HybridIL

Official implementation for HybridIL in the paper ForceMimic: Force-Centric Imitation Learning with Force-Motion Capture System for Contact-Rich Manipulation, accepted by ICRA 2025.

For more information, please visit our project website.

---

Installation

git clone git@github.com:ForceMimic/hybridil.git
cd hybridil

conda create -n hil python=3.8
conda activate hil

# for policy training
pip install -r train_requirements.txt
pip install -e .

# for real robot evaluation
pip install -r eval_requirements.txt
# download flexiv rdk v0.9.1 from https://github.com/flexivrobotics/flexiv_rdk

Data Preparation

Please refer to ForceCapture to collect and process the data. You can download our processed dataset from Google Drive.

Policy Training

python train.py --config configs/train_dp_ftout.json

Robot Evaluation

python node.py --config configs/eval_dp_ftout.json
python eval.py --config configs/eval_dp_ftout.json

---

Acknowledgement

Our policy implementation is based on DexCap, robomimic and Diffusion Policy. Kudos to the authors for their amazing contributions.

Citation

If you find our work useful, please consider citing:

@inproceedings{liu2025forcemimic,
  author={Liu, Wenhai and Wang, Junbo and Wang, Yiming and Wang, Weiming and Lu, Cewu},
  booktitle={2025 IEEE International Conference on Robotics and Automation (ICRA)}, 
  title={ForceMimic: Force-Centric Imitation Learning with Force-Motion Capture System for Contact-Rich Manipulation}, 
  year={2025}
}

License

This repository is released under the MIT license.