A fish school aggregation method, device and medium based on bionic swimming signals

Abstract

The application discloses a fish school gathering method and device based on bionic swimming signals, and a medium, relates to the technical field of fishery gathering control, and comprises the following steps: carrying out fish school situation identification on a target water area, demarcating a gathering operation area along a target fish school distribution range, and forming a fish school situation table; performing contour smoothing on the outer edge of the target fish school recorded in the fish school situation table, arranging a test point along the smoothed outer edge of the target fish school, and outputting bionic test sequences in the test point by bionic robotic fish in sequence; comparing the fish school gathering change and the fish school orientation change before and after the test in the same test response window; registering an acceptable gear when the fish school gathering is enhanced and the fish school orientation is the same; registering an avoidance gear when the fish school outer edge is scattered and the fish school orientation is reversed; and forming a fish school acceptance fingerprint. The application enhances the adaptability of the front edge access, and improves the stability of the transition of the fish school from a natural swimming state to a controlled gathering state.

Description

Technical Field

[0001] This invention relates to the field of fishery agglomeration control technology, and in particular to a method, device and medium for fish agglomeration based on biomimetic swimming signals. Background Technology

[0002] As aquaculture, nearshore stock enhancement and release, and water fishing management develop towards less disturbance and more refined methods, the use of underwater sensing equipment to obtain the outline, density, and direction of movement of fish schools, and then the output of biomimetic robotic fish that mimic the tail wagging, cruising, and swimming movements of fish, to induce, drive, and gather target fish schools, has become an important technical route to replace high-intensity sound and light stimulation and artificial herding methods. Related solutions usually plan the trajectory of the robotic fish based on the current position of the fish school and the target gathering position, and adjust the following state of the fish school by tail wagging frequency, swimming speed, and leading direction during the movement, in order to achieve the slow convergence of the fish school towards the target water area.

[0003] In the aforementioned technologies, bionic robotic fish often approach the outer edge of a school of fish directly according to a predetermined gear and path. They lack the process of probing and identifying the target school's outer edge response and selecting the appropriate gear before entering the school. Because different fish species, densities, levels of fright, and water flow conditions can change the school's acceptance of tail wagging rhythms and approach positions, in scenarios where the leading edge of the school is loose, the orientation is inconsistent, and the outer edge is partially dispersed, if the entry gear is too strong or the entry point falls on a sensitive outer edge section, the school of fish may easily deviate in the opposite direction, disperse locally, or lose its traction direction, thereby affecting the continuity and controllability of subsequent gathering operations. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method for fish agglomeration based on biomimetic swimming signals to solve the problems of mismatched access points at the leading edge of biomimetic robotic fish, easy induction of fish dispersion, and impact on agglomeration stability.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for fish school aggregation based on biomimetic swimming signals, comprising: identifying fish school status in a target water area; delineating an aggregation operation area along the distribution range of the target fish school to form a fish school status table; smoothing the outline of the outer edge of the target fish school recorded in the fish school status table; setting up test points along the smoothed outer edge of the target fish school; a biomimetic robotic fish sequentially outputting a biomimetic test sequence at the test points; comparing the changes in fish school convergence and fish school orientation before and after the test within the same test response window; registering an acceptable level when fish school convergence is enhanced and fish school orientation is converging; registering an avoidance level when the outer edge of the fish school disperses and the fish school orientation is reversed, thus forming a fish school acceptance fingerprint; screening out acceptable levels from the fish school acceptance fingerprint and determining the leading edge access level, and setting the avoidance level... The associated test points are eliminated to form a candidate range for the leading edge. The retained test points are sorted by leading edge priority along the direction of the aggregation operation domain. When the retained test points continuously show the same leading edge direction, edge convergence and gentle turning, the virtual leading edge fish position is marked and bound to the leading edge access position to form a leading edge access task. A companion trajectory is generated around the virtual leading edge fish position in the leading edge access task. The bionic robotic fish enters the companion trajectory according to the leading edge access position. When the leading edge of the target fish group moves in the same direction, the leading angle is slightly adjusted towards the aggregation operation domain to form a leading edge traction table. When the leading edge traction table confirms that the target fish group has entered the aggregation operation domain, the bionic robotic fish swims slowly around the outer edge of the aggregation operation domain. When the target fish group disperses at the outer edge, a low-intensity pull is performed to form a stable aggregation state.

[0007] As a preferred embodiment of the fish school aggregation method based on biomimetic swimming signals described in this invention, the fish school situation table is formed by establishing a working coordinate reference for the target water area, collecting water area perception information and identifying the area where the target fish school is located, determining the outer contour and movement direction of the target fish school, delineating the aggregation operation area along the outer contour of the target fish school towards the aggregation target position, avoiding unsuitable aggregation positions, and organizing the target fish school situation and aggregation operation area accordingly.

[0008] As a preferred embodiment of the fish school aggregation method based on biomimetic swimming signals described in this invention, the outer edge of the target fish school recorded in the fish school situation table is smoothed, and test points are set up along the smoothed outer edge of the target fish school. The biomimetic robotic fish outputs a biomimetic test sequence sequentially at the test points. The specific steps are as follows: Extract the outer contour of the target fish school from the fish school situation table, resample the outer contour of the target fish school with equal arc length, and use Savitzky-Golay smoothing filter to reduce the outer edge spikes and generate a smooth outer contour. Along the smooth outer edge contour towards the gathering operation area, test positions are selected sequentially, and the test positions are kept outside the outer edge of the target fish group to generate a test point sequence. The bionic robotic fish arrives at the current test point according to the test point sequence, maintains a low-disturbance swimming state at the current test point, and uses the duration of the low-disturbance swimming state as the test reference window to obtain the natural outer edge changes of the target fish group within the test reference window. At the current test point, biomimetic swimming is performed sequentially according to the tail swing rhythm gears from low to high to form a biomimetic test sequence, and the duration from the start of output to the stop of output at the current gear is determined as the current test response window.

[0009] As a preferred embodiment of the fish school aggregation method based on biomimetic swimming signals described in this invention, the process of comparing changes in fish school convergence and orientation before and after the trial within the same trial response window, registering an acceptable level when fish school convergence is enhanced and fish school orientation is converging, and registering an avoidance level when fish school outer edges disperse and fish school orientation is reversed, thus forming a fish school acceptance fingerprint, is as follows: Within the current trial response window, the natural outer edge change of the target fish group in the trial reference window is used as the reference outer edge zone. The displacement direction of the target fish group's outer edge relative to the reference outer edge zone is extracted. When the target fish group's outer edge continuously enters towards the center of the group, a net convergence state is generated. When the target fish group's outer edge continuously moves away from the center of the group, a net divergence state is generated. Within the current trial response window, a same-direction discrimination sector is established based on the swimming direction of the bionic robotic fish. The main movement direction of the target fish group within the current trial response window is extracted. When the main movement direction stably falls into the same-direction discrimination sector, a direction convergence state is generated. When the main movement direction continues to deviate from the same-direction discrimination sector and changes in a direction away from the aggregation operation area, a direction reversal state is generated. When the net convergence state and the orientation convergence state occur simultaneously, an acceptable position is registered. When the net divergence state and the orientation reversal state occur simultaneously, an avoidance position is registered. The acceptable positions are continuously verified along the test point sequence to form the fish swarm acceptance fingerprint.

[0010] As a preferred embodiment of the fish agglomeration method based on biomimetic swimming signals described in this invention, the steps of screening out acceptable gear positions by the fish receiving fingerprints and determining the leading edge access gear positions, and eliminating trial points associated with avoiding gear positions to form a leading edge candidate range, are as follows: The acceptable levels that have passed continuous verification are extracted from the fingerprints received by the fish swarm, and the perturbation intensity is checked from low to high according to the tail swing rhythm. The acceptable level that meets the continuous acceptance condition and has the lowest tail swing rhythm is determined as the leading edge access level. Based on the fingerprint positioning registration of the fish swarm, a buffer segment is set with the outer edge segment where the avoidance test point is located as the center, and a leading edge elimination segment is generated. The leading edge elimination segment is deleted from the smooth outer edge contour, and the test points that can correspond to the leading edge access position are retained to generate the leading edge candidate range.

