A multi-robot cooperative cleaning scheduling method and system suitable for a photovoltaic power station

Abstract

This invention relates to the field of robot collaborative management technology, specifically disclosing a multi-robot collaborative cleaning scheduling method and system applicable to photovoltaic power plants. By collecting real-time operating status data and environmental flow field vector data of robots throughout the photovoltaic power plant, the system accurately extracts the operation jet vector from the real-time operating status data. This vector vector is then professionally superimposed with the environmental flow field vector data to construct a diffusion elliptical domain model data with the coordinates of each robot as the focus. This effectively solves the drawback of traditional models that ignore the coupling effect between the environmental flow field and the robot's self-generated flow field, making the prediction of the diffusion range more consistent with the actual operating environment of outdoor photovoltaic power plants. This provides reliable data support for subsequent multi-robot interference prediction, reduces the risk of secondary pollution caused by flow field prediction deviations from the source, and lays a solid foundation for the refined control of multi-robot collaborative operations, ensuring the collaborative consistency of robot operations throughout the entire area.

Description

Technical Field

[0001] This invention belongs to the field of robot collaborative management technology, and relates to a multi-robot collaborative cleaning scheduling method and system suitable for photovoltaic power plants. Background Technology

[0002] As a core component of green energy infrastructure, photovoltaic power plants directly impact the scale and economic benefits of clean energy production through their power generation efficiency. The surface cleanliness of photovoltaic modules is a key factor affecting photoelectric conversion efficiency; dust, snow, and other contaminants can reduce power generation efficiency by up to 15%. Meanwhile, robotic automated cleaning technology, as a modern means to improve operational efficiency, plays an irreplaceable role in the stable operation of large-scale photovoltaic power plants due to its precise path planning and intelligent collaborative operations. This is especially true for ground-mounted power plants with vast areas and complex environments, where the continuity of cleaning operations directly affects the overall profitability of the power plant.

[0003] It is worth noting that photovoltaic power plants are often deployed in open environments with strong winds and significant temperature differences, making their environmental dynamics far more complex than those of indoor or structured environments. Continuous wind disturbances and microclimate differences between module arrays not only accelerate pollutant accumulation but can also interfere with the normal operation of robot sensors through mechanisms such as dust and moisture condensation. This makes environmental adaptability particularly important in the system design of photovoltaic cleaning robots. Especially for large power plants that have already adopted multi-robot collaborative operations, there is a deep coupling relationship between robot trajectory planning, obstacle avoidance logic, and power generation efficiency.

[0004] However, the collaborative management of current photovoltaic cleaning robots still has significant shortcomings. First, traditional control strategies rely excessively on static obstacle avoidance rules, neglecting the influence of environmental fields. Field practice shows that even when robots are equipped with high-precision positioning systems and travel along preset paths, wind interference can still cause cleaning media to drift to adjacent components, resulting in "secondary pollution" or electrical losses. This reflects a design flaw in focusing only on the robot's own movement while ignoring environmental effects such as fluid fields. While existing interrupt-driven collaborative logic ensures basic safety, it fails to reflect the optimality of task scheduling in dynamic environments and lacks a systematic trade-off between energy efficiency and power generation losses during multi-machine interaction. As a result, maintenance strategies are mostly reactive, unable to minimize energy losses through predictive scheduling before efficiency declines, leading to a lag in optimizing the economics of power plant maintenance. Summary of the Invention

[0005] In view of the problems existing in the prior art, the present invention provides a multi-robot collaborative cleaning scheduling method and system applicable to photovoltaic power plants to solve the above-mentioned technical problems.

[0006] To achieve the above and other objectives, the technical solution adopted by the present invention is as follows:

[0007] The first aspect of this invention provides a multi-robot collaborative cleaning scheduling method suitable for photovoltaic power plants, the method comprising:

[0008] Real-time operating status data and environmental flow field vector data of the full-domain robot of the photovoltaic power station are collected. The operation jet vector in the real-time operating status data and the environmental flow field vector data are extracted and vector superimposed to construct a diffusion elliptical domain model data with the coordinates of each robot as the focus.

[0009] Based on the travel speed vector in the real-time running status data, the spatiotemporal trajectory of multiple robots is extrapolated, the expected meeting spatiotemporal coordinate set is calculated, and the diffusion elliptical domain model data is mapped to the expected meeting spatiotemporal coordinate set. The interference coupling coefficient is generated by calculating the overlapping area of ​​the elliptical domains of the robots moving towards each other.

