Marine photovoltaic cleaning method and system based on heterogeneous coupling cluster of cable boats

Abstract

The application provides a sea photovoltaic cleaning method and system based on a heterogeneous coupling cluster of a cable boat, and relates to the technical field of unmanned aerial vehicle system control. Based on the geometric distribution of a photovoltaic array and real-time sea state information, a heterogeneous coupling multi-body dynamics model containing an unmanned aerial vehicle, a water supply cable and an unmanned boat, and a photovoltaic array grid map are constructed. The water supply cable is modeled as a dynamic load with fluid transmission characteristics. The physical boundary constraints of the dynamics model are defined, including operation space constraints and anti-winding constraints for avoiding the spatiotemporal trajectory intersection of the cluster operation time-cable-boat unit. The photovoltaic array to be cleaned is divided into several operation sub-regions. The minimum total energy consumption of the cluster and the balance degree of operation completion time are taken as the joint optimization objectives, a mixed integer linear programming model is constructed, and a double-layer planning architecture is adopted to generate execution instructions, so that the sea photovoltaic array can be cleaned efficiently, continuously and safely.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) system control technology, and in particular to a marine photovoltaic cleaning method and system based on a heterogeneous coupling body cluster of mechanical cable boats. Background Technology

[0002] Offshore photovoltaic (PV) power stations are built in the ocean and can complement marine engineering projects such as offshore wind power. Because they do not occupy land resources and have high power generation efficiency, they have become an important new growth engine for the marine economy in the context of energy structure transformation. However, the high humidity, high salt spray, and frequent seabird activity at sea can easily cause salt crystals, bird droppings, and other dirt to accumulate on the surface of PV modules. This not only significantly reduces PV power generation efficiency but also easily triggers hot spot effects, causing permanent damage to the modules and even inducing fires. Therefore, high-frequency, low-cost unmanned cleaning and maintenance is a core requirement for ensuring the long-term stable operation of offshore PV power stations.

[0003] Existing marine photovoltaic cleaning technologies mainly include manual cleaning, panel crawling robot cleaning, and single / loosely coordinated air-sea unmanned system cleaning. Although they have applications in land or simple marine scenarios, they all reveal significant technical bottlenecks when facing large-scale, far-off marine photovoltaic arrays: crawling robots are affected by the wave swaying of the photovoltaic floating platform, making obstacle crossing and cross-floating unit operations difficult and requiring manual handling; single UAVs have limited liquid carrying capacity and endurance, resulting in a low percentage of effective operating time; single unmanned surface vessels have limited operating angles and poor aiming accuracy; existing air-sea coordination is a loosely coupled mode, which cannot solve the energy and working fluid supply problems during UAV operations, and UAVs have poor wind resistance stability. When multiple units are operating in clusters, accidents such as entanglement and knotting of water supply cables are also prone to occur. Existing path planning and collaborative control methods are also unable to adapt to the operational requirements of this cabled system. Summary of the Invention

[0004] To address the aforementioned issues, this invention proposes a marine photovoltaic cleaning method and system based on a heterogeneous coupled body cluster of unmanned surface vessels, water supply cables, and drones. The method treats the unmanned surface vessel, water supply cable, and drone as a tightly coupled dynamic whole. By establishing a variable mass fluid-multibody coupling model, utilizing topological homotopy theory to construct dynamic anti-entanglement constraints, and employing a coarse-fine collaborative hierarchical planning strategy, the method achieves efficient, continuous, and safe cleaning of marine photovoltaic systems.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] In a first aspect, the present invention provides a marine photovoltaic cleaning method based on a heterogeneous coupling body cluster of a cable-stayed boat, comprising:

[0007] Based on the geometric distribution of the photovoltaic array and real-time sea condition information, a heterogeneous coupled multibody dynamics model including UAVs, water supply cables and unmanned surface vessels, as well as a photovoltaic array grid map, is constructed; among them, the water supply cable is modeled as a dynamic load with fluid transmission characteristics.

[0008] Define the physical boundary constraints of the dynamic model, including the work space constraints and the anti-entanglement constraints to avoid the spatiotemporal trajectory intersection of the cable-boat unit during cluster operations;

[0009] The photovoltaic array to be cleaned is divided into several sub-regions. With the goal of minimizing the total energy consumption of the cluster and balancing the completion time of the tasks, a mixed integer linear programming model is constructed, and a two-level programming architecture is used to generate execution instructions.

[0010] Secondly, the present invention provides a marine photovoltaic cleaning system based on a heterogeneous coupling body cluster of a cable-stayed boat, comprising:

[0011] The scene modeling module is configured to construct a heterogeneous coupled multibody dynamics model, including UAVs, water supply cables, and unmanned surface vessels, as well as a grid map of the photovoltaic array, based on the geometric distribution of the photovoltaic array and real-time sea condition information; among which, the water supply cable is modeled as a dynamic load with fluid transmission characteristics.

[0012] The constraint definition module is configured to define the physical boundary constraints of the dynamic model, including the workspace constraints and the anti-entanglement constraints for avoiding the spatiotemporal trajectory intersection of the cluster operation timing-cable-boat unit;

[0013] The solution module is configured to divide the photovoltaic array to be cleaned into several sub-regions, construct a mixed-integer linear programming model with the joint optimization objectives of minimizing the total energy consumption of the cluster and the balance of the operation completion time, and generate execution instructions using a two-layer programming architecture.

[0014] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the marine photovoltaic cleaning method based on a heterogeneous coupling body cluster of a cable-stayed boat as described in the first aspect.

[0015] Fourthly, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the marine photovoltaic cleaning method based on a heterogeneous coupling body cluster of a cable-stayed boat as described in the first aspect.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0017] (1) This invention constructs a heterogeneous coupled multibody dynamics model of a cable boat that integrates the fluid transmission characteristics of the water supply cable, and combines it with a photovoltaic array grid map to achieve accurate dynamic characterization of the cleaning operation system; by defining work space constraints and anti-entanglement constraints, it avoids safety risks such as equipment collisions and cable entanglements in cluster operations from the bottom layer, and ensures the stability of the operation; with the task allocation aimed at minimizing total energy consumption and balancing operation time, and with the execution instruction generation of a two-layer planning architecture, it realizes the global optimization of cluster cleaning operations and the coordinated matching of air and sea equipment, effectively solves the bottlenecks of endurance and perspective of single equipment cleaning, improves the operation efficiency and intelligence level of marine photovoltaic cleaning, reduces operation and maintenance costs, and ensures the stable and efficient operation of photovoltaic power stations.

[0018] (2) This invention connects the unmanned surface vessel (USV) and the unmanned aerial vehicle (UAV) into a whole for energy and material transmission by tightly coupled water supply cable; using the unmanned surface vessel as a "mobile supply station", it continuously delivers high-pressure cleaning water and electrical energy to the UAV through the cable; this design transforms the traditional "intermittent pulse" operation into "continuous flow scanning" operation, eliminates the supply interruption time, and realizes true all-weather, long-endurance marine operation and maintenance; it breaks through the bottleneck of endurance and load, realizes "unlimited endurance" continuous flow operation, and significantly improves the overall operation efficiency.

