Construction method of multi-underwater-robot cooperative control system

Abstract

The invention discloses a construction method of a multi-underwater robot cooperative control system, which relates to the field of underwater robots, and establishes a multi-parameter coupled data acquisition and preprocessing mechanism by using a Doppler current profiler ADCP, an inertial measurement unit IMU and controller software. A combined algorithm of sliding median filtering and Kalman filtering is adopted, and data disturbance caused by environment noise and attitude drift is effectively removed; and through abnormal point triple standard deviation determination and double adjacent point interpolation completion, stable correction of sampling data is realized. A standardized data set obtained after normalization processing unifies parameter scales of different physical magnitudes, so that the local average flow velocity, the flow velocity fluctuation root mean square and the average acceleration can be directly compared in a feature space. Therefore, according to the method, the high-consistency expression of the water flow disturbance characteristics in time and space is realized, so that the synchronous measurement precision of multiple underwater robots in a non-uniform flow field is remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of underwater robots, in particular to a construction method of a multi-underwater robot cooperative control system. BACKGROUND

[0002] With the rapid development of intelligent robots, multi-underwater robot cooperative operation has become an important development direction of underwater environment monitoring and underwater detection. Especially in the river-lake inlet, nearshore shallow sea and tidal mixing zone, the multi-underwater robot cooperative observation system can realize multi-point synchronous measurement of water flow velocity, flow direction and disturbance characteristics by constructing a three-dimensional space sampling network. However, in such a non-uniform flow field environment, the flow velocity gradient is significant, and the turbulent stratification is obvious. The traditional cooperative control method relying on acoustic positioning or fixed trajectory planning is difficult to maintain the synchronization and spatial consistency among robots.

[0003] At present, in the field of multi-underwater robot environmental observation and data sampling, the traditional cooperative control method usually relies on acoustic positioning systems such as LBL long baseline or USBL ultra-short baseline system to realize the spatial synchronization among multiple robots. However, the acoustic signal is easily affected by turbulent flow disturbance, bottom reflection and temperature-salinity stratification in complex hydrodynamic environment, resulting in large communication delay, positioning error accumulation and high energy consumption. At the same time, the traditional control strategy adopts fixed sampling grid and static path planning, which cannot dynamically adapt to the real-time changes of flow field disturbance characteristics, causing problems such as insufficient sampling accuracy in high disturbance area and redundant sampling in low disturbance area. Especially in the river mouth, nearshore or inland reservoir environment, the microflow layer structure and disturbance scale of the water body change quickly, and the existing system is difficult to maintain the microflow synchronization among multiple robots, resulting in spatio-temporal misalignment and feature drift of the sampling data.

[0004] The main reason for the above situation is that the existing multi-underwater robot system lacks quantitative modeling and synchronous feedback mechanism of "microflow layer disturbance characteristics", and still uses macro flow velocity or fixed trajectory as the basis for cooperation, which cannot reflect the asynchrony of local disturbance in time and space. Once there is a local flow field mutation or disturbance gradient enhancement, the sampling phase among robots will be offset, resulting in accumulation of data fusion error. At the same time, due to the inability to adaptively adjust the path and sampling density, the sampling energy consumption is unevenly distributed, information is lost in high disturbance area and redundant sampling in low disturbance area, which ultimately leads to a decrease in overall cooperative accuracy, system response lag, and even abnormal phenomena such as communication blockage or control drift, which cannot meet the real-time and accuracy requirements of underwater cooperative observation. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a construction method of a multi-underwater robot cooperative control system, which solves the problems mentioned in the background art.

[0006] To achieve the above object, the application is implemented by the following technical solutions: A construction method of a multi-underwater robot cooperative control system, comprising the following steps: S1, in the operation area of the multiple underwater robots, the position flow environment data is collected by the multiple underwater robots respectively and transmitted to the central controller, the flow environment data is statistically processed to obtain local disturbance data, and preprocessing is performed to obtain a standardized data set; S2, based on the standardized data set, the disturbance fingerprint similarity score Sfp of each underwater robot is calculated, and microflow synchronization evaluation is performed; S3, based on the microflow synchronization evaluation result, a sampling grid adaptive reconstruction mechanism is triggered, the sampling grid adaptive reconstruction mechanism determines the fluid disturbance heterogeneity index Gdist of each sampling grid unit according to the spatial distribution of the disturbance fingerprint similarity score Sfp, and calculates the sampling density adjustment coefficient Kgrid accordingly; S4, the central controller divides the disturbance region type and dynamically adjusts the sampling task configuration according to the sampling density adjustment coefficient Kgrid, and then reassigns the underwater robot role according to the divided disturbance region type.

[0007] Preferably, the S1 comprises S11; S11, the operation area is the river-lake mouth area with significant velocity gradient and obvious turbulent stratification, and the multiple underwater robots are arranged to form a three-dimensional monitoring array according to the initial set depth layer and lateral spacing; Each underwater robot determines the fluid flow velocity at the position where the underwater robot is located in real time by the Doppler current profiler ADCP, the inertial measurement unit IMU and the controller software integrated on the body, and obtains the position flow environment data. The position flow environment data includes local flow velocity instantaneous value, acceleration value and flow velocity fluctuation value.

[0008] Preferably, the S1 further comprises S12; S12, the bandwidth underwater acoustic communication module is used for wireless data transmission connection between the multiple underwater robots and the central controller, and the collected position flow environment data is transmitted to the central controller; The bandwidth underwater acoustic communication module comprises a communication time slot number setting unit, an underwater acoustic signal modulation and demodulation unit, and a time synchronization calibration unit. In the preset communication period of 15-30 seconds, each underwater robot locally caches and compressively encodes the position flow environment data obtained in the current sampling window, and sends the data in the bandwidth underwater acoustic communication module through the communication time slot number setting unit according to the communication time slot number allocated by itself. The communication time slot number is distributed by the central controller in the task starting stage by using an initialization broadcast distribution, the time-sharing structure of the initialization broadcast distribution is in the form of a time division multiple access (TDMA) protocol, the duration of a single time slot is 1.5-2 seconds, and the total communication period includes a complete time slot sequence of 2 times the number of online underwater robots; The underwater acoustic signal modulation and demodulation unit performs narrowband transmission at a carrier frequency of 9-15 kHz by using a phase shift keying (PSK) technology, the data bandwidth is 3-5 kbit / s, and the effective communication distance is not less than 1000 meters. The time synchronization calibration unit preposes a 200 ms protection interval before the sending period of each underwater robot and embeds a 32-bit synchronization header code in the communication time slot number.

