Net cage unmanned feeding operation method, device and product

Abstract

The application discloses a net cage unmanned feeding operation method, device and product, relates to the technical field of marine engineering equipment, and the method comprises the following steps: obtaining a feeding task of a target net cage, planning a heading track of an unmanned feeding ship according to a current environmental flow direction, and controlling the unmanned feeding ship to sail to the target net cage in a head current state; when the unmanned feeding ship sails to a preset area, identifying the target net cage through a sensor and obtaining a relative position of the net cage in combination with position prediction information, determining a berthing point based on the relative position of the net cage, controlling the unmanned feeding ship to abut against the berthing point, and executing the feeding task according to a preset feeding strategy after judging that the position of the ship body satisfies feeding conditions; if the feeding operation is completed, controlling an on-board monitoring unmanned aerial vehicle to collect fish school feeding data, and optimizing the feeding strategy according to the fish school feeding data; controlling the on-board monitoring unmanned aerial vehicle to be recovered and planning a sailing track to a next target net cage until all feeding tasks are completed.

Description

Technical Field

[0001] This application relates to the technical field of marine engineering equipment, and in particular to a method, device and product for unmanned feeding of net cages. Background Technology

[0002] As the aquaculture industry develops towards deep-sea and intensive farming, the technical demands for operational efficiency, cost control, and safety in gravity cage aquaculture are becoming increasingly urgent, necessitating the intelligent and unmanned transformation of feeding operations.

[0003] Currently, feeding operations in gravity-fed fish cages generally rely on traditional manual methods. This involves aquaculture workers navigating feeding boats to designated sea areas, visually determining the cage locations, and manually feeding and observing fish schools. However, this manual operation is severely limited by the complex and ever-changing weather and sea conditions at sea, as well as the physiological limits of personnel. It is not only labor-intensive and costly, but also poses significant safety hazards. Furthermore, the accuracy of feeding volume and berthing positioning is highly dependent on individual experience, lacking real-time data feedback and dynamic adjustment mechanisms. This results in low operational efficiency and makes it difficult to meet the urgent needs of large-scale aquaculture for precise feeding and automated operations.

[0004] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main purpose of this application is to provide a method, device and product for unmanned feeding of net cages, which aims to solve the technical problems of low efficiency and poor safety of existing manual feeding methods, which cannot meet the feeding needs of aquaculture for unmanned, precise and efficient operation.

[0006] To achieve the above objectives, this application proposes a method for unmanned feeding of fish cages, the method comprising: Obtain the feeding task of the target cage, plan the course of the unmanned feeding vessel according to the current environmental flow, and control the unmanned feeding vessel to sail towards the target cage in a top-down state; When the unmanned feeding vessel sails to the preset area, it identifies the target cage through sensors and obtains the relative position of the vessel and cage by combining the position prediction information, and determines the berthing point based on the relative position of the vessel and cage. The unmanned feeding vessel is controlled to dock at the mooring point, and after determining that the vessel's position meets the feeding conditions, the feeding task is executed according to the preset feeding strategy. Once the feeding operation is complete, the onboard monitoring drone is controlled to collect data on the fish's feeding activity, and the feeding strategy is optimized based on this data. Control the airborne monitoring drone to recover and plan the flight path to the next target cage until all feeding tasks are completed.

[0007] In one embodiment, the steps of obtaining the feeding task of the target cage, planning the course and trajectory of the unmanned feeding vessel according to the current environmental current, and controlling the unmanned feeding vessel to sail towards the target cage in a top-down state include: If the current feeding task is determined to be the first cage in the operation sequence, the initial course trajectory is planned according to the preset cage number and position priority; If the determination result is that it is not the first cage, the course trajectory will be dynamically planned based on the current position of the hull and the already executed navigation route; Among them, the flow direction feedback data based on the flow direction sensor controls the bow of the unmanned feeding vessel to be opposite to the water flow direction, generating a heading trajectory from the current position to the target cage and sailing in a top-current attitude.

[0008] In one embodiment, the step of controlling the unmanned feeding vessel to navigate towards the target cage in a top-current state further includes: The course trajectory is optimized using the artificial potential field method, which constructs an gravitational potential field around the target cage and a repulsive potential field in the non-target area. The potential field parameters are adjusted according to the relationship between the bow direction of the unmanned feeding vessel and the direction of the water flow, so that the unmanned feeding vessel can automatically adjust its bow direction to point towards the target cage under the action of the resultant force.

[0009] In one embodiment, the step of identifying the target cage using sensors and obtaining the relative position of the cage by combining position prediction information includes: The target cage is identified by a camera to determine its direction, and the target cage is measured by radar. Call the pre-built cage location model and input the current environmental parameters to obtain the predicted location of the target cage; The measured distance and the predicted position are fused using a Kalman filter algorithm to output a high-precision real-time distance between the ship and the container.

[0010] In one embodiment, the steps for constructing the cage location model include: Collect the latitude and longitude location of the net cages and the corresponding environmental parameters at each feeding operation, and establish a net cage location map; The location map of the net cage is trained using a neural network model to obtain the mapping relationship between environmental parameters and net cage location. The predicted location of the target net cage is output by inputting real-time environmental parameters.

[0011] In one embodiment, the step of determining the berthing point based on the relative position of the ship's container includes: Obtain the radius of the target cage and the maximum turning angle of the unmanned feeding vessel, and calculate the berthing start distance; Obtain the real-time distance between the unmanned feeding vessel and the target net cage in the relative position of the vessel and the cage; If the real-time distance is reduced to less than or equal to the docking start distance, then determine whether the docking start condition is met and output the docking start command; In response to the berthing start command, real-time flow direction data is acquired, and the berthing point is the tangent point between the current bow direction of the unmanned feeding vessel and the outer circle of the target net cage. The outer circle of the target net cage is concentric with the target net cage and its radius is a preset distance larger than the radius of the target net cage. During the berthing and approach process, the updated flow direction data is acquired in real time, and the position of the berthing point is dynamically updated according to the relative relationship between the heading and the flow direction.

[0012] In one embodiment, the step of executing the feeding task according to a preset feeding strategy after determining that the hull position meets the feeding conditions includes: Obtain a feeding permission signal, which is generated after the unmanned feeding vessel is within a predetermined feeding range from the target cage and has been sailing stably for a preset time threshold. In response to the feeding permission signal, obtain the current target cage number information and query the pre-stored cage feeding plan table; Extract the preset feeding amount and feed type corresponding to the current target net cage from the net cage feeding plan table; Based on the preset feeding amount and feed type, the feeding equipment is controlled to perform precise feeding operations.

[0013] In one embodiment, the step of controlling the airborne monitoring drone to collect fish feeding data and optimizing the feeding strategy based on the fish feeding data includes: Obtain a feeding completion signal, and in response to the feeding completion signal, generate a drone exit command and send it to the airborne monitoring drone; Control the airborne monitoring drone to fly to and hover over the target cage that has been fed; Acquire images of fish feeding behavior captured by a high-definition camera mounted on the airborne monitoring drone; Extract the characteristics of the intensity of feeding and the amount of feed stored from the images of the fish feeding behavior to generate the current target cage number and the feeding effect evaluation data of the current feeding batch; The feeding strategy parameters corresponding to the target cage are updated based on the feeding effect evaluation data. The feeding strategy parameters include the feeding amount adjustment value and the feeding frequency adjustment value for the next feeding.

