NACA Airfoil optimization with Deep Reinforcement Learning

preview_player
Показать описание
Optimization of a NACA airfoil using Deep Reinforcement Learning (DRL).
Computations : Inviscid panel method
DRL: Proximal policy optimization implemented with Tensorflow 2.0

In this case I decided to use lift coefficient only as the reward, no drag considered.(NB inviscid panel method can only compute lift and pressure coefficient)
It is why at the end I obtain the thick airfoil : NACA8940
Рекомендации по теме
Комментарии
Автор

hi, what is this model being trained on?

aaravdixit
Автор

hi! I am currently working on a similar project in evolutionary optimization, and i was wondering if I can find your codebase somewhere in the web?

patrickbitzer