- Descrição :
Esses conjuntos de dados foram criados com o ambiente PickPlaceCan do simulador de braço robótico robosuite . Os conjuntos de dados humanos foram registrados por um único operador usando o RLDS Creator e um controlador gamepad.
Os conjuntos de dados sintéticos foram registrados usando a biblioteca EnvLogger .
Os conjuntos de dados seguem o formato RLDS para representar etapas e episódios.
Os episódios consistem em 400 etapas. Em cada episódio, uma tag é adicionada quando a tarefa é concluída. Essa tag é armazenada como parte dos metadados da etapa personalizada.
Observe que, devido à dependência do EnvLogger, a geração deste conjunto de dados é atualmente suportada apenas em ambientes Linux.
Código fonte :
tfds.rlds.datasets.robosuite_panda_pick_place_can.RobosuitePandaPickPlaceCanVersões :
-
1.0.0(padrão): versão inicial.
-
Armazenado em cache automaticamente ( documentação ): Não
Chaves supervisionadas (consulte o documento
as_supervised):NoneFigura ( tfds.show_examples ): Não suportado.
Citação :
@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 (configuração padrão)
Descrição da configuração : conjunto de dados gerado por humanos (50 episódios).
Página inicial : https://github.com/google-research/rlds
Tamanho do download :
96.67 MiBTamanho do conjunto de dados :
407.24 MiBDivisões :
| Dividir | Exemplos |
|---|---|
'train' | 50 |
- Estrutura de recursos :
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,
}),
})
- Documentação de recursos :
| Recurso | Aula | Forma | Tipo D | Descrição |
|---|---|---|---|---|
| RecursosDict | ||||
| agente_id | Tensor | corda | ||
| episódio_id | Tensor | corda | ||
| índice_de_episódio | Tensor | int32 | ||
| passos | Conjunto de dados | |||
| etapas/ação | Tensor | (7,) | float64 | |
| passos/desconto | Tensor | float64 | ||
| passos/imagem | Imagem | (Nenhum, Nenhum, 3) | uint8 | |
| passos/é_primeiro | Tensor | bool | ||
| passos/é_último | Tensor | bool | ||
| etapas/is_terminal | Tensor | bool | ||
| etapas/observação | RecursosDict | |||
| etapas/observação/Can_pos | Tensor | (3,) | float64 | |
| passos/observação/Can_quat | Tensor | (4,) | float64 | |
| etapas/observação/Can_to_robot0_eef_pos | Tensor | (3,) | float64 | |
| etapas/observação/Can_to_robot0_eef_quat | Tensor | (4,) | float32 | |
| etapas/observação/estado do objeto | Tensor | (14,) | float64 | |
| etapas/observação/robot0_eef_pos | Tensor | (3,) | float64 | |
| etapas/observação/robot0_eef_quat | Tensor | (4,) | float64 | |
| etapas/observação/robot0_gripper_qpos | Tensor | (2,) | float64 | |
| etapas/observação/robot0_gripper_qvel | Tensor | (2,) | float64 | |
| etapas/observação/robot0_joint_pos_cos | Tensor | (7,) | float64 | |
| etapas/observação/robot0_joint_pos_sin | Tensor | (7,) | float64 | |
| etapas/observação/robot0_joint_vel | Tensor | (7,) | float64 | |
| etapas/observação/robot0_proprio-state | Tensor | (32,) | float64 | |
| passos/recompensa | Tensor | float64 | ||
| passos/tag:colocado | Tensor | bool |
- Exemplos ( tfds.as_dataframe ):
robosuite_panda_pick_place_can/human_images_dc29b40a
Descrição da configuração : conjunto de dados gerado por humanos, incluindo imagens com diferentes ângulos de câmera na observação. Observe que pode levar algum tempo para gerar.
