robosuite_panda_pick_place_can

  • תיאור :

מערכי נתונים אלה נוצרו עם סביבת PickPlaceCan של סימולטור הזרוע הרובוטי של robosuite. מערכי הנתונים האנושיים תועדו על ידי מפעיל יחיד באמצעות ה- RLDS Creator ובקר gamepad.

מערכי הנתונים הסינתטיים תועדו באמצעות ספריית EnvLogger .

מערכי הנתונים פועלים לפי פורמט RLDS כדי לייצג שלבים ופרקים.

פרקים מורכבים מ-400 שלבים. בכל פרק, תג מתווסף עם השלמת המשימה, תג זה מאוחסן כחלק מהמטא נתונים של הצעד המותאם אישית.

שים לב שבגלל התלות ב-EnvLogger, יצירת מערך הנתונים הזה נתמך כרגע רק בסביבות לינוקס.

@misc{ramos2021rlds,
      title={RLDS: an Ecosystem to Generate, Share and Use Datasets in Reinforcement Learning},
      author={Sabela Ramos and Sertan Girgin and Léonard Hussenot and Damien Vincent and Hanna Yakubovich and Daniel Toyama and Anita Gergely and Piotr Stanczyk and Raphael Marinier and Jeremiah Harmsen and Olivier Pietquin and Nikola Momchev},
      year={2021},
      eprint={2111.02767},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}

robosuite_panda_pick_place_can/human_dc29b40a (תצורת ברירת המחדל)

  • תיאור תצורה : מערך נתונים שנוצר על ידי אדם (50 פרקים).

  • דף הבית : https://github.com/google-research/rlds

  • גודל הורדה : 96.67 MiB

  • גודל ערכת נתונים: 407.24 MiB

  • פיצולים :

לְפַצֵל דוגמאות
'train' 50
  • מבנה תכונה :
FeaturesDict({
    'agent_id': string,
    'episode_id': string,
    'episode_index': int32,
    'steps': Dataset({
        'action': Tensor(shape=(7,), dtype=float64),
        'discount': float64,
        'image': Image(shape=(None, None, 3), dtype=uint8),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': FeaturesDict({
            'Can_pos': Tensor(shape=(3,), dtype=float64),
            'Can_quat': Tensor(shape=(4,), dtype=float64),
            'Can_to_robot0_eef_pos': Tensor(shape=(3,), dtype=float64),
            'Can_to_robot0_eef_quat': Tensor(shape=(4,), dtype=float32),
            'object-state': Tensor(shape=(14,), dtype=float64),
            'robot0_eef_pos': Tensor(shape=(3,), dtype=float64),
            'robot0_eef_quat': Tensor(shape=(4,), dtype=float64),
            'robot0_gripper_qpos': Tensor(shape=(2,), dtype=float64),
            'robot0_gripper_qvel': Tensor(shape=(2,), dtype=float64),
            'robot0_joint_pos_cos': Tensor(shape=(7,), dtype=float64),
            'robot0_joint_pos_sin': Tensor(shape=(7,), dtype=float64),
            'robot0_joint_vel': Tensor(shape=(7,), dtype=float64),
            'robot0_proprio-state': Tensor(shape=(32,), dtype=float64),
        }),
        'reward': float64,
        'tag:placed': bool,
    }),
})
  • תיעוד תכונה :
תכונה מעמד צוּרָה Dtype תיאור
FeaturesDict
agent_id מוֹתֵחַ חוּט
פרק_מזהה מוֹתֵחַ חוּט
episode_index מוֹתֵחַ int32
צעדים מערך נתונים
צעדים/פעולה מוֹתֵחַ (7,) לצוף64
צעדים/הנחה מוֹתֵחַ לצוף64
שלבים/תמונה תמונה (אין, אין, 3) uint8
צעדים/הוא_ראשון מוֹתֵחַ bool
צעדים/הוא_אחרון מוֹתֵחַ bool
steps/is_terminal מוֹתֵחַ bool
צעדים/תצפית FeaturesDict
צעדים/תצפית/Can_pos מוֹתֵחַ (3,) לצוף64
צעדים/תצפית/Can_quat מוֹתֵחַ (4,) לצוף64
צעדים/תצפית/Can_to_robot0_eef_pos מוֹתֵחַ (3,) לצוף64
צעדים/תצפית/Can_to_robot0_eef_quat מוֹתֵחַ (4,) לצוף32
צעדים/תצפית/מצב-אובייקט מוֹתֵחַ (14,) לצוף64
צעדים/תצפית/רובוט0_eef_pos מוֹתֵחַ (3,) לצוף64
צעדים/תצפית/רובוט0_eef_quat מוֹתֵחַ (4,) לצוף64
צעדים/תצפית/רובוט0_gripper_qpos מוֹתֵחַ (2,) לצוף64
צעדים/תצפית/רובוט0_gripper_qvel מוֹתֵחַ (2,) לצוף64
צעדים/תצפית/רובוט0_joint_pos_cos מוֹתֵחַ (7,) לצוף64
צעדים/תצפית/רובוט0_joint_pos_sin מוֹתֵחַ (7,) לצוף64
צעדים/תצפית/רובוט0_joint_vel מוֹתֵחַ (7,) לצוף64
צעדים/תצפית/רובוט0_proprio-state מוֹתֵחַ (32,) לצוף64
צעדים/פרס מוֹתֵחַ לצוף64
steps/tag:placed מוֹתֵחַ bool

