SKM 2023 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

BP: Fachverband Biologische Physik

BP 11: Poster Session I

BP 11.68: Poster

Dienstag, 28. März 2023, 12:30–15:30, P1

Optimal navigation of smart active particles in complex landscapes — •Mischa Putzke and Holger Stark — Technische Universität Berlin, Institut für Theoretische Physik, Straße des 17. Juni 135, 10623 Berlin, Germany

The field of active matter, and in particular microswimmers, is finding more and more applications. Synthetic microswimmers are potentially used for microsurgery and the targeted transport of drugs and genes. This requires smart active particles that can process information.

The mentioned applications require the optimal navigation in complex environments where the self-propelled microswimmer only changes its orientation but not its velocity. Machine learning is often used to solve optimization problems. We employ Q-learning to train the agent to move towards a target while it receives information about the direction and distance of the target. To model the smart active particle we use Langevin dynamics.

We show that the microswimmer with its limited information is able to navigate in complex landscapes such as potential barriers and wells but also in vortical flow and find the fastest trajectories. We also show that the navigation optimization is stable against thermal fluctuations by including thermal noise in the orientation during training. For potential wells and vortical flow, it can also be observed that during training not the entire set of existing trajectories is covered by the microswimmer, thus optimal solutions can remain hidden.