A control method of an unmanned ship for a mariculture operation

Abstract

The application belongs to the technical field of marine ranching monitoring, and particularly relates to a control method of an unmanned ship for marine ranching operation. The control method comprises the following steps: S1: acquiring marine environment, biological and non-biological information; S2: selecting a suitable strategy from a basic strategy library or a disturbance strategy library based on the acquired information and the marine ranching task executed by the unmanned ship; S3: executing the marine ranching operation task based on the selected strategy, wherein the marine ranching operation task comprises an inspection task, a feeding task and a detection task; and S4: cyclically executing steps S2-S3 until the marine ranching task of a target area is completed. Through information acquisition, dynamic strategy selection, task execution and cyclic adjustment, the application adapts to the inspection, feeding and detection tasks, copes with complex marine environments, improves operation accuracy, stability and efficiency, enhances system robustness and ensures reliable completion of the task.

Description

Technical Field

[0001] This application belongs to the field of marine ranching monitoring technology, specifically relating to a control method for an unmanned vessel used in marine ranching operations. Background Technology

[0002] In the field of marine ranching operations, the reliability and adaptability of unmanned surface vessel (USV) control systems directly affect operational efficiency, safety, and mission completion quality. Traditional systems often borrow environmental perception and positioning technologies from land-based autonomous driving, relying on multi-sensor fusion solutions such as high-precision maps, GPS, cameras, and LiDAR. However, these technologies have significant limitations in the marine ranching environment: the vast operating area makes high-precision map creation and updating extremely costly and difficult to maintain; GPS signals are easily interfered with in complex sea conditions or near islands and reefs, resulting in insufficient stability; visual information is sparse in the marine environment, and surface reflections, fog, and turbid water affect camera perception; LiDAR has a limited effective detection range, only meeting short-range obstacle avoidance needs and unable to support accurate perception and positioning for large-scale operations. Therefore, long-term navigation and operation of USVs rely more heavily on inertial navigation systems (INS), but these systems have cumulative errors, which can lead to a decline in positioning accuracy over long-term operation, affecting operational accuracy.

[0003] Existing technologies mostly employ reinforcement learning to train unmanned surface vessel (USV) control strategies. However, the sea conditions in marine ranching pose significant challenges. In calm seas, USVs experience minimal external disturbances and maintain stable motion, resulting in a limited range of training samples. This leads to reinforcement learning strategies being prone to overfitting to typical calm sea conditions, exhibiting insufficient learning ability and poor generalization for sudden disturbances such as wind, waves, and currents. When adapting to changing tidal conditions or sudden gusts, retraining the control strategy requires extremely high time and computational resources, making it difficult to meet real-time requirements. Furthermore, in high sea states (such as strong winds, waves, and rapid currents), USV motion is complex and difficult to accurately simulate. Reproducing such extreme scenarios in actual training is challenging, resulting in severely inadequate adaptability of the strategy under strong disturbances. Since marine ranching operations are mostly located nearshore or offshore, significantly affected by monsoons and tides, high sea states occur frequently. If the strategy cannot adapt, it will directly threaten operational safety and may even lead to equipment damage or mission failure.

[0004] In the actual deployment phase, unmanned surface vessels (USVs) face the dual challenges of policy robustness and transferability. On the one hand, existing policies, lacking sufficient training in complex sea conditions, experience a sharp decline in control performance stability when encountering unforeseen external disturbances, making it difficult to guarantee the reliability of core functions such as trajectory tracking accuracy and attitude control of operational equipment. On the other hand, traditional transfer learning methods, when used to improve the policy's adaptability to new sea conditions, are prone to "catastrophic forgetting," meaning that in the process of learning to adapt to new environments, control policies that performed well in calm sea conditions are disrupted, leading to performance degradation in routine operating environments and severely reducing the practicality and reliability of the control system. In summary, existing USV control systems for marine ranching operations have significant shortcomings in environmental perception adaptability, reinforcement learning policy generalization ability, adaptability to complex sea conditions, and policy transfer robustness. Therefore, a control method that can adapt to complex environments and possesses strong robustness and efficient transferability is urgently needed. Summary of the Invention

[0005] The purpose of this application is to provide a control method for unmanned vessels used in marine ranching that can adapt to complex environments, possess strong robustness and efficient migration capabilities.

[0006] The embodiments of this application can be implemented through the following technical solutions:

[0007] A control method for an unmanned surface vessel (USV) used in marine ranching operations, comprising the following steps:

[0008] S1: Acquire marine environmental, biological, and abiotic information;

[0009] S2: Based on the information obtained and the marine ranching task performed by the unmanned vessel, select a suitable strategy from the basic strategy library or the disturbance strategy library;

[0010] S3: Execute the marine ranching operation tasks based on the selected strategy, wherein the marine ranching operation tasks include inspection tasks, feeding tasks, and detection tasks;

[0011] S4: Repeat steps S2 to S3 until the marine ranching task in the target area is completed.

[0012] Further, step S2 includes the following steps:

[0013] S21: Determine whether the marine ranching operation task is an inspection task. If yes, proceed to step S22; otherwise, proceed to step S23.

[0014] S22: Update the basic policies in the basic policy library based on the information obtained, and input the updated basic policies into step S3;

[0015] S23: Determine whether the marine ranch task is a feeding task. If yes, proceed to step S24; otherwise, proceed to step S25.

