CAPABLE：通过行为潜编码实现能力感知的策略自适应

View PDF HTML (experimental)

Abstract:Vision-language-action (VLA) policies assume the embodiment on which they were trained and can fail when a joint fault changes how commanded actions are physically executed. Existing fault-recovery methods often require task-specific retraining, fault labels, explicit diagnosis, or privileged embodiment information. We introduce CAPABLE, a unified capability-aware adaptation framework for frozen VLAs that integrates self-supervised capability inference with residual reinforcement learning. CAPABLE infers capability, how much of the commanded motion each joint actually realizes and how that motion contributes to end-effector behavior, online from command-response history and kinematics using a temporal encoder shared across joints, Jacobian grounding, cross-joint attention, and self-supervised physical prediction. The resulting representation conditions a residual policy that adds bounded corrections to the VLA arm action without fault labels or faulty-joint identifiers. Across 28 LIBERO tasks, CAPABLE raises success on an actuator excluded from fault training from 24.8% to 59.3%, outperforming a parameter-matched global-history baseline by 17.4 points while preserving healthy performance. Leave-one-actuator-out experiments across six joints show that this transfer is not specific to one actuator, and additional evaluations characterize transfer to unseen fault families and demonstrate recovery on a physical Franka Panda. this https URL

Subjects:	Robotics (cs.RO); Machine Learning (cs.LG); Systems and Control (eess.SY)
Cite as:	arXiv:2610.11971 [cs.RO]
	(or arXiv:2610.11971v1 [cs.RO] for this version)
	https://doi.org/10.48550/arXiv.2610.11971 arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Mohammad Khoshnazar [view email]
[v1] Thu, 8 Oct 2026 13:46:52 UTC (3,925 KB)