[0010] The interference coupling coefficient is compared with a preset safety threshold. When the interference coupling coefficient exceeds the preset safety threshold, the safety node coordinate data in the pre-stored photovoltaic module layout map is retrieved, and the drift speed variable of the expected encounter time-space coordinate set is calculated to be translated to the encounter point corresponding to the safety node coordinate data.

[0011] Determine whether the drift speed variable at the meeting point is within the preset allowable speed change range. If so, output a differential drift control command containing the drift speed variable at the meeting point to the associated robot.

[0012] If the drift speed variable at the meeting point exceeds the preset allowable speed range, the diffusion elliptical domain model data is solved in reverse based on the preset safety threshold to calculate the extreme operation intensity value that satisfies the non-interference condition and generate a frequency conversion adjustment command. At the same time, virtual uncleaned area marker data is generated based on the extreme operation intensity value and written into the global task scheduling queue for subsequent path planning calls.

[0013] Another aspect of the present invention provides a multi-robot collaborative cleaning scheduling system suitable for photovoltaic power plants, the system comprising:

[0014] Diffusion model construction module: Collect real-time operating status data and environmental flow field vector data of the robot in the whole field of photovoltaic power station, extract the operation jet vector from the real-time operating status data and perform vector superposition operation with the environmental flow field vector data to construct diffusion elliptical domain model data with the coordinates of each robot as the focus;

[0015] Interference coefficient calculation module: Based on the travel speed vector in the real-time running status data, the module performs spatiotemporal trajectory extrapolation of multiple robots, calculates the expected encounter spatiotemporal coordinate set, maps the diffusion elliptical domain model data to the expected encounter spatiotemporal coordinate set, and generates the interference coupling coefficient by calculating the overlapping area of ​​the elliptical domains of the robots moving towards each other.

[0016] Drift speed generation module: compares the interference coupling coefficient with the preset safety threshold. When the interference coupling coefficient exceeds the preset safety threshold, it retrieves the safety node coordinate data in the pre-stored photovoltaic module layout map and calculates the drift speed variable of the meeting point corresponding to the expected encounter time-space coordinate set translated to the safety node coordinate data.

[0017] Drift speed change judgment module: Determines whether the drift speed variable at the meeting point is within the preset allowable speed change range. If so, it outputs a differential drift control command containing the drift speed variable at the meeting point to the associated robot. If the drift speed variable at the meeting point exceeds the preset allowable speed change range, it performs inverse solution on the diffusion elliptical domain model data based on the preset safety threshold, calculates the extreme operation intensity value that satisfies the non-interference condition, and generates a frequency conversion adjustment command. At the same time, it generates virtual uncleaned area marker data based on the extreme operation intensity value and writes the virtual uncleaned area marker data into the global task scheduling queue for subsequent path planning calls.

[0018] As described above, the multi-robot collaborative cleaning scheduling method and system for photovoltaic power plants provided by the present invention has at least the following beneficial effects:

[0019] 1. This invention collects real-time operating status data and environmental flow field vector data of the full-domain robot in a photovoltaic power station, accurately extracts the operation jet vector from the real-time operating status data, and performs professional vector superposition calculation with the environmental flow field vector data to construct a diffusion elliptical domain model data with the coordinates of each robot as the focus. This breaks through the limitation of traditional control schemes that rely solely on fixed physical distances to judge interference, and realizes dynamic and accurate modeling of the diffusion range of the clean medium. It effectively solves the drawback of traditional models that ignore the coupling effect between the environmental flow field and the robot's self-generated flow field, making the prediction of the diffusion range more consistent with the actual operating environment of outdoor photovoltaic power stations. This not only improves the scientificity and accuracy of flow field modeling, but also provides reliable data support for subsequent multi-machine interference prediction, reducing the risk of secondary pollution caused by flow field prediction deviations from the source. At the same time, it lays a solid foundation for the refined control of multi-machine collaborative operations and ensures the collaborative consistency of the full-domain robot operation.

