Xingyu Lin*, Yufei Wang*, Zixuan Huang, David Held, CoRL 2021. * indicates equal contribution (order by dice rolling).

Website / ArXiv

• 1 Simulation
  • 1.1 Setup
  • 1.2 Train VCD
  • 1.3 Planning with VCD
  • 1.4 Graph Imitation Learning
  • 1.5 Pretrained Models
  • 1.6 Demo

This repository is a subset of SoftAgent cleaned up for VCD. Environment setup for VCD is similar to that of softagent.

1. Install SoftGym. Then, copy softgym as a submodule in this directory by running cp -r [path to softgym] ./. Use the updated softgym on the vcd branch by cd softgym && git checkout vcd
2. You should have a conda environment named softgym. Install additional packages required by VCD, by conda env update --file environment.yml
3. Generate initial environment configurations and cache them, by running python VCD/generate_cached_initial_state.py.
4. Run ./compile_1.0.sh && . ./prepare_1.0.sh to compile PyFleX and prepare other paths.

• Generate the dataset for training by running

  python VCD/main.py --gen_data=1 --dataf=./data/vcd
  

  Please refer to main.py for argument options.

• Train the dynamics model by running

  python VCD/main.py --gen_data=0 --dataf=./data/vcd_dyn
  

• Train the EdgeGNN model by running

  python VCD/main_train_edge.py --gen_data=0 --dataf=./data/vcd_edge
  

python VCD/main_plan.py --edge_model_path={path_to_trained_edge_model}\
                        --partial_dyn_path={path_to_trained_dynamics_model}

An example for loading the model trained for 120 epochs:

python VCD/main_plan.py --edge_model_path ./data/vcd_edge/vsbl_edge_120.pth\
                        --partial_dyn_path ./data/vcd_dyn/vsbl_dyn_120.pth

1. Train dynamics using the full mesh

python VCD/main.py --gen_data=0 --dataf=./data/vcd --train_mode=full

2. Train dynamics using partial point cloud and imitate the teacher model

python VCD/main.py --gen_data=0 --dataf=./data/vcd --train_mode=graph_imit --full_dyn_path={path_to_teacher_model}

Please refer to this page for downloading the pretrained models.

• Dynamics rollout
  
• Planning on square cloth
  

If you find this codebase useful in your research, please consider citing:

@inproceedings{lin2022learning,
  title={Learning visible connectivity dynamics for cloth smoothing},
  author={Lin, Xingyu and Wang, Yufei and Huang, Zixuan and Held, David},
  booktitle={Conference on Robot Learning},
  pages={256--266},
  year={2022},
  organization={PMLR}
}

@inproceedings{corl2020softgym,
 title={SoftGym: Benchmarking Deep Reinforcement Learning for Deformable Object Manipulation},
 author={Lin, Xingyu and Wang, Yufei and Olkin, Jake and Held, David},
 booktitle={Conference on Robot Learning},
 year={2020}
}