toto

  • Keterangan :

Franka menyendok dan menuangkan tugas

Membelah Contoh
  • Struktur fitur :
FeaturesDict({
    'steps': Dataset({
        'action': FeaturesDict({
            'open_gripper': bool,
            'rotation_delta': Tensor(shape=(3,), dtype=float32),
            'terminate_episode': float32,
            'world_vector': Tensor(shape=(3,), dtype=float32),
        }),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': FeaturesDict({
            'image': Image(shape=(480, 640, 3), dtype=uint8),
            'natural_language_embedding': Tensor(shape=(512,), dtype=float32),
            'natural_language_instruction': string,
            'state': Tensor(shape=(7,), dtype=float32, description=numpy array of shape (7,). Contains the robot joint states (as absolute joint angles) at each timestep),
        }),
        'reward': Scalar(shape=(), dtype=float32),
    }),
})
  • Dokumentasi fitur :
Fitur Kelas Membentuk Tipe D Keterangan
FiturDict
tangga Kumpulan data
langkah/tindakan FiturDict
langkah/tindakan/open_gripper Tensor bodoh
langkah/tindakan/rotasi_delta Tensor (3,) float32
langkah/tindakan/terminate_episode Tensor float32
langkah/tindakan/world_vector Tensor (3,) float32
langkah/adalah_pertama Tensor bodoh
langkah/adalah_terakhir Tensor bodoh
langkah/is_terminal Tensor bodoh
langkah/pengamatan FiturDict
langkah/pengamatan/gambar Gambar (480, 640, 3) uint8
langkah/pengamatan/penyematan_bahasa_alami Tensor (512,) float32
langkah/pengamatan/instruksi_bahasa_alami Tensor rangkaian
langkah/pengamatan/keadaan Tensor (7,) float32 susunan bentuk numpy (7,). Berisi status sambungan robot (sebagai sudut sambungan absolut) pada setiap langkah waktu
langkah/hadiah Skalar float32
@inproceedings{zhou2023train,
  author={Zhou, Gaoyue and Dean, Victoria and Srirama, Mohan Kumar and Rajeswaran, Aravind and Pari, Jyothish and Hatch, Kyle and Jain, Aryan and Yu, Tianhe and Abbeel, Pieter and Pinto, Lerrel and Finn, Chelsea and Gupta, Abhinav},
  booktitle={2023 IEEE International Conference on Robotics and Automation (ICRA)},
  title={Train Offline, Test Online: A Real Robot Learning Benchmark},
  year={2023},
 }