locomozione_rlu

  • Descrizione :

RL Unplugged è una suite di benchmark per l'apprendimento per rinforzo offline. RL Unplugged è progettato sulla base delle seguenti considerazioni: per facilitare la facilità d'uso, forniamo ai set di dati un'API unificata che rende facile per il professionista lavorare con tutti i dati nella suite una volta stabilita una pipeline generale.

I set di dati seguono il formato RLDS per rappresentare passaggi ed episodi.

Queste attività sono costituite dalle attività di locomozione del corridoio che coinvolgono l'umanoide CMU, per le quali gli sforzi precedenti hanno utilizzato i dati di motion capture Merel et al., 2019a , Merel et al., 2019b o l'addestramento da zero Song et al., 2020 . Inoltre, il repository DM Locomotion contiene una serie di compiti adattati per adattarsi a un roditore virtuale Merel et al., 2020 . Sottolineiamo che le attività di locomozione DM presentano la combinazione di un controllo continuo impegnativo ad alto DoF insieme alla percezione da ricche osservazioni egocentriche. Per i dettagli su come è stato generato il set di dati, fare riferimento al documento.

Ti consigliamo di provare i metodi RL offline sul set di dati DeepMind Locomotion, se sei interessato a set di dati RL offline molto impegnativi con spazio di azione continuo.

@inproceedings{gulcehre2020rl,
 title = {RL Unplugged: A Suite of Benchmarks for Offline Reinforcement Learning},
 author = {Gulcehre, Caglar and Wang, Ziyu and Novikov, Alexander and Paine, Thomas and G'{o}mez, Sergio and Zolna, Konrad and Agarwal, Rishabh and Merel, Josh S and Mankowitz, Daniel J and Paduraru, Cosmin and Dulac-Arnold, Gabriel and Li, Jerry and Norouzi, Mohammad and Hoffman, Matthew and Heess, Nicolas and de Freitas, Nando},
 booktitle = {Advances in Neural Information Processing Systems},
 pages = {7248--7259},
 volume = {33},
 year = {2020}
}

rlu_locomotion/humanoid_corridor (configurazione predefinita)

  • Dimensione del set di dati: 1.88 GiB

  • Divisioni :

Diviso Esempi
'train' 4.000
  • Struttura delle caratteristiche :
FeaturesDict({
    'episode_id': int64,
    'steps': Dataset({
        'action': Tensor(shape=(56,), dtype=float32),
        'discount': float32,
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': FeaturesDict({
            'walker': FeaturesDict({
                'body_height': Tensor(shape=(1,), dtype=float32),
                'egocentric_camera': Image(shape=(64, 64, 3), dtype=uint8),
                'end_effectors_pos': Tensor(shape=(12,), dtype=float32),
                'joints_pos': Tensor(shape=(56,), dtype=float32),
                'joints_vel': Tensor(shape=(56,), dtype=float32),
                'sensors_accelerometer': Tensor(shape=(3,), dtype=float32),
                'sensors_gyro': Tensor(shape=(3,), dtype=float32),
                'sensors_velocimeter': Tensor(shape=(3,), dtype=float32),
                'world_zaxis': Tensor(shape=(3,), dtype=float32),
            }),
        }),
        'reward': float32,
    }),
    'timestamp': int64,
})
  • Documentazione delle funzionalità :
Caratteristica Classe Forma Tipo D Descrizione
CaratteristicheDict
episodio_id Tensore int64
passi Set di dati
passi/azione Tensore (56,) galleggiante32
passi/sconto Tensore galleggiante32
passi/è_primo Tensore bool
passi/è_ultimo Tensore bool
passi/è_terminale Tensore bool
passi/osservazione CaratteristicheDict
passi/osservazione/camminatore CaratteristicheDict
passi/osservazione/camminatore/altezza_corpo Tensore (1,) galleggiante32
passi/osservazione/camminatore/egocentric_camera Immagine (64, 64, 3) uint8
passi/osservazione/camminatore/end_effectors_pos Tensore (12,) galleggiante32
passi/osservazione/camminatore/articolazioni_pos Tensore (56,) galleggiante32
passi/osservazione/camminatore/articolazioni_vel Tensore (56,) galleggiante32
passi/osservazione/camminatore/sensori_accelerometro Tensore (3,) galleggiante32
passi/osservazione/walker/sensors_gyro Tensore (3,) galleggiante32
passi/osservazione/walker/sensors_velocimeter Tensore (3,) galleggiante32
passi/osservazione/walker/world_zaxis Tensore (3,) galleggiante32
passi/ricompensa Tensore galleggiante32
timestamp Tensore int64

rlu_locomotion/humanoid_gaps

  • Dimensione del set di dati: 4.57 GiB

  • Divisioni :

