Autonomous navigation of smart microswimmers in non-uniform flow fields

We employ a combination of direct numerical simulations and deep reinforcement learning to investigate the autonomous navigation capabilities of smart microswimmers in nonuniform flow conditions, specifically with an applied zig-zag shear flow. The smart microswimmers are equipped with sensors on their body surface to perceive local hydrodynamic signals, i.e., surface stresses, and have the capability of performing torque-free rotation of the propelling axis, such that by mimicking the ciliary beating around their bodies, which is represented by the azimuthal velocity term C1 in the squirmer model. By focusing on a puller-type swimmer, we explore its performance in three distinct navigation tasks: swimming in the flow (1), shear-gradient (2), and vorticity (3) directions. We first investigate the impact of the C1 mode on swimming performance in steady zig-zag shear flow. We then explore the influence of oscillatory shear flow and its convergence to the non-shear flow navigation as the applied frequency increases. Additionally, we extend our methodology to investigate the collective swimming behavior of multiple swimmers in the shear-gradient direction, revealing their ability to swim collectively in a sinusoidal pattern. Finally, we apply our approach to introduce collective behaviors in bulk multi-swimmer dispersions, targeting regimes previously predicted to exhibit non-cohesive behavior.