Bionic fish swarm cooperative control method integrating underwater monitoring and ecological restoration

Abstract

The application relates to a bionic fish swarm cooperative control method integrating underwater monitoring and ecological restoration. In the monitoring process, if a water quality index of a certain point exceeds a threshold value, the bionic fish first starts high-frequency monitoring and sample collection, analyzes the position of a pollution source according to water quality data of monitored points in a monitoring route, plans and changes the route with the pollution source as the terminal point, and performs ecological restoration on the pollution source if restoration is needed. If the water quality indexes of all points are normal in the monitoring process, the bionic fish normally cruise and return the monitoring data. The application realizes efficient monitoring and accurate restoration of the underwater environment, and has great significance for underwater monitoring and restoration work.

Description

TECHNICAL FIELD

[0001] The present application relates to a bionic fish swarm cooperative control method integrating underwater monitoring and ecological restoration, and belongs to the technical field of underwater robots, environmental monitoring and ecological restoration. BACKGROUND

[0002] The importance of water quality monitoring and ecological restoration technology is increasingly prominent. Existing water quality monitoring equipment is mostly fixed equipment or single-function mobile devices, with limited monitoring range and no real-time restoration of pollution. Patent No. CN222379392U proposes a water quality sampling device for river and lake water environment monitoring, which can be moved to different areas for water quality monitoring. However, this device relies on manual pushing and transportation, and does not achieve full automation of movement. Moreover, the monitoring range is limited, and it can only monitor the water quality of a certain depth area on the shore of a well-maintained river or lake. Patent No. CN111289712A proposes a water quality in-situ monitoring device based on an unmanned ship, which can move to different points for water quality monitoring and water sampling. However, this device cannot achieve simultaneous restoration during monitoring and sampling, and the single sampling strategy is difficult to cope with unexpected situations during sampling in natural rivers and lakes. Patent application No. CN116239234A proposes a water body ecological restoration device that combines aeration and dosing. However, this device can only operate in a small range and cannot achieve monitoring and restoration coordination.

[0003] The above defects result in the difficulty of existing water quality monitoring devices in achieving multi-range, stable water quality monitoring and timely restoration measures in river and lake water bodies. Therefore, there is an urgent need for a method that can integrate underwater monitoring and ecological restoration and achieve intelligentization. SUMMARY

[0004] To solve the above problems, the present application discloses a bionic fish swarm cooperative control method integrating underwater monitoring and ecological restoration, and the specific technical solution is as follows:

[0005] A bionic fish swarm cooperative control method integrating underwater monitoring and ecological restoration includes the following processes:

[0006] Step 1: Determine whether there is a monitoring task. When there is no need to monitor water quality, go to Step 2. If water quality needs to be monitored, go to Step 3.

[0007] Step 2: The bionic fish swarm directly follows the set route for regular cruising, detects and records the terrain during cruising, and collects water samples at regular intervals.

[0008] Step 3: Real-time monitoring according to the set monitoring route, the sensor real-time monitors water quality information, and compares the water quality information with the pre-set corresponding water quality index threshold value. If it is abnormal, go to Step 4. If it is normal, go directly to Step 5.

[0009] Step 4: According to the analysis of the water quality data of the monitored points in the monitoring route, the position of the pollution source is determined, and the route is planned and changed with the pollution source as the terminal point, and the position of the pollution source is verified and confirmed;

[0010] Step 5: The control terminal generates a new work path through Q-Learning optimization, and adjusts the bionic fish dynamically based on CNN-LSTM;

[0011] Step 6: Determine whether the pollution needs to be repaired. If it needs to be repaired, go to step 7, and the bionic fish implements in-situ repair. If it does not need to be repaired, cycle steps 1-5;

[0012] Step 7: Determine whether the bionic fish needs to continue working. If it needs to continue working, cycle steps 1-6. If it does not need to continue working, execute the return command.

[0013] Further, the bionic fish group includes a head fish bionic fish and several ordinary bionic fish, each bionic fish is equipped with a data acquisition system, a data processing and instruction center, a navigation and communication module, a driving device and a power supply device, and is also provided with a sampling device and a repair module, the repair module includes a probiotic slow-release capsule warehouse, a chitosan adsorption particle warehouse and an electrolytic aeration unit;

[0014] The data acquisition system includes a camera, an ultrasonic sensor and a water quality sensor;

[0015] The power supply device includes a main power supply device and a backup power supply device.

[0016] Further, the data processing and instruction center of all bionic fish is connected through the communication module and the control center, the control center records and analyzes the position coordinates of each bionic fish and the data transmitted by it, and sends action instructions to each bionic fish, and the driving device adjusts the swimming direction and speed of the bionic fish in time according to the instructions of the control center.

