SoftMAC

Implemetation for our paper "SoftMAC: Differentiable Soft Body Simulation with Forecast-based Contact Model and Two-way Coupling with Articulated Rigid Bodies and Clothes" (accepted to IROS 2024).

[website] [arXiv]

Installation

Soft-rigid and soft-cloth coupling are implemented separately. The underlying cloth simulator DiffClothAI is not open-source yet, but we provide our code for soft-cloth coupling under soft_cloth. Soft-rigid coupling is already supported in this repository. Please follow the instructions:

Step 1: Create conda environment (recommended)

conda create -n softmac python=3.8
conda activate softmac

Step 2: Install Jade

Jade is the underlying articulated rigid body simulator. You can find it here.

Step 3: Install SoftMAC

pip install -r requirements.txt
pip install -e .

Demos

We optimize the action sequences with SoftMAC for each task.

Demo 1: Pour wine

python3 demo_pour.py

Demo 2: Squeeze plasticine

python3 demo_grip.py

Demo 3: Pull door

python3 demo_door.py

Demo 4: Make taco

# run under ./soft_cloth
python3 demo_taco.py

Demo 5: Push towel

# run under ./soft_cloth
python3 demo_hit.py

Update

[2024.1] Add support for velocity control and an example softmac/demo_pour_vel.py. You can activate this mode by setting rigid_velocity_control=True in the configuration file.

Note

Feel free to contact me at minliu2@cs.cmu.edu or create a Github issue if you have questions regarding setting up the repository, running examples or adding new examples.

Citation

If you find our work useful, please consider citing:

@inproceedings{liu2023softmac,
  title={SoftMAC: Differentiable Soft Body Simulation with Forecast-based Contact Model and Two-way Coupling with Articulated Rigid Bodies and Clothes},
  author={Liu, Min and Yang, Gang and Luo, Siyuan and Shao, Lin},
  booktitle={IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) },
  year={2024},
}