nvidia/PhysicalAI-Robotics-Manipulation-SingleArm

Dataset Description:

PhysicalAI-Robotics-Manipulation-SingeArm is a collection of datasets of automatic generated motions of a Franka Panda robot performing operations such as block stacking, opening cabinets and drawers. The dataset was generated in IsaacSim leveraging task and motion planning algorithms to find solutions to the tasks automatically [1, 3]. The environments are table-top scenes where the object layouts and asset textures are procedurally generated [2].
This dataset is available for commercial use.

Dataset Contact(s):

Fabio Ramos ([email protected])
Anqi Li ([email protected])

Dataset Creation Date:

03/18/2025

License/Terms of Use:

cc-by-4.0

Intended Usage:

This dataset is provided in LeRobot format and is intended for training robot policies and foundation models.

Dataset Characterization

Data Collection Method

• Automated
  
• Automatic/Sensors
  
• Synthetic
  

The dataset was generated in IsaacSim leveraging task and motion planning algorithms to find solutions to the tasks automatically [1]. The environments are table-top scenes where the object layouts and asset textures are procedurally generated [2].

Labeling Method

• Not Applicable
  

Dataset Format

Within the collection, there are six datasets in LeRobot format panda-stack-wide, panda-stack-platforms, panda-stack-platforms-texture, panda-open-cabinet-left, panda-open-cabinet-right and panda-open-drawer.

• panda-stack-wide
  The Franka Panda robot picks up the red block and stacks it on top of the green block.
  • action modality: 7d, 6d relative end-effector motion + 1d gripper action
  • observation modalities:
    • observation.state: 53d, including proprioception (robot joint position, joint velocity, end-effector pose) and object poses
    • observation.images.world_camera: 512x512 world camera output stored as mp4 videos
    • observation.images.hand_camera: 512x512 wrist-mounted camera output stored as mp4 videos
• panda-stack-platforms
  The Franka Panda robot picks up a block and stacks it on top of another block in a table-top scene with randomly generated platforms.
  • action modality: 7d, 6d relative end-effector motion + 1d gripper action
  • observation modalities:
    • observation.state: 81d, including proprioception (robot joint position, joint velocity, end-effector pose) and object poses
    • observation.images.world_camera: 512x512 world camera RGB output stored as mp4 videos
    • observation.images.hand_camera: 512x512 wrist-mounted camera RGB output stored as mp4 videos
• panda-stack-platform-texture
  The Franka Panda robot picks up a block and stacks it on top of another block in a table-top scene with randomly generated platforms and random table textures.
  • action modality: 8d, 7d joint motion + 1d gripper action
  • observation modalities:
    • observation.state: 81d, including proprioception (robot joint position, joint velocity, end-effector pose) and object poses
    • observation.images.world_camera: 512x512 world camera RGB output stored as mp4 videos
    • observation.images.hand_camera: 512x512 wrist-mounted camera RGB output stored as mp4 videos
    • observation.depths.world_camera: 512x512 world camera depth output stored as mp4 videos, where 0-255 pixel value linearly maps to depth of 0-6 m
    • observation.depths.hand_camera: 512x512 wrist-mounted camera depth output stored as mp4 videos, where 0-255 pixel value linearly maps to depth of 0-6 m
• panda-open-cabinet-left
  The Franka Panda robot opens the top cabinet of a randomly generated cabinet from left to right.
  • action modality: 8d, 7d joint motion + 1d gripper action
  • observation modalities:
    • observation.state: 81d, including proprioception (robot joint position, joint velocity, end-effector pose) and object poses
    • observation.images.world_camera: 512x512 world camera RGB output stored as mp4 videos
    • observation.images.hand_camera: 512x512 wrist-mounted camera RGB output stored as mp4 videos
    • observation.depths.world_camera: 512x512 world camera depth output stored as mp4 videos, where 0-255 pixel value linearly maps to depth of 0-6 m
    • observation.depths.hand_camera: 512x512 wrist-mounted camera depth output stored as mp4 videos, where 0-255 pixel value linearly maps to depth of 0-6 m
• panda-open-cabinet-right The Franka Panda robot opens the top cabinet of a randomly generated cabinet from right to left.
  • action modality: 8d, 7d joint motion + 1d gripper action
  • observation modalities:
    • observation.state: 81d, including proprioception (robot joint position, joint velocity, end-effector pose) and object poses
    • observation.images.world_camera: 512x512 world camera RGB output stored as mp4 videos
    • observation.images.hand_camera: 512x512 wrist-mounted camera RGB output stored as mp4 videos
    • observation.depths.world_camera: 512x512 world camera depth output stored as mp4 videos, where 0-255 pixel value linearly maps to depth of 0-6 m
    • observation.depths.hand_camera: 512x512 wrist-mounted camera depth output stored as mp4 videos, where 0-255 pixel value linearly maps to depth of 0-6 m
• panda-open-drawer
  The Franka Panda robot opens the top drawer of a randomly generated cabinet.
  • action modality: 8d, 7d joint motion + 1d gripper action
  • observation modalities:
    • observation.state: 81d, including proprioception (robot joint position, joint velocity, end-effector pose) and object poses
    • observation.images.world_camera: 512x512 world camera RGB output stored as mp4 videos
    • observation.images.hand_camera: 512x512 wrist-mounted camera RGB output stored as mp4 videos
    • observation.depths.world_camera: 512x512 world camera depth output stored as mp4 videos, where 0-255 pixel value linearly maps to depth of 0-6 m
    • observation.depths.hand_camera: 512x512 wrist-mounted camera depth output stored as mp4 videos, where 0-255 pixel value linearly maps to depth of 0-6 m

