A Reinforcement Learning Approach To Airfoil Shape Optimization

Airfoil shape optimization has been an active research topic since the 1960s, with numerous techniques emerging to address this problem. Recently, the use of supervised machine learning has proven successful in solving non-linear and high-dimensional problems, making them suited for aerodynamic problems. However, supervised learning methods are limited because they rely on user-provided datasets. A reinforcement learning approach can address these limitations and provide a data-driven approach to a high-dimension exploratory problem. This approach has been demonstrated to be efficient when coupled with deep neural networks, making it a promising candidate to address the airfoil shape optimization problem.