Using Physics Knowledge for Learning Rigid-body Forward Dynamics with Gaussian Process Force Priors

Lucas Rath, Andreas René Geist, Sebastian Trimpe
Proceedings of the 5th Conference on Robot Learning, PMLR 164:101-111, 2022.

Abstract

If a robot’s dynamics are difficult to model solely through analytical mechanics, it is an attractive option to directly learn it from data. Yet, solely data-driven approaches require considerable amounts of data for training and do not extrapolate well to unseen regions of the system’s state space. In this work, we emphasize that when a robot’s links are sufficiently rigid, many analytical functions such as kinematics, inertia functions, and surface constraints encode informative prior knowledge on its dynamics. To this effect, we propose a framework for learning probabilistic forward dynamics that combines physics knowledge with Gaussian processes utilizing automatic differentiation with GPU acceleration. Compared to solely data-driven modeling, the model’s data efficiency improves while the model also respects physical constraints. We illustrate the proposed structured model on a seven joint robot arm in PyBullet. Our implementation of the proposed framework can be found here: https://git.io/JP4Fs

Cite this Paper

BibTeX
@InProceedings{pmlr-v164-rath22a, title = {Using Physics Knowledge for Learning Rigid-body Forward Dynamics with Gaussian Process Force Priors}, author = {Rath, Lucas and Geist, Andreas Ren\'e and Trimpe, Sebastian}, booktitle = {Proceedings of the 5th Conference on Robot Learning}, pages = {101--111}, year = {2022}, editor = {Faust, Aleksandra and Hsu, David and Neumann, Gerhard}, volume = {164}, series = {Proceedings of Machine Learning Research}, month = {08--11 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v164/rath22a/rath22a.pdf}, url = {https://proceedings.mlr.press/v164/rath22a.html}, abstract = {If a robot’s dynamics are difficult to model solely through analytical mechanics, it is an attractive option to directly learn it from data. Yet, solely data-driven approaches require considerable amounts of data for training and do not extrapolate well to unseen regions of the system’s state space. In this work, we emphasize that when a robot’s links are sufficiently rigid, many analytical functions such as kinematics, inertia functions, and surface constraints encode informative prior knowledge on its dynamics. To this effect, we propose a framework for learning probabilistic forward dynamics that combines physics knowledge with Gaussian processes utilizing automatic differentiation with GPU acceleration. Compared to solely data-driven modeling, the model’s data efficiency improves while the model also respects physical constraints. We illustrate the proposed structured model on a seven joint robot arm in PyBullet. Our implementation of the proposed framework can be found here: https://git.io/JP4Fs} }
Endnote
%0 Conference Paper %T Using Physics Knowledge for Learning Rigid-body Forward Dynamics with Gaussian Process Force Priors %A Lucas Rath %A Andreas René Geist %A Sebastian Trimpe %B Proceedings of the 5th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2022 %E Aleksandra Faust %E David Hsu %E Gerhard Neumann %F pmlr-v164-rath22a %I PMLR %P 101--111 %U https://proceedings.mlr.press/v164/rath22a.html %V 164 %X If a robot’s dynamics are difficult to model solely through analytical mechanics, it is an attractive option to directly learn it from data. Yet, solely data-driven approaches require considerable amounts of data for training and do not extrapolate well to unseen regions of the system’s state space. In this work, we emphasize that when a robot’s links are sufficiently rigid, many analytical functions such as kinematics, inertia functions, and surface constraints encode informative prior knowledge on its dynamics. To this effect, we propose a framework for learning probabilistic forward dynamics that combines physics knowledge with Gaussian processes utilizing automatic differentiation with GPU acceleration. Compared to solely data-driven modeling, the model’s data efficiency improves while the model also respects physical constraints. We illustrate the proposed structured model on a seven joint robot arm in PyBullet. Our implementation of the proposed framework can be found here: https://git.io/JP4Fs
APA
Rath, L., Geist, A.R. & Trimpe, S.. (2022). Using Physics Knowledge for Learning Rigid-body Forward Dynamics with Gaussian Process Force Priors. Proceedings of the 5th Conference on Robot Learning, in Proceedings of Machine Learning Research 164:101-111 Available from https://proceedings.mlr.press/v164/rath22a.html.

Related Material