Intelligent control method and system for multi-target multi-mode underwater fishing robot

Abstract

The intelligent control method of the multi-target and multi-mode underwater fishing robot comprises the following steps: firstly, acquiring environment information through a multi-mode sensing unit and carrying out data processing to obtain comprehensive environment data; key features in the comprehensive environment data are extracted through PCA and input into a prediction model constructed based on SVM, and effective assessment of fishing operation is achieved; defining a plurality of optimization objective functions through a multi-objective optimization model, and realizing optimal path selection under a plurality of constraint conditions by applying an NSGA-II algorithm; then the robot is controlled to run to the target area according to the optimal path, a proper fishing mode is selected, and selective fishing of different types of marine organisms is achieved; after fishing is finished, the data feedback optimization module generates feedback information and analyzes the feedback information, and parameters or weight coefficients of the multi-target optimization model are adjusted according to the analysis result so as to be used for optimizing the advancing path during next fishing. The invention further provides an intelligent control system of the multi-target and multi-mode underwater fishing robot.

Description

Technical Field

[0001] This invention belongs to the field of underwater fishing robot control technology, and specifically relates to an intelligent control method and system for multi-target, multi-modal underwater fishing robots. Background Technology

[0002] An underwater fishing robot is an intelligent robot capable of autonomously or semi-autonomously performing search, identification, and fishing operations in underwater environments. It typically includes a structural frame, thrusters, a robotic arm, fishing tools, cameras, sonar, navigation sensors, controllers, power and communication systems, etc. Underwater fishing robot control refers to the automated operation and management of the underwater fishing robot through a series of advanced sensing technologies, intelligent algorithms, and control systems.

[0003] However, in the field of underwater fishing robot control, traditional model building methods are usually based on human experience. They select a small number of features from sonar, optical images, or environmental monitoring data, such as echo intensity, image brightness, water flow velocity, and depth, based on past operational experience. Then, they use linear regression, simple backpropagation neural networks, or traditional statistical models based on threshold rules for environmental assessment or target identification. These methods generally rely on single sensor inputs, have low feature dimensions and limited expressive power, and are difficult to characterize complex factors in the marine environment such as strong noise, changes in illumination, changes in turbidity, and differences in biological behavior. In addition, traditional path planning methods usually plan with the shortest path or minimum energy consumption as the single objective, lacking a comprehensive consideration of factors such as task efficiency, ecological impact, and energy consumption, which makes the prediction results easy to be distorted and affects the quality of decision-making. With advancements in artificial intelligence, sensor technology, and control theory, existing methods have improved upon traditional approaches. These improvements typically involve acquiring environmental information through sonar or optical systems and using methods such as convolutional neural networks, template matching, or sonar feature extraction to identify targets, thus improving perception accuracy. However, due to the reliance on single sensors, multimodal information cannot be fully utilized, and identification accuracy remains limited in complex scenarios such as murky water, low light, or strong light. In path planning, classic single-objective planning methods such as A*, Dijkstra's algorithm, and genetic algorithms are commonly used, focusing on single indicators such as shortest path, minimum energy consumption, or obstacle avoidance efficiency. These methods lack a comprehensive consideration of multiple factors, including ecological impact, task completion, and fishing success rate, resulting in planning outcomes that fail to balance efficiency and eco-friendliness. Regarding control strategies, most currently employ offline trained models with fixed parameters after deployment. This lack of feedback learning and adaptive optimization based on actual fishing results leads to rigid strategies and poor adaptability, making it difficult for underwater fishing robots to perform tasks efficiently and flexibly in complex and changing marine environments, thus reducing the overall success rate and sustainability of operations. Summary of the Invention