[0014] Furthermore, to achieve the above objectives, this application also proposes an unmanned feeding device for net cages, the unmanned feeding device for net cages comprising: The task planning module is used to obtain the feeding task of the target cage, plan the course of the unmanned feeding vessel according to the current environmental flow, and control the unmanned feeding vessel to sail towards the target cage in a top-current state. The berthing decision module is used to identify the target cage through sensors and obtain the relative position of the ship and cage by combining the position prediction information when the unmanned feeding vessel sails to the preset area, and determine the berthing point based on the relative position of the ship and cage. The execution control module is used to control the unmanned feeding vessel to dock at the berthing point, and after determining that the vessel's position meets the feeding conditions, to execute the feeding task according to the preset feeding strategy. The strategy optimization module is used to control the airborne monitoring drone to collect fish feeding data after the feeding operation is completed, and to optimize the feeding strategy based on the fish feeding data. The task loop module is used to control the onboard monitoring drone to recover and plan its flight path to the next target cage until all feeding tasks are completed.

[0015] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the unmanned feeding operation method for net cages as described above.

[0016] This application proposes a method, apparatus, and product for unmanned feeding operations in fish cages. The method includes: acquiring the feeding task of the target fish cage; planning the course and trajectory of the unmanned feeding vessel according to the current environmental flow direction; controlling the unmanned feeding vessel to sail towards the target fish cage in a top-down state; when the unmanned feeding vessel sails to a preset area, identifying the target fish cage through sensors and obtaining the relative position of the vessel and the cage by combining position prediction information; determining the docking point based on the relative position of the vessel and the cage; controlling the unmanned feeding vessel to dock at the docking point; and, after determining that the vessel's position meets the feeding conditions, executing the feeding task according to a preset feeding strategy; if the feeding operation is completed, controlling an airborne monitoring drone to collect fish feeding data and optimizing the feeding strategy based on the fish feeding data; controlling the airborne monitoring drone to recover and plan the navigation trajectory to the next target fish cage, until all feeding tasks are completed. This application employs a technical approach that uses environmental flow direction planning to determine the top current navigation trajectory, integrates sensor and location prediction information to determine the berthing point, and utilizes drones for real-time monitoring and feedback. This approach solves the problems of high labor intensity, high safety risks, reliance on experience leading to poor accuracy, and low operational efficiency in existing manual feeding methods. It achieves fully unmanned and intelligent feeding operations, effectively reduces labor costs, significantly improves operational efficiency and feeding accuracy, and can dynamically optimize feeding strategies based on fish feeding data. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating an embodiment of the unmanned feeding method for fish cages in this application. Figure 2 This is a flowchart illustrating Embodiment 2 of the unmanned feeding operation method for fish cages in this application; Figure 3 The trajectory optimization diagram of the artificial potential field method provided for the unmanned feeding operation method of the net cage in this application; Figure 4 This is a flowchart illustrating Embodiment 3 of the unmanned feeding operation method for fish cages in this application; Figure 5 A diagram illustrating the mooring of a net cage for the unmanned feeding operation method of this application; Figure 6 This is a schematic diagram of the module structure of the unmanned feeding device for net cages according to an embodiment of this application.

[0020] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0022] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0023] Existing manual feeding methods are inefficient and unsafe, and cannot meet the aquaculture industry's demand for unmanned, precise, and efficient feeding operations.

[0024] This application provides a solution that uses technologies such as planning the top current navigation trajectory based on environmental flow direction, fusing sensor and position prediction information to determine the berthing point, and utilizing drones for real-time monitoring and feedback. This solves the problems of high labor intensity, high safety risks, poor accuracy due to reliance on experience, and low operational efficiency in existing manual feeding methods. It realizes the unmanned and intelligent operation of the entire feeding process, effectively reduces labor costs, significantly improves operational efficiency and feeding accuracy, and can dynamically optimize feeding strategies based on fish feeding data.

[0025] Based on this, this application provides a method for unmanned feeding of net cages, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the unmanned feeding method for fish cages in this application.

[0026] In this embodiment, the unmanned feeding operation method for the net cage includes steps S10 to S50: Step S10: Obtain the feeding task of the target cage, plan the course of the unmanned feeding vessel according to the current environmental flow, and control the unmanned feeding vessel to sail towards the target cage in a top-current state.

[0027] It should be noted that the target cage is the specific marine aquaculture facility for which feed needs to be delivered in this operation plan; the feeding task refers to the set of operation instructions that include the geographical coordinates, number, preset feeding amount, and feed type of the target cage; the environmental current direction is the current direction and speed of the seawater flow in the current sea area; the course trajectory refers to the predetermined navigation route of the unmanned feeding vessel from the starting point to the target cage; the head-on state is the navigation attitude of the unmanned feeding vessel, in which its bow faces the opposite direction of the water flow during navigation.

[0028] This embodiment senses the dynamic changes in the marine environment, plans a safe and efficient path that can overcome water flow interference, and controls the ship to sail in a top-current attitude, using the resistance of the water flow to stabilize the ship's attitude and reduce drift.

[0029] In one possible implementation, environmental current direction can be acquired in real time by an acoustic Doppler current profiler (ADCP) installed on the bottom of the unmanned feeding vessel, or by obtaining regional ocean current forecast data through a communication module with a marine weather station, in order to plan longer-term paths.

[0030] In one specific implementation, the system receives a feeding task instruction to proceed to cage A-05. At this time, the ship's onboard current speed sensor detects that the current current direction is due north and the current speed is 1.5 m / s. The system plans an arc-shaped course that points towards cage A-05 while keeping the bow of the ship facing due south (i.e., against the current) and controls the power system to adjust the propeller speed and rudder angle so that the unmanned feeding vessel can maintain a stable attitude against the current and travel towards the target cage.

[0031] Step S20: When the unmanned feeding vessel sails to the preset area, it identifies the target cage through sensors and obtains the relative position of the vessel and cage by combining the position prediction information, and determines the berthing point based on the relative position of the vessel and cage.

[0032] It should be noted that the preset area is a sea area within a specific radius centered on the predicted location of the target cage, used to trigger the vessel to switch from cruise mode to precision operation mode; the location prediction information is the geographical coordinates of the target cage at the current moment, calculated based on historical operation data, sea state parameters, and cage mooring characteristics; the ship-cage relative position refers to the real-time distance, azimuth, and relative elevation difference between the unmanned feeding vessel and the target cage; the berthing point is the optimal location coordinate for the unmanned feeding vessel to finally stop and carry out feeding operations around the target cage.

[0033] This embodiment aims to solve the problem of unstable location caused by the drift of gravity cages at sea due to wind and waves. By combining real-time sensor perception with predictive models, the actual location of the cages can be accurately located, thereby selecting a safe, stable and convenient mooring point for feeding, and ensuring that feeding operations can be carried out accurately.

