A full-coverage cleaning modular control method and system for a photovoltaic cleaning robot

Abstract

The present application provides a modular control method and system for full-coverage cleaning of a photovoltaic cleaning robot. Among them, the method constructs a three-dimensional terrain model of the photovoltaic panel by collecting laser ranging, photovoltaic panel surface temperature distribution, ambient airflow and position relationship data; generates an obstacle distribution map based on the model, and dynamically analyzes the gap width and attachment height parameters of the cleaning area. Based on this, the bending angle and telescopic amplitude of the connecting parts of the cleaning module are adjusted in real time, and the robot posture information is updated synchronously. Combining position data, obstacle maps and the posture of the entire robot, the uncleaned area identification and dynamic coverage path are generated, and the information is shared through a wireless network. The system dynamically allocates the working areas and travel directions of multiple robots to achieve efficient and coordinated full-coverage cleaning of the photovoltaic array. The technical solution provided by the present application realizes intelligent path coordination and full-coverage operation of multiple cleaning robots through multi-source environmental perception and dynamic posture adjustment.

Description

Technical Field

[0001] The present application relates to the technical field of dynamic environment perception and collaborative path planning, and in particular to a full-coverage cleaning modular control method and system for a photovoltaic cleaning robot. Background Art

[0002] In the operation and maintenance of large-scale photovoltaic power plants, the surfaces of photovoltaic panel arrays, exposed to the outdoor environment for extended periods, are constantly subject to issues such as snow accumulation, dust deposition, and the adhesion of bird droppings. These deposits significantly block light transmission and induce hot spot effects, leading to nonlinear degradation of power generation efficiency. To ensure the stability of the power plant's output power, cleaning robots must dynamically adapt to installation angle deviations between different photovoltaic panel units, address panel warping caused by temperature differences due to sunlight, and handle fluctuations in mechanical tolerances of panel gap widths in real time. Furthermore, they must establish a collaborative coverage mechanism for multiple robot clusters in ultra-large arrays and possess the ability to respond instantaneously to sudden obstacles, achieving the coordinated optimization of cleaning completeness and operational efficiency.

[0003] A representative solution in the current field is a preset global path planning system based on laser synchronous positioning and mapping technology. During the initialization phase, the system generates a high-precision static point cloud model of the photovoltaic array through multi-line lidar scanning, and pre-calculates a rasterized traversal path covering the entire working area based on this model. The robot is equipped with an inertial measurement unit to achieve closed-loop feedback on posture during movement, strictly following the preset trajectory to perform cleaning actions, and only triggering emergency braking and avoidance procedures when the robotic arm contact sensor detects a physical collision. The multi-robot cluster periodically broadcasts its own coordinates through a low-power ZigBee mesh network, and adopts a static geographic fence segmentation strategy to allocate basic working areas, forming a preliminary collaborative work framework.

[0004] However, this solution has fundamental technical limitations. The statically constructed point cloud map cannot integrate the dynamic deformation characteristics of photovoltaic panels under real sunlight conditions, which significantly increases the probability of rigid collision between the end effector of the robotic arm and the deformed panel surface. When faced with sudden attachments with temporal and spatial randomness, the preset fixed cleaning path can only perform passive avoidance operations and cannot dynamically reconstruct the cleaning trajectory, resulting in the omission of cleaning of key contaminated areas. The more prominent problem is that the collaborative architecture lacks a global task dynamic scheduling engine. When some robots stop due to motor overheating or battery exhaustion, the system cannot autonomously reconstruct the task allocation map of the uncleaned area, which ultimately leads to the dual degradation of the overall cleaning completeness and operation timeliness of the array. Summary of the Invention

[0005] The present application provides a full-coverage cleaning modular control method and system for a photovoltaic cleaning robot, which is used to solve the problems in the prior art such as lack of dynamic terrain adaptability, delayed response to sudden obstacles, and low efficiency of multi-machine collaboration caused by static environment modeling and fixed path planning.

[0006] In a first aspect, the present application provides a modular control method for full-coverage cleaning of a photovoltaic cleaning robot, comprising:

[0007] Obtain laser ranging data, temperature distribution data on the photovoltaic panel surface, ambient airflow data, and positional relationship data between the photovoltaic panels to construct a three-dimensional terrain model of the photovoltaic panel surface;

[0008] generating an obstacle distribution map corresponding to the photovoltaic panel array based on the three-dimensional terrain model;

[0009] According to the gap width parameters and attachment height parameters of the area cleaned by the photovoltaic cleaning robot in the obstacle distribution map, the bending angle and extension range of the connecting components in the corresponding photovoltaic cleaning robot are dynamically adjusted, and the posture information of the corresponding photovoltaic cleaning robot is updated;

[0010] Based on the position relationship data, the obstacle distribution map and the posture information of all photovoltaic cleaning robots, generating the uncleaned area identification and dynamic coverage path of the cleaning robot;

[0011] The uncleaned area identifier and the coverage path of each cleaning robot are shared through a wireless communication network, and the working areas and travel directions of multiple cleaning robots are dynamically allocated to achieve full coverage cleaning of the photovoltaic panel array.

[0012] Optionally, the step of sharing the uncleaned area identifier and the coverage path of each cleaning robot through a wireless communication network and dynamically allocating the working areas and travel directions of the plurality of cleaning robots to achieve full coverage cleaning of the photovoltaic panel array includes:

[0013] The coordinate set of the uncleaned grid cells in the uncleaned area identification and the coverage path of each cleaning robot are transmitted to the central control unit of all cleaning robots in real time through the wireless communication network;

[0014] In the central control unit, the uncleaned grid cells with high matching degree are preferentially allocated to the corresponding cleaning robots according to the spatial distance of the path extension line, and the remaining uncleaned grid cells are secondary allocated to generate an area allocation mapping table;

[0015] Generate a dedicated work area boundary for each robot based on the area allocation mapping table, and generate a path connection instruction including a steering angle correction value and a speed adjustment value based on the geometric relationship of the robot's current coverage path;

[0016] The boundary of the exclusive working area and the path connection instruction are sent to the driving mechanism of the corresponding robot, driving the corresponding robot to complete the traversal of the uncleaned grid cells in the exclusive working area, and realizing full coverage cleaning of the photovoltaic panel array through the feedback of the working status and the iterative update of the uncleaned area identification.

[0017] Optionally, in the central control unit, uncleaned grid cells with high matching degree are preferentially allocated to corresponding cleaning robots according to the spatial distance of the path extension line, and secondary allocation is performed on the remaining uncleaned grid cells to generate an area allocation mapping table, including:

[0018] In the central control unit, the coordinates of the end points of the coverage paths of the cleaning robots are parsed as the expected positions. For each uncleaned grid cell, the vertical distance from the uncleaned grid cell to the extension line of the current coverage path of each cleaning robot is calculated as a matching parameter. The uncleaned grid cells whose matching parameters are less than a preset threshold are marked as high-matching cells.

[0019] Traverse all combinations of cleaning robots and the uncleaned grid cells, assign the cells with high matching degrees to the cleaning robots corresponding to the extended coverage paths, and update the cell assignment status;

[0020] For the remaining cells whose allocation status has not been updated, the straight-line distance from the expected position of the cleaning robot to the coordinates of the remaining cells is calculated, and each remaining cell is allocated to the available cleaning robot with the smallest straight-line distance;

[0021] The allocation results are integrated to generate an area allocation mapping table that clearly records the responsible cleaning robot for each uncleaned grid cell.

[0022] Optionally, the generating of the uncleaned area identification and the dynamic coverage path of the cleaning robot based on the position relationship data, the obstacle distribution map and the posture information of all photovoltaic cleaning robots includes:

[0023] Obtain the boundary coordinates of the photovoltaic panel array and the position information of all photovoltaic cleaning robots in the position relationship data;

[0024] Divide the cleaning area into a uniform grid cell array according to the boundary coordinates of the photovoltaic panel array, and mark the inaccessible area in the grid cell array according to the obstacle cell coordinate position and obstacle type identifier in the obstacle distribution map;

[0025] Based on the historical posture information of all photovoltaic cleaning robots, the coordinates of the grid cells traversed by the photovoltaic cleaning robots are backtracked to generate a cleaned area map, the cleaned area map is compared with the total grid cells of the photovoltaic panel array, the unmarked grid cells are defined as uncleaned grid cells, and a unique area identifier is assigned to each uncleaned grid cell to generate an uncleaned area identifier;

[0026] Calculating the actual path width that the photovoltaic cleaning robot can pass through by combining the adjusted posture inclination angle of the connecting component and the gap width parameter in the obstacle distribution map;

[0027] Based on the uncleaned area identifier, the set of uncleaned grid unit coordinates is identified, and the current position coordinates in the posture information are used as the starting point. The inaccessible area is avoided, and adjacent uncleaned grid units are connected under the constraint of the actual path width. A continuous moving direction sequence and a moving speed value matching each path segment are generated to generate a dynamic coverage path.