[0020] 2. This invention, based on the travel velocity vector in real-time operating status data, performs comprehensive spatiotemporal trajectory extrapolation for multiple robots, accurately calculates the expected meeting spatiotemporal coordinate set, and then precisely maps the constructed diffusion elliptical domain model data to the expected meeting spatiotemporal coordinate set. By scientifically calculating the overlapping area of ​​the elliptical domains of the robots moving towards each other, an interference coupling coefficient is generated, realizing a quantitative assessment of interference risks in multi-robot collaborative operations. This effectively avoids the shortcomings of traditional methods, which can only qualitatively judge interference but cannot quantify the degree of interference, making interference judgment more objective and accurate. It can predict potential interference risks when multiple robots meet in advance, providing a precise quantitative basis for the formulation of subsequent control strategies. It avoids the decline in work efficiency or the aggravation of secondary pollution caused by blind control, further improving the orderliness and controllability of multi-robot collaborative operations and ensuring the stability of the clean quality of photovoltaic modules.

[0021] 3. This invention accurately compares the interference coupling coefficient with a preset safety threshold. When the interference coupling coefficient exceeds the preset safety threshold, it promptly retrieves the safety node coordinate data from the pre-stored photovoltaic module layout map, calculates the drift speed variable of the meeting point corresponding to the expected encounter time-space coordinate set being translated to the safety node coordinate data, and flexibly outputs different control commands by judging whether the drift speed variable of the meeting point is within the preset allowable speed range. At the same time, it marks the uncleaned areas and includes them in subsequent scheduling, realizing hierarchical and precise control of multi-machine interference. It prioritizes interference-free operation through differential drift to ensure work efficiency. When drift is not possible, it avoids interference and reserves space for re-cleaning through frequency conversion adjustment. This effectively reduces energy consumption, avoids secondary pollution, and ensures the comprehensiveness of cleaning. It also ensures the scientific nature and adaptability of control commands, significantly improving the operation and maintenance efficiency, cleaning quality, and operational reliability of the photovoltaic cleaning robot system, and adapting to the refined operation and maintenance needs of large-scale photovoltaic power plants. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram showing the connections between the steps of the method of the present invention.

[0024] Figure 2 This is a schematic diagram showing the connections of the various modules in the system of the present invention. Detailed Implementation

[0025] The following description, in conjunction with the implementation of this invention, is merely an example and illustration of the concept of this invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the inventive concept or exceed the scope defined in these claims, all of which should fall within the protection scope of this invention.

[0026] In traditional photovoltaic power plant cleaning robot collaborative systems, fixed obstacle avoidance rules and static safety thresholds cannot adapt to changing environmental flow fields and dynamic operational intensities. When multiple robots move towards each other under the influence of wind, the diffusion range of their cleaning media (such as water mist and airflow) changes nonlinearly. The system cannot establish a dynamic correlation between the spatiotemporal relationships between robots and potential pollution risks, leading to a mismatch between trajectory planning logic and real-time interference situation. This static geometric collision avoidance mechanism reduces the continuity of collaborative operations and global energy efficiency, making the status information received by the scheduling system unable to reflect the true risk of "secondary pollution," ultimately affecting the overall quality of cleaning operations and power generation efficiency.

[0027] For example, in a typical sandstorm cleaning scenario, the ambient wind speed increases from an initial 2 m / s to 5 m / s, and the angle between the wind direction and the spray direction of robot A is less than 30 degrees, causing the major axis of its water mist diffusion ellipse to expand by 40%. In this situation, the traditional system still uses a preset fixed safety distance for conflict determination, causing robot B to enter A's diffusion range along its original trajectory. The control commands output by the multi-machine collaborative model are merely simple avoidance or waiting, the cleaning workflow is unexpectedly interrupted, and the dispersed water mist has already caused localized contamination on the surface of the photovoltaic modules it passes through.

[0028] If the above problems are not addressed, misjudgment and failure to avoid dynamic disturbance risks will lead to a misalignment between clean-up plans and actual environmental conditions, exacerbating the waste of water resources and energy. A rigid conflict resolution mechanism will hinder the system from capturing the balance between clean-up efficiency and power generation losses, delaying scheduling adjustments aimed at optimal energy efficiency. The asynchronous processing of environmental vector data and robot states will also cause phase deviations in disturbance field prediction and trajectory extrapolation, reducing the reliability of risk warning and collaborative control, ultimately creating a negative feedback loop that impacts the overall economic efficiency of power plant operation and maintenance.