[0016] S24: Based on the distance between the unmanned vessel and the aquaculture cage, select a suitable strategy from the frontal flow strategy and lateral disturbance zone strategy in the disturbance strategy library and input it into step S3;

[0017] S25: Based on the distance between the unmanned vessel and the aquaculture cage, select a suitable strategy from the upstream strategy, lateral disturbance zone strategy, and wake zone strategy in the disturbance strategy library and input it into step S3.

[0018] Furthermore, in steps S24 and S25:

[0019] The distance between the unmanned vessel and the upstream surface of the aquaculture cage is set with a first threshold and a second threshold, and the first threshold is less than the second threshold. When the distance between the unmanned vessel and the upstream surface of the aquaculture cage is less than the first threshold, the area is the speed-locking stability zone; when the distance between the unmanned vessel and the upstream surface of the aquaculture cage is greater than the first threshold and less than the second threshold, the area is the early deceleration zone.

[0020] The upstream strategy includes a first sub-strategy and a second sub-strategy. When the unmanned vessel is in the early deceleration zone of the aquaculture cage, the first sub-strategy is selected and step S3 is entered; when the unmanned vessel is in the speed-locking and stabilizing zone of the aquaculture cage, the second sub-strategy is selected and step S3 is entered.

[0021] Furthermore, when the unmanned vessel is located in the speed-stabilized zone of the aquaculture cage and the marine ranching operation is a feeding task, the second sub-strategy is fixed-point monitoring;

[0022] When the unmanned vessel is in the speed-stabilized zone of the aquaculture cage and the marine ranching operation is a detection task, the second sub-strategy is to operate at low speed.

[0023] Furthermore, in steps S24 and S25:

[0024] The lateral disturbance zone strategy includes a third sub-strategy. When the unmanned vessel is located in the lateral disturbance zone of the aquaculture cage, the third sub-strategy is selected and step S3 is entered.

[0025] The third sub-strategy is to not adjust the angle of the unmanned vessel within the tracking buffer zone. When the marine ranching operation task is a feeding task, the tracking buffer zone is set within the third threshold range on both sides of the main path; when the marine ranching operation task is a detection task, the tracking buffer zone is set within the third threshold range on the side of the main path closer to the aquaculture cage.

[0026] Preferably, the lateral disturbance zone strategy further includes a fourth sub-strategy: when the unmanned vessel is located in the lateral disturbance zone of the aquaculture cage and leaves the tracking buffer zone, the fourth sub-strategy is selected and step S3 is entered.

[0027] The fourth sub-strategy is a fixed heading offset, the angle of which is determined based on the lateral speed of the unmanned vessel, the heading deviation angle, and the cumulative trajectory deviation.

[0028] Preferably, the wake zone strategy includes a fifth sub-strategy. When the unmanned vessel is located in the wake zone of the aquaculture cage, the fifth sub-strategy is selected and step S3 is entered.

[0029] The fifth sub-strategy involves adjusting the rudder angle compensation and thrust difference of the unmanned vessel. The adjustment of the rudder angle compensation and thrust difference is determined based on the lateral torque experienced by the unmanned vessel.

[0030] Preferably, the wake zone strategy further includes a sixth sub-strategy: when the unmanned vessel is located in the wake zone of the aquaculture cage, the fifth and sixth sub-strategies are selected and the process is entered into step S3.

[0031] The sixth sub-strategy involves periodically making small adjustments to the propulsion of the unmanned vessel.

[0032] Preferably, the wake zone strategy further includes a seventh sub-strategy: when the unmanned vessel is located in the wake zone of the aquaculture cage and the unmanned vessel experiences abnormal drift or acquires abnormal information, the fifth, sixth, and seventh sub-strategies are selected and the process is entered into step S3.

[0033] The seventh sub-strategy involves adjusting both the unmanned vessel's propulsion and rudder angle to their maximum values.

[0034] The control method for unmanned vessels used in marine ranching operations provided by the embodiments of this application has at least the following beneficial effects:

[0035] This application achieves closed-loop control of "perception-decision-execution" through steps S1 to S4. During the process, it flexibly matches strategies from the basic strategy library or the disturbance strategy library based on different task types, ensuring that the unmanned vessel can adopt adaptive strategies for different task requirements in complex marine environments, avoiding operational deviations caused by environmental disturbances, and significantly improving the adaptability and operational accuracy of the unmanned vessel to complex marine ranching scenarios.

[0036] This application employs differentiated strategies tailored to the characteristics of different tasks (e.g., inspection tasks rely on a basic strategy library to ensure path repeatability, while feeding and detection tasks rely on a disturbance strategy library to achieve precise operation), avoiding the resource waste caused by a "one-size-fits-all" approach to control. For example, the feeding task ensures accurate feed delivery and reduces feed loss through fixed-point monitoring; the detection task improves data acquisition quality through low-speed operation while avoiding ineffective energy consumption, ultimately improving the operational efficiency and resource utilization efficiency of the unmanned vessel while ensuring operational effectiveness.

[0037] The unmanned surface vessel in this application can continuously and dynamically adjust its strategy during operation. For example, it can cope with changes in flow velocity by slowing down in advance and locking speed to stabilize at the upstream surface, and offset vortex disturbances by rudder angle compensation and thrust adjustment in the wake region. This effectively solves the problem of hull instability caused by sudden environmental changes (such as sudden changes in flow velocity and vortex interference), and ensures the stable and continuous execution of various tasks (especially highly repetitive inspection tasks and feeding and detection tasks with high precision requirements).

