A review on deep reinforcement learning for fluid mechanics
No TL;DR found
Abstract
Deep reinforcement learning (DRL) has recently been adopted in a wide range\nof physics and engineering domains for its ability to solve decision-making\nproblems that were previously out of reach due to a combination of\nnon-linearity and high dimensionality. In the last few years, it has spread in\nthe field of computational mechanics, and particularly in fluid dynamics, with\nrecent applications in flow control and shape optimization. In this work, we\nconduct a detailed review of existing DRL applications to fluid mechanics\nproblems. In addition, we present recent results that further illustrate the\npotential of DRL in Fluid Mechanics. The coupling methods used in each case are\ncovered, detailing their advantages and limitations. Our review also focuses on\nthe comparison with classical methods for optimal control and optimization.\nFinally, several test cases are described that illustrate recent progress made\nin this field. The goal of this publication is to provide an understanding of\nDRL capabilities along with state-of-the-art applications in fluid dynamics to\nresearchers wishing to address new problems with these methods.\n