This repository contains the python script to parse the LHManip dataset presented in the paper LHManip: A Dataset for Long-Horizon Language-Grounded Manipulation Tasks in Cluttered Tabletop Environments.

To parse the data, you need to download the dataset from here. Then, after extracting the dataset from the downloaded .zip files, run the parse_dataset.py script in the long_horizon_manipulation_dataset folder.

For conciseness, in the provided snippet, we consider the state of the robot as a 23-dimensional array of values comprising all the proprioceptive observation in the order reported in Tab. 3 in the paper. Similarly, we provide the robot action as a 8-dimensional array comprising end-effector position and orientation displacements and the desired gripper opening displacement.

If you find this dataset useful, please consider citing the associated papers:

@INPROCEEDINGS{lhmanip,
  author={F. {Ceola} and L. {Natale} and N. {S}\"underhauf and K. {Rana}},
  booktitle={2024 Robotics: Science and Systems (RSS) Workshop on Data Generation for Robotics},
  title={{LHManip: A Dataset for Long-Horizon Language-Grounded Manipulation Tasks in Cluttered Tabletop Environments}},
  year={2024},
  volume={},
  number={},
  pages={},
  doi={}}

@INPROCEEDINGS{lhmanip,
  author={F. {Ceola} and L. {Natale} and N. {S}\"underhauf and K. {Rana}},
  booktitle={2024 Robotics: Science and Systems (RSS) Workshop on Mechanisms for Mapping Human Input to Robots},
  title={{LHManip: A Dataset for Long-Horizon Language-Grounded Manipulation Tasks in Cluttered Tabletop Environments}},
  year={2024},
  volume={},
  number={},
  pages={},
  doi={}}