[0009] Preferably, the S1 further includes S13. After receiving the position flow environment data uploaded by each underwater robot, the central controller pre-processes the position flow environment data by using a data pre-processing module of the central controller to obtain local disturbance data; the pre-processing includes data denoising and abnormal point elimination operations. The data denoising removes the noise of all parameters in the position flow environment data by using a combination of sliding median filtering and Kalman filtering, wherein the sliding median filtering removes the noise of local flow velocity instantaneous values and flow velocity fluctuation values, and the Kalman filtering removes the noise of acceleration values of three-axis acceleration signals output by an inertial measurement unit (IMU). The abnormal point elimination operation eliminates the abnormal values in the position flow environment data by subtracting the corresponding mean value from all parameters in the position flow environment data after data denoising, and marking and eliminating the output result whose absolute value is greater than 3 times the standard deviation, and then interpolating and completing the data by using a double-neighbor-point mean value, and finally integrating the pre-processed position flow environment data to obtain the local disturbance data. The pre-processed local disturbance data is statistically calculated to obtain the local average flow velocity Vmean of the i-th underwater robot corresponding to each underwater robot, i the local flow velocity fluctuation root mean square Urms of the i-th underwater robot, i and the local average acceleration Amean of the i-th underwater robot. i wherein: The local average flow velocity Vmean is the arithmetic mean value of all local flow velocity instantaneous values collected by the corresponding underwater robot. The local flow velocity fluctuation root mean square Urms is the root mean square value of the mean value of all flow velocity fluctuation values collected by the corresponding underwater robot. The local average acceleration Amean is the average value of all acceleration values collected by the corresponding underwater robot. After the central controller completes the calculation, it normalizes the three types of differences of local average flow velocity Vmean, local flow velocity fluctuation root mean square Urms, and local average acceleration Amean for all underwater robots, performs disturbance feature normalization calculation to eliminate differences between different physical quantities, and obtains a standardized dataset. The standardized dataset includes the normalized scaling parameter ΔV of the local average flow velocity of the i-th underwater robot. i The normalized scaling parameter ΔU of the root mean square of the local velocity fluctuation of the i-th underwater robot. i The normalized scaling parameter ΔA of the local average acceleration of the i-th underwater robot i .

[0010] Preferably, S2 includes S21; S21. Based on the standardized dataset, calculate and output the perturbation fingerprint similarity score Sfp for each underwater robot; the perturbation fingerprint similarity score Sfp is calculated and output using the following algorithm formula; Sfp i =exp(-[(△V i ) 2 +(△U i ) 2 +(△A i ) 2 In the formula, exp represents the exponential function, Sfp i This represents the perturbation fingerprint similarity score of the i-th underwater robot.

[0011] Preferably, S2 further includes S22; S22. Based on the perturbation fingerprint similarity scores Sfp of all underwater robots, calculate the mean to obtain the average value Sfp of the perturbation fingerprint similarity scores. Avg Then, the synchronization stability index Isync is calculated. Set the synchronization stability threshold Isync in the central controller. th Then, the microfluidic synchronization is evaluated with the real-time acquired synchronization stability index Isync to determine the spatial differences in fluid disturbance characteristics among the current multiple underwater robots, and triggering is performed based on the microfluidic synchronization evaluation results; The synchronization stability index Isync is calculated and output using the following algorithm formula; In the formula, N represents the number of underwater robots currently participating in the collaboration; The specific evaluation content of microfluidic synchronization is as follows: When the synchronization stability index Isync < the synchronization stability threshold Isync thIf it is determined that the current group of multiple underwater robots is in a stable and consistent state in the microfluidic disturbance feature space, then the current sampling grid structure and path planning are maintained without adaptive adjustment. When the synchronization stability index Isync ≥ the synchronization stability threshold Isync th When it is determined that there are significant spatial differences in the fluid disturbance characteristics of the current underwater robots, the control system automatically triggers the sampling grid adaptive reconstruction mechanism.

[0012] Preferably, S3 includes S31; S31. After triggering the adaptive reconstruction mechanism of the sampling grid, the central controller determines the fluid disturbance heterogeneity index Gdist of the sampling grid unit based on the spatial distribution of the disturbance fingerprint similarity score Sfp. The specific analysis steps include: S311, S312 and S313. S311. Divide the work area into 20×15 sampling grid units, each sampling grid unit having an area of ​​5m×5m, with an overall coverage area of ​​500m×375m underwater work area; when the work area is larger than the overall coverage area, expand proportionally to 40×30 sampling grid units, maintaining a spatial resolution of 25m for each sampling grid unit. 2 And determine the sampling grid unit to which each sampling point belongs based on the real-time position of the underwater robot; S312. In each sampling grid cell, collect all perturbation fingerprint similarity scores Sfp of the current sampling grid cell, and calculate the mean and standard deviation to obtain the mean and standard deviation of the perturbation fingerprint similarity score Sfp of the j-th sampling grid cell, respectively. Then, traverse all sampling raster cells, and calculate the ratio between the mean of the perturbation fingerprint similarity score Sfp of the j-th sampling raster cell and the standard deviation of the perturbation fingerprint similarity score Sfp of the j-th sampling raster cell to obtain the normalized heterogeneity factor N of each sampling raster cell. S313. Based on the normalized heterogeneity factor N of the current sampling grid cell and the maximum and minimum values ​​among all sampling grid cells, perform linear normalization processing to obtain the fluid disturbance heterogeneity index Gdist of the corresponding sampling grid cell.

[0013] Preferably, S3 further includes S32; S32. The central controller determines the sampling density adjustment coefficient Kgrid for each sampling grid cell based on the obtained fluid disturbance heterogeneity index Gdist. The sampling density adjustment coefficient Kgrid is calculated and output using the following algorithm formula: Kgrid j =1+Cg·Cdist jIn the formula, Kgrid j Cdist represents the sampling density adjustment coefficient of the j-th sampling raster cell. j The expression represents the fluid disturbance heterogeneity index of the j-th sampling grid cell, and Cg represents the sampling density gain coefficient, with a value range of 0.5-1.0.

[0014] Preferably, S4 includes S41; S41. The central controller receives the sampling density adjustment coefficient Kgrid reported by all underwater robots for each sampling grid unit; then it calculates the mean and standard deviation of all sampling density adjustment coefficients Kgrid to obtain the mean KgridJ and the standard deviation KgridB of the sampling density adjustment coefficients Kgrid. The current sampling density adjustment coefficient Kgrid is compared and evaluated with the mean KgridJ and the standard deviation KgridB of the sampling density adjustment coefficient Kgrid. Based on the comparison and evaluation results, the type of disturbance region is determined, and the sampling task configuration is dynamically adjusted based on the determination results. The sampling task configuration includes sampling frequency adjustment, sampling interval adjustment, and robot adjustment. The specific comparative evaluation content is as follows: When the sampling density adjustment coefficient Kgrid≥KgridJ+0.5·KgridB, it is determined to be a first-level disturbance region. At this time, the sampling frequency of the underwater robot in the current sampling grid cell is increased by 40%, and the sampling interval is reduced by 20%. When KgridJ-0.5·KgridB < sampling density adjustment coefficient Kgrid < KgridJ+0.5·KgridB, it is determined to be a secondary disturbance region. At this time, there is no need to configure the sampling task and the original sampling task is maintained. When the sampling density adjustment coefficient Kgrid≤KgridJ-0.5·KgridB, it is determined to be a level three disturbance region. At this time, the sampling frequency of the underwater robot in the current sampling grid cell is reduced by 40%, and the sampling interval is increased by 20%.