[0004] To address the shortcomings of traditional model building methods, such as inaccurate predictions due to improper feature selection or unsuitable algorithms, which negatively impact decision quality, and the limitations of traditional path planning methods that only consider a single objective, this invention provides an intelligent control method for a multi-objective, multi-modal underwater fishing robot. This intelligent control method first collects environmental information through a multi-modal sensing unit and processes the data to obtain comprehensive environmental data, achieving a complete perception of the complex marine environment. Then, it extracts key features from the comprehensive environmental data using PCA and inputs them into a prediction model based on SVM to effectively evaluate fishing operations. Next, it defines multiple optimization objective functions through a multi-objective optimization model and applies the NSGA-II algorithm to achieve optimal path selection under multiple constraints. Then, it controls the robot to run along the optimal path to the target area and selects an appropriate fishing method to selectively fish for different types of marine life. Finally, after fishing, a data feedback optimization module generates and analyzes feedback information, and adjusts the parameters or weight coefficients of the multi-objective optimization model based on the analysis results to optimize the travel path for the next fishing operation, improving the flexibility and adaptability of path planning. Correspondingly, this invention also provides an intelligent control system for a multi-objective, multi-modal underwater fishing robot.

[0005] Regarding the method, the technical solution of this application is as follows:

[0006] A smart control method for a multi-objective, multi-modal underwater fishing robot includes the following steps: Step 1, after the underwater fishing robot is launched, a multi-modal sensing unit deployed on the robot collects information about the surrounding environment and sends it to a data processing module for processing to obtain the current comprehensive environmental data; then, key features in the comprehensive environmental data are extracted using PCA; Step 2, the extracted key features are input into a prediction model based on SVM to predict fishing feasibility, potential risks, and operational effects; then, a multi-objective optimization model defines multiple optimization objective functions based on the prediction results and analyzes them using the NSGA-II algorithm to obtain the optimal... Step 3: Control the underwater fishing robot's navigation and thrusters to move along the optimal path, driving the underwater fishing robot to the target area and locating it; then, collect and process the target information, and combine it with the environmental information of the target area and task priority to determine the fishing method, fish the target, and record the fishing results; Step 4: The data feedback optimization module combines the key features of the comprehensive environmental data, the optimal path, and the fishing results to generate feedback information, and uses a clustering analysis algorithm to analyze the feedback information. Then, based on the analysis results, it adjusts the parameters or weight coefficients of the multi-objective optimization model to optimize the path in the next fishing.

[0007] Compared with existing technologies, the intelligent control method of the multi-objective multimodal underwater fishing robot in this application first collects surrounding environmental information through a multimodal sensing unit and processes the data to obtain comprehensive environmental data, achieving a comprehensive perception of the complex marine environment. This all-round data collection method not only improves the understanding of the environment but also provides a solid foundation for subsequent task planning. Then, key features of the comprehensive environmental data are extracted through PCA and input into a prediction model based on SVM to achieve effective evaluation of fishing operations. This evaluation method can reduce redundant information and improve computational efficiency. Next, multiple optimization objective functions are defined through a multi-objective optimization model, and the NSGA-II algorithm is applied to achieve optimal path selection under multiple constraints, so as to balance task efficiency, environmental protection, and energy conservation. The process involves considering various factors, including the target area, the environment, and task priorities. The robot is then guided to follow the optimal path to the target area and selects the appropriate fishing method based on target information, environmental information, and task priority. This allows for selective fishing of different types of marine life, avoiding potential damage from traditional fishing methods. While improving fishing efficiency, this also reduces the impact on the marine ecosystem, contributing to the maintenance of marine biodiversity. Finally, after fishing, the data feedback optimization module generates and analyzes feedback information. Based on the analysis results, the parameters or weight coefficients of the multi-objective optimization model are adjusted, dynamically optimizing the path for each fishing attempt. This improves the flexibility and adaptability of path planning and allows for continuous strategy improvement based on actual fishing results, thereby enhancing overall operational efficiency and success rate.

[0008] As an optimization, in step 1 of the aforementioned intelligent control method for multi-objective, multi-modal underwater fishing robots, the collected surrounding environmental information may include marine organism density, water flow velocity, seawater temperature, seawater depth, electrical conductivity, and the location and type of the fishing target. The data processing module first standardizes and normalizes the received information to obtain preprocessed data; then, a weighted average method is used to fuse the preprocessed data. This improves data accuracy and consistency, effectively reduces the impact of environmental noise on the data, and ensures that subsequent prediction, assessment, and path selection are based on high-quality data, thereby increasing the success rate of fishing operations.