[0017] Further, during the operation of the bionic fish group, every 2 minutes, all bionic fish broadcast their own heartbeat packets, which include id code, battery voltage and power;

[0018] The bionic fish is detected by the following method:

[0019] Heartbeat packet monitoring: every 2 minutes, the state code is checked through ZigBee broadcast, and if there is no response for 3 times in a row, it is judged to be abnormal;

[0020] Neighbor node detection: ultrasonic ranging is used to scan the neighboring fish within a range of 10m to detect whether the physical displacement is normal;

[0021] Data stream analysis: the data processing system carried by each bionic fish verifies the continuity of the data packet in real time, and if the data jump interval is greater than 5 minutes, it is determined that the transmission is interrupted.

[0022] Further, the specific process of step 3 is that the control center first calculates the distribution of Voronoi cells and generates a monitoring route according to the task area information and the state of the bionic fish device performing the task, and the bionic fish device monitors according to the set monitoring route in a timed or fixed-point manner, and the AI system carried by the bionic fish device processes the terrain and water quality data in real time, and optimizes the monitoring route according to the monitored data and path optimization algorithm;

[0023] During the monitoring process, if the water quality index at a certain point exceeds the threshold value, the bionic fish device first starts high-frequency monitoring and sample collection, the control center uses the CNN-LSTM algorithm to analyze the location of the pollution source according to the water quality data of the monitored points in the monitoring route and the historical data of the region, and plans and changes the route with the pollution source as the terminal, verifies and confirms the location of the pollution source, and then the control center generates a new operation path through Q-Learning optimization, and adjusts the bionic fish group dynamically through the deep learning algorithm CNN-LSTM.

[0024] If the water quality index of each point is normal during the monitoring process, the control center optimizes the operation path and adjusts the dynamic of the bionic fish group according to the data obtained during the operation process.

[0025] Further, during the underwater operation process, when a single bionic fish device fails, the bionic fish itself and its neighbors first determine the type of failure: if it is a non-repairable failure, including bionic fish device body damage, communication loss or insufficient power, the failure is reported to the control terminal, and after obtaining the failure information and positioning, the failure fish is salvaged in time; if a local failure that does not affect navigation and driving occurs, the failure fish directly returns to the set homing point and reports the failure to the terminal, and the failure fish is recovered at the homing point;

[0026] Upon receiving the information that the bionic fish device has failed, the head fish immediately recalculates using the Voronoi algorithm based on the number and location of the existing normal bionic fish devices, generates an optimal adjustment scheme, and issues instructions, and all normal bionic fish devices change the operation path and monitoring range according to the optimal adjustment scheme.

[0027] Further, when the pollution type is dissolved oxygen that is too low, the repair method is to start aeration, and multiple fish ring arrays are used for synchronous aeration;

[0028] When the pollution type is heavy metal exceeding the standard, the repair method is to release magnetic nano adsorbent, and the fish group surrounds the pollution source to form a dynamic barrier;

[0029] When the pollution type is organic matter pollution, the repair method is to release probiotic slow-release capsules, and the responsibility of the region is allocated based on the Voronoi algorithm.

[0030] Furthermore, during the monitoring process, if water quality abnormalities are detected, the bionic fish device reports the location information and water quality information to the lead fish and the control center. The control center combines historical and real-time data and uses deep learning algorithms to calculate and analyze the pollution source location. The lead fish calculates and redistributes the work units, assigning more bionic fish devices to the pollution source and its vicinity for verification, intensive monitoring, and sampling. Based on the water quality index monitoring data, the control center determines in real time whether the monitoring point needs in-situ remediation. If remediation is required, the control center issues corresponding remediation instructions according to the needs. After the remediation is completed, the bionic fish group continues its operation or executes the return command.

[0031] Furthermore, the navigation and communication module uses an underwater acoustic positioning system for navigation and a ZigBee / LoRa dual-mode adaptive module for communication.

[0032] Furthermore, the control center is equipped with various machine learning models, and before use, it requires input of real fish swimming videos, game theory models based on Shapley values, dual Q-Learning, and several emergency and contingency cases.

[0033] The beneficial effects of this invention are:

[0034] This invention enables efficient monitoring and precise restoration of the underwater environment, which is of great significance to underwater monitoring and restoration work.

[0035] The biomimetic fish device designed in this invention has a biomimetic shape, which can reduce disturbance to aquatic organisms, improve the accuracy of monitoring data, and enhance the system's underwater concealment.