[0015] Preferably, S4 further includes S42; S42. After the sampling task configuration is updated, the central controller dynamically assigns roles to the underwater robot based on the disturbance region type of each sampling grid unit; the specific assignment details are as follows: The robot roles include backbone underwater robots, encrypted sampling underwater robots, and standby and self-testing underwater robots; the initial role of the underwater robots is backbone underwater robots. When a Level 1 disturbance area is identified, the underwater robot role of the current sampling grid unit is adjusted to an encrypted sampling underwater robot according to the sampling task configuration. If the sampling load of the encrypted sampling underwater robot increases by 30% after the adjustment, the standby and self-testing underwater robots of the adjacent sampling grid unit are called to enter the current sampling grid unit for collaborative support. The original encrypted sampling underwater robot does not withdraw. If the sampling load is still greater than or equal to 30% after collaborative support, the backbone underwater robot of the adjacent sampling grid unit is called for secondary collaborative support. When the area is identified as a Level 2 disturbance zone, the backbone underwater robot is maintained and serves as a backup support for Level 1 disturbance zones. When the area is identified as a Level 3 disturbance zone, the underwater robot role of the current sampling grid unit is switched to standby and self-testing underwater robot.

[0016] This invention provides a method for constructing a cooperative control system for multiple underwater robots. It offers the following advantages: (1) This method establishes a multi-parameter coupled data acquisition and preprocessing mechanism by utilizing a Doppler current profiler (ADCP), an inertial measurement unit (IMU), and controller software. A combined algorithm of sliding median filtering and Kalman filtering is employed to effectively remove data disturbances caused by environmental noise and attitude drift. Robust correction of the sampled data is achieved through outlier detection using three times the standard deviation and interpolation between neighboring points. The standardized dataset obtained after normalization unifies the parameter scales of different physical quantities, allowing direct comparison of local average flow velocity, root mean square of flow velocity fluctuations, and average acceleration in the feature space. Therefore, this method achieves a highly consistent representation of water flow disturbance characteristics in both time and space, significantly improving the synchronous measurement accuracy of multiple underwater robots in non-uniform flow fields and providing a stable data foundation for subsequent disturbance fingerprint analysis and path control.

[0017] (2) This method introduces a linkage mechanism between the "perturbation fingerprint similarity score Sfp" model and the "fluid perturbation heterogeneity index Gdist - sampling density adjustment coefficient Kgrid". This mechanism uses the Gaussian similarity function and Euclidean distance to calculate the perturbation similarity between underwater robots, and uses this to evaluate the group synchronization stability index Isync; when Isync exceeds the threshold Isyncth, the central controller automatically triggers the sampling grid adaptive reconstruction process. This process dynamically divides the high perturbation region, stable region and low perturbation region according to the spatial distribution characteristics of the perturbation fingerprint similarity score Sfp, and adjusts the sampling frequency, spacing and robot allocation ratio by calculating Kgrid, so as to realize the real-time adjustment of sampling density according to the complexity of fluid perturbation. This mechanism can significantly improve the system's response capability to heterogeneous regions of the flow field, so as to achieve the optimal balance between spatiotemporal resolution and energy consumption efficiency in the sampling task.

[0018] (3) This method proposes a dynamic allocation strategy for underwater robot roles based on the type of disturbed area. The central controller determines the disturbance level of each grid according to the statistical distribution results of the sampling density adjustment coefficient Kgrid, and dynamically adjusts the sampling frequency, sampling interval and number of robots accordingly. At the role allocation level, the system classifies robots into three identities: backbone underwater robots, encrypted sampling underwater robots, and standby and self-testing underwater robots, and introduces a collaborative support mechanism of "standby priority, backbone backup": when the load in a high-disturbance area increases by more than 30%, standby and self-testing underwater robots in adjacent areas are prioritized to participate in collaborative sampling; backbone robots are only called for short-term support when standby resources are insufficient. Through this hierarchical and dynamic scheduling structure, rapid reinforcement and regional self-balancing of sampling tasks can be achieved without destroying the communication backbone, thereby improving the overall system's task completion rate, communication stability and energy efficiency. Attached Figure Description

[0019] Figure 1 This is a schematic diagram illustrating the construction steps of a multi-underwater robot collaborative control system according to the present invention.

[0020] Figure 2 This is a schematic diagram of adaptive reconstruction of the sampling grid. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Example 1 This invention provides a method for constructing a collaborative control system for multiple underwater robots. Please refer to [link / reference]. Figure 1 This includes the following steps: S1. In the operating area of ​​multiple underwater robots, multiple underwater robots collect location water flow environment data and transmit it to the central controller. The water flow environment data is statistically processed to obtain local disturbance data, and preprocessing is performed to obtain a standardized dataset. S2. Based on the standardized dataset, calculate the perturbation fingerprint similarity score Sfp for each underwater robot and evaluate the microfluidic synchronization. S3. Based on the microfluidic synchronization evaluation results, the sampling grid adaptive reconstruction mechanism is triggered. The sampling grid adaptive reconstruction mechanism determines the fluid perturbation heterogeneity index Gdist of each sampling grid unit according to the spatial distribution of the perturbation fingerprint similarity score Sfp, and calculates the sampling density adjustment coefficient Kgrid accordingly. S4. The central controller divides the disturbance area types and dynamically adjusts the sampling task configuration according to the sampling density adjustment coefficient Kgrid, and then reassigns the underwater robot roles according to the divided disturbance area types.

[0023] In this embodiment, the method uses a Doppler current profiler (ADCP) and an inertial measurement unit (IMU) to acquire multiple parameters of local flow velocity and acceleration, reflecting transient disturbance behavior in the microfluidic layer in real time. If the original signal is used directly, it will be affected by bubble interference and attitude drift, resulting in abnormal velocity peaks and pseudo-acceleration responses. Therefore, this implementation introduces a joint filtering and outlier removal mechanism to ensure the acquired local disturbance data has temporal continuity and spatial consistency, guaranteeing stable input for subsequent feature calculations. In S2, the disturbance fingerprint similarity score Sfp is calculated to quantitatively compare microfluidic disturbances among robots. Its physical meaning is equivalent to calculating the "fluid similarity distance" between different sampling points in the disturbance feature space. If a robot drifts due to local vortices in a high-gradient flow field, the disturbance fingerprint similarity score Sfp will significantly decrease, triggering the system's real-time evaluation of the synchronization stability index Isync. This design enables self-synchronization control without GPS, ensuring temporal phase consistency in group sampling. In S3, when the synchronization stability index Isync exceeds a preset threshold, the central controller calculates the fluid disturbance heterogeneity index Gdist and the sampling density adjustment coefficient Kgrid based on the disturbance distribution, enabling the sampling grid to adaptively adjust in real time according to the disturbance strength. Without this adjustment, detail loss would occur in high-disturbance areas due to sparse sampling, while low-disturbance areas would experience increased energy consumption due to repeated sampling. By introducing the Kgrid adaptive mechanism, the system can automatically increase the sampling density in high-disturbance areas and reduce redundant sampling in low-disturbance areas, thereby achieving a dynamic balance between spatial coverage and energy efficiency. In S4, the central controller classifies disturbance areas according to Kgrid and reassigns robot roles: high-disturbance areas are handled by the encrypted sampling underwater robot, stable areas maintain continuous communication and support for the backbone robot, and low-disturbance areas are operated energy-efficiently by standby and self-checking robots. This role-based approach avoids the problem of rigid resource allocation. For example, if all robots operate with the same sampling period, it will lead to communication bandwidth congestion and data redundancy. Through this hierarchical mechanism, the sampling accuracy in high-disturbance areas is improved by about 40%, the overall energy utilization rate is improved by about 20%, and the system communication latency is reduced to 80% of the original level. In summary, this implementation method, by combining disturbance identification, adaptive reconstruction, and role scheduling, achieves group self-coordination and sampling self-equilibrium of multiple underwater robots in dynamic hydrodynamic environments, which has clear physical significance and significant engineering effectiveness.

