robo_ai_u_r5e

  • Sự miêu tả :

Bộ dữ liệu minh họa Universal Robots UR5e dùng để huấn luyện mô hình VLA. Dữ liệu được thu thập từ không gian làm việc thực tế của robot với các chương trình robot được lập trình sẵn.

Tách ra Ví dụ
  • Cấu trúc tính năng :
FeaturesDict({
    'episode_metadata': FeaturesDict({
        'episode_index': int32,
        'file_name': string,
    }),
    'steps': Dataset({
        'absolute_action_mask': Tensor(shape=(7,), dtype=bool),
        'action': Tensor(shape=(7,), dtype=float32),
        'action_normalization_mask': Tensor(shape=(7,), dtype=bool),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'language_instruction': string,
        'observation': FeaturesDict({
            'image': Image(shape=(480, 640, 3), dtype=uint8),
            'robot_state': Tensor(shape=(15,), dtype=float32),
        }),
    }),
})
  • Tài liệu mô tả tính năng :
Tính năng Lớp học Hình dạng Loại D Sự miêu tả
Từ điển tính năng
siêu dữ liệu tập Từ điển tính năng
siêu dữ liệu tập/chỉ mục tập Tensor int32
siêu dữ liệu tập/tên tệp Tensor sợi dây
các bước Tập dữ liệu
các bước/mặt nạ hành động tuyệt đối Tensor (7,) boolean
các bước/hành động Tensor (7,) float32
các bước/mặt nạ chuẩn hóa hành động Tensor (7,) boolean
bước/là_đầu tiên Tensor boolean
bước/là_cuối Tensor boolean
các bước/là_điểm_cuối Tensor boolean
các bước/hướng dẫn ngôn ngữ Tensor sợi dây
các bước/quan sát Từ điển tính năng
các bước/quan sát/hình ảnh Hình ảnh (480, 640, 3) uint8
các bước/quan sát/trạng thái robot Tensor (15,) float32
@article{RoboAIUR5e2025,
  title        = {RoboAI UR5e Training Dataset},
  author       = {Joonas Rouhiainen},
  institution  = {RoboAI Research Center, Satakunta University of Applied Sciences},
  year         = {2025},
  howpublished = {\url{https://sites.google.com/view/roboai-ur5e/home} },
}