[0028] Optionally, generating an obstacle distribution map corresponding to the photovoltaic panel array based on the three-dimensional terrain model includes:

[0029] Analyzing the three-dimensional grid structure in the three-dimensional terrain model, extracting the height coordinate value of each grid cell, calculating the height change difference of the height coordinate values ​​between adjacent grid cells, and identifying grid cells whose height change difference exceeds a preset height change threshold as potential obstacle areas;

[0030] Analyze the temperature value of each grid cell, calculate the degree of deviation between the temperature value and the average temperature value, and define the grid cells with the degree of deviation greater than a preset temperature deviation threshold as temperature abnormal areas;

[0031] Analyze the airflow direction attributes of each grid cell, extract the change trend of the airflow direction value, and identify the grid cells with abnormal change trends as airflow affected areas based on a preset airflow change trend threshold;

[0032] The potential obstacle area, the temperature anomaly area and the airflow affected area are combined and marked as obstacle units, and an obstacle distribution map is generated on the photovoltaic panel array based on the coordinate position of the obstacle unit, combined with the obstacle type identifier and the obstacle height value.

[0033] Optionally, dynamically adjusting the bending angle and extension range of the connecting components of the corresponding photovoltaic cleaning robot according to the gap width parameter and the attachment height parameter of the area cleaned by the photovoltaic cleaning robot in the obstacle distribution map, and updating the posture information of the corresponding photovoltaic cleaning robot, includes:

[0034] Extracting gap width parameters and attachment height parameters of the current cleaning area of ​​the photovoltaic cleaning robot from the obstacle distribution map, and calculating the maximum horizontal expansion and contraction amplitude and the minimum vertical bending angle of the connecting component based on the gap width parameters and the attachment height parameters respectively;

[0035] generating a drive control instruction according to the maximum telescopic amplitude and the minimum bending angle, and sending the drive control instruction to a connecting component drive mechanism of a corresponding photovoltaic cleaning robot;

[0036] The actual extension and contraction range of the connecting component is adjusted to the maximum extension and contraction range by the driving mechanism, and the bending angle is adjusted to the minimum bending angle at the same time;

[0037] Based on the adjusted telescopic amplitude and bending angle, combined with the real-time position coordinates of the corresponding photovoltaic cleaning robot in the position relationship data, its new moving direction coordinates and posture tilt angle are calculated to update the posture information of the corresponding photovoltaic cleaning robot.

[0038] Optionally, acquiring laser ranging data, temperature distribution data on the photovoltaic panel surface, ambient airflow data, and positional relationship data between the photovoltaic panels to construct a three-dimensional terrain model of the photovoltaic panel surface includes:

[0039] The laser distance measuring device installed on the photovoltaic cleaning robot body scans the surface of the photovoltaic panel and continuously obtains the vertical distance between the robot chassis and the photovoltaic panel surface at multiple measuring points as laser distance measurement data;

[0040] The temperature sensor installed on the photovoltaic cleaning robot body synchronously obtains the temperature value of each point on the photovoltaic panel surface to generate temperature distribution data;

[0041] The airflow sensor installed on the photovoltaic cleaning robot body obtains the direction and speed of the ambient airflow near the photovoltaic panel surface to generate ambient airflow data;

[0042] The positional relationship data between the photovoltaic cleaning robot body and the photovoltaic panel is obtained by a position positioning device installed on the photovoltaic cleaning robot body;

[0043] The laser ranging data is converted into height coordinates in three-dimensional space to form a basic spatial shape of the photovoltaic panel surface. The temperature distribution data and the ambient airflow data are associated as additional attributes to the corresponding three-dimensional spatial positions in the basic spatial shape. Based on the position relationship data and the basic spatial shape, a three-dimensional terrain model of the photovoltaic panel surface is constructed.

[0044] In a second aspect, the present application provides a full-coverage cleaning modular control system for a photovoltaic cleaning robot, comprising:

[0045] An acquisition module is used to obtain laser ranging data, temperature distribution data on the surface of photovoltaic panels, ambient airflow data, and positional relationship data between photovoltaic panels to construct a three-dimensional terrain model of the photovoltaic panel surface;

[0046] A generation module, configured to generate an obstacle distribution map corresponding to the photovoltaic panel array based on the three-dimensional terrain model;

[0047] A control module, configured to dynamically adjust the bending angle and extension range of the connecting components of the corresponding photovoltaic cleaning robot according to the gap width parameter and the attachment height parameter of the area cleaned by the photovoltaic cleaning robot in the obstacle distribution map, and update the posture information of the corresponding photovoltaic cleaning robot;

[0048] A planning module, configured to generate a dynamic coverage path and an uncleaned area identifier for the cleaning robot based on the position relationship data, the obstacle distribution map, and the posture information of all photovoltaic cleaning robots;

[0049] The coordination module is used to share the uncleaned area identification and the real-time posture information of each cleaning robot through the wireless communication network, and dynamically allocate the working areas and travel directions of multiple cleaning robots to achieve full coverage cleaning of the photovoltaic panel array.

[0050] In a third aspect, an embodiment of the present application provides a computing device comprising a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a full-coverage cleaning modular control method for a photovoltaic cleaning robot as described in the first aspect above.

[0051] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program. When the computer program is executed by a computer, it implements the full-coverage cleaning modular control method of a photovoltaic cleaning robot as described in the first aspect.

[0052] In the example of the present application, laser ranging data, temperature distribution data on the surface of photovoltaic panels, ambient airflow data and positional relationship data between photovoltaic panels are obtained to construct a three-dimensional terrain model of the surface of photovoltaic panels; based on the three-dimensional terrain model, an obstacle distribution map corresponding to the photovoltaic panel array is generated; according to the gap width parameters and attachment height parameters of the area cleaned by the photovoltaic cleaning robot in the obstacle distribution map, the bending angle and extension amplitude of the connecting parts in the corresponding photovoltaic cleaning robot are dynamically adjusted, and the posture information of the corresponding photovoltaic cleaning robot is updated; based on the positional relationship data, the obstacle distribution map and the posture information of all photovoltaic cleaning robots, the uncleaned area identification and dynamic coverage path of the cleaning robot are generated; the uncleaned area identification and the coverage path of each cleaning robot are shared through a wireless communication network, and the working areas and travel directions of multiple cleaning robots are dynamically allocated to achieve full coverage cleaning of the photovoltaic panel array.

[0053] The technical solution of this application has the following beneficial effects:

[0054] This application dynamically constructs a three-dimensional terrain model of photovoltaic panels by integrating laser ranging, temperature distribution, ambient airflow and position relationship data, and generates an accurate obstacle distribution map based on this, so that the cleaning robot can adaptively adjust the bending angle and extension amplitude of the connecting parts according to the gap width and attachment height parameters obtained in real time, and synchronously update the posture information; at the same time, based on multi-source data and the posture of the entire robot, it generates uncleaned area identification and dynamic coverage paths, and uses wireless communication networks to share information and dynamically allocate the working areas and travel directions of multiple machines, ultimately achieving zero omissions, high-efficiency collaborative full-coverage cleaning of photovoltaic panel arrays under complex working conditions, and completely solving the problems of mechanical collision risks, sudden obstacle response lag and multi-machine collaboration failure caused by static environment modeling in the existing technology.

[0055] Furthermore, the central control unit uses the wireless network to obtain the coverage paths and uncleaned grid coordinate sets of all robots in real time. A spatial distance-priority strategy based on the path extension is used to assign highly matched uncleaned grids to the corresponding robots. A secondary allocation is performed on the remaining grids to generate a regional allocation mapping table. Based on this, the boundaries of the exclusive work area are delineated, and connection instructions containing steering angle correction values ​​and speed adjustment values ​​are generated in combination with the robot's current path geometry. Ultimately, the robot is driven to complete grid traversal within the exclusive area, and full array coverage is achieved through state feedback and iterative updates of uncleaned identification. This mechanism achieves seamless integration of task allocation and robot motion trajectories by dynamically matching the spatial correlation between the path extension and the uncleaned grid, significantly reducing energy consumption during idle travel. The coordinated optimization of exclusive area boundaries and connection instructions ensures continuous operation of multiple robots with synchronized steering angle and speed adjustments, avoiding path conflicts and repeated cleaning. Combined with closed-loop iterative updates of state feedback, the array cleaning completeness rate is significantly improved, completely eliminating the coordination failure problem caused by manual intervention in traditional solutions.

[0056] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0058] Figure 1 A flowchart of a full-coverage cleaning modular control method of a photovoltaic cleaning robot provided by the present application is shown;

[0059] Figure 2 A scene diagram showing a modular control method for full-coverage cleaning of a photovoltaic cleaning robot provided in this application is shown;

[0060] Figure 3 A schematic diagram of the structure of a full-coverage cleaning modular control system of a photovoltaic cleaning robot provided by the present application is shown;

[0061] Figure 4 A schematic structural diagram of a computing device provided by the present application is shown. DETAILED DESCRIPTION

[0062] In order to enable people skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0063] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0064] Research has shown that existing pre-set path solutions based on static laser maps face fundamental limitations in the full-coverage operations of photovoltaic cleaning robots. Their pre-built point cloud models are unable to respond to the dynamic warping and deformation of the panels caused by solar temperature differences, leading to frequent collisions between the robotic arm and the deformed panels. Fixed trajectory planning can only passively avoid unexpected attachments and is unable to autonomously reconstruct the cleaning path, resulting in missed contaminated areas. Multi-machine coordination mechanisms rely on static area divisions, lacking the ability to dynamically schedule global tasks when some robots malfunction, resulting in a dual degradation of cleaning integrity and timeliness. The core of this contradiction lies in the severe mismatch between static environmental modeling and actual dynamic working conditions, necessitating the development of an adaptive collaborative control system that integrates multi-source perception.