Dataset Quantification

Record Count

• panda-stack-wide
  • number of episodes: 10243
  • number of frames: 731785
  • number of RGB videos: 20486 (10243 from world camera, 10243 from hand camera)
• panda-stack-platforms
  • number of episodes: 17629
  • number of frames: 1456899
  • number of RGB videos: 35258 (17629 from world camera, 17629 from hand camera)
• panda-stack-platforms-texture
  • number of episodes: 6303
  • number of frames: 551191
  • number of RGB videos: 12606 (6303 from world camera, 6303 from hand camera)
  • number of depth videos: 12606 (6303 from world camera, 6303 from hand camera)
• panda-open-cabinet-left
  • number of episodes: 1512
  • number of frames: 220038
  • number of RGB videos: 3024 (1512 from world camera, 1512 from hand camera)
  • number of depth videos: 3024 (1512 from world camera, 1512 from hand camera)
• panda-open-cabinet-right
  • number of episodes: 1426
  • number of frames: 224953
  • number of RGB videos: 2852 (1426 from world camera, 1426 from hand camera)
  • number of depth videos: 2852 (1426 from world camera, 1426 from hand camera)
• panda-open-drawer
  • number of episodes: 1273
  • number of frames: 154256
  • number of RGB videos: 2546 (1273 from world camera, 1273 from hand camera)
  • number of depth videos: 2546 (1273 from world camera, 1273 from hand camera)

Total storage: 15.2 GB

Reference(s):

[1] @inproceedings{garrett2020pddlstream,
    title={Pddlstream: Integrating symbolic planners and blackbox samplers via optimistic adaptive planning},
    author={Garrett, Caelan Reed and Lozano-P{\'e}rez, Tom{\'a}s and Kaelbling, Leslie Pack},
    booktitle={Proceedings of the international conference on automated planning and scheduling},
    volume={30},
    pages={440--448},
    year={2020}
}

[2] @article{Eppner2024,
    title = {scene_synthesizer: A Python Library for Procedural Scene Generation in Robot Manipulation},
    author = {Clemens Eppner and Adithyavairavan Murali and Caelan Garrett and Rowland O'Flaherty and Tucker Hermans and Wei Yang and Dieter Fox},
    journal = {Journal of Open Source Software}
    publisher = {The Open Journal},
    year = {2024},
    Note = {\url{https://scene-synthesizer.github.io/}}
}

[3] @inproceedings{curobo_icra23,
    author={Sundaralingam, Balakumar and Hari, Siva Kumar Sastry and
        Fishman, Adam and Garrett, Caelan and Van Wyk, Karl and Blukis, Valts and
        Millane, Alexander and Oleynikova, Helen and Handa, Ankur and
        Ramos, Fabio and Ratliff, Nathan and Fox, Dieter},
    booktitle={2023 IEEE International Conference on Robotics and Automation (ICRA)},
    title={CuRobo: Parallelized Collision-Free Robot Motion Generation},
    year={2023},
    volume={},
    number={},
    pages={8112-8119},
    doi={10.1109/ICRA48891.2023.10160765}
}

Ethical Considerations:

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.

Please report security vulnerabilities or NVIDIA AI Concerns here.