从演示中学习作为概率自动机的任务规范
Learning Task Specifications from Demonstrations as Probabilistic Automata
AI总结：
针对现有从演示学习方法难应对长跨度任务的问题，提出一种高效学习概率确定性有限自动机的方法，可从演示推断子及时序依赖，生成可解释任务规范，经物体操作实验验证，能让机械臂复制多样专家策略并适应环境变化。
摘要：
传统上，为机器人系统指定任务需要编程专业知识、深厚的领域知识以及大量时间投入。虽然从演示中学习提供了一种有前景的替代方案，但现有方法往往难以应对更长时间跨度的任务。为解决这一局限，我们提出一种计算高效的方法，用于学习概率确定性有限自动机（PDFA，probabilistic deterministic finite automata），可直接从演示中捕捉任务结构与专家偏好。该方法推断子目标及其时间依赖关系，生成领域专家易于理解和调整的可解释任务规范。我们通过涉及物体操作任务的实验验证了该方法，展示了其如何让机械臂有效复制多样的专家策略，同时适应不断变化的环境条件。
英文摘要：
Specifying tasks for robotic systems traditionally requires coding expertise, deep domain knowledge, and significant time investment. While learning from demonstration offers a promising alternative, existing methods often struggle with tasks of longer horizons. To address this limitation, we introduce a computationally efficient approach for learning probabilistic deterministic finite automata (PDFA) that capture task structures and expert preferences directly from demonstrations. Our approach infers sub-goals and their temporal dependencies, producing an interpretable task specification that domain experts can easily understand and adjust. We validate our method through experiments involving object manipulation tasks, showcasing how our method enables a robot arm to effectively replicate diverse expert strategies while adapting to changing conditions.
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。