[0065] In response to the above problems, the present invention proposes an adaptive collaborative control scheme based on multi-source perception fusion. It constructs a dynamic three-dimensional terrain model by collecting laser ranging, temperature distribution, ambient airflow, and position relationship data in real time. It dynamically adjusts the bending angle and extension range of the robot joints based on the gap width and attachment height parameters in the obstacle distribution map, and synchronously updates the posture information. At the same time, it combines the posture of the entire cluster with the obstacle map to generate uncleaned area identification and dynamic coverage paths, and shares data through a wireless network. It autonomously allocates the working areas and travel directions of multiple machines based on the spatial matching degree of the path extension line. This scheme fundamentally eliminates the collision risks caused by thermal deformation of the plate surface, achieves zero-omission cleaning of areas with sudden attachments, and ensures the cleaning completeness rate through dynamic reconstruction of global tasks in the event of a single machine failure. Ultimately, it forms a closed-loop control mechanism of environmental perception, dynamic decision-making, and collaborative execution, which completely solves the adaptability defects of static modeling systems under complex working conditions.

[0066] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0067] Figure 1 The present invention provides a flowchart of a modular control method for full coverage cleaning of a photovoltaic cleaning robot, as shown in FIG. Figure 1 As shown, the method includes:

[0068] 101. Obtain laser ranging data, temperature distribution data on the photovoltaic panel surface, ambient airflow data, and positional relationship data between the photovoltaic panels to construct a three-dimensional terrain model of the photovoltaic panel surface;

[0069] Optionally, step 101 may specifically include the following steps:

[0070] 1011. Scan the surface of the photovoltaic panel using a laser ranging device installed on the photovoltaic cleaning robot body, and continuously obtain the vertical distance between the robot chassis and the photovoltaic panel surface at multiple measurement points as laser ranging data;

[0071] 1012. Using a temperature sensor installed on the photovoltaic cleaning robot body, synchronously obtain the temperature value of each point on the surface of the photovoltaic panel to generate temperature distribution data;

[0072] 1013. Obtain the direction and speed of the ambient airflow near the photovoltaic panel surface by using an airflow sensor installed on the photovoltaic cleaning robot body to generate ambient airflow data;

[0073] 1014. Obtaining positional relationship data between the photovoltaic cleaning robot body and the photovoltaic panel through a position positioning device installed on the photovoltaic cleaning robot body;

[0074] 1015. Convert the laser ranging data into height coordinates in three-dimensional space to form a basic spatial shape of the photovoltaic panel surface, associate the temperature distribution data and the ambient airflow data as additional attributes with the corresponding three-dimensional spatial positions in the basic spatial shape, and construct a three-dimensional terrain model of the photovoltaic panel surface based on the position relationship data and the basic spatial shape.

[0075] In the above scheme, laser ranging data refers to a dynamic measurement set reflecting the spatial distance between the robot chassis and the photovoltaic panel surface. It contains vertical distance information based on the light wave's time of flight, including discrete point cloud coordinates and continuous surface interpolation results, and can be used to construct a terrain geometric benchmark. Temperature distribution data refers to a two-dimensional field variable that characterizes the thermodynamic state of the photovoltaic panel surface. It contains temperature gradient information converted from infrared radiation energy, including spatial coordinate-bound temperature values ​​and local thermal deformation correlation characteristics, which can be used to identify areas at risk of panel warping. Ambient airflow data refers to a vector parameter set that describes the near-field fluid motion characteristics of the photovoltaic panel. It contains turbulence correction results for airflow direction angles and velocity values, including the equivalent laminar flow field calculated by the Bernoulli equation, and can be used to predict the migration trajectory of attachments. Position relationship data refers to a six-degree-of-freedom matrix that quantifies the spatial posture relationship of the robot body relative to the photovoltaic panel. It contains real-time solutions of Euclidean distance and yaw angle, including rigid transformation parameters for coordinate system transformation, and can be used to unify the spatial benchmark of multi-source data. A three-dimensional terrain model refers to a dynamic digital surface that integrates geometric forms and physical properties. It contains the temperature field and airflow vector superposition information of the basic spatial shape, including the vertex attribute binding and vector field mapping of the surface mesh, which can provide an environmental twin foundation for adaptive cleaning.

[0076] In the embodiment of the present application, first, in step 1011, a time-of-flight ranging laser radar is used to emit a 905nm pulsed laser along a preset raster scanning path on the surface of the photovoltaic panel. By calculating the time difference between the transmitted and received signals, the vertical distance of each measurement point is calculated according to the formula: ,in is the time difference, and c is the speed of light; a raw discrete point cloud dataset is generated. To eliminate mechanical vibration noise, a Kalman filter is used to smooth 10 consecutive frames of data. The final output is calibrated laser ranging data, which serves as the geometric basis for 3D terrain construction.

[0077] Next, the infrared sensor array integrated in the coaxial position of the laser radar in step 1021 synchronously scans the spatial coordinate points in the laser ranging data. Based on Planck's radiation law, the received 8-14μm band infrared radiation intensity is converted into a temperature value. The specific calculation formula is as follows: ,in is the infrared wavelength, E is the infrared radiation intensity, and C1 and C2 are radiation constants. Through spatial mapping, the temperature values ​​are bound to the corresponding points in the laser ranging data to generate an enhanced point cloud with temperature attributes, establishing a geometric and thermodynamic coupling data layer.

[0078] Ultrasonic anemometers deployed at the four corners of the robot then collected real-time flow field parameters in the laser scanning area. The Reynolds-averaged Navier-Stokes equations were used to correct for the turbulent boundary layer effects on the photovoltaic panel surface. The equations are as follows: ,in , is the velocity vector component, p is the pressure, is the air density, v is the air kinematic viscosity, is the Reynolds stress tensor. The equivalent laminar velocity and direction angle are obtained by solving the problem, and the airflow vector set is output. Bilinear interpolation is used to match it to the spatial grid coordinates of the enhanced point cloud.

[0079] Then, based on the three-base station topology network constructed by the UWB positioning system, the center position of the robot body is solved by the arrival time difference algorithm. Let the base station coordinates be , the signal arrival time difference is , establish the hyperbolic equations: , where (X, Y) is the center coordinate of the robot body, is the coordinate of the mth UWB base station, The Levenberg-Marquardt nonlinear optimization algorithm is used to solve the robot's six-degree-of-freedom posture and complete the unified transformation of the photovoltaic panel coordinate system.

[0080] Finally, the enhanced point cloud is first input into the non-uniform rational B-spline surface generation algorithm to reconstruct a continuous surface, and then the temperature data is written into the surface mesh as vertex attributes, and the airflow vector is converted into the surface tangent field; finally, the rotation matrix R and translation vector T of the pose Pose are used to perform affine transformation to output a three-dimensional terrain model that integrates the geometric shape, temperature field and airflow vector.

[0081] In actual application, when a cleaning robot entered the area of ​​panel 3 of a photovoltaic array, it scanned the panel surface using a time-of-flight (TOF) lidar with a wavelength of 905nm and a grid density of 5cm×5cm. It detected that at coordinates (2.3m, 1.7m), the vertical distance increased from the standard value of 150mm to 163mm due to the expansion of the ice-water mixture. At coordinates (2.5m, 2.1m), the panel surface deflected sharply to 142mm due to a hot spot. A Kalman filter smoothed 10 consecutive frames of data and generated a point cloud of the abnormal area. Simultaneously, a coaxial infrared sensor array converted the radiation intensity in the 8-14μm band to temperature based on Planck's radiation law. The deflected area reached 85°C due to the hot spot effect, while the ice-water mixture area maintained a constant temperature of 0°C, generating a temperature-bound point cloud. A four-corner ultrasonic anemometer then detected a northwesterly wind speed of 7.2 m / s with gusts of 9.5 m / s. After correcting for boundary layer turbulence using the Reynolds-averaged Navier-Stokes equations, an equivalent laminar wind speed of 5.8 m / s and a directional angle of 312° were output, which were then interpolated and mapped to temperature-bound point cloud coordinates. The UWB positioning system then solved the robot's pose, Pose = [2.35 m, 1.92 m, 0.15 m, 5°, 0°, 312°], using the TDOA hyperbola equation, confirming its location in the northeast quadrant of plate 3. Finally, the temperature-bound point cloud was input into a NURBS surface algorithm to reconstruct a continuous surface, marking a 13 mm ridge at (2.3 m, 1.7 m) and an 8 mm dent at (2.5 m, 2.1 m). The corrected airflow vectors were converted to a surface tangent field. An affine transformation was performed based on the rotation matrix R and translation vector T of the pose, outputting a three-dimensional terrain model.

[0082] The overall solution described above, 101, constructs a dynamically evolving three-dimensional terrain model of photovoltaic panels through the simultaneous collection and fusion processing of multi-source sensor data, enabling accurate digital representation of environmental conditions under complex working conditions. This solution overcomes the inherent limitations of static maps by using laser ranging data to establish a high-precision spatial reference. This solution, combined with temperature distribution data, provides real-time perception of thermal deformation trends on the panel surface. It predicts the migration path of attachments using ambient airflow data, and unifies multi-source spatial references based on positional relationship data. The resulting three-dimensional model integrates multi-dimensional attributes such as geometric deformation, temperature field gradients, and airflow vector fields, providing the cleaning robot with the core decision-making basis for millimeter-level terrain adaptability, instantaneous response to sudden obstacles, and multi-machine coordinated scheduling, laying the foundation for environmental perception for fully autonomous closed-loop control.

