Home

Awesome

CPDeform

Code and data for paper Contact Points Discovery for Soft-Body Manipulations with Differentiable Physics at ICLR 2022 (Spotlight).

Alt Text

@InProceedings{li2022contact,
author = {Li, Sizhe and Huang, Zhiao and Du, Tao and Su, Hao and Tenenbaum, Joshua and Gan, Chuang},
title = {{C}ontact {P}oints {D}iscovery for {S}oft-{B}ody {M}anipulations with {D}ifferentiable {P}hysics},
booktitle = {International Conference on Learning Representations (ICLR)},
year = {2022}}

Installation

python3 -m pip install -e .
conda install pyg -c pyg

Data

Download target templates here, and put the folder diff_phys that contains goal shapes onto your machine.

Experiments

Run the following to train an agent that writes "ICLR" on a plasticine board:

python3 scripts/launch_training.py \
--root_dir somewhere_on_your_machine/diff_phys \
--algo cpdeform \ 
--env_name multistage_writer

The experiment results will be stored in the diff_phys folder.

Acknowledgements

Our physics simulation is based on PlasticineLab.