[0038] The differentiated use of the basic strategy library and the disturbance strategy library in this application, combined with real-time information feedback and dynamic strategy adjustment, enables the unmanned surface vessel (USV) to respond quickly to abnormal situations (such as abnormal drift in the wake region or abnormal information acquisition) through specific strategies (such as the maximum thrust and rudder angle adjustment of the seventh sub-strategy), thereby reducing the risk of failure. This hierarchical and dynamic control logic significantly enhances the robustness of the USV operation system, ensuring its reliable operation in the complex and variable environment of marine ranches. Attached Figure Description

[0039] Figure 1 A flowchart of a control method for an unmanned vessel used in marine ranching operations according to this application;

[0040] Figure 2 This is a fluid characteristic analysis diagram of the aquaculture cage in this application;

[0041] Figure 3 This is a flowchart of step S2 of a control method for an unmanned vessel used in marine ranching operations according to this application;

[0042] Figure 4 This is a schematic diagram showing the correspondence between the feeding and detection tasks in this application and the strategies in the disturbance strategy library when they are in different positions relative to the aquaculture cage.

[0043] Figure 5 The image shows a comparison of the trajectories of the PPO algorithm and the control method in this application when the unmanned vessel's trajectory is rectangular.

[0044] Figure 6 This is a comparison of the lateral errors of the PPO algorithm and the control method in this application when the unmanned vessel trajectory is rectangular;

[0045] Figure 7 A comparison of the angle errors between the PPO algorithm and the control method in this application when the unmanned vessel trajectory is rectangular;

[0046] Figure 8 The image shows a trajectory comparison between the PPO algorithm and the control method in this application when the unmanned vessel's trajectory is circular.

[0047] Figure 9 This is a comparison of the lateral errors of the PPO algorithm and the control method in this application when the unmanned vessel's trajectory is circular.

[0048] Figure 10 This is a comparison chart of the angle errors between the PPO algorithm and the control method in this application when the unmanned vessel's trajectory is circular. Detailed Implementation

[0049] The present application will now be further described based on preferred embodiments and with reference to the accompanying drawings.

[0050] The vocabulary used in this specification is for illustrative purposes and is not intended to limit the scope of this application. Unless otherwise expressly specified and limited, the terms "set," "connected," and "linked" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection, a direct connection, or an indirect connection via an intermediate medium; or they can refer to the internal communication between two components. Those skilled in the art will understand the specific meaning of these terms in this application.

[0051] Furthermore, in the description of the embodiments of this application, various components on the drawings have been enlarged or reduced for ease of understanding, but this is not intended to limit the scope of protection of this application.

[0052] This application provides a control method for an unmanned surface vessel (USV) used in marine ranching operations (hereinafter referred to as the "control method"), which is used to control the USV performing marine ranching tasks. Figure 1 The flowchart of the control method in this application is as follows: Figure 1 As shown, the control method includes the following steps:

[0053] S1: Acquire marine environmental, biological, and abiotic information;

[0054] In the unique application scenario of marine ranching, unmanned surface vessels (USVs) serve as the core carriers of intelligent operations, undertaking a series of critical tasks such as feeding, inspection, and detection. Their operational efficiency and accuracy directly affect the production benefits and ecological management level of marine ranches. To achieve autonomous, stable, and efficient operations, USVs are typically equipped with a wide variety of sensing and monitoring devices. These devices act as the USV's "sensory system," enabling them to acquire real-time information on their own operational status and the surrounding marine environment, providing robust data support for dynamic adjustments and operational strategy adaptations.

[0055] In some specific embodiments of this application, the unmanned surface vessel (USV) constructs a comprehensive information acquisition system by integrating inertial navigation systems, high-precision flow field sensors, and multi-axis attitude sensors. The acquisition of sea state data covers key parameters such as wave height, wave period, current speed, and direction. This data directly reflects the dynamic changes in the marine environment and is crucial for the USV to assess the safety of the operating environment, select strategies, and adjust strategies. The acquisition of hull status information includes real-time position coordinates, attitude parameters such as roll / pitch / heading, and motion parameters such as speed and acceleration. This information helps the USV accurately grasp its own operational status, ensuring it remains on the preset operating trajectory. Furthermore, considering the unique disturbances in marine ranching operations, the USV also collects disturbance information such as the turbulence of the flow field around the net cages and hydrodynamic changes in areas where fish congregate. Turbulence in the flow field around the net cages may cause the USV to experience unexpected lateral or impact forces, while hydrodynamic changes caused by fish congregation may affect the USV's navigation stability. After data fusion and analysis, the above information will be fed back to the unmanned vessel's control system in real time. If this information is abnormal or exceeds the preset threshold, it will directly affect the unmanned vessel's navigation attitude, speed control, and operational accuracy, thereby adversely affecting the overall operation process. Therefore, accurate and real-time information collection is a prerequisite for ensuring the reliable operation of unmanned vessels in the complex environment of marine ranches.