Diviso Esempi
'train' 8.000
  • Struttura delle caratteristiche :
FeaturesDict({
    'episode_id': int64,
    'steps': Dataset({
        'action': Tensor(shape=(56,), dtype=float32),
        'discount': float32,
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': FeaturesDict({
            'walker': FeaturesDict({
                'body_height': Tensor(shape=(1,), dtype=float32),
                'egocentric_camera': Image(shape=(64, 64, 3), dtype=uint8),
                'end_effectors_pos': Tensor(shape=(12,), dtype=float32),
                'joints_pos': Tensor(shape=(56,), dtype=float32),
                'joints_vel': Tensor(shape=(56,), dtype=float32),
                'sensors_accelerometer': Tensor(shape=(3,), dtype=float32),
                'sensors_gyro': Tensor(shape=(3,), dtype=float32),
                'sensors_velocimeter': Tensor(shape=(3,), dtype=float32),
                'world_zaxis': Tensor(shape=(3,), dtype=float32),
            }),
        }),
        'reward': float32,
    }),
    'timestamp': int64,
})
  • Documentazione delle funzionalità :
Caratteristica Classe Forma Tipo D Descrizione
CaratteristicheDict
episodio_id Tensore int64
passi Set di dati
passi/azione Tensore (56,) galleggiante32
passi/sconto Tensore galleggiante32
passi/è_primo Tensore bool
passi/è_ultimo Tensore bool
passi/è_terminale Tensore bool
passi/osservazione CaratteristicheDict
passi/osservazione/camminatore CaratteristicheDict
passi/osservazione/camminatore/altezza_corpo Tensore (1,) galleggiante32
passi/osservazione/camminatore/egocentric_camera Immagine (64, 64, 3) uint8
passi/osservazione/camminatore/end_effectors_pos Tensore (12,) galleggiante32
passi/osservazione/camminatore/articolazioni_pos Tensore (56,) galleggiante32
passi/osservazione/camminatore/articolazioni_vel Tensore (56,) galleggiante32
passi/osservazione/camminatore/sensori_accelerometro Tensore (3,) galleggiante32
passi/osservazione/walker/sensors_gyro Tensore (3,) galleggiante32
passi/osservazione/walker/sensors_velocimeter Tensore (3,) galleggiante32
passi/osservazione/walker/world_zaxis Tensore (3,) galleggiante32
passi/ricompensa Tensore galleggiante32
timestamp Tensore int64

rlu_locomotion/humanoid_walls

  • Dimensione del set di dati: 2.36 GiB

  • Divisioni :

