Plug-and-Play Benchmarking of Reinforcement Learning Algorithms for Large-Scale Flow Control

Jannis Becktepe, Aleksandra Franz, Nils Thuerey, Sebastian Peitz
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:7244-7286, 2026.

Abstract

Reinforcement learning (RL) has shown promising results in active flow control (AFC), yet progress in the field remains difficult to assess as existing studies rely on heterogeneous observation and actuation schemes, numerical setups, and evaluation protocols. Current AFC benchmarks attempt to address these issues but heavily rely on external computational fluid dynamics (CFD) solvers, are not fully differentiable, and provide limited 3D and multi-agent support. To overcome these limitations, we introduce FluidGym, the first standalone, fully differentiable benchmark suite for RL in AFC. Built entirely in PyTorch on top of the GPU-accelerated PICT solver, FluidGym runs in a single Python stack, requires no external CFD software, and provides standardized evaluation protocols. We present baseline results with PPO, SAC, DPC, and TD-MPC, and release all environments, datasets, and trained models as public resources. FluidGym enables systematic comparison of control methods, establishes a scalable foundation for future research in learning-based flow control, and is available at github.com/safe-autonomous-systems/fluidgym.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-becktepe26a, title = {Plug-and-Play Benchmarking of Reinforcement Learning Algorithms for Large-Scale Flow Control}, author = {Becktepe, Jannis and Franz, Aleksandra and Thuerey, Nils and Peitz, Sebastian}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {7244--7286}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/becktepe26a/becktepe26a.pdf}, url = {https://proceedings.mlr.press/v306/becktepe26a.html}, abstract = {Reinforcement learning (RL) has shown promising results in active flow control (AFC), yet progress in the field remains difficult to assess as existing studies rely on heterogeneous observation and actuation schemes, numerical setups, and evaluation protocols. Current AFC benchmarks attempt to address these issues but heavily rely on external computational fluid dynamics (CFD) solvers, are not fully differentiable, and provide limited 3D and multi-agent support. To overcome these limitations, we introduce FluidGym, the first standalone, fully differentiable benchmark suite for RL in AFC. Built entirely in PyTorch on top of the GPU-accelerated PICT solver, FluidGym runs in a single Python stack, requires no external CFD software, and provides standardized evaluation protocols. We present baseline results with PPO, SAC, DPC, and TD-MPC, and release all environments, datasets, and trained models as public resources. FluidGym enables systematic comparison of control methods, establishes a scalable foundation for future research in learning-based flow control, and is available at github.com/safe-autonomous-systems/fluidgym.} }
Endnote
%0 Conference Paper %T Plug-and-Play Benchmarking of Reinforcement Learning Algorithms for Large-Scale Flow Control %A Jannis Becktepe %A Aleksandra Franz %A Nils Thuerey %A Sebastian Peitz %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-becktepe26a %I PMLR %P 7244--7286 %U https://proceedings.mlr.press/v306/becktepe26a.html %V 306 %X Reinforcement learning (RL) has shown promising results in active flow control (AFC), yet progress in the field remains difficult to assess as existing studies rely on heterogeneous observation and actuation schemes, numerical setups, and evaluation protocols. Current AFC benchmarks attempt to address these issues but heavily rely on external computational fluid dynamics (CFD) solvers, are not fully differentiable, and provide limited 3D and multi-agent support. To overcome these limitations, we introduce FluidGym, the first standalone, fully differentiable benchmark suite for RL in AFC. Built entirely in PyTorch on top of the GPU-accelerated PICT solver, FluidGym runs in a single Python stack, requires no external CFD software, and provides standardized evaluation protocols. We present baseline results with PPO, SAC, DPC, and TD-MPC, and release all environments, datasets, and trained models as public resources. FluidGym enables systematic comparison of control methods, establishes a scalable foundation for future research in learning-based flow control, and is available at github.com/safe-autonomous-systems/fluidgym.
APA
Becktepe, J., Franz, A., Thuerey, N. & Peitz, S.. (2026). Plug-and-Play Benchmarking of Reinforcement Learning Algorithms for Large-Scale Flow Control. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:7244-7286 Available from https://proceedings.mlr.press/v306/becktepe26a.html.

Related Material