[0056] S2: Based on the information acquired and the marine ranching mission performed by the unmanned vessel, select an appropriate strategy from the basic strategy library or the disturbance strategy library;

[0057] S3: Execute the marine ranching operation tasks based on the selected strategy, wherein the marine ranching operation tasks include inspection tasks, feeding tasks, and detection tasks;

[0058] In the unique operating environment of marine ranching, aquaculture cages are, besides environmental factors, the most significant factor affecting unmanned vessel operations. Specifically, such as... Figure 2As shown, aquaculture cages alter water flow paths: when water approaches the cages, its velocity gradually decreases due to obstruction, creating stagnant zones on the cage surface; as water flows around the sides of the cages, its velocity increases, forming wake vortices and low-speed backflow zones behind them; simultaneously, the permeability of the cage mesh causes changes in the water velocity gradient, i.e., the velocity decreases near the cages while increasing on the sides. Furthermore, if the unmanned surface vessel's (USV) speed is close to the wake vortex shedding frequency of the cages, it may induce vortex-induced vibrations, exacerbating the vessel's attitude fluctuations. Moreover, the concentration of fish in the cages also affects the USV's trajectory; the biologically induced currents generated by the fish's movement cause slight trajectory deviations, an effect more pronounced when the USV operates at low speeds.

[0059] In this application, the marine ranching operations include inspection, feeding, and testing tasks.

[0060] Inspection missions are typically conducted in relatively stable sea conditions, requiring unmanned surface vessels (USVs) to repeat specific inspection routes consistently and accurately day after day. For example, USVs need to patrol along the outer boundary of the ranch at fixed times each day, systematically inspecting the entire ranch area. This repetitiveness and regularity necessitate that USVs possess "knowledge retention" capabilities to ensure that, during long-term operations, they do not lose the mastered routine inspection route strategies (i.e., the basic strategies in the basic strategy library) due to training for subsequent complex tasks, thus maintaining stable and accurate patrol performance and ensuring the continuity and reliability of daily inspection data for the marine ranch.

[0061] The core of the feeding task is to precisely deliver feed to the fish in the aquaculture cages. This must be strictly controlled within specific feeding areas such as the upstream face and the lateral disturbance zone, while the wake zone is not considered a feeding area. When feeding upstream, the unmanned vessel can maintain a certain distance from the cages by utilizing the water flow, thus avoiding hydrodynamic interference from the fish. However, when feeding in the lateral disturbance zone, the unmanned vessel will inevitably be affected by this interference.

[0062] The inspection task focuses on the aquaculture cages themselves and the condition of the fish inside (such as cage damage, fish growth, signs of disease, etc.), requiring unmanned vessels to approach specific areas such as the upstream side, lateral disturbance area, and wake of the cages for close-range observation.

[0063] Therefore, in feeding and inspection tasks, it is necessary to select appropriate strategies from the disturbance strategy library based on the distance of the unmanned vessel relative to the aquaculture cage; while inspection tasks have different requirements for strategies due to their own characteristics.

[0064] Furthermore, such as Figure 3 As shown, step S2 includes the following steps:

[0065] S21: Determine whether the marine ranching operation task is an inspection task. If yes, proceed to step S22; otherwise, proceed to step S23.

[0066] S22: Update the basic policies in the basic policy library based on the information obtained, and input the updated basic policies into step S3;

[0067] S23: Determine whether the marine ranch task is a feeding task. If yes, proceed to step S24; otherwise, proceed to step S25.

[0068] S24: Based on the distance between the unmanned vessel and the aquaculture cage, select a suitable strategy from the frontal flow strategy and lateral disturbance zone strategy in the disturbance strategy library and input it into step S3;

[0069] S25: Based on the distance between the unmanned vessel and the aquaculture cage, select a suitable strategy from the upstream strategy, lateral disturbance zone strategy, and wake zone strategy in the disturbance strategy library and input it into step S3.

[0070] It should be noted that the basic strategies in the basic strategy library are the routine path strategies formed by the unmanned vessel through historical training. Moreover, these basic strategies are not fixed and will be dynamically updated according to various information such as sea conditions, hull status and disturbances acquired by the unmanned vessel in real time, so as to adapt to the subtle changes in the marine ranch environment and ensure the continuous accuracy of routine inspection tasks.

[0071] It is precisely because the feeding task has strict requirements for the accuracy of feed delivery and the detection task has special requirements for close-range observation of the status of the cages and fish that the appropriate strategy needs to be selected from the disturbance strategy library according to the specific position of the unmanned vessel relative to the aquaculture cage in different task scenarios.

[0072] Figure 4 This diagram illustrates the correspondence between the feeding and detection tasks in this application and the strategies in the disturbance strategy library when they are located at different positions relative to the aquaculture cage. Solid circles in the diagram represent corresponding strategies and tasks, while hollow circles represent no correspondence.

[0073] Furthermore, on the upstream side of the aquaculture cage, since the unmanned vessel needs to move towards the direction of the water flow, the direct impact of the water flow makes it face a large propulsion resistance. When the unmanned vessel goes around the front edge of the cage and enters the subsequent area, the water flow velocity will drop suddenly. This rapid change in flow velocity can easily cause the unmanned vessel to suddenly accelerate or overshoot due to the sudden reduction in resistance.

[0074] Based on the impact of the water flow on the upstream face of the aquaculture cage on the propulsion of the unmanned vessel, special settings are required for the upstream face strategy in steps S24 and S25, as follows:

[0075] The distance between the unmanned vessel and the upstream surface of the aquaculture cage is set with a first threshold and a second threshold, and the first threshold is less than the second threshold. When the distance between the unmanned vessel and the upstream surface of the aquaculture cage is less than the first threshold, the area is a speed-locking stability zone; when the distance between the unmanned vessel and the upstream surface of the aquaculture cage is greater than the first threshold and less than the second threshold, the area is an early deceleration zone.