[0034] In one possible implementation, the location prediction information is obtained by using a Kalman filter algorithm to fit and predict the historical drift trajectory of the cage.

[0035] In one specific implementation, when the unmanned feeding vessel enters a preset area within 100 meters of the center point of the target net cage, the lidar at the bow of the vessel is activated to scan the floating pipe structure of the net cage. At the same time, the system calls the prediction model to know that the net cage has shifted 2.5 meters to the northeast due to the influence of the southwest current. The system performs weighted fusion of the point cloud data detected by the radar and the predicted shift to calculate the precise relative position of the vessel and the net cage, and selects the location 3 meters downstream of the net cage as the optimal berthing point.

[0036] Step S30: Control the unmanned feeding vessel to dock at the berthing point, and after determining that the vessel's position meets the feeding conditions, execute the feeding task according to the preset feeding strategy.

[0037] It should be noted that the hull position refers to the real-time coordinates and heading attitude of the unmanned feeding vessel in the geographic coordinate system or relative to the target net cage coordinate system; the feeding conditions refer to the pre-set safety thresholds and status standards that allow the feeding equipment to be started, such as the relative distance error range, the hull roll amplitude threshold, or the length of time to maintain stability; the preset feeding strategy refers to the set of rules for the feed feeding rate, total amount, and feeding pattern formulated for different aquaculture species, growth stages, water temperature conditions, or weather conditions.

[0038] This embodiment aims to ensure the safety and effectiveness of feeding operations. Feeding can only be carried out when the vessel's attitude is stable and its position is accurate, avoiding inaccurate feed placement or equipment damage due to hull movement. Precise feeding is conducted according to a scientific strategy to improve feed utilization. When a feeding command is triggered, the motor speed and discharge port opening of the feeding mechanism are controlled according to the strategy.

[0039] In one specific implementation, the system controls the unmanned feeding vessel to adjust its power output and uses the side thruster to correct the vessel's position, so that the vessel is precisely moored at the mooring point 2.5 meters away from the edge of the net cage. At this time, the monitoring system determines that the position error of the vessel is within 0.3 meters and the sway amplitude of the vessel in the waves is less than 3 degrees. This stable state lasts for 6 seconds. The system determines that the feeding conditions are met and then starts the feeder according to the preset feeding strategy to carry out the feeding operation at a rate of 50 kilograms per minute.

[0040] Step S40: If the feeding operation is completed, control the airborne monitoring drone to collect fish feeding data and optimize the feeding strategy based on the fish feeding data.

[0041] It should be noted that airborne monitoring drones are unmanned aerial vehicles mounted on unmanned feeding boats, capable of autonomous take-off, hovering flight, and image acquisition; fish feeding data reflects the activity status, aggregation degree, water surface wave characteristics, and feed remaining status of fish during the feeding process; optimizing feeding strategies refers to the process of adjusting and correcting the feeding amount, feeding duration, feeding frequency, or feed ratio based on the actual feeding feedback of the fish.

[0042] This embodiment automatically launches a drone for aerial inspection after feeding, collecting visual data and transmitting it back to the processing unit. Image recognition algorithms analyze the fish's activity level and uneaten feed, generating strategy adjustment parameters. This embodiment aims to construct a closed-loop feedback mechanism for feeding operations, utilizing the drone's high-altitude perspective to capture details of fish feeding that are difficult to observe manually. Intelligent analysis evaluates the effectiveness of the current feeding, dynamically adjusting the aquaculture strategy to achieve refined aquaculture, improve feed conversion rates, and reduce water pollution.

[0043] In one specific implementation, after the feeding operation is completed, the control system issues a command to open the hangar door, and the drone automatically takes off and hovers 20 meters above the net cage to take pictures. The images transmitted back by the high-definition camera show that the fish are feeding fiercely and there is no leftover food on the water surface. Based on this, the system determines that the current feeding amount is appropriate and maintains the next feeding amount in the optimized feeding strategy. If the image shows that the fish are feeding slowly and the feed is scattered, the strategy will automatically reduce the next feeding amount by 10%.

[0044] Step S50: Control the airborne monitoring drone to recover and plan the flight path to the next target cage until all feeding tasks are completed.

[0045] It should be noted that planning the navigation trajectory to the next target cage refers to calculating the optimal path from the current work point to the next cage to be fed in the task list, taking into account factors such as the shortest distance, obstacle avoidance safety, and requirements for working against the current; all feeding tasks refer to the list of all cages that need to be fed in this operation plan.

[0046] In this embodiment, after completing the feeding and feedback of the current cage, the drone equipment is retrieved first to ensure flight safety. Then, the task list is checked for any unexecuted targets. If so, a path planning algorithm is activated to generate a heading to the next target. If all targets have been completed, a return or standby path is planned. This solution achieves automated workflow for continuous multi-cage operations, improves operational efficiency, ensures the complete execution of the work sequence, reduces manual intervention, and enables all-weather unmanned operation.

[0047] In one possible implementation, the recovery process can employ visual identification-based precise guidance and positioning to assist the drone's landing, or it can use a robotic arm for grasping and a damping net for recovery, to adapt to the swaying conditions of the sea surface.

[0048] In one specific implementation, after the UAV completes data collection and transmission, the system generates a return guidance signal based on the real-time position of the unmanned feeding vessel, controls the UAV to land safely on the helipad at the stern of the vessel, the mechanical locking device automatically locks the UAV and starts charging, then the system searches the task list and finds that cage B-12 has not yet been fed, immediately plans a top-current navigation trajectory from the current position to cage B-12, and controls the unmanned feeding vessel to go to the next work point.

[0049] Furthermore, the steps of acquiring the feeding task of the target cage, planning the course and trajectory of the unmanned feeding vessel according to the current environmental current, and controlling the unmanned feeding vessel to sail towards the target cage in a top-down state include: If the current feeding task is determined to be the first cage in the operation sequence, the initial course trajectory is planned according to the preset cage number and position priority; If the determination result is that it is not the first cage, the course trajectory will be dynamically planned based on the current position of the hull and the already executed navigation route; Among them, the flow direction feedback data based on the flow direction sensor controls the bow of the unmanned feeding vessel to be opposite to the water flow direction, generating a heading trajectory from the current position to the target cage and sailing in a top-current attitude.

[0050] It should be noted that, in this embodiment, the unmanned feeding vessel is a surface operation robot with autonomous navigation capabilities; the top-current state refers to the vessel's bow facing the opposite direction of the water flow; the first net cage is the first feeding target in the operation sequence; the position priority is the access order set according to geographical proximity; the initial course trajectory refers to the path from the starting point to the first target net cage; non-first net cages refer to other feeding targets in the operation sequence besides the first one; flow direction feedback data refers to the real-time water flow information collected and output by the sensors; water flow direction refers to the specific direction of water flow.

