Project page: https://michaelx.io/parc
Tested with Ubuntu 22.04
Install IsaacGym: https://developer.nvidia.com/isaac-gym
Make sure to install it within a conda environment with python 3.8.19 (Other versions may also work, but not tested).
Install requirements:
conda activate parc
pip install -r requirements.txt
and it should be good to go. If pytorch is not being able to detect CUDA, try reinstalling:
pip install torch==1.13.1+cu117 -f https://download.pytorch.org/whl/torch_stable.html
Download the datasets from the initial iteration and each stage of PARC, as well as the models trained in the 3rd PARC iteration. These files are loaded with anim/motion_lib.py and anim/kin_char_model.py. You can view them with motion_forge.py, by editing the "motion_filepath" param in motionscope/motion_forge_config.yaml https://1sfu-my.sharepoint.com/:f:/g/personal/mxa23_sfu_ca/Et16uLMFxoRKouibvBa7LbwBEmX5_iI5a8dZyiMc0wmSTA?e=ihma1b The password is "PARC".
The PARC training loop consists of 4 main stages, which are individually run by:
python parc_1_train_gen.py --config path/to/config
python parc_2_kin_gen.py --config path/to/config
python parc_3_tracker.py --config path/to/config
python parc_4_phys_record.py --config path/to/config
These modules are independent and flexibly configured using configuration files. However, to aid in setting up all the configuration files for a PARC iteration, we provide the following script:
parc_0_setup_iter.py
- PARC Guide (Coming Soon)
- Motionscope Guide (Coming Soon)
- Iteration 4 and 5 models, datasets
- Motion file format refactoring
- Faster inference code
- Full codebase cleanup and refactoring
If you find PARC helpful, please consider citing:
@inproceedings{xu2025parc,
author = {Xu, Michael and Shi, Yi and Yin, KangKang and Peng, Xue Bin},
title = {PARC: Physics-based Augmentation with Reinforcement Learning for Character Controllers},
year = {2025},
booktitle={SIGGRAPH 2025 Conference Papers (SIGGRAPH '25 Conference Papers)}
}Please also consider citing MimicKit, the codebase that PARC was built on:
@misc{MimicKit,
title = {MimicKit},
author = {Peng, Xue Bin},
year = {2025},
publisher = {GitHub},
journal = {GitHub repository},
howpublished = {\url{https://github.com/xbpeng/MimicKit}},
}If you find Motionscope (previous name: Motion Forge, update coming soon) helpful, please consider citing:
@software{Xu_Michael_Motionscope,
author = {Xu, Michael and Peng, Xue Bin},
title = {{motionscope}},
year = {2025},
month = aug,
version = {0.0.1},
url = {https://github.com/mshoe/PARC},
license = {MIT}
}Motionscope is built on top of a wonderful tool called Polyscope. Please consider citing Polyscope here:
@misc{polyscope,
title = {Polyscope},
author = {Nicholas Sharp and others},
note = {www.polyscope.run},
year = {2019}
}
Also, consider citing these important papers which PARC's motion tracker and motion generator builds upon:
DeepMimic
@article{
2018-TOG-deepMimic,
author = {Peng, Xue Bin and Abbeel, Pieter and Levine, Sergey and van de Panne, Michiel},
title = {DeepMimic: Example-guided Deep Reinforcement Learning of Physics-based Character Skills},
journal = {ACM Trans. Graph.},
issue_date = {August 2018},
volume = {37},
number = {4},
month = jul,
year = {2018},
issn = {0730-0301},
pages = {143:1--143:14},
articleno = {143},
numpages = {14},
url = {http://doi.acm.org/10.1145/3197517.3201311},
doi = {10.1145/3197517.3201311},
acmid = {3201311},
publisher = {ACM},
address = {New York, NY, USA},
keywords = {motion control, physics-based character animation, reinforcement learning},
} Motion Diffusion Model
@inproceedings{
tevet2023human,
title={Human Motion Diffusion Model},
author={Guy Tevet and Sigal Raab and Brian Gordon and Yoni Shafir and Daniel Cohen-or and Amit Haim Bermano},
booktitle={The Eleventh International Conference on Learning Representations },
year={2023},
url={https://openreview.net/forum?id=SJ1kSyO2jwu}
}