A method and system for evaluating fish school cohesion and coordination based on a bionic robotic fish

Abstract

This invention discloses a method and system for evaluating fish swarm cohesion and coordination based on a biomimetic robotic fish. It is applicable to the quantitative evaluation of the fish-attracting effect of fish passage facilities in water conservancy projects, solving problems such as fragmented traditional evaluation indicators, failure to consider internal and external relationships within the fish swarm, and insufficient connection with engineering criteria. The system comprises six modules, including video acquisition, target detection, and tracking. First, it extracts trajectory data of the biomimetic robotic fish and real fish from experimental videos, calculating basic behavioral parameters such as fish-to-fish / fish-to-robotic fish distance and group polarity. After statistical analysis using time windows, distance thresholds are set based on fish body length for normalization. Combining power function nonlinear mapping and fixed / adaptive weight strategies, a comprehensive index of cohesion and coordination is constructed. Time-weighted aggregation yields the overall result, and a grade determination is made. This method automates the entire evaluation process, and the indicators can be linked to engineering criteria, improving the objectivity and repeatability of the evaluation and providing a scientific basis for optimizing biomimetic robotic fish-attracting strategies.

Description

Technical Field

[0001] This invention relates to the field of fish farming, specifically to a method and system for evaluating the cohesion and coordination of fish groups based on biomimetic robotic fish. Background Technology

[0002] With the construction of large and medium-sized water conservancy projects in my country, river connectivity has been disrupted, hindering migratory fish during their upstream migration, spawning, and habitat relocation. Constructing fish passage facilities has become an important measure to restore river connectivity. To improve the fish-attracting effect of facilities such as fishways and fish collection systems, engineering projects typically employ methods such as water flow, sound, electric fields, and light to induce and gather fish schools. However, these traditional fish-attracting methods have limited controllability and specificity.

[0003] Biomimetic robotic fish can mimic the shape and swimming style of real fish, creating flow fields and movement trajectories similar to those of real fish in water, and are considered a promising new method for attracting fish. Existing research mostly focuses on whether the robotic fish swims well or whether real fish approach it, using single or a few indicators such as average distance, dwell time, speed, and polarity to describe the fish-attracting effect. However, there is a lack of a unified evaluation method specifically addressing the "cohesion and coordination of fish groups induced by biomimetic robotic fish."

[0004] In the prior art, patent CN113421345B provides a method for simulating swarm navigation of biomimetic robotic fish based on deep reinforcement learning technology. It constructs a 3D fish swarm environment model using the Unity3D engine and designs a biomimetic robotic fish agent model comprising perception, motion, and decision-making. A curiosity mechanism is introduced into the reward function, and a distributed training framework is built using the PPO2 algorithm to train the agent, achieving 3D simulation of swarm navigation of biomimetic robotic fish. This solves the problems of poor convergence in traditional artificial fish swarm algorithms and the fact that existing simulations are mostly 2D environments. It enables the virtual fish swarm to learn swarm behavior close to nature, and the model can be deployed to real robotic fish. Furthermore, it allows for swarm navigation control by adjusting the speed of the robotic fish.

[0005] Patent CN119105386B discloses a biomimetic robotic fish control method and system based on fish body wave optimization. Addressing the problems of large head swaying and poor robustness in traditional controllers, this method uses sensors to acquire the head swaying state and tail joint motion parameters in real time. It then combines this with model reference adaptive control to estimate fluid dynamic parameters. These parameters are incorporated into a linearized kinematic model to establish an optimal control model, using the optimal tail joint acceleration as the control variable to achieve stable control of the robotic fish. This effectively reduces head swaying, improves the robotic fish's robustness to parameter changes, and allows for adaptation to different tail fin-driven biomimetic robotic fish by simply modifying joint parameters.

[0006] However, the existing technology has the following drawbacks:

[0007] There is no quantitative evaluation system for the fish-attracting effect of bionic robotic fish. CN113421345B only implements the simulation and control strategy training of robotic fish swarm navigation, and CN119105386B only optimizes the stability of the robotic fish's own swimming. Neither of them has constructed a method for evaluating the fish swarm behavior in the scenario of bionic robotic fish inducing real fish, and therefore cannot quantitatively judge the fish-attracting effectiveness of robotic fish.

[0008] It is impossible to measure the collaborative behavior between the fish school and the bionic robotic fish. Existing technologies all focus on the robotic fish itself, without considering the internal behavioral characteristics of real fish schools, nor paying attention to the dependence and following relationship between real fish and robotic fish. It is impossible to distinguish different fish-attracting states such as "the robotic fish moves well but the fish school does not follow" and "the robotic fish leads the fish school to gather closely and swim synchronously".