[0036] In this invention, the biomimetic fish device and control center are equipped with and use a variety of algorithms. The application of the Voronoi algorithm can realize the optimized division and dynamic adjustment of the task area, ensuring full coverage without omission. CNN-LSTM can combine historical data to efficiently analyze and locate pollution sources, and combine Q-Learning method to optimize and adjust the movement line of the biomimetic fish group, thereby improving the operation efficiency of the biomimetic fish group.

[0037] When the bionic fish swarm is operating, it periodically reports its own status and judges the status of neighboring fish. If the bionic fish device malfunctions, the bionic fish swarm can dynamically adjust the allocation of the operating area, and the operator can obtain relevant information and handle it in a timely manner.

[0038] After detecting water quality anomalies, bionic fish swarm monitoring can locate and verify the source of pollution, and intensively monitor water quality and collect samples to improve data accuracy.

[0039] The control center can determine whether a precise repair instruction is needed based on real-time water quality indicators, and the bionic fish swarm will implement in-situ repair measures to achieve precise in-situ repair and monitor and determine the repair effect. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the structure of the present invention.

[0041] List of reference numerals in the attached diagram: 01—Ultrasonic probe, 02—Water quality sensor, 03—Camera, 04—Electrolytic aeration unit, 05—Chitosan adsorption particle chamber, 06—Probiotic sustained-release capsule chamber, 07—Data processing and command center, 08—Navigation and communication module, 09—Power supply equipment, 10—Water sample collection unit, 11—Tail drive device.

[0042] Figure 2 This is a flowchart of the conventional operation of the biomimetic fish school system of the present invention.

[0043] Figure 3 This is a schematic diagram of the initial distribution of the biomimetic fish swarm in Embodiment 1 of the present invention. Detailed Implementation

[0044] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0045] Combined with appendix Figure 1This embodiment of a biomimetic fish system integrating underwater monitoring and ecological restoration consists of multiple biomimetic fish devices, including a head fish and ordinary member fish. The shape and structure of the biomimetic fish devices are designed as follows: (1) The shape of the biomimetic fish devices is similar to that of common fish in rivers and lakes, so as to reduce the disturbance to underwater organisms and improve their concealment in the natural environment. The outer shell is made of thermoplastic polyurethane elastomer rubber material with good waterproof and strong properties, which can effectively protect the internal components and simulate the flexibility and elasticity of real fish; (2) Multiple sensors are installed at the head and torso of the biomimetic fish devices, among which The ultrasonic sensor on the head can detect the terrain, the water quality sensor is used to monitor temperature, pH, turbidity, dissolved oxygen, heavy metal content, and organic matter content, and the camera acquires underwater images to analyze aquatic organisms and assist in judging the terrain; (3) The bionic fish device has a retractable sampling device on its abdomen, which can complete the retraction operation according to the instructions of the control module. When it reaches the designated position, the sampling device extends and opens the sampling port to collect the water sample into the storage bottle. After a water sample is collected, a new storage bottle is automatically replaced. The collected water sample is sealed to ensure the stability and purity of the water sample during the transmission process. (4) The back of the bionic fish device is equipped with three repair modules, namely an electrolytic aeration unit, a chitosan adsorption particle chamber, and a probiotic slow-release chamber, which operate according to repair needs; (5) The tail of the bionic fish device is a drive device, which adopts a multi-joint drive system to achieve efficient underwater propulsion by simulating the tail fin swing of fish; (6) The power supply equipment provides power support for the entire device. It is located in the middle and rear of the bionic fish body and uses a high-performance lithium battery to ensure its long-term underwater endurance. It is also equipped with an energy management system to optimize energy use efficiency and extend service life; (7) The bionic fish device is equipped with a data processing and command center to perform simple processing and judgment on the water quality and terrain information obtained by the bionic fish device; (8) The bionic fish is also equipped with a navigation module and a communication module. The navigation module adopts an underwater acoustic positioning system to achieve accurate underwater positioning and path planning. The communication module adopts ZigBee / LoRa dual-mode adaptive technology to transmit the data and device status obtained by the bionic fish device to the ground control center in real time, and at the same time receive the command from the control center. Meanwhile, the status information of the bionic fish device is also transmitted to the adjacent bionic fish through the communication module.

[0046] The ground control center is equipped with various machine learning models and, before use, inputs real fish swimming videos, dual Q-Learning data, regional historical monitoring data, and several emergency and contingency cases to improve the control center's data processing efficiency and decision-making accuracy.

