How plankton can exploit hydrodynamic cues for vertical migration: a deep reinforcement learning approach

Migration up and down the water column is an important task for various types of plankton. While they are carried by the flow, many species are able to swim and sense the surrounding flow. How can they exploit these hydrodynamic cues to migrate more efficiently? We use deep reinforcement learning to address this question. We train a microswimmer at responding to flow velocity gradients so as to maximize long-term vertical displacement. We give a physical interpretation to the learned behaviour in several complex flows and compare it to an approximately optimal solution recently derived analytically [1]. These results allow us to quantify the advantage given by flow sensing for the vertical migration of planktonic organisms.

[1] R. Monthiller, A. Loisy, M. A. Koehl, B. Favier, and C. Eloy. Surfing on turbulence: a strategy for planktonic navigation. Physical Review Letters 129(6), 064502 (2022).