Schooling Hydrodynamics of 300 Fish

We present 2D and 3D Direct Numerical Simulations of the collective behaviour of 300 swimmers.

The study relies on solving the incompressible Navier Stokes equations using Adaptive Mesh Refinement 

techniques, efficiently deployed on massively parallel computer architectures. The coordinated swimming

is  achieved through synchronised control of the individual swimmer motions in response to the flow environment 

by the Remember and Forget Experience Replay Reinforcement Learning framework developed in our group.

We examine the variation of hydrodynamic benefits with a progressive increase of the school size.

We demonstrate the importance of controlling the vorticity field generated by the synchronised swimmers and 

present hydrodynamic quantities (pressure, drag) for the individual swimmers and the school.