aloha_handover_box_joint_pos_rl_orig
收藏资源简介:
该数据集由LeRobot创建,是一个专门用于双机械臂操作任务的机器人控制数据集。它包含50个训练集,总计11,829帧数据,帧率为25fps,以Parquet格式存储,并包含150个视频文件。数据集结构丰富,特征包括观测状态(如左右机械臂的关节角度和夹爪状态)、动作(与观测状态对应的控制命令)、以及来自多个视角的视频观测(如背部摄像头和左右腕部摄像头,分辨率分别为480x640和240x320)。此外,还包含奖励、成功标志、时间戳、帧索引、集索引等元数据。该数据集适用于机器人学习、强化学习和模仿学习任务,支持训练模型进行机械臂控制。
This dataset was developed by LeRobot, and is a specialized robotic control dataset tailored for dual-arm manipulation tasks. It consists of 50 training splits, with a total of 11,829 frames at a frame rate of 25 fps, stored in Parquet format, and includes 150 video files. The dataset has a rich structure, with features including observational states (such as joint angles and gripper states of the left and right robotic arms), actions (control commands corresponding to the observational states), and video observations from multiple viewpoints (e.g., the back camera and left/right wrist cameras, with resolutions of 480×640 and 240×320 respectively). Additionally, it encompasses metadata such as rewards, success flags, timestamps, frame indices, and split indices. This dataset is applicable to robotic learning, reinforcement learning, and imitation learning tasks, and supports the training of models for robotic arm control.
数据集概述
- 数据集名称: aloha_handover_box_joint_pos_rl_orig
- 许可证: Apache-2.0
- 任务类别: 机器人学 (Robotics)
- 标签: LeRobot
数据集描述
该数据集使用 LeRobot 框架创建,旨在为机器人强化学习提供支持。
数据集结构
- 代码库版本: v3.0
- 机器人类型: 未指定
- 总片段数 (Episodes): 50
- 总帧数 (Frames): 11,829
- 总任务数 (Tasks): 1
- 块大小 (Chunks): 1000
- 帧率 (FPS): 25
- 数据划分 (Splits): 训练集
train包含全部 50 个片段 (索引 0 至 49)
数据特征 (Features)
| 特征名称 | 数据类型 | 形状 | 说明 |
|---|---|---|---|
observation.state |
float32 | (14,) | 机器人状态,包含 7 个左臂关节位置/夹爪状态和 7 个右臂关节位置/夹爪状态 |
action |
float32 | (14,) | 动作空间,与 observation.state 结构一致 |
observation.images.back |
video | (480, 640, 3) | 后部摄像头视图 (AV1 编码, 25fps) |
observation.images.wrist_left |
video | (240, 320, 3) | 左腕部摄像头视图 (AV1 编码, 25fps) |
observation.images.wrist_right |
video | (240, 320, 3) | 右腕部摄像头视图 (AV1 编码, 25fps) |
next.reward |
float32 | (1,) | 下一步奖励值 |
next.success |
bool | (1,) | 下一步成功标志 |
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 |
reward |
float32 | (1,) | 当前奖励值 |
mc_return |
float32 | (1,) | 蒙特卡洛回报 |
数据存储
- 数据路径:
data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet(Parquet 格式) - 视频路径:
videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4(MP4 格式) - 数据文件大小: 约 100 MB
- 视频文件大小: 约 200 MB



