LodeStar：基于人类演示合成数据增强的长时程灵巧操作
LodeStar: Long-horizon Dexterity via Synthetic Data Augmentation from Human Demonstrations
AI总结：
针对长时程灵巧操作任务数据获取难的问题，提出LodeStar框架，通过基础模型分解人类演示技能，结合强化学习生成合成数据集训练技能，再用SRT策略链接受训技能，显著提升任务性能与稳健性。
摘要：
开发能够以人类水平灵巧性稳健执行长时程操作任务的机器人系统具有挑战性，这类任务既需要物理灵巧性，又需要无缝衔接操作技能，同时还要稳健处理环境变化。尽管模仿学习是一种有前景的方法，但获取全面数据集耗费资源。本研究提出学习框架与系统LodeStar，它利用现成基础模型将任务演示自动分解为语义有意义的技能，并通过强化学习从少量人类演示中生成多样化的合成演示数据集。这些仿真增强数据集支持稳健的技能训练，Skill Routing Transformer（SRT）策略能有效将所学技能链接起来，执行复杂的长时程操作任务。对三个具有挑战性的真实世界长时程灵巧操作任务的实验评估表明，与之前的基线相比，我们的方法显著提升了任务性能和稳健性。视频可在lodestar-robot.github.io查看。
英文摘要：
Developing robotic systems capable of robustly executing long-horizon manipulation tasks with human-level dexterity is challenging, as such tasks require both physical dexterity and seamless sequencing of manipulation skills while robustly handling environment variations. While imitation learning offers a promising approach, acquiring comprehensive datasets is resource-intensive. In this work, we propose a learning framework and system LodeStar that automatically decomposes task demonstrations into semantically meaningful skills using off-the-shelf foundation models, and generates diverse synthetic demonstration datasets from a few human demos through reinforcement learning. These sim-augmented datasets enable robust skill training, with a Skill Routing Transformer (SRT) policy effectively chaining the learned skills together to execute complex long-horizon manipulation tasks. Experimental evaluations on three challenging real-world long-horizon dexterous manipulation tasks demonstrate that our approach significantly improves task performance and robustness compared to previous baselines. Videos are available at lodestar-robot.github.io.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
