A biomimetic robotic fish dynamic induction and fish-gathering system based on fish behavioral characteristics

Abstract

This invention discloses a biomimetic robotic fish dynamic induction and fish-gathering system based on fish behavioral characteristics. The system includes a dynamic leadership transfer module, a fish swarm response monitoring module, and a leadership transfer execution module. The system quantifies the intensity of the fish swarm response by real-time monitoring of fish density distribution and movement direction consistency. When the response intensity corresponding to the current leader is detected to be lower than a set threshold, the leadership transfer process is automatically initiated. The system selects a new leader with the best expected effect from multiple biomimetic robotic fish based on a predictive evaluation mechanism and achieves a smooth transition by coordinating the behavioral characteristic intensity of the old and new leaders. This invention can dynamically optimize the allocation of leadership based on the real-time response of the fish swarm, effectively maintaining the continuity and stability of the induction effect and improving fish-gathering efficiency.

Description

Technical Field

[0001] The invention relates to the field of fishery equipment technology, specifically to a biomimetic robotic fish dynamic induction and fish collection system based on fish behavioral characteristics. Background Technology

[0002] With the increasing demand for fisheries resource management and ecological protection, the development of efficient and environmentally friendly fish-collecting technologies is of great significance. Biomimetic robotic fish systems based on fish behavioral characteristics provide a new technological approach for fish-collecting operations. By simulating natural fish behavior, they can guide fish schools in a targeted manner and can be widely applied in resource surveys, ecological monitoring, and fisheries production.

[0003] Existing biomimetic robotic fish attraction technologies primarily rely on pre-defined behavioral patterns or simple environmental feedback mechanisms, which have significant limitations. First, most systems employ fixed leader individuals and static behavioral strategies, lacking the ability to dynamically adjust based on the real-time responses of the fish swarm. Once the fish swarm adapts to the current robotic fish, the attraction effect decreases significantly. Second, current technologies lack precise quantitative assessments of fish swarm responses, typically relying on a single parameter such as swarm density while neglecting crucial behavioral indicators like consistency in movement direction. Furthermore, leadership transfer mechanisms are relatively simple, often employing rotation or random selection, failing to consider the varying attraction effects of different robotic fish on the swarm. These shortcomings make it difficult for the system to maintain a stable long-term attraction effect in complex aquatic environments, limiting its practical application value.

[0004] Therefore, this paper proposes a biomimetic robotic fish dynamic induction and fish gathering system based on fish behavioral characteristics to address the following problems existing in biomimetic robotic fish gathering systems: how to dynamically optimize the allocation of leadership based on the real-time response of the fish swarm; how to establish an accurate evaluation system for the intensity of the fish swarm response; how to achieve a smooth transition of leadership to maintain a continuous induction effect; and how to continuously optimize system performance through historical data analysis. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a biomimetic robotic fish dynamic induction and fish collection system based on fish behavioral characteristics, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a biomimetic robotic fish dynamic induction and fish-gathering system based on fish behavioral characteristics, comprising: The dynamic leadership transfer module is used to automatically select another bionic robotic fish that can elicit a stronger fish response as the new leader when the fish response intensity of the current bionic robotic fish that is the leader is lower than a set threshold, based on the real-time monitored fish response intensity. The fish school response monitoring module quantifies the intensity of the fish school response by analyzing the fish school density distribution and the consistency of movement direction. The fish school density distribution is obtained by calculating the ratio of the number of fish in a unit area to the total number of fish in the total monitoring area, and the consistency of movement direction is obtained by statistically analyzing the angle distribution between the swimming direction of the fish and the movement direction of the bionic robotic fish. The leadership transfer execution module is used to maintain the original leader's current behavioral characteristics during the leadership transfer process, while controlling the new leader to gradually increase the intensity of their behavioral characteristics, so as to achieve a smooth transition of leadership.

[0007] Preferably, the calculation of the fish school reaction intensity adopts a weighted comprehensive evaluation method, in which different weight coefficients are assigned to the fish school density distribution and the consistency of movement direction, and the final reaction intensity value is obtained by linear weighted summation, wherein the weight coefficient of fish school density distribution is set to 0.6-0.8, and the weight coefficient of movement direction consistency is set to 0.2-0.4.