robosuite_panda_pick_place_can/human_images_dc29b40a

  • תיאור תצורה : מערך נתונים שנוצר על ידי אדם, כולל תמונות עם זוויות מצלמה שונות בתצפית. שימו לב שייתכן שייקח זמן מה להפיק.

  • דף הבית : https://www.tensorflow.org/datasets/catalog/robosuite_panda_pick_place_can

  • גודל הורדה : 10.95 GiB

  • גודל מערך נתונים : 7.53 GiB

  • פיצולים :

לְפַצֵל דוגמאות
'train' 50
  • מבנה תכונה :
FeaturesDict({
    'agent_id': string,
    'episode_id': string,
    'episode_index': int32,
    'steps': Dataset({
        'action': Tensor(shape=(7,), dtype=float64),
        'discount': float64,
        'image': Image(shape=(None, None, 3), dtype=uint8),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': FeaturesDict({
            'Can_pos': Tensor(shape=(3,), dtype=float64),
            'Can_quat': Tensor(shape=(4,), dtype=float64),
            'Can_to_robot0_eef_pos': Tensor(shape=(3,), dtype=float64),
            'Can_to_robot0_eef_quat': Tensor(shape=(4,), dtype=float32),
            'agentview_image': Image(shape=(256, 256, 3), dtype=uint8),
            'birdview_image': Image(shape=(256, 256, 3), dtype=uint8),
            'object-state': Tensor(shape=(14,), dtype=float64),
            'robot0_eef_pos': Tensor(shape=(3,), dtype=float64),
            'robot0_eef_quat': Tensor(shape=(4,), dtype=float64),
            'robot0_eye_in_hand_image': Image(shape=(256, 256, 3), dtype=uint8),
            'robot0_gripper_qpos': Tensor(shape=(2,), dtype=float64),
            'robot0_gripper_qvel': Tensor(shape=(2,), dtype=float64),
            'robot0_joint_pos_cos': Tensor(shape=(7,), dtype=float64),
            'robot0_joint_pos_sin': Tensor(shape=(7,), dtype=float64),
            'robot0_joint_vel': Tensor(shape=(7,), dtype=float64),
            'robot0_proprio-state': Tensor(shape=(32,), dtype=float64),
            'robot0_robotview_image': Image(shape=(256, 256, 3), dtype=uint8),
        }),
        'reward': float64,
        'tag:placed': bool,
    }),
})
  • תיעוד תכונה :
תכונה מעמד צוּרָה Dtype תיאור
FeaturesDict
agent_id מוֹתֵחַ חוּט
פרק_מזהה מוֹתֵחַ חוּט
episode_index מוֹתֵחַ int32
צעדים מערך נתונים
צעדים/פעולה מוֹתֵחַ (7,) לצוף64
צעדים/הנחה מוֹתֵחַ לצוף64
שלבים/תמונה תמונה (אין, אין, 3) uint8
צעדים/הוא_ראשון מוֹתֵחַ bool
צעדים/הוא_אחרון מוֹתֵחַ bool
steps/is_terminal מוֹתֵחַ bool
צעדים/תצפית FeaturesDict
צעדים/תצפית/Can_pos מוֹתֵחַ (3,) לצוף64
צעדים/תצפית/Can_quat מוֹתֵחַ (4,) לצוף64
צעדים/תצפית/Can_to_robot0_eef_pos מוֹתֵחַ (3,) לצוף64
צעדים/תצפית/Can_to_robot0_eef_quat מוֹתֵחַ (4,) לצוף32
steps/observation/agentview_image תמונה (256, 256, 3) uint8
צעדים/תצפית/תמונת_ציפור תמונה (256, 256, 3) uint8
צעדים/תצפית/מצב-אובייקט מוֹתֵחַ (14,) לצוף64
צעדים/תצפית/רובוט0_eef_pos מוֹתֵחַ (3,) לצוף64
צעדים/תצפית/רובוט0_eef_quat מוֹתֵחַ (4,) לצוף64
צעדים/תצפית/רובוט0_עין_בתמונה_יד תמונה (256, 256, 3) uint8
צעדים/תצפית/רובוט0_gripper_qpos מוֹתֵחַ (2,) לצוף64
צעדים/תצפית/רובוט0_gripper_qvel מוֹתֵחַ (2,) לצוף64
צעדים/תצפית/רובוט0_joint_pos_cos מוֹתֵחַ (7,) לצוף64
צעדים/תצפית/רובוט0_joint_pos_sin מוֹתֵחַ (7,) לצוף64
צעדים/תצפית/רובוט0_joint_vel מוֹתֵחַ (7,) לצוף64
צעדים/תצפית/רובוט0_proprio-state מוֹתֵחַ (32,) לצוף64
steps/observation/robot0_robotview_image תמונה (256, 256, 3) uint8
צעדים/פרס מוֹתֵחַ לצוף64
steps/tag:placed מוֹתֵחַ bool