[0029] Faced with the aforementioned problems, this invention first considers how to establish a dynamic correlation mechanism between flow field parameters and robot operating status. Traditional systems use fixed physical distances to judge interference and a single control mode to deal with complex scenarios, resulting in flow field modeling deviations, inaccurate interference prediction, and an inability to adapt to dynamic operating environments. To address this, this invention attempts to couple environmental flow field vector data with robot operating status data, constructing a dynamic diffusion elliptical domain model through vector superposition to achieve accurate prediction of interference range. Further analysis reveals that relying solely on single data or static models is insufficient to capture the nonlinear fluctuations of the environmental flow field and the dynamic interference of multi-robot operations. It is necessary to introduce full-domain data acquisition and spatiotemporal trajectory extrapolation, combined with interference coupling coefficients to achieve quantitative assessment of interference risk. By designing a hierarchical control mechanism, differential drift is prioritized for interference-free operation. When drift is not possible, frequency conversion adjustment is used to avoid interference and reserve space for rewashing, allowing the control strategy to adaptively adjust according to the operating scenario, thereby solving the problems of delayed interference prediction, rigid control, and poor cleaning quality.

[0030] After introducing the basic concept of the present invention, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0031] Example 1:

[0032] Please see Figure 1 As shown, a multi-robot collaborative cleaning scheduling method suitable for photovoltaic power plants is proposed, which includes:

[0033] Real-time operating status data and environmental flow field vector data of the robot in the entire photovoltaic power station are collected. The operation jet vector in the real-time operating status data and the environmental flow field vector data are extracted and vector superimposed to construct a diffusion elliptical domain model data with the coordinates of each robot as the focus.

[0034] Preferably, the data for constructing a diffusion elliptical domain model with each robot's coordinates as the focus includes:

[0035] Real-time location coordinates, cleaning intensity, and initial spray direction vector are extracted from real-time operation status data, and environmental wind speed scalar and environmental wind direction vector are separated from environmental flow field vector data.

[0036] The initial injection direction vector and the environmental wind direction vector are vector synthesized to generate a synthetic diffusion direction vector that characterizes the actual drift trend of pollutants.

[0037] Based on the combination relationship between the cleaning operation intensity value and the environmental wind speed scalar, the corresponding diffusion distance parameter is matched in the preset influence range mapping relationship, and the diffusion distance parameter is defined as the major axis data and minor axis data of the geometric model;

[0038] Using real-time location coordinate data as spatial anchor points, the spatial orientation angle is determined based on the synthetic diffusion direction vector, and a closed geometric boundary is generated by combining the major axis data and the minor axis data, thereby constructing the diffusion elliptical domain model data.

[0039] Preferably, the corresponding diffusion distance parameter is matched in the preset influence range mapping relationship, specifically including:

[0040] A weighted correlation calculation is performed between the cleaning operation intensity value and the environmental wind speed scalar to generate a longitudinal drift factor that characterizes the degree of splashing in the downwind direction, and the cleaning operation intensity value is extracted separately as a lateral dispersion factor that characterizes its own radiation range.

[0041] The longitudinal drift factor and the lateral diffusion factor are used as index keys, respectively. The data are retrieved from the preset distance mapping database to extract the physical length values ​​corresponding to each factor value, and the longitudinal extension distance data and the lateral coverage distance data are generated respectively.

[0042] The longitudinal extension distance data is assigned as the major axis data of the geometric model, and the lateral coverage distance data is assigned as the minor axis data of the geometric model, thereby completing the size definition of the diffusion elliptical domain model data.

[0043] In one specific embodiment, for each robot in operation, the current physical position coordinates are parsed from the real-time operating status data. and the work intensity value that characterizes the cleaning effect. ;

[0044] In practice, the robot uses water power to clean the photovoltaic power station. The work intensity value here corresponds to the water pump spray pressure, and the unit is Bar.

[0045] Simultaneously, the initial injection direction vector of the robot nozzle is extracted. Simultaneously separate the environmental wind speed scalar from the environmental flow field vector data. With environmental wind direction vector .

[0046] Next, all vectors are mapped to a unified two-dimensional Cartesian coordinate system, and the composite diffusion direction vector is calculated. The calculation formula is as follows:

[0047] in, This is the direction vector of the synthetic diffusion. The initial injection velocity vector of the nozzle, whose modulus is determined by the working intensity. It is derived from Bernoulli's equation; This represents the ambient wind speed vector. The air resistance coupling coefficient is used to characterize the degree to which water droplets are pulled by wind. The density ratio of air to water. air density, The density of water;

[0048] Based on the calculation Using inverse trigonometric functions Calculate the deflection angle of the elliptical domain relative to true north. . and Synthesize diffusion direction vectors separately The x-axis and y-axis components in a two-dimensional Cartesian coordinate system;

[0049] The system first normalizes the input variables and then calculates the longitudinal drift factor. and lateral diffusion factor .

