robosuite_panda_pick_place_can

  • Description:

These datasets have been created with the PickPlaceCan environment of the robosuite robotic arm simulator. The human datasets were recorded by a single operator using the RLDS Creator and a gamepad controller.

The synthetic datasets have been recorded using the EnvLogger library.

Episodes consist of 400 steps. In each episode, a tag is added when the task is completed, this tag is stored as part of the custom step metadata.

@misc{google-research, title={RLDS},
url={https://github.com/google-research/rlds}, journal={GitHub},
author={S. Ramos, S. Girgin et al.} }

robosuite_panda_pick_place_can/human_dc29b40a (default config)

  • Config description: Human generated dataset (50 episodes).

  • Download size: 96.67 MiB

  • Dataset size: 407.24 MiB

  • Splits:

Split Examples
'train' 50
  • Features:
FeaturesDict({
    'agent_id': tf.string,
    'episode_id': tf.string,
    'episode_index': tf.int32,
    'steps': Dataset({
        'action': Tensor(shape=(7,), dtype=tf.float64),
        'discount': tf.float64,
        'image': Image(shape=(None, None, 3), dtype=tf.uint8),
        'is_first': tf.bool,
        'is_last': tf.bool,
        'is_terminal': tf.bool,
        'observation': FeaturesDict({
            'Can_pos': Tensor(shape=(3,), dtype=tf.float64),
            'Can_quat': Tensor(shape=(4,), dtype=tf.float64),
            'Can_to_robot0_eef_pos': Tensor(shape=(3,), dtype=tf.float64),
            'Can_to_robot0_eef_quat': Tensor(shape=(4,), dtype=tf.float32),
            'object-state': Tensor(shape=(14,), dtype=tf.float64),
            'robot0_eef_pos': Tensor(shape=(3,), dtype=tf.float64),
            'robot0_eef_quat': Tensor(shape=(4,), dtype=tf.float64),
            'robot0_gripper_qpos': Tensor(shape=(2,), dtype=tf.float64),
            'robot0_gripper_qvel': Tensor(shape=(2,), dtype=tf.float64),
            'robot0_joint_pos_cos': Tensor(shape=(7,), dtype=tf.float64),
            'robot0_joint_pos_sin': Tensor(shape=(7,), dtype=tf.float64),
            'robot0_joint_vel': Tensor(shape=(7,), dtype=tf.float64),
            'robot0_proprio-state': Tensor(shape=(32,), dtype=tf.float64),
        }),
        'reward': tf.float64,
        'tag:placed': tf.bool,
    }),
})

robosuite_panda_pick_place_can/synthetic_stochastic_sac_afe13968

  • Config description: Synthetic dataset generated by a stochastic agent trained with SAC (200 episodes).

  • Download size: 144.44 MiB

  • Dataset size: 622.86 MiB

  • Splits:

Split Examples
'train' 200
  • Features:
FeaturesDict({
    'agent_id': tf.string,
    'episode_id': tf.string,
    'episode_index': tf.int32,
    'steps': Dataset({
        'action': Tensor(shape=(7,), dtype=tf.float32),
        'discount': tf.float64,
        'image': Image(shape=(None, None, 3), dtype=tf.uint8),
        'is_first': tf.bool,
        'is_last': tf.bool,
        'is_terminal': tf.bool,
        'observation': FeaturesDict({
            'Can_pos': Tensor(shape=(3,), dtype=tf.float32),
            'Can_quat': Tensor(shape=(4,), dtype=tf.float32),
            'Can_to_robot0_eef_pos': Tensor(shape=(3,), dtype=tf.float32),
            'Can_to_robot0_eef_quat': Tensor(shape=(4,), dtype=tf.float32),
            'object-state': Tensor(shape=(14,), dtype=tf.float32),
            'robot0_eef_pos': Tensor(shape=(3,), dtype=tf.float32),
            'robot0_eef_quat': Tensor(shape=(4,), dtype=tf.float32),
            'robot0_gripper_qpos': Tensor(shape=(2,), dtype=tf.float32),
            'robot0_gripper_qvel': Tensor(shape=(2,), dtype=tf.float32),
            'robot0_joint_pos_cos': Tensor(shape=(7,), dtype=tf.float32),
            'robot0_joint_pos_sin': Tensor(shape=(7,), dtype=tf.float32),
            'robot0_joint_vel': Tensor(shape=(7,), dtype=tf.float32),
            'robot0_proprio-state': Tensor(shape=(32,), dtype=tf.float32),
        }),
        'reward': tf.float64,
        'tag:placed': tf.bool,
    }),
})