SKM 2023 – wissenschaftliches Programm
BP 11.1: Poster
Dienstag, 28. März 2023, 12:30–15:30, P1
Reinforcement Learning: Optimizing Target-search in a homogeneous environment — •Harpreet Kaur, Michele Caraglio, and Thomas Franosch — Institute for Theoretical Physics, Universität Innsbruck, Innsbruck, Austria.
The target-search problem is an interdisciplinary problem comprising several scales, ranging from bacteria looking for food to robots collecting garbage. Generally, in target search we make decisions in an uncertain and often complex environment with the aim of finding a target as efficiently as possible. The key feature that efficient searching agents have in common is the ability to self-propel. Being able to develop efficient search strategies is crucial, as the time needed to discover a target is often a limiting resource. Here, we address the problem of how a smart microswimmer finds a randomly located target in a homogeneous environment by resorting on machine-learning techniques, particularly Reinforcement Learning. We aim to show that learned strategies are optimal and enable minimization of the search time. Also, our work will provide a better understanding of bacteria behavior and biological foraging.