[0011] As a preferred embodiment of the fish school aggregation method based on biomimetic swimming signals described in this invention, the specific steps of forming the leading edge access task are as follows: Project the retained test points within the candidate range of the leading edge onto the smooth outer edge contour, and take the end closest to the entrance of the clustering operation domain as the sorting starting point, arrange the retained test points according to the arc length direction to generate the leading edge sequence. Check the movement direction of the outer edge segment corresponding to the retained test point along the leading edge sequence, and generate a candidate segment in the same direction when the movement direction continuously points to the clustering operation area. Check the retraction state of the outer edge segment corresponding to the retained test point within the candidate segment in the same direction, and generate a candidate segment for merging when the outer edge segment continues to enter the center of the group. Three-point sliding verification is performed on the reserved test points within the candidate segment. When the outer edge segment transitions continuously without sharp bends, a virtual leading edge calibration segment is generated. Then, a safe approach distance is offset to the outside of the target fish group along the outer normal direction of the middle outer edge segment of the virtual leading edge calibration segment to mark the virtual leading edge fish position. The virtual leading edge fish position is then bound to the leading edge access position to form a leading edge access task.

[0012] As a preferred embodiment of the fish agglomeration method based on biomimetic swimming signals described in this invention, the process involves generating a companion trajectory around the virtual leading edge fish position in the leading edge access task. The biomimetic robotic fish enters the companion trajectory according to the leading edge access level. When the target fish school moves in the same direction at the leading edge, it finely adjusts its leading angle towards the agglomeration operation area to form a leading edge traction table. The specific steps are as follows: The virtual leading edge fish position is extracted from the leading edge access task. The trajectory starting point is determined by the current position of the bionic robotic fish. A tangential approach point is set outside the virtual leading edge fish position. A cubic B-spline is used to smoothly fit the trajectory starting point, the tangential approach point and the virtual leading edge fish position to generate a continuous turning companion trajectory. The leading edge of the escort trajectory is corrected, the distance relationship between the escort trajectory and the outer edge of the target fish group is checked, and the trajectory segments that deviate from the safe escort distance zone are pulled back along the outer edge normal direction to generate an effective escort trajectory. After the bionic robotic fish enters the effective companion trajectory, it switches to the leading edge access position by ramp transition and moves along the effective companion trajectory by forward viewpoint tracking to generate the companion entry state. While the target fish are in the same direction as the leading edge of the school, check the state of movement of the leading edge. When the state of movement of the leading edge is established, use a small angle limit method to finely adjust the leading angle towards the gathering operation area and generate the leading edge traction table.

[0013] As a preferred embodiment of the fish school aggregation method based on biomimetic swimming signals described in this invention, the specific steps for forming a stable aggregation state are as follows: The stable leading angle is retrieved from the leading edge traction table to track the position change of the outer edge segment of the target fish group relative to the entrance of the gathering operation area, and a gathering entry state is generated when the main body of the outer edge of the target fish group enters the inner side of the gathering operation area. When the aggregation state is established, the outer edge boundary of the aggregation operation area is sampled with equal arc length and smoothed with curvature to generate a low-speed circular path, and the bionic robotic fish is controlled to move along the low-speed circular path at the circular gear to generate the outer edge circular state. In the outer edge swimming state, an outward dispersion discrimination zone is set up, and the outward dispersion edge segment is confirmed by the hysteresis judgment method. The bionic robotic fish is controlled to deflect at a small angle towards the center of the aggregation operation area in the pull-back position. When the outward dispersion edge segment retracts to the inside of the aggregation operation area, a stable aggregation state is generated.

[0014] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the fish agglomeration method based on biomimetic swimming signals as described in the first aspect of the present invention.

[0015] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the fish school aggregation method based on biomimetic swimming signals as described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: by setting up test points on the outer edge of the target fish group and outputting a biomimetic test sequence, the net gathering state, net dispersing state, orientation convergence state, and orientation reversal state are identified within the same test response window, forming a fish group acceptance fingerprint. Then, the leading edge access level is screened and the avoidance test points are eliminated, so that the biomimetic robotic fish can determine the access strength and access range based on the actual response of the fish group, which enhances the adaptability of the leading edge access and improves the stability of the fish group transitioning from a natural swimming state to a controlled gathering state. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a method for fish school aggregation based on biomimetic swimming signals.

[0019] Figure 2 A flowchart for fish swarm situation identification and probing modeling.

[0020] Figure 3 A flowchart for fingerprint formation and leading edge access determination for fish swarms.

[0021] Figure 4 A flowchart for constructing the leading edge access task and escort traction.

[0022] Figure 5 A comparison chart showing the deviation in the main movement direction of the fish swarm under different access methods. Detailed Implementation

[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0024] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0025] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0026] Reference Figures 1-5 As one embodiment of the present invention, this embodiment provides a method for fish school aggregation based on biomimetic swimming signals, including the following steps: S1. Conduct fish school situation identification in the target waters, delineate the gathering operation area along the distribution range of the target fish school, and form a fish school situation table.

[0027] S1.1 Establish operational coordinate benchmarks for the target water area, collect water area perception information and identify the area where the target fish school is located, determine the outer contour and movement direction of the target fish school, then delineate the gathering operation area along the outer contour of the target fish school towards the gathering target location, and avoid unsuitable gathering locations, and organize the target fish school situation with the gathering operation area to form a fish school situation table.

[0028] Furthermore, a work coordinate reference is established using stable positioning points as a benchmark. The direction towards the target location is determined as the main work direction, and the continuously collected water sensing information is uniformly mapped to the work coordinate reference. Based on the spatial proximity of fish positions at the same collection time and the continuous movement relationship between adjacent collection times, valid fish positions belonging to the same target fish group are merged, and isolated positions separated from the main space of the fish group are excluded, thus forming the area where the target fish group is located.

[0029] Outer edge points are extracted along the continuous boundary between the target fish school area and the surrounding waters. The adjacency relationship of the outer edge points is determined by the spatial sampling resolution of the water sensing device. Starting from any outermost outer edge point, adjacent outer edge points are continuously tracked along a fixed direction until the starting outer edge point is returned. The sequence of outermost outer edge points that closes the beginning and end and continuously encloses the target fish school area is determined as the outer edge contour of the target fish school. When there are local depressions, protrusions or asymmetrical shapes, the tracking is still carried out continuously along the actual boundary position, and regular geometric contours are not used as substitutes.

[0030] At each effective acquisition moment, the positions of all valid fish in the target fish swarm are read, and the average lateral and longitudinal positions are determined respectively. The corresponding coordinate positions are then used as the center of the target fish swarm. The fixed effective output period of the water sensing device and the fixed number of consecutive frames used for continuous target tracking are read. The effective acquisition moments corresponding to the number of consecutive frames form the movement orientation sampling window, and the sampling period and sampling window length are kept consistent throughout the same aggregation operation.

[0031] The target fish school's center position is arranged according to the sampling time sequence within the movement direction sampling window. The direction from the center position at the start sampling time to the center position at the end sampling time determines the candidate movement direction. Then, using the candidate movement direction as a reference, the leading and trailing edge positions of the target fish school's outer contour are determined, and the overall changes of the leading and trailing edge positions within the movement direction sampling window are verified. When both the leading and trailing edge positions are consistent with the candidate movement direction, the candidate movement direction is determined as the target fish school's movement direction. When only a short-term local oscillation occurs and the overall direction does not deviate from the candidate movement direction, the candidate movement direction is not changed. When the leading or trailing edge position continuously deviates from the candidate movement direction, the current movement direction sampling window is determined as the direction conflict range, and the target fish school's location, outer contour, and movement direction are re-determined based on subsequent water area perception information.

[0032] After confirming the outer contour of the target fish school and its direction of movement, the gathering operation area is determined by extending along the outer contour of the target fish school towards the gathering target location. Shallow water areas, underwater obstacle locations, locations affected by strong currents, and prohibited operation boundaries are excluded to form the gathering operation area. The area where the target fish school is located, the outer contour of the target fish school, the location of the center of the target fish school, the direction of movement of the target fish school, the gathering target location, the gathering operation area, and unsuitable gathering locations are organized according to the operation coordinate reference to form a fish school situation table.

[0033] It should also be noted that the effective acquisition time is the acquisition time when the water area sensing device completes one water area sensing information output according to a fixed effective output cycle, and the corresponding water area sensing information can simultaneously obtain the effective fish body position and the outer contour of the target fish school; the time interval between adjacent effective acquisition times is determined by the fixed effective output cycle of the water area sensing device. The number of effective acquisition times included in the moving orientation sampling window is directly read from the number of consecutive frames pre-fixed in the continuous target tracking configuration of the water area sensing device, and is not changed temporarily according to the judgment result; the total sampling duration of the moving orientation sampling window is determined by the actual time length between the first effective acquisition time and the last effective acquisition time, and remains fixed in the same aggregation operation. The fixed effective output cycle, spatial sampling resolution, and fixed number of consecutive frames for continuous target tracking of the water area sensing device are all read and registered in the fish school situation recognition parameter record before the aggregation operation begins, in order to ensure that different moving orientation sampling windows use a consistent time scale and spatial scale.

