Adaptive-Control-Oriented Meta-Learning for Nonlinear Systems

This repository accompanies the paper "Adaptive-Control-Oriented Meta-Learning for Nonlinear Systems" [1].

Getting started

Ensure you are using Python 3. Clone this repository and install the packages listed in requirements.txt. In particular, this code uses JAX.

Reproducing results

Training data, trained parameters, and test results are all conveniently saved in this repository, since it can take a while to re-generate them. To simply produce Figures 2, 3, and 4 in [1], run the command python plots.py.

Training data can be generated with the command python generate_data.py.

Parameters can then be trained (for multiple training set sizes and random seeds) with the command ./train.sh. This will take a while.

Finally, test results for Figures 3 and 4 in [1] can be produced with the commands python test_single.py and ./test.sh, respectively. This may also take a while.

Citing this work

Please use the following bibtex entry to cite this work.

@INPROCEEDINGS{RichardsAzizanEtAl2021,
author    = {Richards, S. M. and Azizan, N. and Slotine, J.-J. E. and Pavone, M.},
title     = {Adaptive-control-oriented meta-learning for nonlinear systems},
booktitle = {Robotics: Science and Systems},
year      = {2021},
note      = {In press. Available at \url{https://arxiv.org/abs/2103.04490}},
}

References

[1] S. M. Richards, N. Azizan, J.-J. E. Slotine, and M. Pavone. Adaptive-control-oriented meta-learning for nonlinear systems. In Robotics: Science and Systems, 2021. In press. Available at https://arxiv.org/abs/2103.04490.