MCPcopy Index your code
hub / github.com/amazon-far/ResMimic

github.com/amazon-far/ResMimic @main

Chat with this repo
repository ↗ · DeepWiki ↗ · + Follow
1,197 symbols 2,838 edges 97 files 248 documented · 21%
What it actually does AI analysis from the code graph — generated when you open this
loading…
README

ResMimic: From General Motion Tracking to Humanoid Whole-Body Loco-Manipulation via Residual Learning

Siheng Zhao, Yanjie Ze, Yue Wang, C. Karen Liu†, Pieter Abbeel†, Guanya Shi†, Rocky Duan†
† Equal Advising · 2025

Website · arXiv · Video


Table of Contents


Installation

One-time setup (creates environment and installs dependencies):

source setup_dev.sh

Each time you open a new terminal, activate the environment:

source source_dev_setup.sh

Usage

Training

Train the residual policy with the provided config. You need to pass your Weights & Biases entity (username or team):

./train.sh <wandb_entity>

Example: ./train.sh my_username

Evaluation

Run evaluation (play) for a trained run. From the repository root:

./eval.sh <wandb_run_id> <checkpoint_iter>

Example: ./eval.sh abc123def 5000 runs evaluation for run abc123def at checkpoint iteration 5000.

Optional: uncomment --record_video in eval.sh to save videos.


Citation and Contact

If you find this work useful, please cite:

@article{zhao2025resmimic,
title={ResMimic: From General Motion Tracking to Humanoid Whole-body Loco-Manipulation via Residual Learning}, 
author={Siheng Zhao and Yanjie Ze and Yue Wang and C. Karen Liu and Pieter Abbeel and Guanya Shi and Rocky Duan},
year={2025},
journal= {arXiv preprint arXiv:2510.05070}
}

And also consider citing the related works:

@article{ze2025twist2,
title={TWIST2: Scalable, Portable, and Holistic Humanoid Data Collection System},
author= {Yanjie Ze and Siheng Zhao and Weizhuo Wang and Angjoo Kanazawa and Rocky Duan and Pieter Abbeel and Guanya Shi and Jiajun Wuand C. Karen Liu},
year= {2025},
journal= {arXiv preprint arXiv:2511.02832}
}

Questions: sihengz@usc.edu


Acknowledgments

This codebase is built on TWIST2.

Security

See CONTRIBUTING for more information.

License

This project is licensed under the Apache-2.0 License.

This repository builds upon third-party open-source projects, including legged_gym and rsl_rl. These components have been modified to support our use case. The original LICENSE files in their respective subdirectories are retained and continue to govern those portions of the code. All rights and attribution remain with the original authors in accordance with their licenses.

Core symbols most depended-on inside this repo

format
called by 53
pose/pose/utils/logger.py
numpy
called by 50
pose/pose/utils/motion_lib.py
clone
called by 41
pose/pose/poselib/poselib/skeleton/skeleton3d.py
quat_rotate_inverse
called by 30
pose/pose/utils/isaacgym_torch_utils.py
torch_rand_float
called by 22
pose/pose/utils/isaacgym_torch_utils.py
reindex
called by 21
legged_gym/legged_gym/envs/base/legged_robot.py
load
called by 19
rsl_rl/rsl_rl/runners/on_policy_dagger_runner.py
add_roughness
called by 16
legged_gym/legged_gym/gym_utils/terrain.py

Shape

Method 770
Class 215
Function 212

Languages

Python100%

Modules by API surface

pose/pose/poselib/poselib/skeleton/skeleton3d.py79 symbols
legged_gym/legged_gym/envs/base/humanoid_mimic.py63 symbols
legged_gym/legged_gym/envs/base/legged_robot.py59 symbols
pose/pose/poselib/poselib/visualization/plt_plotter.py44 symbols
pose/pose/poselib/poselib/visualization/simple_plotter_tasks.py41 symbols
legged_gym/legged_gym/envs/g1/g1_mimic_distill_config.py40 symbols
rsl_rl/rsl_rl/datasets/motion_loader.py36 symbols
pose/pose/poselib/poselib/core/rotation3d.py35 symbols
rsl_rl/rsl_rl/utils/utils.py30 symbols
rsl_rl/rsl_rl/modules/actor_critic_transformer.py30 symbols
pose/pose/utils/motion_lib.py28 symbols
legged_gym/legged_gym/envs/base/legged_robot_config.py28 symbols

For agents

$ claude mcp add ResMimic \
  -- python -m otcore.mcp_server <graph>

⬇ download graph artifact