[0034] Unsuitable gathering locations are determined by applying water area sensing information to the operational coordinate baseline and by checking each location for insufficient water depth, underwater obstacles, water flow deviation from the gathering target location, and prohibited operation boundaries.

[0035] S2. Smooth the outline of the target fish group recorded in the fish group situation table, and set up test points along the smoothed target fish group outer edge. The bionic robotic fish outputs a bionic test sequence at the test points in sequence. Compare the changes in fish group gathering and fish group orientation before and after the test within the same test response window. When the fish group gathering is enhanced and the fish group orientation tends to be similar, register the acceptable level. When the fish group disperses at the outer edge and the fish group orientation is reversed, register the avoidance level, thus forming a fish group acceptance fingerprint.

[0036] S2.1 Extract the outer contour of the target fish school from the fish school situation table, resample the outer contour of the target fish school with equal arc length, and use Savitzky-Golay smoothing filter to reduce the outer edge spikes and generate a smooth outer contour.

[0037] Furthermore, using the operational coordinate benchmark as a reference, the outer contour of the target fish school, which has already been organized in the fish school situation table, is called up. The enclosure relationship between the outer contour of the target fish school and the area where the target fish school is located is checked. Isolated boundary points that do not continuously enclose the area where the target fish school is located are eliminated, and the outer edge points that can be continuously extended along the outside of the target fish school are retained. Then, the arc length distance between adjacent outer edge points is calculated along the extension direction of the outer contour of the target fish school, and the outer edge points are reselected on the outer contour of the target fish school according to equal arc length intervals. The coordinates of the resampled positions that fall between adjacent outer edge points are determined by linear interpolation, so that the outer contour of the target fish school after resampling has a uniform point spacing. A sliding window is formed by resampled adjacent outer edge points. Savitzky-Golay smoothing filter is used to perform local polynomial fitting on the coordinates of the outer edge points within the sliding window. This preserves the continuous variation trend of the outer edge contour of the target fish group along the arc length direction, and weakens the short-distance protruding outer edge spikes and local sharp points caused by acquisition jitter. After completing the Savitzky-Golay smoothing filter, the filtered outer edge points are continuously checked to ensure that adjacent outer edge points can still enclose the area where the target fish group is located. The continuously checked outer edge points are then connected according to the extension direction of the outer edge contour of the target fish group to generate a smooth outer edge contour.

[0038] It should also be noted that Savitzky-Golay smoothing filter is a smoothing filtering method. The process involves setting a sliding window on the coordinate sequence of the outer edge points of the target fish school's outer contour. Within each sliding window, a least-squares polynomial fitting is used to preserve the local variation trend of the outer edge point coordinate sequence. Then, the fitted coordinates corresponding to the center position of the sliding window are used to replace the original coordinates, thereby reducing short-distance spikes and acquisition jitter, while preserving the overall curvature of the target fish school's outer contour as much as possible.

[0039] S2.2. Select trial positions sequentially along the smooth outer edge contour toward the aggregation operation area, and keep the trial positions outside the outer edge of the target fish group to generate a trial point sequence. The bionic robotic fish arrives at the current trial point according to the trial point sequence, maintains a low-disturbance swimming state at the current trial point, and uses the duration of the low-disturbance swimming state as the trial reference window to obtain the natural outer edge changes of the target fish group within the trial reference window.

[0040] Furthermore, using the smooth outer edge contour as the selection benchmark, and combining it with the direction of the gathering operation area in the fish school situation table, starting from the end of the smooth outer edge contour that is close to the entrance of the gathering operation area, the direction of the outer edge is checked segment by segment according to the extension sequence of the smooth outer edge contour. A safe approach distance is offset on the outside of each smooth outer edge contour in the direction away from the center of the target fish school. Points that are still on the side of the gathering operation area after offset and do not fall into unsuitable gathering areas are used as trial positions. When the offset point is close to the inner side of the outer edge of the target fish group, it continues to move outward along the direction away from the center of the target fish group to the outer side of the outer edge of the target fish group, so that the test position and the outer edge of the target fish group maintain an interval that allows the bionic robotic fish to swim with low disturbance; after completing the selection of the entire smooth outer edge contour, the test positions are arranged in the order of the smooth outer edge contour extending from one end of the gathering operation area entrance to the other end, generating a sequence of test points.

[0041] The bionic robotic fish arrives at the current test point sequentially according to the test point sequence. At the current test point, it adopts a low-amplitude tail swing, low-speed cruising, and directional maintenance to form a low-disturbance swimming state, so that the bionic robotic fish stays outside the outer edge of the target fish group without actively pressing the outer edge of the target fish group. The duration of the low-disturbance swimming state is used as the test reference window. Water perception information is continuously collected within the test reference window, and the natural displacement direction and natural contour changes of the outer edge of the target fish group under the low-disturbance swimming state are extracted from the water perception information. The changes of the outer edge of the target fish group within the test reference window without superimposed bionic test sequence stimulation are taken as the natural outer edge changes of the target fish group.

[0042] S2.3. At the current test point, perform biomimetic swimming sequentially according to the tail swing rhythm gears from low to high to form a biomimetic test sequence, and determine the duration from the start of output to the stop of output at the current gear as the current test response window.

[0043] Furthermore, when performing biomimetic swimming at the current probing point, the biomimetic robotic fish first maintains itself at the current probing point on the outer edge of the target fish school, and then outputs biomimetic swimming actions sequentially according to the tail sway rhythm level from low to high. Before each tail sway rhythm level is output, the biomimetic robotic fish maintains the orientation of the current probing point stable, and then continues to perform biomimetic swimming at the current tail sway rhythm level. After completing the output of the current tail sway rhythm level, it stops the current tail sway rhythm level and enters the output of the next tail sway rhythm level. After completing all biomimetic swimming actions sequentially from low to high tail sway rhythm levels, a biomimetic probing sequence is formed. At the same time, the duration from the start of output to the stop of output of each current level is determined as the current probing response window, which is used for subsequent comparison of changes in the outer edge of the target fish school and changes in the orientation of the target fish school.

[0044] S2.4 Within the current trial response window, the natural outer edge change of the target fish group in the trial reference window is used as the reference outer edge zone. The displacement direction of the target fish group's outer edge relative to the reference outer edge zone is extracted. When the target fish group's outer edge continuously enters towards the center of the group, a net convergence state is generated. When the target fish group's outer edge continuously moves away from the center of the group, a net divergence state is generated.

[0045] Furthermore, within the current trial response window, the natural outer edge changes of the target fish group in the trial reference window are first placed into the working coordinate reference, and the outer boundary range formed by the natural swing of the target fish group's outer edge within the trial reference window is organized into the reference outer edge band. Subsequently, the outer edge of the target fish group is continuously extracted within the current trial response window, and the outer edge of the target fish group in the current trial response window and the reference outer edge band are placed under the same working coordinate reference for position verification. The verification process uses the center of the target fish group as the direction reference. When the outer edge of the target fish group continuously enters the side closer to the center of the target fish group relative to the reference outer edge band, it indicates that after the current gear output, the outer edge of the target fish group does not swing naturally, but shows an inward retraction trend, generating a net contraction state, which is used to confirm that the current tail swing rhythm gear has a contraction induction effect on the outer edge of the target fish group, improving the accuracy of the acceptable gear registration. When the outer edge of the target fish group continuously extends away from the center of the target fish group relative to the reference outer edge zone, it indicates that the outer edge of the target fish group does not swing naturally after the current gear output, but shows an outward diffusion trend, generating a net outward diffusion state. This is used to confirm that the current tail swing rhythm gear has an outward diffusion disturbance effect on the outer edge of the target fish group, enhancing the reliability of avoidance gear registration and leading edge candidate range elimination.

[0046] S2.5 Within the current trial response window, establish a same-direction discrimination sector based on the swimming direction of the bionic robotic fish, extract the main movement direction of the target fish group within the current trial response window, generate a direction convergence state when the main movement direction stably falls into the same-direction discrimination sector, and generate a direction reversal state when the main movement direction continuously deviates from the same-direction discrimination sector and changes in a direction away from the aggregation operation area.