[0024] Example 2 Please see Figure 1 Specifically: S1 includes S11; S11. The work area is the river mouth area with significant velocity gradient and obvious turbulent stratification. Multiple underwater robots are arranged to form a three-dimensional monitoring array according to the initially set depth layers and lateral spacing. Each underwater robot uses an ADCP (Advanced Doppler Current Profiler), an IMU (Inertial Measurement Unit), and controller software integrated on its body to measure the fluid velocity at its location in real time and obtain location water flow environment data. Locational water flow environment data includes local instantaneous velocity values, acceleration values, and velocity fluctuation values; The instantaneous local velocity value is acquired by a Doppler current profiler (ADCP), specifically representing the instantaneous local velocity value measured at the k-th sampling time along the main direction of the water flow at the location of the i-th underwater robot in the water body. Acceleration values ​​are acquired via an inertial measurement unit (IMU). The flow velocity fluctuation value is obtained by performing root mean square processing on the average value of the instantaneous local flow velocity of the i-th underwater robot through the controller software.

[0025] S1 also includes S12; S12. Multiple underwater robots and the central controller are connected wirelessly via a bandwidth underwater acoustic communication module to transmit the collected location and water flow environment data to the central controller. The bandwidth underwater acoustic communication module includes a communication time slot numbering setting unit, an underwater acoustic signal modulation and demodulation unit, and a time synchronization calibration unit. Each underwater robot locally caches and compresses the location water flow environment data acquired in the current sampling window within a preset communication period of 15 to 30 seconds, and transmits the data in the time-division underwater acoustic channel of the bandwidth underwater acoustic communication module according to its assigned communication time slot number through the communication time slot number setting unit. The communication time slot number is assigned by the central controller during the task startup phase using an initial broadcast. The time-division structure of the initial broadcast assignment adopts the Time Division Multiple Access (TDMA) protocol. The duration of a single time slot is 1.5 to 2 seconds, and the total communication cycle includes a complete time slot sequence that is twice the number of online underwater robots in seconds. The underwater acoustic signal modulation and demodulation unit uses phase shift keying (PS) technology to perform narrowband transmission at a carrier frequency of 9kHz-15kHz, with a data bandwidth of 3-5kbit / s and an effective communication distance of not less than 1000 meters. To prevent signal interference between multiple robots, the time synchronization calibration unit adds a 200-millisecond protection interval before the transmission time of each underwater robot and embeds a 32-bit synchronization header code in the communication time slot number for the central controller to perform data packet start detection.

[0026] S1 also includes S13; S13. After receiving the location and water flow environment data uploaded by each underwater robot, the central controller preprocesses the location and water flow environment data through the central controller's data preprocessing module to obtain local disturbance data; the preprocessing includes data denoising and outlier removal operations. Data denoising is achieved by using a combination of sliding median filtering and Kalman filtering to remove noise from all parameters in the location flow environment data. Sliding median filtering denoises the instantaneous and fluctuating values ​​of local flow velocities, while Kalman filtering denoises the acceleration values ​​of the triaxial acceleration signals output by the inertial measurement unit (IMU). The outlier removal operation is based on the mean and standard deviation of all parameters in the denoised location flow environment data. The mean of each parameter in the flow environment data is subtracted from the mean. If the absolute value of the output result is greater than 3 times the standard deviation, it is marked as an outlier and removed. Then, the mean of two neighboring points is used for interpolation to complete the data. Finally, the preprocessed location flow environment data is integrated to obtain local disturbance data. Statistical calculations were performed on the preprocessed local disturbance data to obtain the local average flow velocity Vmean for the i-th underwater robot for each underwater robot. i The root mean square (Urms) of local flow velocity fluctuations of the i-th underwater robot i The local average acceleration Amean of the i-th underwater robot i ,in: The local average velocity Vmean is the arithmetic mean of all the instantaneous local velocities collected by the corresponding underwater robot; The root mean square (Urms) of local velocity fluctuations is the root mean square value of the average of all velocity fluctuation values ​​collected by the corresponding underwater robot. The local average acceleration Amean is the average of all acceleration values ​​collected by the corresponding underwater robot; After the central controller completes the calculation, it normalizes the three types of differences of local average flow velocity Vmean, local flow velocity fluctuation root mean square Urms, and local average acceleration Amean for all underwater robots, performs disturbance feature normalization calculation to eliminate differences between different physical quantities, and obtains a standardized dataset. The standardized dataset includes the normalized scaling parameter ΔV of the local average flow velocity for the i-th underwater robot. i The normalized scaling parameter ΔU of the root mean square of the local velocity fluctuation of the i-th underwater robot. i The normalized scaling parameter ΔA of the local average acceleration of the i-th underwater robot i .

[0027] In this embodiment, the method selects the river inlet region, characterized by significant velocity gradients and distinct turbulent stratification, as the operating area. Multiple underwater robots are arranged according to depth layers and lateral spacing to construct a three-dimensional monitoring array. This array effectively avoids the problem of mixed sampling of flow characteristics between different layers. Because the velocity differences and shearing are significant in the inlet region, stratified sampling can easily lead to distorted disturbance signals, reducing the comparability of subsequent calculations. The instantaneous local velocity values ​​are measured using a Doppler current profiler (ADCP), and acceleration values ​​are measured using an inertial measurement unit (IMU). Combined with the controller software, the velocity fluctuation values ​​are calculated, capturing the transient disturbance characteristics of the water body in real time and achieving a comprehensive characterization of mainstream energy, local turbulence intensity, and inertial response. Then, in S12, multiple underwater robots transmit data to the central controller via a bandwidth underwater acoustic communication module. A TDMA time-division multiplexing structure is used for communication scheduling, with time slots of 1.5 to 2 seconds and a guard interval of 200 milliseconds. This avoids signal collisions and errors in shallow water multipath environments, as simultaneous signal transmission by multiple robots can cause data frame overlap, leading to communication delays and false alarms. By introducing a 32-bit synchronization header, precise packet boundary detection is achieved, enabling the central controller to ensure time synchronization and data integrity within a kilometer-scale communication range, physically eliminating the risk of communication drift between multiple robots. In S13, the central controller performs joint filtering and anomaly removal on the acquired raw signals through a data preprocessing module. Sliding median filtering removes instantaneous spike noise, and Kalman filtering dynamically smooths the IMU output signal; the combination of these two methods suppresses high-frequency noise while preserving effective disturbance trends. Furthermore, statistical removal based on three standard deviations automatically identifies non-physical anomalies and uses interpolation with the mean of two neighboring points to ensure the continuity of time-series data. Normalization maps the local average velocity Vmean, root mean square velocity fluctuation Urms, and average acceleration Amean to a standardized dataset of the same scale, making different physical quantities comparable and avoiding the dominant effect of dimensional differences on subsequent disturbance feature analysis. In summary, through the collaborative design of hierarchical sampling, time-division communication, and filtering normalization, this implementation successfully achieved high-precision disturbance capture and synchronous data acquisition in underwater environments with complex flow fields and unstable channels. This design effectively solves the common problems of sampling distortion, communication conflicts, and data drift in traditional systems, enabling multiple underwater robot groups to achieve physical time synchronization, data consistency, and disturbance feature alignment. This lays a highly reliable data foundation for subsequent disturbance fingerprint similarity scoring and adaptive control, significantly improving the overall measurement accuracy and robustness of the system.

