AdaManip

This is the official codebase for AdaManip: Adaptive Articulated Object Manipulation Environments and Policy Learning (ICLR 2025)

Setup

1. Create a Conda environment

conda create -n adamanip python=3.8
conda activate adamanip

2. Install PyTorch & IsaacGym

pip install torch==1.13.1+cu117 torchvision==0.14.1+cu117 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/cu117

Download IsaacGym from Nvidia Website and Install the package following the instructions of the official documentation.

3. Install Pointnet++ & pytorch3d

• Install Pointnet2_Pytorch following its instructions.
• Git clone the pytorch3d and install it locally.

git clone https://github.com/facebookresearch/pytorch3d.git
cd pytorch3d && pip install -e .

4. Install other dependencies

pip install -r requirements.txt

Download Assets

Download all assets from this Google Drive link.

Place the downloaded assets folder in the root directory of this repository:

AdaManip/
├── assets/
├── scripts/
├── cfg/
├── ...

Run Experiments

All scripts for data collection, policy training, and evaluation are provided in the scripts folder. The following example demonstrates usage with the open microwave task:

Data Collection

sh scripts/microwave/collect_mv.sh

Policy Training

sh scripts/microwave/diffusion_train_mv_manip.sh

Evaluation

sh scripts/microwave/eval_mv_model.sh

Note: Configuration files are located in the cfg folder. Please ensure the model checkpoint path is correctly set before running evaluation.

BibTeX

If you found AdaManip useful, please consider citing:

@inproceedings{wang2025adamanip,
    title={AdaManip: Adaptive Articulated Object Manipulation Environments and Policy Learning},
    author={Wang, Yuanfei and Zhang, Xiaojie and Wu, Ruihai and Li, Yu and Shen, Yan and Wu, Mingdong and He, Zhaofeng and Wang, Yizhou and Dong, Hao},
    booktitle={International Conference on Learning Representations},
    year={2025},
    url={https://openreview.net/forum?id=Luss2sa0vc}
  }

Contact

If you have any suggestion or questions, please get in touch at yuanfei_wang@pku.edu.cn