[0051] This embodiment distinguishes the status of tasks within the operational sequence, flexibly selects path planning strategies, and combines real-time hydrological information to plan an optimal route that meets both operational sequence requirements and ensures vessel navigation stability, thereby improving the efficiency and safety of feeding operations. The system continuously receives feedback data from flow direction sensors, calculates the current vector in real time, and adjusts the bow of the unmanned feeding vessel to always maintain the opposite direction of the current. Ultimately, it generates a course trajectory that guides the unmanned feeding vessel to stably approach the target net cage with a head-on attitude against the current, effectively counteracting the effects of current drift.

[0052] In one possible implementation, the preset cage numbers and location priorities can be determined by finding the shortest path based on the straight-line distance between cages, or by setting time priorities based on the feeding time window of the farmed fish. Meanwhile, the flow direction sensor can be a shipborne acoustic Doppler current profiler (ADCP) or a marine monitoring buoy connected via a communication module to obtain water flow data over a wider area.

[0053] In one specific implementation, the system obtains the daily feeding task list and first determines that the target cage to be operated on is numbered A01. After comparison, it is confirmed as the first cage in the operation sequence. At this time, based on the coordinates of each cage in the preset map data, the system plans an initial course trajectory from the starting base station to cage A01 according to the principle of proximity. During the voyage, the shipborne current direction sensor monitors the current marine environment in real time. The current direction is from north to south, and the current speed is 1.2 m / s. The system calculates the top current vector and controls the bow of the unmanned feeding vessel to be adjusted to face north. The vessel then sails south along the planned trajectory to cage A01 with a top current attitude. Subsequently, after the task of cage A01 is completed, the system determines that the next target cage B05 is not the first cage. Based on the real-time position coordinates of the unmanned feeding vessel around cage A01, combined with the navigation route data from A01 to A02 that has been executed previously, an optimal adjustment trajectory to cage B05 is dynamically planned. The vessel continues to maintain a top-current navigation state based on real-time flow feedback to ensure that the vessel always resists the influence of the water flow smoothly.

[0054] Furthermore, referring to Figure 2 The second embodiment of the unmanned feeding operation method for fish cages in this application provides a flowchart, based on the above. Figure 2 The embodiment shown further refines the step S10, "controlling the unmanned feeding vessel to navigate towards the target cage in a top-current state," by including steps A201-A202: Step A201: The heading trajectory is optimized using the artificial potential field method. An gravitational potential field is constructed around the target cage, and a repulsive potential field is constructed in the non-target area.

[0055] It should be noted that, in this embodiment, the artificial potential field method is a path planning algorithm that simulates the robot's motion in the environment as motion under force in a virtual force field. It aims to treat the target point as a gravitational source that generates attraction and obstacles as a repulsive source that generates repulsion. The gravitational potential field is a virtual force field region centered on the target net cage that attracts the unmanned feeding vessel within a certain range, and its gravitational force is usually inversely proportional to the distance. The non-target area is the sea area in the navigation environment other than the location of the target net cage, which includes obstacles, boundaries or other unsafe areas. The repulsive potential field is a virtual force field region constructed around obstacles in the non-target area that repels the unmanned feeding vessel to avoid collision.

[0056] This embodiment establishes a gravitational potential field at the location of the target net cage based on electronic nautical charts or real-time environmental data, generating a continuous pulling force on the unmanned feeding vessel. Simultaneously, a repulsive potential field is established around known obstacles or predicted risk areas outside the target region. When the unmanned feeding vessel approaches these areas, it generates thrust. By calculating the vector sum of the gravitational and repulsive forces, the unmanned feeding vessel is guided to adjust its course with minimal resistance. By introducing an artificial potential field method to smooth and optimize the initially planned course trajectory, the problem of local optima, tortuous paths, or untimely obstacle avoidance that may exist in traditional path planning is solved. This results in planning an optimal path that safely avoids obstacles and smoothly approaches the target.

[0057] like Figure 3 The diagram shows a trajectory optimization diagram using an artificial potential field method. The potential function of the gravitational potential field can be a quadratic function or a conical function to adjust the rate at which gravity changes with distance. The construction of the repulsive potential field not only considers the distance factor, but can also dynamically adjust the range of repulsive force based on the danger level of obstacles or the current sailing speed of the unmanned feeding vessel. For example, the faster the speed, the larger the range of repulsive force is set to ensure sufficient braking distance.

[0058] In one specific implementation, based on environmental detection data, a gravitational potential field is constructed at the center point of the target net cage at coordinates (X1, Y1), with a gravitational coefficient set to 10. Simultaneously, the system detects a reef obstacle at coordinates (X2, Y2) on the preset route and constructs a repulsive potential field within a 50-meter radius around the reef, with a repulsive coefficient set to 15. During navigation, the unmanned feeding vessel is propelled forward by the gravitational pull from the target net cage. When it approaches the reef area and enters the repulsive potential field, the repulsive force rapidly increases, forcing the unmanned feeding vessel to automatically veer to the left to avoid the reef. Subsequently, under the influence of gravity, it smoothly returns to the route pointing towards the target net cage.

[0059] Step A202: Adjust the potential field parameters according to the relationship between the bow direction of the unmanned feeding vessel and the direction of the water flow, and control the unmanned feeding vessel to automatically adjust its bow direction to the target cage under the action of the resultant force.

[0060] It should be noted that in this embodiment, the potential field parameters are the coefficients, gain or attenuation factors, etc., used in the artificial potential field method to calculate the magnitude of attraction and repulsion; the resultant force is the vector sum of the target attraction and obstacle repulsion experienced by the unmanned feeding boat in the virtual force field; pointing towards the target cage is the orientation of the boat's bow directly aligned with or tending towards the target cage.

[0061] This embodiment monitors the angular relationship between the current heading of the unmanned feeding vessel and the actual water flow direction in real time. When the water flow is detected causing the vessel to yaw or lose its heading, the potential field parameters are adjusted (e.g., increasing the gravitational component pointing towards the target net cage, or adjusting the magnitude of the lateral repulsive force) to change the direction and magnitude of the resultant force acting on the vessel. This resultant force drives the rudder and propulsion systems of the unmanned feeding vessel, automatically correcting the bow and ensuring it always points towards the target net cage while approaching it with a head-on attitude. This application aims to incorporate environmental current factors into the control loop of path planning, dynamically adjusting potential field parameters to counteract the interference of the water flow on the vessel's attitude, ensuring that the unmanned feeding vessel can maintain stability while sailing against the current in complex sea conditions and accurately align with the target net cage.

[0062] In one specific implementation, as the unmanned feeding vessel approaches the target net cage, it encounters a crosscurrent flowing from left to right, forcing the bow to veer to the right. At this point, the line connecting the bow and the target net cage deviates by 15 degrees. Upon detecting this, the system determines that the restoring torque needs to be increased. Therefore, it decreases the attenuation coefficient of the repulsive potential field and increases the coefficient of the gravitational potential field pointing towards the target net cage. The adjusted resultant force vector changes direction, generating a leftward corrective force. In response to this resultant force, the system automatically drives the rudder to steer to the left. Under the action of the resultant force, the unmanned feeding vessel gradually returns to its original position, with its bow pointing towards the target net cage and sailing against the current.