[0028] Example 3 Please see Figure 1 Specifically: S2 includes S21; S21. Based on the standardized dataset, calculate and output the perturbation fingerprint similarity score Sfp for each underwater robot; the perturbation fingerprint similarity score Sfp is calculated and output using the following algorithm formula; Sfp i =exp(-[(△V i ) 2 +(△U i ) 2 +(△A i ) 2 In the formula, exp represents the exponential function, Sfp i This represents the perturbation fingerprint similarity score of the i-th underwater robot; The formula originates from the Euclidean distance metric theory in mathematics and the Gaussian similarity function in probability and statistics. In a physical sense, the model is equivalent to calculating the distance similarity between each underwater robot and the reference robot in the "perturbation feature space", and normalizing the distance using an exponential function so that the similarity value is in the interval (0,1]. This formula maps three key parameters of water disturbance (flow velocity, flow velocity fluctuation, and acceleration) to a three-dimensional disturbance feature space. Assuming a standardized dataset with normalized differences in each dimension, the disturbance feature distance is defined as . Substituting this into the Gaussian similarity function, this formula constitutes the core expression for the disturbance fingerprint similarity score. The output of the disturbance fingerprint similarity score Sfp is a dimensionless parameter with a value range of 0-1. Among them, when the perturbation fingerprint similarity score Sfp is close to 1, it means that the underwater robot is in the same microfluidic perturbation environment as the reference underwater robot; When the perturbation fingerprint similarity score Sfp is significantly lower than 0.6, it indicates that the fluid perturbation characteristics of the underwater robot are significantly different from those of the reference point.

[0029] S2 also includes S22; S22. Based on the perturbation fingerprint similarity scores Sfp of all underwater robots, calculate the mean to obtain the average value Sfp of the perturbation fingerprint similarity scores. Avg Then, the synchronization stability index Isync is calculated. Set the synchronization stability threshold Isync in the central controller. th Then, the microfluidic synchronization is evaluated with the real-time acquired synchronization stability index Isync to determine the spatial differences in fluid disturbance characteristics among the current multiple underwater robots, and triggering is performed based on the microfluidic synchronization evaluation results; The synchronization stability index Isync is calculated and output using the following algorithm formula; In the formula, N represents the number of underwater robots currently participating in the collaboration; The specific evaluation content of microfluidic synchronization is as follows: When the synchronization stability index Isync < the synchronization stability threshold Isync th If it is determined that the current group of multiple underwater robots is in a stable and consistent state in the microfluidic disturbance feature space, then the current sampling grid structure and path planning are maintained without adaptive adjustment. When the synchronization stability index Isync ≥ the synchronization stability threshold Isync th When it is determined that there are significant spatial differences in the fluid disturbance characteristics of the current underwater robots, the control system automatically triggers the sampling grid adaptive reconstruction mechanism.

[0030] In this embodiment, the perturbation fingerprint similarity score Sfp is calculated based on a standardized dataset, using the Gaussian similarity function exp[-((△Vi)]. 2 +(△U i ) 2 +(△Ai) 2 The distance of the three-dimensional perturbation features is normalized and mapped. The core significance of this design is that in actual underwater operating environments, the physical quantities such as flow velocity and acceleration measured by robots at different depths and positions vary greatly. If the linear distance is directly compared, it will lead to high-dimensional features dominating the overall similarity, causing the evaluation results to be distorted. By compressing and mapping the Euclidean distance to the (0,1] interval using an exponential function, any deviation of a single parameter will be smoothed out overall, thus ensuring a stable output of similarity. Its physical meaning can be understood as follows: the closer Sfp is to 1, the more consistent the microfluidic disturbance environment of the robot is with that of the reference robot; while when Sfp is significantly lower than 0.6, it indicates that the fluid disturbance characteristics at the robot's location exhibit "interlayer drift" or "shear differentiation," meaning that local differences have occurred in the hydrodynamic environment. This design is equivalent to establishing a "distance scale in the disturbance space" for group operations, enabling the central controller to identify the disruption of the uniformity of the microfluidic layer in real time. Subsequently, in S22, by averaging the disturbance fingerprint similarity scores Sfp of all underwater robots and calculating the synchronization stability index Isync, the central controller can achieve a global-level judgment of microfluidic consistency. If Isync is lower than the synchronization stability threshold Isync... thIf the Isync threshold is reached, it indicates that the population is in a state of coordinated disturbance, requiring no adjustment of the sampling structure. If the Isync threshold is exceeded, it indicates a significant difference in the population's fluid disturbance, and the central controller automatically triggers the adaptive reconstruction mechanism of the sampling grid. This judgment logic avoids the drawbacks of triggering adjustments at fixed time intervals or with manual thresholds, because in natural flow fields, disturbance changes usually exhibit localized abruptness, and overly frequent grid adjustments can introduce computational and communication loads. By introducing the Isync dynamic triggering mechanism, the system only performs adaptive reconstruction when disturbance differences are significant, thus achieving "on-demand change" of the sampling structure. Overall, this design based on the disturbance fingerprint similarity score Sfp and the synchronization stability index Isync gives the system intelligent characteristics of self-sensing, self-judgment, and self-triggering. Physically, Sfp reflects the similarity of local disturbances, while Isync reflects the synchronization stability of global disturbances. The combination of the two is equivalent to establishing a statistical measurement system for population stability within the "disturbance space." Using this method, the system can promptly identify the microfluidic splitting trend when the water flow gradient changes drastically, and actively optimize the sampling grid to dynamically match the sampling density distribution with the disturbance intensity, thereby effectively improving the spatiotemporal resolution of flow field reconstruction and the system's energy efficiency.