[0047] Furthermore, within the current trial response window, taking the leading direction of the bionic robotic fish during the current output gear as the central orientation, the leading direction angle of the bionic robotic fish during the current tail sway rhythm output gear is extracted, and the allowable orientation deviation range is expanded to both sides with the leading direction angle as the center, forming directional boundaries on both sides; the fan-shaped range between the two directional boundaries is expanded towards the gathering operation area to form a same-direction discrimination sector, which is used to check whether the main movement orientation of the target fish group within the current trial response window is consistent with the leading direction of the bionic robotic fish.

[0048] Within the current trial response window, the changes in the outer edge position of the target fish school are extracted according to the continuous acquisition time, and the main movement direction is determined based on the displacement direction of the target fish school center during the continuous acquisition time. Then, the main movement direction is checked against the same direction discrimination sector. When the main movement direction remains within the same direction discrimination sector within the current trial response window, it indicates that the target fish school is moving in the same direction as the lead direction of the bionic robotic fish, generating a direction convergence state. When the main movement direction deviates from the same direction discrimination sector within the current trial response window, and the main movement direction changes towards the direction away from the gathering operation area, it indicates that the current tail swing rhythm gear causes the target fish school to have a reverse deviation trend, generating a direction reversal state. This is used to confirm that the current tail swing rhythm gear causes the main movement direction of the target fish school to deviate from the lead direction of the bionic robotic fish and move away from the gathering operation area, thus enhancing the reliability of the avoidance gear registration.

[0049] S2.6 When the net convergence state and the orientation convergence state occur simultaneously, register the acceptable position; when the net divergence state and the orientation reversal state occur simultaneously, register the avoidance position. And continuously verify the acceptable position along the test point sequence to form the fish swarm acceptance fingerprint.

[0050] Furthermore, after verifying the current test response window of the current gear at the current test point, the net convergence state and the orientation convergence state are placed in the same current test response window for corresponding verification. When the net convergence state comes from the outer edge of the target fish group continuously entering the center of the group under the current gear, and the orientation convergence state comes from the main movement orientation under the current gear stably falling into the same direction discrimination sector, the current gear is registered as an acceptable gear. The acceptable gear is the tail sway rhythm gear corresponding to the simultaneous occurrence of the net convergence state and the orientation convergence state in the same test response window. It is used to mark the tail sway rhythm gear that can make the outer edge of the target fish group converge and make the orientation of the target fish group converge, providing a basis for determining the leading edge access gear.

[0051] The net outward dispersion state and the orientation reversal state are placed in the same current trial response window for corresponding verification. When the net outward dispersion state is caused by the outer edge of the target fish group continuously moving away from the center of the group under the current gear, and the orientation reversal state is caused by the main movement orientation continuously deviating from the same direction discrimination sector and changing towards the direction away from the aggregation operation area under the current gear, the current gear is registered as the avoidance gear. The avoidance gear is the tail wagging rhythm gear registered when the net outward dispersion state and the orientation reversal state occur simultaneously in the same trial response window. It is used to mark the tail wagging rhythm gear that is likely to cause the outer edge of the target fish group to disperse and cause the target fish group to reverse its orientation, providing a basis for eliminating the leading edge candidate range.

[0052] The registration status of acceptable positions in adjacent test points is checked one by one according to the order of the test point sequence. When the same tail sway rhythm position is continuously registered as an acceptable position in adjacent test points, and no avoidance position corresponding to the same tail sway rhythm position is registered in adjacent test points, the acceptable position is retained. When the same tail sway rhythm position is only registered as an acceptable position in a single test point, and no continuous acceptance relationship is formed between adjacent test points, it is not considered as an acceptable position that has passed continuous review. The acceptable positions, avoidance positions, and continuously reviewed acceptable positions are arranged according to the test point sequence to form the fish school acceptance fingerprint. It should also be noted that the fish swarm acceptance fingerprint is a fish swarm response characterization formed after registering acceptable levels, avoidance levels, and continuously verified acceptable levels along the test point sequence. It is used to reflect the target fish swarm's acceptance and avoidance of different tail wagging rhythm levels, and provides a basis for subsequent screening of acceptable levels, determination of leading edge access levels, and elimination of test points associated with avoidance levels.

[0053] S3. The fish swarm receives fingerprints to screen out acceptable gear positions and determines the leading edge access gear position. The test points associated with the avoidance gear position are eliminated to form the leading edge candidate range. The retained test points are sorted in order of leading edge priority along the direction of the aggregation operation domain. When the retained test points continuously show the leading edge in the same direction, the edge converges and the turning is gentle, the virtual leading edge fish position is marked and the virtual leading edge fish position is bound to the leading edge access gear position to form the leading edge access task.

[0054] S3.1 Extract acceptable gears that have passed continuous verification from the fingerprints received by the fish swarm, and check the disturbance intensity from low to high according to the tail swing rhythm. Determine the acceptable gear that meets the continuous acceptance condition and has the lowest tail swing rhythm as the leading edge access gear.

[0055] Furthermore, using the order of the test point sequence as the verification order, each tail swing rhythm position is searched for that is continuously registered as an acceptable position in adjacent test points. The registration content of the same tail swing rhythm position in adjacent test points is checked accordingly. When the same tail swing rhythm position corresponds to the net convergence state and the orientation convergence state in adjacent test points, and the same tail swing rhythm position does not correspond to the net divergence state and the orientation reversal state in adjacent test points, the same tail swing rhythm position is regarded as an acceptable position that has passed continuous verification.

[0056] Acceptable levels are arranged in a fixed order from low to high according to the tail sway rhythm levels and have been continuously verified. The order of the tail sway rhythm levels indicates the magnitude of the disturbance. First, it is checked whether the first acceptable level meets the continuous acceptance condition. Then, the subsequent acceptable levels are checked in turn. When the first acceptable level has been stably corresponding to the net convergence state and the orientation convergence state in the adjacent test points and there is no corresponding avoidance level, the first acceptable level in the tail sway rhythm level sequence that meets the continuous acceptance condition is determined as the leading edge access level. This level is used to determine the bionic swimming intensity used when the bionic robotic fish enters the leading edge of the target fish group to accompany and guide it, so as to avoid the bionic robotic fish using an excessively strong tail sway rhythm that causes the target fish group to scatter.

[0057] It should also be noted that the continuous acceptance condition refers to the fact that the same tail swing rhythm position continuously corresponds to the net convergence state and the orientation convergence state in adjacent test points, and the same tail swing rhythm position does not correspond to the net divergence state and the orientation reversal state in adjacent test points; the continuous acceptance condition is determined by checking the registration content of the same tail swing rhythm position item by item along the test point sequence.

[0058] S3.2. Based on the fingerprint positioning registration of the fish swarm, a buffer segment is set with the outer edge segment where the avoidance test point is located as the center, and a leading edge elimination segment is generated. The leading edge elimination segment is deleted from the smooth outer edge contour, and the test points that can correspond to the leading edge access position are retained to generate the leading edge candidate range.

[0059] Furthermore, the gear registration information corresponding to each test point is checked item by item along the test point sequence, and the test points with the avoidance gear are marked as avoidance test points. Then, the avoidance test points are projected onto the smooth outer edge contour, the corresponding outer edge segment on the smooth outer edge contour is found, and the outer edge segment where the avoidance test point is located is used as the center. Buffer segments are set on both sides along the extension direction of the smooth outer edge contour so that the outer edge segments near the avoidance test point that are prone to triggering net divergence state and orientation reverse deviation state are included in the leading edge removal range, generating the leading edge removal segment.

[0060] According to the arc length position of the leading edge removal segment in the smooth outer edge contour, the corresponding outer edge segment is cut off, and the test points falling into the leading edge removal segment are simultaneously removed; then, the correspondence between the test points and the leading edge access position is checked along the remaining smooth outer edge contour, and only the test points that can correspond to the leading edge access position and have not been registered as avoidance positions are retained, so that the retained test points can both avoid the fish-sensitive outer edge segment and support the execution of the leading edge access position, thus generating the leading edge candidate range.

[0061] S3.3 Project the retained test points within the candidate range of the leading edge onto the smooth outer edge contour, and take the end closest to the entrance of the clustering operation domain as the sorting starting point, arrange the retained test points according to the arc length direction to generate the leading edge sequence.