Diviso Esempi
'train' 4.000
  • Struttura delle caratteristiche :
FeaturesDict({
    'episode_id': int64,
    'steps': Dataset({
        'action': Tensor(shape=(56,), dtype=float32),
        'discount': float32,
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': FeaturesDict({
            'walker': FeaturesDict({
                'body_height': Tensor(shape=(1,), dtype=float32),
                'egocentric_camera': Image(shape=(64, 64, 3), dtype=uint8),
                'end_effectors_pos': Tensor(shape=(12,), dtype=float32),
                'joints_pos': Tensor(shape=(56,), dtype=float32),
                'joints_vel': Tensor(shape=(56,), dtype=float32),
                'sensors_accelerometer': Tensor(shape=(3,), dtype=float32),
                'sensors_gyro': Tensor(shape=(3,), dtype=float32),
                'sensors_velocimeter': Tensor(shape=(3,), dtype=float32),
                'world_zaxis': Tensor(shape=(3,), dtype=float32),
            }),
        }),
        'reward': float32,
    }),
    'timestamp': int64,
})
  • Documentazione delle funzionalità :
Caratteristica Classe Forma Tipo D Descrizione
CaratteristicheDict
episodio_id Tensore int64
passi Set di dati
passi/azione Tensore (56,) galleggiante32
passi/sconto Tensore galleggiante32
passi/è_primo Tensore bool
passi/è_ultimo Tensore bool
passi/è_terminale Tensore bool
passi/osservazione CaratteristicheDict
passi/osservazione/camminatore CaratteristicheDict
passi/osservazione/camminatore/altezza_corpo Tensore (1,) galleggiante32
passi/osservazione/camminatore/egocentric_camera Immagine (64, 64, 3) uint8
passi/osservazione/camminatore/end_effectors_pos Tensore (12,) galleggiante32
passi/osservazione/camminatore/articolazioni_pos Tensore (56,) galleggiante32
passi/osservazione/camminatore/articolazioni_vel Tensore (56,) galleggiante32
passi/osservazione/camminatore/sensori_accelerometro Tensore (3,) galleggiante32
passi/osservazione/walker/sensors_gyro Tensore (3,) galleggiante32
passi/osservazione/walker/sensors_velocimeter Tensore (3,) galleggiante32
passi/osservazione/walker/world_zaxis Tensore (3,) galleggiante32
passi/ricompensa Tensore galleggiante32
timestamp Tensore int64

rlu_locomotion/rodent_bowl_escape

  • Dimensione del set di dati: 16.46 GiB

  • Divisioni :

Diviso Esempi
'train' 2.000
  • Struttura delle caratteristiche :
FeaturesDict({
    'episode_id': int64,
    'steps': Dataset({
        'action': Tensor(shape=(38,), dtype=float32),
        'discount': float32,
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': FeaturesDict({
            'walker': FeaturesDict({
                'appendages_pos': Tensor(shape=(15,), dtype=float32),
                'egocentric_camera': Image(shape=(64, 64, 3), dtype=uint8),
                'joints_pos': Tensor(shape=(30,), dtype=float32),
                'joints_vel': Tensor(shape=(30,), dtype=float32),
                'sensors_accelerometer': Tensor(shape=(3,), dtype=float32),
                'sensors_gyro': Tensor(shape=(3,), dtype=float32),
                'sensors_touch': Tensor(shape=(4,), dtype=float32),
                'sensors_velocimeter': Tensor(shape=(3,), dtype=float32),
                'tendons_pos': Tensor(shape=(8,), dtype=float32),
                'tendons_vel': Tensor(shape=(8,), dtype=float32),
                'world_zaxis': Tensor(shape=(3,), dtype=float32),
            }),
        }),
        'reward': float32,
    }),
    'timestamp': int64,
})
  • Documentazione delle funzionalità :
Caratteristica Classe Forma Tipo D Descrizione
CaratteristicheDict
episodio_id Tensore int64
passi Set di dati
passi/azione Tensore (38,) galleggiante32
passi/sconto Tensore galleggiante32
passi/è_primo Tensore bool
passi/è_ultimo Tensore bool
passi/è_terminale Tensore bool
passi/osservazione CaratteristicheDict
passi/osservazione/camminatore CaratteristicheDict
passi/osservazione/camminatore/appendici_pos Tensore (15,) galleggiante32
passi/osservazione/camminatore/egocentric_camera Immagine (64, 64, 3) uint8
passi/osservazione/camminatore/articolazioni_pos Tensore (30,) galleggiante32
passi/osservazione/camminatore/articolazioni_vel Tensore (30,) galleggiante32
passi/osservazione/camminatore/sensori_accelerometro Tensore (3,) galleggiante32
passi/osservazione/walker/sensors_gyro Tensore (3,) galleggiante32
passi/osservazione/camminatore/sensori_tocco Tensore (4,) galleggiante32
passi/osservazione/walker/sensors_velocimeter Tensore (3,) galleggiante32
passi/osservazione/deambulatore/tendons_pos Tensore (8,) galleggiante32
passi/osservazione/camminatore/tendini_vel Tensore (8,) galleggiante32
passi/osservazione/walker/world_zaxis Tensore (3,) galleggiante32
passi/ricompensa Tensore galleggiante32
timestamp Tensore int64

rlu_locomotion/rodent_gaps

  • Dimensione del set di dati: 8.90 GiB

  • Divisioni :

