Skip to content

We are implementing differentiable models of robot manipulators, which allows us to learn typically assumed to be known models of robots for control and motion planning.

License

Notifications You must be signed in to change notification settings

EricCousineau-TRI/differentiable-robot-model

 
 

Repository files navigation

differentiable robot model

CircleCI Code style: black

Differentiable and learnable robot model. Our differentiable robot model implements computations such as forward kinematics and inverse dynamics, in a fully differentiable way. We also allow to specify
parameters (kinematics or dynamics parameters), which can then be identified from data (see examples folder).

Currently, our code should work with any kinematic trees. This package comes with wrappers specifically for:

  • TriFinger Edu
  • Kuka iiwa
  • Franka Panda
  • Allegro Hand
  • Fetch Arm
  • a 2-link toy robot

You can find the documentation here: Differentiable-Robot-Model Documentation

Installation

Requirements: python>= 3.7

clone this repo and install from source:

git clone [email protected]:facebookresearch/differentiable-robot-model.git
cd differentiable-robot-model
python setup.py develop

Examples

2 examples scripts show the learning of kinematics parameters

python examples/learn_kinematics_of_iiwa.py

and the learning of dynamics parameters

python examples/learn_dynamics_of_iiwa.py

L4DC paper and experiments

the notebook experiments/l4dc-sim-experiments shows a set of experiments that are similar to what we presented in our L4DC paper

@InProceedings{pmlr-v120-sutanto20a, 
    title = {Encoding Physical Constraints in Differentiable Newton-Euler Algorithm}, 
    author = {Sutanto, Giovanni and Wang, Austin and Lin, Yixin and Mukadam, Mustafa and Sukhatme, Gaurav and Rai, Akshara and Meier, Franziska}, 
    pages = {804--813}, 
    year = {2020},
    editor = {Alexandre M. Bayen and Ali Jadbabaie and George Pappas and Pablo A. Parrilo and Benjamin Recht and Claire Tomlin and Melanie Zeilinger}, 
    volume = {120}, 
    series = {Proceedings of Machine Learning Research}, 
    address = {The Cloud}, month = {10--11 Jun}, 
    publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v120/sutanto20a/sutanto20a.pdf},
    url = {http://proceedings.mlr.press/v120/sutanto20a.html}, 
}

Testing

running pytest in the top-level folder will run our differentiable robot model tests, which compare computations against pybullet.

Code Contribution

We enforce linters for our code. The formatting test will not pass if your code does not conform.

To make this easy for yourself, you can either

  • Add the formattings to your IDE
  • Install the git pre-commit hooks by running
    pip install pre-commit
    pre-commit install

For Python code, use black.

To enforce this in VSCode, install black, set your Python formatter to black and set Format On Save to true.

To format manually, run: black .

License

differentiable-robot-model is released under the MIT license. See LICENSE for additional details about it. See also our Terms of Use and Privacy Policy.

About

We are implementing differentiable models of robot manipulators, which allows us to learn typically assumed to be known models of robots for control and motion planning.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 62.6%
  • Jupyter Notebook 37.3%
  • CMake 0.1%