[0062] Furthermore, using the work coordinate benchmark as a unified reference, each retained test point is mapped onto the smooth outer edge contour along the direction close to the smooth outer edge contour, and the outer relationship between the retained test point and the smooth outer edge contour is checked to ensure that the retained test point is still located outside the outer edge of the target fish group. Determine the starting point for sorting from the end of the smooth outer edge contour closest to the entrance of the gathering operation area. Check the corresponding position of each reserved test point on the smooth outer edge contour along the arc length direction. Place the reserved test points closer to the starting point in the first position, and place the reserved test points further away from the starting point along the arc length direction in the last position. This ensures that the arrangement order of the reserved test points reflects the order in which the target fish school enters the gathering operation area from the leading edge, thus generating the leading edge sequence.

[0063] S3.4. Check the movement direction of the outer edge segment corresponding to the retained test point along the leading edge sequence, and generate a candidate segment in the same direction when the movement direction continuously points to the aggregation operation area. Check the retraction state of the outer edge segment corresponding to the retained test point within the candidate segment in the same direction, and generate a candidate segment for convergence when the outer edge segment continues to enter the center of the group.

[0064] Furthermore, using the order of the reserved test points in the leading edge sequence as the verification order, the corresponding position of each reserved test point on the smooth outer edge contour is used as the positioning reference for the outer edge segment. The outer edge segment corresponding to the reserved test point that can represent the changes in the leading edge of the target fish group is selected. Then, the position change direction of the outer edge segment corresponding to the reserved test point is checked during continuous acquisition, and the position change direction is compared with the direction of the aggregation operation area. When the position change direction of the outer edge segment corresponding to the adjacent reserved test point is continuously towards the aggregation operation area, and there is no movement change away from the aggregation operation area, the outer edge segments that continuously meet the direction requirements are sorted into candidate segments in the same direction.

[0065] Using the center of the target fish population as a reference for the retraction direction, the positional change of the outer edge segment corresponding to the retained test point relative to the smooth outer edge contour during continuous sampling is checked item by item along the candidate segments in the same direction. When the outer edge segment corresponding to the retained test point continues to move towards the center of the target fish population within the candidate segments in the same direction, and the trend of the outer edge segment entering is consistent with the acceptable level registration content corresponding to the front edge access level, the candidate segments in the same direction that continuously meet the retraction requirements are sorted into the convergence candidate segments.

[0066] S3.5. Perform a three-point sliding check on the remaining test points within the candidate section, and generate a virtual leading edge calibration section when the outer edge section transitions continuously without sharp bends. Then, offset a safe approach distance to the outside of the target fish group along the outer normal direction of the middle outer edge section of the virtual leading edge calibration section to calibrate the virtual leading edge fish position, and bind the virtual leading edge fish position with the leading edge access position to form a leading edge access task.

[0067] Furthermore, the tangential direction of the outer edge segment corresponding to each retained test point within the candidate converging segment, along the direction of increasing arc length of the leading edge sequence, is determined as the orientation of the corresponding outer edge segment. Starting from the second retained test point in the candidate converging segment, the previous retained test point, the current retained test point, and the next retained test point are formed into a current three-point verification group. The first adjacent orientation difference between the outer edge segment corresponding to the previous retained test point and the outer edge segment corresponding to the current retained test point, and the second adjacent orientation difference between the outer edge segment corresponding to the current retained test point and the outer edge segment corresponding to the next retained test point are determined respectively. The two adjacent orientation differences are then verified against the allowable range for smooth turning.

[0068] When both the first and second adjacent orientation differences are within the allowable range for smooth turning, the outer edge segment corresponding to the currently retained test point is registered as a three-point check passed outer edge segment; when any adjacent orientation difference exceeds the allowable range for smooth turning, the outer edge segment corresponding to the currently retained test point is registered as a three-point check failed outer edge segment. After completing the current three-point check group, the currently retained test point is moved to the next retained test point along the leading edge sequence, and the corresponding previous and next retained test points are reselected based on the moved current retained test point to form the next three-point check group. This process is repeated until the last retained test point in the converging candidate segments that can form a three-point check group, completing the three-point sliding check.

[0069] Merge consecutive three-point checks along the leading edge sequence through the outer edge segments. When consecutively passed outer edge segments continue to move in the same direction as the aggregation operation area and continue to enter the target fish school center side, and the orientation difference between corresponding adjacent outer edge segments is within the allowable range of gentle turning, the corresponding consecutive outer edge segments are marked as virtual leading edge marking segments. When the three-point check fails to pass the outer edge segment and is located between consecutively passed outer edge segments, terminate the merging of the current consecutive outer edge segments at the corresponding position, and restart the continuous merging from the subsequent three-point check passed outer edge segments.

[0070] Select the outer edge segment of the middle of the virtual leading edge calibration segment, and determine the outer normal direction towards the outside of the target fish group based on the tangential direction of the outer edge segment along the leading edge sequence. Offset a safe approach distance to the outside of the target fish group along the outer normal direction. Mark the points that are still outside the outer edge of the target fish group after offset and have not fallen into unsuitable gathering positions as virtual leading edge fish positions. Bind the virtual leading edge fish positions to the corresponding leading edge access positions, so that the bionic robotic fish enters the companion position from the outside of the leading edge of the target fish group according to the leading edge access positions, thus forming the leading edge access task.

[0071] It should also be noted that the three-point sliding check is performed by sequentially selecting the previous retained test point, the current retained test point, and the next retained test point according to the leading edge sequence to form a three-point check group. The orientation differences between the previous outer edge segment and the current outer edge segment, and between the current outer edge segment and the next outer edge segment are compared respectively. If the orientation differences of two adjacent segments do not exceed the allowable range for smooth turning, the outer edge segment corresponding to the current retained test point is retained; otherwise, it is not retained. After completing one check, the current retained test point is moved to the next retained test point along the leading edge sequence, and the next three-point check group is formed again, until all retained test points within the candidate segment that can form a three-point check group are checked.

[0072] The allowable range for smooth steering is determined based on the orientation change of the outer edge segment under stable folding conditions, the positioning error of the outer edge points, and the minimum turning radius of the bionic robotic fish. The orientation difference between adjacent outer edge segments is statistically analyzed when stable folding occurs without orientation reversal, and the maximum orientation difference is determined as the upper limit of the basic orientation range. Using the two endpoints of the outer edge segment as positioning references, the extreme endpoint position that causes the outer edge segment's orientation to deviate the most is determined within the allowable positioning error range of the outer edge points. The angle between the extreme outer edge direction and the original outer edge direction is determined as the positioning orientation tolerance. Based on this, the upper limit of the basic orientation range is extended to the side with the larger orientation difference, forming the positioning correction upper limit. The corresponding extreme turning arc is then determined using the minimum turning radius of the bionic robotic fish and the arc length distance between adjacent outer edge segments. The steering angle corresponding to the extreme turning arc is determined as the maximum steering change range. The positioning correction upper limit and the maximum steering change range are compared, and the smaller one is taken as the upper limit of the allowable range for smooth steering. When the orientation difference between adjacent outer edge segments does not exceed the upper limit of the allowable range for smooth steering, it is determined as smooth steering; otherwise, it is determined as sharp turning.

[0073] S4. Generate a companion trajectory around the virtual leading edge fish position in the leading edge access task. The bionic robotic fish enters the companion trajectory according to the leading edge access level. When the target fish group moves in the same direction at the leading edge, it finely adjusts the leading angle towards the gathering operation area to form a leading edge traction table.

[0074] S4.1 Extract the virtual leading edge fish position from the leading edge access task, determine the trajectory starting point based on the current position of the bionic robotic fish, set the tangential approach point outside the virtual leading edge fish position, and use cubic B-splines to smoothly fit the trajectory starting point, tangential approach point and virtual leading edge fish position to generate a continuous turning companion trajectory.

[0075] Furthermore, the virtual leading edge fish position is extracted in the leading edge access task, and the outer relationship between the virtual leading edge fish position and the outer edge of the target fish group is checked with the operation coordinate reference. The current position of the bionic robotic fish is placed in the same operation coordinate reference, and the current position of the bionic robotic fish is used as the trajectory starting point. The outer edge tangential direction is extracted according to the virtual leading edge calibration segment where the virtual leading edge fish position is located. A position that allows the bionic robotic fish to approach the leading edge of the target fish group in the forward direction is selected outside the virtual leading edge fish position along the outer edge tangential direction. The tangential approach point is determined so that the trajectory starting point, the tangential approach point and the virtual leading edge fish position form a spatial sequence of gradually approaching the leading edge of the target fish group from the outside.

