My current research focuses on deep learning in the area of liquid simulations. The basic idea is to accelerate and refine the simulations with machine learning techniques. Accordingly, part of my research deals with learning reduced representations of fluid simulations or investigating upsampling methods for particle-based data. In recent work, I have also tried to increase the robustness of learning fluid behavior using physically induced biases.
Since 2018 I’m a Ph.D. candidate in Nils Thuerey’s group.
Contact
E-mail: lukas.prantl (at) tum.de
Phone: +49.89.289.19451
Fax: +49.89.289.19462
Room: 02.13.060
Publications
- Lukas Prantl, Boris Bonev, and Nils Thuerey, “Generating Liquid Simulations with Deformation-aware Neural Networks“, ICLR 2019, last update March 2019 [Project]
- Nils Thuerey, Konstantin Weissenow, Lukas Prantl, and Xiangyu Hu, “Deep Learning Methods for Reynolds-Averaged Navier-Stokes Simulations of Airfoil Flows“, AIAA (2019) [Project]
- Lukas Prantl, Nuttapong Chentanez, Stefan Jeschke, and Nils Thuerey, “Tranquil Clouds: Neural Networks for Learning Temporally Coherent Features in Point Clouds“, ICLR 2020, last update January 2020 [Project]
- Benjamin Ummenhofer, Lukas Prantl, Nils Thuerey, and Vladlen Koltun, “Lagrangian Fluid Simulation with Continuous Convolutions“, ICLR 2020, last update December 2019 [Project]
- Lukas Prantl, Jan Bender, Tassilo Kugelstadt, and Nils Thuerey, “Wavelet-based Loss for High-frequency Interface Dynamics“, arXiv 2022, last update November 2022
- Lukas Prantl, Benjamin Ummenhofer, Vladlen Koltun, and Nils Thuerey, “Guaranteed Conservation of Momentum for Learning Particle-based Fluid Dynamics“, NeurIPS 2022, last update November 2022 [Project]
Theses
- Lukas Prantl, “Deep Learning Methods for Simulation of Liquids”, Doctoral Thesis, Technical University of Munich, September 2022. (English)
- Lukas Prantl, “Neural Networks for Particle-based Fluid Data”, M.Sc. Thesis, Technical University of Munich, April 2018. (English)
- Lukas Prantl, “Interactive Interpolation of Fluid Simulations”, B.Sc. Thesis, Technical University of Munich, August 2015. (English)
Talks and Posters
- “Generating Liquid Simulations with Deformation-aware Neural Networks”, poster presentation, ICLR, May 2019
- “Tranquil Clouds: Neural Networks for Learning Temporally Coherent Features in Point Clouds”, spotlight and poster presentation, ICLR, April 2020
- “Guaranteed Conservation of Momentum for Learning Particle-based Fluid Dynamics”, oral and poster presentation, NeurIPS, December 2022
Teaching
- Winter 2019/20: Seminar Deep Learning in Physics
- Summer 2019: Grundlagen Algorithmen und Datenstrukturen
- Winter 2018/19: Seminar Deep Learning in Physics
Supervised Theses
- “Adversarial training of airfoil flow-predictions” – Daniel Matter, Bachelor’s Thesis, 2019