← 研踪 PaperTrailarXiv:2606.31043 · v4
Warp: Structured Output-Space Adaptation for Frozen Robot Policies
首次提交 2026-06-30 10:25
版本更新 2026-11-04 09:14
雷达观测 2026-11-05 11:11
摘要
Robot policies that remain broadly competent after deployment can degrade when dynamics, contact, geometry, actuation, or sensing differ from source conditions. We study post-hoc adaptation through structured output-space adapters: compact transformations of samples from a frozen base policy. We introduce Warp, a context-conditioned monotone bijective spline that enables nonlinear sample-dependent reshaping while preserving per-dimension ordering; Translation and Affine provide location-only and location--scale references. Across four ManiSkill tasks with controlled deployment shifts, Warp achieves the strongest recovery in a matched evolution-strategy (ES) adapter study, outperforming controls for spline parameterization, location--scale adaptation, and generic nonlinear capacity. Under PPO, the ordering becomes task dependent, showing that richer adapter structure does not automatically translate into better adaptation under a different optimizer. Finally, we test the practical benefit of this approach on physical SO101 peg insertion. Within an interaction-limited RLPD/SAC pipeline, Warp achieves comparable success to a strong stochastic residual while reducing median successful cycle time by approximately 30%. These results identify structured output-space adaptation as a useful design space for adapting frozen robot policies.
作者备注：8 pages, 4 figures
观测到的版本
相邻记录可能跳过中间版本；首次发现 v2 及以上也会标为版本更新。
v4 · 2026-11-04 09:14 · 最新观测版本
Warp: Structured Output-Space Adaptation for Frozen Robot Policies
Robot policies that remain broadly competent after deployment can degrade when dynamics, contact, geometry, actuation, or sensing differ from source conditions. We study post-hoc adaptation through structured output-space adapters: compact transformations of samples from a frozen base policy. We introduce Warp, a context-conditioned monotone bijective spline that enables nonlinear sample-dependent reshaping while preserving per-dimension ordering; Translation and Affine provide location-only and location--scale references. Across four ManiSkill tasks with controlled deployment shifts, Warp achieves the strongest recovery in a matched evolution-strategy (ES) adapter study, outperforming controls for spline parameterization, location--scale adaptation, and generic nonlinear capacity. Under PPO, the ordering becomes task dependent, showing that richer adapter structure does not automatically translate into better adaptation under a different optimizer. Finally, we test the practical benefit of this approach on physical SO101 peg insertion. Within an interaction-limited RLPD/SAC pipeline, Warp achieves comparable success to a strong stochastic residual while reducing median successful cycle time by approximately 30%. These results identify structured output-space adaptation as a useful design space for adapting frozen robot policies.
查看此版本解读 →查看此版本原文 ↗