SPNets: Differentiable Fluid Dynamics for Deep Neural Networks

Connor Schenck, Dieter Fox
Proceedings of The 2nd Conference on Robot Learning, PMLR 87:317-335, 2018.

Abstract

In this paper we introduce Smooth Particle Networks (SPNets), a framework for integrating fluid dynamics with deep networks. SPNets adds two new layers to the neural network toolbox: ConvSP and ConvSDF, which enable computing physical interactions with unordered particle sets. We use these layers in combination with standard neural network layers to directly implement fluid dynamics inside a deep network, where the parameters of the network are the fluid parameters themselves (e.g., viscosity, cohesion, etc.). Because SPNets are implemented as a neural network, the resulting fluid dynamics are fully differentiable. We then show how this can be successfully used to learn fluid parameters from data, perform liquid control tasks, and learn policies to manipulate liquids.

Cite this Paper

BibTeX
@InProceedings{pmlr-v87-schenck18a, title = {SPNets: Differentiable Fluid Dynamics for Deep Neural Networks}, author = {Schenck, Connor and Fox, Dieter}, booktitle = {Proceedings of The 2nd Conference on Robot Learning}, pages = {317--335}, year = {2018}, editor = {Billard, Aude and Dragan, Anca and Peters, Jan and Morimoto, Jun}, volume = {87}, series = {Proceedings of Machine Learning Research}, month = {29--31 Oct}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v87/schenck18a/schenck18a.pdf}, url = {https://proceedings.mlr.press/v87/schenck18a.html}, abstract = {In this paper we introduce Smooth Particle Networks (SPNets), a framework for integrating fluid dynamics with deep networks. SPNets adds two new layers to the neural network toolbox: ConvSP and ConvSDF, which enable computing physical interactions with unordered particle sets. We use these layers in combination with standard neural network layers to directly implement fluid dynamics inside a deep network, where the parameters of the network are the fluid parameters themselves (e.g., viscosity, cohesion, etc.). Because SPNets are implemented as a neural network, the resulting fluid dynamics are fully differentiable. We then show how this can be successfully used to learn fluid parameters from data, perform liquid control tasks, and learn policies to manipulate liquids. } }
Endnote
%0 Conference Paper %T SPNets: Differentiable Fluid Dynamics for Deep Neural Networks %A Connor Schenck %A Dieter Fox %B Proceedings of The 2nd Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2018 %E Aude Billard %E Anca Dragan %E Jan Peters %E Jun Morimoto %F pmlr-v87-schenck18a %I PMLR %P 317--335 %U https://proceedings.mlr.press/v87/schenck18a.html %V 87 %X In this paper we introduce Smooth Particle Networks (SPNets), a framework for integrating fluid dynamics with deep networks. SPNets adds two new layers to the neural network toolbox: ConvSP and ConvSDF, which enable computing physical interactions with unordered particle sets. We use these layers in combination with standard neural network layers to directly implement fluid dynamics inside a deep network, where the parameters of the network are the fluid parameters themselves (e.g., viscosity, cohesion, etc.). Because SPNets are implemented as a neural network, the resulting fluid dynamics are fully differentiable. We then show how this can be successfully used to learn fluid parameters from data, perform liquid control tasks, and learn policies to manipulate liquids.
APA
Schenck, C. & Fox, D.. (2018). SPNets: Differentiable Fluid Dynamics for Deep Neural Networks. Proceedings of The 2nd Conference on Robot Learning, in Proceedings of Machine Learning Research 87:317-335 Available from https://proceedings.mlr.press/v87/schenck18a.html.

Related Material