CoRL2026-CSI/SO101-cap_close_drawer_10fps_30epi_8to2

该数据集是一个机器人控制数据集，使用LeRobot创建，专注于让机器人执行关闭抽屉的任务。数据集包含30个episodes，总计7057帧，采样频率为10fps。数据来源于so101_follower机器人，涵盖单个任务。数据集特征包括机器人关节状态（6维，如肩部平移、肩部提升、肘部弯曲等）、动作（6维，与状态对应）、来自顶部和左腕摄像头的图像观测（分辨率480x640，3通道RGB视频）、末端执行器位置和姿态（6维，包括xyz坐标和roll/pitch/yaw角度）、二值化夹爪状态、技能相关数据（如自然语言描述、验证问题、技能类型、进度、目标位置等）、子任务信息（如自然语言描述、对象名称、目标位置）、以及时间戳、帧索引、episode索引等元数据。数据以parquet格式存储，视频以mp4格式存储，适用于机器人学习、模仿学习或强化学习研究。This dataset is a robotics control dataset created using LeRobot, focusing on the task of closing a drawer with a robot. It contains 30 episodes, totaling 7057 frames, with a sampling rate of 10fps. The data originates from the so101_follower robot and covers a single task. Dataset features include robot joint states (6-dimensional, such as shoulder pan, shoulder lift, elbow flex, etc.), actions (6-dimensional, corresponding to states), image observations from top and left wrist cameras (resolution 480x640, 3-channel RGB video), end-effector position and orientation (6-dimensional, including xyz coordinates and roll/pitch/yaw angles), binary gripper state, skill-related data (e.g., natural language descriptions, verification questions, skill type, progress, goal positions), subtask information (e.g., natural language descriptions, object names, target positions), and metadata such as timestamps, frame indices, and episode indices. The data is stored in parquet format, with videos in mp4 format, suitable for robotics learning, imitation learning, or reinforcement learning research.