[0009] As an optimization, in the aforementioned intelligent control method for multi-objective, multi-modal underwater fishing robots, the prediction model is trained using historical data: First, key features are extracted from historical comprehensive environmental data using PCA and divided into training and testing sets; then, the data in the training set is input into the prediction model for iterative training, enabling the model to continuously learn the relationship between fishing feasibility, potential risks, and operational effects under different environmental combinations until a predetermined number of training iterations is reached; finally, the data in the testing set is input into the trained model to evaluate its performance, obtain the optimal weight model, and complete the training.

[0010] As an optimization, in step 3 of the aforementioned intelligent control method for multi-target multimodal underwater fishing robots, when the underwater fishing robot approaches the target area, it combines the absolute position coordinates provided by the ultra-short baseline positioning system with the centimeter-level relative position deviation provided by the terrain matching algorithm, and uses data fusion technology to achieve spatial positioning of the underwater fishing robot. Then, the target area is scanned by the optical camera and sonar in the multimodal perception unit to acquire visible light images and sonar data. The data processing module performs denoising, enhancement, and registration processing, fusing optical and acoustic information. Next, the YOLOv8-seg algorithm is applied for target detection and instance segmentation, identifying and locating the spatial coordinates, contour features, and biological species of the fishing targets, generating a target list containing target location, size, category, and motion status. Finally, combined with environmental information such as current water flow speed, seabed type, visibility, and task priority, a suitable fishing method is determined. Furthermore, the fishing method includes electric field induction and vacuum adsorption, with corresponding fishing devices being a vacuum adsorption device and an electric field induction device. The vacuum adsorption device includes an interconnected flexible silicone suction cup and a miniature vacuum pump; the electric field induction device includes an electrically connected ring electrode array and a low-voltage pulsed DC power supply. Using non-contact vacuum adsorption technology and bio-friendly electric field induction technology for fishing not only reduces physical damage and stress responses to target organisms but also significantly improves the selectivity and accuracy of fishing, avoiding the high accidental catch rate and ecological damage problems caused by traditional mechanical grasping or trawl fishing.

[0011] For the system, the technical solution of this application is as follows:

[0012] An intelligent control system for a multi-objective, multi-modal underwater fishing robot is provided to implement the aforementioned intelligent control method for the multi-objective, multi-modal underwater fishing robot. It includes a data processing module, a prediction model, a multi-objective optimization model, a fishing execution module, and a data feedback optimization module. The data processing module processes information collected by the multi-modal sensing unit on the underwater fishing robot. The prediction model predicts fishing feasibility, potential risks, and operational effects based on input information. The multi-objective optimization model defines multiple optimization objective functions based on the prediction results and analyzes them using the NSGA-II algorithm to obtain the optimal travel path. The fishing execution module controls the underwater fishing robot's navigation and thrusters to move along the optimal travel path, driving the robot to the target area and locating it. Simultaneously, it determines the fishing method based on target information, environmental information of the target area, and task priority. The data feedback optimization module generates feedback information after each fishing operation, analyzes the feedback information using a clustering analysis algorithm, and adjusts the parameters or weight coefficients of the multi-objective optimization model based on the analysis results.

[0013] Furthermore, it also includes an energy management module and a system maintenance module. The energy management module is used to detect the energy usage of the underwater fishing robot and to collect wave energy and thermal energy from the marine environment to power the underwater fishing robot. The system maintenance module is used to periodically check the operating status of the underwater fishing robot and to provide a prompt when any abnormality is detected in a component, allowing staff to perform corresponding repair operations. By regularly checking the operating status of the underwater fishing robot and promptly repairing faults based on feedback information, the service life of the equipment can be effectively extended and downtime reduced. Attached Figure Description

[0014] Figure 1 This is a flowchart of the intelligent control method for the multi-objective, multi-modal underwater fishing robot of this application. Detailed Implementation

[0015] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0016] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0017] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0018] Example 1:

