📄 中文摘要
动作空间的规范在基于模仿的机器人操作策略学习中起着关键作用，深刻影响策略学习的优化格局。尽管近期的研究主要集中在扩大训练数据和模型能力上，但动作空间的选择仍然依赖于经验法则或传统设计，导致对机器人策略设计理念的理解模糊。为了解决这一模糊性，进行了一项大规模系统的实证研究，确认动作空间对机器人策略学习具有显著且复杂的影响。研究沿时间和空间轴对动作设计空间进行了剖析，促进了对这些选择如何影响策略学习的结构化分析。
揭开机器人操作策略的动作空间设计之谜
出处: Demystifying Action Space Design for Robotic Manipulation Policies
发布: 2026年3月2日
📄 English Summary
Demystifying Action Space Design for Robotic Manipulation Policies
The specification of the action space plays a crucial role in imitation-based robotic manipulation policy learning, fundamentally shaping the optimization landscape of policy learning. While recent advances have focused heavily on scaling training data and model capacity, the choice of action space remains guided by ad-hoc heuristics or legacy designs, leading to an ambiguous understanding of robotic policy design philosophies. To address this ambiguity, a large-scale and systematic empirical study was conducted, confirming that the action space has significant and complex impacts on robotic policy learning. The study dissects the action design space along temporal and spatial axes, facilitating a structured analysis of how these choices govern policy learning.