[0047] The following is a specific embodiment of the structure of a sampling device, a probiotic sustained-release capsule chamber, a chitosan adsorption particle chamber, and an electrolytic aeration unit:

[0048] The specific structure of the sampling device is as follows:

[0049] The sampling device includes a sampling unit and a sample storage unit. The sampling unit comprises a retractable miniature sampling head, a miniature corrosion-resistant peristaltic pump, and a fluid pipeline. The retractable miniature sampling head is equipped with a one-way valve. When not in operation, the sampling head retracts into a biomimetic fish shell; when in operation, it extends via a miniature servo motor. The sample storage unit is a rotatable sample tray with multiple independent and sealed sample vials evenly distributed circumferentially. The sample tray is driven by a stepper motor. The unit also includes an automatic cap opening and closing mechanism, which opens the cap of the sampling vial when it rotates to the sampling position and closes it immediately after sampling.

[0050] After receiving the sampling command, the sampling device uses a stepper motor to drive the sample tray, rotating an unused empty sample bottle to the sampling station. The sealing cap opening mechanism opens the sealing cap of the sample bottle. The micro servo motor extends the sampling head, the peristaltic pump starts, and a specific volume of target water sample is extracted and injected into the sample bottle through the pipeline. After sampling is completed, the peristaltic pump stops, the sealing cap closes tightly, and the sampling head retracts. The sample tray rotates one position to prepare the next empty bottle for the sampling station.

[0051] Beneficial effects: It enables automated and orderly sampling, and ensures that the sample bottles are sealed during non-sampling periods, effectively preventing sample volatilization, external contamination, or cross-contamination between samples, thereby improving the accuracy and reliability of sampling data.

[0052] The specific structures of the probiotic sustained-release capsule chamber and the chitosan adsorption particle chamber are as follows:

[0053] The probiotic sustained-release capsule chamber and chitosan adsorbent particle chamber are used to store and release sustained-release capsules / chitosan particles. Essentially, they are controlled storage chambers, their main feature being the isolation of the drug from water. Both the probiotic sustained-release capsule chamber and the chitosan adsorbent particle chamber have the same structure, employing a dual-chamber structure of a main storage chamber + a buffer release chamber. The main storage chamber is a sealed dry chamber used to store unused capsules or particles. Its bottom outlet is connected to the buffer release chamber below via an electrically operated sealing butterfly valve. The buffer release chamber is a small-capacity cavity with a double-opening door at the bottom driven by a micro-motor. This door has excellent sealing performance and only opens at the moment of release.

[0054] Taking the release of probiotic sustained-release capsules as an example: After receiving the release command, the electric butterfly valve at the bottom of the main storage chamber is first opened to a specific degree or for a specific time, so that a certain number of capsules fall into the buffer release chamber; then the electric butterfly valve is immediately closed, completely isolating the main storage chamber from the external environment; the double-door at the bottom of the buffer release chamber is quickly opened, and the probiotic sustained-release capsules are released into the water. After the release is completed, the door is immediately closed.

[0055] Beneficial effects: Ensuring that only the buffer release chamber is in direct contact with external water greatly reduces the risk of the main storage chamber becoming damp, compacted, or ineffective due to water ingress, thus guaranteeing the reliability of long-term operation.

[0056] The specific structure of the electrolytic aeration unit is as follows:

[0057] The electrolysis aeration unit includes an electrolytic cell, a power supply and management area, and a gas-liquid separation and discharge system. The electrolytic cell employs a compact plate or honeycomb design, with electrode materials consisting of a base metal oxide coated anode and a stainless steel cathode. The power supply and management module integrates a high-voltage micro DC-DC converter, which can boost the main power supply voltage of the biomimetic fish to the operating voltage required for water electrolysis; it also has a built-in intelligent control chip that can adjust the output power and control the oxygen production rate according to instructions. The hydrogen and oxygen mixture generated during electrolysis naturally precipitates from the top of the electrolytic cell and collects in a micro gas collection chamber above the unit; the gas collection chamber guides the gas through a thin aeration pipe to the microporous aeration head on the surface of the biomimetic fish for discharge.

[0058] Beneficial effects: When the bionic fish detects low dissolved oxygen in a small area, it generates and releases oxygen-rich microbubbles, instantly improving the dissolved oxygen level in the local water around the bionic fish, achieving "precision aeration." The electrolytic cell design maximizes the electrode surface area while minimizing volume. The electrode material has good catalytic efficiency and corrosion resistance, extending its service life.

