arXiv cs.AI· Tyler Ga Wei Lum, Kushal Kedia, C. Karen Liu, Jeannette Bohg·· 2 天前AI 评分29
Play2Perfect：灵巧操作预训练中哪些因素对精密装配至关重要？
Play2Perfect: What Matters in Dexterous Play Pretraining for Precise Assembly?
AI 导读
研究提出 Play2Perfect，一个通过多样化物体与目标上的"玩耍"进行任务无关预训练的 RL 框架，再微调适配精密装配。该先验比从零 RL 训练的样本效率高 33 倍，即使后者提供密集多阶段奖励。方法实现零样本 sim-to-real 迁移，在 0.5 mm 接触间隙的紧密插入任务上成功率达 60%，长时程多部件装配和拧螺丝成功率超 50%。
正文 · 原文
Abstract:Multi-fingered robots promise the speed and dexterity of human hands, yet challenging problems such as precise assembly have remained out of reach. These tasks are contact-rich, making data collection for imitation learning difficult, and sparse-reward, making direct exploration with reinforcement learning (RL) intractable. Consequently, prior work has made progress by structuring the problem with specialized grippers, tool attachments, and environment fixtures. In this work, we argue that before a robot can perfect precise assembly, it must first learn to play. We further ask the question: what factors in the process of learning to play matter for precise assembly? We propose Play2Perfect, an RL framework for task-agnostic pretraining through play on diverse objects and goals, which is then perfected on precise assembly. The goal of play is to acquire reusable manipulation priors, such as grasping, in-hand reorientation and pose reaching. Finetuning then adapts this general prior to assembly, focusing exploration on the final contact-rich, high-precision interactions needed for success. We systematically study key design choices in play pretraining, including object diversity, training objective, trajectory diversity, and goal precision. We show that our prior is 33x more sample-efficient than RL training from scratch, even when provided with dense, multi-stage rewards. We demonstrate zero-shot sim-to-real transfer, achieving 60% success on tight insertions with only 0.5 mm contact clearance, and over 50% success on long-horizon multi-part assembly and screwing.
| Comments: | 26 pages, 19 figures, 4 tables. Project page: this https URL |
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2606.26428 [cs.RO] |
| (or arXiv:2606.26428v4 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2606.26428 arXiv-issued DOI via DataCite |
Submission history
From: Tyler Lum [view email]
[v1]
Wed, 24 Jun 2026 22:39:12 UTC (28,422 KB)
[v2]
Thu, 9 Jul 2026 17:59:45 UTC (28,422 KB)
[v3]
Tue, 8 Sep 2026 04:22:00 UTC (31,854 KB)
[v4]
Thu, 15 Oct 2026 09:00:05 UTC (31,872 KB)
来源：arXiv cs.AI · arxiv.org