arXiv cs.AI· Jayjun Lee, Jessica Yin, Asif Rana, Nicholas Blauch, Octi Zhang, Sam Mady, Mohak Bhardwaj, Nima Fazeli, Nathan Ratliff, Karl Van Wyk, Ankur Handa·· 12 小时前AI 评分42
ADEPT：用强化学习预训练与后训练加速高自由度机器人灵巧操作
ADEPT: Accelerating Dexterity via Pre-Training and Post-Training using Reinforcement Learning
AI 导读
ADEPT 是一个大规模强化学习框架，先在通用物体重摆任务上预训练灵巧策略，再以其为先验后训练下游策略，避免每个新任务从零学习技能。该方法结合行为克隆蒸馏、critic 预热与保守的 on-policy 更新，解决朴素 RL 微调导致预训练能力快速退化的问题，并引入关节空间 Geometric Fabric 保障运动学灵巧性。
正文 · 原文
Authors:Jayjun Lee, Jessica Yin, Asif Rana, Nicholas Blauch, Octi Zhang, Sam Mady, Mohak Bhardwaj, Nima Fazeli, Nathan Ratliff, Karl Van Wyk, Ankur Handa
Abstract:We introduce Accelerating Dexterity via Pre-Training (ADEPT), a large-scale reinforcement learning (RL) framework for learning sim-to-real transferable dexterity across high degree-of-freedom (DoF) robot embodiments that can solve long-horizon tasks directly from raw visuo-tactile perception. ADEPT pretrains a dexterous policy on a generic object reposing task, then post-trains downstream policies with this pretrained behavior as a prior. ADEPT enables learning new behaviors that are otherwise difficult to discover from scratch on multi-fingered robots and avoids learning the same set of skills over again for every new downstream task. The pretrained policy zero-shots the reposing phase of downstream tasks, but naïve RL fine-tuning rapidly degrades this capability during transfer. We address this with a stable post-training recipe combining behavior-cloning distillation, critic warm-up, and conservative on-policy updates. To safely exploit the full kinematic dexterity, we introduce a joint-space Geometric Fabric that mediates between the RL policy and the robot. We distill post-trained teachers into perceptive students that zero-shot sim-to-real transfer on two embodiments: a 23 DoF Kuka-Allegro with two RGB cameras, and a 29 DoF Flexiv-Sharpa with two RGB cameras and five vision-based tactile sensors, and can solve long-horizon tasks from challenging initial states with dexterity at human-level speed.
| Comments: | Accepted at CoRL 2026 | Project page: this https URL |
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2608.19182 [cs.RO] |
| (or arXiv:2608.19182v2 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2608.19182 arXiv-issued DOI via DataCite |
Submission history
From: Jayjun Lee [view email]
[v1]
Wed, 19 Aug 2026 17:55:06 UTC (18,995 KB)
[v2]
Fri, 16 Oct 2026 07:12:42 UTC (18,991 KB)
来源：arXiv cs.AI · arxiv.org