[0008] Preferably, the triggering conditions for the transfer of leadership include at least one of the following: The fish population's response intensity remained below the threshold for 3-5 consecutive monitoring periods; The fish population's response intensity dropped sharply and fell below 70% of the safety threshold within a single monitoring period; The monitoring cycle lasts 5-10 seconds and is dynamically adjusted according to the aquatic environment.

[0009] Preferably, the selection process of the new leader adopts a predictive evaluation mechanism. Based on the historical performance data of each bionic robotic fish, the system uses a multiple linear regression model to predict the intensity of the fish group response that it may elicit when it acts as the leader. The regression model uses the behavioral characteristic parameters of the bionic robotic fish as independent variables and the historical fish group response intensity as dependent variables, and obtains the prediction function by fitting it using the least squares method.

[0010] Preferably, the prediction and evaluation mechanism further includes: Establish a database linking the behavioral characteristics of biomimetic robotic fish with the intensity of fish swarm responses; Based on time series analysis, the expected response intensity of each biomimetic robotic fish in the next 3-5 cycles is calculated. The bionic robotic fish with the highest expected response intensity was chosen as the new leader.

[0011] Preferably, the leadership transfer execution module further includes a transition control submodule, which is used for: Within 10-20 seconds of the new leader assuming leadership responsibilities, continuously monitor changes in the intensity of the fish's response. If the response intensity does not reach 80% of the expected improvement, the leadership transfer process will be restarted. Set the ratio of the behavioral characteristics intensity of the original leader and the new leader during the transition period, and this ratio changes linearly over time.

[0012] Preferably, it also includes a multi-level leadership structure management module, which sets up a main leader and 1-2 auxiliary leaders. When the fish swarm response intensity corresponding to the main leader decreases, the auxiliary leaders first supplement and strengthen it. If the response intensity still cannot be improved within 2-3 monitoring cycles, then a complete transfer of leadership is executed.

[0013] Preferably, it also includes a leadership transfer record analysis module, which records relevant data of each leadership transfer, including the reasons for the transfer, the basis for selecting the new leader, and the changes in the effects after the transfer. The module analyzes the effectiveness of the leadership transfer strategy through data mining algorithms and dynamically optimizes the transfer threshold parameters.

[0014] Preferably, the leadership transfer record analysis module specifically includes: The data acquisition unit records the changes in the intensity of the fish's reaction before and after each transfer, the transfer time, and the success rate. The analysis and evaluation unit uses cluster analysis to classify transfer records and identify the optimal transfer strategy under different environmental conditions. The parameter optimization unit automatically adjusts the trigger threshold for leadership transfer and the parameters of the prediction model based on the analysis results.

[0015] Preferably, it also includes a behavioral feature coordination and control module, which coordinates the behavioral feature expressions of multiple bionic robotic fish during the leadership transfer process to ensure that the induction effect on the fish group is maintained during the transition. This is achieved in the following ways: Establish a behavioral characteristic intensity mapping function to define the relationship between the changes in the behavioral characteristic intensity of the original leader and the new leader; Set a minimum behavioral characteristic strength guarantee value during the transition period to prevent the inducement effect from being interrupted due to the transfer of leadership; The duration of the transition period is dynamically adjusted based on aquatic environmental parameters.

[0016] Compared with the prior art, the beneficial effects of the present invention include: This invention achieves intelligent optimization of leadership allocation for biomimetic robotic fish by establishing a dynamic leadership transfer mechanism based on the intensity of fish swarm responses. The system continuously monitors the density distribution and movement direction consistency of the fish swarm, quantifies response intensity using a weighted evaluation method, and automatically initiates a leadership transfer process when a decrease in the response intensity of the current leader is detected. By analyzing the historical performance data of each biomimetic robotic fish through a predictive evaluation mechanism, a new leader with the best expected effect is selected, and the behavioral characteristic intensity of the old and new leaders is coordinated during the transition period to ensure the continuity of the induction effect. This approach effectively maintains the fish swarm's attention to the biomimetic robotic fish and significantly improves the stability of fish gathering efficiency.

