rollout_pick_n_place_dagger_r1
收藏资源简介:
该数据集是一个用于机器人学研究的演示数据集,由LeRobot项目创建。数据集记录了机器人执行拾取与放置(pick and place)任务的过程。数据采用时间序列格式组织,包含机器人动作指令、关节状态观测以及来自两个摄像头(前置视角和带标注视角)的视觉观测视频。具体而言,每个数据样本包含一个6维的浮点型动作向量,对应肩部平移、肩部抬升、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置;一个6维的浮点型观测状态向量,反映上述关节的实时位置;两个分辨率为480x640的RGB视频观测流(帧率为30fps,使用AV1编解码器);以及时间戳、帧索引、回合索引、样本索引和任务索引等元数据。数据集总规模为1个任务回合,包含788个时间步(帧)的数据。数据以Parquet文件格式存储,适用于机器人模仿学习(如DAgger算法)、策略学习、状态估计等研究任务。
This dataset is a demonstration dataset for robotics research, created by the LeRobot project. It records the process of robots performing pick-and-place tasks. The data is organized in time-series format, including robot action commands, joint state observations, and visual observation videos from two cameras (front view and annotated view). Specifically, each data sample contains a 6-dimensional floating-point action vector corresponding to shoulder translation, shoulder elevation, elbow flexion, wrist flexion, wrist rotation, and gripper position; a 6-dimensional floating-point observation state vector reflecting the real-time positions of the above-mentioned joints; two RGB video observation streams with a resolution of 480×640 (30 fps frame rate, using the AV1 codec); as well as metadata such as timestamps, frame indices, episode indices, sample indices, and task indices. The total scale of the dataset is 1 task episode, containing 788 time steps (frames) of data. The data is stored in Parquet file format, which is applicable to research tasks such as robotic imitation learning (e.g., DAgger algorithm), policy learning, and state estimation.
数据集概述
- 数据集名称:
rollout_pick_n_place_dagger_r1 - 许可证: Apache-2.0
- 任务类别: 机器人 (robotics)
- 标签: LeRobot
- 创建工具: 本数据集使用 LeRobot 创建
数据集结构
- 数据格式: Parquet 文件 (位于
data/*/*.parquet) - 数据总帧数: 788 帧
- 总任务数: 1
- 总集数: 1
- 帧率: 30 FPS
- 数据文件大小: 约 100 MB
- 视频文件大小: 约 200 MB
- 机器人类型: so_follower
特征 (Features)
| 特征名 | 数据类型 | 形状 | 说明 |
|---|---|---|---|
action |
float32 | (6,) | 机器人动作: shoulder_pan.pos, shoulder_lift.pos, elbow_flex.pos, wrist_flex.pos, wrist_roll.pos, gripper.pos |
observation.state |
float32 | (6,) | 机器人状态 (与动作维度相同) |
observation.images.front |
video | (480, 640, 3) | 前置摄像头视频 (AV1 编码, 30 FPS) |
observation.images.annotated |
video | (480, 640, 3) | 标注版视频 (AV1 编码, 30 FPS) |
timestamp |
float32 | (1,) | 时间戳 |
frame_index |
int64 | (1,) | 帧索引 |
episode_index |
int64 | (1,) | 集索引 |
index |
int64 | (1,) | 全局索引 |
task_index |
int64 | (1,) | 任务索引 |
数据集划分
- 训练集: 0:1 (全部数据用于训练)
文件路径结构
- 数据路径:
data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet - 视频路径:
videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4
可视化
您可以通过 此处 在线可视化本数据集。




