Dataset Viewer

The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

RoboLab Motion-Planning Code

This Hugging Face dataset repository is used as a code-only snapshot for RoboLab plus the RoboLab motion-planning baseline. Generated datasets and rendered demos are intentionally not included here.

Included

  • robolab/: RoboLab Python package code.
  • robolab_mp/: motion-planning baseline, GR00T exporter, geometry helpers, video verifier.
  • examples/: CLI entrypoints for running motion planning and exporting GR00T datasets.
  • documentation/: how to run and reproduce the verified motion-planning tasks.
  • docs/: RoboLab documentation copied from the working tree.
  • pyproject.toml, requirements.txt, license and project metadata.

Excluded

  • output/: generated demos and datasets.
  • assets/: large RoboLab simulation assets; use the full local/project asset checkout when running Isaac tasks.
  • .git/, caches, and generated bytecode.

Environment

The verified local environment used here was:

/home/horde/miniconda3/envs/env_uwlab

Run commands from the repository root after restoring/installing the required RoboLab assets:

cd RoboLab

If you have reviewed and accepted the NVIDIA Omniverse EULA, run Isaac headless with:

OMNI_KIT_ACCEPT_EULA=YES

Run Verified Motion Planning

OMNI_KIT_ACCEPT_EULA=YES /home/horde/miniconda3/bin/conda run -n env_uwlab   python examples/run_motion_planning.py   --headless   --task BananaInBowlTask MustardInLeftBinTask   --video-mode all   --output-dir output/mp_grasp_stability_yaw_verified   --max-steps 460   --verify-videos

Expected verified tasks:

  • BananaInBowlTask
  • MustardInLeftBinTask

Generate GR00T-Format Data

After motion-planning outputs exist:

/home/horde/miniconda3/envs/env_uwlab/bin/python examples/export_motion_planning_gr00t.py   --input output/mp_grasp_stability_yaw_verified   --output output/mp_grasp_stability_yaw_verified_gr00t   --fps 15   --only-successful   --video-key front   --video-key wrist

The exporter writes GR00T-flavored LeRobot v2 data with:

meta/info.json
meta/modality.json
meta/tasks.jsonl
meta/episodes.jsonl
meta/stats.json
meta/relative_stats.json
data/chunk-000/episode_*.parquet
videos/chunk-000/observation.images.front/episode_*.mp4
videos/chunk-000/observation.images.wrist/episode_*.mp4

Documentation

Start with:

  • documentation/motion_planning_run.md
  • documentation/reproduce_motion_planning_results.md
  • docs/motion_planning.md

Notes

This repo is for code only. The generated banana GR00T dataset is uploaded separately to:

https://huggingface.co/datasets/yqi19/robolab_data
Downloads last month
276