Developing escape strategies for copepods using Q-learning

Abstract

Copepods are a type of small crustaceans with the ability to move using powerful jumps. This project seeks to develop a model for the dynamics of these copepods in the turbulent flows of the oceans and then apply Q-learning to find good navigation strategies. Contained in this project is also the development of a statistical model of a turbulent flow. It is found that using Q-learning it is possible for the simulated copepods to learn strategies that allow them to avoid dangerous areas almost equally as well as more simple strategies that require more information.