Human Pouring Control

Code to reproduce the figures for "How to pour a cup of coffee" Niteesh Midlagajni, Roland W. Fleming, Constantin A. Rothkopf

Contents

This repo contains the analysis and plotting notebooks used to generate the behavioral and modeling figures in the paper. Preprocessed data is hosted on Zenodo and downloaded automatically when you run the notebooks.

• behavioural_plots.ipynb — behavioral figures (Figures 1, 2, 3).
• model_plots.ipynb — SINDYc dynamics fits, iLQG controller fits, and inverse optimal control results: recovered parameters, model comparison, and the signal-dependent-noise ablation (Figures 4 and 5).
• controller_helper.py — SINDYc forward dynamics equations and iLQG simulation wrappers for each container–vessel combination.
• utils.py — shared data loading and plotting utilities.

Data

Preprocessed data is hosted on Zenodo: 10.5281/zenodo.22163061. The first cell of each notebook downloads and extracts it into data/ automatically if it isn't already present.

Dependencies

iLQG solver come from nioc-pouring, our fork of the nioc library that includes all the container-vessel pouring environments. This is installed automatically as a project dependency.

Setup

This project uses uv to install and run.

git clone https://github.com/RothkopfLab/human-pouring-control.git
cd human-pouring-control
uv sync
uv run jupyter lab

Open behavioural_plots.ipynb or model_plots.ipynb and run.

Raw Data Playback

For replaying and visualizing the raw mocap, scale, and gaze data, see pouring_data_playback.

Citation

@article{midlagajni2026pouring,
  title={How to pour a cup of coffee},
  author={Midlagajni, Niteesh and Fleming, Roland W. and Rothkopf, Constantin A.},
  journal={bioRxiv},
  year={2026},
  doi={10.64898/2026.08.26.746627},
  url={https://www.biorxiv.org/content/10.64898/2026.08.26.746627v1}
}