[0059] Combined with appendix Figure 2A biomimetic fish swarm collaborative control method integrating underwater monitoring and ecological restoration is proposed. Its main operational process is as follows: During routine operations, if water quality monitoring is not required, the biomimetic fish swarm directly cruises along the route set by the operator, detecting and recording terrain and collecting water samples periodically. If water quality monitoring is required, the control center first calculates and allocates Voronoi units and generates a monitoring route based on the operational area information and the status of the biomimetic fish devices performing the task. The biomimetic fish devices monitor regularly or at fixed points according to the set monitoring route, and the onboard AI system processes terrain and water quality data in real time, optimizing and adjusting the monitoring route based on the monitored data and path optimization algorithms. During monitoring, if the water quality index at a certain point exceeds the threshold, the biomimetic fish devices first activate high-frequency monitoring and sample collection. The control center uses the CNN-LSTM algorithm to analyze the location of the pollution source based on the water quality data of the monitored points along the monitoring route and historical data of the area, planning and changing the route with the pollution source as the endpoint. After verifying and confirming the location of the pollution source, the control center generates a new operational path through Q-Learning optimization and dynamically adjusts the biomimetic fish swarm using the deep learning algorithm CNN-LSTM. If the water quality indicators at all monitoring points are normal during the monitoring process, the control center will optimize the operational path and adjust the dynamics of the bionic fish swarm based on the acquired data. During the operation, the lead fish in the bionic fish swarm system uses the Voronoi algorithm, with the input sampling range as the constraint boundary, to process and generate the operational paths and responsibility areas of all bionic fish devices in real time. All members of the bionic fish swarm adjust their routes and dynamics according to the lead fish's planning and allocation. Simultaneously, during the monitoring process, the control center will determine in real time whether the monitoring points require in-situ remediation based on the water quality indicator monitoring data. If remediation is required, the control center will issue corresponding remediation instructions as needed. After the remediation is completed, the bionic fish swarm will continue its operation or execute a return command.

[0060] Example 1: The following is an example of fault handling during actual operation of the present invention:

[0061] The biomimetic fish swarm collaborative control method integrating underwater monitoring and ecological restoration can promptly handle the faulty fish and replan the operation path and range of the normal biomimetic fish devices when a single biomimetic fish device malfunctions during operation. This embodiment provides an example of biomimetic fish swarm system handling and redistribution when multiple biomimetic fish swarms malfunction during zoned operation.

[0062] The bionic fish device achieves perception and response through a three-level linkage mechanism: heartbeat packet monitoring, neighbor node monitoring, and data stream analysis, determining whether other bionic fish devices are operating normally. The specific operations and judgment methods are shown in Table 1. During the operation of the bionic fish swarm, every 2 minutes, all bionic fish devices broadcast their own heartbeat packets, including ID codes, battery voltages, and power levels.

[0063] Table 1. Judging the Operating Status of the Bionic Fish Device

[0064]

[0065] In a mission with regional topographic survey and water quality monitoring as its core objectives, 14 biomimetic fish devices formed a monitoring network covering a rectangular water area of ​​approximately 3 km × 2 km. This water area was divided into three zones: an ecological protection zone, an industrial sewage discharge zone, and a waterway zone. The IDs of the 14 biomimetic fish devices were F01 to F14, with F7 being the lead fish.

[0066] In this monitoring mission, significant differences exist in monitoring value, environmental risk, and node capabilities across different regions. Based on the principles of allocating more monitoring resources to areas with high ecological value, increasing coverage redundancy in high-pollution-risk areas, and assigning greater responsibility to high-performance nodes, a weighted Voronoi diagram was used to delineate monitoring areas. Weights were calculated during the initial pattern distribution to quantify the monitoring capabilities or priorities of nodes. Nodes with higher weights will receive larger Voronoi units, thus assuming greater monitoring responsibility. The weight calculation method is as follows:

[0067] weight =α*Vi +β*Ci + γ*Ri - δ*Ei

[0068] Where α, β, γ, and δ are weighting coefficients, and satisfy α+β+γ+δ=1. V is the regional value coefficient, C is the node capability index, R is the environmental risk index, E is the energy consumption rate, and i represents each fish. For example, each fish has its own number, so i represents the number of each fish.

[0069] V is the regional value coefficient, automatically assigned based on the regional type, ranging from 0.8 to 1.5, and calculated using (Python):

[0070] # Automatic value assignment based on region type

[0071] def get_value_factor(zone_type):

[0072] mapping = {

[0073] "RARE_SPECIES_HABITAT": 1.5,

[0074] "WATER_INTAKE": 1.3,

[0075] "INDUSTRIAL_DRAIN": 1.2,

[0076] "SHIPPING_LANE": 1.0,

[0077] "GENERAL_WATER": 0.8

[0078] }

[0079] return mapping.get(zone_type, 1.0)