[0031] Example 4 Please see Figure 1 and Figure 2 Specifically: S3 includes S31; S31. After triggering the adaptive reconstruction mechanism of the sampling grid, the central controller determines the fluid disturbance heterogeneity index Gdist of the sampling grid unit based on the spatial distribution of the disturbance fingerprint similarity score Sfp. The specific analysis steps include: S311, S312 and S313. S311. Divide the work area into 20×15 sampling grid units, each sampling grid unit having an area of ​​5m×5m, with an overall coverage area of ​​500m×375m underwater work area; when the work area is larger than the overall coverage area, expand proportionally to 40×30 sampling grid units, maintaining a spatial resolution of 25m for each sampling grid unit. 2 And determine the sampling grid unit to which each sampling point belongs based on the real-time position of the underwater robot; S312. In each sampling grid cell, collect all perturbation fingerprint similarity scores Sfp of the current sampling grid cell, and calculate the mean and standard deviation to obtain the mean and standard deviation of the perturbation fingerprint similarity score Sfp of the j-th sampling grid cell, respectively. Then, traverse all sampling raster cells, and calculate the ratio between the mean of the perturbation fingerprint similarity score Sfp of the j-th sampling raster cell and the standard deviation of the perturbation fingerprint similarity score Sfp of the j-th sampling raster cell to obtain the normalized heterogeneity factor N of each sampling raster cell. S313. Based on the normalized heterogeneity factor N of the current sampling grid cell and the maximum and minimum values ​​among all sampling grid cells, perform linear normalization processing to obtain the fluid disturbance heterogeneity index Gdist of the corresponding sampling grid cell.

[0032] S3 also includes S32; S32. The central controller determines the sampling density adjustment coefficient Kgrid for each sampling grid cell based on the obtained fluid disturbance heterogeneity index Gdist. The sampling density adjustment coefficient Kgrid is calculated and output using the following algorithm formula: Kgrid j =1+Cg·Cdist j In the formula, Kgrid j Cdist represents the sampling density adjustment coefficient of the j-th sampling raster cell. j This represents the fluid disturbance heterogeneity index of the j-th sampling grid cell. Cg represents the sampling density gain coefficient, which ranges from 0.5 to 1.0 and is used to control the sensitivity of the sampling density to changes in disturbance heterogeneity. The specific value is set by the user.

[0033] In this embodiment, method S3 achieves adaptive reconstruction of the sampling grid through a two-step closed loop from "heterogeneity metric Gdist to density coefficient Kgrid": First, the working area is divided into 20×15 (expanded to 40×30 if necessary, maintaining 25m) 2The submersible robot (Sfp) is discretized into sampling grids, and samples are assigned to each grid based on the real-time position of the underwater robot. Within each grid, the mean and standard deviation of the perturbation fingerprint similarity score Sfp are calculated, and a normalized heterogeneity factor is constructed using the standard deviation / mean. This factor is then linearly normalized across the entire domain to obtain the fluid perturbation heterogeneity index Gdist. The purpose of this is to use "relative dispersion" to measure the local micro-fluid layer inhomogeneity, which is physically equivalent to "the consistency of perturbation fingerprints within the same small scale," avoiding the neglect of turbulent splitting by only considering the mean. Subsequently, the central controller calculates the sampling density adjustment coefficient Kgrid based on the fluid perturbation heterogeneity index Gdist, making the sampling density monotonically coupled with the perturbation complexity. For example, when vortex core migration occurs near a curved bank or inflow shear zone, Sfp increases and disperses within the grid, Gdist rises, and then Kgrid rises, promptly increasing the sampling frequency of that grid, reducing the sampling interval, and guiding reinforcements, thereby avoiding "information omissions" in high perturbation areas. Conversely, in areas with lower Gdist, the density is maintained or reduced to suppress redundancy and energy consumption. The real-time objective of this process is to dynamically match the spatial resolution with the heterogeneity of the microfluidic layer, achieving the effect of "sampling where it is most needed." Its true physical meaning lies in characterizing local turbulence intensity and stratification fragmentation using statistical dispersion, and directly translating this into a sampling strategy. The resulting benefits are: significantly improved detail fidelity in high-gradient regions, significantly reduced redundancy in low-gradient regions, and simultaneous improvement in overall coverage and energy efficiency. Furthermore, the continuous mapping of Kgrid avoids communication congestion and path jitter caused by frequent on / off adjustments, thus supporting more stable and accurate subsequent role allocation and group synchronization control.

[0034] Example 5 Please see Figure 1 Specifically: S4 includes S41; S41. The central controller receives the sampling density adjustment coefficient Kgrid reported by all underwater robots for each sampling grid unit; then it calculates the mean and standard deviation of all sampling density adjustment coefficients Kgrid to obtain the mean KgridJ and the standard deviation KgridB of the sampling density adjustment coefficients Kgrid. The current sampling density adjustment coefficient Kgrid is compared and evaluated with the mean KgridJ and the standard deviation KgridB of the sampling density adjustment coefficient Kgrid. Based on the comparison and evaluation results, the type of disturbance region is determined, and the sampling task configuration is dynamically adjusted based on the determination results. The sampling task configuration includes sampling frequency adjustment, sampling interval adjustment, and robot adjustment. The specific comparative evaluation content is as follows: When the sampling density adjustment coefficient Kgrid≥KgridJ+0.5·KgridB, it is determined to be a first-level disturbance region, i.e. a high disturbance region. At this time, the sampling frequency of the underwater robot in the current sampling grid cell is increased by 40%, and the sampling interval is reduced by 20%. When KgridJ-0.5·KgridB < sampling density adjustment coefficient Kgrid < KgridJ+0.5·KgridB, it is determined to be a second-level disturbance region, i.e. a stable region. At this time, there is no need to configure the sampling task and the original sampling task is maintained. When the sampling density adjustment coefficient Kgrid≤KgridJ-0.5·KgridB, it is determined to be a level three disturbance region, i.e. a low disturbance region. At this time, the sampling frequency of the underwater robot in the current sampling grid cell is reduced by 40%, and the sampling interval is increased by 20%.

[0035] S4 also includes S42; S42. After the sampling task configuration is updated, the central controller dynamically assigns roles to the underwater robot based on the disturbance region type of each sampling grid unit; the specific assignment details are as follows: The robot roles include backbone underwater robots, encrypted sampling underwater robots, and standby and self-testing underwater robots; the initial role of the underwater robots is backbone underwater robots. It should be noted that the robot role here does not refer to the model of the underwater robot, but rather to the identity of the underwater robot. In the process of role allocation, there is no need to re-deploy underwater robots for the current task. Instead, the existing underwater robots are dynamically assigned roles and tasks are adjusted. When a Level 1 disturbance area is identified, the underwater robot role of the current sampling grid unit is adjusted to an encrypted sampling underwater robot according to the sampling task configuration. If the sampling load of the encrypted sampling underwater robot increases by 30% after the adjustment, the standby and self-testing underwater robots of the adjacent sampling grid unit are called to enter the current sampling grid unit for collaborative support. The original encrypted sampling underwater robot does not withdraw. If the sampling load is still greater than or equal to 30% after collaborative support, the backbone underwater robot of the adjacent sampling grid unit is called for secondary collaborative support. When the area is identified as a Level 2 disturbance zone, the backbone underwater robot is maintained and serves as a backup support for Level 1 disturbance zones. When the area is identified as a Level 3 disturbance zone, the underwater robot role of the current sampling grid unit is switched to standby and self-testing underwater robot.

