CoRL2026-CSI/SO101-PickandPlace_Ours_30epi_8to2

该数据集是一个机器人操作数据集，专门用于训练和评估机器人技能。数据集包含30个完整的情节（episodes），总计9974帧数据，覆盖1个任务。数据采集自SO101跟随者机器人（so101_follower），帧率为10 fps。数据集以Parquet格式存储，总数据文件大小为100 MB，视频文件大小为200 MB。数据特征包括：机器人关节状态（6维浮点数）、动作（6维浮点数）、两个摄像头图像（顶部和左腕，分辨率480x640，3通道RGB视频）、末端执行器位置（6维浮点数表示XYZ和RPY）、夹爪状态（1维浮点数）、技能的自然语言描述、技能验证问题、技能类型、技能进度、技能目标位置（关节和机器人坐标系）、子任务的自然语言描述、对象名称、目标位置（3维浮点数），以及时间戳、帧索引、情节索引、索引和任务索引等元数据。数据集适用于机器人学习、强化学习、计算机视觉和自然语言处理等研究领域。This dataset is a robotics manipulation dataset designed for training and evaluating robot skills. It contains 30 complete episodes, totaling 9974 frames, covering 1 task. The data was collected from an SO101 follower robot (so101_follower) at a frame rate of 10 fps. The dataset is stored in Parquet format, with a total data file size of 100 MB and video file size of 200 MB. Features include: robot joint states (6-dimensional float), actions (6-dimensional float), two camera images (top and left wrist, resolution 480x640, 3-channel RGB video), end-effector position (6-dimensional float for XYZ and RPY), gripper state (1-dimensional float), natural language descriptions of skills, skill verification questions, skill types, skill progress, skill goal positions (joint and robot coordinate systems), natural language descriptions of subtasks, object names, target positions (3-dimensional float), and metadata such as timestamp, frame index, episode index, index, and task index. The dataset is suitable for research in robotics learning, reinforcement learning, computer vision, and natural language processing.