Reinforcement Learning and Optimal Experimental Design for Modeling and Control of Fish Schooling Hydrodynamics

Open access
Date
2023Type
- Doctoral Thesis
ETH Bibliography
yes
Altmetrics
Abstract
This doctoral thesis investigates the modeling and control of fish hydrodynamics. The study consists of two main components: computational modeling of flow fields and sensory cues, and understanding the optimality principles driving fish behavior. The thesis emphasizes the incorporation of hydrodynamic interactions to accurately represent the fish’s environment. The research utilizes reinforcement learning, Bayesian inference, and high-performance computing to analyze natural behavior and flow fields. The insights gained from this study have potential applications in autonomous robot swimmers and may inspire new experiments in biology. Efficient and scalable implemen- tations of computation fluid dynamics, reinforcement learning, and Bayesian inference algorithms are developed to address the computational challenge and pave the way for future advancements in the field. Show more
Permanent link
https://doi.org/10.3929/ethz-b-000637048Publication status
publishedExternal links
Search print copy at ETH Library
Publisher
ETH ZurichOrganisational unit
03499 - Koumoutsakos, Petros (ehemalig) / Koumoutsakos, Petros (former)
More
Show all metadata
ETH Bibliography
yes
Altmetrics