This is the official implementation for AdaptNet: Policy Adaptation for Physics-Based Character Control. [arXiv] [Youtube] (SIGGRAPH Asia'23, TOG)
This implementation is based on [webpage] [code]
Composite Motion Learning with Task Control [arXiv] [Youtube] (SIGGRAPH'23, TOG)
A GAN-Like Approach for Physics-Based Imitation Learning and Interactive Character Control [arXiv] [Youtube] (SCA'21, PACMCGIT)
We recommend to install all the requirements through Conda by
$ conda create --name <env> --file requirements.txt -c pytorch -c conda-forge
Download IsaacGym Pr4 from the official site and install it via pip.
$ python main.py <configure_file> --meta <pretrained_meta_policy> --ckpt <checkpoint_dir> --test
We provide pretrained policy models in pretrained
folder. To evaluate a pretrained policy, e.g. please run
$ python main.py config/config_run_lowfriction.py \
--meta pretrained/locomotion_run --ckpt pretrained/run_lowfriction \
--test
$ python main.py config/config_walk_lowfriction.py \
--meta pretrained/locomotion_walk --ckpt pretrained/walk_lowfriction \
--test
$ python main.py config/config_terrain.py \
--meta pretrained/locomotion_walk --ckpt pretrained/walk_terrain \
--test
$ python main.py config/config_walk_jaunty.py \
--meta pretrained/locomotion_walk --ckpt pretrained/walk_jaunty \
--test
$ python main.py config/config_walk_stoop.py \
--meta pretrained/locomotion_walk --ckpt pretrained/walk_stoop \
--test
If you use the code or provided motions for your work, please consider citing our papers:
@article{adaptnet,
author = {Xu, Pei and Xie, Kaixiang and Andrews, Sheldon and Kry, Paul G and Neff, Michael and McGuire, Morgan and Karamouzas, Ioannis and Zordan, Victor},
title = {{AdaptNet}: Policy Adaptation for Physics-Based Character Control},
journal = {ACM Transactions on Graphics},
publisher = {ACM New York, NY, USA},
year = {2023},
volume = {42},
number = {6},
doi = {10.1145/3618375}
}
@article{composite,
author = {Xu, Pei and Shang, Xiumin and Zordan, Victor and Karamouzas, Ioannis},
title = {Composite Motion Learning with Task Control},
journal = {ACM Transactions on Graphics},
publisher = {ACM New York, NY, USA},
year = {2023},
volume = {42},
number = {4},
doi = {10.1145/3592447}
}
@article{iccgan,
author = {Xu, Pei and Karamouzas, Ioannis},
title = {A {GAN}-Like Approach for Physics-Based Imitation Learning and Interactive Character Control},
journal = {Proceedings of the ACM on Computer Graphics and Interactive Techniques},
publisher = {ACM New York, NY, USA},
year = {2021},
volume = {4},
number = {3},
pages = {1--22},
doi = {10.1145/3480148}
}