[0019] (3) This invention innovatively introduces the theory of topological homotopy and the separation axis theorem (SAT); at the path planning layer, the projection of each water supply cable on the sea surface is modeled as a dynamic "geometric no-fly zone", and a "spatiotemporal anti-entanglement fence" for cluster operation is constructed; this mechanism guarantees the topological invariant of any two sets of heterogeneous units on the spatiotemporal trajectory from a mathematical perspective, that is, it physically eliminates the possibility of cable crossing, making it possible for large-scale clusters to operate in parallel in dense photovoltaic arrays, and solving the problem of safety of heterogeneous clusters operating in narrow waterways.

[0020] (4) The present invention establishes a high-fidelity coupled dynamic model containing variable mass fluid characteristics and designs a tension feedforward compensation controller; the system senses and predicts the wave response of the unmanned surface vessel in real time, converts the tension fluctuations that the cable will generate into feedforward signals, and instructs the UAV to adjust the rotor collective pitch and flight attitude in advance; by changing passive interference to active utilization, the tension feedforward compensation is used to achieve centimeter-level fixed-altitude cleaning under severe sea conditions.

[0021] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0022] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute a limitation thereof.

[0023] Figure 1 A main flowchart of a marine photovoltaic cleaning method based on a heterogeneous coupling body cluster of a cable-stayed boat provided in an embodiment of the present invention;

[0024] Figure 2 A schematic diagram of the overall structure of a photovoltaic cleaning system for implementing a marine photovoltaic cleaning method based on a heterogeneous coupling body cluster of a cable boat, as provided in an embodiment of the present invention;

[0025] Figure 3 The overall flowchart of the multi-constraint marine photovoltaic cleaning collaborative planning method provided in the embodiments of the present invention is shown below.

[0026] Figure 4 This is a schematic diagram of the force analysis of the tightly coupled dynamic model of "unmanned surface vessel-variable mass flexible cable-unmanned aerial vehicle" provided in an embodiment of the present invention; wherein, It is an inertial coordinate system; , For cable tension at different locations;

[0027] Figure 5 This is a schematic diagram of the anti-entanglement constraint principle of clusters based on topological homotopy class provided in an embodiment of the present invention; wherein, (a) represents an illegal intersection path (topological entanglement), and (b) represents a legal homotopy path (topological separation);

[0028] Figure 6 This is a schematic diagram illustrating the effect of hierarchical collaborative path planning for "boat-to-machine sweeping" provided in an embodiment of the present invention;

[0029] Figure 7 This is a block diagram of a system control architecture including tension feedforward compensation provided in an embodiment of the present invention;

[0030] Figure 8 This invention provides an architecture diagram of a marine photovoltaic cleaning system based on a heterogeneous coupling body cluster of a cable-stayed boat.

[0031] Among them, 1-unmanned surface vessel; 2-unmanned aerial vehicle; 3-water supply cable; 4-marine photovoltaic array; 41-operation waterway; 5-sea surface; 6-cleaning and spraying components; 7-intelligent servo winch; 8-unmanned surface vessel sensing unit. Detailed Implementation

[0032] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0033] Example 1

[0034] like Figure 1As shown in the figure, this embodiment discloses a marine photovoltaic cleaning method based on a heterogeneous coupler cluster of cable-stayed boats, including the following steps:

[0035] S1: Based on the geometric distribution of the photovoltaic array and real-time sea condition information, a heterogeneous coupled multibody dynamics model including UAVs, water supply cables and unmanned surface vessels, as well as a photovoltaic array grid map are constructed; among them, the water supply cable is modeled as a dynamic load with fluid transmission characteristics.

[0036] S2: Define the physical boundary constraints of the dynamic model, including workspace constraints and anti-entanglement constraints to avoid the intersection of the spatiotemporal trajectories of the cable-boat unit during cluster operations;

[0037] S3: Divide the photovoltaic array to be cleaned into several sub-regions. With the goal of minimizing the total energy consumption of the cluster and balancing the completion time of the tasks, construct a mixed integer linear programming model and use a two-layer programming architecture to generate execution instructions.

[0038] A schematic diagram of the overall structure of the marine photovoltaic cleaning system based on a heterogeneous coupling body cluster of a cable-stayed boat provided in this embodiment of the invention is shown below. Figure 2 As shown in the figure, with sea level 5 as the reference, the figure shows the spatial relationship of the unmanned surface vessel 1, the drone 2, the water supply cable 3, the marine photovoltaic array 4, the cleaning and spraying assembly 6, the intelligent servo winch 7, and the unmanned surface vessel sensing unit 8, as well as the layout of each key hardware module in the system.

[0039] Next, combined Figure 3 This embodiment provides a detailed description of a marine photovoltaic cleaning method based on a heterogeneous coupling body cluster of a cable boat.

[0040] Step S1: High-fidelity environment modeling and initialization of heterogeneous couplers.

[0041] S1.1 Multi-source data acquisition:

[0042] Using a 16-line 3D lidar (Velodyne VLP-16) and RTK-GNSS positioning module mounted on an unmanned surface vessel (USV), the absolute coordinates and geometric distribution of the marine photovoltaic array were obtained, constructing a resolution of [resolution missing]. A rasterized environment map. Specifically, this includes the following steps:

[0043] (1) Spatiotemporal synchronization and alignment of multi-source data:

[0044] Because LiDAR and RTK-GNSS have different data sampling frequencies (typically 10-20Hz for LiDAR and 1-10Hz for RTK-GNSS) and there are installation position deviations, strict calibration and synchronization are required.

[0045] First, time synchronization is performed, using either hard triggering or a soft synchronization mechanism based on GPS time. This is based on the LiDAR acquisition time. Using this as a baseline, the precise pose of the unmanned surface vessel at that moment is calculated using interpolation algorithms (such as linear interpolation or spherical linear interpolation SLERP). ,in, This represents the three-dimensional position coordinates of the unmanned surface vessel in the geodetic coordinate system. Represents the three-dimensional attitude angles of the unmanned surface vessel. This indicates transpose.

[0046] Next, extrinsic parameter calibration is performed, and the rotation matrix of the LiDAR coordinate system relative to the GNSS / IMU coordinate system is pre-calibrated. Translation vector And the conversion relationship between the GNSS antenna and the center of the ship.

[0047] (2) Laser point cloud data preprocessing:

[0048] The raw point cloud data contains sea surface clutter, rain and fog noise, and dynamic obstacles, and needs to be filtered:

[0049] Straight-through filtering: Based on the estimated height range of the photovoltaic array, set the Z-axis threshold (e.g., 0.5m to 5m above sea level) to remove sea surface echoes (water surface) and excessively high aerial noise.

[0050] Distance filtering: The effective ranging range of LiDAR (e.g., 0.5m-50m) is truncated, and points that are obstructed by the ship itself and too far away sparse points are removed.

[0051] Statistical Outlier Removal (SOR): For each point, calculate its average distance to neighboring points and remove outlier noise points (such as false points generated by birds or water mist).

[0052] (3) Point cloud coordinate transformation and stitching:

[0053] Transform the preprocessed local point cloud from the LiDAR coordinate system to a global geographic coordinate system (such as the UTM coordinate system): For the first... Any point in a frame point cloud Using the synchronized hull pose and extrinsic parameter matrix, its coordinates in the global coordinate system are calculated. : ;

[0054] in, and These are the rotation matrix and translation vector of the unmanned surface vessel in the world coordinate system, respectively.