[0017] This invention employs a multi-level leadership architecture and a gradual transition control strategy, effectively avoiding fluctuations in the induced effect caused by changes in leadership. The system establishes a collaborative working mechanism between a primary leader and assistant leaders. When the primary leader's effectiveness declines, the assistant leader first provides supplementary reinforcement; if the effect is still unsatisfactory, a full transfer of leadership is then executed. During the transition phase, a smooth handover process is established by precisely controlling the gradient of changes in the intensity of behavioral characteristics, while a minimum intensity guarantee value is set to prevent interruption of the induced effect. This design ensures that the system maintains stable performance under various environmental conditions, improving the reliability of fish-gathering operations.

[0018] By establishing a leadership transfer record analysis module, this invention achieves a self-optimization function for system parameters. This module records in detail the reasons for each transfer, the selection criteria, and the changes in effects. It uses data mining algorithms to analyze the optimal strategy under different environmental conditions and automatically adjusts trigger thresholds and prediction model parameters. This continuous optimization mechanism based on actual operational data enables the system to gradually adapt to the characteristics of specific aquatic environments, continuously improving fish-attracting efficiency and reducing the workload of manual debugging and maintenance. Attached Figure Description

[0019] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0020] Figure 1 This is a flowchart illustrating the overall system workflow of the present invention.

[0021] Figure 2 This is a detailed flowchart of the fish swarm response monitoring method of the present invention.

[0022] Figure 3 This is a diagram of the dynamic leadership transfer mechanism of the present invention.

[0023] Figure 4 This is a diagram illustrating the multi-level architecture and optimization mechanism of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] As attached Figures 1 to 4 The system shown is a biomimetic robotic fish dynamic induction and aggregating system based on fish behavioral characteristics. The system achieves adaptive optimization of the fish aggregating effect by establishing an intelligent dynamic leadership transfer mechanism.

[0026] The dynamic leadership transfer module serves as the decision-making center, intelligently determining the timing and target of leadership changes based on real-time monitoring data. The fish swarm response monitoring module acts as the sensing front end, accurately quantifying the behavioral response characteristics of the fish swarm through multi-source sensor data fusion technology. The leadership transfer execution module serves as the control end, ensuring a smooth transition of behavioral characteristics between the old and new leaders. These modules interact via industrial Ethernet based on the TCP / IP protocol, using a unified JSON format for data transmission, forming a complete closed-loop control system.

[0027] During system cold start, a deep self-test process is performed, including sensor network synchronous calibration, actuator response testing, and communication quality assessment. The underwater camera array undergoes dark field correction and color balancing, while the acoustic array completes propagation loss compensation settings. Each biomimetic robotic fish performs a full-stroke self-test to verify the response characteristics of the propulsion system, light-emitting unit, and sound-emitting device. The control center synchronously loads a behavioral characteristic knowledge base and establishes a dynamic parameter mapping table, including monitoring cycle parameters, reaction intensity threshold parameters, and transition time parameters under different water quality conditions. The environmental baseline establishment module collects initial water profile data and generates environmental feature vectors, providing a benchmark reference for subsequent adaptive adjustments.

[0028] The system operates on the basis of fish gregarious behavior. By continuously monitoring the intensity of the fish swarm's response to the current leader, it automatically triggers a leadership transfer process when a decline in the induction effect is detected. It selects the most promising new leader from multiple candidate bionic robotic fish and maintains the continuity of the induction effect through a gradual transition strategy based on a time function. This design effectively solves the problem of induction effect decay caused by fish swarm adaptation in traditional fixed leadership models, while overcoming the uncertainty of random rotation strategies.

[0029] Furthermore, the system employs a weighted comprehensive evaluation method to calculate the intensity of the fish school's response. This method quantifies the complex behavioral characteristics of the fish school into a calculable comprehensive index. In specific implementation, .

[0030] in As a comprehensive response intensity output value, its physical meaning is a quantitative expression of the overall attention and willingness of the fish swarm to the current bionic robotic fish leader. This parameter directly reflects the effectiveness of the induction effect.

[0031] The parameter representing the fish density distribution is obtained by real-time analysis of the video stream captured by the underwater camera through the image processing unit. The specific calculation process includes three steps: background subtraction, target detection, and density estimation. This parameter reflects the degree of spatial aggregation of the fish swarm around the bionic robotic fish.

