A Differentiable Recipe for Learning Visual Non-Prehensile Planar Manipulation

This repository contains the source code of the paper A Differentiable Recipe for Learning Visual Non-Prehensile Planar Manipulation, CoRL 2021.

Content

The code contains five sets of experiments:

1. Fully connneced Neural Network Trained on robot actions (Exp_NN)
2. Fully connneced Neural Network Trained on differnetiable simulator (Exp_NNM)
3. Differentiable Pipaline Trained on machanical parameters (Exp_MDR)
4. Differentiable Pipaline Trained on machanical parameters and CVX Layer (Exp_CVX)
5. Differentiable Pipaline Trained on end to end with simulation (Exp_DLM)

Installation

Install the dependencies:

1. Pytorch 1.4.0 (https://pytorch.org/)
2. CVXPy Layers (https://github.com/cvxgrp/cvxpylayers)
3. PyGame (https://www.pygame.org/wiki/GettingStarted)
4. PyODE (http://pyode.sourceforge.net/)

To install the simulator, also install:

cd ./src/diffsim_lcp
python setup.py install --user

Tested in Python 3.7.7

Citing

If you find this repository helpful in your publications, please cite the following:

@inproceedings{aceituno2021corl,
    title={A Differentiable Recipe for Learning Visual Non-Prehensile Planar Manipulation },
    author={B. Aceituno, and A. Rodriguez, and S. Tulsiani, and A. Gupta, and M. Mukadam},
    booktitle={CoRL},
    year={2021}
}

License

This repository is licensed under the MIT License.