[0083] 102. Generate an obstacle distribution map corresponding to the photovoltaic panel array based on the three-dimensional terrain model;

[0084] Optionally, step 102 may specifically include the following steps:

[0085] 1021. Analyze the three-dimensional grid structure in the three-dimensional terrain model, extract the height coordinate value of each grid cell, calculate the height change difference of the height coordinate values ​​between adjacent grid cells, and identify the grid cells whose height change difference exceeds a preset height change threshold as potential obstacle areas;

[0086] 1022. Analyze the temperature value of each grid cell, calculate the degree of deviation between the temperature value and the average temperature value, and define the grid cells where the degree of deviation is greater than a preset temperature deviation threshold as temperature abnormality areas;

[0087] 1023. Analyze the airflow direction attributes of each grid cell, extract the change trend of the airflow direction value, and identify the grid cells with abnormal change trends as airflow impact areas based on a preset airflow change trend threshold;

[0088] 1024. Combine the potential obstacle area, the temperature anomaly area, and the airflow impact area and mark them as obstacle units. Based on the coordinate positions of the obstacle units, combined with the obstacle type identifier and the obstacle height value, generate an obstacle distribution map on the photovoltaic panel array.

[0089] In the above scheme, the height change difference refers to the absolute differential value reflecting the vertical coordinate offset between adjacent grid cells, including the mutation amplitude and spatial gradient direction information of the elevation data, which is used to identify physical surface protrusions or depressions. The degree of temperature deviation refers to the normalized deviation intensity of the temperature value of a single grid cell from the global average temperature of the array, including the characteristics of thermodynamic diffusion anomalies, which can indicate the risk of local hot spot effects or structural deformation. The airflow direction attribute refers to a set of vector parameters that describes the state of airflow movement in the grid cell, including the time-series change rate of the directional angle and the turbulence intensity spectrum characteristics, which are used to evaluate the degree of interference of wind-induced mechanical vibration on cleaning stability. The obstacle type identifier refers to a classification coding label generated based on the association of multi-source sensor data, including obstacle cause attributes and risk level information, to support the generation of differentiated obstacle avoidance strategies. In the embodiment of the present application, the three-dimensional terrain model is first subjected to spatial grid discretization processing through step 1021, and the height gradient modulus of the 8 neighborhoods of each grid cell is calculated using the Sobel edge detection operator. In the specific implementation, based on the height coordinate value of the center point Perform a two-dimensional convolution operation with the preset convolution kernel to generate the X / Y direction gradient component and synthesize the gradient modulus value When the gradient modulus between adjacent cells continuously exceeds the preset height change threshold, it is determined that there is a physical height mutation in the area, and the marker coordinate set P and height difference parameters are output to complete the identification of geometric deformations such as snow accumulation or component misalignment.

[0090] Next, based on the grid topology framework established in step 1021, the temperature field distribution data collected by the infrared thermal imager is called. The sliding window statistical strategy is used to calculate the local and global temperature means, and the unit temperature anomaly intensity is quantified by the normalized deviation formula. The normalized deviation formula is as follows: When the deviation value exceeds the preset temperature deviation threshold, it is determined that there is a thermodynamic anomaly, and the marker coordinate set T and deviation parameters are output to accurately capture the local warping deformation of the panel caused by hot spot effects or uneven sunlight.

[0091] Then, in step 1023, a joint frequency-domain and time-domain analysis is performed on the airflow direction time series data collected by the ultrasonic anemometer within a unified grid space. The main frequency amplitude characteristics of the directional angle sequence are extracted through a fast Fourier transform, and the time-domain directional angle change rate is simultaneously calculated. When the main frequency amplitude and change rate jointly exceed a preset airflow change trend threshold, strong turbulence interference is determined, and a marker coordinate set F and turbulence intensity parameters are output, enabling dynamic perception of the cleaning stability risk area caused by gusts.

[0092] Finally, the three sets of coordinates P, T, and F are input into the spatial clustering fusion engine, and the DBSCAN algorithm is used to merge spatial overlapping units with a preset neighborhood radius. Obstacle type identifiers are assigned based on the dominant features. The height mutation dominant area is assigned identifier 1 to represent a physical obstacle, the temperature deviation dominant area is assigned identifier 2 to represent a thermal deformation area, and the airflow anomaly dominant area is assigned identifier 3 to represent an airflow interference area. The final associated height difference , temperature deviation , turbulence intensity Parameters such as the obstacle distribution map with multi-dimensional attributes are generated to provide a basis for cause classification and decision-making for the dynamic posture adjustment of the robot.

[0093] In actual application, in a certain midday operation scenario, strong wind and sand invasion formed local sand dunes in the northeast area of ​​the photovoltaic array, and the height difference of the B3 panel group reached 8.2mm. The surface temperature of the D7 panel group on the southwest side rose to 52°C due to the hot spot effect, exceeding the array average temperature of 35.3°C. At the same time, gusts of wind encountered in the northwest corner caused the airflow direction angle to fluctuate at a rate of 38° per second. The system calculated the grid gradient modulus in region B3 using the Sobel operator and detected that the height difference between adjacent cells, at 8.2 mm, exceeded the 5 mm threshold. It was marked as a physical obstacle and its coordinates were recorded. Subsequently, the temperature field data for cell D7 was analyzed, and the normalized deviation value (NDI) was calculated to be 4.77, exceeding the threshold of 2.5. This area was identified as a thermal deformation zone and its coordinates were annotated. Fast Fourier transform analysis of the airflow direction sequence in the northwest corner of the grid revealed a dominant frequency amplitude of 0.92 m / s and a directional angle change rate of 38° / s, both exceeding the thresholds of 0.8 m / s and 30° / s, marking it as an airflow interference zone. Finally, DBSCAN clustering was used to fuse the coordinates of the three regional types. Region B3 was assigned identifier 1, associated with a height difference of 8.2 mm; region D7 was assigned identifier 2, associated with a temperature of 52°C; and the wind disturbance zone was assigned identifier 3, associated with a turbulence intensity of 34.96 (° / s·m / s). This generated a multidimensional attribute map containing 12 obstacle cells. The composite obstacle overlapping the boundary between B3 and D7 was automatically identified as a mixed thermal-physical type.

[0094] The 102-point overall solution, through multi-physics field coupling analysis, enables accurate perception and cause classification of complex obstacles on photovoltaic panel surfaces. Its core value lies in the integration of geometric deformation gradients, thermodynamic deviations, and airflow dynamic characteristics to construct a multi-dimensional decision benchmark, eliminating the risk of misjudgment of physical protrusions and thermal deformation in single-sensor detection. An obstacle type identifier system based on dominant causes provides the robot with differentiated posture adjustment strategies, such as applying height compensation for physical obstacles and initiating temperature control protection for thermally deformed areas. At the same time, spatial clustering and fusion technology integrates discrete obstacle features into a unified topological map, supporting the subsequent path planning module to achieve millimeter-level terrain adaptation and zero-missing cleaning, completely resolving the perception mismatch problem of static environment modeling under dynamic conditions.

[0095] 103. Dynamically adjust the bending angle and extension range of the connecting components of the corresponding photovoltaic cleaning robot according to the gap width parameter and the attachment height parameter of the area cleaned by the photovoltaic cleaning robot in the obstacle distribution map, and update the posture information of the corresponding photovoltaic cleaning robot;

[0096] Optionally, step 103 may specifically include the following steps:

[0097] 1031. Extracting a gap width parameter and an attachment height parameter of the current cleaning area of ​​the photovoltaic cleaning robot from the obstacle distribution map, and calculating a maximum horizontal extension and a minimum vertical bending angle of the connecting component based on the gap width parameter and the attachment height parameter.

[0098] 1032. Generate a drive control instruction according to the maximum telescopic range and the minimum bending angle, and send the drive control instruction to a connecting component drive mechanism of a corresponding photovoltaic cleaning robot;

[0099] 1033. Adjust the actual extension and contraction range of the connecting component to the maximum extension and contraction range through the driving mechanism, and adjust the bending angle to the minimum bending angle.

[0100] 1034. Based on the adjusted telescopic amplitude and bending angle, combined with the real-time position coordinates of the corresponding photovoltaic cleaning robot in the position relationship data, its new moving direction coordinates and posture tilt angle are calculated to update the posture information of the corresponding photovoltaic cleaning robot.

[0101] In the above scheme, the gap width parameter refers to the actual physical distance measurement of the passable gap between adjacent photovoltaic panel units, which can be used to determine the lateral extension space required for the robot to cross the gap. The attachment height parameter refers to the characteristic value of the thickness of the deposits on the surface of the photovoltaic panel in the vertical direction, which can be used to plan the obstacle-crossing and lifting trajectory of the cleaning mechanism. The connecting component refers to the reconfigurable mechanical link between the robot body and the cleaning actuator, which includes a multi-stage telescopic link and a pitch and yaw two-degree-of-freedom joint, including real-time load stress distribution fed by strain gauges, which can be used to dynamically adapt to panel warping and gap changes. Posture information refers to the set of spatial state descriptions of the robot in the photovoltaic array, which can be used for posture constraint modeling for global path planning. The drive control instruction refers to the digital control signal of the motion parameters of the connecting component, which includes a motion parameter set consisting of the maximum telescopic amplitude and the minimum bending angle, and can be converted into a motor speed signal by the controller.