[0032] The sensor network adopts a heterogeneous multi-source architecture, comprising an optical observation layer, an acoustic sensing layer, and a motion sensing layer. The optical layer deploys a combination of wide-angle monitoring cameras and zoom tracking cameras, responsible for large-area monitoring and tracking of key targets, respectively. The acoustic layer is equipped with a multi-band sonar array, overcoming water absorption effects through frequency diversity technology. The motion sensing layer collects the motion state of the biomimetic robotic fish through an inertial measurement unit. The data fusion center employs a spatiotemporal registration algorithm to unify sensing data from different sources into a common coordinate system, improving sensing reliability through confidence weighting.

[0033] The directional consistency coefficient is derived from fish swarm trajectory data collected by multibeam sonar. It is obtained by calculating the variance of the angle between the swimming direction of individual fish and the swimming direction of the bionic robotic fish. This parameter reflects the level of synchronization between the behavior of the fish swarm and the behavior of the leader.

[0034] Weighting coefficient and The design is based on fish behavioral studies, which show that spatial aggregation has a higher weight than directional synchronization during induction. The advantage of this multi-parameter fusion assessment method is that it can comprehensively capture multiple dimensions of fish school responses, avoiding the one-sidedness of single-indicator assessments. By weighted summation, behavioral parameters of different dimensions are unified into standardized response intensity values, providing a reliable data foundation for subsequent decision-making.

[0035] Furthermore, the leadership transfer triggering mechanism adopts an intelligent decision-making model based on multi-condition judgment. The purpose of this design is to ensure timeliness while avoiding false triggering.

[0036] The system maintains a circular buffer that stores response intensity data for the most recent 32 monitoring periods, a design that takes into account the temporal correlation of fish behavior.

[0037] A circular buffer enables intelligent data lifecycle management, employing a data update strategy based on information entropy. The buffer is divided into multiple logical areas: a real-time data area stores the latest collected raw data, a feature data area stores extracted behavioral feature vectors, and a statistical data area maintains sliding window statistics. The data quality management module monitors data integrity indicators in real time, initiating a data reconstruction process when the data loss rate exceeds a threshold. The buffer also implements data version management, supporting historical state retrospection and decision-making process auditing.

[0038] The specific implementation process of the dynamic adjustment mechanism for the monitoring cycle is as follows: the environmental sensing unit obtains water transparency data through a turbidity sensor, obtains water flow velocity data through an electromagnetic current meter, and then determines the current monitoring cycle parameters according to the preset environmental classification rules.

[0039] In clear, slow-flowing water environments, fish behavior is relatively stable, and the following methods are employed: A longer monitoring cycle can reduce the system's computational load; in turbid, fast-flowing environments, fish behavior changes rapidly, requiring the use of... A shorter monitoring cycle can improve the system response speed.

[0040] In the dual-path architecture that triggers the judgment, the trend analysis path detects continuous trends through a sliding window algorithm. The trend of reaction intensity changes within a cycle is judged based on the principle of the cumulative effect of fish school adaptation. When the reaction intensity is consistently below the threshold of 0.6, it indicates that the induction effect of the current leader is systematically decaying.

[0041] The anomaly detection path monitors the rate of change in reaction intensity in real time between adjacent weeks using differential calculations. In response to sudden environmental disturbances or fish fright, when the intensity decreases by more than [a certain percentage], [the path will detect the anomaly]. Furthermore, a transfer is triggered immediately when the absolute value is below 0.42.

[0042] This multi-level triggering mechanism works by first judging trends to ensure the stability of the system response, and then combining anomaly detection to ensure timely handling of emergencies, thus balancing the accuracy and timeliness of the system response.

[0043] Furthermore, the new leader selection mechanism employs a predictive evaluation model based on historical data, introducing statistical prediction methods into the leadership decision-making process of the biomimetic robotic fish. Multiple linear regression model. Each variable in the text has a clear meaning: This indicates the predicted response intensity and serves as the basis for decision-making in the model output; The standardized motion speed parameter reflects the swimming vitality level of the biomimetic robotic fish; The standardized steering frequency parameters reflect the complexity of the motion trajectory. This is a standardized luminous intensity parameter, corresponding to the intensity of the visual stimulus. Regression coefficient. The least squares method is used to fit historical data, and the calculation process involves minimizing the sum of squared prediction errors by solving a system of normal equations.