[0009] The lack of quantitative data for engineering optimization of the control / navigation strategy of the robotic fish makes it impossible to provide a basis for such optimization. The navigation strategy of CN113421345B and the control method of CN119105386B both lack scientific evaluation indicators to verify their actual effects in fish-attracting scenarios. They cannot provide objective data support for parameter tuning and strategy iteration of the two technologies, nor can they be aligned with the actual engineering criteria of "excellent / good / medium / poor fish-attracting effect".

[0010] The technology has not been effectively integrated with engineering applications for fish attraction. Existing technological achievements only focus on the simulation or control of the robotic fish itself, without considering the actual fish attraction needs of fish passage facilities in water conservancy projects. Therefore, it cannot provide a basis for decision-making in selecting the optimal biomimetic robotic fish control strategy and navigation mode for engineering projects.

[0011] Therefore, developing a quantitative evaluation system for fish-attracting effects has become a pressing technical problem to be solved in this field. Summary of the Invention

[0012] To address the above problems, this invention proposes a method for evaluating the cohesion and coordination of fish schools based on a biomimetic robotic fish. The method includes the following steps: S1. Video acquisition begins, capturing experimental video signals including the bionic robotic fish and real fish; S2. The target detection and tracking module processes the video signal to obtain the spatial coordinates and timestamps of the bionic robotic fish and each real fish in each frame, forming trajectory data; S3. The behavior parameter calculation module calculates the basic behavior parameters for each frame based on the trajectory data, including the distance between real fish and real fish nearest neighbors, the distance between real fish and bionic robotic fish, the polarity of the real fish group arrangement, the speed synchronization of real fish, and the polarity of the real fish and bionic robotic fish arrangement. S4. The time window division module divides the entire experimental time axis into several time windows according to the set time length ΔT, and averages or statistically analyzes the basic behavioral parameters obtained in step S3 within each time window. S5. Calculate the cohesion index for each time window using a cohesion and coordination calculation algorithm. and coordination indicators And calculate the overall cohesion index. and coordination indicators ; The specific algorithm for calculating cohesion and coordination is as follows: S5.1 Calculates the basic statistics for each time window; The calculation of the basic statistics within the m-th time window includes: S5.1.1 Calculate the average of the nearest neighbor distances between real fish and real fish. and standard deviation ; S5.1.2 Calculate the average nearest neighbor distance of the real fish – bionic robotic fish ; S5.1.3 Calculate the polarity d36 of the true fish population arrangement, with a value range of 0 to 1; S5.1.4 Calculating the speed synchronization of a real fish group Its numerical range is 0 to 1; S5.1.5 Calculating the Polarity of Real Fish – Bionic Robotic Fish Arrangement Its numerical range is 0 to 1.

[0013] S5.2 Based on fish body length L Determine the distance behavior threshold and normalize the distance class parameters; S5.1.6 in Within a time window, the average distance between real fish and their nearest neighbors is obtained. The average distance between the real fish and the bionic robotic fish was calculated to obtain... .

[0014] S5.2.1 Set the closeness threshold for the internal nearest neighbor distance. and loose threshold , where k1 and k2 are dimensionless coefficients, satisfying 0.5≤k1≤3, 4≤k2≤10, and k2>k1; S5.2.2 Setting the threshold for close proximity between the real fish and the bionic robotic fish and loose threshold ,in k3 and k4 are dimensionless coefficients that satisfy 0.5≤k3≤3, 4≤k4≤10, and k4>k3; S5.2.3 By performing interval normalization, the baseline value of cohesion within the fish swarm is obtained. : ; in, For the first Mean nearest neighbor distance of a group of real fish within a time window; For tightness threshold; Loose threshold; This represents the average body length of a true fish. , The coefficient is dimensionless. This represents the baseline value for cohesion within the fish population.

[0015] S5.2.4 By performing interval normalization, the basic values ​​of the spatial dependence of real fish on the bionic robotic fish are obtained. : ; in, For the first The average distance between real fish and bionic robotic fish within a time window; For tight thresholds; Loose threshold; This represents the average body length of a true fish. , The coefficient is dimensionless. This represents the basic value of the spatial dependence of a real fish on a bionic robotic fish.

[0016] S5.3 calculates the cohesion component within the fish swarm, the spatial attachment component between the fish and the robotic fish, and the spatial uniformity component to form cohesion-related components. The method for calculating the cohesion-related components is as follows: S5.3.1 According to Define the cohesion component within the fish school , = ; S5.3.2 According to Define the spatial dependence components of a real fish on a biomimetic robotic fish. , = ; S5.3.3 Set a reasonable range for the standard deviation of the nearest neighbor distance. ,in The fish are in a relatively uniform and compact state. For fish groups in a relatively discrete state, Normalization is performed to obtain the spatial homogeneity component. : ; in, Indicates the first The standard deviation of the nearest neighbor distance between real fish within a time window; This represents the lower limit of the standard deviation when the fish population is in a relatively uniform and compact state. This represents the upper limit of the standard deviation when the fish population is in a relatively discrete state, and ; Indicates the first The spatial uniformity component within each time window ranges from 0 to 1.

