robosuite_panda_pick_place_can

  • وصف :

تم إنشاء مجموعات البيانات هذه باستخدام بيئة PickPlaceCan الخاصة بمحاكاة الذراع الآلية robosuite . تم تسجيل مجموعات البيانات البشرية بواسطة مشغل واحد باستخدام RLDS Creator ووحدة تحكم لوحة الألعاب.

تم تسجيل مجموعات البيانات الاصطناعية باستخدام مكتبة EnvLogger .

تتبع مجموعات البيانات تنسيق RLDS لتمثيل الخطوات والحلقات.

تتكون الحلقات من 400 خطوة. في كل حلقة، تتم إضافة علامة عند اكتمال المهمة، ويتم تخزين هذه العلامة كجزء من البيانات التعريفية للخطوة المخصصة.

لاحظ أنه نظرًا لتبعية EnvLogger، فإن إنشاء مجموعة البيانات هذه مدعوم حاليًا في بيئات Linux فقط.

@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 (التكوين الافتراضي)

  • وصف التكوين : مجموعة البيانات التي أنشأها الإنسان (50 حلقة).

  • الصفحة الرئيسية : https://github.com/google-research/rlds

  • حجم التحميل : 96.67 MiB

  • حجم مجموعة البيانات : 407.24 MiB

  • الإنشقاقات :

ينقسم أمثلة
'train' 50
  • هيكل الميزة :
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,
    }),
})
  • وثائق الميزة :
ميزة فصل شكل نوع D وصف
المميزاتDict
agent_id الموتر خيط
معرف الحلقة الموتر خيط
Episode_index الموتر int32
خطوات مجموعة البيانات
الخطوات/الإجراء الموتر (7،) float64
الخطوات/الخصم الموتر float64
الخطوات/الصورة صورة (لا شيء، لا شيء، 3) uint8
الخطوات/is_first الموتر منطقي
الخطوات/is_last الموتر منطقي
الخطوات/is_terminal الموتر منطقي
الخطوات/الملاحظة المميزاتDict
الخطوات/الملاحظة/Can_pos الموتر (3،) float64
الخطوات/الملاحظة/Can_quat الموتر (4،) float64
الخطوات/الملاحظة/Can_to_robot0_eef_pos الموتر (3،) float64
الخطوات/الملاحظة/Can_to_robot0_eef_quat الموتر (4،) float32
الخطوات/الملاحظة/حالة الكائن الموتر (14،) float64
الخطوات/الملاحظة/robot0_eef_pos الموتر (3،) float64
الخطوات/الملاحظة/robot0_eef_quat الموتر (4،) float64
الخطوات/الملاحظة/robot0_gripper_qpos الموتر (2،) float64
الخطوات/الملاحظة/robot0_gripper_qvel الموتر (2،) float64
الخطوات/الملاحظة/robot0_joint_pos_cos الموتر (7،) float64
الخطوات/الملاحظة/robot0_joint_pos_sin الموتر (7،) float64
الخطوات/الملاحظة/robot0_joint_vel الموتر (7،) float64
الخطوات/الملاحظة/robot0_proprio-state الموتر (32،) float64
خطوات/مكافأة الموتر float64
الخطوات/العلامة:وضعت الموتر منطقي

robosuite_panda_pick_place_can/human_images_dc29b40a

  • وصف التكوين : مجموعة البيانات التي أنشأها الإنسان، بما في ذلك الصور ذات زوايا الكاميرا المختلفة في المراقبة. لاحظ أن الأمر قد يستغرق بعض الوقت للإنشاء.

  • الصفحة الرئيسية : https://www.tensorflow.org/datasets/catalog/robosuite_panda_pick_place_can

  • حجم التحميل : 10.95 GiB

  • حجم مجموعة البيانات : 7.53 GiB

  • الإنشقاقات :