[0063] Furthermore, the step of identifying the target cage through sensors and obtaining the relative position of the ship and cage by combining position prediction information includes A301-A303: Step A301: Use a camera to identify the target cage to determine its direction, and use radar to obtain the distance to the target cage.

[0064] It should be noted that direction refers to the azimuth or angle information of the target net cage relative to the unmanned feeding vessel; the measured distance refers to the straight-line distance between the unmanned feeding vessel and the target net cage calculated by the radar through transmitting and receiving signals.

[0065] In this embodiment, the camera is controlled to scan the sea area in front, and the features of the target cage (such as floating pipes and frames) are extracted from the background through image recognition algorithms. The pixel position of the target in the image is calculated and then converted into azimuth angle. At the same time, the radar is controlled to emit detection waves, and the distance between the two is determined based on the echo time or phase change, thereby initially constructing the relative spatial position relationship between the unmanned feeding vessel and the target cage.

[0066] In one possible implementation, the camera can be a wide-angle high-definition camera to expand the field of view, or an infrared camera with night vision capabilities to adapt to low-light environments; the radar can be a lidar to obtain high-precision point cloud distance data, or a millimeter-wave radar to adapt to the detection needs in rainy and foggy weather.

[0067] Step A302: Call the pre-built cage location model and input the current environmental parameters to obtain the predicted location of the target cage.

[0068] It should be noted that the pre-built cage location model refers to an algorithm model or database based on historical data and physical laws, used to calculate or estimate the spatial location of the cage under different environmental conditions; the current environmental parameters refer to real-time environmental data that affects the cage's location drift, including wind speed, wind direction, current speed, current direction, and wave height; the predicted location is the theoretical geographical coordinate of the target cage, calculated based on the environmental parameters and the cage location model.

[0069] This embodiment can acquire current wind and current data in real time from meteorological sensors or marine monitoring networks, input these parameters into a pre-trained cage position model, and the model, combined with the characteristics of the cage's mooring system, calculates the cage's offset under the action of environmental forces, thereby outputting the predicted position.

[0070] Additionally, it should be noted that the cage location model is a neural network model trained on a large amount of historical drift data, which can fit the relationship between environmental parameters and location changes through deep learning.

[0071] In one specific implementation, the system detects that the current sea area is affected by a southwest wind with a wind speed of level 6, accompanied by a northeast current of 1.2 m / s. By inputting these current environmental parameters into a pre-built cage location model, the model calculates and analyzes to predict that the target cage will shift approximately 3.5 meters northeastward relative to its designated anchorage point under the combined influence of wind and current, and thus outputs the predicted location coordinates of the target cage that include this shift.

[0072] Step A303: The measured distance and the predicted position are fused using a Kalman filter algorithm to output a high-precision real-time distance between the ship and the container.

[0073] It should be noted that the Kalman filter algorithm is a recursive filtering algorithm that uses data-driven and state estimation methods to make optimal estimates of the system state through measurement data containing noise; the high-precision real-time distance between the ship and the target net cage is the real-time relative distance information between the unmanned feeding ship and the target net cage obtained after fusion processing with minimal error.

[0074] This embodiment aims to combine the accuracy of sensor measurements with the regularity of model predictions, eliminate random noise in single sensor data and cumulative errors in model calculations, and output high-confidence relative position data to provide a reliable basis for accurate berthing.

[0075] In one possible implementation, the Kalman filter algorithm can employ standard Kalman filtering for linear systems, or extended Kalman filtering (EKF) and unscented Kalman filtering (UKF) to adapt to highly nonlinear marine environment motion models.

[0076] In one specific implementation, the radar measured the distance to the target cage as 80.5 meters, while the model predicted the location, which was converted to a distance of 79.2 meters. The Kalman filter algorithm was used to fuse these two values. Considering that the current wind and waves were large, causing fluctuations in radar measurements, while the model prediction was relatively stable, the algorithm gave the model prediction a slightly higher weight. After calculation, the high-precision real-time distance of the ship cage was output as 79.6 meters, which is closer to the actual situation than relying solely on radar or solely on the model.

[0077] This application utilizes cameras and radar to acquire the direction and measured distance of the target net cage, and combines this with a net cage position model based on environmental parameters to obtain the predicted position. A Kalman filter algorithm is then used to fuse the measured distance and predicted position, thereby outputting a high-precision real-time distance between the ship and the cage. This scheme, by fusing multi-sensor measured data with environmental model prediction information, effectively compensates for the inaccuracies of single-sensor measurements in complex sea conditions and the biases in model predictions, significantly improving the measurement accuracy, robustness, and environmental adaptability of the relative position of the ship and the cage.

[0078] Furthermore, the steps for constructing the cage location model include A401-A402: Step A401: Collect the latitude and longitude location of the net cage and the corresponding environmental parameters at each feeding operation, and establish a net cage location map; Step A402: Train the location map of the cage using a neural network model to obtain the mapping relationship between environmental parameters and cage location, and output the predicted location of the target cage by inputting real-time environmental parameters.

[0079] It should be noted that, in this embodiment, the latitude and longitude position of the net cage refers to the absolute coordinates of the net cage in the Earth coordinate system; the corresponding time refers to the specific time point when the data was collected; environmental parameters refer to meteorological and hydrological data that affect the floating state of the net cage, including wind speed, wind direction, current speed, current direction, and wave height; the net cage location map is a data set formed by associating the net cage location data recorded in historical operations with the environmental parameters at the same time; training refers to the process of adjusting the internal weight parameters of the neural network using a large amount of historical sample data so that it can fit the nonlinear relationship between the input and output; the mapping relationship refers to the functional correspondence between the change in environmental parameters and the drift of the net cage position; and the real-time environmental parameters refer to the instantaneous wind field and flow field data acquired by the sensor at the current moment.

[0080] In this embodiment, during the long-term execution of feeding operations, historical data including location and environment are continuously accumulated to build a rich map of cage locations. The nonlinear fitting capability of neural networks is used to uncover the environment-location correlation patterns hidden behind the data. Through training, a mature prediction model is formed. In practical applications, only the current real-time environmental parameters need to be input to quickly output the predicted location of the target cage based on the learned mapping relationship.

[0081] In one possible implementation, the neural network model can employ a Long Short-Term Memory (LSTM) network to better capture the impact of the time-series features of environmental parameters on the cumulative drift of the cage's position. The LSTM includes an input layer that receives a sequence of multidimensional environmental parameters arranged in chronological order, along with the historical position data of the cage from the previous time step. The LSTM hidden layer consists of a series of memory units connected in series. Each memory unit contains a forget gate (determining which information to discard from the unit state), an input gate (determining which new environmental parameters to store in the unit state), an output gate (determining what value to output as the hidden state for the next time step), and a cell state (responsible for retaining the cumulative displacement of the cage over time). A fully connected output layer maps the vector containing time-dependent features output by the LSTM hidden layer to the final prediction result, i.e., the predicted latitude and longitude coordinates of the target cage or its offset relative to the anchor point. Alternatively, a backpropagation (BP) neural network combined with a genetic algorithm can be used for weight optimization to improve the model's convergence speed and prediction accuracy. The BP neural network includes an input layer (receiving real-time environmental parameters), hidden layers (one or more neurons using activation functions such as Sigmoid or ReLU to perform nonlinear transformations on the input data and extract high-dimensional features), and an output layer (outputting the predicted target cage location coordinates). Specifically, a genetic algorithm is first used to roughly search for the optimal weight parameters globally, which are then assigned to the BP network as initial values. The BP algorithm (gradient descent) is then used for local fine-tuning training, thereby quickly and accurately obtaining the mapping relationship between environmental parameters and cage location.