[0017] when This indicates that the spacing between individual fish is relatively small, and the spatial distribution of the fish school is relatively uniform. ; when This indicates that the spacing between individual fish varies greatly, and the spatial distribution of the fish school is relatively uneven. ; when At that time, the spatial homogeneity component was calculated using linear interpolation. C uni (m) .

[0018] S5.4 performs nonlinear normalization on parameters related to directional consistency and velocity consistency to form coordination-related components; The specific method for nonlinear normalization is as follows: S5.4.1 Polarity of True Fish Group Arrangement Power function mapping is performed to obtain the directional coordination components. : ; in >1, used to enhance sensitivity under high directional consistency conditions; S5.4.2 Synchronization of Speed ​​in a Group of True Fish By performing power function mapping, the velocity coordination components are obtained. : ; in >1, used to enhance sensitivity under high-speed consistency conditions: S5.4.3 Polarity of the arrangement of real fish and bionic robotic fish Perform power function mapping to obtain the external follower component. : ; in >1, used to enhance the influence of real fish on the orientation-following behavior of biomimetic robotic fish in high consistency regions.

[0019] S5.5 Determine cohesion indicators based on preset or adaptive strategies. The weights are then calculated and summed. The The specific method for determining the weights and performing a weighted sum is as follows: S5.5.1 Set the weights of the internal cohesion component, spatial dependence component, and spatial homogeneity component to w1, w2, and w3, respectively, and satisfy the following conditions: ; S5.5.2 In fixed weight mode, It is a constant; In S5.5.3, under adaptive weight mode, it can be based on... The average distance between the real fish and the bionic robotic fish is adjusted by w2, for example, when... Increase when the distance is less than the preset distance threshold D1 ,when When the distance exceeds the preset threshold D2, decrease w2 and adjust accordingly. .

[0020] S5.5.4 Cohesion index for the m-th time window for: ; in, C m Indicates the first m A time window's cohesion index; C int (m) Indicates the first m The cohesion component within the fish population within a time window; C att (m) Indicates the first m The spatial dependence of a real fish on a bionic robotic fish within a time window. C uni (m) Indicates the first m Spatial uniformity component within a time window; , , These represent the weighting coefficients corresponding to the cohesion component, spatial dependence component, and spatial evenness component within the fish school, respectively, satisfying... ,and , , .

[0021] The cohesion index C mIt is used to comprehensively characterize the degree of aggregation within a real fish group, the degree of attachment of real fish to the bionic robotic fish, and the uniformity of the spatial distribution of the fish group. The larger the value, the higher the overall cohesion of the fish group.

[0022] S5.6 Determine coordination indicators based on preset or adaptive strategies. Q m The weights are then calculated and summed. The Q m The specific method for determining the weights and performing a weighted sum is as follows: S5.6.1 Set the weights of the direction coordination component, velocity coordination component, and external following component as follows: And satisfy ; In S5.6.2, under the fixed weight mode, v1, v2, and v3 are constants, for example... ; In S5.6.3, under adaptive weight mode, it can be based on... Size adjustment v3, when Increase v3 when it exceeds the preset threshold. Decrease v3 when the value is below a preset threshold; S5.6.4 The coordination index of the m-th time window for: ; in, Q m This represents the coordination index for the m-th time window; Q dir (m) This represents the directional coordination component within the m-th time window; Q spd (m) This represents the velocity coordination component within the m-th time window; Q follow (m) This represents the external following component within the m-th time window; , , These represent the weighting coefficients corresponding to the direction coordination component, velocity coordination component, and external following component, respectively, satisfying... ,and , , The aforementioned coordination indicators Q mIt is used to comprehensively characterize the consistency of a group of real fish in terms of swimming direction, speed changes, and response to the bionic robotic fish. The larger the value, the stronger the coordination of the fish group.

[0023] S5.7 for all time windows C m , Q m By aggregating time, an overall cohesion index is obtained. and coordination indicators And determine the level based on the threshold.

[0024] The calculation of the overall cohesion index and coordination indicators The method is as follows: S5.7.1 Suppose the experiment is divided into M time windows, and the time weight of the m-th time window is... ,satisfy: ; in, This represents the total number of time windows obtained from the experimental division; Indicates the time window number, and ; Indicates the first The time weights corresponding to each time window satisfy the following conditions: Furthermore, the sum of the time weights of all time windows is 1.

[0025] When the lengths of each time window are the same, we can take... When different time windows are of different importance, different weights can be assigned to each time window according to the evaluation needs.