[0044] In the actual prediction process, the system first performs z-score standardization on the current behavioral parameters of the candidate bionic robotic fish to eliminate the influence of dimensions before substituting them into the trained regression equation to calculate the expected response intensity. Scientific decision-making is achieved by establishing a statistical relationship between behavioral parameters and induction effects. This quantitatively evaluates the potential induction capabilities of each candidate bionic robotic fish, avoiding the uncertainty of subjective selection. Simultaneously, the linearity of the regression model ensures computational efficiency and meets real-time requirements.

[0045] The model maintenance system implements full lifecycle management, including version control, performance monitoring, and online updates. The training data management module implements data quality control, ensuring the representativeness of the training set through diversity sampling. Model validation employs temporal cross-validation to avoid overfitting. The update decision module triggers a model update based on prediction bias trend analysis, when the average absolute percentage error exceeds a warning threshold for multiple consecutive periods. Historical model versions are retained during the update process, supporting a fast rollback mechanism.

[0046] Furthermore, the integration of time series analysis capabilities enhances the temporal dimension of forecasting and evaluation. (ARIMA model) The introduction of this technology enables the system to capture the dynamic changes in the fish's response.

[0047] in This represents an autoregressive operator that reflects the continued influence of historical response strength on the current value; It is a difference operator used to eliminate the nonstationarity of sequences; It is a moving average operator that characterizes the persistent impact of historical prediction errors.

[0048] The time series modeling process automates parameter optimization, comprising four stages: data preprocessing, model selection, parameter estimation, and model validation. The preprocessing stage detects and handles missing and outlier values ​​to ensure data quality. The model selection stage compares the merits of different parameter combinations using information criteria. The parameter estimation stage employs robust estimation methods to reduce the impact of outliers. The validation stage evaluates the model's generalization ability through out-of-sample prediction, ensuring the model's reliability in practical applications.

[0049] Model parameters were determined using maximum likelihood estimation, and the prediction time domain was set to the future. A monitoring cycle is used to balance forecast accuracy and timeliness.

[0050] The system weightedly fuses the trend prediction results from the ARIMA model with the parameter prediction results from the regression model. This dual prediction mechanism combines the advantages of cross-sectional data analysis and time series analysis, taking into account both the individual characteristics of the biomimetic robotic fish and the dynamic patterns of fish school responses. Multi-angle prediction improves the comprehensiveness and reliability of decision-making, avoiding the limitations of a single prediction method.

[0051] Furthermore, the leadership transfer execution module adopts a gradual transition control strategy based on a time function to ensure the continuity of the induction effect during the leadership handover process.

[0052] Intensity control algorithm The parameter settings are based on: This ensures that the new leader has sufficient visibility in the initial phase. Transition time corresponding to full power operation The setting is based on the physiological characteristics of fish's attention shift.

[0053] Complementary functions This ensures the symmetry of the power shift between the old and new leaders. The monitoring unit's detection settings at key time points are as follows: Time-based testing ensured the effectiveness verification during the transition period. Timely inspection confirms handover completion. Time-based detection provides an additional safety margin.

[0054] The intensity control system adopts a hierarchical adjustment architecture, including a decision-making layer, a coordination layer, and an execution layer. The decision-making layer generates the desired intensity curve, the coordination layer resolves coupling conflicts between multiple actuators, and the execution layer achieves precise output control. Each actuator is equipped with an online calibration function to correct input-output characteristics in real time. The safety monitoring module continuously monitors the system status and automatically switches to degrade mode when actuator saturation or abnormal response is detected to ensure safe system operation.

[0055] This gradual control strategy avoids abrupt changes in fish perception by smoothing out intensity variations, maintains the continuity of the induction signal, and effectively prevents fish dispersal that may occur during leadership changes.

[0056] Furthermore, the multi-level leadership architecture provides system redundancy and performance gradients, extending the single leadership model into a collaborative guidance system. The primary leader undertakes the main guidance tasks, and the intensity of its behavioral characteristics remains at full power, which is the foundational level for the normal operation of the system.

[0057] When the detected response intensity of the primary leader drops to the threshold of 0.6 At that time, the first assistant leader began with The intensity participates in the induction. If after After one monitoring cycle, the response intensity still did not improve, and a second auxiliary leader joined and... The intensity of their coordinated work forms a triple-layered induction protection.