[0036] In this embodiment, the method achieves online resource optimization through a closed loop of "statistical judgment, task adjustment and role grouping": The central controller first performs global mean KgridJ and standard deviation KgridB statistics based on the sampling density adjustment coefficient Kgrid of each sampling grid unit, and then divides the region into three categories of high disturbance, stable and low disturbance using relative thresholds (Kgrid≥KgridJ+0.5·KgridB, between the two, or ≤KgridJ-0.5·KgridB). This is because a fixed threshold is difficult to cover the disturbance scale of different water bodies and different time periods. The relative scale of "mean ± standard deviation" can adapt to the current flow field baseline. The purpose of this implementation is to allow the density adjustment to automatically normalize with environmental fluctuations. When a disturbance is identified as high, the sampling frequency is increased, the sampling interval is reduced, and additional personnel are deployed locally to avoid "information omissions" at shear bands / vortex core migration points. When a disturbance is identified as low, the frequency is reduced, the interval is widened, and redundancy is released to prevent bandwidth and energy consumption from being occupied by low-value sampling. In stable areas, the nominal configuration is maintained to ensure the continuity of the backbone link and time reference. Subsequently, roles are grouped according to the new task configuration: in high-disturbance areas, the system switches to encrypted sampling underwater robots on-site, and priority is given to drawing standby and self-testing underwater robots from nearby low-disturbance areas for support. Backbone underwater robots are only deployed when standby resources are insufficient or response latency exceeds the standard to ensure that the backbone relay density is not depleted (avoiding energy consumption and communication interruptions caused by long-distance maneuvers). Its physical significance lies in treating Kgrid as a mapping of local disturbance heterogeneity to "sampling information gains," with statistical thresholds decoupling the global mean field from the local bias field, and role grouping precisely targeting "information needs" to "information hotspots." The direct effects of this are: a significant improvement in detail capture rate in high gradient regions, a significant decrease in redundant sampling in low gradient regions, controlled communication collisions and latency jitter, and stable group synchronization; thus, under the same energy consumption and bandwidth constraints, higher spatiotemporal resolution and better task completion rate can be achieved.

[0037] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a collaborative control system for multiple underwater robots, characterized in that: Includes the following steps: S1. In the operating area of ​​multiple underwater robots, multiple underwater robots collect location water flow environment data and transmit it to the central controller. The water flow environment data is statistically processed to obtain local disturbance data, and preprocessing is performed to obtain a standardized dataset. S2. Based on the standardized dataset, calculate the perturbation fingerprint similarity score Sfp for each underwater robot and evaluate the microfluidic synchronization. S3. Based on the microfluidic synchronization evaluation results, the sampling grid adaptive reconstruction mechanism is triggered. The sampling grid adaptive reconstruction mechanism determines the fluid perturbation heterogeneity index Gdist of each sampling grid unit according to the spatial distribution of the perturbation fingerprint similarity score Sfp, and calculates the sampling density adjustment coefficient Kgrid accordingly. S4. The central controller divides the disturbance area types and dynamically adjusts the sampling task configuration according to the sampling density adjustment coefficient Kgrid, and then reassigns the underwater robot roles according to the divided disturbance area types.

2. The method for constructing a multi-underwater robot cooperative control system according to claim 1, characterized in that: S1 includes S11; S11. The work area is a river mouth area with significant velocity gradient and obvious turbulent stratification. Multiple underwater robots are arranged to form a three-dimensional monitoring array according to the initially set depth layers and lateral spacing. Each underwater robot uses an ADCP (Advanced Doppler Current Profiler), an IMU (Inertial Measurement Unit), and controller software integrated on its body to measure the fluid velocity at its location in real time and obtain location water flow environment data. The location-specific water flow environment data includes instantaneous local flow velocity values, acceleration values, and flow velocity fluctuation values.

3. The method for constructing a multi-underwater robot cooperative control system according to claim 1, characterized in that: S1 further includes S12; S12. Multiple underwater robots and the central controller are connected wirelessly via a bandwidth underwater acoustic communication module to transmit the collected location and water flow environment data to the central controller. The bandwidth underwater acoustic communication module includes a communication time slot numbering setting unit, an underwater acoustic signal modulation and demodulation unit, and a time synchronization calibration unit. Each underwater robot locally caches and compresses the location water flow environment data acquired in the current sampling window within a preset communication period of 15 to 30 seconds, and transmits the data in the time-division underwater acoustic channel of the bandwidth underwater acoustic communication module according to its assigned communication time slot number through the communication time slot number setting unit. The communication time slot number is assigned by the central controller during the task startup phase using an initialization broadcast. The time-division structure of the initialization broadcast assignment adopts the Time Division Multiple Access (TDMA) protocol. The duration of a single time slot is 1.5 to 2 seconds, and the total communication cycle includes a complete time slot sequence that is twice the number of seconds of the online underwater robot. The underwater acoustic signal modulation and demodulation unit uses phase shift keying (PS) technology to perform narrowband transmission at a carrier frequency of 9kHz-15kHz, with a data bandwidth of 3-5kbit / s and an effective communication distance of not less than 1000 meters. The time synchronization calibration unit sets a 200-millisecond protection interval before the time period sent by each underwater robot and embeds a 32-bit synchronization header code in the communication time slot number.

4. The method for constructing a multi-underwater robot cooperative control system according to claim 3, characterized in that: S1 also includes S13; S13. After receiving the location and water flow environment data uploaded by each underwater robot, the central controller preprocesses the location and water flow environment data through the central controller's data preprocessing module to obtain local disturbance data; the preprocessing includes data denoising and outlier removal operations. The data denoising is achieved by using a combination of sliding median filtering and Kalman filtering to remove noise from all parameters in the location water flow environment data. Specifically, sliding median filtering denoises the instantaneous values ​​and fluctuations of local flow velocity, while Kalman filtering denoises the acceleration values ​​of the triaxial acceleration signals output by the inertial measurement unit (IMU). The outlier removal operation is based on the mean and standard deviation of all parameters in the denoised location water flow environment data. The mean of each parameter in the water flow environment data is subtracted from the mean. If the absolute value of the output result is greater than 3 times the standard deviation, it is marked as an outlier and removed. Then, the mean of two neighboring points is used for interpolation to complete the data. Finally, the preprocessed location water flow environment data is integrated to obtain local disturbance data. Statistical calculations were performed on the preprocessed local disturbance data to obtain the local average flow velocity Vmean for the i-th underwater robot for each underwater robot. i The root mean square (Urms) of local flow velocity fluctuations of the i-th underwater robot i The local average acceleration Amean of the i-th underwater robot i ,in: The local average velocity Vmean is the arithmetic mean of all the instantaneous local velocities collected by the corresponding underwater robot; The root mean square (Urms) of local velocity fluctuations is the root mean square value of the average of all velocity fluctuation values ​​collected by the corresponding underwater robot. The local average acceleration Amean is the average of all acceleration values ​​collected by the corresponding underwater robot; After the central controller completes the calculation, it normalizes the three types of differences of local average flow velocity Vmean, local flow velocity fluctuation root mean square Urms, and local average acceleration Amean for all underwater robots, performs disturbance feature normalization calculation to eliminate differences between different physical quantities, and obtains a standardized dataset. The standardized dataset includes the normalized scaling parameter ΔV of the local average flow velocity of the i-th underwater robot. i The normalized scaling parameter ΔU of the root mean square of the local velocity fluctuation of the i-th underwater robot. i The normalized scaling parameter ΔA of the local average acceleration of the i-th underwater robot i .