[0082] In one specific implementation, after each feeding task is completed, the system automatically records the confirmed GPS latitude and longitude coordinates of the net cage during that operation, and simultaneously records environmental data such as wind direction (northeast), wind speed (level 5), current speed (1.2 m / s), and current direction (north). These paired data are stored to form a massive net cage location map. Subsequently, the system uses a BP neural network algorithm to iterate and train this net cage location map thousands of times, establishing a mapping model between environmental parameters and net cage positions. In one operation, when the sensor detects that the current sea surface wind direction has changed to southeast and the current speed has increased to 1.5 m / s, the system inputs this set of real-time environmental parameters into the trained model. The model quickly determines through internal calculations that the net cage will drift to the northwest to a certain extent and outputs a corrected predicted position of the target net cage. The unmanned feeding vessel then docks based on this position, achieving high-precision positioning.

[0083] Furthermore, referring to Figure 4 The third embodiment of the unmanned feeding operation method for fish cages in this application provides a flowchart, based on the above. Figure 4 The embodiment shown further refines the step of "determining the berthing point based on the relative position of the ship's container" in step S20, including steps A501 to A504: Step A501: Obtain the radius of the target net cage and the maximum turning angle of the unmanned feeding vessel, and calculate the berthing start distance.

[0084] It should be noted that the radius refers to the shortest distance from the geometric center of the target net cage to its edge. If it is a square net cage, it refers to the equivalent radius or a specific side length parameter. The maximum turning angle refers to the maximum rate of change of angle or the limit turning angle that the unmanned feeding vessel can achieve by adjusting its course through the rudder or propeller at the current speed. The berthing start distance refers to the distance threshold at which the unmanned feeding vessel must start deceleration or turning operations at a specific distance from the target net cage in order to safely and smoothly transition to the berthing attitude.

[0085] This embodiment is used to scientifically calculate the timing of initiating the berthing operation based on the physical dimensions of the target cage and the motion performance of the unmanned feeding vessel, to prevent collisions caused by insufficient turning due to excessive distance, or to prevent operational efficiency from being affected by excessive distance.

[0086] In one possible implementation, the berthing start distance is:

[0087] Where d is the radius of the net cage and θ is the maximum turning angle of the unmanned feeding vessel.

[0088] Step A502: Obtain the real-time distance between the unmanned feeding vessel and the target net cage in the relative position of the vessel and the cage.

[0089] It should be noted that the real-time distance is the linear distance between the reference point of the unmanned feeding vessel (such as the center of the vessel or the position of the GPS antenna) and the reference point of the target net cage (such as the center or edge of the net cage) at the current moment. In this embodiment, the distance parameter can be extracted from the data packets of the relative position of the vessel and the net cage by continuously reading the data packets of the relative position of the vessel and the net cage through the sensor module.

[0090] Step A503: If the real-time distance is reduced to less than or equal to the docking start distance, determine whether the docking start condition is met and output the docking start command.

[0091] It should be noted that, in addition to meeting the distance requirement, berthing initiation conditions must also meet auxiliary conditions to ensure berthing safety, such as whether the ship's speed has decreased to a preset range and whether the power system is functioning normally. This embodiment aims to accurately capture the optimal intervention time for berthing; the berthing process can only be initiated when the distance requirement is met and other conditions are also permissible, thereby avoiding misoperation.

[0092] Specifically, the real-time distance is compared with the berthing start distance to determine whether the real-time distance has narrowed to less than or equal to the berthing start distance. If the result is yes, the current ship status is further checked to see if it meets the berthing start conditions. If all conditions are met, a berthing start command is generated and output.

[0093] In one possible implementation, the berthing initiation conditions may also include determining whether the current sea state level is within the permissible range for operation, or determining whether there are dynamic obstacles such as interfering vessels around the target cage.

[0094] Step A504: In response to the berthing start command, real-time flow direction data is obtained, and the berthing point is the tangent point between the current bow direction of the unmanned feeding vessel and the outer circle of the target net cage. The outer circle of the target net cage is concentric with the target net cage and its radius is a preset distance larger than the radius of the target net cage.

[0095] It should be noted that real-time flow data refers to the current flow direction and velocity information of seawater; bow direction is the geographical azimuth of the bow of the unmanned feeding vessel; the outer circle of the target cage is a virtual circular geometric boundary with the center of the target cage as the center and a radius larger than the actual radius of the target cage; the tangent point is the geometric point where the bow ray of the unmanned feeding vessel is tangent to the outer circle of the target cage; the preset distance is a buffer interval added to the actual radius of the target cage for operational safety.

[0096] The overall purpose of this step is to use the principle of geometric tangents, combined with the direction of water flow, to determine a precise berthing point that can ensure operational safety and is not subject to severe water flow interference.

[0097] like Figure 5The diagram shows a net cage mooring. In this embodiment, after receiving the instruction, the latest environmental data is immediately obtained. Using the current heading of the unmanned feeding vessel as a reference, a tangent point is found on the outer circle of the constructed virtual target net cage. This tangent point is used as the mooring point, so that the vessel can make smooth contact along the tangent direction when mooring. The design of the outer circle ensures that the vessel will not have a physical collision with the net cage.

[0098] In one possible implementation, the preset distance can be set according to the launching range of the feeding device, for example, set to 3 to 5 meters, or dynamically adjusted according to the current wave height; the higher the wave height, the greater the preset distance.

[0099] Step A505: During the berthing and approach process, the updated flow direction data is acquired in real time, and the position of the berthing point is dynamically updated according to the relative relationship between the heading and the flow direction.

[0100] It should be noted that the updated flow direction data is the latest water flow direction and speed information obtained by the system during the approach process, which is continuously or periodically refreshed; the relative relationship refers to the angle or azimuth difference between the heading and the flow direction.

[0101] This embodiment aims to address the dynamic changes in the marine environment, overcome the positional deviation caused by changes in water flow during berthing, and ensure that the unmanned feeding vessel can always accurately reach the predetermined position.

[0102] In one specific implementation, as the unmanned feeding vessel approaches the target cage, the monitoring system detects that the current direction suddenly changes from northeast to due east, causing a change in the original head-on attitude. The system immediately recalculates the tangent point position based on the updated current direction data and the current relative relationship between the bow and the current direction, corrects the position of the mooring point 2 meters eastward along the edge of the cage, and controls the unmanned feeding vessel to continue sailing to the new mooring point.