Diviso Esempi
'train' 2.000
  • Struttura delle caratteristiche :
FeaturesDict({
    'episode_id': int64,
    'steps': Dataset({
        'action': Tensor(shape=(38,), dtype=float32),
        'discount': float32,
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': FeaturesDict({
            'walker': FeaturesDict({
                'appendages_pos': Tensor(shape=(15,), dtype=float32),
                'egocentric_camera': Image(shape=(64, 64, 3), dtype=uint8),
                'joints_pos': Tensor(shape=(30,), dtype=float32),
                'joints_vel': Tensor(shape=(30,), dtype=float32),
                'sensors_accelerometer': Tensor(shape=(3,), dtype=float32),
                'sensors_gyro': Tensor(shape=(3,), dtype=float32),
                'sensors_touch': Tensor(shape=(4,), dtype=float32),
                'sensors_velocimeter': Tensor(shape=(3,), dtype=float32),
                'tendons_pos': Tensor(shape=(8,), dtype=float32),
                'tendons_vel': Tensor(shape=(8,), dtype=float32),
                'world_zaxis': Tensor(shape=(3,), dtype=float32),
            }),
        }),
        'reward': float32,
    }),
    'timestamp': int64,
})
  • Documentazione delle funzionalità :
Caratteristica Classe Forma Tipo D Descrizione
CaratteristicheDict
episodio_id Tensore int64
passi Set di dati
passi/azione Tensore (38,) galleggiante32
passi/sconto Tensore galleggiante32
passi/è_primo Tensore bool
passi/è_ultimo Tensore bool
passi/è_terminale Tensore bool
passi/osservazione CaratteristicheDict
passi/osservazione/camminatore CaratteristicheDict
passi/osservazione/camminatore/appendici_pos Tensore (15,) galleggiante32
passi/osservazione/camminatore/egocentric_camera Immagine (64, 64, 3) uint8
passi/osservazione/camminatore/articolazioni_pos Tensore (30,) galleggiante32
passi/osservazione/camminatore/articolazioni_vel Tensore (30,) galleggiante32
passi/osservazione/camminatore/sensori_accelerometro Tensore (3,) galleggiante32
passi/osservazione/walker/sensors_gyro Tensore (3,) galleggiante32
passi/osservazione/camminatore/sensori_tocco Tensore (4,) galleggiante32
passi/osservazione/walker/sensors_velocimeter Tensore (3,) galleggiante32
passi/osservazione/deambulatore/tendons_pos Tensore (8,) galleggiante32
passi/osservazione/camminatore/tendini_vel Tensore (8,) galleggiante32
passi/osservazione/walker/world_zaxis Tensore (3,) galleggiante32
passi/ricompensa Tensore galleggiante32
timestamp Tensore int64

rlu_locomotion/rodent_mazes

  • Dimensione del set di dati: 20.71 GiB

  • Divisioni :

