Abstract:The autonomous swimming of fish in a complex flow environment is a nonlinear and intricate system, which is the focus and challenge in various fields. This study proposed a novel simulation framework for artificial intelligence fish. It employed a high-precision immersed boundary-lattice Boltzmann coupling scheme to simulate the interactions between fish and flow in real time, and utilized the soft actor-critic (SAC) deep reinforcement learning algorithm for fish brain decision-making module, which was further divided into a vision-based directional navigation and a lateral line-based flow perception modules, each matched with its corresponding macro-action space. The flow features were extracted using a deep neural network based on a multi-classification algorithm from the data perceived by the lateral line and were linked to the fish actions. The predation swimming and the various Kármán gait swimming were explored in terms of training, simulation, and generalization. Numerical results demonstrated significant advantages in the convergence speed and training efficiency of the SAC algorithm. Owing to the closed-loop “perceive-feedback-memory” mode, intelligent fish can respond in real-time to changes in flow fields based on reward-driven requirements and experience, and the accumulated experience can be directly utilized in other flow fields, and its adaptability, model training efficiency, and generalization were substantially improved.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: how to simulate the adaptive behavior of fish in complex flow environments by combining deep reinforcement learning (DRL) and fluid - structure coupling techniques. Specifically, the research aims to construct a new simulation framework for the autonomous swimming behavior of intelligent fish, in order to achieve real - time feedback control of fish in the flow environment, improve their adaptability and model training efficiency, and enhance their generalization ability. ### Specific manifestations of the problem 1. **Limited action space**: - Existing methods can only achieve fish turning actions and single - frequency tail movements, and cannot simulate the diverse actions of real fish. - This limits the swimming simulation of fish in still water or flow fields with similar hydrodynamic characteristics. 2. **Limitations of flow perception**: - The "perception - feedback - memory" pattern of fish is a closed - loop system. If the flow field information cannot be effectively associated with the response actions of fish, the acquisition of hydrodynamic information based on the numerical simulation platform will lose most of its significance. 3. **Poor model generalization ability**: - In existing methods, the swimming strategies learned by intelligent fish in a specific flow field are difficult to transfer to unfamiliar environments and need to be retrained. ### Solutions To solve the above problems, the research proposes the following improvement measures: 1. **Introduce the lateral line perception module**: - Simulate the function of the fish lateral line, and in addition to visual perception, define a multi - element macroscopic action space corresponding to lateral line perception. 2. **Use the SAC algorithm and DNN**: - The soft actor - critic (SAC) algorithm based on the maximum entropy (ME) objective and the deep neural network (DNN) based on the multi - classification (MC) algorithm achieve the learning and accumulation of fish experience, thus forming a closed - loop fish perception - feedback - memory pattern. 3. **Design a controller with robustness and strong adaptability**: - Finally, an intelligent fish autonomous swimming behavior controller with good robustness, strong adaptability and superior generalization ability is obtained. ### Conclusion Through these improvements, the research shows the significant advantages of the proposed framework in complex flow environments, especially in terms of convergence speed and training efficiency. This provides valuable insights for the development of new intelligent bionic autonomous underwater vehicles (AUVs), and provides new methods for the research of fish ethology and other disciplines.

A numerical simulation research on fish adaption behavior based on deep reinforcement learning and fluid–structure coupling: Implementation of the “perceive-feedback-memory” control system

A numerical simulation method of fish adaption behavior based on deep reinforcement learning and fluid-structure coupling-realization of some lateral line functions

A numerical study of fish adaption behaviors in complex environments with a deep reinforcement learning and immersed boundary–lattice Boltzmann method

Design and implementation of a biomimetic robotic fish with 3D locomotion

A Data-Driven Motion Control Approach for a Robotic Fish

Motion Control of a Multi-Joint Robotic Fish Based on Biomimetic Learning

A numerical simulation of target-directed swimming for a three-link bionic fish with deep reinforcement learning

A GIM-Based Biomimetic Learning Approach for Motion Generation of a Multi-Joint Robotic Fish

Fish Can Save Energy Via Proprioceptive Sensing.

Conceptual Design and Simulation of a BCF Swimming Soft Robotic Fish for Deep-Sea*

Flow pattern investigation of bionic fish by immersed boundary–lattice Boltzmann method and dynamic mode decomposition

FishGym: A High-Performance Physics-based Simulation Framework for Underwater Robot Learning

Efficient Navigation of a Robotic Fish Swimming Across the Vortical Flow Field

Learning for Attitude Holding of a Robotic Fish: An End-to-End Approach With Sim-to-Real Transfer

From Simulation to Reality: A Learning Framework for Fish-Like Robots to Perform Control Tasks

CFD Based Parameter Tuning for Motion Control of Robotic Fish.

Enhancing efficiency and propulsion in bio-mimetic robotic fish through end-to-end deep reinforcement learning

Learning Agile Swimming: An End-to-End Approach without CPGs

Deep-reinforcement-learning-based self-organization of freely undulatory swimmers

Design, Modeling, and Visual Learning-Based Control of Soft Robotic Fish Driven by Super-Coiled Polymers

A Rapid Behavior Simulation Approach for Fishes