DynAMO: multi-agent reinforcement learning for dynamic anticipatory mesh optimization with applications to hyperbolic conservation laws. (English) Zbl 1536.65098
Summary: We introduce DynAMO, a reinforcement learning paradigm for Dynamic Anticipatory Mesh Optimization. Adaptive mesh refinement is an effective tool for optimizing computational cost and solution accuracy in numerical methods for partial differential equations. However, traditional adaptive mesh refinement approaches for time-dependent problems typically rely only on instantaneous error indicators to guide adaptivity. As a result, standard strategies often require frequent remeshing to maintain accuracy. In the DynAMO approach, multi-agent reinforcement learning is used to discover new local refinement policies that can anticipate and respond to future solution states by producing meshes that deliver more accurate solutions for longer time intervals. By applying DynAMO to discontinuous Galerkin methods for the linear advection and compressible Euler equations in two dimensions, we demonstrate that this new mesh refinement paradigm can outperform conventional threshold-based strategies while also generalizing to different mesh sizes, remeshing and simulation times, and initial conditions.
MSC:
| 65M50 | Mesh generation, refinement, and adaptive methods for the numerical solution of initial value and initial-boundary value problems involving PDEs |
| 65M60 | Finite element, Rayleigh-Ritz and Galerkin methods for initial value and initial-boundary value problems involving PDEs |
| 35L65 | Hyperbolic conservation laws |
Keywords:
adaptive mesh refinement; finite element methods; scientific machine learning; reinforcement learning; hyperbolic conservation lawsReferences:
| [1] | Takaki, Tomohiro; Fukuoka, Toshimichi; Tomita, Yoshihiro, Phase-field simulation during directional solidification of a binary alloy using adaptive finite element method, J. Cryst. Growth, 283, 1, 263-278, 2005 |
| [2] | Berger, Lorenz; Bordas, Rafel; Kay, David; Tavener, Simon, A stabilized finite element method for finite-strain three-field poroelasticity, Comput. Mech., 60, 1, 51-68, 2017 · Zbl 1386.74134 |
| [3] | Kumar, Sarvesh; Oyarzúa, Ricardo; Ruiz-Baier, Ricardo; Sandilya, Ruchi, Conservative discontinuous finite volume and mixed schemes for a new four-field formulation in poroelasticity, ESAIM: Math. Model. Numer. Anal., 54, 1, 273-299, 2020 · Zbl 1511.65114 |
| [4] | Kitzmann, Daniel; Bolte, Jan; Beate, A.; Patzer, C., Discontinuous Galerkin finite element methods for radiative transfer in spherical symmetry, Astron. Astrophys., 595, A90, 2016 |
| [5] | Van de Vosse, F. N.; De Hart, J.; Van Oijen, C. H.G. A.; Bessems, D.; Gunther, T. W.M.; Segal, A.; Wolters, B. J.B. M.; Stijnen, J. M.A.; Baaijens, F. P.T., Finite-element-based computational methods for cardiovascular fluid-structure interaction, J. Eng. Math., 47, 335-368, 2003 · Zbl 1057.74047 |
| [6] | Becker, Roland; Rannacher, Rolf, An optimal control approach to a posteriori error estimation in finite element methods, Acta Numer., 10, 1-102, May 2001 · Zbl 1105.65349 |
| [7] | Tyson, William C.; Swirydowicz, Katarzyna; Derlaga, Joseph M.; Roy, Christopher J.; de Sturler, Eric, Improved functional-based error estimation and adaptation without adjoints, (46th AIAA Fluid Dynamics Conference, June 2016, American Institute of Aeronautics and Astronautics) |
| [8] | Tyson, William C.; Roy, Christopher J., A higher-order error estimation framework for finite-volume CFD, J. Comput. Phys., 394, 632-657, October 2019 · Zbl 1452.76136 |
| [9] | Tyson, William C.; Yan, Gary K.; Roy, Christopher J.; Ollivier-Gooch, Carl F., Relinearization of the error transport equations for arbitrarily high-order error estimates, J. Comput. Phys., 397, Article 108867 pp., November 2019 · Zbl 1453.65282 |
| [10] | Wang, Hongyu; Roy, Christopher J., Error transport equations implementation for discontinuous Galerkin methods, J. Comput. Phys., 474, Article 111760 pp., February 2023 · Zbl 07640532 |
| [11] | Moosavi, Azam; Ştefănescu, Răzvan; Sandu, Adrian, Multivariate predictions of local reduced-order-model errors and dimensions, Int. J. Numer. Methods Eng., 113, 3, 512-533, October 2017 · Zbl 07874306 |
