student project of computational science · bow riding or wake drafting? training virtual swimmers...

1
Best practices in marine hitchhiking: bow riding or wake drafting? Training virtual swimmers to find optimal strategies to minimize their effort when swimming near bluff bodies. Fish live their lives in a perpetually unsteady environment, where flow is set into motion and interacts with a multitude of natural and human pro- cesses. Since their birth, fish must learn to cope with this environment, to stabilize their posture and maintain their swimming trajectory. At the same time fish learn to identify situa- tions where swimming effort can be reduced by exploiting pressure differ- ences or shear forces in the flow. In this project we propose employing machine learning algorithms to simu- late this learning process and teach virtual swimmers to reduce their swimming effort by exploiting bluff moving bodies in the flow. We envi- sion the swimmers employing two behaviors observed in nature: bow riding and wake drafting. This would lead to interesting comparison of swimming strategies. The project will provide experience in massively parallel interaction of fluid dynamics solvers and distributed ma- chine learning algorithms. PREREQUISITES Programming in C/C++ Basic Fluid mechanics Understanding of Machine Learning CONTACT Prof. Petros Koumoutsakos Guido Novati e-mail: [email protected] [email protected] Student Project CS E LAB Chair of Computational Science - D-MAVT WWW.CSE-LAB.ETHZ.CH In the CSE Lab, we combine computational methods, computer science tools and do- main specific knowledge to solve scientific and engineering problems in areas such as Fluid Mechanics, Nanotechnology and Life Sciences. The core computational compe- tences of our group are in particle methods and in stochastic optimization techniques. Motivated by challenges in application fields, we focus on identifying the common ele- ments among computational techniques and on formulating common methodological, algorithmic and software structures that facilitate their further development. CHAIR OF COMPUTATIONAL SCIENCE Description: (top) Fish swimming between the vortices shed by a bluff body (Liao et al , 2003). (right) Dolphins swimming in the bow wave of a ship (credit: www.- dolphindock.com.au).

Upload: others

Post on 27-Mar-2020

2 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Student Project OF COMPUTATIONAL SCIENCE · bow riding or wake drafting? Training virtual swimmers to find optimal strategies to minimize their effort when swimming near bluff bodies

Best practices in marine hitchhiking: bow riding or wake drafting?Training virtual swimmers to find optimal strategies to minimize their effort when swimming near bluff bodies. Fish live their lives in a perpetually unsteady environment, where flow is set into motion and interacts with a multitude of natural and human pro-cesses. Since their birth, fish must learn to cope with this environment, to stabilize their posture and maintain their swimming trajectory. At the same time fish learn to identify situa-tions where swimming effort can be reduced by exploiting pressure differ-ences or shear forces in the flow.

In this project we propose employing machine learning algorithms to simu-late this learning process and teach virtual swimmers to reduce their swimming effort by exploiting bluff moving bodies in the flow. We envi-sion the swimmers employing two behaviors observed in nature: bow riding and wake drafting. This would lead to interesting comparison of swimming strategies.The project will provide experience in massively parallel interaction of fluid dynamics solvers and distributed ma-chine learning algorithms.

PREREQUISITESProgramming in C/C++Basic Fluid mechanicsUnderstanding of Machine Learning

CONTACTProf. Petros Koumoutsakos Guido Novatie-mail: [email protected] [email protected]

Stud

ent P

roje

ct

CSELABChair of Computational Science - D-MAVT

WWW.CSE-LAB.ETHZ.CH

In the CSE Lab, we combine computational methods, computer science tools and do-main specific knowledge to solve scientific and engineering problems in areas such as Fluid Mechanics, Nanotechnology and Life Sciences. The core computational compe-tences of our group are in particle methods and in stochastic optimization techniques.Motivated by challenges in application fields, we focus on identifying the common ele-ments among computational techniques and on formulating common methodological, algorithmic and software structures that facilitate their further development.

CH

AIR

OF

CO

MPU

TATI

ON

AL

SCIE

NC

E

Description: (top) Fish swimming

between the vortices shed by a

bluff body (Liao et al, 2003).

(right) Dolphins swimming in the

bow wave of a ship (credit: www.-

dolphindock.com.au).