This is the official PyTorch implementation of the paper "Towards Adaptable Humanoid Control via Adaptive Motion Tracking" by

Tao Huang, Huayi Wang, Junli Ren, Kangning Yin, Zirui Wang, Xiao Chen, Feiyu Jia, Wentao Zhang, Junfeng Long, Jingbo Wang†, Jiangmiao Pang†

Clone this repository:

git clone https://github.com/InternRobotics/AdaMimic.git cd AdaMimic

Create a conda environment:

conda env create -f conda_env.yml conda activate adamimic

Download and install Isaac Gym:

cd isaacgym/python && pip install -e .

Install rsl_rl (PPO implementation) and legged gym:

cd rsl_rl && pip install -e . && cd .. cd legged_gym && pip install -e . && cd .. 

Training stage 1:

python legged_gym/scripts/train.py +dataset=g1_dof27/${task} +algorithm=adamimic/stage1 

Here, ${task} can be one of task in this list.

Training stage 2:

python legged_gym/scripts/train.py +dataset=g1_dof27/${task} +algorithm=adamimic/stage2 checkpoint_path=${path/to/stage1_ckpt} 

The ${path/to/stage1_ckpt} should be replaced with checkpoints trained in the stage 1.

Play policies:

python legged_gym/scripts/play.py +dataset=g1_dof27/${task} +algorithm=adamimic/stage2 resume_path=${path/to/stage2_ckpt} 

Train baselines:

python legged_gym/scripts/train.py +dataset=g1_dof27/${task} +algorithm=${baseline}

All configurations of ${baseline} are implemented in this folder.

Play policies

python legged_gym/scripts/play.py +dataset=g1_dof27/${task} +algorithm=${baseline} resume_path=${path/to/baseline_ckpt} 

For any questions, please feel free to email taou.cs13@gmail.com. We will respond to it as soon as possible.

This repository is built upon the support and contributions of the following open-source projects. Special thanks to:

• legged_gym and HIMLoco: The foundation for training and running codes.
• rsl_rl: Reinforcement learning algorithm implementation.
• ASAP: Motion tracking implementation.
• AMP for hardware: AMP implementation.
• GVHMR: SMPL motion reconstruction algorithom.

If you find our work useful, please consider citing:

@article{huang2025adaptive, title={Towards Adaptable Humanoid Control via Adaptive Motion Tracking}, author={Huang, Tao and Wang, Huayi and Ren, Junli and Yin, Kangning and Wang, Zirui and Chen, Xiao and Jia, Feiyu and Zhang, Wentao and Long, Jungfeng and Wang, Jingbo and Pang, Jiangmiao}, year={2025} } 

The code is licensed under the CC BY-NC-SA 4.0 International License . Commercial use is not allowed without explicit authorization.