[0050] Longitudinal drift factor It primarily represents the distance of water mist splashing downwind, and its calculation formula uses weighted correlation operations:

[0051]

[0052] in, and These are the normalized cleaning intensity and wind speed scalars, respectively; As the initial velocity weight of the jet, As the weighting factor for wind load, the squared treatment of the wind speed term here is based on the physical law in fluid mechanics that dynamic pressure is proportional to the square of velocity, which means that the effect of increasing wind speed on drift distance exhibits a non-linear and dramatic increase.

[0053] Lateral diffusion factor This mainly represents the width of the nozzle coverage, and the calculation formula is:

[0054]

[0055] The square root relationship is used here because, with a fixed nozzle orifice diameter, the coverage width generated by the injection angle gradually decreases as the pressure increases. is the nozzle structure constant, a dimensionless parameter. The nozzle structure constant is related to the nozzle geometry and affects the calculation of lateral dispersion.

[0056] In obtaining dimensionless and Then, it is used as an index key value and input into a preset distance mapping database, which is fitted with historical wind tunnel experimental data. The factors are then converted into physical lengths using a linear regression function.

[0057]

[0058]

[0059] in, For the major axis data of the geometric model, For minor axis data; and This refers to the basic spray size under calm conditions; and Dimensional conversion coefficients are used to convert dimensionless quantities into linear quantities. and Convert to physical length. This indicates the maximum extension distance of the water mist diffusion area in the synthesis direction; This indicates the maximum diffusion width of the water mist diffusion area perpendicular to the synthesis direction.

[0060] Finally, using the robot's real-time position coordinates Let be one focus of the ellipse, and construct a closed diffusion elliptic domain model using parametric equations. For any point within the elliptic domain... Its boundary determination conditions are satisfied as follows:

[0061]

[0062] This inequality describes a center located at... The major axis is The minor axis is And the major axis has rotated relative to the X-axis. The solid elliptical region of the corner. The system marks the set of all coordinate points that meet this condition as the diffusion elliptical domain model data of the robot at the current moment. This model dynamically quantifies the "unsafe area" that may be affected by water mist and pollutants during operation.

[0063] The longitudinal extension distance and lateral coverage distance data are derived by calculating the longitudinal drift factor and the lateral dispersion factor. This is because directly using raw wind speed and pressure data cannot intuitively characterize the spatial distribution of pollutants, and the dominant physical forces in different directions (downwind and lateral) are completely different. Longitudinal drift is mainly dominated by the aerodynamic drag generated by wind load, and the dynamic pressure is proportional to the square of the wind speed. Therefore, the square term of the wind speed is introduced when constructing the longitudinal drift factor to capture the nonlinear characteristic of the exponential growth of drift distance under strong winds, while also superimposing the initial momentum provided by the initial jet velocity. Lateral dispersion is mainly dominated by the nozzle atomization angle and internal turbulence. As pressure increases, the atomized particles become finer, but air resistance also increases dramatically, leading to a slower width expansion rate. Therefore, the square root relationship of pressure is used in the lateral dispersion factor to simulate this diminishing marginal effect.

[0064] This approach aligns with the breakup and transport characteristics of liquid jets in crosswinds. By encapsulating these complex physical mechanisms into two independent dimensionless factors—"longitudinal" and "lateral"—a feature descriptor for an intermediate state is essentially constructed. The influence of ambient wind (dominantly longitudinal) and the influence of equipment operating parameters (dominantly lateral) are mathematically separated, facilitating model debugging.

[0065] Based on the travel velocity vector in the real-time operating status data, the spatiotemporal trajectory of multiple robots is extrapolated, the expected spatiotemporal coordinate set of the encounter is calculated, and the diffusion elliptical domain model data is mapped to the expected spatiotemporal coordinate set of the encounter. The interference coupling coefficient is generated by calculating the overlapping area of ​​the elliptical domains of the robots moving towards each other.

[0066] Preferably, calculating the interference coupling coefficient by the overlapping area of ​​the elliptical domains of the opposing robots includes:

[0067] Extract the travel speed vector and real-time position coordinate data from the real-time operating status data, calculate the relative speed and relative distance values ​​of the robots traveling in opposite directions, estimate the expected meeting time based on the ratio of the relative distance value to the relative speed value, and determine the corresponding expected meeting position point in combination with the expected meeting time point. Combine the expected meeting time point and the expected meeting position point into a set of expected meeting spatiotemporal coordinates.