[0076] The different strategies in the upstream strategy correspond to the speed-locking stabilization zone and the early deceleration zone. Specifically, the upstream strategy includes a first sub-strategy and a second sub-strategy. When the unmanned vessel is in the early deceleration zone of the aquaculture cage, the first sub-strategy is selected and step S3 is entered; when the unmanned vessel is in the speed-locking stabilization zone of the aquaculture cage, the second sub-strategy is selected and step S3 is entered.

[0077] In some specific embodiments of this application, the first threshold is set to 10 meters and the second threshold is set to 15 meters. When the unmanned vessel is in the advance deceleration zone 10-15 meters away from the front edge of the aquaculture cage, the "slow mode" is activated and the upper limit of propulsion power is limited (the maximum thrust does not exceed 50% of the rated value). When it is in the speed-locking and stabilization zone 10 meters away from the front edge of the aquaculture cage, the target speed is set.

[0078] Meanwhile, due to the different requirements of feeding and detection tasks, the second sub-strategy in the lock-speed stability zone differs—for feeding tasks, the second sub-strategy is fixed-point monitoring to ensure stable feed placement; for detection tasks, the second sub-strategy is low-speed operation to ensure clear observation details.

[0079] In some specific embodiments of this application, the speed of the unmanned surface vessel is determined based on the following formula in the speed-locked stabilization region:

[0080] Among them, S mod As the mode switching factor, when S mod When V = 1, cmd For the speed of the motion detection task, when S mod When V = 0, cmd For the speed of the fixed-point feeding task; V max V represents the maximum speed for the motion detection task. min The minimum compensation velocity for the fixed-point feeding task is α, where α is the disturbance weight of the water flow velocity, and ||V current || represents the water flow velocity, β represents the disturbance weight of the lateral disturbance, and a lat For lateral disturbance, γ is the disturbance weight for the unmanned vessel's heading deviation, and θ is the lateral disturbance. dev For the unmanned vessel's heading deviation, K cruise K is the velocity gain coefficient for the motion detection task. hold For the deviation-velocity gain coefficient of the fixed-point feeding task, ||P err|| is the generalized bias (when it is a motion detection task), ||P err || represents the path tracking deviation; when it is a fixed-point feeding task, ||P err || represents the deviation of the fixed point position.

[0081] Furthermore, by using the same set of interference parameters α||V current ||+βa lat +γθ dev Simultaneously applying the speeds of both the mobile detection task and the fixed-point feeding task allows the physical mechanisms by which factors such as water flow, lateral disturbances, and heading deviations affect the two task modes to be shared. This ensures that different tasks maintain consistent force logic and motion response patterns when facing the same disturbances, providing a foundation for unified optimization and adaptation of strategies.

[0082] When S mod When = 1, it corresponds to the speed setting of the motion detection task. The formula at this time is:

[0083]

[0084] As can be seen from the formula, the strength of the interference parameter is inversely related to the cruise speed: the stronger the interference parameter (i.e., the larger the denominator in the formula), the lower the cruise speed, thus achieving interference avoidance; while when the interference is weak, the cruise speed increases accordingly, thus ensuring efficient movement.

[0085] When S mod When = 0, the speed setting corresponds to the fixed-point feeding task, and the formula is:

[0086]

[0087] As can be seen from the formula, the greater the position deviation of the unmanned vessel, the greater the compensation speed, enabling it to quickly respond to position deviations and make adjustments; the smaller the position deviation, the more it maintains the current speed to counteract persistent minor disturbances; at the same time, the stronger the disturbance parameter (i.e., the larger the denominator in the formula), the slower the compensation speed increases, which can effectively avoid overshoot due to excessive adjustment.

[0088] Furthermore, in the lateral disturbance zone of the aquaculture cage, the water flow mainly flows around the side of the cage. This flow characteristic leads to strong shear force and vortices that are close to the edge of the cage in this area, which in turn has a significant impact on unmanned vessels passing by and can easily cause obvious lateral deviation (i.e., sideslip).

[0089] Based on the interference caused by the aforementioned water flow characteristics to the propulsion of the unmanned vessel, special settings are required in steps S24 and S25 for the lateral disturbance zone strategy, as follows:

[0090] The lateral disturbance zone strategy includes a third sub-strategy. When the unmanned vessel is located in the lateral disturbance zone of the aquaculture cage, the third sub-strategy is selected and step S3 is entered.

[0091] The third sub-strategy is to not adjust the angle of the unmanned vessel within the tracking buffer zone. When the marine ranching operation task is a feeding task, the tracking buffer zone is set within the third threshold range on both sides of the main path; when the marine ranching operation task is a detection task, the tracking buffer zone is set within the third threshold range on the side of the main path closer to the aquaculture cage.

[0092] In summary, the core idea of ​​the third sub-strategy is to combine the characteristics of aquaculture cages and fish schools, and adopt a two-level trajectory zone mode of "main path + tracking buffer zone" in the lateral disturbance zone, without requiring the unmanned vessel to perform precise centerline tracking, so as to adapt to complex flow field environments.