[0055] By accumulating consecutive frames, discrete single-frame point clouds are pieced together into a dense global point cloud map.

[0056] (4) Generation of 2.5D probabilistic raster maps:

[0057] Projecting the 3D global point cloud onto a 2D plane to construct a resolution of... A raster map.

[0058] Specifically, initialize the grid. This is based on the operating area of ​​the photovoltaic array (e.g., ...). (meter), initialize a A two-dimensional matrix, where Each grid cell Store two values, probability of occupancy and height information .

[0059] Next, point cloud projection is performed and statistics are collected. This involves iterating through every point in the global point cloud. Calculate its corresponding raster index : ;in, , These represent the minimum x-coordinate and minimum y-coordinate of the work area in the global coordinate system, respectively.

[0060] For points falling into the same grid cell, record their maximum height. and minimum height .like If the height exceeds a set threshold (such as the height of a photovoltaic panel support), then the grid is considered to have a three-dimensional obstacle.

[0061] Update the occupancy probability using the Bresenham ray tracing algorithm or a simple point cloud statistical method. If the point cloud density (number of points) within a certain grid cell exceeds a threshold... If the grid is identified as an "obstacle" (such as a photovoltaic floating platform or support structure), then it is set... (or a high probability value). If a laser beam passes through a grid but does not produce an echo at that location, the occupancy probability of that grid is reduced according to Bayes' theorem until it approaches 0 (representing a passable waterway).

[0062] The generated original raster map is subjected to morphological dilation, with the dilation radius set to the safe radius of the unmanned surface vessel (e.g., half the width of the vessel plus a safety margin) to construct a cost map for path planning.

[0063] Through the above steps, the system ultimately outputs a high-precision raster map containing obstacle areas (photovoltaic arrays), passable areas (operational waterway 41), and unknown areas, which can be directly used by subsequent path planning algorithms. This enables accurate differentiation between marine photovoltaic waterways and photovoltaic arrays, providing reliable environmental support for cleaning path planning and improving the safety and coverage efficiency of cleaning operations.

[0064] S1.2 Variable Mass Tightly Coupled Dynamics Modeling:

[0065] Traditional cable models typically treat them as massless springs or simple flexible cables, ignoring the inertial effects caused by internal fluid transport. However, in high-pressure cleaning scenarios using cleaning spray components, the high-speed water flow generates significant Coriolis and centrifugal forces, causing the cable to exhibit variable stiffness characteristics during motion. This can even lead to a "whiplash effect" when the water pump starts and stops. The sudden change in water flow during pump startup and shutdown causes the cable to initially tense and then suddenly rebound, resulting in a violent, high-speed swing at the end, much like a whip. The force analysis of the tightly coupled dynamic model of the "unmanned surface vessel-variable mass flexible cable-UAV" is as follows: Figure 4 As shown;

[0066] To address the aforementioned issues, this embodiment introduces the variable mass lumped mass method, fully considering dynamic factors such as inertial effects, Coriolis forces, centrifugal forces, and momentum thrust caused by fluid transport within the cable. A tightly coupled dynamic model of "unmanned surface vessel-variable mass flexible cable-UAV" is constructed, achieving an accurate description of the cable's variable stiffness characteristics and whiplash effect under high-pressure cleaning conditions. This provides a high-fidelity simulation foundation and dynamic coupling boundary conditions for subsequent controller design. Specifically:

[0067] (1) Lumped Mass Method:

[0068] Discretize the continuous water supply cable into Nodes and A massless spring-damper element.

[0069] Among them, the The position vector of each node is The quality is Connect the nodes and A cable segment is defined as a unit, the length of which is The tangential unit vector is .

[0070] (2) Calculation of additional mass:

[0071] Treating the water supply cable as a variable mass system, nodes Total mass It includes not only the quality of the cable sheath, but also the quality of the fluid inside the pipe and the additional quality of the external fluid:

[0072] ;

[0073] in, The quality of the cable material itself (Kevlar / polyurethane, etc.). This refers to the water flow quality within the pipe. When the pump is shut off (empty pipe), When the water pump is turned on (full pipe), ; This indicates the density of the fluid transported within the pipe. This indicates the cross-sectional area of ​​the water supply channel inside the water supply cable. This indicates that after discretizing the entire cable, the first... The physical length of each line segment unit. The model needs to simulate the water filling process, i.e. It is time This function causes the system's inertia matrix to change over time. Adding mass to the hydrodynamics: when the cable is partially submerged in seawater, the accelerated movement of the cable will cause the surrounding seawater to move. This virtual mass needs to be added (usually taken as 1.0 times the mass of the displaced water volume).

[0074] This embodiment, based on the mass of the cable itself, further introduces dynamic variable mass water flow mass inside the pipe and external hydrodynamic added mass, which can accurately reflect the time-varying characteristics of the system inertia during the start-up and shutdown of the water pump. At the same time, it considers the hydrodynamic added mass effect brought about by the movement of the surrounding seawater when the cable is partially submerged, thus improving the model's accuracy in depicting the dynamic response of the cable under real sea conditions.

[0075] (3) Force analysis and dynamic equations:

[0076] For any intermediate node ( Based on Newton's second law, establish the following equation:

[0077] ;

[0078] in, For elastic tension, For gravity and buoyancy, For environmental loads, Additional force generated for fluid transport within the pipe.

[0079] Specifically, elastic tension Based on the Kelvin-Voigt viscoelastic model: ,in For the first The tension vector of each spring-damped unit. For tensile stiffness, The damping coefficient is... Original length. Node The resultant tension is: .

[0080] Gravity and buoyancy Specifically:

[0081] ;

[0082] in, This indicates the density of seawater. This represents the volume of seawater displaced by the node; if the node is above the water surface, the buoyancy term is 0; if it is underwater, the mass of the displaced seawater must be deducted.

[0083] Environmental load Including air resistance in the air section and the hydrodynamic load of the underwater part Air resistance :

[0084] ;

[0085] in, Indicates air density, Indicates the air drag coefficient. This represents the windward projected area of ​​the cable node in the direction perpendicular to the relative wind speed. This represents the velocity vector of the cable node relative to the surrounding air. The hydrodynamic load is represented by the Morison equation, which includes drag and inertial force terms.

[0086] Additional force generated by fluid transport within the pipe This describes the dynamic characteristics that distinguish a "water pipe" from an "ordinary rope." According to the dynamics theory of fluid transport pipelines, the forces exerted by the fluid on the pipe wall mainly consist of three components:

[0087] ;

[0088] in, Coriolis force for fluid (corresponding to) Figure 4 middle ), Centrifugal force of fluid (corresponding to) Figure 4 middle ), It is momentum thrust.

[0089] Specifically, fluid Coriolis force This indicates that when the cable rotates (angular velocity) When the fluid inside the pipe rotates with the pipe, its angular momentum changes, thus exerting a reaction force on the pipe wall. The expression is:

[0090] ;

[0091] Or unfolded into a unit volume form:

[0092] ;

[0093] in, Indicates the first The Coriolis force vector of the fluid acting on a cable segment; Indicates the first The mass of water flow contained within each micro-element node; Indicates the first The angular velocity vector of each cable node; This represents the velocity vector of water flowing inside the water supply cable. Indicates the density of the fluid (water) being transported within the pipe; This indicates the effective cross-sectional area inside the water supply cable; Indicates the first The physical length of a discretized cable segment. This represents the relative flow velocity vector of the fluid inside the pipe relative to the pipe wall.

