taccap-g1-Clamp-fixture-loosen-clip-secure-to-edge-tighten-clip-0722
收藏资源简介：
该数据集使用LeRobot框架创建，专注于机器人操作任务，具体任务为Taccap-g1-Clamp-fixture-loosen-clip-secure-to-edge-tighten-clip，涉及夹具的松开、固定到边缘并夹紧的操作。数据来自一个双TacCap夹爪机器人系统，包含28个完整episodes，总计147,185帧，帧率为30fps。数据特征包括：20维的动作和状态观测（左右机械臂的TCP位姿和夹爪位置），以及多路视觉观测——左右手腕摄像头（分辨率480x640）和左右触觉传感器图像（分辨率400x700）。数据以parquet格式存储，视频以MP4格式存储，总数据量约300MB。数据集适用于机器人学习、模仿学习及触觉感知研究。
This dataset is created using the LeRobot framework and focuses on robot manipulation tasks. It includes a task named Taccap-g1-Clamp-fixture-loosen-clip-secure-to-edge-tighten-clip, which involves loosening a fixture, securing it to an edge, and tightening a clip. The data comes from a dual TacCap gripper robot system, containing 28 complete episodes with a total of 147,185 frames at a frame rate of 30fps. Data features include: 20-dimensional action and state observations (TCP poses and gripper positions for both left and right robotic arms), and multi-view visual observations—left and right wrist cameras (resolution 480x640) and left and right tactile sensor images (resolution 400x700). The data is stored in parquet format, with videos in MP4 format, and the total data size is approximately 300MB. The dataset is suitable for robot learning, imitation learning, and tactile perception research.
数据集概述
- 数据集名称: taccap-g1-Clamp-fixture-loosen-clip-secure-to-edge-tighten-clip-0722
- 许可证: Apache-2.0
- 任务类别: 机器人学 (Robotics)
- 创建工具: LeRobot (https://github.com/huggingface/lerobot)
- 机器人类型:
bi_taccap_gripper（双工位触觉夹爪）
数据集规模
| 指标 | 数值 |
|---|---|
| 总回合数 (Episodes) | 28 |
| 总帧数 (Frames) | 147,185 |
| 任务数量 | 1 |
| 帧率 (FPS) | 30 |
| 数据文件大小 | 100 MB |
| 视频文件大小 | 200 MB |
| 块大小 (Chunks) | 1000 |
- 数据集分割: 所有 28 回合全部用于训练 (
train: 0:28)
数据结构
数据文件与视频文件路径
- 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
特征 (Features)
| 特征名称 | 数据类型 | 形状 | 说明 |
|---|---|---|---|
action |
float32 | [20] | 20维机械臂末端执行器位姿与夹爪开合指令 |
observation.state |
float32 | [20] | 20维机器人状态观测（同动作空间结构） |
observation.images.left_tactile_left |
video | [400, 700, 3] | 左触觉传感器左视角彩色视频 (h264, 30fps) |
observation.images.left_tactile_right |
video | [400, 700, 3] | 左触觉传感器右视角彩色视频 (h264, 30fps) |
observation.images.left_wrist |
video | [480, 640, 3] | 左手腕摄像头彩色视频 (h264, 30fps) |
observation.images.right_tactile_left |
video | [400, 700, 3] | 右触觉传感器左视角彩色视频 (h264, 30fps) |
observation.images.right_tactile_right |
video | [400, 700, 3] | 右触觉传感器右视角彩色视频 (h264, 30fps) |
observation.images.right_wrist |
video | [480, 640, 3] | 右手腕摄像头彩色视频 (h264, 30fps) |
timestamp |
float32 | [1] | 时间戳 |
frame_index |
int64 | [1] | 帧索引 |
episode_index |
int64 | [1] | 回合索引 |
index |
int64 | [1] | 全局索引 |
task_index |
int64 | [1] | 任务索引 |
动作/状态空间维度说明
action和observation.state均包含 20 个维度，对应左右机械臂的 TCP（工具中心点）位置 (x, y, z)、姿态 (r1~r6) 和夹爪开合位置 (pos)：left_tcp.x~left_tcp.r6(7维)left_gripper.pos(1维)right_tcp.x~right_tcp.r6(7维)right_gripper.pos(1维)
视觉传感器规格
| 摄像头 | 分辨率 (高度×宽度) | 编码格式 | 帧率 |
|---|---|---|---|
| 左/右触觉左视角 | 400 × 700 | H.264 (yuv420p) | 30 fps |
| 左/右触觉右视角 | 400 × 700 | H.264 (yuv420p) | 30 fps |
| 左/右手腕摄像头 | 480 × 640 | H.264 (yuv420p) | 30 fps |
所有视频均无深度信息 (非深度图)、无音频。
引用
bibtex @misc{xense-taccap-lerobot, author = {XenseRobotics Team}, title = {LeRobot-Xense: LeRobot with Xense Tactile Robotics Support}, howpublished = {url{https://github.com/Vertax42/xense-taccap-lerobot}}, year = {2026} }