[0093] Furthermore, the establishment of tracking buffer zones is closely related to the type of marine ranching operations, and their range is defined by a third threshold:

[0094] For feeding tasks, since it is necessary to cover a certain range on the side of the cage to ensure that the feed is evenly distributed, the tracking buffer zone is set within the third threshold range on both sides of the main path.

[0095] For the detection task, because the accuracy requirements for observing the side status of the cage are higher, it is necessary to get closer to the side of the cage. The tracking buffer zone is set within the third threshold range of the main path close to the side of the aquaculture cage.

[0096] In some specific embodiments of this application, the third threshold is set to 0.5 meters. Accordingly, when performing a feeding task, the tracking buffer zone is within ±0.5 meters of the main path (i.e., 0.5 meters on each side of the main path); when performing a detection task, the tracking buffer zone is within -0.5 meters of the main path (where the negative sign indicates the side closer to the aquaculture cage), to meet the requirements of close-range accurate detection.

[0097] In some preferred embodiments of this application, the lateral disturbance zone strategy includes a fourth sub-strategy in addition to the third sub-strategy. The applicable scenario is: when the unmanned vessel is in the lateral disturbance zone of the aquaculture cage and its trajectory deviates due to factors such as water flow disturbance and fish school induction flow, and it leaves the preset tracking buffer zone, the fourth sub-strategy is selected and step S3 is entered.

[0098] The fourth sub-strategy is fixed heading offset, which adjusts the unmanned surface vessel's (USV) trajectory by setting a stable heading offset angle to bring it back to a reasonable operating range. The heading offset angle is not determined by a single parameter, but rather by comprehensively considering multiple factors, including the USV's current lateral velocity (reflecting the dynamic trend of lateral drift), heading deviation angle (reflecting the degree of deviation between the actual and target headings), and cumulative trajectory deviation (recording the total trajectory offset over a period of time). This ensures the accuracy and effectiveness of the offset adjustment and better addresses the complex flow field interference in the lateral disturbance zone.

[0099] In some specific embodiments of this application, the heading offset correction angle is calculated using the following formula:

[0100]

[0101] Among them, K p K is the gain parameter for lateral velocity proportional adjustment. i K is the gain parameter for lateral velocity integral adjustment. d K is the gain parameter adjusted by the differential of the lateral velocity. d′ K is the gain parameter for adjusting the heading deviation ratio. o The gain parameter v is adjusted to account for the cumulative trajectory deviation. y Let θ be the lateral velocity of the unmanned vessel. dev Let d be the heading deviation angle of the unmanned vessel. off This represents the cumulative trajectory deviation of the unmanned vessel.

[0102] In the wake zone of aquaculture cages, the water flow exhibits strong vortex and turbulence characteristics, with high turbulence intensity and extremely unstable hydrodynamic direction. This complex flow field environment can cause unmanned vessels to produce irregular yaws and shaking, and may even cause the heading control system to temporarily fail, significantly affecting the navigation stability of the unmanned vessels.

[0103] Given the significant interference caused by the complex water flow characteristics in the wake region to the propulsion of the unmanned vessel, special settings for the wake region strategy are required in step S25, as follows:

[0104] The wake zone strategy includes a fifth sub-strategy. When the unmanned vessel is located in the wake zone of the aquaculture cage, select the fifth sub-strategy and enter step S3.

[0105] The fifth sub-strategy involves adjusting the rudder angle compensation and thrust difference of the unmanned vessel. The adjustment of the rudder angle compensation and thrust difference is based on the lateral moment experienced by the unmanned vessel. This is because the vortex disturbance in the wake region will generate a lateral moment on the hull, and this lateral moment is the main reason for the irregular yaw of the hull.

[0106] Specifically, assume the disturbance moment acting on the hull is M.w The corresponding angular velocity disturbance is w z The relationship between the two can be expressed as:

[0107]

[0108] Among them, I z Let c be the moment of inertia of the unmanned surface vessel about its vertical axis. r This represents the corresponding damping coefficient.

[0109] Furthermore, to counteract the aforementioned disturbance torque, it is necessary to design a rudder angle compensation δ and a thrust difference ΔT between the left and right propellers to generate a reverse control torque. The relationship between these two parameters can be expressed as follows:

[0110] M c =M δ ·δ+K T ·T;

[0111] Among them, M δ K is the lateral moment coefficient corresponding to a unit rudder angle. T It is the lateral moment coefficient corresponding to the unit thrust difference, used to quantify the control weight of rudder angle compensation and thrust difference on lateral moment.

[0112] and control

[0113] This application enables the control torque generated by rudder angle compensation and thrust difference to be equal in magnitude and opposite in direction to the disturbance torque experienced by the hull.

[0114] By using the above formulas and real-time data such as hull attitude and angular velocity collected by sensors, the rudder angle compensation δ and the thrust difference ΔT between the left and right propellers can be dynamically calculated and adjusted in real time. This effectively counteracts the disturbance torque in the wake region, significantly improves the irregular yaw and shaking of the hull, and achieves attitude stability of the unmanned vessel when operating in the wake region.

[0115] In some preferred embodiments of this application, considering that vortex disturbances may lead to cumulative drift when the unmanned vessel operates continuously in the wake zone of the aquaculture cage, the wake zone strategy includes a sixth sub-strategy in addition to the fifth sub-strategy. When the unmanned vessel is in the wake zone, both the fifth and sixth sub-strategies must be selected and both input into step S3.