[0026] S5.7.2 The overall cohesion index C is: ; in, C Indicators representing overall cohesion; C m Indicates the first A time window's cohesion index; a m Indicates the first The time weight corresponding to each time window; M This indicates the total number of time windows.

[0027] The overall cohesion index C Cohesion indicators for each time window C m According to the corresponding time weight a m The weighted sum is used to comprehensively characterize the overall cohesion of the fish swarm throughout the entire experiment. The larger the value, the stronger the overall cohesion of the fish swarm.

[0028] The overall coordination index Q in S5.7.3 is: ; in, Q Indicates overall coordination indicators; Q m Indicates the first Coordination indicators for each time window; a m Indicates the first The time weight corresponding to each time window; M This indicates the total number of time windows.

[0029] The overall coordination index Q Coordination indicators for each time window Q m According to the corresponding time weight a m The weighted sum is used to comprehensively characterize the consistency of the fish group in terms of swimming direction, speed changes, and response to the bionic robotic fish throughout the entire experiment. The larger the value, the stronger the overall coordination of the fish group.

[0030] S6. Will C , Q The results of the evaluation are output and stored. If it is necessary to continue evaluating new experiments, return to step S1; otherwise, the process ends.

[0031] The grading results are divided into three levels: excellent, good, and poor.

[0032] Based on the aforementioned method for evaluating fish school cohesion and coordination using biomimetic robotic fish, a system for evaluating fish school cohesion and coordination based on biomimetic robotic fish is provided, specifically including: Video acquisition module. Acquires experimental video signals including those from the bionic robotic fish and real fish; Target detection and tracking module. This module performs target detection and multi-target tracking on the video signals acquired by the video acquisition module, obtaining the spatial trajectories of the bionic robotic fish and individual real fish over time. Behavioral parameter calculation module. Based on the trajectory data output by the target detection and tracking module, it calculates basic behavioral parameters such as the nearest neighbor distance between real fish, the distance between real fish and bionic robotic fish, the polarity of the real fish group arrangement, the speed synchronization of real fish, and the polarity of the real fish and bionic robotic fish arrangement. Time window segmentation module. The entire experimental timeline is divided into several time windows, and the basic behavioral parameters output by the behavioral parameter calculation module are statistically summarized within each time window; Cohesion and Coordination Calculation Module. Based on the basic behavioral parameters obtained from the time window segmentation module, the fish swarm cohesion index for each time window is calculated using a preset or adaptive normalization and weighting algorithm. and coordination indicators And further obtain the overall cohesion index. and coordination indicators ; Results storage module. For each experiment , And its time-varying sequence is stored and displayed.

[0033] The video acquisition module can use a top-mounted camera to acquire experimental video images from a top-down perspective.

[0034] Compared with the prior art, the beneficial effects of the present invention include: (1) Filling the technological gap in the quantitative evaluation of the fish-attracting effect of biomimetic robotic fish. For the first time, a unified evaluation framework based on the dual indicators of "cohesion and coordination" was constructed, solving the problem that existing technologies cannot quantitatively evaluate the fish-attracting effectiveness of biomimetic robotic fish, and providing a standardized method for verifying the effectiveness of related control and navigation technologies of biomimetic robotic fish.

[0035] (2) Improve the comprehensiveness and accuracy of the evaluation of fish attraction effect. Simultaneously depicting the cohesion state within the fish school and the attachment and following behavior of real fish to the robotic fish, it can accurately distinguish different fish attraction states, avoiding the one-sidedness of only focusing on the movement performance of the robotic fish itself while ignoring the actual fish attraction effect, and the evaluation results are more in line with the actual needs of the project.

[0036] (3) Achieve direct alignment between evaluation methods and engineering fish-attracting criteria. By classifying quantitative indicators into fish-attracting effect levels that can be directly used in engineering, a decision-making basis is provided for the selection and optimization of biomimetic robotic fish control strategies and navigation modes in water conservancy engineering fish passage facilities, promoting the transition of biomimetic robotic fish technology from laboratory research to practical engineering applications.

[0037] (4) Improve the objectivity and repeatability of the evaluation of fish attraction effect. The automated index calculation process replaces manual frame-by-frame statistics, effectively reducing human subjective error. The evaluation results are repeatable and comparable, and can be used as a unified standard for horizontal comparison between different bionic robotic fish technologies and different fish attraction strategies. Attached Figure Description

[0038] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0039] Figure 1This is a schematic diagram of a fish swarm cohesion and coordination evaluation system based on a biomimetic robotic fish.

[0040] Figure 2 This is a schematic diagram of the experimental scene and coordinate system of the biomimetic robotic fish – a real fish.

[0041] Figure 3 This is a flowchart of a method for evaluating the cohesion and coordination of a fish school based on a biomimetic robotic fish.

[0042] Figure 4 This is a schematic diagram showing the relationship between the distance and speed of a real fish and its nearest neighbor, as well as a bionic robotic fish.

