Source and Target Datasets

该数据集包含了4000个状态图像，这些图像是在两个模拟环境域中采用随机行为政策拍摄的。在这些环境中，智能体的目标是到达一个目标位置。源数据集和目标数据集中的观察结果在像素级别上有显著的差异。该数据集用于评估少样本策略迁移方法的有效性，其中包括智能体在导航至目标位置的环境中的图像。该数据集对于理解在观察结果不匹配的不同条件下，所提出方法的性能表现至关重要。每个源数据集和目标数据集都包含了4000张图像，该任务旨在研究机器人导航中的少样本策略迁移。

This dataset comprises 4000 state images collected via random behavioral policies across two simulated environment domains. In these environments, the agent's goal is to navigate to a target position. Observations in the source and target datasets show significant pixel-level discrepancies. This dataset is utilized to evaluate the effectiveness of few-shot policy transfer methods, with images sourced from environments where agents perform navigation tasks to reach target positions. It is critical for understanding the performance of proposed methods under different conditions involving mismatched observations. Each of the source and target datasets contains 4000 images, and this task aims to investigate few-shot policy transfer in robot navigation.