HyeonseokE/SO101-pnp_ma_80epi
收藏资源简介:
该数据集是一个用于机器人学习的数据集,专门针对so101_follower机器人类型创建。数据集包含80个episodes,总计78,128帧,涉及一个任务。数据特征丰富,包括机器人状态观测(6个关节位置:肩部平移、肩部提升、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置)、动作(相同6个关节位置)、来自顶部和左腕摄像头的视频观测(分辨率480x640,30fps,H.264编码)、末端执行器位置(XYZ坐标和RPY姿态)、二进制夹爪状态、技能信息(自然语言描述、验证问题、类型、进度、目标关节和笛卡尔位置)、子任务信息(自然语言描述、对象名称、目标XYZ位置)以及时间戳、帧索引、episode索引等元数据。数据集适用于机器人模仿学习、强化学习或技能学习任务,支持多模态输入(状态和图像)。数据以Parquet文件格式存储,视频以MP4格式存储,遵循Apache 2.0许可证。
This dataset is designed for robotics learning, specifically created for the so101_follower robot type. It contains 80 episodes with a total of 78,128 frames and involves one task. The dataset features rich data attributes, including robot state observations (6 joint positions: shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, and gripper positions), actions (same 6 joint positions), video observations from top and left wrist cameras (resolution 480x640, 30fps, H.264 codec), end-effector positions (XYZ coordinates and RPY orientation), binary gripper state, skill information (natural language descriptions, verification questions, types, progress, target joint and Cartesian positions), subtask information (natural language descriptions, object names, target XYZ positions), as well as metadata such as timestamps, frame indices, and episode indices. The dataset is suitable for robotics imitation learning, reinforcement learning, or skill learning tasks, supporting multimodal inputs (state and images). Data is stored in Parquet file format, with videos in MP4 format, and is licensed under Apache 2.0.