[0094] This force will impede the bending vibration of the cable. When the drone maneuvers rapidly, causing the cable to swing violently, the Coriolis force will generate a significant lateral load.

[0095] fluid centrifugal force This means that centrifugal force is generated as the fluid follows the curved cable. Mathematically, this manifests as a reduction in the effective stiffness of the cable. High-speed fluids tend to "straighten" the pipe.

[0096] ;

[0097] in, Indicates the mass of the fluid within the node; The vector representing the rotational angular velocity of the node; The radius of curvature vector of a node relative to its instantaneous center of rotation.

[0098] In discrete models, this is usually incorporated into the tension term correction, meaning that the higher the flow velocity, the "stiffer" the cable appears or the more easily it becomes unstable.

[0099] Momentum thrust It only acts on the end of the cable (at the drone nozzle). The reaction force generated when the water is ejected:

[0100] ;

[0101] in, Indicates the density of the ejected fluid; This indicates the effective cross-sectional area of ​​the cleaning nozzle; The absolute exit velocity of the water flow when it leaves the nozzle is given. The square of the velocity term indicates that the thrust is proportional to the water flow velocity.

[0102] This is a follower force, always directed in the opposite direction of the nozzle, acting directly on the drone's body.

[0103] This embodiment introduces elastic tension, environmental load, Coriolis force, centrifugal force, and momentum thrust generated by fluid transmission within the pipe to construct a complete cable force model, which can effectively describe the influence of high-speed water flow on cable stiffness and vibration characteristics, and support the simulation reproduction of dynamic phenomena such as the "whiplash effect".

[0104] (4) Boundary conditions and heterogeneous coupling:

[0105] The "tight coupling" of the model is reflected in the handling of boundary conditions:

[0106] First node ( - Boat cable coupling: It is constrained to the position of the unmanned surface vessel's cable guide hole.

[0107] ;

[0108] Here, and The reaction force of the cable on the boat, calculated from the dynamic model of the unmanned surface vessel, is: ; This represents the position vector of the cable's first node in the global inertial coordinate system. This represents the real-time position vector of the unmanned surface vessel's center of mass in the global coordinate system; This is the rotation matrix of the unmanned surface vessel in the global coordinate system, used for coordinate system transformation; This represents the local fixed coordinate vector of the cable guide hole relative to the center of mass of the unmanned surface vessel.

[0109] End node ( - Cable coupling: It is confined to the drone's mounting point.

[0110] ;

[0111] The tension of the cable on the drone and the reaction force of water spray This external force is input into the six-degree-of-freedom dynamic equations of the UAV. The translational equation describes the acceleration of the UAV's center of mass in the global coordinate system, expressed as:

[0112] ;

[0113] in, The total mass of the drone itself and its payload; Let the linear acceleration of the UAV's center of mass in the global coordinate system be denoted as . (correspond Figure 4 middle () represents the total lift vector generated by the multi-rotor system; For air resistance and sea wind disturbance loads; This represents the physical tension vector between the cable end node and the UAV mounting point. This represents the reaction thrust vector when high-pressure water is ejected. The rotation equation describes the change in angular velocity of the UAV in the body coordinate system, expressed as:

[0114] ;

[0115] in, Here is the rotational inertia matrix of the UAV; These are represented as the angular acceleration and angular velocity vectors of the UAV, respectively. This is expressed as the control torque generated by the rotor; Expressed as aerodynamic drag torque; It is represented as the offset vector of the cable attachment point relative to the center of mass of the UAV; It is represented as the offset vector of the cleaning nozzle relative to the center of mass of the drone.

[0116] This embodiment establishes bidirectional coupled boundary conditions of displacement, velocity, and force between the boat, cable, and engine, thereby realizing tightly coupled dynamic modeling of a multi-rigid-body system and a flexible variable-mass cable. This provides a unified simulation platform foundation for the collaborative control and dynamic stability analysis of heterogeneous systems.

[0117] Step S2: Considering the poor wind resistance stability of UAVs in high sea states, and the tendency for water supply cables to become entangled, knotted, or even pulled during dynamic towing in multi-unit swarm operations, and the difficulty of adapting existing path planning and collaborative control methods to the motion constraints and obstacle avoidance requirements of cabled systems, this embodiment constructs multi-dimensional operational constraints and topological anti-entanglement restrictions to avoid safety risks such as equipment collisions and cable entanglement in swarm operations from the bottom layer, ensuring operational stability; a schematic diagram of the swarm anti-entanglement constraint principle based on topological homotopy is shown below. Figure 5 As shown, the specific steps and algorithm implementation are as follows:

[0118] S2.1 Constructing job space constraints based on physical geometry limitations:

[0119] The system first defines the UAV's permissible state space based on the physical properties of the heterogeneous coupler to prevent equipment damage or operational failure.

[0120] (1) Coupling constraint between UAV operating altitude and tension

[0121] Considering the physical length of the water supply cable It is fixed, and the drone must keep the cable under "slight tension" to maintain the liquid supply; the drone's permissible altitude... The following geometric inequalities must be satisfied:

[0122] ;

[0123] in, To prevent the nozzle from touching the photovoltaic panel, the minimum safe height is set as follows: ; and These are the position coordinates of the drone and the unmanned surface vessel on the horizontal plane, respectively. The safety redundancy factor (for example, a value of...) This is used to allow for the necessary sag of the cable catenary, preventing the cable from being stretched and broken.

[0124] (2) Relative yaw sector constraints

[0125] To prevent the towed cable from cutting the photovoltaic panel support laterally when the unmanned surface vessel turns or the drone moves sideways, the angle of the cable relative to the bow needs to be limited.

[0126] Define the bow vector of the unmanned surface vessel as: Define the projection vector of the cable onto the horizontal plane as... Calculate the relative angle. :

[0127] ;

[0128] The constraints are: If detected As it approaches the boundary, the control system will forcibly limit the drone's lateral maneuvering range.

[0129] S2.2 Constructing a dynamic topology anti-entanglement constraint based on homotopy classes

[0130] To achieve intrinsic safety in multi-machine collaborative planning, the system reduces the three-dimensional cable entanglement problem to a two-dimensional dynamic line segment intersection determination problem, and uses the homotopy class theory to ensure that the topological relationship between cables remains unchanged at any time (i.e., always remains in a separated state).

[0131] (1) Geometric Abstraction and Projection

[0132] For any two heterogeneous units (denoted as units) in the cluster and unit ), at each discrete time step of the planning algorithm :

[0133] Extraction unit Cable projection line segment Its endpoints are the coordinates of the unmanned surface vessel. and drone coordinates ;

[0134] Extraction unit Cable projection line segment Its endpoints are and .

[0135] (2) Fast Repulsion and Crossing Experiments Based on Vector Cross Product

[0136] To efficiently determine line segments and To determine whether they intersect (i.e., whether topological entanglement occurs), the following two-level decision logic is executed:

[0137] Level 1: Rapid Rejection Test

[0138] Determine based on line segment and This checks if the diagonal rectangular regions overlap. If they do not overlap, they cannot intersect, and subsequent calculations are skipped. The judgment logic is as follows:

[0139] ;

[0140] in, Represents a logical OR operation.

