Artificial swimmers in concentration gradients: Simulation and learning

Resumo

The coupled problem of hydrodynamics and solute transport for artificial microswimmers is studied,
with the Reynolds number set to zero and Peclet numbers (Pe) ranging from 0 to 100. The adopted method is ́
the numerical simulation of the problem with a finite element code based upon the FEniCS library. Details of the
second-order treatment of the time-evolving geometry are presented and shown to be essential for the basic physics
to be respected. The code is first applied to compute the effective solute intake of several artificial swimmers
as functions of the Peclet number. The results confirm that no significant gain in solute intake is achieved by ́
swimming if Pe is smaller than 10. We also consider the swimmers as learning agents inside a fluid that has
a concentration gradient in the far field. We couple the simulations with reinforcement learning processes and
investigate the ability of the agents to learn to move towards the region of higher concentration. The results
demonstrate that microscopic organisms need to solve a challenging learning problem to migrate efficiently when
exposed to chemical inhomogeneities.