[0043] Figure 5 It is a cohesion indicator Calculation flowchart.

[0044] Figure 6 It is a coordination indicator Calculation flowchart.

[0045] Figure 7 It is a cohesion indicator Coordination Indicators A schematic graph illustrating how time changes.

[0046] In the diagram: 1 is the video acquisition module; 2 is the target detection and tracking module; 3 is the behavior parameter calculation module; 4 is the time window division module; 5 is the cohesion and coordination calculation module; 6 is the result storage module; 7 is the camera; 8 is the bionic robotic fish; 9 is the real fish; 10 is the experimental tank; 11 is the computer; 12 is the velocity vector of the bionic robotic fish; 13 is the velocity vector of the target real fish; 14 is the target real fish. Detailed Implementation

[0047] The present invention discloses a method and system for evaluating the cohesion and coordination of fish schools based on biomimetic robotic fish, which is implemented as follows: Example 1 like Figure 1 As shown, the present invention provides a fish school cohesion and coordination evaluation system, including a video acquisition module 1, a target detection and tracking module 2, a behavior parameter calculation module 3, a time window division module 4, a cohesion and coordination calculation module 5, and a result storage module 6.

[0048] The video acquisition module 1 is used to acquire experimental video signals containing a bionic robotic fish 8 and a real fish 9 through a camera 7; the target detection and tracking module 2 is used to perform target detection and multi-target tracking on the bionic robotic fish 8 and the real fish 9 in the experimental video to obtain their trajectory data; the behavior parameter calculation module 3 is used to calculate basic behavior parameters based on the trajectory data; the time window division module 4 is used to divide the experimental process into time windows according to a preset time length and perform statistics; the cohesion and coordination calculation module 5 is used to calculate the cohesion index and coordination index of each time window, and further obtain the overall cohesion index and overall coordination index; and the result storage module 6 is used to output and store the evaluation results.

[0049] like Figure 2 As shown, camera 7 is positioned above experimental tank 10 to capture video footage of the movement of the biomimetic robotic fish 8 and real fish 9 within the tank. Computer 11 is connected to camera 7 to receive the experimental video and perform processing such as target detection, trajectory tracking, behavioral parameter calculation, and evaluation index output. In this embodiment, the biomimetic robotic fish 8 and multiple real fish 9 are placed together in experimental tank 10, and their movement trajectories in a two-dimensional plane coordinate system are acquired by camera 7, serving as the data basis for subsequent evaluation of cohesion and coordination.

[0050] Example 2 Based on the system described in Embodiment 1, such as Figure 3 As shown, this embodiment provides a method for evaluating the cohesion and coordination of a fish school based on a biomimetic robotic fish, including the following steps: Step S1: Acquire experimental video signals The video acquisition module 2 acquires experimental video signals containing the bionic robotic fish 8 and the real fish 9. Preferably, a camera 7 positioned above the experimental tank 10 is used to capture overhead footage, obtaining a video sequence of the motion of the bionic robotic fish 8 and the real fish 9 during the experiment.

[0051] Step S2: Target Detection and Tracking The target detection and tracking module 2 performs frame-by-frame target detection and multi-target tracking on the video signal acquired in step S1, distinguishes between the bionic robotic fish 8 and each real fish 9, and obtains their spatial coordinates and timestamps at each moment to form the spatiotemporal trajectory data of the bionic robotic fish 8 and each real fish 9.

[0052] Step S3: Calculate basic behavioral parameters The behavior parameter calculation module 3 calculates basic behavior parameters based on the trajectory data obtained in step S2. The basic behavior parameters include: (1) The nearest neighbor distance between real fish; (2) Distance between real fish and bionic robotic fish; (3) Polarity of the arrangement of true fish groups; (4) Synchronization of speed in a group of real fish; (5) Polarity of arrangement of real fish and bionic robotic fish.

[0053] like Figure 4 As shown, taking target fish 14 as an example, the distance between target fish 14 and its nearest neighbor fish 9 is denoted as... The distance between the target real fish 14 and the bionic robotic fish 8 is denoted as . The velocity vector of the biomimetic robotic fish 8 is shown in direction 12, denoted as V. m The velocity vector of the target fish 14 is in the direction shown in Figure 13, denoted as V. i The angle between the two is denoted as Based on the aforementioned distance and direction parameters, we can further calculate the average and standard deviation of the nearest neighbor distance between real fish, the average distance between real fish and bionic robotic fish, the polarity of the real fish group arrangement, the synchronization of the real fish group speed, and the polarity of the real fish and bionic robotic fish arrangement, among other basic behavioral parameters.

[0054] Step S4: Time window division and statistics Time window division module 5 divides the entire experimental timeline according to the set time length. The time window is divided into several time windows, and the basic behavioral parameters obtained in step S3 are averaged or statistically analyzed within each time window to obtain the basic statistics for each time window.