| [12] | Drohmann, Martin; Carlberg, Kevin, The ROMES method for statistical modeling of reduced-order-model error, SIAM/ASA J. Uncertain. Quantificat., 3, 1, 116-145, January 2015 · Zbl 1322.65029 |
| [13] | Dyck, D. N.; Lowther, D. A.; McFee, S., Determining an approximate finite element mesh density using neural network techniques, IEEE Trans. Magn., 28, 2, 1767-1770, 1992 |
| [14] | Paszyński, Maciej; Grzeszczuk, Rafał; Pardo, David; Demkowicz, Leszek, Deep learning driven self-adaptive hp finite element method, (International Conference on Computational Science, 2021, Springer), 114-121 |
| [15] | Służalec, Tomasz; Grzeszczuk, Rafał; Rojas, Sergio; Dzwinel, Witold; Paszyński, Maciej, Quasi-optimal hp-finite element refinements towards singularities via deep neural network prediction, Comput. Math. Appl., 142, 157-174, July 2023 · Zbl 1538.65550 |
| [16] | Gillette, Andrew; Keith, Brendan; Petrides, Socratis, Learning robust marking policies for adaptive mesh refinement, SIAM J. Sci. Comput., 46, A264-A289, 2023 · Zbl 1532.65119 |
| [17] | Bohn, Jan; Feischl, Michael, Recurrent neural networks as optimal mesh refinement strategies, Comput. Math. Appl., 97, 61-76, 2021 · Zbl 1524.65759 |
| [18] | Chen, Guodong; Fidkowski, Krzysztof, Output-based error estimation and mesh adaptation using convolutional neural networks: application to a scalar advection-diffusion problem, (AIAA Scitech 2020 Forum, 2020), 1143 |
| [19] | Chen, Guodong; Fidkowski, Krzysztof J., Output-based adaptive aerodynamic simulations using convolutional neural networks, Comput. Fluids, 223, Article 104947 pp., 2021 · Zbl 1521.76280 |
| [20] | Chakraborty, Ayan; Wick, Thomas; Zhuang, Xiaoying; Rabczuk, Timon, Multigoal-oriented dual-weighted-residual error estimation using deep neural networks, 2021 |
| [21] | Roth, Julian; Schröder, Max; Wick, Thomas, Neural network guided adjoint computations in dual weighted residual error estimation, SN Appl. Sci., 4, 2, 1-17, 2022 |
| [22] | Chedid, R.; Najjar, N., Automatic finite-element mesh generation using artificial neural networks-part I: prediction of mesh density, IEEE Trans. Magn., 32, 5, 5173-5178, 1996 |
| [23] | Pfaff, Tobias; Fortunato, Meire; Sanchez-Gonzalez, Alvaro; Battaglia, Peter W., Learning mesh-based simulation with graph networks, 2020, arXiv preprint |
| [24] | Huang, Keefe; Krügener, Moritz; Brown, Alistair; Menhorn, Friedrich; Bungartz, Hans-Joachim; Hartmann, Dirk, Machine learning-based optimal mesh generation in computational fluid dynamics, 2021, arXiv preprint |
| [25] | Zhang, Zheyan; Wang, Yongxing; Jimack, Peter K.; Meshingnet, He Wang, MeshingNet: A new mesh generation method based on deep learning, (International Conference on Computational Science, 2020, Springer), 186-198 |
| [26] | Song, Wenbin; Zhang, Mingrui; Wallwork, Joseph G.; Gao, Junpeng; Tian, Zheng; Sun, Fanglei; Piggott, Matthew D.; Chen, Junqing; Shi, Zuoqiang; Chen, Xiang, M2N: mesh movement networks for PDE solvers, 2022, arXiv preprint |
| [27] | Chan, Chiu Ling; Scholz, Felix; Takacs, Thomas, Locally refined quad meshing for linear elasticity problems based on convolutional neural networks, 2022, arXiv preprint |
| [28] | Yang, Jiachen; Dzanic, Tarik; Petersen, Brenden; Kudo, Jun; Mittal, Ketan; Tomov, Vladimir; Camier, Jean-Sylvain; Zhao, Tuo; Zha, Hongyuan; Kolev, Tzanio; Anderson, Robert; Faissol, Daniel, Reinforcement learning for adaptive mesh refinement, (Proceedings of the 26th International Conference on Artificial Intelligence and Statistics. Proceedings of the 26th International Conference on Artificial Intelligence and Statistics, 25-27 Apr 2023. Proceedings of the 26th International Conference on Artificial Intelligence and Statistics. Proceedings of the 26th International Conference on Artificial Intelligence and Statistics, 25-27 Apr 2023, Proceedings of Machine Learning Research, vol. 206, 2023, PMLR), 5997-6014 |
| [29] | Yang, Jiachen; Mittal, Ketan; Dzanic, Tarik; Petrides, Socratis; Keith, Brendan; Petersen, Brenden; Faissol, Daniel; Anderson, Robert, Multi-agent reinforcement learning for adaptive mesh refinement, (Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems. Proceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems, AAMAS ’23, 2023), 14-22 |