[0019] See Figure 1This embodiment provides an intelligent control method for a multi-objective, multi-modal underwater fishing robot, which is implemented based on an intelligent control system for the robot. The intelligent control system includes a data processing module, a prediction model, a multi-objective optimization model, a fishing execution module, a data feedback optimization module, an energy management module, and a system maintenance module. The data processing module processes information collected by the multi-modal sensing unit on the underwater fishing robot. The prediction model predicts fishing feasibility, potential risks, and operational effects based on input information. The multi-objective optimization model defines multiple optimization objective functions based on the prediction results and analyzes them using the NSGA-II algorithm to obtain the optimal travel path. The fishing execution module controls the underwater fishing robot's navigation and thrusters to move along the optimal travel path, driving the underwater fishing robot... Upon reaching the target area, the system determines the fishing method based on target information, environmental information of the target area, and task priority. The data feedback optimization module generates feedback information after each fishing operation, analyzes the feedback information using a clustering analysis algorithm, and adjusts the parameters or weight coefficients of the multi-objective optimization model based on the analysis results. The energy management module collects wave energy and thermal energy from the marine environment to power the underwater fishing robot and dynamically allocates energy based on the robot's energy usage. The system maintenance module periodically monitors the underwater fishing robot's operating status and alerts staff when any abnormalities are detected, allowing for appropriate repairs.

[0020] The intelligent control method specifically includes the following steps.

[0021] Step 1: After the underwater fishing robot is launched, a multimodal sensing unit deployed on the robot collects information about the surrounding environment and sends it to the data processing module for processing. Specifically, the multimodal sensing unit includes optical cameras, sonar, CTD sensors, current meters, etc., deployed at the front and around the underwater fishing robot. The collected environmental information includes marine organism density, water current velocity, seawater temperature, seawater depth, conductivity, and the location and type of the fishing target. The data processing module first standardizes and normalizes the received information to obtain preprocessed data; then, it uses a weighted average method to fuse the preprocessed data to obtain the current comprehensive environmental data. Finally, it extracts the key feature set F from the comprehensive environmental data using PCA (Principal Component Analysis).

[0022] Step 2 involves inputting the extracted key feature set F into the prediction model to predict fishing feasibility, potential risks (e.g., robot malfunction, communication interruption, target damage, ecological disturbance, etc.) and operational effects (e.g., fishing success rate, operation rate, task completion time, energy consumption, etc.), obtaining the prediction results. The prediction model is built based on SVM, and its expression is: ;in, It is the first The feature set of a sample It is the first The true evaluation results of a sample It is the loss function. Prediction results. In the formula, It is the first The weight coefficients of each feature, It is the first Each feature value.

[0023] The predictive model is trained using historical data: First, key features are extracted from historical comprehensive environmental data using PCA, and divided into training and test sets. Then, data from the training set is input into the predictive model for iterative training, allowing the model to continuously learn the relationship between fishing feasibility, potential risks, and operational effectiveness under different environmental combinations until a predetermined number of training iterations (or loss function convergence) is reached. Next, data from the test set is input into the trained model to evaluate its performance, obtain the optimal weight model, and complete the training. Furthermore, after each fishing operation, the intelligent control system accumulates the fishing feedback information into the historical dataset, updates the training samples, and allows for subsequent retraining of the model, refining the model weights, and enabling the model to continuously evolve in practice to adapt to the changing underwater environment.

[0024] In this embodiment, the PCA method is used to extract the key feature set and a prediction model based on SVM is adopted. This not only reduces computational complexity but also improves the generalization ability of the model. The combination of PCA and SVM enables the prediction model of this invention to provide accurate prediction results in the face of the ever-changing marine environment, providing a solid foundation for subsequent decision-making and helping to formulate more scientific and reasonable course of action.

[0025] The prediction results are then input into a multi-objective optimization model, which defines multiple objective functions. The NSGA-II algorithm is then used for analysis to obtain the optimal travel path; the expression is: ;in, It is a decision variable. It's the task completion rate. It is an ecological impact factor. It is energy consumption. It is the ecological impact weighting coefficient. It is the energy consumption weighting coefficient. Indicates the travel path.