[0103] This application calculates the berthing start distance and combines it with real-time distance to determine the trigger for berthing commands. It uses real-time current data and the vessel's bow to determine the tangent point on the virtual outer circle of the target net cage as the berthing point, and dynamically updates the berthing point position during the approach based on the relative relationship between the bow and the current. This solution, through the setting of a virtual concentric outer circle and a dynamic update mechanism, effectively ensures safe collision avoidance during berthing, achieves adaptive adjustment to complex ocean current environments, and significantly improves the safety, accuracy, and stability of unmanned feeding vessel berthing operations.

[0104] Furthermore, the step of executing the feeding task according to the preset feeding strategy after determining that the hull position meets the feeding conditions includes A601-A604: Step A601: Obtain the feeding permission signal. The feeding permission signal is generated after the unmanned feeding vessel is within a predetermined feeding range from the target cage and has been sailing stably for a preset time threshold. Step A602: In response to the feeding permission signal, obtain the number information of the current target cage and query the pre-stored cage feeding plan table; Step A603: Extract the preset feeding amount and feed type corresponding to the current target cage from the cage feeding plan table; Step A604: Based on the preset feeding amount and feed type, control the feeding equipment to perform precise feeding operations.

[0105] It should be noted that in this embodiment, the predetermined feeding range is a circular operating area set with the center of the target cage as the center; continuous stable navigation is the motion state in which the unmanned feeding vessel maintains a stable course and the sway amplitude is below the allowable threshold within the predetermined range; the pre-stored cage feeding plan table refers to a data table in the database that records the types and amounts of feed corresponding to different numbered cages; the feeding equipment is a mechanical device installed on the ship to scatter feed into the cage.

[0106] This embodiment is used to establish a safe and precise feeding control process. By strictly monitoring the stability of the vessel before feeding, it avoids feed spilling outside the cage or equipment failure due to hull swaying. At the same time, it automatically matches the corresponding feed parameters according to the specific cage identity, so as to realize the refinement and automation of aquaculture management.

[0107] In one possible implementation, the preset time threshold can be a dynamic value that is automatically adjusted according to the current sea state level. For example, the threshold time can be appropriately extended when the sea state is bad to ensure absolute stability, while the time can be shortened when the sea state is good to improve efficiency. The cage feeding schedule can also be remotely updated in real time via a wireless communication module to adapt to flexible aquaculture needs.

[0108] Furthermore, the steps of controlling the airborne monitoring drone to collect fish feeding data and optimizing the feeding strategy based on the fish feeding data include A701-A705: Step A701: Obtain the feeding completion signal, generate a drone exit command in response to the feeding completion signal, and send it to the airborne monitoring drone; Step A702: Control the airborne monitoring drone to fly over the target cage that has been fed and hover there; Step A703: Obtain images of fish feeding behavior captured by the high-definition camera mounted on the airborne monitoring drone; Step A704: Extract the characteristics of the intensity of feeding and the characteristic value of the amount of feed stored from the image of the feeding behavior of the fish school, and generate the current target cage number and the feeding effect evaluation data of the current feeding batch; Step A705: Update the feeding strategy parameters corresponding to the target cage according to the feeding effect evaluation data. The feeding strategy parameters include the feeding amount adjustment value and the feeding frequency adjustment value for the next feeding.

[0109] It should be noted that, in this embodiment, the image of fish feeding behavior refers to visual data captured by a camera that reflects the activity of fish competing for food on or underwater; the intensity of fish feeding refers to indicators extracted from the image to characterize the level of fish's feeding activity, such as fish density, frequency of leaping out of the water, or swimming speed; the characteristic value of feed storage refers to an indicator obtained based on image analysis to characterize the amount of uneaten feed in the cage area; the current target cage number is the unique identification code of the monitored cage; the current feeding batch is the timestamp or sequence number of this feeding operation; the feeding effect evaluation data refers to a set of data that comprehensively evaluates whether the feeding was appropriate and the fish's appetite by combining the intensity of feeding and the amount of remaining feed; the feeding amount adjustment value refers to the amount of feed added or reduced based on the baseline feeding amount; and the feeding frequency adjustment value refers to the time value for extending or shortening the feeding interval.

[0110] This embodiment, by immediately dispatching an airborne drone to the target area for fixed-point observation after confirming the completion of the feeding task, uses computer vision technology to intelligently analyze the activity level of the fish and the amount of remaining feed from the captured images, thereby generating quantitative evaluation data. Based on this data, the system automatically adjusts the next feeding plan for the net cage, forming a closed-loop control of "execution-monitoring-evaluation-optimization". This solves the problem of traditional blind feeding not being able to adjust the strategy according to the actual appetite of the fish, and achieves efficient use of feed and precise management of the fish.

[0111] In one possible implementation, the characteristic value of feed storage can be estimated by identifying specific color distribution areas of feed on the water surface and their area proportions; in addition, the update of feeding strategy parameters can also be combined with trend analysis of historical data. For example, if the feeding intensity is low for several consecutive times, not only can the feeding amount be adjusted, but an early warning of abnormal fish health can also be triggered.

[0112] In one specific implementation, after the unmanned feeding vessel delivers feed to the target cage numbered "N-05" and sends a feeding completion signal, the system immediately generates a drone takeoff command, controlling the onboard monitoring drone to automatically take off and hover 5 meters above cage "N-05". The drone activates its high-definition camera to capture video streams, from which the system extracts images of fish feeding behavior. Image recognition algorithms analyze the images to determine that the fish are densely packed and frequently leaping out of the water, calculating the intensity of feeding as "high". Simultaneously, only a very small amount of feed is detected floating on the surface, indicating a "low" feed storage level. This generates feeding effect evaluation data containing cage number "N-05" and current batch information, determining that the fish have a strong appetite but the feed delivery is slightly insufficient. Based on this, the system automatically updates the feeding strategy parameters corresponding to cage "N-05", setting the next feeding amount to "increase by 2 kg" and maintaining the feeding frequency unchanged to ensure that the next feeding meets the fish's growth needs.

[0113] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the unmanned feeding operation method of the cages in this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0114] This application also provides an unmanned feeding device for fish cages; please refer to... Figure 6 The unmanned feeding device for the net cages includes: The task planning module 10 is used to obtain the feeding task of the target cage, plan the course trajectory of the unmanned feeding vessel according to the current environmental flow, and control the unmanned feeding vessel to sail towards the target cage in the top of the current. The berthing decision module 20 is used to identify the target cage through sensors and obtain the relative position of the ship and cage by combining the position prediction information when the unmanned feeding vessel sails to the preset area, and determine the berthing point based on the relative position of the ship and cage. The execution control module 30 is used to control the unmanned feeding vessel to dock at the berthing point, and after determining that the position of the vessel meets the feeding conditions, to execute the feeding task according to the preset feeding strategy. The strategy optimization module 40 is used to control the airborne monitoring drone to collect fish feeding data and optimize the feeding strategy based on the fish feeding data if the feeding operation is completed. The task loop module 50 is used to control the onboard monitoring drone to recover and plan the flight path to the next target cage until all feeding tasks are completed.