| [30] | Freymuth, Niklas; Dahlinger, Philipp; Würth, Tobias; Kärger, Luise; Neumann, Gerhard, Swarm reinforcement learning for adaptive mesh refinement, 2023 |
| [31] | Foucart, Corbin; Charous, Aaron; Lermusiaux, Pierre F. J., Deep reinforcement learning for adaptive mesh refinement, J. Comput. Phys., 491, Article 112381 pp., 2023 · Zbl 07771296 |
| [32] | Hesthaven, Jan S.; Warburton, Tim, Nodal Discontinuous Galerkin Methods, 2008, Springer: Springer New York · Zbl 1134.65068 |
| [33] | Rusanov, V. V., The calculation of the interaction of non-stationary shock waves and obstacles, USSR Comput. Math. Math. Phys., 1, 2, 304-320, January 1962 |
| [34] | Davis, S. F., Simplified second-order Godunov-type methods, SIAM J. Sci. Stat. Comput., 9, 3, 445-473, May 1988 · Zbl 0645.65050 |
| [35] | Sutton, Richard S.; Barto, Andrew G., Reinforcement Learning: An Introduction, 2018, MIT Press · Zbl 1407.68009 |
| [36] | Puterman, Martin L., Markov Decision Processes: Discrete Stochastic Dynamic Programming, 2014, John Wiley & Sons · Zbl 0829.90134 |
| [37] | Oliehoek, Frans A.; Amato, Christopher, A Concise Introduction to Decentralized POMDPs, 2016, Springer · Zbl 1355.68005 |
| [38] | Schulman, John; Wolski, Filip; Dhariwal, Prafulla; Radford, Alec; Klimov, Oleg, Proximal policy optimization algorithms, 2017, arXiv preprint |
| [39] | Tan, Ming, Multi-agent reinforcement learning: independent vs. cooperative agents, (Proceedings of the Tenth International Conference on Machine Learning, 1993), 330-337 |
| [40] | Chang, Yu-Han; Ho, Tracey; Kaelbling, Leslie P., All learning is local: multi-agent learning in global reward games, (Advances in Neural Information Processing Systems, 2004), 807-814 |
| [41] | Anderson, R.; Andrej, J.; Barker, A.; Bramwell, J.; Camier, J.-S.; Cerveny, J.; Dobrev, V.; Dudouit, Y.; Fisher, A.; Kolev, Tz.; Pazner, W.; Stowell, M.; Tomov, V.; Akkerman, I.; Dahm, J.; Medina, D.; Zampini, S., MFEM: a modular finite element methods library, Comput. Math. Appl., 81, 42-74, 2021 · Zbl 1524.65001 |
| [42] | pymfem, PyMFEM: modular finite element methods [software], 2022 |
| [43] | Cockburn, Bernardo; Shu, Chi-Wang, Runge-Kutta discontinuous Galerkin methods for convection-dominated problems, J. Sci. Comput., 16, 3, 173-261, 2001 · Zbl 1065.76135 |
| [44] | Barth, Timothy; Jespersen, Dennis, The design and application of upwind schemes on unstructured meshes, (27th Aerospace Sciences Meeting, January 1989, American Institute of Aeronautics and Astronautics) |
| [45] | Carstensen, Carsten; Feischl, Michael; Page, Marcus; Praetorius, Dirk, Axioms of adaptivity, Comput. Math. Appl., 67, 6, 1195-1253, 2014 · Zbl 1350.65119 |
| [46] | Zienkiewicz, Olgierd Cecil; Zhu, Jian Zhong, The superconvergent patch recovery and a posteriori error estimates. Part 1: the recovery technique, Int. J. Numer. Methods Eng., 33, 7, 1331-1364, 1992 · Zbl 0769.73084 |
| [47] | Zienkiewicz, Olgierd Cecil; Zhu, Jian Zhong, The superconvergent patch recovery and a posteriori error estimates. Part 2: error estimates and adaptivity, Int. J. Numer. Methods Eng., 33, 7, 1365-1382, 1992 · Zbl 0769.73085 |
| [48] | Liang, Eric; Liaw, Richard; Nishihara, Robert; Moritz, Philipp; Fox, Roy; Goldberg, Ken; Gonzalez, Joseph; Jordan, Michael; Stoica, Ion, RLlib: abstractions for distributed reinforcement learning, (International Conference on Machine Learning, 2018, PMLR), 3053-3062 |
| [49] | Schulz-Rinne, Carsten W.; Collins, James P.; Glaz, Harland M., Numerical solution of the Riemann problem for two-dimensional gas dynamics, SIAM J. Sci. Comput., 14, 6, 1394-1414, November 1993 · Zbl 0785.76050 |
| [50] | Liska, Richard; Wendroff, Burton, Comparison of several difference schemes on 1D and 2D test problems for the Euler equations, SIAM J. Sci. Comput., 25, 3, 995-1017, January 2003 · Zbl 1096.65089 |
| [51] | Lax, Peter D.; Liu, Xu-Dong, Solution of two-dimensional Riemann problems of gas dynamics by positive schemes, SIAM J. Sci. Comput., 19, 2, 319-340, January 1998 · Zbl 0952.76060 |
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. In some cases that data have been complemented/enhanced by data from zbMATH Open. This attempts to reflect the references listed in the original paper as accurately as possible without claiming completeness or a perfect matching.