[0102] In the embodiment of the present application, the gap width parameter and the attachment height parameter of the current working area are first extracted by the rasterized obstacle map analysis algorithm, wherein the gap width parameter The height parameter of the attachment is obtained by calculating the spatial difference of the edge point set generated by the laser ranging data. The fusion temperature distribution and optical texture data are generated by the height field reconstruction algorithm. Then the parameters are input into the kinematic inverse model, and the maximum horizontal expansion and contraction amplitude is By the formula Solve, where is the mechanism shrinkage coefficient, For safety margin, the minimum bending angle in the vertical direction Then through Determine, among them is the reference clearance, For example, when sand accumulation between the plates causes the gap width parameter to be reduced to 9cm and the attachment height parameter reaches 2.5cm, the system calculates the maximum extension range. The minimum bending angle is 8.1 5.7°.

[0103] Next, in step 1032, the maximum telescopic range is set. With minimum bending angle This data frame is encapsulated as a CANopen-compliant data frame. The telescopic target value is written to register 0x6010 with 0.1mm accuracy, the bending angle is stored at address 0x6020 with 0.01° resolution, and the real-time execution flag 0x0001 is appended. This data frame is segmented into 32-bit packets using a time-division multiple access wireless transmission mechanism and sent to the target robot joint controller via the IEEE 802.15.4 protocol at preset time slots. A CRC-32 checksum is used during transmission to ensure data integrity, ensuring accurate delivery of commands within a 50ms delay. For example, when the command {telescopic target: 81mm, bending angle: 5.7°} is sent to robot #5, the controller verifies the command and immediately updates the motor control registers.

[0104] The telescopic mechanism's linear motor then performs closed-loop position control based on the maximum telescopic range. A PID algorithm dynamically adjusts the PWM duty cycle to achieve ±0.05mm positioning accuracy. Simultaneously, the pitch joint servo executes the minimum bend angle command. When the angle error exceeds 5°, bang-bang control is used for rapid approach to the target. When the error is less than 5°, an adaptive PID algorithm is used for fine tracking. An absolute encoder provides 0.01° real-time feedback. This process completes within 300ms, for example, precisely positioning the telescopic link to 81mm of travel while simultaneously rotating the joint to 5.7° and locking it.

[0105] Finally, through step 1034, based on the adjusted robot arm endpoint coordinates , through the homogeneous transformation matrix Fusion of RTK positioning data , solve the global position, attitude tilt angle According to the IMU quaternion and the board normal vector Calculate, where n is the surface normal vector and z is the unit vector in the direction of gravity. The updated pose information is written to the global state database at a frequency of 10Hz. For example, after fusing the new position of the robotic arm (205.7, 88.3, ​​41.6) with the RTK reference (203.2, 86.1, 38.5), the global pose {position: (204.5, 87.2, 40.1), tilt: 5.7°} is output, with a spatial consistency error within 2cm.

[0106] In actual applications, in the operation and maintenance scenario of a photovoltaic power station after a sandstorm, the #5 cleaning robot detected that the width of the gap between the panels had been reduced to 9cm due to sand accumulation. At the same time, optical texture analysis identified that the height of mud stains on the panel surface was 2.5cm. Based on this, the system calculated the maximum horizontal telescopic range to be 7.2cm, the contraction coefficient of the mechanism was 0.9, and the safety margin was 0.9cm. The minimum vertical bending angle was calculated to be 10.4°, the reference clearance was 3cm, and the arm length was 30cm. The drive command {telescopic target: 72mm, bending angle: 10.4°} was generated and encapsulated into a CANopen data frame for transmission via LoRa. After receiving the command, the #5 robot joint controller positioned the linear motor of the telescopic mechanism to 72mm stroke through PID closed-loop control within 300ms, and the synchronous pitch joint servo rotated to 10.4° and locked. Then, the new arm end coordinates were used. The global pose (233.17, 86.1, 39.38) is calculated by fusing the RTK base coordinates (203.2, 86.1, 38.5) and the IMU yaw angle of 5°. At the same time, the attitude tilt angle of 5.7° is calculated based on the plate surface normal vector (0.1, 0.0, 0.995). Finally, the millimeter-level adaptive fitting of the robotic arm to the deformed plate surface is achieved, eliminating the collision risk under sand accumulation conditions.

[0107] The overall solution of 103 mentioned above analyzes the gap width and attachment height parameters in the obstacle distribution map in real time, dynamically calculates the optimal expansion and contraction range and bending angle of the connecting components, generates precise drive instructions to control the reconstruction of the mechanism; and simultaneously integrates high-precision positioning data to update the robot's posture information, forming a closed-loop control chain that includes terrain perception, mechanism adjustment and state feedback. Its core value lies in enabling the cleaning robot to autonomously adapt to dynamic working conditions such as thermal deformation, gap fluctuations and sudden attachments of photovoltaic panels, completely eliminating the risk of mechanical structure collision; through millimeter-level fitting control, it ensures continuous and stable contact between the cleaning mechanism and the panel surface, avoiding the omission of key contaminated areas; the real-time update of posture information provides a reliable state benchmark for path planning, ultimately achieving zero-intervention continuous operation in complex terrain, greatly improving the robustness and full coverage reliability of the cleaning system.

[0108] 104. Generate an uncleaned area identifier and a dynamic coverage path of the cleaning robot based on the position relationship data, the obstacle distribution map, and the posture information of all photovoltaic cleaning robots;

[0109] Optionally, step 104 may specifically include the following steps:

[0110] 1041. Obtaining the boundary coordinates of the photovoltaic panel array and the position information of all photovoltaic cleaning robots in the position relationship data;

[0111] 1042. Divide the cleaning area into a uniform grid cell array according to the boundary coordinates of the photovoltaic panel array, and mark the inaccessible area in the grid cell array according to the obstacle cell coordinate positions and obstacle type identifiers in the obstacle distribution map;

[0112] 1043. Based on the historical position information of all photovoltaic cleaning robots, the coordinates of the grid cells traversed by the photovoltaic cleaning robots are backtracked to generate a cleaned area map. The cleaned area map is compared with the total grid cells of the photovoltaic panel array, and unmarked grid cells are defined as uncleaned grid cells. A unique area identifier is assigned to each uncleaned grid cell to generate an uncleaned area identifier.

[0113] 1044. Calculate an actual path width that the photovoltaic cleaning robot can pass through by combining the adjusted posture tilt angle of the connecting component and the gap width parameter in the obstacle distribution map;

[0114] 1045. Based on the uncleaned grid unit coordinate set in the uncleaned area identifier, taking the current position coordinates in the posture information as the starting point, avoiding the impassable area, connecting adjacent uncleaned grid units under the constraint of the actual path width, generating a continuous travel direction sequence and matching the travel speed value of each path segment to generate a dynamic coverage path.

[0115] In the above scheme, the impassable area refers to the physical space restriction unit marked in the obstacle distribution map, including areas where the height of attachments exceeds the standard and areas with insufficient structural gaps, which can be used to constrain the motion path planning of the cleaning robot to avoid mechanical damage. The uncleaned area identifier refers to the digital mark of the contaminated area generated based on historical posture backtracking, including the set of grid unit coordinates not covered by the robot trajectory and its unique encoding, which can be used for target positioning and priority determination of the dynamic task allocation system. The actual path width refers to the passability parameter of the robot mechanism in complex terrain, which can be used to constrain the physical feasibility of path generation. The continuous travel direction sequence refers to the optimized trajectory description quantity connecting the uncleaned grids, which can be used to drive the robot to perform high-precision coverage operations. The coverage path refers to the optimized travel trajectory that meets the terrain constraints, including a continuous direction angle sequence and a curvature adaptive velocity vector, which can be used to drive the robot to traverse the uncleaned area with high completion.

[0116] In this embodiment, the high-precision positioning system first acquires the coordinates of the PV array boundary vertices and the real-time poses of all robots in step 1041, establishing a global Cartesian coordinate system as a spatial reference. For example, a power plant accurately calibrates the array boundary X∈[0,200m], Y∈[0,150m] and tracks the robot R1 pose (35.2m, 82.7m, 3.5°) in real time.

[0117] Next, in step 1042, an adaptive quadtree grid decomposition algorithm is used based on the aforementioned coordinate system to divide the cleaning area into a 20cm×20cm grid cell array. Combined with the coordinate attribute mapping table in the obstacle distribution map, areas with excessive attachments, structural restrictions, and dynamic deformation are marked as impassable areas, generating a grid map matrix with attribute labels. For example, the grid at coordinates (82.3m, 45.1m) is marked as "BF" due to the detection of 4.2cm of bird droppings.

[0118] Next, in step 1043, the robot's historical pose time series database is called and the Bresenham line filling algorithm is used to discretize the continuous trajectory into a grid coverage sequence. A binary cleaned area map is generated by comparing the full array grid state matrix. Uncovered grids are assigned a unique identifier of "NS-row number-column number" to form an uncovered area identification set. For example, the grid at the uncovered coordinates (41.5m, 36.2m) of robot R2's trajectory is encoded as "NS-2075-1810" and added to the uncovered area set.

[0119] Then, the robot mechanism parameters are integrated with the real-time terrain data to read the pitch angle of the robot arm. , extract the obstacle gap width d, and use the envelope circle model formula Calculate the minimum passable width ,in is the width of the robot body, is the pitch angle of the manipulator arm. When the measured gap d is greater than or equal to the minimum passable width, a pass permission sign is generated. For example, the pitch angle of the manipulator arm at the coordinates (75.4m, 92.1m) It is 8.2°, the obstacle gap width d is 28 cm, the calculated minimum passable width is 62.3 cm, and the gap width of 65 cm is greater than the minimum passable width and is marked as passable.