[0080] C is the node capability index, which is calculated by assigning different weights to sensor status, battery status, CPU performance, and reliability. The index ranges from 0 to 1, and the calculation method is as follows:

[0081] Ci = 0.4 \times \frac{S_{sensor}}{S_{max}} + 0.3 \times \frac{B_{battery}}{B_{max}} + 0.2 \times P_{cpu} + 0.1 \times R_{reliability}

[0082] R is the environmental risk index, calculated based on historical pollution data, ranging from 1 to 2. The calculation method is as follows:

[0083] risk_factor = min(2.0, 1.0 + log10(pollution_events + 1))

[0084] E is the energy consumption rate, calculated as a relative value of distance and power, ranging from 0.1 to 0.5. The calculation method is as follows:

[0085] Ei = 0.3 \times \frac{D_{distance}}{D_{max}} + 0.7 \times \frac{P_{power}}{P_{max}}

[0086] Through multi-objective optimization, α=0.5, β=0.3, γ=0.15, and δ=0.05 were determined. The weights of different biomimetic fish devices were calculated and adjusted to determine the Voronoi unit allocation. Figure 3 The initial distribution pattern of the biomimetic fish swarm task execution is shown. Under this distribution, F02, F03, and F04 have the highest weights, F07 and F08 are relatively high, F01 and F11 have moderate weights, and F05, F06, F09, F10, F12, F13, and F14 have relatively low weights. During the operation, the weights are updated based on monitoring data and the operating status of the biomimetic fish device. If the node weights change significantly, the Voronoi units are recalculated and adjusted.

[0087] Upon reaching the designated location, the bionic fish device began patrolling and monitoring. During this process, bionic fish device F04 was damaged, rendering it unable to operate or communicate normally. F03 and F08, upon discovering that F04 had failed to respond three times consecutively, immediately sent signals to the lead fish and the control center, reporting the location of F04's last communication before the heartbeat was lost. Upon receiving the information about F04's abnormality, the lead fish immediately performed a Voronoi reassignment. Ultimately, F03 and F08 took over 60% and 40% of the remaining monitoring area of ​​F04, respectively, and allocated some of the original monitoring area to neighboring fish. After the reassignment, the working area of ​​all member fish in the bionic fish swarm increased by 5% to 15%.

[0088] Example 2: The following is an example of pollution treatment during the actual operation of the present invention:

[0089] A biomimetic fish swarm collaborative control method integrating underwater monitoring and ecological restoration triggers intensive monitoring and sampling when water quality anomalies are detected, and analyzes and verifies the location of pollution sources. In-situ remediation can also be implemented when the environment requires it. Table 2 lists the remediation methods and system collaboration mechanisms of the biomimetic fish swarm under three pollution conditions. This embodiment presents the response process after detecting water quality anomalies.

[0090] Table 2 In-situ remediation methods and synergistic mechanisms

[0091]

[0092] In a water quality monitoring mission, ten biomimetic fish-like devices formed a monitoring network, operating in a strip-like, equidistant distribution pattern. The devices were numbered F01 to F10, with F05 as the lead fish. The distance between adjacent fish was 100 m, and the river channel was 20 m wide. During operation, the speed of the fish-like devices was approximately 2 m / min, the monitoring frequency was 4 min / time, and the sampling frequency was 10 min / time. During operation, F08 detected that the dissolved oxygen content at a certain location was 1 / 3 lower than the threshold. It immediately packaged its coordinates, sensor readings, and timestamp, uploaded them to the control center via the communication module, and broadcast a simple alarm signal to neighboring units F07 and F09.

[0093] After receiving the detailed data from F08, the control center first retrieves the recent historical time-series data of F07, F08, and F09, the water quality data in the historical database of the area, the diffusion model and water flow and wind direction data, and the information such as shoreline, water depth, and sewage outlet location provided by the electronic map. The system automatically creates a 200m×20m analysis grid centered on anomaly point F08. It aligns and normalizes the real-time and historical time-series data of F07, F08, and F09, forming a temporal tensor X_temporal with dimensions of (3 observation points, 5 time steps, and 4 water quality parameters). Spatial Kriging interpolation is performed on the spatial coordinates of the three biomimetic fish and their current dissolved oxygen detection values ​​to generate a 100×10 grid "water quality spatial snapshot," forming a (20, 20, 1) spatial tensor X_spatial. X_spatial is input into a lightweight CNN with the structure Conv2D( filters=8, kernel_size=3, activation='relu' )→MaxPooling2D()→Flatten(), which identifies the concentration gradient change from F08 to F07 and outputs a spatial feature vector. The X_temporal input is fed into an LSTM with the structure LSTM(units=16, return_sequences=False) to analyze the dynamic changes in pollutant concentration and output a temporal feature vector. The two feature vectors output from the CNN and LSTM are concatenated and fed into a fully connected layer for regression prediction: Dense(units=32, activation='relu') → Dense(units=2, activation='linear'). The most probable pollution source coordinates are output as (40, 10), predicting the pollution source to be located approximately 40m upstream of F08 on the bank, with a confidence level of 85%. The control center issues instructions for F07, F08, and F09 to proceed to the predicted location at a speed of 10m / min for confirmation and intensive monitoring.