[0055] In this embodiment, let If the total experiment duration is Then it can be divided into There are several time windows. Within each time window, the average and standard deviation of the nearest neighbor distance between real fish, the average distance between real fish and bionic robotic fish, the polarity of the real fish group arrangement, the speed synchronization of the real fish group, and the polarity of the real fish and bionic robotic fish arrangement are statistically analyzed.

[0056] Step S5: Calculate the cohesion and coordination indices for each time window. The cohesion and coordination calculation module 5 calculates the cohesion index for each time window based on the basic statistics within each time window, using preset normalization, nonlinear mapping, and weighted synthesis methods. C m and coordination indicators Q m .

[0057] Among them, cohesion index C m This indicator is used to comprehensively characterize the degree of aggregation within the real fish population, the spatial dependence of real fish on the bionic robotic fish, and the uniformity of the fish population's spatial distribution within the m-th time window; coordination index Q mThis is used to comprehensively characterize the consistency of the fish swarm in terms of swimming direction, speed changes, and response to the bionic robotic fish within the m-th time window.

[0058] The cohesion index of each time window C m For the specific construction method, please refer to Example 3, where the coordination index of each time window is described. Q m For the specific construction method, please refer to Example 4.

[0059] Step S6: Time-weighted aggregation and result output Cohesion index for all time windows C m and coordination indicators Q m By performing time-weighted aggregation, the overall cohesion index is obtained. C and overall coordination indicators Q The system determines the level based on a preset threshold and outputs and stores the overall evaluation results and their time-varying sequence in the result storage module 7.

[0060] The experiment is divided into M time windows, and the time weight of the m-th time window is... a m ,satisfy a m ≥0 and the sum of all time weights is 1. When the lengths of each time window are the same, equal weights are applied. The overall cohesion index... C and overall coordination indicators Q For a detailed explanation of the time aggregation method, please refer to Example 5.

[0061] Through the above steps, a quantitative evaluation of the cohesion and coordination of fish groups induced by biomimetic robotic fish can be achieved, providing an evaluation basis for optimizing the control strategy, navigation mode, and fish-attracting effect of biomimetic robotic fish.

[0062] Example 3 Based on Example 2, this example calculates the cohesion index for each time window.

[0063] like Figure 5 As shown, within the m-th time window, cohesion-related components are constructed based on the nearest neighbor distance between real fish, the distance between real fish and the bionic robotic fish, and the standard deviation of the nearest neighbor distance between real fish and real fish. These components are then further weighted and synthesized to obtain the cohesion index for that time window. C m .

[0064] First, based on the fish's body length Set a threshold for distance behavior and normalize the distance class parameters. Set a tight threshold for the nearest neighbor distance between real fish. and loose threshold ,in, , The coefficient is dimensionless and satisfies , ,and Set a threshold for the closeness of the distance between the real fish and the bionic robotic fish. and loose threshold ,in, , The coefficient is dimensionless and satisfies , ,and .

[0065] In the Within a time window, the average distance between real fish and their nearest neighbors. By performing interval normalization, the baseline value of cohesion within the fish swarm is obtained. Average distance between real fish and bionic robotic fish By performing interval normalization, the basic values ​​of the spatial dependence of real fish on the bionic robotic fish are obtained. Furthermore, the cohesion component C within the fish school is defined based on the baseline value of cohesion within the fish school. int (m), that is The spatial dependence component C of a real fish on a biomimetic robotic fish is defined based on the basic value of spatial dependence. att (m), that is .

[0066] At the same time, a reasonable range for the standard deviation of the distance between real fish and their nearest neighbors should be set. and for the first Standard deviation of the nearest neighbor distance between real fish within a time window Normalization yields the spatial homogeneity component C. uni (m). Among them, This represents the lower limit of the standard deviation when the fish population is in a relatively uniform and compact state. This represents the upper limit of the standard deviation when the fish population is in a relatively discrete state, and C uni The value of (m) ranges from 0 to 1. The larger the value, the smaller the fluctuation in the distance between individual fish and the more uniform the spatial distribution of the fish group.

[0067] After obtaining the three cohesion-related components mentioned above, they are weighted and synthesized. The weights of the fish swarm internal cohesion component, spatial dependence component, and spatial homogeneity component are set as follows: , , ,satisfy ,and , , In the fixed-weight mode, , , This is a preset constant; in adaptive weight mode, it can be based on the average distance between the real fish and the bionic robotic fish. Adjustment For example, when Less than the preset distance threshold Time increases ,when Greater than the preset distance threshold Time decrease and adjust accordingly. and .