[0076] A cubic B-spline is used to smoothly fit the trajectory starting point, tangential approach point, and virtual leading edge fish position. The cubic B-spline uses piecewise cubic polynomials to continuously connect the path from the trajectory starting point to the tangential approach point and the path from the tangential approach point to the virtual leading edge fish position, so that the path position and tangential direction remain continuous between adjacent path segments. After the smooth fitting is completed, it is checked whether the smooth fitted path is located outside the outer edge of the target fish school and whether the smooth fitted path avoids unsuitable gathering positions. If the smooth fitted path can smoothly reach the virtual leading edge fish position from the trajectory starting point and maintain continuous turning along the way, a continuously turning companion trajectory is generated.

[0077] It should also be noted that cubic B-splines, also known as cubic B-spline curves, are a curve fitting and smoothing interpolation method based on piecewise cubic polynomials. The curve shape is determined by the sequence of control points and the sequence of nodes. Cubic B-splines continuously connect the paths between multiple control points using piecewise cubic polynomials, ensuring a smooth transition in path position, tangential direction, and curvature. In the generation of the swimming trajectory of biomimetic robotic fish, cubic B-splines are used to connect the trajectory starting point, the tangential approach point, and the virtual leading edge fish position, enabling the biomimetic robotic fish to approach the leading edge of the target fish school in a continuous turning manner, reducing the disturbance caused by sharp turns to the target fish school.

[0078] S4.2. Correct the leading edge of the escort trajectory, check the distance relationship between the escort trajectory and the outer edge of the target fish group, and revert the trajectory segments that deviate from the safe escort distance zone along the outer edge normal direction to generate an effective escort trajectory.

[0079] Furthermore, using the outer edge of the target fish school as a reference, trajectory segments are selected sequentially along the extension of the escort trajectory. Each trajectory segment is placed within the same operational coordinate reference corresponding to the outer edge of the target fish school. The distance from the trajectory segment to the outer edge of the target fish school is calculated and checked against the safe escort distance zone. When the distance of the trajectory segment is less than the inner boundary of the safe escort distance zone, it is pulled back along the outer edge normal direction of the outer edge of the target fish school, moving away from the outer edge of the target fish school. When the distance of the trajectory segment is greater than the outer boundary of the safe escort distance zone, it is pulled back along the outer edge normal direction of the outer edge of the target fish school, moving closer to the outer edge of the target fish school, until the trajectory segment enters the safe escort distance zone. After completing the distance check and pullback of all trajectory segments of the escort trajectory, the connection direction of adjacent trajectory segments is checked for continuity, so that the escort trajectory after pullback still maintains continuous turning, generating a valid escort trajectory.

[0080] The distance from the trajectory segment to the outer edge of the target fish school is calculated using the following expression: ; in: It is the escort route number 1 The minimum distance from a trajectory segment to the outer edge of the target fish school, in units of length; It is a point The horizontal coordinate in the work coordinate datum is measured in units of length. It is a point The horizontal coordinate in the work coordinate datum is measured in units of length. It is a point The vertical coordinate in the work coordinate datum is measured in units of length. It is a point The vertical coordinate in the work coordinate datum is measured in units of length. It is a point Located in the escort route Trajectory segment superior; It is a point The continuous outer edge line corresponding to the outer edge of the target fish school superior; It is the sequence number of the trajectory segment in the escort route; It is the escort route number 1 Any point on a trajectory segment; It is any point on the continuous outer edge line corresponding to the outer edge of the target fish school; It is the marker corresponding to the outer edge of the target fish school; The above expression is used to characterize the escort trajectory of the first... Trajectory segment The continuous outer edge line corresponding to the outer edge of the target fish school The shortest spatial interval between them can express the degree of fit between the trajectory segment and the outer edge of the target fish school as a length quantity. This provides a distance basis for subsequent verification of whether the escort trajectory is within the safe escort distance zone; the coordinates in the expression all fall under the same working coordinate reference, and the horizontal and vertical coordinate differences are both length quantities, forming It is still a length quantity, with consistent dimensions.

[0081] It should also be noted that the interval between the bionic robotic fish and the outer edge of the target fish school is gradually changed along the outer edge of the leading edge. The minimum interval corresponding to the absence of net outward dispersion and orientation reversal is determined as the initial inner boundary, and the maximum interval corresponding to the continued formation of net convergence and orientation convergence is determined as the initial outer boundary. The maximum normal distance from the positioning reference point of the bionic robotic fish to the outermost contour on the side facing the target fish school is used to determine the size margin of the robotic fish. The maximum positioning error of the positioning device is used to determine the positioning margin, and the maximum normal drift distance caused by water flow disturbance during two adjacent position corrections is used to determine the water flow disturbance margin. The total length formed by connecting the three types of margins end to end along the distance direction is determined as the safety correction distance. The initial inner boundary is adjusted away from the outer edge of the target fish school according to the safety correction distance, and the initial outer boundary is adjusted closer to the outer edge of the target fish school. The range between the two adjusted boundaries is determined as the safe companion distance zone. When the adjusted inner boundary is within the outer boundary, the safe companion distance zone is determined to be valid; otherwise, the initial inner boundary and the initial outer boundary are redefined.

[0082] S4.3 After the bionic robotic fish enters the effective companion trajectory, it switches to the leading edge access position by ramp transition and moves along the effective companion trajectory by forward viewpoint tracking to generate the companion entry state.

[0083] Furthermore, after the bionic robotic fish enters the effective escort trajectory, it transitions smoothly along a ramp according to the sequence of the current swimming gear to the leading edge access gear, allowing the tail wagging rhythm to switch smoothly from the current swimming gear to the leading edge access gear, avoiding sudden changes in the tail wagging rhythm that could disturb the leading edge of the target fish school. After completing the leading edge access gear switch, a forward-looking point is selected ahead of the effective escort trajectory, the guidance direction from the current position of the bionic robotic fish to the forward-looking point is calculated, and the heading of the bionic robotic fish is corrected according to the guidance direction, so that the bionic robotic fish gradually approaches the virtual leading edge fish position along the effective escort trajectory. When the position of the bionic robotic fish falls within the allowable deviation range of the effective escort trajectory, and the tail wagging rhythm of the bionic robotic fish remains stable at the leading edge access gear, the escort track entry state is generated.

[0084] The guiding direction from the current position of the bionic robotic fish to the forward viewpoint is calculated using the following expression: ; in: The current location of the bionic robotic fish Pointing to the forward viewpoint The guiding direction angle, with the dimension of angle; It is the forward viewpoint The vertical coordinate in the work coordinate datum is measured in units of length. The current location of the bionic robotic fish The vertical coordinate in the work coordinate datum is measured in units of length. It is the forward viewpoint The horizontal coordinate in the work coordinate datum is measured in units of length. The current location of the bionic robotic fish The horizontal coordinate in the work coordinate datum is measured in units of length. It is the forward viewpoint on the effective escort route; This is the current position of the bionic robotic fish; It should also be noted that the ramp transition method refers to a smooth switching method that gradually transitions the operating gear from the current swimming gear to the target gear. Usually, according to the gear sequence, amplitude change limit and time continuity requirements, the tail swing rhythm, propulsion amplitude and swimming speed gradually approach the target gear in multiple consecutive moments, avoiding sudden changes in motion and sudden increases in water disturbance caused by instantaneous gear changes.

[0085] Forward-looking point tracking refers to selecting a forward-looking point in front of the current position of the bionic robotic fish along a predetermined trajectory, and using the direction from the current position of the bionic robotic fish to the forward-looking point as the guiding direction. Typically, by continuously updating the position of the forward-looking point and the guiding direction, the bionic robotic fish can move smoothly along the trajectory, reducing lateral swaying and sharp turns when approaching the trajectory.

[0086] S4.4. While the target fish group is in the same direction as the leading edge of the fish group, check the leading edge movement status in the same direction. When the leading edge movement status is established, use a small angle limit method to finely adjust the leading angle towards the gathering operation area and generate the leading edge traction table.