[0141] Level 2: Vector Cross Product Crossing Experiment

[0142] If fast repulsion is used, the relative positions between line segments can be determined using the cross product of vectors. The cross product operation of two-dimensional vectors is defined. .

[0143] First, determine The two endpoints Is it located in Both sides of the line:

[0144] ;

[0145] ;

[0146] like ,illustrate straddle The extension line. Next, perform a reverse check to determine... The two endpoints Is it located in Both sides of the line:

[0147] ;

[0148] ;

[0149] like ,illustrate straddle The extension of .

[0150] If and only if and When both conditions are met, determine the line segment. and "Topological entanglement" occurs, such as Figure 5 As shown in (a); conversely, if the condition is not met, it is determined to be "topological separation", such as Figure 5 As shown in (b).

[0151] This embodiment establishes a topological state discrimination criterion for cable paths using homotopy class theory, explicitly defines the judgment conditions for "illegal crossing paths" (topological entanglement), and classifies their complements as "legal homotopy paths" (topological separation), thereby achieving accurate differentiation between the two types of path states at the topological level and providing a clear classification basis for subsequent constraint execution.

[0152] S2.3 Constraint Enforcement and Penalty Function Construction

[0153] In the path planning algorithm, the above geometric decision is transformed into a numerical cost function to guide heterogeneous units to actively avoid entanglement.

[0154] (1) Calculation of minimum distance between line segments: When no intersection occurs, calculate the minimum distance between line segments. and The shortest Euclidean distance between This distance is obtained by calculating the minimum of the perpendicular distance from the point to the line segment and the distance between the endpoints.

[0155] (2) Potential field penalty function Design and construct an anti-tangle penalty term of the following form, and add it to the overall optimization objective function. middle:

[0156] ;

[0157] in, Let the total energy consumption cost function of the system be . This is the job time cost function; Defined as the obstacle potential field function:

[0158] ;

[0159] ;

[0160] In this formula, , This represents any two distinct cable segments. The preset security topology distance threshold; This is the repulsion gain coefficient.

[0161] Through the above mechanism, when the planned path of the heterogeneous unit causes the cable distance to approach... At this point, the penalty value increases sharply; once a prediction crosses, the penalty value tends to infinity, thus forcing the planner to abandon the current homotopic path and search for a "tangle-free" cooperative trajectory that maintains the topological separation of the cables, such as... Figure 5 As shown in (b).

[0162] Based on topology classification, this embodiment constructs a penalty function for illegal intersection paths and a guidance mechanism for legal homotopy paths, respectively. This transforms topology constraints into objective terms that can be embedded in the path planning optimization model, enabling the planner to actively avoid entanglement risks and tend towards separation paths during the iteration process, thus realizing the computable execution of topology constraints in numerical optimization.

[0163] This embodiment constructs two types of topological constraints, "illegal cross paths" and "legal homotopy paths," to identify and avoid entanglement risks and guide and optimize feasible paths, respectively. This provides underlying anti-entanglement protection and dynamic adaptability for cable cluster operations at the path planning level.

[0164] Step S3: Implementation of Global Cooperative Task Allocation

[0165] Global collaborative task allocation aims to solve the combinatorial optimization problem of "multi-machine, multi-task". This embodiment adopts the energy-efficient mixed-integer linear programming (MILP) method, and the specific implementation process is as follows:

[0166] S3.1 Discretization and Attribute Mapping of the Work Area

[0167] In order to transform the continuous physical space into a mathematical model that the algorithm can process, the marine photovoltaic array is first rasterized.

[0168] Task block division: Divide the photovoltaic array area to be cleaned. Divided into A set of rectangular task blocks. .

[0169] Set the size of each task block to (Corresponding to one row or half row of photovoltaic array), this size is adapted to the single dwell time range of a single heterogeneous coupler.

[0170] Extract each task block Geometric center coordinates .

[0171] Dirt Attribute Mapping: Utilizing the dirt distribution heatmap obtained from the initial drone inspection, for each task block... Assign a dirtiness level coefficient .in, This indicates that the item is clean and requires no washing. It indicates extreme dirtiness.

[0172] Define the cleaning time for a single block ,in Based on the cleaning time, This is the dirt gain coefficient.

[0173] S3.2 Constructing a Mixed Integer Linear Programming (MILP) Model

[0174] Assuming the heterogeneous coupled body cluster contains A set of units. The following optimization model is established:

[0175] (1) Definition of decision variables: binary decision variables , indicating the first Is the heterogeneous unit responsible for executing the [number]th [unit]? Task blocks:

[0176] ;

[0177] (2) Objective Function Construction The core objective of this invention is to minimize the total energy consumption of the cluster to complete all tasks. The total energy consumption consists of three parts: unmanned surface vessel propulsion energy consumption, unmanned aerial vehicle flight energy consumption, and high-pressure water pump operation energy consumption.

[0178] ;

[0179] The specific energy consumption calculations are as follows:

[0180] Unmanned surface vessel propulsion energy consumption : Travel with the unmanned surface vessel from the current location (or the previous mission point) to the mission point. distance Proportional.

[0181] ;

[0182] Drone flight energy consumption The value is related to the length of the scanning path and the hovering time of the drone within the task block, and is affected by the degree of dirt.

[0183] ;

[0184] in, This indicates the duration of the cleaning operation.

[0185] Energy consumption of water pump operation It depends directly on the cleaning time and pump pressure.

[0186] ;

[0187] in The energy conversion efficiency coefficient. Rated power of each component ( Rated power for unmanned surface vessels, Rated power for drones, (Rated power of the water pump).

[0188] This embodiment focuses on minimizing total energy consumption while covering energy consumption from three categories: unmanned surface vessels, drones, and water pumps. It can achieve optimal global energy consumption and lowest operating costs for heterogeneous clusters.

[0189] (3) Setting Constraints

[0190] Task completeness constraint: Ensure that each task block It must be executed by one and only one heterogeneous unit.

[0191] ;

[0192] Where M is the total number of task blocks.

[0193] Job balance constraint: To prevent an overload of tasks in one unit from causing a decrease in overall efficiency, the difference in the number of tasks between units is limited.

[0194] ;

[0195] in, This is the threshold for the allowed number of tasks.

[0196] Macroscopic safety distance constraint (spatial isolation): To ensure that each unit maintains a safe distance on a macroscopic level and avoids overlapping work areas, this embodiment introduces a centroid repulsion constraint. Calculate the... The geometric centroid of all task blocks assigned to each unit :

[0197] ;

[0198] The task centroid of any two units is required to be ( , Distance greater than the safety threshold : .

[0199] The constraint settings in this embodiment, through the triple constraints of task completeness, job balance, and macroscopic safety distance, ensure full task coverage, load balance, and no job collisions, thereby improving system stability and execution efficiency.

[0200] In MILP solutions, this nonlinear constraint is typically achieved by presetting the "preferred working sector" for each element or by adding a "distance penalty term" to the objective function. That is, if the element... Performing a task that is too far from its initial position will result in a huge penalty.

[0201] S3.3 Model Solving and Route Generation

[0202] Solution algorithm: Considering The problem may be quite complex, as it falls under the NP-hard category. This embodiment uses the Gurobi optimizer or a genetic algorithm (GA) to solve it.