[0116] The core of the sixth sub-strategy is to periodically make small adjustments to the propulsion of the unmanned vessel. Through these subtle changes in thrust, the position of the hull can be precisely adjusted, thereby breaking the continuous cumulative effects of vortex disturbances and enhancing the robustness of the unmanned vessel in complex flow field environments.

[0117] In some specific embodiments of this application, the propulsion force is:

[0118]

[0119] Where T0 is the average thrust of the unmanned vessel while maintaining a low speed, which is the basic power output; A is the micro-amplitude, used to control the magnitude of thrust adjustment; and f is the vibration frequency, which determines the rate of periodic change of thrust. The phase is used to set the initial state for thrust adjustment.

[0120] This formula enables regular fluctuations in propulsion, allowing unmanned vessels to maintain their basic operational trajectory while making minor positional corrections, effectively offsetting the cumulative effects of vortex disturbances.

[0121] In some preferred embodiments of this application, considering that when the unmanned vessel operates in the wake zone of aquaculture cages, it may experience abnormal drift due to intensified vortex disturbances, sensor interference from turbulence, or other reasons, or encounter abnormal information acquisition situations (such as interruption of sea state data acquisition, abrupt changes in hull attitude parameters, distortion of cage status monitoring signals, etc.), the wake zone strategy includes a seventh sub-strategy in addition to the fifth and sixth sub-strategies. When the unmanned vessel experiences the aforementioned abnormal drift or information anomalies, the fifth, sixth, and seventh sub-strategies must be selected simultaneously and step S3 must be entered. The seventh sub-strategy involves adjusting the unmanned vessel's propulsion and rudder angle to their maximum values ​​to quickly escape the complex disturbance environment through strong power output, preventing the abnormal state from continuing to escalate.

[0122] In some specific embodiments of this application, when I w >I thresh or At that time, the thrust will be rapidly increased to T max Adjust the rudder angle to δ max ;

[0123] Among them, I w I is the physical quantity of interference intensity. thresh The threshold for interference intensity. Let be the real-time position vector of the unmanned vessel at time t. Let d be the desired position vector of the unmanned vessel at time t. thresh This is the position deviation threshold.

[0124] S4: Repeat steps S2 to S3 until the marine ranching task in the target area is completed.

[0125] Step S4 is a cyclical execution mechanism. During the marine ranching operation, steps S2 (information collection and environmental assessment) and S3 (strategy selection and execution) are continuously and repeatedly executed. By constantly sensing the unmanned vessel's own status, surrounding sea conditions, and position relative to the aquaculture cages in real time, the operation strategy is dynamically adjusted to adapt to environmental changes until all operational tasks within the target area are completed (such as full coverage of inspection paths, achieving the required feeding amount, and complete collection of detection data). This ensures the continuity, adaptability, and final quality of task execution.

[0126] To verify the actual effectiveness of the unmanned vessel control method for marine ranching operations proposed in this application, Figures 5-10 Verification is carried out using specific embodiments: In this embodiment, for the two common structural forms of aquaculture cages, circular and rectangular, the traditional PPO algorithm (proximal strategy optimization algorithm) and the control method of this application (this method in...) are compared. Figures 5-10 Multi-dimensional comparative tests were conducted on the two control schemes (the ones labeled "fusion strategy algorithm" in Table 1) to comprehensively evaluate their performance differences.

[0127] Table 1 Comparison of experimental results of the fusion strategy algorithm and the PPO algorithm under different test cages.

[0128]

[0129] The test data clearly demonstrates the following conclusions: In a lightly disturbed marine environment (such as low-intensity current disturbances and small wave effects), for circular aquaculture cages, the fusion strategy algorithm, compared to the PPO algorithm, reduces the root mean square error of path tracking by 36%, the root mean square error of heading by 58%, and the rudder angle variance by 64%. For rectangular aquaculture cages, these three key performance indicators are reduced by 58%, 55%, and 70%, respectively. This series of data clearly shows that the fusion strategy algorithm exhibits significantly better performance than the PPO algorithm in both core aspects: path tracking accuracy (reducing position deviation) and heading stability (suppressing attitude fluctuations).

[0130] Further analysis reveals the advantages of the fusion strategy algorithm: During the training phase, the algorithm constructs an efficient sample selection mechanism that fully utilizes key samples with high value for strategy optimization (such as attitude adjustment data under complex flow fields) while effectively suppressing redundant and invalid samples (such as repetitive routine data in stable environments). This significantly improves the efficiency of training data utilization, enabling higher strategy convergence accuracy with the same amount of training data, laying a precise control foundation for subsequent practical operations. Furthermore, in practical applications, when facing environmental disturbances such as current disturbances and vortex effects common in the marine environment, as well as dynamic uncertainties caused by fish movement, the fusion strategy algorithm exhibits stronger environmental adaptability. It can more quickly correct ship attitude deviations caused by external disturbances by dynamically adjusting control parameters in real time, preventing the accumulation and expansion of deviations.

[0131] In summary, the advantages of the fusion strategy algorithm are not only reflected in the significant improvement of quantitative indicators such as path tracking and heading stability, but also in the multiple practical values ​​demonstrated in real-world application scenarios: First, the higher sample utilization rate shortens the algorithm training cycle and improves the actual training efficiency; second, the adaptability to different cage shapes (circular and rectangular) reflects stronger strategy generalization; third, the stable performance in a mild disturbance environment further proves its robustness to complex marine environments, providing strong support for the reliable operation of unmanned vessels in marine ranches.