[0068] Then the first Cohesion index of a time window Represented as: ; in, Used for comprehensive characterization of the first Within a given time window, the degree of aggregation within the real fish population, the degree of spatial dependence of the real fish on the bionic robotic fish, and the uniformity of the spatial distribution of the fish population are all considered. The larger the value, the higher the overall cohesion of the fish population.

[0069] Example 4 Based on Example 2, this example calculates the coordination index for each time window.

[0070] like Figure 6 As shown, this embodiment uses the polarity of a real fish group arrangement. Synchronization of speed in a group of real fish And real fish - bionic robotic fish arrangement polarity Using the input as input, the direction coordination component, velocity coordination component, and external following component are obtained through power function nonlinear mapping, and then further weighted and synthesized to obtain the coordination index for the m-th time window. Q m .

[0071] First, a power function mapping is performed on the polarity d36 of the true fish population arrangement to obtain the directional coordination components. Where d36 ranges from 0 to 1, and the power exponent of the directional coordination component mapping is... ,and Power function mapping can enhance the representation effect in regions of high directional consistency, making the coordinated behavior of real fish groups in a state of high directional consistency more prominent.

[0072] Secondly, the speed synchronization of a group of real fish By performing power function mapping, the velocity coordination component Q is obtained.spd (m); where, The value range of is 0 to 1, and the power exponent of the velocity coordination component mapping is . ,and Power function mapping can enhance the representation effect in high-speed synchronization regions, making the coordinated behavior of fish groups more sensitive when the consistency of speed changes is high.

[0073] Secondly, regarding the polarity of the arrangement of real fish and bionic robotic fish. Perform a power function mapping to obtain the external follower component Q. follow (m); where, The value range is 0 to 1, and the power exponent of the external follower component mapping is ,and By using power function mapping, the influence of real fish on the orientation-following behavior of the biomimetic robotic fish can be enhanced in the high consistency region, making the representation of external following behavior more sensitive in the coordination evaluation.

[0074] After obtaining the three coordination-related components mentioned above, they are weighted and synthesized. The weights of the direction coordination component, velocity coordination component, and external following component are set as follows: , , ,satisfy ,and , , In the fixed-weight mode, , , This is a preset constant; in adaptive weight mode, it can be adjusted based on the external follower component Q. follow (m) size adjustment When Q follow (m) increases when it exceeds the preset threshold. When Q follow (m) decreases when it is below the preset threshold. and adjust accordingly. and .

[0075] Then the coordination index of the m-th time window Q m Represented as: ; in, Q m This value is used to comprehensively characterize the consistency of the real fish group in terms of swimming direction, speed changes, and response to the bionic robotic fish within the m-th time window. The larger the value, the stronger the overall coordination of the fish group.

[0076] Example 5 Based on Examples 2 to 4, this example performs time-weighted aggregation of cohesion and coordination indices for all time windows to obtain the overall cohesion index. C and overall coordination indicators Q The time-weighted aggregation and engineering level determination are consistent with the description of step S6 in the claims.

[0077] Suppose the experiment is divided into M time windows, and the time weight of the m-th time window is... a m ,satisfy: ; Where M represents the total number of time windows obtained from the experiment, and m represents the time window number. a m This represents the time weight corresponding to the m-th time window. When all time windows have the same length, we can take... .

[0078] The overall cohesion index C is represented as: ; Among them, C m This represents the cohesion index for the m-th time window. The overall cohesion index C is obtained by weighting and summing the cohesion indices of each time window according to their corresponding time weights, and is used to comprehensively characterize the overall cohesion of the fish population throughout the entire experiment.

[0079] The overall coordination index Q is represented as: ; Among them, Q m This represents the coordination index for the m-th time window. The overall coordination index q is obtained by weighting and summing the coordination indices of each time window according to their corresponding time weights, and is used to comprehensively characterize the consistency of the fish population in terms of swimming direction, speed changes, and response to the bionic robotic fish throughout the entire experiment.

[0080] In this embodiment, the calculation results of five time windows obtained under a certain control strategy are statistically analyzed to obtain the following results: Figure 7 The curves showing the changes in the indicators are illustrated. The results indicate that both the cohesion and coordination indices increase overall with the increase of the time window number m. After the third time window, both exceed 0.6, with the final overall cohesion index reaching approximately 0.78 and the overall coordination index reaching approximately 0.70. This demonstrates that the control strategy can gradually enhance the cohesion and cooperative behavior of the fish swarm towards the biomimetic robotic fish.

[0081] According to the needs of engineering applications, the threshold for judging the fish-attracting effect can be set based on the overall cohesion index C and the overall coordination index Q. For example, when both C and Q are greater than or equal to 0.8, the fish-attracting effect can be judged as "excellent"; when both C and Q are between 0.6 and 0.8, the fish-attracting effect can be judged as "good"; when either index is lower than 0.6, the fish-attracting effect can be judged as "poor". In this embodiment, the overall cohesion index is about 0.78 and the overall coordination index is about 0.70, so the fish-attracting effect of the bionic robotic fish under this control strategy can be judged as "good".