[0026] In this embodiment, by defining multiple optimization objectives (task completion, ecological impact factor, and energy consumption) and using the NSGA-II / Pareto frontier optimization algorithm, the optimal solution can be found under multiple constraints. This not only maximizes task efficiency but also minimizes the impact on the ecological environment, achieving a win-win situation for both economic benefits and environmental protection.

[0027] Step 3: Control the underwater fishing robot's navigation and thrusters to move along the optimal path, propelling the robot to the target area. As the robot approaches the target area, spatial positioning is achieved through data fusion technology, combining the absolute position coordinates provided by the Ultra-Short Baseline (USBL) positioning system with the centimeter-level relative position deviation provided by the terrain matching algorithm. Next, the target area is scanned using the optical camera and sonar in the multimodal sensing unit to acquire visible light images and sonar data. The data processing module then performs denoising, enhancement, and registration processing, fusing optical and acoustic information. Finally, the YOLOv8-seg algorithm is applied for target detection and instance segmentation, identifying and locating the spatial coordinates, contour features, and biological species of the target. A target list containing the target's location, size, category, and motion status is generated. Combining current water flow velocity (collected by a current meter), substrate type, visibility, and other environmental information, as well as task priority, the optimal fishing method is selected through a preset fishing strategy decision function to catch the target, and the fishing results are recorded. Among them, the water flow velocity is obtained by the current meter; the substrate type and visibility are obtained from the fusion processing results of visible light images and sonar data; the task priority is set according to actual needs, including but not limited to the following factors: the economic value level of the target organism, the importance of the work area, the urgency of the task corresponding to the current energy surplus level; and the level of work instructions set by the operator.

[0028] In this embodiment, the fishing methods include electric field induction and vacuum adsorption, and the corresponding fishing devices employ a vacuum adsorption device and an electric field induction device. The vacuum adsorption device includes an interconnected flexible silicone suction cup and a miniature vacuum pump; the electric field induction device includes an electrically connected ring electrode array and a low-voltage pulsed DC power supply, with the ring electrode array arranged around the robot's fishing port. Using non-contact vacuum adsorption technology and bio-friendly electric field induction technology for fishing not only reduces physical damage and stress responses to target organisms but also significantly improves the selectivity and accuracy of fishing, avoiding the high accidental catch rate and ecological damage problems caused by traditional mechanical gripping or trawl fishing.

[0029] The expression for the fishing strategy decision function is: ;in, Scoring fishing strategies For biological species, For the target size, For water flow velocity, For visibility, As a task priority, to These are the weighting coefficients. , , , These are, respectively, the category adaptation function, the size adaptation function, the environment adaptation function, and the task requirement function. When When choosing a non-contact vacuum adsorption technology, select that method. When choosing a bio-friendly electric field induction technique, select that technique. At the same time, a hybrid mode of collaborative operation is adopted; for each target, specific fishing operations are carried out in accordance with the fishing strategy; The settings should be configured according to the actual situation. In this embodiment, Set to 2. Set to 5.

[0030] When using non-contact vacuum adsorption technology, the vacuum adsorption device is activated, forming a sealed cavity between the flexible silicone suction cup and the target surface. A micro vacuum pump generates negative pressure to attract the target object, and the adsorption force is calculated in real time during the process. The expression is: ;in, It is the actual adsorption force. It is the adsorption area. It is the pressure difference between the inside and outside of the suction cup. It is the contact efficiency coefficient, determined by the surface roughness of the target and the adhesion of the suction cup, and its expression is: ;in, It is the surface roughness of the target (the surface roughness can be calculated by analyzing the visible light images and sonar data of the target area, and by using the Hertz contact model and the JKR model). This is a material correction factor; the maximum safe adsorption force is preset in this embodiment. And adjust in real time through pressure sensor feedback. ,make sure ≤ To avoid causing physical damage to the target.