[0068] Read the diffusion elliptic domain model data, keep its geometric shape parameters and diffusion direction parameters unchanged, and virtually translate its geometric center from the real-time position coordinate data at the current moment to the expected meeting point, thereby constructing the virtual interference field distribution data at the future moment;

[0069] Geometric intersection calculations are performed on the virtual interference field distribution data of the opposing robots to identify the common coverage area of ​​the two robots on the same plane, and the area value of the common coverage area is calculated.

[0070] The area value of the public coverage area is compared with the preset benchmark area threshold and normalized to output the quantized interference coupling coefficient, which is used to characterize the risk of secondary pollution between robots.

[0071] Preferably, virtual interference field distribution data for future moments is constructed, specifically including:

[0072] The diffusion elliptical domain model data is analyzed to separate the geometric morphological features representing the coverage area and the diffusion direction vector representing the jet direction.

[0073] The coordinate difference between the real-time location coordinate data and the expected meeting point is calculated to generate a spatial displacement vector connecting the current point and the future point.

[0074] The spatial displacement vector is superimposed onto the original center coordinates of the diffusion elliptical domain model data to obtain the updated virtual center coordinates.

[0075] The virtual center coordinates are recombined with the unchanged geometric morphological feature data and diffusion direction vector to generate virtual disturbance field distribution data to characterize the pollution impact range at future moments.

[0076] In one specific embodiment, the system extracts real-time position coordinate data, including timestamps, from two target robots (denoted as Robot A and Robot B) that are on the same travel path or have a potential collision tendency. With the velocity vector The Euclidean distance between the two was calculated using the principles of relative kinematics. and relative velocity vector Based on the assumption of uniform linear motion, a formula for predicting the meeting time is constructed. ,in The current system time. The cosine of the angle between the velocity vectors of the two vehicles is used to correct for the encounter component when they are not traveling head-on. If the denominator approaches zero, it is determined that there is no risk of encounter; this is combined with the calculated relative time increment. The predicted meeting points of the two robots at future moments are calculated using vector extrapolation. ,Will and The combination is encapsulated as a set of expected encounter spacetime coordinates.

[0077] Then, the diffusion elliptical domain model data is read, preserving its core geometric features (major axis L, minor axis W) and the diffusion direction vector (deflection angle). (The coordinates remain unchanged, and a virtual translation operation based on rigid body transformation logic is performed; specifically, the coordinates from the current real-time position are calculated.) Pointing to the expected meeting point Spatial displacement vector This vector is linearly superimposed onto the geometric center coordinates of the original elliptical domain model to generate new virtual center coordinates. This process constructs virtual interference field distribution data for the moment when the two robots meet in the future. The physical significance of this step is to simulate where the pollutant diffusion cloud carried by the robots will be located at the moment of the meeting if the robots continue to drive in the current state. This is a feedforward risk prediction mechanism that avoids the lag in traditional feedback control where pollution occurs before avoidance.

[0078] Subsequently, the virtual disturbance field distribution data of the two robots (i.e., two parameterized elliptical regions) were analyzed. and Perform geometric intersection operations to identify the common coverage area of ​​the two objects on the same two-dimensional plane, and calculate the area of ​​this common coverage area. Considering that analytical solutions involve solving high-order equations, resulting in excessive computation, the Monte Carlo integration method or grid discretization method is preferred for estimation here. That is, a density of [density value missing] is generated in the potential overlapping region. Given a test point set, count the number of sample points that simultaneously satisfy the constraints of two elliptic equations. ,but This area value intuitively reflects the spatial extent in which the water mist or splashed pollutants from the two robots physically mix at the moment of encounter.

[0079] Finally, the calculated public coverage area value will be... With the preset allowable overlap area threshold Perform comparison and normalization processing, and output the quantized interference coupling coefficient. Its calculation formula is set as follows .

[0080] In this formula:

[0081] The preset threshold for the allowable overlap area is set based on the robot's self-cleaning capability. It is preferably set to 5%-10% of the standard working ellipse area of ​​a single robot, meaning "slight edge overlap that can be tolerated without affecting the quality of the work".

[0082] As an environmental severity factor, when the environmental wind speed sensor detects that the wind speed exceeds the preset warning value, this factor increases linearly with the wind speed. It is used to artificially amplify the risk coefficient under severe weather conditions and improve the system's conservatism.

