arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2611.09030cs.ROcs.AIcs.LG
收藏

智能体机器人学中模型、技能、反馈与课程的对照研究

A Controlled Study of Models, Skills, Feedback, and Curriculum in Agentic Robotics

UC Berkeley(加州大学伯克利分校);Bosch(博世公司);NVIDIA(英伟达公司)

Kaiyuan Chen, Xinyue Zhu, Shuangyu Xie, Uksang Yoo, Thi Thu Phuong Luong, Marcus Gualtieri, Christian Juette, Liu Ren, Spencer Huang, Ken Goldberg

AI总结:

本研究通过涵盖9个任务、631种配置、28000次部署的受控模拟实验,考察智能体机器人学的自学习循环,发现瓶颈在于信息解读而非获取,成功循环依赖智能体的失败归因、结构推理与代码重构能力。

摘要:

智能体机器人学要求编码智能体从模拟与真实世界的物理交互中实现并迭代改进机器人策略。我们通过一项受控模拟研究考察智能体机器人学的自学习循环,该研究涵盖9个操作任务、631种不同实验配置以及28000次编码智能体的部署,系统地改变模型能力、技能先验、反馈结构与课程设置。实验数据表明,瓶颈往往并非获取足够信息,而是将其解读为正确的策略更新。在这些因素中,成功的自学习循环取决于智能体归因失败、推理任务相关空间与几何结构,以及选择性重构代码以改进策略的有效性。

英文摘要:

Agentic robotics asks coding agents to implement and iteratively improve robot policies from simulated and real-world physical interactions. We study the self-learning loop of agentic robotics through a controlled simulation study spanning nine manipulation tasks, 631 different experiment configurations, and 28K coding agent rollouts, systematically varying model capability, skill priors, feedback structure, and curriculum. The experiment data suggests that the bottleneck is often not the access to sufficient information, but interpreting it into the right policy update. Across these factors, a successful self-learning loop depends on how effectively agents attribute failures, reason task-relevant spatial and geometric structure, and selectively restructure code to improve policies.

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。

↑