Página inicial : https://www.tensorflow.org/datasets/catalog/robosuite_panda_pick_place_can
Tamanho do download :
10.95 GiBTamanho do conjunto de dados :
7.53 GiBDivisões :
| Dividir | Exemplos |
|---|---|
'train' | 50 |
- Estrutura de recursos :
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,
}),
})
- Documentação de recursos :
| Recurso | Aula | Forma | Tipo D | Descrição |
|---|---|---|---|---|
| RecursosDict | ||||
| agente_id | Tensor | corda | ||
| episódio_id | Tensor | corda | ||
| índice_de_episódio | Tensor | int32 | ||
| passos | Conjunto de dados | |||
| etapas/ação | Tensor | (7,) | float64 | |
| passos/desconto | Tensor | float64 | ||
| passos/imagem | Imagem | (Nenhum, Nenhum, 3) | uint8 | |
| passos/é_primeiro | Tensor | bool | ||
| passos/é_último | Tensor | bool | ||
| etapas/is_terminal | Tensor | bool | ||
| etapas/observação | RecursosDict | |||
| etapas/observação/Can_pos | Tensor | (3,) | float64 | |
| passos/observação/Can_quat | Tensor | (4,) | float64 | |
| etapas/observação/Can_to_robot0_eef_pos | Tensor | (3,) | float64 | |
| etapas/observação/Can_to_robot0_eef_quat | Tensor | (4,) | float32 | |
| etapas/observação/agentview_image | Imagem | (256, 256, 3) | uint8 | |
| passos/observação/birdview_image | Imagem | (256, 256, 3) | uint8 | |
| etapas/observação/estado do objeto | Tensor | (14,) | float64 | |
| etapas/observação/robot0_eef_pos | Tensor | (3,) | float64 | |
| etapas/observação/robot0_eef_quat | Tensor | (4,) | float64 | |
| etapas/observação/robot0_eye_in_hand_image | Imagem | (256, 256, 3) | uint8 | |
| etapas/observação/robot0_gripper_qpos | Tensor | (2,) | float64 | |
| etapas/observação/robot0_gripper_qvel | Tensor | (2,) | float64 | |
| etapas/observação/robot0_joint_pos_cos | Tensor | (7,) | float64 | |
| etapas/observação/robot0_joint_pos_sin | Tensor | (7,) | float64 | |
| etapas/observação/robot0_joint_vel | Tensor | (7,) | float64 | |
| etapas/observação/robot0_proprio-state | Tensor | (32,) | float64 | |
| etapas/observação/robot0_robotview_image | Imagem | (256, 256, 3) | uint8 | |
| passos/recompensa | Tensor | float64 | ||
| passos/tag:colocado | Tensor | bool |
- Exemplos ( tfds.as_dataframe ):
robosuite_panda_pick_place_can/synthetic_stochastic_sac_afe13968
Descrição da configuração : Conjunto de dados sintético gerado por um agente estocástico treinado com SAC (200 episódios).
Página inicial : https://github.com/google-research/rlds
Tamanho do download :
144.44 MiBTamanho do conjunto de dados :
622.86 MiBDivisões :
| Dividir | Exemplos |
|---|---|
'train' | 200 |
- Estrutura de recursos :
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,
}),
})
- Documentação de recursos :
| Recurso | Aula | Forma | Tipo D | Descrição |
|---|---|---|---|---|
| RecursosDict | ||||
| agente_id | Tensor | corda | ||
| episódio_id | Tensor | corda | ||
| índice_de_episódio | Tensor | int32 | ||
| passos | Conjunto de dados | |||
| etapas/ação | Tensor | (7,) | float32 | |
| passos/desconto | Tensor | float64 | ||
| passos/imagem | Imagem | (Nenhum, Nenhum, 3) | uint8 | |
| passos/é_primeiro | Tensor | bool | ||
| passos/é_último | Tensor | bool | ||
| etapas/is_terminal | Tensor | bool | ||
| etapas/observação | RecursosDict | |||
| etapas/observação/Can_pos | Tensor | (3,) | float32 | |
| passos/observação/Can_quat | Tensor | (4,) | float32 | |
| etapas/observação/Can_to_robot0_eef_pos | Tensor | (3,) | float32 | |
| etapas/observação/Can_to_robot0_eef_quat | Tensor | (4,) | float32 | |
| etapas/observação/estado do objeto | Tensor | (14,) | float32 | |
| etapas/observação/robot0_eef_pos | Tensor | (3,) | float32 | |
| etapas/observação/robot0_eef_quat | Tensor | (4,) | float32 | |
| etapas/observação/robot0_gripper_qpos | Tensor | (2,) | float32 | |
| etapas/observação/robot0_gripper_qvel | Tensor | (2,) | float32 | |
| etapas/observação/robot0_joint_pos_cos | Tensor | (7,) | float32 | |
| etapas/observação/robot0_joint_pos_sin | Tensor | (7,) | float32 | |
| etapas/observação/robot0_joint_vel | Tensor | (7,) | float32 | |
| etapas/observação/robot0_proprio-state | Tensor | (32,) | float32 | |
| passos/recompensa | Tensor | float64 | ||
| passos/tag:colocado | Tensor | bool |
- Exemplos ( tfds.as_dataframe ):