[0083] like If the value is greater than 1, it indicates that the risk of secondary pollution at the time of future encounter has exceeded the safety boundary, and the system needs to immediately trigger path planning adjustment. This formula, by introducing environmental factor correction, transforms the purely geometric area ratio into a dynamic risk assessment index with environmental adaptability, ensuring the robustness of the discrimination standard under different meteorological conditions.

[0084] The interference coupling coefficient is compared with a preset safety threshold. When the interference coupling coefficient exceeds the preset safety threshold, the coordinate data of the safe node in the pre-stored photovoltaic module layout map is retrieved, and the drift velocity variable of the meeting point corresponding to the expected encounter time-space coordinate set is calculated.

[0085] Preferably, calculating the drift velocity variable of the meeting point corresponding to the expected meeting time-space coordinate set translated to the safe node coordinate data includes:

[0086] The interference coupling coefficient is compared with the preset safety threshold. If the interference coupling coefficient is determined to be greater than the preset safety threshold, the path replanning logic is activated and the pre-stored photovoltaic module layout map is read.

[0087] Using the location coordinates in the expected encounter time-space coordinate set as the search center, the coordinate points of the non-power generation area are traversed in the photovoltaic module layout map, and the point with the smallest Euclidean distance from the location coordinates is selected and determined as the safe node coordinate data.

[0088] Calculate the spatial displacement vector between the coordinate data of the safe node and its position coordinates, and extract the time parameter from the set of expected meeting time spatial coordinates;

[0089] Based on the ratio of spatial displacement vector to time parameter, inverse deduction is performed to calculate the additional velocity component required to achieve position offset, and the additional velocity component is output as the drift velocity variable at the meeting point.

[0090] Preferably, the reverse deduction logic based on the ratio of spatial displacement vector to time parameter is as follows:

[0091] Extract the remaining time value from the time parameter, establish a rate calculation model with the spatial displacement vector as the numerator and the remaining time value as the denominator, and calculate the theoretical offset velocity vector per unit time through the ratio.

[0092] The theoretical offset velocity vector is analyzed to obtain velocity amplitude data and direction guidance data. The velocity amplitude data is defined as the rate gain value required to achieve spatial crossing.

[0093] The rate gain value is associated with and encapsulated with the direction guidance data to generate an additional velocity component.

[0094] It should be added that the preset logic of the rate calculation model is as follows:

[0095] Construct a kinematic sample set containing the robot's historical maneuver trajectories. For each historical maneuver trajectory, extract the physical distance vector between the trajectory's starting point and ending point as a distance reference sample for the model input, and extract the time taken to complete the trajectory as a time reference sample. At the same time, mark the attitude stability value during the trajectory operation as a validity verification label.

[0096] The distance reference sample is divided into multiple gradient displacement intervals according to the magnitude. Within each displacement interval, standard time records with validity verification labels higher than the preset stability threshold are selected. The ratio of the distance reference sample to the standard time record is calculated to generate the ideal rate guidance value corresponding to each interval.

[0097] A vector division function structure is constructed with the spatial displacement vector as the dividend variable and the remaining time value as the divisor variable. A safety redundancy coefficient is introduced to compensate for mechanical inertia. The ideal rate guidance value is fitted and calibrated using the least squares method to determine the convergence value of the safety redundancy coefficient, thereby generating a rate calculation model.

[0098] In one specific embodiment, the interference coupling coefficient calculated in real time is... With preset safety threshold Perform numerical comparison. If determined... The system immediately activates its path replanning logic, retrieving a pre-stored photovoltaic module layout map. This map, stored in high-precision grid or vector node format, clearly defines the photovoltaic module coverage area and maintenance access routes. The system then uses the location coordinates from the expected encounter time-space coordinate set. As the retrieval center, the K-nearest neighbor algorithm is used to traverse all non-power generation area coordinates in the maintenance and repair channel layer of the map, and filter out those that are related to the center. The point with the smallest Euclidean distance between the nodes is locked as the safe node coordinate data. .

[0099] Next, the system construction starts from point to Spatial displacement vector It also extracts the time parameters from the expected meeting time-space coordinate set and calculates the current time. Distance from the expected meeting time Remaining time value .

[0100] In order to calculate accurately within a finite Completed internally The additional velocity required for displacement was calculated in reverse using a pre-trained rate calculation model in this embodiment. This is because in actual operation, robots experience mechanical inertia, ground friction fluctuations, and motor response lag; simple linear calculations often result in the actual position arriving later than the theoretical position.

