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

DY: Fachverband Dynamik und Statistische Physik

DY 8: Active Matter I (joint session DY/CPP/BP)

DY 8.2: Vortrag

Montag, 12. März 2018, 10:15–10:30, BH-N 243

Active systems learning at the microscale — •Santiago Muiños-Landin¹, Keyan Ghazi-Zahedi², and Frank Cichos¹ — ¹Molecular Nanophotonics. University of Leipzig. Institut for Experimental Physics I — ²Information Theory of Cognitive Systems, Max Planck Institute for Mathematics in the Sciences

Living organisms are able to sense and process information about the environment they live in. They are also able to update this information in order to contruct solutions for real life problems such as finding food or avoiding danger. This active adaption process that in the long run drives the evolution of species is the result of a short time scale evolution of the knowledge of an organism that we know as learning. At the microscale the learning is hampered by stochasticity given that the intrinsic Brownian noise makes critical to build a feedback between stimulus and action. Here, we present a system based on a self-themophoretic microswimmer that allows the application of artifical intelligence algorithms at the microscale. Using reinforcement learning we show that even under noise conditions a system is able to learn how to optimize a simple navigation task. We study the influence of noise and the situation where multiple agents can share information to carry out specific tasks. This way we show how adaptation and intelligent collective behavior can be studied in artificial microswimmers systems.