Under development...

Active/passive control of flow around bodies using reinforcement learning

This is a widespread test problem of flow around a cylinder in a channel with active control using two small jets at the sides of the body [Rabault et al., J. Fluid Mech., 2019]. The control strategy for drag reduction is approximated by an artificial neural network, which uses velocity probes near the body as input. The coefficients of the artificial neural network are found using deep reinforcement learning (proximal policy optimization method).

Under development...