Diviso Esempi
'train' 2.000
  • Struttura delle caratteristiche :
FeaturesDict({
    'episode_id': int64,
    'steps': Dataset({
        'action': Tensor(shape=(38,), dtype=float32),
        'discount': float32,
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': FeaturesDict({
            'walker': FeaturesDict({
                'appendages_pos': Tensor(shape=(15,), dtype=float32),
                'egocentric_camera': Image(shape=(64, 64, 3), dtype=uint8),
                'joints_pos': Tensor(shape=(30,), dtype=float32),
                'joints_vel': Tensor(shape=(30,), dtype=float32),
                'sensors_accelerometer': Tensor(shape=(3,), dtype=float32),
                'sensors_gyro': Tensor(shape=(3,), dtype=float32),
                'sensors_touch': Tensor(shape=(4,), dtype=float32),
                'sensors_velocimeter': Tensor(shape=(3,), dtype=float32),
                'tendons_pos': Tensor(shape=(8,), dtype=float32),
                'tendons_vel': Tensor(shape=(8,), dtype=float32),
                'world_zaxis': Tensor(shape=(3,), dtype=float32),
            }),
        }),
        'reward': float32,
    }),
    'timestamp': int64,
})
  • Documentazione delle funzionalità :
Caratteristica Classe Forma Tipo D Descrizione
CaratteristicheDict
episodio_id Tensore int64
passi Set di dati
passi/azione Tensore (38,) galleggiante32
passi/sconto Tensore galleggiante32
passi/è_primo Tensore bool
passi/è_ultimo Tensore bool
passi/è_terminale Tensore bool
passi/osservazione CaratteristicheDict
passi/osservazione/camminatore CaratteristicheDict
passi/osservazione/camminatore/appendici_pos Tensore (15,) galleggiante32
passi/osservazione/camminatore/egocentric_camera Immagine (64, 64, 3) uint8
passi/osservazione/camminatore/articolazioni_pos Tensore (30,) galleggiante32
passi/osservazione/camminatore/articolazioni_vel Tensore (30,) galleggiante32
passi/osservazione/camminatore/sensori_accelerometro Tensore (3,) galleggiante32
passi/osservazione/walker/sensors_gyro Tensore (3,) galleggiante32
passi/osservazione/camminatore/sensori_tocco Tensore (4,) galleggiante32
passi/osservazione/walker/sensors_velocimeter Tensore (3,) galleggiante32
passi/osservazione/deambulatore/tendons_pos Tensore (8,) galleggiante32
passi/osservazione/camminatore/tendini_vel Tensore (8,) galleggiante32
passi/osservazione/walker/world_zaxis Tensore (3,) galleggiante32
passi/ricompensa Tensore galleggiante32
timestamp Tensore int64

rlu_locomotion/rodent_two_touch

  • Dimensione del set di dati: 23.05 GiB

  • Divisioni :

Diviso Esempi
'train' 2.000
  • Struttura delle caratteristiche :
FeaturesDict({
    'episode_id': int64,
    'steps': Dataset({
        'action': Tensor(shape=(38,), dtype=float32),
        'discount': float32,
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': FeaturesDict({
            'walker': FeaturesDict({
                'appendages_pos': Tensor(shape=(15,), dtype=float32),
                'egocentric_camera': Image(shape=(64, 64, 3), dtype=uint8),
                'joints_pos': Tensor(shape=(30,), dtype=float32),
                'joints_vel': Tensor(shape=(30,), dtype=float32),
                'sensors_accelerometer': Tensor(shape=(3,), dtype=float32),
                'sensors_gyro': Tensor(shape=(3,), dtype=float32),
                'sensors_touch': Tensor(shape=(4,), dtype=float32),
                'sensors_velocimeter': Tensor(shape=(3,), dtype=float32),
                'tendons_pos': Tensor(shape=(8,), dtype=float32),
                'tendons_vel': Tensor(shape=(8,), dtype=float32),
                'world_zaxis': Tensor(shape=(3,), dtype=float32),
            }),
        }),
        'reward': float32,
    }),
    'timestamp': int64,
})
  • Documentazione delle funzionalità :
Caratteristica Classe Forma Tipo D Descrizione
CaratteristicheDict
episodio_id Tensore int64
passi Set di dati
passi/azione Tensore (38,) galleggiante32
passi/sconto Tensore galleggiante32
passi/è_primo Tensore bool
passi/è_ultimo Tensore bool
passi/è_terminale Tensore bool
passi/osservazione CaratteristicheDict
passi/osservazione/camminatore CaratteristicheDict
passi/osservazione/camminatore/appendici_pos Tensore (15,) galleggiante32
passi/osservazione/camminatore/egocentric_camera Immagine (64, 64, 3) uint8
passi/osservazione/camminatore/articolazioni_pos Tensore (30,) galleggiante32
passi/osservazione/camminatore/articolazioni_vel Tensore (30,) galleggiante32
passi/osservazione/camminatore/sensori_accelerometro Tensore (3,) galleggiante32
passi/osservazione/walker/sensors_gyro Tensore (3,) galleggiante32
passi/osservazione/camminatore/sensori_tocco Tensore (4,) galleggiante32
passi/osservazione/walker/sensors_velocimeter Tensore (3,) galleggiante32
passi/osservazione/deambulatore/tendons_pos Tensore (8,) galleggiante32
passi/osservazione/camminatore/tendini_vel Tensore (8,) galleggiante32
passi/osservazione/walker/world_zaxis Tensore (3,) galleggiante32
passi/ricompensa Tensore galleggiante32
timestamp Tensore int64