[0031] When using bio-friendly electric field induction technology, the electric field induction device is activated, and the low-voltage pulsed DC power supply is turned on, causing the ring electrode array to generate a low-intensity directional electric field gradient. This utilizes the electrotaxis of aquatic organisms to guide them towards the fishing port. The formula for calculating the electric field strength is: ;in, It is the effective electric field strength. Indicates the applied voltage. This indicates the distance between the positive and negative electrodes. This indicates the distance between the target and the center of the electrode. It is the seawater attenuation coefficient, expressed as: ;in, It is an empirical constant. Where is the electrical conductivity of the water body, and T is the water temperature. The electric field parameters are dynamically adjusted according to the electrical sensitivity threshold of the target organism: for electrically sensitive species (such as fish), the following settings are used: For species with low susceptibility (such as crustaceans), the level can be appropriately increased to [missing information]. However, it is always kept within a safe range that does not cause muscle rigidity or tissue damage.

[0032] When using the hybrid mode, the electric field is first activated to drive the target to the adsorption area. After the target comes to rest, the vacuum adsorption mode is switched to complete the capture, thus achieving the synergy of the two technologies.

[0033] In this embodiment, by establishing an accurate mathematical model of adsorption force and electric field strength, and by adaptively adjusting the operating parameters in combination with target characteristics and environmental conditions, the fishing process can be made intelligent, flexible, and eco-friendly.

[0034] Step 4: The data feedback optimization module combines key features of the comprehensive environmental data, the optimal travel path, and the fishing results (including target type, quantity, fishing time, fishing location, fishing method used, suction force or electric field strength parameters, fishing success rate, etc.) to generate feedback information, expressed as: ;in, It is the result of fishing. It is a set of key features. It is the travel path; then cluster analysis algorithms are used to process the feedback information. The system performs analysis and obtains the results. Based on the analysis results, it adjusts the parameters or weight coefficients of the multi-objective optimization model to optimize the travel path in the next fishing attempt. At this point, one fishing attempt ends. Steps 1-4 are repeated to fish for the next target, and the adjusted multi-objective optimization model is used to generate the optimal travel path in step 2. Furthermore, throughout the fishing process, the system can evaluate the performance of the fishing methods used based on the fishing results under different environmental conditions, and adjust the weight coefficients of the fishing strategy decision function accordingly (for example, if a specific fishing method is ineffective under certain environmental conditions, the weight coefficient corresponding to that method is reduced; conversely, the weight coefficient of a well-performing fishing method is increased). Then, in the next fishing attempt, the adjusted fishing strategy decision function is used in step 3.

[0035] In this embodiment, the fishing results are analyzed by cluster analysis algorithm, and the parameters or weight coefficients in the multi-objective optimization model are adjusted according to the analysis results. This can dynamically optimize the path, which not only improves the flexibility and adaptability of path planning, but also continuously improves the strategy based on the actual fishing results, thereby improving the overall operation efficiency and success rate.

[0036] In this embodiment, during the use of the underwater fishing robot, the energy management module monitors the energy usage of the underwater fishing robot in real time, as expressed by: ;in, It is energy efficiency. This refers to the energy consumed during the mission. This refers to the total available energy; it also assesses wave energy in the current ocean environment. and temperature difference energy The availability and acquisition potential are expressed as: ; ;in, It is the density of seawater. It is gravitational acceleration. It’s the wave height, It is a wave period. It is the conversion efficiency coefficient. It's a temperature difference. It is heat flow (these parameters are collected in real time by the underwater fishing robot during actual operation through its onboard environmental monitoring instruments, including CTD sensors, flow meters, pressure sensors, temperature difference measurement modules, float-type or IMU-based wave sensing components and other related equipment).

[0037] Then, based on the energy usage analysis results and the harvesting potential of wave energy and thermal energy (energy contained in the temperature difference between the upper and lower layers of seawater), when the availability of wave energy and thermal energy is high, wave energy is collected through the ocean wave energy harvesting module, and temperature difference is collected through the thermal energy harvesting module (such as a thermal battery), and converted into electrical energy to power the robot, avoiding excessive battery consumption. Simultaneously, a dynamic energy allocation strategy can be implemented to adjust the energy supply of each sub-component in real time according to different task requirements and environmental conditions (when wave energy is abundant, priority is given to providing energy to high-energy-consuming equipment; when thermal energy is relatively stable, energy is provided to low-energy-consuming equipment), expressed as: ;in, It is time Energy distribution at that time It is the first The weighting coefficient of each sub-component It is the first Energy requirements of each component.