[0132] The specific embodiments of this application have been described in detail above. For those skilled in the art, several improvements and modifications can be made to this application without departing from the principle of this application, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. A control method of an unmanned ship for a mariculture operation, for controlling an unmanned ship that performs a mariculture operation task, characterized by, The method comprises the following steps: S1: obtaining marine environment, biological and non-biological information; S2: selecting appropriate strategies from a basic strategy library or a disturbance strategy library based on the obtained information and the marine ranching task performed by the unmanned ship; S3: performing the marine ranching task based on the selected strategy, wherein the marine ranching task includes a patrol task, a feeding task and a detection task; S4: repeatedly performing steps S2-S3 until the marine ranching task in the target area is completed; The step S2 comprises the following steps: S21: determining whether the marine ranching task is a patrol task, if yes, performing step S22, otherwise performing step S23; S22: updating the basic strategy in the basic strategy library based on the obtained information, and inputting the updated basic strategy into step S3; S23: determining whether the marine ranching task is a feeding task, if yes, performing step S24, otherwise performing step S25; S24: selecting appropriate strategies from the disturbance strategy library of the flow face strategy, the lateral disturbance zone strategy and the wake zone strategy based on the distance of the unmanned ship relative to the net cage, and inputting the selected strategies into step S3; S25: selecting appropriate strategies from the disturbance strategy library of the flow face strategy, the lateral disturbance zone strategy and the wake zone strategy based on the distance of the unmanned ship relative to the net cage, and inputting the selected strategies into step S3; Wherein, the flow face strategy is to set a lock speed stable zone and an early deceleration zone for the unmanned ship relative to the flow face of the net cage, corresponding to the first sub-strategy and the second sub-strategy, the lateral disturbance zone strategy includes the third sub-strategy and the fourth sub-strategy, the third sub-strategy is not to adjust the angle of the unmanned ship within the tracking buffer zone, and the fourth sub-strategy is to adjust the trajectory by using a fixed heading bias when the unmanned ship leaves the tracking buffer zone; the wake zone strategy includes the fifth sub-strategy, and the fifth sub-strategy is to adjust the rudder angle compensation and the thrust difference based on the lateral moment received by the unmanned ship.

2. The control method of an unmanned ship for a mariculture farm according to claim 1, characterized by, In the steps S24 and S25: The distance of the unmanned ship relative to the flow face of the net cage is provided with a first threshold value and a second threshold value, and the first threshold value is less than the second threshold value; wherein, when the distance between the unmanned ship and the flow face of the net cage is less than the first threshold value, the region is a lock speed stable zone; when the distance between the unmanned ship and the flow face of the net cage is greater than the first threshold value and less than the second threshold value, the region is an early deceleration zone; When the unmanned ship is located in the early deceleration zone of the net cage, the first sub-strategy is selected and input into step S3; when the unmanned ship is located in the lock speed stable zone of the net cage, the second sub-strategy is selected and input into step S3.

3. The control method of the unmanned ship for marine ranching according to claim 2, wherein: When the unmanned ship is located in the lock speed stable zone of the net cage and the marine ranching task is a feeding task, the second sub-strategy is a fixed point guard; When the unmanned ship is located in the lock speed stable zone of the net cage and the marine ranching task is a detection task, the second sub-strategy is a low-speed operation.

4. The control method of an unmanned ship for a mariculture farm according to claim 1, characterized by, In the steps S24 and S25: When the unmanned ship is located in the lateral disturbance zone of the net cage, the third sub-strategy is selected and input into step S3; When the mariculture operation task is a feeding task, the tracking buffer zone is set within a third threshold range on both sides of the main path; when the mariculture operation task is a detection task, the tracking buffer zone is set within a third threshold range on the side of the main path close to the net cage.

5. The control method of the unmanned ship for mariculture operation according to claim 4, characterized in that: When the unmanned ship is located in the lateral wake area of the net cage and leaves the tracking buffer zone, a fourth sub-strategy is selected and input into step S3; Wherein, the fourth sub-strategy is a fixed heading offset, and the angle of the heading offset is determined based on the lateral speed of the unmanned ship, the heading deviation angle and the cumulative trajectory deviation.

6. The control method of an unmanned ship for a mariculture operation according to claim 1, characterized by, In the step S25: When the unmanned ship is located in the wake area of the net cage, a fifth sub-strategy is selected and input into step S3.

7. The control method of an unmanned ship for a mariculture operation according to claim 6, characterized by, In the step S25: The wake area strategy also includes a sixth sub-strategy, when the unmanned ship is located in the wake area of the net cage, the fifth sub-strategy and the sixth sub-strategy are selected and input into step S3; Wherein, the sixth sub-strategy is to periodically adjust the propulsion of the unmanned ship in a small range.

8. The control method of an unmanned ship for a mariculture operation according to claim 7, characterized by, In the step S25: The wake area strategy also includes a seventh sub-strategy, when the unmanned ship is located in the wake area of the net cage, and the unmanned ship appears abnormal drift or the acquired information is abnormal, the fifth sub-strategy, the sixth sub-strategy and the seventh sub-strategy are selected and input into step S3; Wherein, the seventh sub-strategy is to adjust the propulsion and rudder angle of the unmanned ship to the maximum value.

Images