IoU, versus 45.7% and 0.689 for the strongest prior baseline.
- Detects failed grasps, rolls back, and retries before continuing
- Preserves successful grasps and avoids previous failed locations
- Needs no recovery demonstrations or base-policy retraining
- 75.2% success and 0.837 IoU versus 45.7% and 0.689
From the original article
1 min read at arxiv.orgWe present FoldBack, a self-correcting masked generative policy for long-horizon garment folding. Existing long-trajectory policies may continue after a missed or slipped grasp even when the garment has not reached the intended configuration. We structure FoldBack's recovery mechanisms around three inference-time decisions: when to refine and verify, how to roll back, and where and how to retry.
By Lipeng Zhuang · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.