[0120] Finally, in step 1045, the robot's current position is used as the starting point, and a constraint-based algorithm is used to search for a path in the uncleaned mark set, avoiding the marked grids and ensuring that the distance between the center of the continuous grids of the path is ≥ the minimum passable width. , and according to the path curvature radius Dynamically set the speed. When the curvature radius is ≥3m, the robot moves at the highest safe speed. When the curvature radius is <1m, the robot moves at the safety critical speed. When 1m<curvature radius<3m, the robot moves at the highest safe speed. , where 0.1 is the speed adjustment coefficient. The final output is a dynamic coverage path containing a sequence of path points, direction angles, and speed vectors. For example, the path planned to the "NS-2075-1810" grid is =1.8m curve section automatically reduces speed to 0.62m / s.

[0121] In actual applications, during the midday operation and maintenance of a 50MW photovoltaic power station, the center of the array was locally bulged due to the temperature difference caused by sunlight. The thermal deformation height at coordinate (40,25) was 4.3cm. At the same time, bird droppings suddenly occurred in the northeast corner, and the height of the attachment in the area between coordinates (5,5) and (5,7) was 5.1cm. The system first obtains the array boundary coordinates [(0,0), (100,0), (100,50), (0,50)] and the real-time pose of robot R1 (10,20,15°); then divides the cleaning area into a 20cm×20cm grid array, annotates (5,5)-(5,7) as "guano not passable" and (40,25) as "thermal deformation limit"; then traces back the historical trajectory of R1 covering the area (0,0)-(10,20) and generates the uncleaned marks {NS-5-5, NS-5-6, NS-5-7, NS-40-25}; then, based on the pitch angle of R1 robot arm 12° and the plate gap width of 28cm, the minimum passable width of the body is calculated to be 62cm; finally, with the current position of R1 (10,20) as the starting point, constraint A is used to The algorithm plans the path, avoids the (5,5)-(5,7) and (40,25) grids, connects adjacent uncleaned grids to generate a trajectory, and =2.1m curved section automatically reduces speed to 0.69m / s, and restores to 0.8m / s on straight section, achieving zero omission cleaning of polluted areas and zero mechanical collision throughout the process.

[0122] The 104-robot solution, described above, uses dynamic 3D terrain modeling and obstacle distribution mapping to provide real-time perception of thermal deformation and unexpected attachment distribution in photovoltaic panels. It adaptively adjusts the robot's joints based on gap width and attachment height parameters, achieving millimeter-level terrain conformity and collision-free motion. It combines the entire cluster's position information to accurately identify uncleaned areas, driving dynamic reconstruction of coverage paths to ensure zero missed cleaning of contaminated areas. By leveraging wireless communication networks to share global state, it autonomously allocates tasks to multiple robots based on the spatial matching of path extensions. In the event of a single robot failure, it automatically reconstructs the task map, forming a closed-loop control mechanism of environmental perception, dynamic decision-making, and coordinated execution. Ultimately, this system overcomes the adaptability limitations of static modeling systems and achieves efficient, full-coverage cleaning of photovoltaic arrays under complex working conditions.

[0123] 105. Share the uncleaned area identifier and the coverage path of each cleaning robot through the wireless communication network, and dynamically allocate the working areas and travel directions of multiple cleaning robots to achieve full coverage cleaning of the photovoltaic panel array.

[0124] Optionally, step 105 may specifically include the following steps:

[0125] 1051. Transmitting the uncleaned grid unit coordinate set in the uncleaned area identifier and the coverage path of each cleaning robot to the central control unit of all cleaning robots in real time via the wireless communication network;

[0126] 1052. In the central control unit, the uncleaned grid cells with a high matching degree are preferentially allocated to the corresponding cleaning robots based on the spatial distance of the path extension line, and the remaining uncleaned grid cells are secondary allocated to generate an area allocation mapping table.

[0127] Among them, step 1034 may specifically include the following processes: in the central control unit, the coordinates of the end point of the coverage path of each cleaning robot are parsed as the expected position, and for each uncleaned grid unit, the vertical distance from the uncleaned grid unit to the extension line of the current coverage path of each cleaning robot is calculated as a matching parameter, and the uncleaned grid unit whose matching parameter is less than a preset threshold is marked as a high-matching unit; all combinations of cleaning robots and the uncleaned grid units are traversed, and the high-matching units are assigned to the cleaning robots to which the corresponding coverage path extension lines belong and the unit allocation status is updated; for the remaining units whose allocation status is not updated, the straight-line distance from the expected position of the cleaning robot to the coordinates of the remaining units is calculated, and each remaining unit is assigned to the available cleaning robot with the smallest straight-line distance; the allocation results are integrated to generate an area allocation mapping table that clearly records the responsible cleaning robot for each uncleaned grid unit.

[0128] 1053. Generate a dedicated work area boundary for each robot based on the area allocation mapping table, and generate a path connection instruction including a steering angle correction value and a speed adjustment value based on the geometric relationship of the robot's current coverage path;

[0129] 1054. Send the boundary of the exclusive working area and the path connection instruction to the driving mechanism of the corresponding robot, drive the corresponding robot to complete the traversal of the uncleaned grid cells in the exclusive working area, and realize full coverage cleaning of the photovoltaic panel array through the feedback of the working status and the iterative update of the uncleaned area identification.

[0130] In the above scheme, the path extension line refers to a virtual trajectory segment linearly extrapolated from the end point of the cleaning robot's current coverage path along its motion direction. Essentially, it represents a spatial projection of the robot's motion trend and is used to predict the spatial correlation between uncleaned grid cells and the robot's natural trajectory. The matching parameter, defined as the scalar value of the perpendicular distance from the center point of the uncleaned grid cell to the path extension line, serves as the core decision-making basis for dynamic task allocation. The area allocation mapping table uses a hash data structure to store the mapping between uncleaned grid cell coordinates and the ID of the responsible cleaning robot, providing the basis for defining responsibility boundaries for multi-robot collaborative operations. The dedicated work area boundary is a convex hull polygonal boundary generated from the set of uncleaned grid cells assigned to a specific robot. Its geometric properties form a closed, continuous spatial domain, which is used to define the physical scope of a single robot's cleaning operation and avoid multi-robot path conflicts. The path connection instruction, a composite control instruction set consisting of steering angle correction values ​​and speed adjustment values, ensures smooth trajectory transitions during work area switching.

[0131] In this embodiment, a central control unit continuously receives the coverage path coordinate sequence and the uncleaned grid unit coordinate set uploaded by each cleaning robot via a wireless communication network. The coverage path is composed of time-series coordinate points generated by the robot's local SLAM module, and the end point coordinates represent the current motion trend. The uncleaned grid unit coordinate set is dynamically updated with a resolution of 0.1m×0.1m based on a gridded map of the photovoltaic array, laying the data foundation for dynamic task allocation.

[0132] Next, the coordinates of the end points of each robot's coverage path are analyzed in step 1502, and the equation of the path extension line is extrapolated along the direction of motion. For each uncleaned grid cell , calculate the vertical distance to the extension line , if the vertical distance Cells with distances less than a preset threshold of 0.5m are marked as highly matched. The system traverses all robot-grid combinations, assigning highly matched cells to the corresponding extended line robots and updating their status indicators. For remaining unassigned cells, the Euclidean distance to each robot's endpoint is calculated and assigned to the available robot with the shortest distance. Finally, the allocation results are integrated to create a region allocation map, using a hash structure to clearly define the responsibility for each grid cell.

[0133] Then, in step 1053, the responsibility grid set of each robot in the area allocation mapping table is extracted, and the convex hull vertex is calculated using the Graham scanning algorithm to generate the exclusive work area polygon boundary. Then combine the robot's current path end point Tangential angle With exclusive area entry point The exclusive area entry point is the nearest vertex on the boundary, and the heading angle correction value is calculated using the following formula: . At the same time, based on the distance from the entry point Generate exponential speed curve Final packaging 、 and boundary coordinates It is a path connection instruction set.

[0134] Finally, the central unit transmits the designated area boundaries and connection instructions to the corresponding robot drive controller via the wireless network. The robot first adjusts its steering angle and speed according to the connection instructions to enter the designated area, then traverses the assigned grid using a modified serpentine trajectory. After each grid is cleaned, a status flag is transmitted back to the central unit, triggering a dynamic update of the uncleared flag and initiating a new round of allocation until the entire PV array is covered.

[0135] In practical applications, during the cleaning operation of a 200m×100m photovoltaic array, three robots initially ran along a preset path. A sudden strong wind caused the accumulation of dust in the area B with coordinates [50-70, 30-50] to form an uncleaned grid. At the same time, Robot2 stopped due to battery exhaustion. The central control unit received the coverage paths of Robot1 and Robot3 and the set of uncleaned grid coordinates including area B through the wireless network. The direction angle of Robot1's end point (48, 32) was 30°, and the direction angle of Robot3's end point (80, 40) was 120°. By calculating the perpendicular distance from each grid to the path extension line, for example The distance from grid (52,33) to the extension line of Robot1 is 0.3m < 0.5m threshold, marked as a high-matching unit, and the grids [50-60,30-40] in area B are assigned to Robot1; the remaining grids [60-70,30-50] are assigned to Robot3 according to the Euclidean distance, for example, the distance from grid (65,45) to the end point of Robot3 is 18.0m < 20.5m to Robot1; the convex hull boundary of the exclusive area of ​​Robot1 {(50,30), (60,30), (60,40), (50,40)} is generated and the steering correction angle of the entry point (50,35) is calculated =15° and speed adjustment curve , and generated corresponding instructions for Robot3 at the same time; the two robots smoothly drove into the exclusive area according to the instructions, completed the grid traversal in a serpentine path and sent back the status in real time. After the central unit updated the uncleaned mark, the full-area coverage rate reached 100%.