[0058] The intensity adjustment of the auxiliary leader adopts a fuzzy control algorithm, which dynamically calculates the compensation value based on the deviation of the response intensity and its rate of change.

[0059] The intelligent compensation controller implements an adaptive adjustment mechanism, and the fuzzy rule base supports online updates. The knowledge base maintenance module records historical control effects and optimizes rule weights through data mining. Membership function parameters are dynamically adjusted based on actual operating data to improve control accuracy. The rule inference engine supports uncertain reasoning, maintaining control smoothness when handling boundary conditions. The performance evaluation module continuously monitors control performance, providing feedback for system optimization.

[0060] The principle is to achieve precise replenishment of the inductive force through intelligent adjustment. The multi-level architecture ensures the system's operating efficiency under normal conditions, provides emergency protection in case of failure, and avoids resource waste through a gradient activation mechanism.

[0061] Furthermore, the leadership transfer record analysis module enables the system to continuously self-optimize, allowing it to learn from experience. K-means clustering algorithm. This enables the system to automatically discover patterns in historical data, among which The number of clusters is represented by the elbow rule; Data points that include features such as environmental parameters and transfer effects; As the cluster center, it represents the typical characteristics of this type of transition event.

[0062] The gradient descent method is used in the parameter optimization unit. The system parameters are approximated to the optimal solution through iterative updates, where the learning rate... Control the update step size, It indicates the direction of parameter adjustment. The working principle of this self-optimization mechanism is to discover the optimal parameter configuration through data mining. Its advantage lies in enabling the system to automatically adapt to the environmental characteristics of specific water areas, reducing the need for manual parameter adjustment.

[0063] Furthermore, the three specialized processing units of the record analysis module constitute a complete data processing chain. The data acquisition unit records transfer logs in JSON format, a design that facilitates data serialization and transmission.

[0064] The improved K-means algorithm used in the analysis and evaluation unit assesses clustering quality through silhouette coefficients, ensuring the reasonableness of the classification results. The parameter optimization unit employs the Q-learning algorithm. The system is guided to explore the optimal strategy through a reward mechanism, where the state... Includes environmental conditions and system status, actions Decision-making based on corresponding parameters, and rewards Based on the transfer effect setting, this reinforcement learning mechanism accumulates decision-making experience through trial and error. Its advantage lies in its ability to autonomously explore the parameter space and find the long-term optimal strategy.

[0065] Furthermore, the behavioral feature coordination and control module achieves refined management of the transition process through an environment adaptive algorithm. (Sine function) Used to describe the variation pattern of behavioral characteristic intensity, where the amplitude coefficient Adjusted according to water turbidity, and increased in turbid water environments to ensure signal significance; angular frequency Control the rate of change, set according to the fish's reaction sensitivity; phase Ensure the consistency of the change in the strength of the old and new leaders; offset. Set a minimum strength guarantee value.

[0066] Transition period duration formula In the middle, the reference time Based on research on fish behavioral adaptation cycles, the adjustment coefficient was determined. The flow rate was obtained through experimental optimization. Data is acquired in real time through sensors. The working principle of this environmental adaptive coordinated control is to dynamically adjust parameters to match environmental conditions. Its advantage lies in ensuring that the system maintains optimal transition performance under different operating conditions.

[0067] To provide a more complete picture of the system workflow, the following section details the collaborative operation of each module in specific application scenarios.

[0068] During the system initialization phase, the dynamic leadership transfer module loads the historical database and model parameters, the fish swarm response monitoring module performs sensor calibration and data channel testing, and the leadership transfer execution module completes the biomimetic robotic fish's self-check and communication link establishment.

[0069] Once the system is operational, the fish swarm response monitoring module will... Frequency-based data acquisition is used to calculate reaction intensity values ​​in real time via a parallel processing pipeline. Assuming that at a certain monitoring moment, the image processing unit calculates the fish density distribution parameters... The acoustic analysis unit yields the directional consistency coefficient. The reaction intensity is then calculated using the weighted formula. .

