面向机器人操作的奖励机推理
Reward Machine Inference for Robotic Manipulation
AI总结：
该研究提出一种从机器人操作视觉演示中直接学习奖励机的LfD方法，无需预定义命题与稀疏奖励先验，联合学习结构与关键转移事件，可准确捕捉任务结构并助力RL智能体学习最优策略。
摘要：
从演示中学习（LfD）与强化学习（RL）已使机器人智能体能够完成复杂任务。奖励机（Reward Machines, RMs）通过结构化高层任务信息，增强了RL在长时程场景下训练策略的能力。本文提出一种新型LfD方法，可直接从机器人操作任务的视觉演示中学习RMs。与现有方法不同，该方法无需预定义命题，也无需底层稀疏奖励信号的先验知识，而是联合学习RM结构并识别驱动RM状态间转移的关键高层事件。我们在基于视觉的操作任务上验证了方法，结果表明推理得到的RM可准确捕捉任务结构，使RL智能体能有效学习最优策略。
英文摘要：
Learning from Demonstrations (LfD) and Reinforcement Learning (RL) have enabled robot agents to accomplish complex tasks. Reward Machines (RMs) enhance RL's capability to train policies over extended time horizons by structuring high-level task information. In this work, we introduce a novel LfD approach for learning RMs directly from visual demonstrations of robotic manipulation tasks. Unlike previous methods, our approach requires no predefined propositions or prior knowledge of the underlying sparse reward signals. Instead, it jointly learns the RM structure and identifies key high-level events that drive transitions between RM states. We validate our method on vision-based manipulation tasks, showing that the inferred RM accurately captures task structure and enables an RL agent to effectively learn an optimal policy.
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。