[0203] Input: Task block coordinate matrix, dirt level vector, and initial position of heterogeneous units.

[0204] Output: Optimal solution matrix .

[0205] Reference docking point sequence generation: Based on the solution results, for each heterogeneous unit Extract the subset of tasks it is responsible for. .

[0206] Serialization: Utilizing the nearest neighbor algorithm for solving the Traveling Salesman Problem (TSP), an unordered subset of tasks is serialized. Arranged into an ordered sequence of tasks. .

[0207] Path smoothing: Connect the center points of each task block to generate the global baseline cruise route (GlobalWaypoints) for the unmanned surface vessel.

[0208] Example of Result from Example S3.4

[0209] Photovoltaic array is Row Columns, total Each task block, input There are two heterogeneous couplers (Unit A, Unit B). After the above calculation steps, the system outputs the following allocation instructions:

[0210] Unit A: Task Assignment Block (Corresponding to rows 1-5 in physical space). The reference berthing point sequence is distributed from west to east along the third drainage channel.

[0211] Unit B: Assigning Task Blocks (Corresponding to rows 6-10 in physical space). The reference berthing point sequence is distributed from west to east along the 8th drainage channel. Safety verification: The operating trajectories of Unit A and Unit B always maintain an interval of at least 3 rows in space, satisfying the macroscopic safety distance constraint.

[0212] Furthermore, a hierarchical collaborative path planning for "boat-to-machine sweeping" was developed.

[0213] This step aims to address the significant differences in mobility between large-inertia unmanned surface vessels (USVs) and highly maneuverable unmanned aerial vehicles (UAVs) by achieving spatiotemporal synchronization of heterogeneous units through a "macro-micro" two-layer architecture. A schematic diagram of the layered collaborative path planning effect is shown below. Figure 6 As shown.

[0214] (1) Upper layer (USV layer): Global route planning based on seakeeping optimization

[0215] As a mobile base station for the system, the unmanned surface vessel (USV) needs to not only avoid obstacles in its path planning, but also optimize its course to reduce hull swaying (especially roll) caused by waves, thereby stabilizing the cable base.

[0216] Environmental Gridding and Cost Mapping: Based on the grid map constructed in step S1, the waterways between photovoltaic arrays are defined as feasible areas. An improved approach is adopted. The algorithm performs the search.

[0217] Construction of wave resistance cost function: Defining nodes Total agency value . This represents the actual path length from the starting point to the current node. This is a heuristic estimate of the Euclidean distance from the current node to the destination. (Wave-oriented penalty term, core feature):

[0218] ;

[0219] in, For unmanned surface vessels at nodes The planned heading angle at the location; For real-time monitoring of the main wave direction; This is the roll sensitivity coefficient.

[0220] Physical meaning: When the heading is perpendicular to the wave direction (cross waves), The cost is greatest when the course is parallel to the wave direction (with the wave / against the wave). The cost is minimized when the wave-cutting or zigzag navigation path is used to avoid cross waves.

[0221] (2) Lower layer (UAV layer): Dynamic coverage scan under the moving base

[0222] The unmanned boat was moving at high speed While moving along the planned path, the planned drone will perform operations that cover the entire hull relative to the vessel.

[0223] Definition of moving coordinate system: Establishing the ship's coordinate system The origin is located at the unmanned surface vessel's cable guide hole. The axis runs along the bow of the boat.

[0224] Reciprocating scan path generation: In This generates a series of scan waypoints. The scan line spacing is then set. (Slightly smaller than the spray width to ensure overlap), scan length (Limited by the length of the photovoltaic panels). The relative motion trajectory of the drone is described as follows:

[0225] ;

[0226] in It is a lateral scanning motion. This is the compensation amount for the ship's feed motion. This refers to the working height.

[0227] (3) Speed ​​coupling matching model of motorboat

[0228] To achieve continuous cleaning with "zero waiting time," the speeds of the machine and the vessel must be strictly matched. The following time equivalence equation is established:

[0229] Drone single cycle time The time required for the drone to complete a line scan and make a turn.

[0230] ;

[0231] in, For the flight speed of the drone operation, The average time taken for a drone to make a "U-turn" between two lines (typically 1000 meters) ).

[0232] Unmanned surface vessel feed matching: in Within a given time period, the distance the unmanned surface vessel (USV) travels forward must be exactly equal to a scan line spacing. .

[0233] ;

[0234] in, This refers to the forward speed of the unmanned surface vessel.

[0235] Coupled control law: Calculates the optimal reference speed command for the unmanned surface vessel. :

[0236] ;

[0237] The edge computing terminal sends the command to the unmanned surface vessel's underlying propulsion controller to ensure that the two remain dynamically synchronized.

[0238] Furthermore, robust control execution is based on tension feedforward.

[0239] This step addresses the disturbance transmission chain of "waves-hull-cable-UAV" by designing a model-predictive feedforward compensation mechanism to actively isolate wave interference. The system control architecture including tension feedforward compensation is as follows: Figure 7 As shown.

[0240] (1) Wave disturbance state prediction based on EKF: The extended Kalman filter (EKF) is used to make short-term predictions of the heave motion of the ship, thus solving the sensor lag problem.

[0241] State vector definition: Let the system state be... , representing the ship's heave displacement, heave velocity, equivalent wave height, and vertical wave velocity, respectively. State transition equation (based on a second-order wave model):

[0242] ;

[0243] ;

[0244] in, Let k be the process noise of the system at time k. Main frequency for ocean waves Sampling time. Prediction execution: based on the current time. IMU observations, iteratively predicting the future The speed of heave after stepping .

[0245] (2) Estimation of cable tension fluctuation

[0246] Based on the variable mass dynamics model established in step S1, the estimated value of tension fluctuation caused by hull heave is simplified. Using the Kelvin-Voigt viscoelastic model:

[0247]

[0248] in, The equivalent stiffness and damping coefficient of the water-filled cable (updated in real time by model parameters); This represents the predicted vertical displacement change of the hull. To predict the step size.

[0249] (3) Dual-end collaborative feedforward compensation strategy

[0250] The estimated tension fluctuation The control is decomposed into feedforward control variables, which are applied to both the UAV and the intelligent servo winch. Thrrust Feedforward:

[0251] The UAV flight controller, based on the received position loop PID output, adds feedforward thrust:

[0252] ;

[0253] in, This represents the output control quantity of the UAV position loop PID controller. Indicates the feedforward control gain. For the cable zenith angle, The drone's tilt angle. Logic: When it is predicted that the cable will be tightened (tension increased), the drone increases its throttle in advance to prevent a drop in altitude.

[0254] Active Heave Compensation (AHC): Instructs the intelligent winch to perform active heave compensation, and its speed control commands... for:

[0255] ;

[0256] in, The predicted heave / sink rate of the ship; For the compensation gain coefficient (take) ); To maintain a slightly tensioned relaxation velocity term. Logic: When the hull is lifted by the waves, the winch quickly pulls in the cable; when the hull sinks, the winch quickly releases the cable, thereby physically cutting off the transmission path of wave energy to the drone.