[0070] The multi-source decision-making system implements an evidence theory fusion framework, integrating three information sources: regression prediction, time series prediction, and expert rules. Each information source outputs a prediction result and its confidence level, which are then synthesized into a final decision using DS evidence theory. A conflict detection module identifies discrepancies between different information sources, initiating expert intervention when the conflict level is too high. A decision tracing module records the contribution of each information source, providing a basis for subsequent optimization.

[0071] When the system continuously monitors When the reaction intensity data for each cycle are all below the threshold of 0.6, the dynamic leadership transfer module initiates the decision-making process. The prediction and evaluation unit extracts historical data of each candidate bionic robotic fish from the behavioral feature database, assuming that there are currently... There are three candidate robotic fish with standardized behavioral parameters: Candidate A (0.8, 0.6, 0.7), Candidate B (0.7, 0.8, 0.5), and Candidate C (0.6, 0.7, 0.8). Substituting these parameters into the regression equation... The system calculates predicted values ​​and simultaneously uses the time series analysis unit to predict future trends based on the ARIMA model. After comprehensive evaluation, the system selects candidate A as the new leader.

[0072] The leadership transfer execution module then initiates the transition procedure. The system coordinates the strength changes between the old and new leaders. The monitoring unit verifies the transition effect at key time points to ensure the response strength reaches the expected level. After the handover is complete, the new leader begins full-power operation, and the system records various parameters of this transition for subsequent analysis and optimization.

[0073] The fault-tolerant management system implements a tiered emergency response, activating corresponding contingency plans based on the severity of the anomaly. Minor anomalies trigger adaptive adjustments, with the system automatically correcting parameters to restore normal operation. Moderate anomalies initiate a refactoring process, reallocating system resources to maintain core functionality. Severe anomalies execute safety isolation, disconnecting the faulty unit from the system while maintaining overall operation. Detailed handling logs are recorded for all anomaly events, supporting post-event analysis and system improvement.

[0074] The entire workflow demonstrates the complete technical chain of the system to achieve dynamic optimization of leadership through intelligent perception, predictive evaluation, and coordinated control. The close cooperation of each module ensures the continuous and stable fish swarm induction effect.

[0075] The system health management platform enables predictive maintenance, anticipating potential failures through performance trend analysis. Key performance indicators are continuously monitored, including response latency, decision accuracy, and resource utilization. The maintenance decision support system optimizes maintenance plans and resource allocation based on equipment lifespan models and operational status data. The knowledge management module accumulates maintenance experience, forming a fault diagnosis knowledge base and improving the intelligence level of system operation and maintenance.

[0076] Finally, the following points should be noted: First, in the description of this application, it should be noted that, unless otherwise specified and limited, the terms "installation", "connection", and "linkage" should be interpreted broadly, and can be mechanical or electrical connections, or internal connections between two components, or direct connections. "Up", "down", "left", "right", etc. are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may change. Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other. In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A biomimetic robotic fish dynamic induction and fish-gathering system based on fish behavioral characteristics, characterized in that, Includes the following modules: The dynamic leadership transfer module is used to automatically select another bionic robotic fish that can elicit a stronger fish response as the new leader when the fish response intensity of the current bionic robotic fish that is the leader is lower than a set threshold, based on the real-time monitored fish response intensity. The fish school response monitoring module quantifies the intensity of the fish school response by analyzing the fish school density distribution and the consistency of movement direction. The fish school density distribution is obtained by calculating the ratio of the number of fish in a unit area to the total number of fish in the total monitoring area, and the consistency of movement direction is obtained by statistically analyzing the angle distribution between the swimming direction of the fish and the movement direction of the bionic robotic fish. The multi-level leadership structure management module provides assistant leaders, and priority is given to selecting from among the assistant leaders when leadership is transferred. The leadership transfer execution module is used to maintain the original leader's continued execution of current behavioral characteristics during the leadership transfer process, while controlling the new leader to gradually increase the intensity of their behavioral characteristics, so as to achieve a smooth transition of leadership. The leadership transfer record analysis module records relevant data for each leadership transfer. The behavioral characteristic coordination and control module coordinates the behavioral characteristics of multiple bionic robotic fish during the leadership transfer process, ensuring that the induction effect on the fish group is maintained during the transition.

