Reinforcement Learning

Control of perturbed and
heated channel flows

An RL framework was developed for perturbed and heated channel flows. The approach learned to manipulate the flow and thermal fields through active control of the channel's shape.

[PAPER]

Automated nasal cavity
surgery planning

RL was combined with CFD to automatically explore and optimize surgical modifications for obstructed nasal airways based on their impact on flow and temperature.

[PAPER]

HydroGym RL platform
for fluid dynamics

An open, solver-independent RL platform for fluid dynamics was developed, providing standardized environments for flow-control transfer learning research.

[PAPER]

Delayed thermal control
in pulsatile stenotic pipe flow

A dimensionless delay criterion based on the advective dead time was developed to predict policy transfer across dimensions and grid resolutions.