ViaCatalyst/maniskill-pickcube-motionplanning-lerobot-v3

ManiSkill PickCube motion-planning — verified LeRobot v3 reference subset

Community conversion produced by ViaCatalyst BYOD. This repository is not an official upstream release and is not affiliated with the ManiSkill authors.

This is a compact, provenance-complete conversion of the first 10 trajectories in the pinned ManiSkill PickCube-v1 motion-planning HDF5 file. It is a reproducible LeRobot v3 reference dataset and conversion-quality example.

At a glance

Property	Value
LeRobot format	v3.0
Robot	Panda
Task	Pick up the cube and move it to the goal
Episodes	10
Frames	726
Duration	36.30 seconds
FPS	20
Observation modality	Low-dimensional state
Action dimension	8
State dimension	70
Official LeRobot reader	Passed with LeRobot 0.6.0

Features and source mapping

LeRobot feature	dtype	shape	Source mapping
action	float32	[8]	Direct per-frame copy from each selected trajectory's actions dataset, with a float32 cast only
observation.state	float32	[70]	Deterministic numeric state vector produced from the ManiSkill trajectory state fields; names are documented in meta/info.json
episode_index	int64	[1]	Preserved HDF5 trajectory boundary
frame_index	int64	[1]	Zero-based position inside the trajectory
timestamp	float32	[1]	frame_index / 20 seconds
task_index	int64	[1]	Maps to the PickCube task in meta/tasks.parquet

Original-action preservation

The conversion reads the original actions array for each selected trajectory and writes each row directly to action. It does not replay a learned policy, regenerate actions, interpolate action values, or relabel trajectories. Episode boundaries are retained; timestamps are normalized to the declared 20 FPS.

The upstream repository's illustrative sample video is not an aligned stream in the selected trajectory HDF5 file. This release therefore makes no image, depth, audio, or video modality claim.

Validation evidence

The complete machine-readable evidence is in validation-report.json. All seven critical automated checks passed:

• Dataset metadata counts
• Parquet schema and frame count
• Episode boundaries and timestamp regularity
• Feature dimensions and finite values
• LeRobot v3 relational metadata
• License and provenance completeness
• Official LeRobot reader smoke test

The readiness score is 92/100 — Review recommended. One medium-confidence statistical signal is disclosed rather than hidden: observation.state has a robust-outlier frame rate of 64.19%. This is not proof of corruption; inspect feature distributions in the validation report before training.

Source, revision, and integrity

• Source dataset: haosulab/ManiSkill_Demonstrations
• Pinned source revision: d674485bbffdd533914e52d272fdda34c0515608
• Source file: demos/PickCube-v1/motionplanning/trajectory.h5
• Source file size: 29,349,195 bytes
• Source SHA-256: 03ca60546541a7f18321d9d32721f0254bc75217828c0cadacd217d0c014576a
• Converter adapter: maniskill adapter 1.0.0
• Converter code revision: 4230bc4b56abe85116b202398b1a625a81c1c55c

Additional audit files: provenance.json, bundle-manifest.json, bundle-manifest.external.json, and UPSTREAM_LICENSE.md.

License and attribution

The pinned upstream dataset card declares Apache-2.0. This conversion retains that identifier and the original citation. Conversion does not transfer ownership, create affiliation, or replace upstream terms. Users remain responsible for reviewing source data and any separately licensed assets in their own workflows.

Intended use

• Validating LeRobot v3 readers and data pipelines
• Small offline imitation-learning or educational experiments
• Reproducible testing of HDF5-to-LeRobot conversion
• Comparing schema, provenance, and validation tooling

Limitations

• This is a community conversion, not an official upstream release.
• It contains only the first 10 trajectories from the pinned upstream file, not the complete ManiSkill demonstration set.
• It is a low-dimensional state-action release, not a multimodal dataset.
• Statistical quality signals require domain-aware review before training.
• Automated validation does not measure policy performance, task success, or simulator replay success.

Load with LeRobot

from lerobot.datasets.lerobot_dataset import LeRobotDataset

dataset = LeRobotDataset("ViaCatalyst/maniskill-pickcube-motionplanning-lerobot-v3")
print(dataset.meta.total_episodes, dataset.meta.total_frames)

Conversion tooling

Converted and validated with the ViaCatalyst BYOD Processing Platform, a free workflow for converting robotics datasets to LeRobot format. For high-volume datasets, contact ViaCatalyst support through the platform.

Citation

Please cite the original ManiSkill work:

@article{taomaniskill3,
  title={ManiSkill3: GPU Parallelized Robotics Simulation and Rendering for Generalizable Embodied AI},
  author={Tao, Stone and Xiang, Fanbo and Shukla, Arth and others},
  journal={Robotics: Science and Systems},
  year={2025}
}