[0094] After F07, F08, and F09 arrived at the target point, they intensively monitored the water quality at and near the target location, confirming the existence of a hidden sewage outlet. The control center then activated the Voronoi and Q-Learning algorithms to regenerate the global operational path. Based on the current positions of all bionic fish, the location of the pollution source, and the priority of uncompleted monitoring points in each area, the new operational area for each bionic fish and the movement instructions to be assigned to each bionic fish were calculated. The reward function consisted of a high-priority area coverage reward, an energy-saving reward, and a collision avoidance penalty. After the algorithm ran, it output the globally optimal new operational plan: F07, F08, and F09 formed a ring array around the pollution source, performing aeration, feeding, and effect monitoring; the operational range of the other bionic fish was expanded by 20%-40%, and their paths were readjusted to ensure full coverage of the original area with the highest efficiency.

[0095] Those skilled in the art will understand that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless defined as herein.

[0096] Based on the above-described preferred embodiments of the present invention, and through the foregoing description, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A biomimetic fish swarm collaborative control method integrating underwater monitoring and ecological restoration, characterized in that, Includes the following processes: Step 1: Use a weighted Voronoi diagram to divide the monitoring area. During the initial pattern distribution process, calculate the weights to quantify the monitoring capabilities or priorities of the nodes. Nodes with higher weights will receive larger Voronoi units, thus undertaking more monitoring responsibilities. The following is the calculation method for the weights: weight =α×Vi +β×Ci + γ×Ri - δ×Ei Where α, β, γ, and δ are weighting coefficients, satisfying α+β+γ+δ=1; V is the regional value coefficient, automatically assigned based on the regional type; C is the node capability index, calculated by assigning different weights to sensor status, battery status, CPU performance, and reliability; R is the environmental risk index, calculated based on historical pollution data; and E is the energy consumption rate, calculated using the relative values ​​of distance and power. Determine whether there is a monitoring task in each monitoring area. If there is no need to monitor water quality, proceed to step 2. If water quality needs to be monitored, proceed to step 3. Step 2: The bionic fish swarm directly follows the set monitoring route for regular patrols. During the patrol, it detects and records the terrain and collects water samples at regular intervals. Specifically, the control center first calculates and allocates Voronoi units and generates monitoring routes based on the information of the work area and the status of the bionic fish device performing the task. The bionic fish device monitors at regular intervals or fixed points according to the set monitoring route, and the onboard AI system processes the terrain and water quality data in real time, and optimizes and adjusts the monitoring route based on the monitored data and path optimization algorithm. Step 3: Monitor water quality in real time according to the set monitoring route. The sensor monitors water quality information in real time and compares the water quality information with the preset corresponding water quality index threshold. If there is an abnormality, proceed to step 4; if it is normal, proceed directly to step 5. Step 4: During the monitoring process, if the water quality index at a certain point exceeds the threshold, the location of the pollution source is analyzed based on the water quality data of the monitored points along the monitoring route. The bionic fish device first starts high-frequency monitoring and sample collection. The control center uses the CNN-LSTM algorithm to analyze the location of the pollution source based on the water quality data of the monitored points along the monitoring route and the historical data of the area. The route is planned and changed with the pollution source as the endpoint. After verifying and confirming the location of the pollution source, the control center generates a new operation path through Q-Learning optimization and dynamically adjusts the bionic fish group through the deep learning algorithm CNN-LSTM. Step 5: If the water quality indicators at each monitoring point are normal during the monitoring process, the control center will optimize the operation path and adjust the dynamics of the bionic fish group based on the acquired data during the operation. Specifically, the control center will generate a new operation path through Q-Learning and adjust the dynamics of the bionic fish group based on CNN-LSTM. Step 6: Determine if the pollution needs to be repaired. If repair is needed, proceed to Step 7, where the bionic fish swarm performs in-situ repair. If repair is not needed, repeat Steps 1-5. Step 7: Determine whether the bionic fish swarm needs to continue operations. If it needs to continue, repeat steps 1-6. If it does not need to continue, execute the return command.

