stanford_kuka_multimodal_dataset_converted_externally_to_rlds

  • Sự miêu tả :

Chèn chốt Kuka iiwa với phản hồi lực

Tách ra Ví dụ
'train' 3.000
  • Cấu trúc tính năng :
FeaturesDict({
    'episode_metadata': FeaturesDict({
    }),
    'steps': Dataset({
        'action': Tensor(shape=(4,), dtype=float32),
        'discount': Scalar(shape=(), dtype=float32),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'language_embedding': Tensor(shape=(512,), dtype=float32),
        'language_instruction': Text(shape=(), dtype=string),
        'observation': FeaturesDict({
            'contact': Tensor(shape=(50,), dtype=float32),
            'depth_image': Tensor(shape=(128, 128, 1), dtype=float32),
            'ee_forces_continuous': Tensor(shape=(50, 6), dtype=float32),
            'ee_orientation': Tensor(shape=(4,), dtype=float32),
            'ee_orientation_vel': Tensor(shape=(3,), dtype=float32),
            'ee_position': Tensor(shape=(3,), dtype=float32),
            'ee_vel': Tensor(shape=(3,), dtype=float32),
            'ee_yaw': Tensor(shape=(4,), dtype=float32),
            'ee_yaw_delta': Tensor(shape=(4,), dtype=float32),
            'image': Image(shape=(128, 128, 3), dtype=uint8),
            'joint_pos': Tensor(shape=(7,), dtype=float32),
            'joint_vel': Tensor(shape=(7,), dtype=float32),
            'optical_flow': Tensor(shape=(128, 128, 2), dtype=float32),
            'state': Tensor(shape=(8,), dtype=float32),
        }),
        'reward': Scalar(shape=(), dtype=float32),
    }),
})
  • Tài liệu tính năng :
Tính năng Lớp học Hình dạng loại D Sự miêu tả
Tính năngDict
tập_siêu dữ liệu Tính năngDict
bước Tập dữ liệu
bước/hành động Tenxơ (4,) phao32 Hoạt động của rô-bốt, bao gồm [3x vị trí EEF, 1x đóng/mở kẹp cầm].
bước/giảm giá Vô hướng phao32 Giảm giá nếu được cung cấp, mặc định là 1.
bước/is_first Tenxơ bool
bước/is_last Tenxơ bool
bước/is_terminal Tenxơ bool
các bước/ngôn ngữ_embedding Tenxơ (512,) phao32 Nhúng ngôn ngữ Kona. Xem https://tfhub.dev/google/universal-sentence-encoding-large/5
các bước/ngôn ngữ_instruction Chữ sợi dây Giảng dạy ngôn ngữ.
bước/quan sát Tính năngDict
bước/quan sát/liên hệ Tenxơ (50,) phao32 Thông tin liên hệ của robot.
bước/quan sát/độ sâu_image Tenxơ (128, 128, 1) phao32 Quan sát camera độ sâu chính.
bước/quan sát/ee_forces_continuous Tenxơ (50, 6) phao32 Lực tác động cuối của robot.
bước/quan sát/ee_orientation Tenxơ (4,) phao32 Robot định hướng tác động cuối quaternion.
bước/quan sát/ee_orientation_vel Tenxơ (3,) phao32 Tốc độ định hướng của bộ phận tác động cuối của robot.
bước/quan sát/ee_position Tenxơ (3,) phao32 Vị trí tác động cuối của robot.
bước/quan sát/ee_vel Tenxơ (3,) phao32 Tốc độ tác động cuối của robot.
bước/quan sát/ee_yaw Tenxơ (4,) phao32 Robot tác động cuối ngáp.
bước/quan sát/ee_yaw_delta Tenxơ (4,) phao32 Robot tác động cuối cùng là yaw delta.
bước/quan sát/hình ảnh Hình ảnh (128, 128, 3) uint8 Quan sát RGB của camera chính.
bước/quan sát/joint_pos Tenxơ (7,) phao32 Vị trí khớp robot.
bước/quan sát/joint_vel Tenxơ (7,) phao32 Vận tốc khớp của robot.
bước/quan sát/quang_flow Tenxơ (128, 128, 2) phao32 Luồng quang.
bước/quan sát/trạng thái Tenxơ (số 8,) phao32 Thông tin về khả năng cảm nhận của robot, [7x vị trí khớp, 1x mở/đóng tay cầm].
bước/phần thưởng Vô hướng phao32 Phần thưởng nếu được cung cấp, 1 ở bước cuối cùng cho bản demo.
@inproceedings{lee2019icra,
  title={Making sense of vision and touch: Self-supervised learning of multimodal representations for contact-rich tasks},
  author={Lee, Michelle A and Zhu, Yuke and Srinivasan, Krishnan and Shah, Parth and Savarese, Silvio and Fei-Fei, Li and  Garg, Animesh and Bohg, Jeannette},
  booktitle={2019 IEEE International Conference on Robotics and Automation (ICRA)},
  year={2019},
  url={https://arxiv.org/abs/1810.10191}
}