[0115] The unmanned feeding device for net cages provided in this application, employing the unmanned feeding method for net cages in the above embodiments, can solve the technical problems of low efficiency and poor safety of existing manual feeding methods, failing to meet the feeding needs of aquaculture for unmanned, precise, and efficient operations. Compared with the prior art, the beneficial effects of the unmanned feeding device for net cages provided in this application are the same as those of the unmanned feeding method for net cages provided in the above embodiments, and other technical features in the unmanned feeding device for net cages are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0116] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described unmanned feeding operation method for net cages.

[0117] The computer program product provided in this application can solve the technical problems of low efficiency and poor safety of existing manual feeding methods, which cannot meet the feeding needs of aquaculture for unmanned, precise and efficient operations. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the unmanned feeding operation method for net cages provided in the above embodiments, and will not be repeated here.

[0118] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for unmanned feeding of net cages, characterized in that, The method includes: Obtain the feeding task of the target cage, plan the course of the unmanned feeding vessel according to the current environmental flow, and control the unmanned feeding vessel to sail towards the target cage in a top-down state; When the unmanned feeding vessel sails to the preset area, it identifies the target cage through sensors and obtains the relative position of the vessel and cage by combining the position prediction information, and determines the berthing point based on the relative position of the vessel and cage. The unmanned feeding vessel is controlled to dock at the mooring point, and after determining that the vessel's position meets the feeding conditions, the feeding task is executed according to the preset feeding strategy. Once the feeding operation is complete, the onboard monitoring drone is controlled to collect data on the fish's feeding activity, and the feeding strategy is optimized based on this data. Control the airborne monitoring drone to recover and plan the flight path to the next target cage until all feeding tasks are completed.

2. The unmanned feeding operation method for net cages as described in claim 1, characterized in that, The steps of acquiring the feeding task of the target cage, planning the course and trajectory of the unmanned feeding vessel according to the current environmental current, and controlling the unmanned feeding vessel to sail towards the target cage in a top-down state include: If the current feeding task is determined to be the first cage in the operation sequence, the initial course trajectory is planned according to the preset cage number and position priority; If the determination result is that it is not the first cage, the course trajectory will be dynamically planned based on the current position of the hull and the already executed navigation route; Among them, the flow direction feedback data based on the flow direction sensor controls the bow of the unmanned feeding vessel to be opposite to the water flow direction, generating a heading trajectory from the current position to the target cage and sailing in a top-current attitude.

3. The unmanned feeding operation method for net cages as described in claim 2, characterized in that, The step of controlling the unmanned feeding vessel to navigate towards the target cage in a top-current state also includes: The course trajectory is optimized using the artificial potential field method, which constructs an gravitational potential field around the target cage and a repulsive potential field in the non-target area. The potential field parameters are adjusted according to the relationship between the bow direction of the unmanned feeding vessel and the direction of the water flow, so that the unmanned feeding vessel can automatically adjust its bow direction to point towards the target cage under the action of the resultant force.

4. The unmanned feeding operation method for net cages as described in claim 1, characterized in that, The step of identifying the target cage using sensors and obtaining the relative position of the cage by combining position prediction information includes: The target cage is identified by a camera to determine its direction, and the target cage is measured by radar. Call the pre-built cage location model and input the current environmental parameters to obtain the predicted location of the target cage; The measured distance and the predicted position are fused using a Kalman filter algorithm to output a high-precision real-time distance between the ship and the container.

5. The unmanned feeding operation method for net cages as described in claim 4, characterized in that, The steps for constructing the cage location model include: Collect the latitude and longitude location of the net cages and the corresponding environmental parameters at each feeding operation, and establish a net cage location map; The location map of the net cage is trained using a neural network model to obtain the mapping relationship between environmental parameters and net cage location. The predicted location of the target net cage is output by inputting real-time environmental parameters.

6. The unmanned feeding method for net cages as described in claim 1, characterized in that, The step of determining the berthing point based on the relative position of the ship's container includes: Obtain the radius of the target cage and the maximum turning angle of the unmanned feeding vessel, and calculate the berthing start distance; Obtain the real-time distance between the unmanned feeding vessel and the target net cage in the relative position of the vessel and the cage; If the real-time distance is reduced to less than or equal to the docking start distance, then determine whether the docking start condition is met and output the docking start command; In response to the berthing start command, real-time flow direction data is acquired, and the berthing point is the tangent point between the current bow direction of the unmanned feeding vessel and the outer circle of the target net cage. The outer circle of the target net cage is concentric with the target net cage and its radius is a preset distance larger than the radius of the target net cage. During the berthing and approach process, the updated flow direction data is acquired in real time, and the position of the berthing point is dynamically updated according to the relative relationship between the heading and the flow direction.

7. The unmanned feeding operation method for net cages as described in claim 1, characterized in that, The steps of executing the feeding task according to the preset feeding strategy after determining that the ship's position meets the feeding conditions include: Obtain a feeding permission signal, which is generated after the unmanned feeding vessel is within a predetermined feeding range from the target cage and has been sailing stably for a preset time threshold. In response to the feeding permission signal, obtain the current target cage number information and query the pre-stored cage feeding plan table; Extract the preset feeding amount and feed type corresponding to the current target net cage from the net cage feeding plan table; Based on the preset feeding amount and feed type, the feeding equipment is controlled to perform precise feeding operations.

8. The unmanned feeding operation method for net cages as described in claim 1, characterized in that, The steps of controlling the airborne monitoring drone to collect fish feeding data and optimizing the feeding strategy based on the fish feeding data include: Obtain a feeding completion signal, and in response to the feeding completion signal, generate a drone exit command and send it to the airborne monitoring drone; Control the airborne monitoring drone to fly to and hover over the target cage that has been fed; Acquire images of fish feeding behavior captured by a high-definition camera mounted on the airborne monitoring drone; Extract the characteristics of the intensity of feeding and the amount of feed stored from the images of the fish feeding behavior to generate the current target cage number and the feeding effect evaluation data of the current feeding batch; The feeding strategy parameters corresponding to the target cage are updated based on the feeding effect evaluation data. The feeding strategy parameters include the feeding amount adjustment value and the feeding frequency adjustment value for the next feeding.

9. A device for unmanned feeding of net cages, characterized in that, The unmanned feeding device for the cages includes: The task planning module is used to obtain the feeding task of the target cage, plan the course of the unmanned feeding vessel according to the current environmental flow, and control the unmanned feeding vessel to sail towards the target cage in a top-current state. The berthing decision module is used to identify the target cage through sensors and obtain the relative position of the ship and cage by combining the position prediction information when the unmanned feeding vessel sails to the preset area, and determine the berthing point based on the relative position of the ship and cage. The execution control module is used to control the unmanned feeding vessel to dock at the berthing point, and after determining that the vessel's position meets the feeding conditions, to execute the feeding task according to the preset feeding strategy. The strategy optimization module is used to control the airborne monitoring drone to collect fish feeding data after the feeding operation is completed, and to optimize the feeding strategy based on the fish feeding data. The task loop module is used to control the onboard monitoring drone to recover and plan its flight path to the next target cage until all feeding tasks are completed.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the unmanned feeding operation method for net cages as described in any one of claims 1 to 8.

Images