[0136] The overall solution described above, 105, significantly reduces the frequency of robot steering idle turns and energy loss through a mechanism that prioritizes the spatial relevance of path extension lines. Convex hull optimization modeling of the boundaries of dedicated work areas, combined with dual-parameter connection instructions for steering angle correction and speed adjustment values, enables smooth trajectory transitions and conflict-free motion coordination for multiple robots to new work areas. Real-time feedback of work status and dynamic, iterative updates of uncleared markers form a closed-loop control architecture. In disturbance scenarios such as sudden obstacle attachment and single-machine failure shutdowns, task allocation map reconstruction and real-time path optimization are autonomously triggered to ensure the system's strong anti-interference capabilities. Ultimately, a collaborative control paradigm of global dynamic perception, distributed efficient execution, and closed-loop autonomous evolution is formed, achieving the core goals of high robustness, high efficiency, and zero omissions in full coverage cleaning of photovoltaic arrays in complex dynamic environments.

[0137] The following is a complete example for steps 101 to 105. Figure 2As shown in the figure, during the summer operation and maintenance of a 100MW photovoltaic power station, three cleaning robots (R1, R2, and R3) started their work. The onboard 16-line laser radar scanned the photovoltaic array at a frequency of 10Hz, acquiring laser ranging point cloud data with millimeter-level precision. At the same time, the infrared thermal imager captured the panel surface temperature distribution at a sampling rate of 5Hz. When the central control unit detected that area A had formed a high temperature area of ​​60℃ due to the hot spot effect (i.e., Figure 2 The high temperature area A), the adjacent area B is only 35℃ (i.e. Figure 2 The ultrasonic anemometer collects real-time ambient airflow data, detecting a prevailing northwest wind of 3.5 m / s. Combined with the UWB positioning module's output of the robot's positional relationship with the photovoltaic panel, the central control unit integrates the four-dimensional data stream and uses the TSDF algorithm to construct a dynamic three-dimensional terrain model, reflecting panel surface deformation in real time. A 4.2 cm localized bulge was detected in the high-temperature area A, while a 1.8 mm gap widening occurred in the low-temperature area B due to thermal stress.

[0138] Then, based on the 3D terrain model, three key obstacle parameters were extracted: component gap width, attachment height, and deformation slope. It was found that the gap in the low-temperature area B increased from the standard 20mm to 21.8mm, and in area C, bird droppings were detected to form a 12mm bulge (i.e. Figure 2 The guano area (C) in the figure is shown, and the edge of the high-temperature area (A) forms an 18° inclination. A 5 cm³ obstacle distribution map is generated through voxel rasterization. Red highlights indicate guano deposits, blue grids indicate areas with abnormal gap widths, and yellow gradients indicate deformation gradients. The map is stored in matrix form, with each grid containing a triplet of attributes: gap width, deposit height, and deformation index.

[0139] Secondly, when robot R1 enters the guano area C, the system analyzes the obstacle map and finds that the height parameter of the attachment in the corresponding grid is 12mm. It then dynamically adjusts its six-degree-of-freedom manipulator, increases the pitch angle of the bending joint K3 by 25° so that the roller brush fits tightly against the raised surface, and extends the telescopic cylinder S2 by 8mm to increase the downforce. The posture information is updated synchronously, and the coordinate system of the manipulator end is changed from Adjust to When R2 reaches high-temperature area A, based on the 4.2cm rise data, it actively retracts support leg L1 by 3cm to prevent chassis scraping and increases the drive wheel torque distribution ratio to 65% to overcome the slope resistance. Posture updates are broadcast to the central unit in real time, forming a baseline for the cluster's motion status.

[0140] Then, the central unit fuses the position relationship data, obstacle map and the whole robot posture, and identifies the uncleaned areas through the spatiotemporal conflict detection algorithm. The dust area D is not included in the original path due to the strong wind and the new dust adhesion; the area E forms a 3×4m blind area due to the mechanical failure of R3 (i.e. Figure 2Dynamic coverage paths are generated based on the improved wavefront method: an S-shaped trajectory is planned for R1 to bypass the guano concentration area C, and a Z-shaped trajectory is generated for R2 to climb the raised area A. Dust area D and fault blind area E are marked as high-priority uncleaned areas.

[0141] Finally, the identification of the uncleaned area and the path data of each robot are shared through the dedicated network, including the coordinates of the dust area D [120-125, 45-50] and the fault blind area E [80-83, 30-33]. The central unit then performs dynamic allocation, assigning the dust area D to R1, which is closest to the path extension line, and the fault blind area E to R2 based on the Euclidean distance. The dedicated work area boundary and connection instructions are generated, and the steering angle of R1 is corrected. =-12°, speed curve ; R2 gets the steering angle =8°. Following the command, the robot smoothly enters the new area. R1 uses a spiral path to clean area D, and R2 uses a reciprocating path to cover area E. Status feedback triggers iterative updates of flags, ultimately achieving full coverage.

[0142] Figure 3 The present invention provides a schematic diagram of a modular control system for full-coverage cleaning of a photovoltaic cleaning robot. Figure 3 As shown, the system includes:

[0143] An acquisition module 31 is used to acquire laser ranging data, temperature distribution data on the surface of the photovoltaic panels, ambient airflow data, and positional relationship data between the photovoltaic panels, and to construct a three-dimensional terrain model of the photovoltaic panel surface;

[0144] A generating module 32 is configured to generate an obstacle distribution map corresponding to the photovoltaic panel array based on the three-dimensional terrain model;

[0145] A control module 33 is configured to dynamically adjust the bending angle and extension range of the connecting components of the corresponding photovoltaic cleaning robot according to the gap width parameter and the attachment height parameter of the area cleaned by the photovoltaic cleaning robot in the obstacle distribution map, and update the posture information of the corresponding photovoltaic cleaning robot;

[0146] A planning module 34 is configured to generate a dynamic coverage path and an uncleaned area identifier of the cleaning robot based on the position relationship data, the obstacle distribution map, and the posture information of all photovoltaic cleaning robots;

[0147] The coordination module 35 is used to share the uncleaned area identification and the real-time posture information of each cleaning robot through the wireless communication network, and dynamically allocate the working areas and travel directions of multiple cleaning robots to achieve full coverage cleaning of the photovoltaic panel array.

[0148] Figure 3The photovoltaic cleaning robot's full coverage cleaning modular control system can be executed Figure 1 The implementation principle and technical effects of the modular control method for full-coverage cleaning of a photovoltaic cleaning robot described in the illustrated embodiment will not be elaborated on here. The specific manner in which each module and unit performs operations in the modular control system for full-coverage cleaning of a photovoltaic cleaning robot in the above embodiment has been described in detail in the embodiments of the method and will not be elaborated on here.

[0149] In one possible design, Figure 3 The modular control system for full coverage cleaning of a photovoltaic cleaning robot of the embodiment shown can be implemented as a computing device, such as Figure 4 As shown, the computing device may include a storage component 41 and a processing component 42;

[0150] The storage component 41 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 42 .

[0151] The processing component 42 is used for the above Figure 1 The embodiment provides a modular control method for full-coverage cleaning of a photovoltaic cleaning robot.

[0152] The processing component 42 may include one or more processors to execute computer instructions to perform all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0153] The storage component 41 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0154] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0155] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0156] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0157] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0158] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The illustrated embodiment provides a modular control method for full-coverage cleaning of a photovoltaic cleaning robot.

[0159] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0160] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0161] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0162] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A modular control method for full-coverage cleaning of a photovoltaic cleaning robot, characterized in that: include: Obtain laser ranging data, temperature distribution data on the photovoltaic panel surface, ambient airflow data, and positional relationship data between the photovoltaic panels to construct a three-dimensional terrain model of the photovoltaic panel surface; generating an obstacle distribution map corresponding to the photovoltaic panel array based on the three-dimensional terrain model; According to the gap width parameters and attachment height parameters of the area cleaned by the photovoltaic cleaning robot in the obstacle distribution map, the bending angle and extension range of the connecting components in the corresponding photovoltaic cleaning robot are dynamically adjusted, and the posture information of the corresponding photovoltaic cleaning robot is updated; Based on the position relationship data, the obstacle distribution map and the posture information of all photovoltaic cleaning robots, generating the uncleaned area identification and dynamic coverage path of the cleaning robot; Sharing the uncleaned area identifier and the coverage path of each cleaning robot through a wireless communication network, dynamically allocating the working areas and travel directions of multiple cleaning robots to achieve full coverage cleaning of the photovoltaic panel array; The method of sharing the uncleaned area identifier and the coverage path of each cleaning robot through a wireless communication network and dynamically allocating the working areas and travel directions of multiple cleaning robots to achieve full coverage cleaning of the photovoltaic panel array includes: The coordinate set of the uncleaned grid cells in the uncleaned area identification and the coverage path of each cleaning robot are transmitted to the central control unit of all cleaning robots in real time through the wireless communication network; In the central control unit, the uncleaned grid cells with high matching degree are preferentially allocated to the corresponding cleaning robots according to the spatial distance of the path extension line, and the remaining uncleaned grid cells are secondary allocated to generate an area allocation mapping table; Generate a dedicated work area boundary for each robot based on the area allocation mapping table, and generate a path connection instruction including a steering angle correction value and a speed adjustment value based on the geometric relationship of the robot's current coverage path; The boundary of the exclusive working area and the path connection instruction are sent to the driving mechanism of the corresponding robot, driving the corresponding robot to complete the traversal of the uncleaned grid cells in the exclusive working area, and realizing full coverage cleaning of the photovoltaic panel array through the feedback of the working status and the iterative update of the uncleaned area identification.