[0257] This specific embodiment is a collaborative planning method for marine photovoltaic cleaning based on tightly coupled water supply cables, which overcomes the harsh marine environment, breaks through the bottleneck of single equipment endurance, achieves continuous supply of energy and cleaning media, and effectively avoids the risk of cluster cable entanglement. This method enables all-weather, high-efficiency, unmanned intelligent operation and maintenance of marine photovoltaic power plants.

[0258] Example 2

[0259] This embodiment provides a marine photovoltaic cleaning system based on a heterogeneous coupler cluster of cable-stayed boats, including:

[0260] The scene modeling module is configured to construct a heterogeneous coupled multibody dynamics model, including UAVs, water supply cables, and unmanned surface vessels, as well as a grid map of the photovoltaic array, based on the geometric distribution of the photovoltaic array and real-time sea condition information; among which, the water supply cable is modeled as a dynamic load with fluid transmission characteristics.

[0261] The constraint definition module is configured to define the physical boundary constraints of the dynamic model, including the workspace constraints and the anti-entanglement constraints for avoiding the spatiotemporal trajectory intersection of the cluster operation timing-cable-boat unit;

[0262] The solution module is configured to divide the photovoltaic array to be cleaned into several sub-regions, construct a mixed-integer linear programming model with the joint optimization objectives of minimizing the total energy consumption of the cluster and the balance of the operation completion time, and generate execution instructions using a two-layer programming architecture.

[0263] Based on the planning layer logical framework comprised of the aforementioned scenario modeling module, constraint definition module, and solution module, this embodiment further concretizes it into a complete engineering implementation system at the physical execution and real-time control level. Through a distributed multi-source heterogeneous sensing module, a coupled modeling and solution module, a cluster collaborative planning module, an intelligent winch execution module, and a robust flight control module, the aforementioned abstract model and constraints are mapped into a practically deployable perception-decision-control closed loop, such as... Figure 8 As shown.

[0264] The system adopts a hierarchical distributed architecture, and physically consists of a heterogeneous intelligent agent platform deployed on the sea surface and remote control terminals deployed on shore or mother ships.

[0265] Specifically, 1. Distributed multi-source heterogeneous sensing module

[0266] This module is responsible for constructing the high-precision spatiotemporal state space of the entire system, and adopts a three-in-one sensor layout of "sea-air-cable":

[0267] (1) Unmanned surface vessel (USV) sensing unit:

[0268] Positioning and orientation subsystem: Integrated dual-antenna RTK-GNSS receiver (main antenna located at the bow, secondary antenna at the stern, baseline length...) ),by Frequency output of the ship's absolute latitude and longitude coordinates and high-precision bow angle (Heading accuracy is better than) ).

[0269] Inertial navigation subsystem: Equipped with an industrial-grade fiber optic gyroscope IMU (zero bias stability) ),by Frequency acquisition of six-degree-of-freedom motion data of the ship's hull in waves (especially heave) Horizontal rocking , swaying This provides raw data for wave compensation.

[0270] Environmental perception subsystem: Equipped with a 16-line mechanical 3D LiDAR (Velodyne VLP-16), horizontal field of view. Vertical field of view It is used to construct local high-precision point cloud maps of photovoltaic arrays and operating waterways in real time.

[0271] (2) Unmanned Aerial Vehicle (UAV) Sensing Unit:

[0272] Precision height control subsystem: Millimeter-wave radar (operating frequency) integrated at the end of the spray boom. It has the ability to penetrate water mist and measures the vertical distance between the nozzle and the photovoltaic panel surface in real time. Measurement accuracy .

[0273] Visual perception subsystem: Front-facing binocular depth camera (Intel RealSense series) is used to identify the boundary and surface dirt thermal map of photovoltaic panels and assist in visual odometry (VIO) positioning.

[0274] (3) Cable status sensing unit:

[0275] Force sensor: An S-type high-frequency tension / compression sensor is connected in series at the intelligent winch cable guide, with a sampling frequency of... Range Real-time feedback of cable tension .

[0276] Length sensor: A multi-turn absolute encoder integrated into the winch motor, which calculates the real-time release length of the cable based on the transmission ratio. .

[0277] Spatiotemporal synchronization interface: All the sensor data mentioned above are synchronized with the unmanned surface vessel's main control computer at the sub-microsecond level via the PTP (Precision Time Protocol, IEEE 1588) protocol to eliminate timestamp drift between heterogeneous sensors.

[0278] 2. Coupled Modeling and Solving Module

[0279] This module serves as the system's "digital twin brain," deployed on an edge computing workstation mounted on the unmanned surface vessel (configured with NVIDIA Jetson AGX Orin, AI computing power 275 TOPS).

[0280] Variable mass physics engine: Built-in variable mass heterogeneous coupling dynamics model described in step S1 of embodiment one.

[0281] Real-time calculation: Engine subscribes to water pump flow feedback (Used to update fluid mass distribution), winch retraction and extension length And the motion status of the USV.

[0282] Coriolis force calculation: based on formula The system calculates in real time the additional Coriolis force and centrifugal force exerted by the high-speed water flow inside the pipe on each node of the cable, and updates the generalized mass matrix of the system. Coriolis force matrix .

[0283] Output: The frequency is used to output the predicted tension vector of the cable end on the drone to the control module. And the towing torque on the unmanned surface vessel.

[0284] 3. Cluster Collaborative Planning Module

[0285] This module runs on top of the ROS 2 (Robot Operating System) middleware and contains two core sub-engines:

[0286] (1) Topology anti-tangle engine:

[0287] Input: Real-time location coordinates of all heterogeneous units (USV+UAV) in the cluster.

[0288] Algorithm execution: Run the homotopy-type vector cross product algorithm described in step S2 of Example 1 in real time. Calculate the relative positional relationship between the projected line segments of any two cables.

[0289] Virtual geofence generation: When the shortest distance between two line segments is detected. At that time, a high potential energy repulsion field is dynamically generated and superimposed on the cost map of path planning as a virtual fence.

[0290] (2) Hierarchical path planner:

[0291] Global layer: Based on the geometric constraints of the photovoltaic array and the MILP task allocation results, a global waypoint sequence is generated for the unmanned surface vessel.

[0292] Local layer: Executes the "boat-to-machine scanning" algorithm described in Example 1. Optimizes the unmanned surface vessel's course based on the direction of the ocean waves. According to the velocity coupling formula Simultaneously generate propulsion speed commands for the unmanned surface vessel and reciprocating scanning trajectories for the unmanned aerial vehicle.

[0293] 4. Intelligent winch execution module

[0294] This module is the physical actuator that implements "tight coupling":

[0295] (1) Hardware composition: includes high power density servo motor, precision planetary reducer, cable arrangement mechanism and conductive slip ring.

[0296] (2) Control Mode:

[0297] Mode A: Constant Tension Follow-up: During normal operation, the cable tension is maintained at... The micro-tension range automatically expands and contracts as the drone moves.

[0298] Mode B: Active wave compensation: When the IMU detects vertical acceleration of the hull. Triggered at time. The controller receives the predicted hull heave rate calculated in step S3. Execute speed command ( As a compensation coefficient, take 0.9), drive the drum to rotate rapidly in the opposite direction, physically isolating the effect of waves on the cable length.