2. The biomimetic robotic fish dynamic induction and fish-gathering system based on fish behavioral characteristics according to claim 1, characterized in that, The fish school reaction intensity was calculated using a weighted comprehensive evaluation method, which assigned different weight coefficients to the fish school density distribution and the consistency of movement direction, and obtained the final reaction intensity value through linear weighted summation. The weight coefficient of fish school density distribution was set to 0.6-0.8, and the weight coefficient of movement direction consistency was set to 0.2-0.

4.

3. The biomimetic robotic fish dynamic induction and fish-gathering system based on fish behavioral characteristics according to claim 1, characterized in that, The triggering conditions for the transfer of leadership include at least one of the following: The fish population's response intensity remained below the threshold for 3-5 consecutive monitoring periods; The fish population's response intensity dropped sharply and fell below 70% of the safety threshold within a single monitoring period; The monitoring cycle lasts 5-10 seconds and is dynamically adjusted according to the aquatic environment.

4. The biomimetic robotic fish dynamic induction and fish-gathering system based on fish behavioral characteristics according to claim 1, characterized in that, The process of re-selecting a leader from multiple bionic robotic fish adopts a predictive evaluation mechanism. Based on the historical performance data of each bionic robotic fish, the system uses a multiple linear regression model to predict the intensity of the fish swarm response that it may elicit when it acts as the leader. The regression model uses the behavioral characteristic parameters of the bionic robotic fish as independent variables and the historical fish swarm response intensity as the dependent variable, and obtains the prediction function by fitting it using the least squares method.

5. The biomimetic robotic fish dynamic induction and fish-gathering system based on fish behavioral characteristics according to claim 4, characterized in that, The prediction and evaluation mechanism also includes: Establish a database linking the behavioral characteristics of biomimetic robotic fish with the intensity of fish swarm responses; Based on time series analysis, the expected response intensity of each biomimetic robotic fish in the next 3-5 cycles is calculated. The bionic robotic fish with the highest expected response intensity was chosen as the new leader.

6. The biomimetic robotic fish dynamic induction and fish-gathering system based on fish behavioral characteristics according to claim 1, characterized in that, The leadership transfer execution module also includes a transition control submodule, which is used for: Within 10-20 seconds of the new leader assuming leadership responsibilities, continuously monitor changes in the intensity of the fish's response. If the response intensity does not reach 80% of the expected improvement, the leadership transfer process will be restarted. Set the ratio of the behavioral characteristics intensity of the original leader and the new leader during the transition period, and this ratio changes linearly over time.

7. The biomimetic robotic fish dynamic induction and fish-gathering system based on fish behavioral characteristics according to claim 1, characterized in that, The multi-level leadership structure management module sets up a main leader and 1-2 auxiliary leaders. When the fish swarm response intensity corresponding to the main leader decreases, the auxiliary leaders first supplement and strengthen it. If the response intensity still cannot be improved within 2-3 monitoring cycles, then a complete transfer of leadership is executed.

8. The biomimetic robotic fish dynamic induction and fish-gathering system based on fish behavioral characteristics according to claim 1, characterized in that, The leadership transfer record analysis module records relevant data for each leadership transfer, including the reasons for the transfer, the basis for selecting the new leader, and the changes in the effects after the transfer. It analyzes the effectiveness of the leadership transfer strategy through data mining algorithms and dynamically optimizes the transfer threshold parameters.

9. A biomimetic robotic fish dynamic induction and fish-gathering system based on fish behavioral characteristics according to claim 1, characterized in that, The leadership transfer record analysis module specifically includes: The data acquisition unit records the changes in the intensity of the fish's reaction before and after each transfer, the transfer time, and the success rate. The analysis and evaluation unit uses cluster analysis to classify transfer records and identify the optimal transfer strategy under different environmental conditions. The parameter optimization unit automatically adjusts the trigger threshold for leadership transfer and the parameters of the prediction model based on the analysis results.

10. A biomimetic robotic fish dynamic induction and fish-gathering system based on fish behavioral characteristics according to claim 1, characterized in that, The behavioral feature coordination and control module is implemented in the following ways: Establish a behavioral characteristic intensity mapping function to define the relationship between the changes in the behavioral characteristic intensity of the original leader and the new leader; Set a minimum behavioral characteristic strength guarantee value during the transition period to prevent the inducement effect from being interrupted due to the transfer of leadership; The duration of the transition period is dynamically adjusted based on aquatic environmental parameters.

Images