CoRL2026-CSI/SO101-CloseDrawer_Ours_30epi
收藏资源简介:
这是一个使用LeRobot创建的机器人学习数据集,专门用于训练和评估机器人操作任务,特别是关闭抽屉的动作。数据集包含30个完整的情节(episodes),总计7970帧数据,帧率为10fps。数据以Parquet格式存储,并附带MP4格式的视频文件。数据集的特征包括机器人关节状态观察(6维浮点数)、动作(6维浮点数)、两个摄像头视角的图像观察(顶部和左腕,分辨率480x640,RGB视频)、末端执行器位置(6维浮点数)、二进制夹爪状态、技能相关的自然语言描述和验证问题、技能类型和进度、目标位置(关节和笛卡尔坐标)、子任务的自然语言描述、对象名称和目标位置,以及时间戳、帧索引、情节索引等元数据。数据集适用于机器人控制、模仿学习或强化学习研究。
This is a robotics learning dataset created using LeRobot, designed for training and evaluating robot manipulation tasks, specifically focusing on closing drawer actions. The dataset includes 30 complete episodes, with a total of 7970 frames at a frame rate of 10fps. Data is stored in Parquet format and accompanied by MP4 video files. Features include robot joint state observations (6-dimensional floats), actions (6-dimensional floats), image observations from two camera views (top and left wrist, 480x640 resolution, RGB video), end-effector position (6-dimensional floats), binary gripper state, skill-related natural language descriptions and verification questions, skill type and progress, goal positions (joint and Cartesian coordinates), subtask natural language descriptions, object names and target positions, as well as metadata such as timestamps, frame indices, and episode indices. The dataset is suitable for robotics control, imitation learning, or reinforcement learning research.