2. The method according to claim 1, characterized in that In the central control unit, the uncleaned grid cells with high matching degree are preferentially allocated to the corresponding cleaning robots according to the spatial distance of the path extension line, and secondary allocation is performed on the remaining uncleaned grid cells to generate an area allocation mapping table, including: In the central control unit, the coordinates of the end points of the coverage paths of the cleaning robots are parsed as the expected positions. For each uncleaned grid cell, the vertical distance from the uncleaned grid cell to the extension line of the current coverage path of each cleaning robot is calculated as a matching parameter. The uncleaned grid cells whose matching parameters are less than a preset threshold are marked as high-matching cells. Traverse all combinations of cleaning robots and the uncleaned grid cells, assign the cells with high matching degrees to the cleaning robots corresponding to the extended coverage paths, and update the cell assignment status; For the remaining cells whose allocation status has not been updated, the straight-line distance from the expected position of the cleaning robot to the coordinates of the remaining cells is calculated, and each remaining cell is allocated to the available cleaning robot with the smallest straight-line distance; The allocation results are integrated to generate an area allocation mapping table that clearly records the responsible cleaning robot for each uncleaned grid cell.

3. The method according to claim 1, characterized in that The generating of the uncleaned area identification and the dynamic coverage path of the cleaning robot based on the position relationship data, the obstacle distribution map and the posture information of all photovoltaic cleaning robots includes: Obtain the boundary coordinates of the photovoltaic panel array and the position information of all photovoltaic cleaning robots in the position relationship data; Divide the cleaning area into a uniform grid cell array according to the boundary coordinates of the photovoltaic panel array, and mark the inaccessible area in the grid cell array according to the obstacle cell coordinate position and obstacle type identifier in the obstacle distribution map; Based on the historical posture information of all photovoltaic cleaning robots, the coordinates of the grid cells traversed by the photovoltaic cleaning robots are backtracked to generate a cleaned area map, the cleaned area map is compared with the total grid cells of the photovoltaic panel array, the unmarked grid cells are defined as uncleaned grid cells, and a unique area identifier is assigned to each uncleaned grid cell to generate an uncleaned area identifier; Calculating the actual path width that the photovoltaic cleaning robot can pass through by combining the adjusted posture inclination angle of the connecting component and the gap width parameter in the obstacle distribution map; Based on the uncleaned area identifier, the set of uncleaned grid unit coordinates is identified, and the current position coordinates in the posture information are used as the starting point. The inaccessible area is avoided, and adjacent uncleaned grid units are connected under the constraint of the actual path width. A continuous moving direction sequence and a moving speed value matching each path segment are generated to generate a dynamic coverage path.

4. The method according to claim 1, wherein Generating an obstacle distribution map corresponding to the photovoltaic panel array based on the three-dimensional terrain model includes: Analyzing the three-dimensional grid structure in the three-dimensional terrain model, extracting the height coordinate value of each grid cell, calculating the height change difference of the height coordinate values ​​between adjacent grid cells, and identifying grid cells whose height change difference exceeds a preset height change threshold as potential obstacle areas; Analyze the temperature value of each grid cell, calculate the degree of deviation between the temperature value and the average temperature value, and define the grid cells with the degree of deviation greater than a preset temperature deviation threshold as temperature abnormal areas; Analyze the airflow direction attributes of each grid cell, extract the change trend of the airflow direction value, and identify the grid cells with abnormal change trends as airflow affected areas based on a preset airflow change trend threshold; The potential obstacle area, the temperature anomaly area and the airflow affected area are combined and marked as obstacle units, and an obstacle distribution map is generated on the photovoltaic panel array based on the coordinate position of the obstacle unit, combined with the obstacle type identifier and the obstacle height value.

5. The method according to claim 1, wherein The method dynamically adjusts the bending angle and extension range of the connecting components of the photovoltaic cleaning robot according to the gap width parameter and the attachment height parameter of the area cleaned by the photovoltaic cleaning robot in the obstacle distribution map, and updates the posture information of the photovoltaic cleaning robot, including: Extracting gap width parameters and attachment height parameters of the current cleaning area of ​​the photovoltaic cleaning robot from the obstacle distribution map, and calculating the maximum horizontal expansion and contraction amplitude and the minimum vertical bending angle of the connecting component based on the gap width parameters and the attachment height parameters respectively; generating a drive control instruction according to the maximum telescopic amplitude and the minimum bending angle, and sending the drive control instruction to a connecting component drive mechanism of a corresponding photovoltaic cleaning robot; The actual extension and contraction range of the connecting component is adjusted to the maximum extension and contraction range by the driving mechanism, and the bending angle is adjusted to the minimum bending angle at the same time; Based on the adjusted telescopic amplitude and bending angle, combined with the real-time position coordinates of the corresponding photovoltaic cleaning robot in the position relationship data, its new moving direction coordinates and posture tilt angle are calculated to update the posture information of the corresponding photovoltaic cleaning robot.

6. The method according to claim 1, wherein The method of acquiring laser ranging data, temperature distribution data on the surface of the photovoltaic panels, ambient airflow data, and positional relationship data between the photovoltaic panels to construct a three-dimensional terrain model of the photovoltaic panel surface includes: The laser distance measuring device installed on the photovoltaic cleaning robot body scans the surface of the photovoltaic panel and continuously obtains the vertical distance between the robot chassis and the photovoltaic panel surface at multiple measuring points as laser distance measurement data; The temperature sensor installed on the photovoltaic cleaning robot body synchronously obtains the temperature value of each point on the photovoltaic panel surface to generate temperature distribution data; The airflow sensor installed on the photovoltaic cleaning robot body obtains the direction and speed of the ambient airflow near the photovoltaic panel surface to generate ambient airflow data; The positional relationship data between the photovoltaic cleaning robot body and the photovoltaic panel is obtained by a position positioning device installed on the photovoltaic cleaning robot body; The laser ranging data is converted into height coordinates in three-dimensional space to form a basic spatial shape of the photovoltaic panel surface. The temperature distribution data and the ambient airflow data are associated as additional attributes to the corresponding three-dimensional spatial positions in the basic spatial shape. Based on the position relationship data and the basic spatial shape, a three-dimensional terrain model of the photovoltaic panel surface is constructed.

7. A full coverage cleaning modular control system for a photovoltaic cleaning robot, characterized in that: include: An acquisition module is used to obtain laser ranging data, temperature distribution data on the photovoltaic panel surface, ambient airflow data, and positional relationship data between the photovoltaic panels to construct a three-dimensional terrain model of the photovoltaic panel surface; A generation module, configured to generate an obstacle distribution map corresponding to the photovoltaic panel array based on the three-dimensional terrain model; A control module, configured to dynamically adjust the bending angle and extension range of the connecting components of the corresponding photovoltaic cleaning robot according to the gap width parameter and the attachment height parameter of the area cleaned by the photovoltaic cleaning robot in the obstacle distribution map, and update the posture information of the corresponding photovoltaic cleaning robot; A planning module, configured to generate a dynamic coverage path and an uncleaned area identifier for the cleaning robot based on the position relationship data, the obstacle distribution map, and the posture information of all photovoltaic cleaning robots; A coordination module is used to share the uncleaned area identification and the real-time posture information of each cleaning robot through a wireless communication network, and dynamically allocate the working areas and travel directions of multiple cleaning robots to achieve full coverage cleaning of the photovoltaic panel array; The method of sharing the uncleaned area identifier and the coverage path of each cleaning robot through a wireless communication network and dynamically allocating the working areas and travel directions of multiple cleaning robots to achieve full coverage cleaning of the photovoltaic panel array includes: The coordinate set of the uncleaned grid cells in the uncleaned area identification and the coverage path of each cleaning robot are transmitted to the central control unit of all cleaning robots in real time through the wireless communication network; In the central control unit, the uncleaned grid cells with high matching degree are preferentially allocated to the corresponding cleaning robots according to the spatial distance of the path extension line, and the remaining uncleaned grid cells are secondary allocated to generate an area allocation mapping table; Generate a dedicated work area boundary for each robot based on the area allocation mapping table, and generate a path connection instruction including a steering angle correction value and a speed adjustment value based on the geometric relationship of the robot's current coverage path; The boundary of the exclusive working area and the path connection instruction are sent to the driving mechanism of the corresponding robot, driving the corresponding robot to complete the traversal of the uncleaned grid cells in the exclusive working area, and realizing full coverage cleaning of the photovoltaic panel array through the feedback of the working status and the iterative update of the uncleaned area identification.

8. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a full-coverage cleaning modular control method for a photovoltaic cleaning robot as described in any one of claims 1 to 6.

9. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, a full-coverage cleaning modular control method of a photovoltaic cleaning robot as described in any one of claims 1 to 6 is implemented.

Images