[0038] This embodiment employs a wave energy-thermal energy conversion composite power supply method, and optimizes energy allocation based on the robot's current energy usage. This fully leverages the advantages of renewable energy in the marine environment, improves energy efficiency, and reduces dependence on external power sources. This not only reduces operating costs but also helps achieve long-term sustainable underwater operations.

[0039] In this embodiment, the system maintenance module is based on feedback information. Surrounding environment data collected by the multimodal sensing unit (Including marine organism density, water flow velocity, seawater temperature, seawater depth, electrical conductivity, and the location and type of fishing targets, etc.), the operation status of the underwater fishing robot is evaluated using a predefined health assessment model.

[0040] Specifically: First, a fault diagnosis algorithm is used to obtain parameter values ​​that deviate from the normal range. The expression is: ;in, It is the diagnosis result. It is a conditional probability, representing the probability given feedback information. and environmental data In the event of a malfunction The probability; then based on the diagnosis results A repair plan is developed, and then repair operations are carried out step by step according to the plan. For example, when irreversible damage to a component is detected and replacement is required (when the parameters of a component deviate significantly from the normal operating range and have not recovered for a long time, and the fault diagnosis algorithm confirms that the fault is irreparable, the system will determine that the component has irreversible damage), the system will trigger an alarm and suggest that the component needs to be replaced. The staff will then use a replacement part that meets the specifications, strictly follow the manufacturer's guidelines for installation, and clean the surface and interior of the underwater fishing robot of dirt and marine organisms; or check whether the current software version is the latest version. If not, the latest firmware update will be downloaded and installed via wireless network, and the error logs will be analyzed to locate and repair the code segment that caused the underwater fishing robot to crash or malfunction. After the repair is completed, the working performance of the robot's robotic arm will be checked by simulating a fishing task to obtain the inspection results after the repair.

[0041] By regularly checking the operational status of underwater fishing robots and promptly repairing faults based on feedback, the service life of the equipment can be effectively extended and downtime reduced. In addition, updating software versions and fixing code segments via wireless network can ensure the latest status and technological advancement of the intelligent control system, thereby enhancing the system's stability and reliability.

[0042] Example 2:

[0043] This embodiment provides a computer device, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the intelligent control method of the multi-target multimodal underwater fishing robot in Embodiment 1.

[0044] The computer device can be a terminal, including a processor, memory, communication interface, display screen, and input device connected via a system bus. The processor provides computing and control capabilities; the memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs, while the internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media; the communication interface is used for wired or wireless communication with external terminals, with wireless communication achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies; the display screen can be an LCD screen or an e-ink display screen; the input device can be a touch layer covering the display screen, or buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse, etc.

[0045] Example 3:

[0046] This embodiment provides a storage medium on which a computer program is stored. When executed by a processor, this program implements the intelligent control method for the multi-target, multi-modal underwater fishing robot in Embodiment 1. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0047] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An intelligent control method for a multi-target multi-modal underwater fishing robot, characterized in that, The method comprises the following steps: Step 1: After the underwater fishing robot is launched, the surrounding environment information is collected by the multi-modal perception unit arranged on the robot, and is sent to the data processing module for processing to obtain the current comprehensive environment data; then the key features in the comprehensive environment data are extracted by PCA; Step 2: The extracted key features are input into the prediction model based on SVM to predict the fishing feasibility, potential risks and operation effect; then, multiple optimization objective functions are defined according to the prediction results by the multi-objective optimization model, and the optimal travel path is obtained by analyzing and processing based on the NSGA-II algorithm; Step 3: The navigation and propeller of the underwater fishing robot are controlled to move according to the optimal travel path, so that the underwater fishing robot reaches the target area and is positioned; then, the target information is collected and processed, and the fishing method is determined in combination with the environmental information of the target area and the task priority to fish the target and record the fishing results; Step 4: The feedback information is generated by the data feedback optimization module in combination with the key features of the comprehensive environment data, the optimal travel path and the fishing results, and the feedback information is analyzed by using the clustering analysis algorithm, and then the parameters or weight coefficients of the multi-objective optimization model are adjusted according to the analysis results for optimizing the travel path in the next fishing.