2. The biomimetic fish swarm collaborative control method integrating underwater monitoring and ecological restoration according to claim 1, characterized in that, The biomimetic fish swarm includes a headed fish biomimetic fish and several ordinary biomimetic fish. Each biomimetic fish is equipped with a data acquisition system, a data processing and command center, a navigation and communication module, a drive device and a power supply device. It is also equipped with a sampling device and a repair module. The repair module includes a probiotic slow-release capsule chamber, a chitosan adsorption particle chamber and an electrolytic aeration unit. The data acquisition system includes a camera, an ultrasonic sensor, and a water quality sensor; The power supply equipment includes main power supply equipment and backup power supply equipment.

3. The biomimetic fish swarm collaborative control method integrating underwater monitoring and ecological restoration according to claim 2, characterized in that, All the data processing and command centers for the bionic fish are connected to the control center through a communication module. The control center records and analyzes the position coordinates of each bionic fish and the data it transmits, and sends motion commands to each bionic fish. The drive device adjusts the swimming direction and speed of the bionic fish in a timely and accurate manner according to the commands from the control center.

4. The biomimetic fish swarm collaborative control method integrating underwater monitoring and ecological restoration according to claim 1, characterized in that, During the operation of the bionic fish swarm, every 2 minutes, all bionic fish broadcast their own heartbeat packets, which include ID code, battery voltage, and power level. The following methods can be used to detect whether the bionic fish is malfunctioning: Heartbeat monitoring: Verification is performed every 2 minutes via ZigBee broadcast status code; failure to respond after 3 consecutive attempts indicates an anomaly. Neighbor node detection: Ultrasonic ranging is used to scan nearby fish within a 10m range to detect whether the physical displacement is normal; Data flow analysis: The data processing system on each bionic fish verifies the continuity of data packets in real time. If the data jump interval is greater than 5 minutes, the transmission is considered interrupted.

5. The biomimetic fish swarm collaborative control method integrating underwater monitoring and ecological restoration according to claim 1, characterized in that, During underwater operations, when a single bionic fish device malfunctions, it and its neighboring bionic fish first determine the type of malfunction: if it is an irreparable malfunction, which includes damage to the bionic fish device body, loss of communication, or insufficient power, then the malfunction is reported to the control terminal. After obtaining the malfunction information and location, the malfunctioning fish is retrieved in a timely manner; if a local malfunction occurs that does not affect navigation and drive, the malfunctioning fish will return directly to the set return point and report the malfunction to the terminal, and the malfunctioning fish will be retrieved at the return point. Upon receiving information that the bionic fish device has malfunctioned, the lead fish immediately uses the Voronoi algorithm to recalculate based on the number and location of the existing normal bionic fish devices, generates the optimal adjustment plan, and issues instructions. All normally operating bionic fish devices change their operating paths and monitoring ranges according to the optimal adjustment plan.

6. The biomimetic fish swarm collaborative control method integrating underwater monitoring and ecological restoration according to claim 1, characterized in that, When the pollution type is low dissolved oxygen, the remediation method is to start aeration and simultaneously aerate the multi-fish ring array. When the pollution type is heavy metal exceeding the standard, the remediation method is to release magnetic nano-adsorbents, and the fish surround the pollution source to form a dynamic barrier. When the pollution type is organic pollution, the remediation method is to release probiotic slow-release capsules and allocate regional responsibility based on the Voronoi algorithm.

7. The biomimetic fish swarm collaborative control method integrating underwater monitoring and ecological restoration according to claim 1, characterized in that, During the monitoring process, if water quality abnormalities are detected, the bionic fish device reports the location information and water quality information to the lead fish and the control center. The control center combines historical and real-time data and uses deep learning algorithms to calculate and analyze the pollution source location. The lead fish calculates and redistributes the work units, assigning more bionic fish devices to the pollution source and its vicinity for verification, intensive monitoring, and sampling. Based on the water quality index monitoring data, the control center determines in real time whether the monitoring point needs in-situ remediation. If remediation is required, the control center issues the corresponding remediation command according to the needs. After the remediation is completed, the bionic fish group continues to operate or executes the return command.

8. The biomimetic fish swarm collaborative control method integrating underwater monitoring and ecological restoration according to claim 2, characterized in that, The navigation and communication module uses an underwater acoustic positioning system for navigation and a ZigBee / LoRa dual-mode adaptive module for communication.

9. The biomimetic fish swarm collaborative control method integrating underwater monitoring and ecological restoration according to claim 1, characterized in that, The control center is equipped with various machine learning models and requires input of real fish swimming videos, game theory models based on Shapley values, dual Q-Learning, and several emergency and contingency cases before use.

Images