A numerical simulation research on fish adaption behavior based on deep reinforcement learning and fluid–structure coupling: The refuge–predation behaviors of intelligent fish under varying environmental pressure DOI
Tao Li, Chunze Zhang, Guibin Zhang

et al.

Physics of Fluids, Journal Year: 2024, Volume and Issue: 36(12)

Published: Dec. 1, 2024

The study of fish swimming behavior and locomotion mechanisms holds substantial scientific engineering significance. With the rapid progression artificial intelligence, integration intelligence with high-precision numerical simulation methods presents a novel highly efficient tool for investigating behavior. In this paper, we proposed perception model that more closely reflects natural logic. By introducing individual vision partially visibility model, physics-based visual system mirrored sensory capabilities live was developed. Furthermore, through construction flow using conventional neural networks, enhanced intelligent fish's ability to detect unsteady hydrodynamic parameters via lateral line. validity new demonstrated experiments, which hunts complex moving targets in flow. Finally, applied refuge/predation behaviors coral reef under varying pressures. results indicated significantly impact survival strategies high velocity, environments, shaping distinct evolutionary decision-making traits. These insights may help understand niche competition different conditions.

Language: Английский

The influence of wall effects on self-propelled performance of brown trout swimming DOI Creative Commons
Guang Yang, Wenjie Li, Hongbo Du

et al.

Physics of Fluids, Journal Year: 2024, Volume and Issue: 36(4)

Published: April 1, 2024

When benthic fish engage in predation, they often swim near the riverbank or close to hard rocks, where are subjected combined effects of side and walls. This study focuses on brown trout employs a three-dimensional numerical model simulate process accelerating from stationary state cruising under influence wall effects. A self-developed subroutine algorithm is applied solve various hydrodynamic parameters swimming. By varying distance between fish's center gravity wall, this explores self-propelled performance efficiency swimming affected by sidewall as well also reveals mechanism that impact body/caudal fin (BCF) mode. The results demonstrate when less than 0.5 times body length fish, can enhance speed thrust, but will reduce. Closer proximity leads increased power consumption decreased efficiency, which disadvantageous for findings reveal unstable experienced offer insights designing biomimetic underwater vehicles leverage creating habitats support BCF

Language: Английский

Citations

5
Conducting eco-hydraulic simulation experiments using embodied intelligent fish DOI Open Access
Tao Li, Chunze Zhang, Guibin Zhang

et al.

Physics of Fluids, Journal Year: 2025, Volume and Issue: 37(3)

Published: March 1, 2025

The design and optimization of fishways other fish-passage facilities are one the critical issues in hydraulic engineering. Traditional methods using physical experiments for fishway face challenges such as uncontrollability fish behavior, limited non-intrusive measurement techniques, nonlinear scale effects. Numerical simulation can provide performance information by analyzing flow velocity, turbulence energy, patterns, but fail to account active responses hydrodynamic characteristics environment. In this study, a research paradigm embodied intelligent optimize eco-hydraulic was attempted. core is platform based on deep reinforcement learning (DRL) immersed boundary–lattice Boltzmann (IB-LB) coupling algorithm. Based platform, endowing with biological tendencies biometric features related perception/feedback/decision-making at individual scale, achieve multimodal perception autonomous decision-making complex digital fields potential habits live fish. Using proposed framework, conduct end-to-end training then deploy trained virtual vertical-slot ecological experiments. Comparative analyses were conducted three different geometric shapes. results demonstrated that new evaluates structural through adaptive response behavior environment, which quantitative guidance from terms passage path, time, energy consumption, etc. This study belonged an individual-scale twin innovative attempt utilize lifeforms simulation-based experimental research.

Language: Английский

Citations

0
Learning obstacle avoidance and predation in complex reef environments with deep reinforcement learning DOI
Ji Hou, Caili He, Tao Li

et al.

Bioinspiration & Biomimetics, Journal Year: 2024, Volume and Issue: 19(5), P. 056014 - 056014

Published: July 18, 2024

The reef ecosystem plays a vital role as habitat for fish species with limited swimming capabilities, serving not only sanctuary and food source but also influencing their behavioral tendencies. Understanding the intricate mechanism through which adeptly navigate moving targets within environments complex water flow, all while evading obstacles maintaining stable postures, has remained challenging prominent subject in realms of behavior, ecology, biomimetics alike. An integrated simulation framework is used to investigate predation problems environments, combining deep reinforcement learning algorithms (DRL) high-precision fluid-structure interaction numerical methods-immersed boundary lattice Boltzmann method (lB-LBM). Soft Actor-Critic (SAC) algorithm improve intelligent fish's capacity random exploration, tackling multi-objective sparse reward challenge inherent real-world scenarios. Additionally, shaping tailored its action purposes been developed, capable capturing outcomes trend characteristics effectively. convergence robustness advantages elucidated this paper are showcased two case studies: one addressing randomly hydrostatic flow field, other focusing on counter-current foraging capture drifting food. A comprehensive analysis was conducted influence significance various types decision-making processes environments.

Language: Английский

Citations

2
A numerical simulation research on fish adaption behavior based on deep reinforcement learning and fluid–structure coupling: The refuge–predation behaviors of intelligent fish under varying environmental pressure DOI
Tao Li, Chunze Zhang, Guibin Zhang

et al.

Physics of Fluids, Journal Year: 2024, Volume and Issue: 36(12)

Published: Dec. 1, 2024

The study of fish swimming behavior and locomotion mechanisms holds substantial scientific engineering significance. With the rapid progression artificial intelligence, integration intelligence with high-precision numerical simulation methods presents a novel highly efficient tool for investigating behavior. In this paper, we proposed perception model that more closely reflects natural logic. By introducing individual vision partially visibility model, physics-based visual system mirrored sensory capabilities live was developed. Furthermore, through construction flow using conventional neural networks, enhanced intelligent fish's ability to detect unsteady hydrodynamic parameters via lateral line. validity new demonstrated experiments, which hunts complex moving targets in flow. Finally, applied refuge/predation behaviors coral reef under varying pressures. results indicated significantly impact survival strategies high velocity, environments, shaping distinct evolutionary decision-making traits. These insights may help understand niche competition different conditions.

Language: Английский

Citations

1