[0087] Furthermore, under the escorting trajectory, the positional changes of the outer edge segment of the target fish school's leading edge in the operational coordinate reference are continuously acquired, and the movement direction of the outer edge segment of the target fish school's leading edge is compared with the forward direction of the effective escorting trajectory and the direction of the gathering operational area. When the outer edge segment of the target fish school's leading edge moves towards the gathering operational area during continuous acquisition time, and the outer edge segment of the target fish school's leading edge does not show any outward divergence from the effective escorting trajectory, the leading edge's unidirectional movement state is confirmed. After the leading edge's unidirectional movement state is confirmed, the current leading angle of the bionic robotic fish is used as the starting angle, and the leading angle is finely adjusted towards the gathering operational area in stages according to the small angle limit method, so that each change in the leading angle is kept within the allowable range of the small angle level, and the stable leading angle is recorded when the leading edge of the target fish school continues to move in the same direction, generating a leading edge traction table.

[0088] like Figure 5 As shown, the curve corresponding to the method of the present invention gradually decreases and tends to stabilize after the leading edge access, indicating that under the combined effect of the leading edge access level, the virtual leading edge fish position, the effective escort trajectory, and the small angle lead-swim fine adjustment, the main movement direction of the target fish school gradually turns towards the direction of the gathering operation area; the fixed level direct access method shows deviation rebound in both the access stage and the traction stage, indicating that fixed intensity access is prone to causing orientation reversal; although the fixed path escort method reduces some sudden changes, there are still fluctuations, indicating that relying solely on the smooth path cannot replace the access adaptation processing of the fish school accepting fingerprints and the virtual leading edge fish position.

[0089] The fixed-position direct access method does not involve setting up test points or recognizing fish responses. The bionic robotic fish directly selects a fixed tail swing rhythm position, approaches the target fish school from the outer edge of the front edge, and leads and guides the fish along the direction of the gathering operation area. Throughout the access process, the bionic robotic fish does not determine whether the outer edge of the fish school is gathering or dispersing, nor does it determine whether the main movement direction of the fish school is reversed. The tail swing rhythm and access position remain unchanged.

[0090] The fixed-path escort method first generates a smooth escort path based on the current position of the bionic robotic fish, the leading edge position of the target fish group, and the direction of the aggregation operation area. Then, the bionic robotic fish approaches the leading edge of the target fish group along the smooth path and follows the outer edge of the fish group within a set escort distance. This method focuses on controlling the continuity of the path and the stability of the escort distance, but does not test different tail swing rhythms, does not form a fish group acceptance fingerprint, and does not delete sensitive outer edge segments that trigger outward dispersion.

[0091] S5. When the front edge traction table confirms that the target fish group has entered the gathering operation area, the bionic robotic fish swims slowly around the outer edge of the gathering operation area. When the target fish group disperses to the outer edge, it performs a low-intensity pullback to form a stable gathering state.

[0092] S5.1. Retrieve the stable leading angle from the leading edge traction table, track the positional change of the outer edge segment of the target fish school relative to the entrance of the gathering operation area, and generate the gathering entry state when the outer edge of the target fish school enters the inner side of the gathering operation area.

[0093] Furthermore, a stable leading angle is retrieved from the leading-edge traction table, and the bionic robotic fish is controlled to maintain traction along the leading edge of the target fish school at a stable leading angle. Within the operational coordinate reference, the positional changes of the outer edge segment of the target fish school's leading edge are continuously extracted, and the positional relationship between the outer edge segment of the target fish school's leading edge and the entrance of the gathering operation area is checked to confirm whether the outer edge segment of the target fish school's leading edge gradually moves from the outside of the gathering operation area entrance to the inside of the gathering operation area entrance. Then, the relationship between the outer body of the target fish school's outer edge and the boundary of the gathering operation area is checked along the outline of the target fish school's outer edge. When the outer body of the target fish school's outer edge is continuously located inside the gathering operation area, and the outer edge segment of the target fish school's leading edge does not show any change of retreating towards the outside of the gathering operation area, the gathering entry state is generated.

[0094] It should also be noted that the stable leading angle is obtained by continuously checking the same-direction movement of the outer edge of the target fish group during the accompanying track entry state, and by registering the leading angle currently maintained by the bionic robotic fish in the leading edge traction table when the outer edge of the target fish group continues to move towards the gathering operation area without any outward changes.

[0095] S5.2 When the aggregation entry state is established, perform equal arc length sampling and curvature smoothing on the outer edge boundary of the aggregation operation domain to generate a low-speed circular path, and control the bionic robotic fish to move along the low-speed circular path at the circular gear to generate the outer edge circular state.

[0096] Furthermore, after the aggregation state is established, the arc length distance between adjacent boundary points is calculated along the extension direction of the outer edge of the aggregation operation area, using the outer boundary of the aggregation operation area as the path reference. Boundary sampling points are then reselected at uniform arc length intervals to ensure that the boundary sampling points are evenly distributed on the outer edge of the aggregation operation area. Subsequently, the curvature of the local bending changes of the boundary sampling points is smoothed, and the positions of abrupt turning changes are corrected to continuous transition positions. The smoothed boundary sampling points are checked to ensure that they are still within the circumferential range between the inner side of the outer edge of the aggregation operation area and the outer side of the target fish group. The smoothed boundary sampling points are connected in the extension order of the outer edge of the aggregation operation area to generate a low-speed circumferential path. The bionic robotic fish is controlled to switch to the circumferential mode and move slowly along the low-speed circumferential path, so that the bionic robotic fish remains circumferential outside the outer edge of the target fish group, generating the outer edge circumferential state.

[0097] S5.3. Set up an outward dispersion discrimination zone in the outer edge swimming state, use the hysteresis judgment method to confirm the outward dispersion edge segment, and control the bionic robotic fish to deflect at a small angle towards the center of the aggregation operation area in the pull-back position. When the outward dispersion edge segment retracts to the inside of the aggregation operation area, a stable aggregation state is generated.

[0098] Furthermore, in the outer perimeter circling state, the bionic robotic fish expands outward along the outer boundary of the gathering operation area to form an outward dispersion discrimination zone. During the low-speed circling movement, the positional change of the outer perimeter of the target fish group relative to the outward dispersion discrimination zone is continuously tracked. A hysteresis judgment method is used to confirm the outer perimeter of the target fish group. When the outer perimeter of the target fish group continuously exceeds the outer range of the outward dispersion discrimination zone, the outward dispersion edge segment is confirmed. When the outer perimeter of the target fish group swings briefly but does not continuously exceed the outer range of the outward dispersion discrimination zone, the outward dispersion edge segment is not confirmed. After confirming the outward dispersion edge segment, the bionic robotic fish is controlled to switch from the circling gear to the pull-back gear and deflects at a small angle towards the center of the gathering operation area, so that the bionic robotic fish forms a low-intensity pull-back traction along the outer perimeter of the target fish group. When the outward dispersion edge segment continuously retracts to the inner side of the gathering operation area, and the outer perimeter of the target fish group does not exceed the outer range of the outward dispersion discrimination zone again, a stable gathering state is generated.

[0099] This embodiment also provides a computer device applicable to the fish school aggregation method based on biomimetic swimming signals, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the fish school aggregation method based on biomimetic swimming signals proposed in the above embodiment.

[0100] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0101] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the method for fish agglomeration based on biomimetic swimming signals as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0102] In summary, this invention achieves this by: deploying test points on the outer edge of the target fish swarm and outputting a biomimetic test sequence; identifying net convergence, net divergence, convergence of orientation, and divergence of orientation within the same test response window to form a fish swarm acceptance fingerprint; and then screening the leading edge access levels and eliminating avoidance test points. This allows the biomimetic robotic fish to determine the access strength and access range based on the actual response of the fish swarm, enhancing the adaptability of the leading edge access and improving the stability of the fish swarm transitioning from a natural swimming state to a controlled aggregation state.

