sisl / mechamodlearn

Structured framework for learning mechanical systems in PyTorch

MechaModLearn

Authors: @kunalmenda and @rejuvyesh

This library provides a structured framework for learning mechanical systems in PyTorch.

---

Installation

Requires Python3.

git clone https://github.com/sisl/mechamodlearn.git
cd mechamodlearn
pip install -e .

Usage

Example experiments are placed in experiments directory.

To run the Simple experiment:

python experiments/simple.py --logdir /tmp/simplelog

References

---

If you found this library useful in your research, please consider citing our paper:

@article{gupta2019mod,
    title={A General Framework for Structured Learning of Mechanical Systems},
    author={Gupta, Jayesh K. and Menda, Kunal and Manchester, Zachary and Kochenderfer, Mykel},
    journal={arXiv preprint arXiv:1902.08705},
    year={2019}
}

mechamodlearn's Issues

Need your suggestion for extension of this work

Dear Contributors, the work presented in the paper is really exciting. I am starting my Ph.D. research in a similar line of work, where we are exploring hybrid models (analytical + neural network) for learning the dynamics of robotic systems. I would like to know whether this model can be extended for 6-dof serial manipulators?

Thanks.

module 'resource' not available in windows

Hi,
Thank you for the great work and impressive paper.
I tried to run the example code as in readme.md, but an error message occurred that,

no module named 'resource'

After googling a bit I found that

1. 'resource' is not available in Windows OS (I'm using Windows 10)

Also by examining the code I found that

1. It seems it only checks peak memory usage in my computer, log the value and treat it as one of the metric, so I just commented them out in the code,
2. Runs perfectly. :D

Still, I am quite new in python and in this field so I wonder whether you have any idea to edit/modify the code so that any windows user would not encounter this problem.

Thanks!