A Numerical Study of Fish Adaption Behaviors by Deep Reinforcement Learning and Immersed Boundary–Lattice Boltzmann Method

Download files
Access & Terms of Use
open access
Copyright: Zhu, Yi
The aim of this thesis is to study the adaption behaviors of self-propelled fish in complex environments. In order to do so, a numerical framework is first developed. In this framework, fish swimming in a viscous incompressible flow is simulated with the immersed boundary--lattice Boltzmann method (IB--LBM). Furthermore, a deep recurrent Q-network (DRQN) is incorporated with the IB--LBM to train the fish model to adapt its motion to optimally achieve a specific task, such as prey capture, rheotaxis and Kármán gaiting. Compared to existing models for fish, this work incorporates the fish position, velocity and acceleration into the state space in the DRQN; and it considers the amplitude and frequency action space as well as the historical effects. This framework makes use of the high computational efficiency of the IB--LBM which is of crucial importance for the effective coupling with learning algorithms. Test cases including point-to-point swimming in quiescent flow and position holding both in a uniform stream and a Kármán vortex street have been conducted to show the effectiveness of the proposed framework. With the proposed method, the effect of vision, superficial neuromast (SN) and canal neuromast (CN) in position holding swimming in a uniform flow are then investigated. It is found that the fish is able to hold position with all those sensory methods. The control with vision is most accurate while the control with CN information is least accurate. In addition, the combination of vision, SN and CN will not improve the control with only vision, but the combination of SN and CN outperforms SN or CN alone. The effect of the undulation frequency on fish's behavior in a Kármán vortex street is finally investigated. Result shows that the swimming is most stable and efficient when the fish is synchronizing its tail-beat frequency with the vortex shedding frequency. Higher undulation frequency will decrease the hydrodynamic efficiency, and lower undulation frequency will decrease swimming stability. The effect of the scale of the vortex street on fish behavior is also investigated. Smaller vortex street makes the swimming more stable and less efficient, while larger vortex street makes the swimming unstable but hydrodynamically efficient.
Persistent link to this record
Link to Publisher Version
Link to Open Access Version
Additional Link
Conference Proceedings Editor(s)
Other Contributor(s)
Corporate/Industry Contributor(s)
Publication Year
Resource Type
Degree Type
PhD Doctorate
UNSW Faculty
download public version.pdf 51.64 MB Adobe Portable Document Format
Related dataset(s)