aloha_sim_transfer_cube_human

该数据集是一个正在开发中的机器人学测试数据集，专为ALOHA环境设计。数据集基于LeRobot项目，并进行了修改以支持ALOHA环境及其特定的观察空间要求。主要修改包括增加了对`environment_state`的支持，修改了环境以返回12维状态向量，并更新了文档以反映观察空间的变化。数据集包含50个episodes，20000帧，1个任务和50个视频。数据以parquet格式存储，视频以mp4格式存储。数据集的结构包括观察图像、状态、环境状态、动作等多个特征。环境状态包含12维向量，描述了环境中立方体的状态，包括中心坐标、尺寸、颜色、刚度等属性。

This is an under-development robotics test dataset specifically designed for the ALOHA environment. It is based on the LeRobot project, and has been modified to support the ALOHA environment and its specific observation space requirements. The main modifications include adding support for `environment_state`, revising the environment to return a 12-dimensional state vector, and updating the documentation to reflect the changes in the observation space. The dataset contains 50 episodes, 20,000 frames, 1 task, and 50 videos. The data is stored in Parquet format, while the videos are stored in MP4 format. The dataset structure includes multiple features such as observation images, states, environment states, and actions. The environment state consists of a 12-dimensional vector that describes the states of the cubes in the environment, including attributes such as center coordinates, dimensions, color, and stiffness.

原始信息汇总

数据集概述

基本信息

• 数据集名称: aloha_sim_transfer_cube_human
• 主页: https://tonyzhaozh.github.io/aloha/
• 论文: https://arxiv.org/abs/2304.13705
• 许可证: MIT
• 任务类别: 机器人学
• 标签: LeRobot, aloha

数据集结构

• 总集数: 50
• 总帧数: 20000
• 总任务数: 1
• 总视频数: 50
• 总块数: 1
• 块大小: 1000
• 帧率: 50 fps
• 训练集划分: 0:50

数据文件

• 数据路径: data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet
• 视频路径: videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4

特征描述

1. observation.images.top

   • 类型: 视频
   • 形状: [480, 640, 3]
   • 帧率: 50.0
   • 编解码器: av1
   • 像素格式: yuv420p
   • 深度图: 否
   • 音频: 无

2. observation.state

   • 类型: float32
   • 形状: [14]
   • 电机名称:
     • left_waist
     • left_shoulder
     • left_elbow
     • left_forearm_roll
     • left_wrist_angle
     • left_wrist_rotate
     • left_gripper
     • right_waist
     • right_shoulder
     • right_elbow
     • right_forearm_roll
     • right_wrist_angle
     • right_wrist_rotate
     • right_gripper

3. action

   • 类型: float32
   • 形状: [14]
   • 电机名称: 同observation.state

4. 其他特征

   • episode_index: int64, 形状[1]
   • frame_index: int64, 形状[1]
   • timestamp: float32, 形状[1]
   • next.done: bool, 形状[1]
   • index: int64, 形状[1]
   • task_index: int64, 形状[1]

引用

bibtex @article{Zhao2023LearningFB, title={Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware}, author={Tony Zhao and Vikash Kumar and Sergey Levine and Chelsea Finn}, journal={RSS}, year={2023}, volume={abs/2304.13705}, url={https://arxiv.org/abs/2304.13705} }