[0299] 5. Robust Flight Control Module

[0300] This module is deployed on the UAV's flight control computer (Pixhawk 6X) and executes the strategy described in step S3 of Example 1:

[0301] (1) Cascaded control architecture: Outer loop (position loop): Generates the desired velocity based on the scanning trajectory deviation. Inner loop (attitude / angular rate loop): Performs high-frequency attitude stabilization.

[0302] (2) Tension feedforward compensation: Receives the predicted tension fluctuation value from the coupled solution module. Map it to thrust feedforward. This is directly superimposed on the output of the inner loop controller.

[0303] ;

[0304] Effect: Before the waves cause the cable to suddenly tighten Increase rotor speed in advance to ensure that the drone's altitude relative to the solar panel remains stable in sea state 4-5. Within this range, to prevent collisions or cleaning failure.

[0305] In this embodiment, the scene modeling module constructs a heterogeneous coupled multibody dynamics model including UAVs, water supply cables, and unmanned surface vessels (USVs) based on the geometric distribution of the photovoltaic array and real-time sea conditions. The water supply cable is modeled as a dynamic load with fluid transmission characteristics, accurately reproducing the system behavior under complex sea conditions. The constraint definition module provides reliable physical boundaries for the spatiotemporal trajectory intersections of multiple UAV-cable-vessel units through spatiotemporal constraints and entanglement avoidance constraints, effectively mitigating operational risks. The solution module divides the array into sub-regions, aiming to minimize the total energy consumption of the cluster and balance the completion time of operations. It uses a two-layer planning architecture to generate execution instructions, achieving optimized resource allocation and efficient task distribution. Thus, a complete heterogeneous coupled cluster operation architecture for UAVs, cables, and USVs is constructed based on scene modeling, constraint definition, and solution optimization, enabling efficient collaboration of marine photovoltaic cleaning tasks.

[0306] Example 3

[0307] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the marine photovoltaic cleaning method based on a heterogeneous coupling body cluster of a cable-stayed boat as described in Embodiment 1 above.

[0308] Example 4

[0309] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the marine photovoltaic cleaning method based on a heterogeneous coupling body cluster of a cable boat as described in Embodiment 1 above.

[0310] The steps or modules involved in Embodiments 2 to 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0311] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for cleaning marine photovoltaic systems based on a heterogeneous coupling system cluster of cable-stayed boats, characterized in that, include: Based on the geometric distribution of the photovoltaic array and real-time sea state information, a heterogeneous coupled multibody dynamics model including a UAV, a water supply cable, and an unmanned surface vessel (USV), as well as a photovoltaic array grid map, is constructed. Specifically, the water supply cable is modeled as a dynamic load with fluid transmission characteristics. This involves discretizing the cable into nodes and units between nodes. The dynamic equations of each node include elastic tension, gravity and buoyancy, environmental loads, and additional forces generated by fluid transmission within the pipe. These additional forces include Coriolis force, centrifugal force, and momentum thrust. The construction of the photovoltaic array grid map specifically includes: using a lidar mounted on the USV... The process involves acquiring raw point cloud data of the marine environment; filtering the raw point cloud data based on the estimated height range, effective ranging range, and outliers of the photovoltaic array to obtain local point cloud data; using the pose and extrinsic parameter calibration matrix of the unmanned surface vessel (USV), transforming the local point cloud data from the lidar coordinate system to the global geographic coordinate system; accumulating discrete single-frame point clouds through continuous frames to create a dense global point cloud map; projecting the global point cloud map onto a two-dimensional plane and constructing an initial grid map based on the operating area of ​​the photovoltaic array; identifying obstacles based on the point cloud density within the grid, and generating a grid map containing obstacle areas, passable areas, and unknown areas. Define the physical boundary constraints of the dynamic model, including the work space constraints and the anti-entanglement constraints to avoid the spatiotemporal trajectory intersection of the cable-boat unit during cluster operations; The photovoltaic array to be cleaned is divided into several sub-regions. With the goal of minimizing the total energy consumption of the cluster and balancing the completion time of the tasks, a mixed integer linear programming model is constructed, and a two-level programming architecture is used to generate execution instructions.

2. The marine photovoltaic cleaning method based on a heterogeneous coupling body cluster of a cable-stayed boat as described in claim 1, characterized in that, The operational space constraints include the UAV operational altitude and tension coupling constraints and the relative yaw angle sector constraints. Specifically, the anti-entanglement constraint is constructed for cluster operations based on topological homotopy, defining the projection of each pair of machine-cable-boat units onto the sea surface as a dynamic no-fly zone, ensuring that any two groups of units do not intersect each other in the spatiotemporal trajectory.

3. The marine photovoltaic cleaning method based on a heterogeneous coupling body cluster of a cable-stayed boat as described in claim 1, characterized in that, After constructing the mixed-integer linear programming model, the process also includes solving the model, allocating sub-regions to each heterogeneous coupled entity, and generating global baseline routes and docking point sequences for each unmanned surface vessel.

4. The marine photovoltaic cleaning method based on a heterogeneous coupling body cluster of a cable-stayed boat as described in claim 1, characterized in that, The two-tier planning architecture specifically includes: The upper layer is used to plan collision-free paths for unmanned surface vessels in the photovoltaic array waterways and to optimize the heading angle based on the direction of the waves to reduce rolling. The lower layer is used to plan the UAV to perform reciprocating full-coverage scanning with the current length of the water supply cable as the dynamic radius while the unmanned surface vessel advances at low speed along the planned path; It also includes a coordination layer, which is used to calculate the coupling matching ratio between the UAV scanning speed and the UAV propulsion speed to ensure that the cable is always kept in a slightly tensioned catenary state during the cleaning process.

5. The marine photovoltaic cleaning method based on a heterogeneous coupling body cluster of a cable-stayed boat as described in claim 1, characterized in that, It also includes real-time acquisition of the unmanned surface vessel's six degrees of freedom motion data and cable end tension; introduces a tension feedforward compensation mechanism to predict the cable's dragging interference force on the UAV at the next moment using the UAV's motion state, and adjusts the UAV's attitude and rotor speed in advance to counteract the systemic disturbances caused by ocean waves.

6. A marine photovoltaic cleaning system based on a heterogeneous coupler cluster of a cable-stayed boat, based on the marine photovoltaic cleaning method based on a heterogeneous coupler cluster of a cable-stayed boat as described in claim 1, characterized in that, include: The scene modeling module is configured to construct a heterogeneous coupled multibody dynamics model, including UAVs, water supply cables, and unmanned surface vessels, as well as a grid map of the photovoltaic array, based on the geometric distribution of the photovoltaic array and real-time sea condition information; among which, the water supply cable is modeled as a dynamic load with fluid transmission characteristics. The constraint definition module is configured to define the physical boundary constraints of the dynamic model, including the workspace constraints and the anti-entanglement constraints for avoiding the spatiotemporal trajectory intersection of the cluster operation timing-cable-boat unit; The solution module is configured to divide the photovoltaic array to be cleaned into several sub-regions, construct a mixed-integer linear programming model with the joint optimization objectives of minimizing the total energy consumption of the cluster and the balance of the operation completion time, and generate execution instructions using a two-layer programming architecture.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the marine photovoltaic cleaning method based on a heterogeneous coupling body cluster of a cable boat as described in any one of claims 1-5.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the marine photovoltaic cleaning method based on a heterogeneous coupling body cluster of a cable boat as described in any one of claims 1-5.

Images