Physics Foundation Models for Efficient Control of Fluid Flows
Abstract
We investigate whether a physics foundation model (PFM) pretrained for PDE forecasting can be repurposed for closed-loop fluid-flow control. We construct a PFM-based controller by attaching lightweight Gaussian actor and critic heads to frozen pretrained representations and train the policy using proximal policy optimization (PPO). We evaluate the approach on HydroGym's two-dimensional Kolmogorov-flow environment using vorticity observations at 64×64 and 128×128 resolutions and Reynolds numbers Re=100 and 200. Across these settings, we compare the frozen PFM controller against a baseline MLP policy trained with the same PPO framework. The PFM controller achieves consistently higher reward while requiring smaller executed action magnitudes over training, indicating improved control performance with reduced control effort. These results provide initial evidence that representations learned through self-supervised next-frame prediction can be transferred directly to closed-loop fluid flow control. Project page: https://mahindrautela.github.io/morph.
Related
- Physics-informed neural networks for viscoelastic fluid flows around a cylinder in a two-dimensional channel
- Real-time quantification of fluid flows around bubbles during directional solidification
- From Dye to Dynamics: Quantitative Measurements of Fluid Flows from Vortex Geometry
- Physics-informed, boundary-constrained Gaussian process regression for the reconstruction of fluid flow fields
- Similarities between the speed limit in relativity and the sound barrier in inviscid fluid flows
BibTeX
@misc{rautela2026physics,
title = {{Physics Foundation Models for Efficient Control of Fluid Flows}},
author = {Mahindra Rautela and Shaifalee Saxena and Alexander Scheinker},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.15173},
eprint = {2610.15173},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.15173},
}
RIS
TY - UNPB
TI - Physics Foundation Models for Efficient Control of Fluid Flows
AU - Mahindra Rautela
AU - Shaifalee Saxena
AU - Alexander Scheinker
PY - 2026
DA - 2026/10/10
PB - arXiv
DO - 10.48550/arxiv.2610.15173
UR - https://arxiv.org/abs/2610.15173
AB - We investigate whether a physics foundation model (PFM) pretrained for PDE forecasting can be repurposed for closed-loop fluid-flow control. We construct a PFM-based controller by attaching lightweight Gaussian actor and critic heads to frozen pretrained representations and train the policy using proximal policy optimization (PPO). We evaluate the approach on HydroGym's two-dimensional Kolmogorov-flow environment using vorticity observations at 64×64 and 128×128 resolutions and Reynolds numbers Re=100 and 200. Across these settings, we compare the frozen PFM controller against a baseline MLP policy trained with the same PPO framework. The PFM controller achieves consistently higher reward while requiring smaller executed action magnitudes over training, indicating improved control performance with reduced control effort. These results provide initial evidence that representations learned through self-supervised next-frame prediction can be transferred directly to closed-loop fluid flow control. Project page: https://mahindrautela.github.io/morph.
ER -