robosuite_panda_pick_place_can/synthetic_stochastic_sac_afe13968

  • תיאור תצורה : מערך נתונים סינתטי שנוצר על ידי סוכן סטוכסטי מאומן עם SAC (200 פרקים).

  • דף הבית : https://github.com/google-research/rlds

  • גודל הורדה : 144.44 MiB

  • גודל ערכת נתונים: 622.86 MiB

  • פיצולים :

לְפַצֵל דוגמאות
'train' 200
  • מבנה תכונה :
FeaturesDict({
    'agent_id': string,
    'episode_id': string,
    'episode_index': int32,
    'steps': Dataset({
        'action': Tensor(shape=(7,), dtype=float32),
        'discount': float64,
        'image': Image(shape=(None, None, 3), dtype=uint8),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'observation': FeaturesDict({
            'Can_pos': Tensor(shape=(3,), dtype=float32),
            'Can_quat': Tensor(shape=(4,), dtype=float32),
            'Can_to_robot0_eef_pos': Tensor(shape=(3,), dtype=float32),
            'Can_to_robot0_eef_quat': Tensor(shape=(4,), dtype=float32),
            'object-state': Tensor(shape=(14,), dtype=float32),
            'robot0_eef_pos': Tensor(shape=(3,), dtype=float32),
            'robot0_eef_quat': Tensor(shape=(4,), dtype=float32),
            'robot0_gripper_qpos': Tensor(shape=(2,), dtype=float32),
            'robot0_gripper_qvel': Tensor(shape=(2,), dtype=float32),
            'robot0_joint_pos_cos': Tensor(shape=(7,), dtype=float32),
            'robot0_joint_pos_sin': Tensor(shape=(7,), dtype=float32),
            'robot0_joint_vel': Tensor(shape=(7,), dtype=float32),
            'robot0_proprio-state': Tensor(shape=(32,), dtype=float32),
        }),
        'reward': float64,
        'tag:placed': bool,
    }),
})
  • תיעוד תכונה :
תכונה מעמד צוּרָה Dtype תיאור
FeaturesDict
agent_id מוֹתֵחַ חוּט
פרק_מזהה מוֹתֵחַ חוּט
episode_index מוֹתֵחַ int32
צעדים מערך נתונים
צעדים/פעולה מוֹתֵחַ (7,) לצוף32
צעדים/הנחה מוֹתֵחַ לצוף64
שלבים/תמונה תמונה (אין, אין, 3) uint8
צעדים/הוא_ראשון מוֹתֵחַ bool
צעדים/הוא_אחרון מוֹתֵחַ bool
steps/is_terminal מוֹתֵחַ bool
צעדים/תצפית FeaturesDict
צעדים/תצפית/Can_pos מוֹתֵחַ (3,) לצוף32
צעדים/תצפית/Can_quat מוֹתֵחַ (4,) לצוף32
צעדים/תצפית/Can_to_robot0_eef_pos מוֹתֵחַ (3,) לצוף32
צעדים/תצפית/Can_to_robot0_eef_quat מוֹתֵחַ (4,) לצוף32
צעדים/תצפית/מצב-אובייקט מוֹתֵחַ (14,) לצוף32
צעדים/תצפית/רובוט0_eef_pos מוֹתֵחַ (3,) לצוף32
צעדים/תצפית/רובוט0_eef_quat מוֹתֵחַ (4,) לצוף32
צעדים/תצפית/רובוט0_gripper_qpos מוֹתֵחַ (2,) לצוף32
צעדים/תצפית/רובוט0_gripper_qvel מוֹתֵחַ (2,) לצוף32
צעדים/תצפית/רובוט0_joint_pos_cos מוֹתֵחַ (7,) לצוף32
צעדים/תצפית/רובוט0_joint_pos_sin מוֹתֵחַ (7,) לצוף32
צעדים/תצפית/רובוט0_joint_vel מוֹתֵחַ (7,) לצוף32
צעדים/תצפית/רובוט0_proprio-state מוֹתֵחַ (32,) לצוף32
צעדים/פרס מוֹתֵחַ לצוף64
steps/tag:placed מוֹתֵחַ bool