berkeley_cable_routing

  • Descripción :

Colocación del cable en abrazaderas sobre la mesa

Dividir Ejemplos
'test' 165
'train' 1.482
  • Estructura de características :
FeaturesDict({
    'steps': Dataset({
        'action': FeaturesDict({
            '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=(128, 128, 3), dtype=uint8),
            'natural_language_embedding': Tensor(shape=(512,), dtype=float32),
            'natural_language_instruction': string,
            'robot_state': Tensor(shape=(7,), dtype=float32),
            'top_image': Image(shape=(128, 128, 3), dtype=uint8),
            'wrist225_image': Image(shape=(128, 128, 3), dtype=uint8),
            'wrist45_image': Image(shape=(128, 128, 3), dtype=uint8),
        }),
        'reward': Scalar(shape=(), dtype=float32),
    }),
})
  • Documentación de funciones :
Característica Clase Forma tipo D Descripción
FuncionesDict
pasos Conjunto de datos
pasos/acción FuncionesDict
pasos/acción/rotation_delta Tensor (3,) flotador32 Velocidad angular con respecto al eje z.
pasos/acción/terminar_episodio Tensor flotador32
pasos/acción/world_vector Tensor (3,) flotador32 Velocidad en XYZ.
pasos/es_primero Tensor booleano
pasos/es_último Tensor booleano
pasos/es_terminal Tensor booleano
pasos/observación FuncionesDict
pasos/observación/imagen Imagen (128, 128, 3) uint8
pasos/observación/incrustación_lenguaje_natural Tensor (512,) flotador32
pasos/observación/instrucción_lenguaje_natural Tensor cadena
pasos/observación/robot_state Tensor (7,) flotador32
pasos/observación/top_image Imagen (128, 128, 3) uint8
pasos/observación/wrist225_image Imagen (128, 128, 3) uint8
pasos/observación/wrist45_image Imagen (128, 128, 3) uint8
pasos/recompensa Escalar flotador32
  • Citación :
@article{luo2023multistage,
  author    = {Jianlan Luo and Charles Xu and Xinyang Geng and Gilbert Feng and Kuan Fang and Liam Tan and Stefan Schaal and Sergey Levine},
  title     = {Multi-Stage Cable Routing through Hierarchical Imitation Learning},
  journal   = {arXiv pre-print},
  year      = {2023},
  url       = {https://arxiv.org/abs/2307.08927},
}