Discovering optimal survival strategies for plankton

• Main Author: Foo, Chang Lin
• Other Authors: Marcos
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Mathematics and analysis::Simulations
• Online Access: https://hdl.handle.net/10356/166814
• Institution: Nanyang Technological University

Diel vertical migration (DVM) is the migration cycle of zooplankton that came about as a form of adaptation to the environment. Studies done on DVM typically involve field observations or mathematical models to imitate DVM behavior. Field observations are typically very costly and time-consuming due to the large time and spatial scale of DVM. As such, mathematical models are considered to be a more efficient approach. This project aims to develop a framework to model DVM and considered chemotaxis and negative phototaxis as the underlying movement patterns in DVM. A biased random walk model and a deep reinforcement learning (RL) approach were explored in this project and their relative performances were compared. The RL models were shown to have superior performance over the biased random walk, as the RL models were able to maximize the chemical gradient exposure while avoiding high light intensity exposure depending on the simulation’s time step. However, the RL approach requires more computational resources and time to train an effective model. Curriculum learning was then proposed to reduce the time taken to train these RL models. Finally, the problem designed in this project can be made more complex in further studies by considering more environmental conditions or introducing more complex physics to more accurately model the movement of planktonic organisms in water.