2. The intelligent control method of the multi-target multi-modal underwater fishing robot according to claim 1, characterized in that: In the step 1, the data processing module first performs standardization and normalization processing on the received information to obtain preprocessed data; then the weighted average method is used for data fusion of the preprocessed data. 3.The intelligent control method of the multi-target multi-modal underwater fishing robot according to claim 2, characterized in that: The collected surrounding environment information includes marine organism density, water flow velocity, seawater temperature, seawater depth, electrical conductivity, and position and type of the fishing target. 4.The intelligent control method of the multi-target multi-modal underwater fishing robot according to claim 2, characterized in that: The prediction model is trained by historical data: first, the key features are extracted from the historical comprehensive environment data by PCA and are divided into training set and test set; then the data in the training set are input into the model for iterative training, so that the model continuously learns the relationship between the fishing feasibility, potential risks and operation effect under different environment combinations until a predetermined training number is reached; then the data in the test set are input into the trained model to evaluate the performance of the model, obtain the optimal weight model, and the training is completed. 5.The intelligent control method of the multi-target multi-modal underwater fishing robot according to claim 1, characterized in that: When the underwater fishing robot approaches the target area, the absolute position coordinates provided by the ultra-short baseline positioning system and the centimeter-level relative position deviation provided by the terrain matching algorithm are combined to realize the spatial positioning of the underwater fishing robot through data fusion technology. 6.The intelligent control method of the multi-target multi-modal underwater fishing robot according to claim 5, characterized in that: In the step 3, the target area is scanned by the optical camera and sonar in the multi-modal perception unit to obtain visible light images and sonar data, which are denoised, enhanced and registered by the data processing module to fuse the optical and acoustic information; then the YOLOv8-seg algorithm is applied for target detection and instance segmentation to generate a target list containing target position, size, category and motion state, and the appropriate fishing method is determined in combination with the current water flow velocity, bottom type, visibility and task priority.

7. The intelligent control method of the multi-target multi-modal underwater fishing robot according to claim 6, characterized in that: The fishing method comprises electric field induction and vacuum adsorption, and the corresponding fishing device is a vacuum adsorption device and an electric field induction device; the vacuum adsorption device comprises a flexible silica gel suction cup and a micro vacuum pump connected with each other; and the electric field induction device comprises an annular electrode array and a low-voltage pulse direct current power supply connected with each other.

8. The intelligent control system of the multi-target multi-modal underwater fishing robot, used for realizing the intelligent control method of the multi-target multi-modal underwater fishing robot according to claim 1; characterized in that: The system comprises a data processing module, a prediction model, a multi-objective optimization model, a fishing execution module and a data feedback optimization module; the data processing module is used for processing information collected by a multi-modal sensing unit on the underwater fishing robot; the prediction model is used for predicting fishing feasibility, potential risks and operation effects according to input information; The multi-objective optimization model is used for defining multiple optimization objective functions according to the prediction results, and analyzing and processing based on an NSGA-II algorithm to obtain an optimal travel path; the fishing execution module is used for controlling the navigation and propeller of the underwater fishing robot to move according to the optimal travel path, driving the underwater fishing robot to reach a target area and position, and then determining a fishing method according to target information, environmental information of the target area and a task priority; and the data feedback optimization module is used for generating feedback information after each fishing, analyzing the feedback information by using a clustering analysis algorithm, and adjusting parameters or weight coefficients of the multi-objective optimization model according to the analysis results.

9. The intelligent control system of multi-target multi-modal underwater fishing robot according to claim 8, characterized in that: The system further comprises an energy management module; the energy management module is used for detecting energy usage conditions of the underwater fishing robot, and collecting wave energy and temperature difference energy in a marine environment to provide power for the underwater fishing robot.

10. The intelligent control system of the multi-target multi-modal underwater fishing robot according to claim 9, characterized in that: The system further comprises a system maintenance module; the system maintenance module is used for regularly detecting running conditions of the underwater fishing robot, and prompting when detecting that a component is abnormal, so that a staff member performs corresponding repair operations.

Images