Clutter-Aware Spill-Free Liquid Transport via Learned Dynamics

This is the repository for training the Spill-Free Classifier (SFC) model presented in the Clutter-Aware Spill-Free Liquid Transport via Learned Dynamics paper (link).

Data

Data Directory

• Mocap: Data collected from individuals handling different container scenarios captured using a motion capture system
• Panda: Data collected from the panda robotic arm handling different container scenarios. (Some files are currently missing)

Setup

git clone git@github.com:HIRO-group/trajectory-model.git
conda create -n "sfrrt"
conda activate sfrrt
cd trajectory-model
pip install -e .
mkdir data/

Copy the downloaded data in the data/ dir to execute the following commands.

Train Spill-Free Classifer (SFC)

python scripts/train_SFC.py

Predict Using SFC

python scripts/predict_SFC.py

Calculate Container Tilt Angle

python scripts/calculate_tilt_angle.py

Visualize Mocap Trajectories

python scripts/plot_mocap_trajectory.py

Visualize Panda Trajectories

python scripts/plot_panda_trajectory.py

Consume Data Streamed By Natnet from Mocap

python scripts/consume_data_from_mocap.py

Convert Joint Angles to Cartesian

python scripts/convert_panda_joints_to_cartesian.py

Citation

@inproceedings{abderezaei2024clutterawarespillfreeliquidtransport,
  title={Clutter-Aware Spill-Free Liquid Transport via Learned Dynamics}, 
  author={Abderezaei, Ava and Pasricha, Anuj and Klausenstock, Alex and Roncone, Alessandro},
  year={2024},
  booktitle={2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
}