openarm-sim-groot-v2
收藏资源简介:
该数据集使用LeRobot工具创建,是一个用于机器人学习的仿真数据集,基于双机械臂(bimanual_openarm)平台,代码库版本为v2.1,采用Apache-2.0许可证。数据规模包含2个episodes、总计1176帧、4个独立任务和6个视频片段,所有数据均划分为训练集。核心特征包括:观测状态为16维浮点数向量,表示左右机械臂各7个关节及左右手指各1个关节的角度;动作为16维浮点数向量,与观测状态对应的关节控制指令;多视角图像观测包含来自头部、左腕和右腕三个视角的RGB视频,分辨率均为480x640,帧率为30 fps,存储为MP4格式;元数据包括时间戳、帧索引、episode索引、全局索引和任务索引。该数据集适用于机器人模仿学习、强化学习、行为克隆以及多模态机器人控制策略的研究与开发。
This dataset is constructed using the LeRobot toolkit, serving as a simulation dataset for robot learning research. It is built on the bimanual_openarm robotic platform, with codebase version v2.1 and licensed under the Apache-2.0 license. In terms of data scale, this dataset includes 2 episodes, totaling 1176 frames, 4 independent tasks, and 6 video clips. The data is organized in chunks, with all samples allocated to the training split. The core features of this dataset are as follows: 1) Observation state: A 16-dimensional floating-point vector representing the joint angles of 7 joints for each of the left and right robotic arms, plus 1 joint angle for each of the left and right fingers; 2) Action: A 16-dimensional floating-point vector serving as joint control commands corresponding to the observation state; 3) Multi-view image observations: RGB videos captured from three perspectives (head, left wrist, and right wrist), all with a resolution of 480x640, a frame rate of 30 fps, and stored in MP4 format; 4) Metadata: Timestamps, frame indices, episode indices, global indices, and task indices. This dataset is applicable to research and development in robotic imitation learning, reinforcement learning, behavioral cloning, and multimodal robotic control strategy development.
数据集概述:openarm-sim-groot-v2
- 数据集地址:https://huggingface.co/datasets/20-wasa/openarm-sim-groot-v2
- 许可证:Apache-2.0
- 任务类别:机器人学(Robotics)
- 标签:LeRobot
数据集规模与结构
- 代码库版本:v2.1
- 机器人类型:双机械臂开放式机械臂(bimanual_openarm)
- 总片段数:18
- 总帧数:9278
- 总任务数:4
- 总视频数:54
- 总数据块数:1
- 数据块大小:1000
- 帧率:30 FPS
- 训练/测试划分:全部18个片段用于训练(train: "0:18")
数据特征
数据文件格式为 Parquet,视频文件格式为 MP4(H.264 编码,分辨率 480x640,30 FPS)。每条数据包含以下特征:
| 特征名称 | 数据类型 | 形状 | 说明 |
|---|---|---|---|
observation.state |
float32 | [16] | 机器人关节状态:左臂7个关节 + 右臂7个关节 + 左手指关节 + 右手指关节 |
action |
float32 | [16] | 动作指令:与状态维度一致(左右臂各7关节 + 左右手指各1关节) |
observation.images.head |
video | [480,640,3] | 头部摄像头视频(H.264,480x640,30 FPS) |
observation.images.left_wrist |
video | [480,640,3] | 左腕摄像头视频(H.264,480x640,30 FPS) |
observation.images.right_wrist |
video | [480,640,3] | 右腕摄像头视频(H.264,480x640,30 FPS) |
timestamp |
float32 | [1] | 时间戳 |
frame_index |
int64 | [1] | 帧索引 |
episode_index |
int64 | [1] | 片段索引 |
index |
int64 | [1] | 全局索引 |
task_index |
int64 | [1] | 任务索引 |
数据集创建工具
该数据集使用 LeRobot 创建。
可视化工具
可通过 可视化空间 在线查看数据集内容。