[0103] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for fish school aggregation based on biomimetic swimming signals, characterized in that, include: Conduct fish school situation identification in the target waters, delineate the aggregation operation area along the distribution range of the target fish school, and form a fish school situation table; The outline of the target fish group recorded in the fish group situation table is smoothed. Test points are set up along the smoothed target fish group outer edge. The bionic robotic fish outputs a bionic test sequence at the test points in sequence. The changes in fish group gathering and fish group orientation before and after the test are compared in the same test response window. When the fish group gathering is enhanced and the fish group orientation tends to be similar, the acceptable level is registered. When the fish group disperses at the outer edge and the fish group orientation is reversed, the avoidance level is registered, thus forming a fish group acceptance fingerprint. The acceptable gear is screened out by the fingerprint of the fish swarm and the leading edge access gear is determined. The test points associated with the avoidance gear are eliminated to form the leading edge candidate range. The remaining test points are sorted in order of leading edge priority along the direction of the aggregation operation domain. When the remaining test points continuously show the leading edge in the same direction, the edge converges and the turning is gentle, the virtual leading edge fish position is marked and the virtual leading edge fish position is bound to the leading edge access gear to form the leading edge access task. A companion trajectory is generated around the virtual leading edge fish position in the leading edge access task. The bionic robotic fish enters the companion trajectory according to the leading edge access level. When the target fish group moves in the same direction at the leading edge, the leading angle is finely adjusted towards the gathering operation area to form a leading edge traction table. When the front edge traction table confirms that the target fish group has entered the gathering operation area, the bionic robotic fish swims slowly around the outer edge of the gathering operation area. When the target fish group disperses to the outer edge, it performs a low-intensity pullback to form a stable gathering state.

2. The method for fish school aggregation based on biomimetic swimming signals as described in claim 1, characterized in that, The fish swarm situation table is formed by establishing a work coordinate benchmark for the target water area, collecting water area perception information and identifying the area where the target fish swarm is located, determining the outer contour and movement direction of the target fish swarm, and then delineating the gathering operation area along the outer contour of the target fish swarm towards the gathering target location, avoiding unsuitable gathering locations, and organizing the target fish swarm situation and gathering operation area accordingly.

3. The method for fish school aggregation based on biomimetic swimming signals as described in claim 2, characterized in that, The outline of the target fish group recorded in the fish group situation table is smoothed, and test points are set up along the smoothed outer edge of the target fish group. The bionic robotic fish outputs a bionic test sequence at the test points in sequence. The specific steps are as follows: Extract the outer contour of the target fish school from the fish school situation table, resample the outer contour of the target fish school with equal arc length, and use Savitzky-Golay smoothing filter to reduce the outer edge spikes and generate a smooth outer contour. Along the smooth outer edge contour towards the gathering operation area, test positions are selected sequentially, and the test positions are kept outside the outer edge of the target fish group to generate a test point sequence. The bionic robotic fish arrives at the current test point according to the test point sequence, maintains a low-disturbance swimming state at the current test point, and uses the duration of the low-disturbance swimming state as the test reference window to obtain the natural outer edge changes of the target fish group within the test reference window. At the current test point, biomimetic swimming is performed sequentially according to the tail swing rhythm gears from low to high to form a biomimetic test sequence, and the duration from the start of output to the stop of output at the current gear is determined as the current test response window.

4. The method for fish school aggregation based on biomimetic swimming signals as described in claim 3, characterized in that, Within the same trial response window, the changes in fish school convergence and orientation before and after the trial are compared. When fish school convergence increases and orientation becomes more similar, an acceptable level is registered. When fish school disperses at the outer edge and orientation reverses, an avoidance level is registered, thus forming a fish school acceptance fingerprint. The specific steps are as follows: Within the current trial response window, the natural outer edge change of the target fish group in the trial reference window is used as the reference outer edge zone. The displacement direction of the target fish group's outer edge relative to the reference outer edge zone is extracted. When the target fish group's outer edge continuously enters towards the center of the group, a net convergence state is generated. When the target fish group's outer edge continuously moves away from the center of the group, a net divergence state is generated. Within the current trial response window, a same-direction discrimination sector is established based on the swimming direction of the bionic robotic fish. The main movement direction of the target fish group within the current trial response window is extracted. When the main movement direction stably falls into the same-direction discrimination sector, a direction convergence state is generated. When the main movement direction continues to deviate from the same-direction discrimination sector and changes in a direction away from the aggregation operation area, a direction reversal state is generated. When the net convergence state and the orientation convergence state occur simultaneously, an acceptable position is registered. When the net divergence state and the orientation reversal state occur simultaneously, an avoidance position is registered. The acceptable positions are continuously verified along the test point sequence to form the fish swarm acceptance fingerprint.

5. The method for fish school aggregation based on biomimetic swimming signals as described in claim 4, characterized in that, The process of identifying acceptable gear positions by receiving fingerprints from fish swarms and determining leading edge access gear positions, and eliminating trial points associated with avoided gear positions to form a leading edge candidate range, involves the following specific steps: The acceptable levels that have passed continuous verification are extracted from the fingerprints received by the fish swarm, and the perturbation intensity is checked from low to high according to the tail swing rhythm. The acceptable level that meets the continuous acceptance condition and has the lowest tail swing rhythm is determined as the leading edge access level. Based on the fingerprint positioning registration of the fish swarm, a buffer segment is set with the outer edge segment where the avoidance test point is located as the center, and a leading edge elimination segment is generated. The leading edge elimination segment is deleted from the smooth outer edge contour, and the test points that can correspond to the leading edge access position are retained to generate the leading edge candidate range.

6. The method for fish school aggregation based on biomimetic swimming signals as described in claim 1, characterized in that, The specific steps for forming the leading edge access task are as follows: Project the retained test points within the candidate range of the leading edge onto the smooth outer edge contour, and take the end closest to the entrance of the clustering operation domain as the sorting starting point, arrange the retained test points according to the arc length direction to generate the leading edge sequence. Check the movement direction of the outer edge segment corresponding to the retained test point along the leading edge sequence, and generate a candidate segment in the same direction when the movement direction continuously points to the clustering operation area. Check the retraction state of the outer edge segment corresponding to the retained test point within the candidate segment in the same direction, and generate a candidate segment for merging when the outer edge segment continues to enter the center of the group. Three-point sliding verification is performed on the reserved test points within the candidate segment. When the outer edge segment transitions continuously without sharp bends, a virtual leading edge calibration segment is generated. Then, a safe approach distance is offset to the outside of the target fish group along the outer normal direction of the middle outer edge segment of the virtual leading edge calibration segment to mark the virtual leading edge fish position. The virtual leading edge fish position is then bound to the leading edge access position to form a leading edge access task.

7. The method for fish school aggregation based on biomimetic swimming signals as described in claim 6, characterized in that, The process involves generating a companion trajectory around the virtual leading edge fish position in the leading edge access task. The bionic robotic fish enters the companion trajectory according to the leading edge access level. When the target fish school moves in the same direction at the leading edge, it finely adjusts its leading angle towards the gathering operation area to form a leading edge traction table. The specific steps are as follows: The virtual leading edge fish position is extracted from the leading edge access task. The trajectory starting point is determined by the current position of the bionic robotic fish. A tangential approach point is set outside the virtual leading edge fish position. A cubic B-spline is used to smoothly fit the trajectory starting point, the tangential approach point and the virtual leading edge fish position to generate a continuous turning companion trajectory. The leading edge of the escort trajectory is corrected, the distance relationship between the escort trajectory and the outer edge of the target fish group is checked, and the trajectory segments that deviate from the safe escort distance zone are pulled back along the outer edge normal direction to generate an effective escort trajectory. After the bionic robotic fish enters the effective companion trajectory, it switches to the leading edge access position by ramp transition and moves along the effective companion trajectory by forward viewpoint tracking to generate the companion entry state. While the target fish are in the same direction as the leading edge of the school, check the state of movement of the leading edge. When the state of movement of the leading edge is established, use a small angle limit method to finely adjust the leading angle towards the gathering operation area and generate the leading edge traction table.

8. The method for fish school aggregation based on biomimetic swimming signals as described in claim 1, characterized in that, The specific steps for forming a stable aggregated state are as follows: The stable leading angle is retrieved from the leading edge traction table to track the position change of the outer edge segment of the target fish group relative to the entrance of the gathering operation area, and a gathering entry state is generated when the main body of the outer edge of the target fish group enters the inner side of the gathering operation area. When the aggregation state is established, the outer edge boundary of the aggregation operation area is sampled with equal arc length and smoothed with curvature to generate a low-speed circular path, and the bionic robotic fish is controlled to move along the low-speed circular path at the circular gear to generate the outer edge circular state. In the outer edge swimming state, an outward dispersion discrimination zone is set up, and the outward dispersion edge segment is confirmed by the hysteresis judgment method. The bionic robotic fish is controlled to deflect at a small angle towards the center of the aggregation operation area in the pull-back position. When the outward dispersion edge segment retracts to the inside of the aggregation operation area, a stable aggregation state is generated.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the fish school aggregation method based on biomimetic swimming signals as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the fish agglomeration method based on biomimetic swimming signals as described in any one of claims 1 to 8.

Images