Reinforcement learning for automatic quadrilateral mesh generation: a soft actor-critic approach

Abstract

This paper proposes, implements, and evaluates a Reinforcement Learning (RL) based computational framework for automatic mesh generation. Mesh generation, as one of six basic research directions identified in NASA Vision 2030, is an important area in computational geometry and plays a fundamental role in numerical simulations in the area of finite element analysis (FEA) and computational fluid dynamics (CFD). Existing mesh generation methods suffer from high computational complexity, low mesh quality in complex geometries, and speed limitations. By formulating the mesh generation as a Markov decision process (MDP) problem, we are able to use soft actorcritic, a state-of-the-art RL algorithm, to learn the meshing agent's policy from trials automatically, and achieve a fully automatic mesh generation system without human intervention and any extra clean-up operations, which are typically needed in current commercial software. In our experiments and comparison with a number of representative commercial software, our system demonstrates promising performance with respect to generalizability, robustness, and effectiveness.

Multi-Agent Reinforcement Learning for Adaptive Mesh Refinement

“…[10] proposed a local single-agent RL approach whereby the agent makes a decision for one randomly-selected element at each step. At training time, the global solution is updated every time a Other work at the intersection of FEM and deep learning include reinforcement learning for generating a fixed (nonadaptive) mesh [26], unsupervised clustering for marking and p-refinement [33], and supervised learning for target resolution prediction [23], error estimation [37], and mesh movement [30].…”