5. The method for constructing a multi-underwater robot cooperative control system according to claim 4, characterized in that: S2 includes S21; S21. Based on the standardized dataset, calculate and output the perturbation fingerprint similarity score Sfp for each underwater robot; the perturbation fingerprint similarity score Sfp is calculated and output using the following algorithm formula; Sfp i =exp(-[(△V i ) 2 +(△U i ) 2 +(△A i ) 2 In the formula, exp represents the exponential function, Sfp i This represents the perturbation fingerprint similarity score of the i-th underwater robot.

6. The method for constructing a multi-underwater robot cooperative control system according to claim 4, characterized in that: S2 further includes S22; S22. Based on the perturbation fingerprint similarity scores Sfp of all underwater robots, calculate the mean to obtain the average value Sfp of the perturbation fingerprint similarity scores. Avg Then, the synchronization stability index Isync is calculated. Set the synchronization stability threshold Isync in the central controller. th Then, the microfluidic synchronization is evaluated with the real-time acquired synchronization stability index Isync to determine the spatial differences in fluid disturbance characteristics among the current multiple underwater robots, and triggering is performed based on the microfluidic synchronization evaluation results; The synchronization stability index Isync is calculated and output using the following algorithm formula; In the formula, N represents the number of underwater robots currently participating in the collaboration; The specific evaluation content of microfluidic synchronization is as follows: When the synchronization stability index Isync < the synchronization stability threshold Isync th If it is determined that the current group of multiple underwater robots is in a stable and consistent state in the microfluidic disturbance feature space, then the current sampling grid structure and path planning are maintained without adaptive adjustment. When the synchronization stability index Isync ≥ the synchronization stability threshold Isync th When it is determined that there are significant spatial differences in the fluid disturbance characteristics of the current underwater robots, the control system automatically triggers the sampling grid adaptive reconstruction mechanism.

7. The method for constructing a multi-underwater robot cooperative control system according to claim 6, characterized in that: S3 includes S31; S31. After triggering the adaptive reconstruction mechanism of the sampling grid, the central controller determines the fluid disturbance heterogeneity index Gdist of the sampling grid unit based on the spatial distribution of the disturbance fingerprint similarity score Sfp. The specific analysis steps include: S311, S312 and S313. S311. Divide the work area into 20×15 sampling grid units, each sampling grid unit having an area of ​​5m×5m, with an overall coverage area of ​​500m×375m underwater work area; when the work area is larger than the overall coverage area, expand proportionally to 40×30 sampling grid units, maintaining a spatial resolution of 25m for each sampling grid unit. 2 And determine the sampling grid unit to which each sampling point belongs based on the real-time position of the underwater robot; S312. In each sampling grid cell, collect all perturbation fingerprint similarity scores Sfp of the current sampling grid cell, and calculate the mean and standard deviation to obtain the mean and standard deviation of the perturbation fingerprint similarity score Sfp of the j-th sampling grid cell, respectively. Then, traverse all sampling raster cells, and calculate the ratio between the mean of the perturbation fingerprint similarity score Sfp of the j-th sampling raster cell and the standard deviation of the perturbation fingerprint similarity score Sfp of the j-th sampling raster cell to obtain the normalized heterogeneity factor N of each sampling raster cell. S313. Based on the normalized heterogeneity factor N of the current sampling grid cell and the maximum and minimum values ​​among all sampling grid cells, perform linear normalization processing to obtain the fluid disturbance heterogeneity index Gdist of the corresponding sampling grid cell.

8. The method for constructing a multi-underwater robot cooperative control system according to claim 7, characterized in that: S3 further includes S32; S32. The central controller determines the sampling density adjustment coefficient Kgrid for each sampling grid cell based on the obtained fluid disturbance heterogeneity index Gdist. The sampling density adjustment coefficient Kgrid is calculated and output using the following algorithm formula: Kgrid j =1+Cg·Cdist j In the formula, Kgrid j Cdist represents the sampling density adjustment coefficient of the j-th sampling raster cell. j The expression represents the fluid disturbance heterogeneity index of the j-th sampling grid cell, and Cg represents the sampling density gain coefficient, with a value range of 0.5-1.

0.

9. The method for constructing a multi-underwater robot cooperative control system according to claim 8, characterized in that: S4 includes S41; S41. The central controller receives the sampling density adjustment coefficient Kgrid reported by all underwater robots for each sampling grid unit; then it calculates the mean and standard deviation of all sampling density adjustment coefficients Kgrid to obtain the mean KgridJ and the standard deviation KgridB of the sampling density adjustment coefficients Kgrid. The current sampling density adjustment coefficient Kgrid is compared and evaluated with the mean KgridJ and the standard deviation KgridB of the sampling density adjustment coefficient Kgrid. Based on the comparison and evaluation results, the type of disturbance region is determined, and the sampling task configuration is dynamically adjusted based on the determination results. The sampling task configuration includes sampling frequency adjustment, sampling interval adjustment, and robot adjustment. The specific comparative evaluation content is as follows: When the sampling density adjustment coefficient Kgrid≥KgridJ+0.5·KgridB, it is determined to be a first-level disturbance region. At this time, the sampling frequency of the underwater robot in the current sampling grid cell is increased by 40%, and the sampling interval is reduced by 20%. When KgridJ-0.5·KgridB < sampling density adjustment coefficient Kgrid < KgridJ+0.5·KgridB, it is determined to be a secondary disturbance region. At this time, there is no need to configure the sampling task and the original sampling task is maintained. When the sampling density adjustment coefficient Kgrid≤KgridJ-0.5·KgridB, it is determined to be a level three disturbance region. At this time, the sampling frequency of the underwater robot in the current sampling grid cell is reduced by 40%, and the sampling interval is increased by 20%.

10. The method for constructing a multi-underwater robot cooperative control system according to claim 9, characterized in that: S4 also includes S42; S42. After the sampling task configuration is updated, the central controller dynamically assigns roles to the underwater robot based on the disturbance region type of each sampling grid unit; the specific assignment details are as follows: The robot roles include backbone underwater robots, encrypted sampling underwater robots, and standby and self-testing underwater robots; the initial role of the underwater robots is backbone underwater robots. When a Level 1 disturbance area is identified, the underwater robot role of the current sampling grid unit is adjusted to an encrypted sampling underwater robot according to the sampling task configuration. If the sampling load of the encrypted sampling underwater robot increases by 30% after the adjustment, the standby and self-testing underwater robots of the adjacent sampling grid unit are called to enter the current sampling grid unit for collaborative support. The original encrypted sampling underwater robot does not withdraw. If the sampling load is still greater than or equal to 30% after collaborative support, the backbone underwater robot of the adjacent sampling grid unit is called for secondary collaborative support. When the area is identified as a Level 2 disturbance zone, the backbone underwater robot is maintained and serves as a backup support for Level 1 disturbance zones. When the area is identified as a Level 3 disturbance zone, the underwater robot role of the current sampling grid unit is switched to standby and self-testing underwater robot.

Images