Claims

1. A method for evaluating the cohesion and coordination of fish schools based on biomimetic robotic fish, characterized in that, Includes the following steps: S1 collects experimental videos containing both biomimetic robotic fish and real fish; S2 processes the video to obtain spatiotemporal trajectory data of the bionic robotic fish and various real fish; S3 calculates the basic behavioral parameters; S4 divides time windows according to the set duration and calculates the basic parameters in each window; S5 normalizes distance-related parameters by setting a distance behavior threshold based on fish body length, performs power function nonlinear mapping on direction and velocity-related parameters, and calculates cohesion and coordination indicators for each time window by combining fixed / adaptive weight strategies. S6 performs time-weighted aggregation of indicators from each time window to obtain the overall indicator and completes the engineering level determination; at the same time, it depicts the cohesion within the fish swarm and the fish's dependence on and following of the robotic fish, realizing a quantitative evaluation of the fish-attracting effect.

2. The method for evaluating fish school cohesion and coordination based on biomimetic robotic fish according to claim 1, characterized in that, The basic behavioral parameters of step S3 specifically include five categories: real fish-real fish nearest neighbor distance, real fish-bionic robot fish distance, real fish group arrangement polarity, real fish speed synchronization, and real fish-bionic robot fish arrangement polarity, taking into account both the behavior within the fish group and the external interaction behavior between the fish and the robot fish.

3. The method for evaluating fish school cohesion and coordination based on biomimetic robotic fish according to claim 1, characterized in that, The distance behavior thresholds set in step S5 include the nearest neighbor distance within the real fish, the close threshold and the loose threshold for the fish-machine fish distance, and the loose threshold coefficient is greater than the close threshold coefficient, thereby achieving fish body feature adaptation normalization of distance parameters.

4. The method for evaluating fish school cohesion and coordination based on biomimetic robotic fish according to claim 1, characterized in that, The cohesion index in step S5 is a weighted composite of the cohesion component within the fish swarm, the spatial dependence component of the real fish on the bionic robotic fish, and the spatial uniformity component. The spatial uniformity component is obtained by normalizing the standard deviation of the nearest neighbor distance of the real fish within a reasonable range.

5. The method for evaluating fish school cohesion and coordination based on biomimetic robotic fish according to claim 1, characterized in that, The adaptive weighting strategy in step S5 is as follows: in the cohesion index, the spatial dependence component weight is adjusted according to the average distance between the real fish and the bionic robotic fish; in the coordination index, the external following component weight is adjusted according to the arrangement polarity of the real fish and the robotic fish, so as to realize the dynamic adaptation of weights to the behavior state of the fish group.

6. The method for evaluating fish school cohesion and coordination based on biomimetic robotic fish according to claim 1, characterized in that, In step S5, the power coefficients of the power function nonlinear mapping are all greater than 1, which respectively map the polarity of the real fish group arrangement, speed synchronization, and fish-robot fish arrangement polarity to obtain directional coordination, speed coordination, and external following components, thereby enhancing the evaluation sensitivity under high consistency behavior state.

7. The method for evaluating the cohesion and coordination of a fish school based on a biomimetic robotic fish according to claim 1, characterized in that, The time-weighted aggregation and grade determination in step S6 are as follows: Assume the experiment is divided into M time windows, and the time weight of the m-th time window is... ,satisfy ; Overall cohesion index , Indicates the first Cohesion indicators for each time window; overall coordination indicators , Indicates the first Coordination indicators for each time window; The engineering level is determined by setting thresholds based on the values ​​of C and Q.

8. The method for evaluating fish school cohesion and coordination based on biomimetic robotic fish according to claim 1, characterized in that, The engineering level determination in step S6 is to set a threshold based on the quantitative indicators of overall cohesion and coordination, and divide the fish-attracting effect into three levels: excellent, good, and poor.

9. A system for evaluating the cohesion and coordination of a fish school based on a biomimetic robotic fish, implementing the method of any one of claims 1-8, characterized in that, It includes six functional modules connected in sequence: video acquisition module, target detection and tracking module, behavior parameter calculation module, time window division module, cohesion and coordination calculation module, and result storage module. These modules work together to automate the entire process from video acquisition to evaluation result storage.

10. The fish school cohesion and coordination evaluation system based on biomimetic robotic fish according to claim 9, characterized in that, The video acquisition module uses a top-down camera to capture experimental videos and establishes a two-dimensional rectangular coordinate system on the horizontal plane of the experimental water body. The camera is connected to the back-end computer to realize real-time / offline processing of trajectory data, ensuring the accuracy of spatial coordinate acquisition between the bionic robotic fish and real fish.

Images