ينقسم أمثلة
'train' 50
  • هيكل الميزة :
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,
    }),
})
  • وثائق الميزة :
ميزة فصل شكل نوع D وصف
المميزاتDict
agent_id الموتر خيط
معرف الحلقة الموتر خيط
Episode_index الموتر int32
خطوات مجموعة البيانات
الخطوات/الإجراء الموتر (7،) float64
الخطوات/الخصم الموتر float64
الخطوات/الصورة صورة (لا شيء، لا شيء، 3) uint8
الخطوات/is_first الموتر منطقي
الخطوات/is_last الموتر منطقي
الخطوات/is_terminal الموتر منطقي
الخطوات/الملاحظة المميزاتDict
الخطوات/الملاحظة/Can_pos الموتر (3،) float64
الخطوات/الملاحظة/Can_quat الموتر (4،) float64
الخطوات/الملاحظة/Can_to_robot0_eef_pos الموتر (3،) float64
الخطوات/الملاحظة/Can_to_robot0_eef_quat الموتر (4،) float32
الخطوات/الملاحظة/agentview_image صورة (256، 256، 3) uint8
الخطوات/الملاحظة/birdview_image صورة (256، 256، 3) uint8
الخطوات/الملاحظة/حالة الكائن الموتر (14،) float64
الخطوات/الملاحظة/robot0_eef_pos الموتر (3،) float64
الخطوات/الملاحظة/robot0_eef_quat الموتر (4،) float64
الخطوات/الملاحظة/robot0_eye_in_hand_image صورة (256، 256، 3) uint8
الخطوات/الملاحظة/robot0_gripper_qpos الموتر (2،) float64
الخطوات/الملاحظة/robot0_gripper_qvel الموتر (2،) float64
الخطوات/الملاحظة/robot0_joint_pos_cos الموتر (7،) float64
الخطوات/الملاحظة/robot0_joint_pos_sin الموتر (7،) float64
الخطوات/الملاحظة/robot0_joint_vel الموتر (7،) float64
الخطوات/الملاحظة/robot0_proprio-state الموتر (32،) float64
الخطوات/الملاحظة/robot0_robotview_image صورة (256، 256، 3) uint8
خطوات/مكافأة الموتر float64
الخطوات/العلامة:وضعت الموتر منطقي

robosuite_panda_pick_place_can/synthetic_stochastic_sac_afe13968

  • وصف التكوين : مجموعة البيانات الاصطناعية التي تم إنشاؤها بواسطة وكيل عشوائي تم تدريبه باستخدام SAC (200 حلقة).

  • الصفحة الرئيسية : https://github.com/google-research/rlds

  • حجم التحميل : 144.44 MiB

  • حجم مجموعة البيانات : 622.86 MiB

  • الإنشقاقات :

ينقسم أمثلة
'train' 200
  • هيكل الميزة :
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,
    }),
})
  • وثائق الميزة :
ميزة فصل شكل نوع D وصف
المميزاتDict
agent_id الموتر خيط
معرف الحلقة الموتر خيط
Episode_index الموتر int32
خطوات مجموعة البيانات
الخطوات/الإجراء الموتر (7،) float32
الخطوات/الخصم الموتر float64
الخطوات/الصورة صورة (لا شيء، لا شيء، 3) uint8
الخطوات/is_first الموتر منطقي
الخطوات/is_last الموتر منطقي
الخطوات/is_terminal الموتر منطقي
الخطوات/الملاحظة المميزاتDict
الخطوات/الملاحظة/Can_pos الموتر (3،) float32
الخطوات/الملاحظة/Can_quat الموتر (4،) float32
الخطوات/الملاحظة/Can_to_robot0_eef_pos الموتر (3،) float32
الخطوات/الملاحظة/Can_to_robot0_eef_quat الموتر (4،) float32
الخطوات/الملاحظة/حالة الكائن الموتر (14،) float32
الخطوات/الملاحظة/robot0_eef_pos الموتر (3،) float32
الخطوات/الملاحظة/robot0_eef_quat الموتر (4،) float32
الخطوات/الملاحظة/robot0_gripper_qpos الموتر (2،) float32
الخطوات/الملاحظة/robot0_gripper_qvel الموتر (2،) float32
الخطوات/الملاحظة/robot0_joint_pos_cos الموتر (7،) float32
الخطوات/الملاحظة/robot0_joint_pos_sin الموتر (7،) float32
الخطوات/الملاحظة/robot0_joint_vel الموتر (7،) float32
الخطوات/الملاحظة/robot0_proprio-state الموتر (32،) float32
خطوات/مكافأة الموتر float64
الخطوات/العلامة:وضعت الموتر منطقي