[0101] The construction process of the velocity calculation model is as follows: First, a kinematic sample set containing the robot's historical maneuver trajectories is established, and for each historical trajectory... Extract the physical distance vectors of its start and end points. As a distance reference sample, extract the actual time consumption. As a time reference sample, the attitude stability values ​​measured by the IMU are recorded. The value ranges from 0 to 1, with the value of 1 being the most stable.

[0102] To ensure the robustness of the model, the system first... Filter out abnormal data such as bumps or slippage, and retain only... High-quality samples, among which This represents the preset minimum tolerance threshold. Subsequently, the distance to the reference sample is divided into multiple gradient intervals according to the magnitude, and the ideal rate steering value is calculated within each interval. .

[0103] Based on this, a safety redundancy coefficient is introduced. The structure of the vector division function:

[0104] in, This is the inertial compensation term. To determine... To determine the convergence value, the system uses a nonlinear least squares method to fit historical data, aiming to minimize the mean square error between the theoretical calculation time and the actual historical calculation time. Through training, the optimal solution is calculated for different distance intervals. value.

[0105] Ultimately, the system matches the corresponding distance based on the displacement of the current task. Substituting into the above formula, the theoretical offset velocity vector per unit time can be calculated. The system analyzes the vector and extracts its magnitude as the rate gain value. Extract its unit direction vector as direction guidance data. And repackage the two as a drift velocity variable at the meeting point. .

[0106] The drift velocity variable at the meeting point is a control parameter with velocity dimensions. Its physical meaning is: to ensure that the robot's actual position is exactly offset from the safe node when it reaches the meeting point, an additional velocity component needs to be added to the robot's original cruising speed. This calculation process considers both spatial geometric constraints and compensates for system dynamic errors through historical data fitting, ensuring the accuracy and safety of the drift maneuver.

[0107] Determine whether the drift speed variable at the meeting point is within the preset allowable speed change range. If so, output a differential drift control command containing the drift speed variable at the meeting point to the associated robot.

[0108] Preferably, the output differential drift control command includes the drift speed variable at the meeting point, including:

[0109] The rated power parameters and minimum stable operating speed parameters of the robot chassis drive system are retrieved, and the feasible physical boundaries for speed adjustment are defined based on the rated power parameters and minimum stable operating speed parameters, thereby constructing a preset allowable speed range.

[0110] The numerical amplitude of the drift speed variable at the meeting point is verified to be inclusive of the preset allowable speed change range. When the verification result confirms the inclusive relationship, the associated robot in the current task is locked and its real-time travel speed data is read.

[0111] The real-time travel speed data and the drift speed variable at the meeting point are algebraically weighted to calculate the corrected target cruise rate that meets the preset safety node arrival requirements, and the corrected target cruise rate is converted into the corresponding motor drive frequency signal.

[0112] The motor drive frequency signal is associated with and packaged with the device identification code of the associated robot to generate differential drift control commands and send them to the execution control unit.

[0113] In one specific embodiment, when the system calculates the drift velocity variable at the meeting point... Next, the rated power parameters of the robot chassis drive system need to be retrieved. With minimum stable operating rate parameter The minimum stable operating rate parameter characterizes the lower limit of the speed at which the motor overcomes static friction without experiencing low-frequency vibration. Based on the law of conservation of energy and the ground resistance model, the maximum theoretical cruising speed of the robot under its current load is calculated. The calculation formula is as follows:

[0114]

[0115] in Let m be the efficiency coefficient of the transmission system, m be the total mass of the robot, and g be the acceleration due to gravity. The coefficient of rolling friction on the surface of the photovoltaic panel. The current wind resistance is used to construct the feasible physical boundary for speed adjustment, i.e., the preset allowable speed variation range. .

[0116] Then, the drift speed variable at the meeting point (Including velocity gain amplitude) With direction ) and the current real-time travel speed data of the robot Perform vector superposition simulation and calculate the predicted combined velocity modulus after superposition. The system performs interval inclusion verification, that is, it determines... Whether it falls Within the closed interval; if the verification result confirms an inclusion relationship, i.e. If the drift maneuver is within the hardware's capabilities, the system locks onto the associated robot in the current task and reads its real-time speed data. The system performs an algebraic weighted calculation on the real-time speed data and the drift speed variable at the meeting point to calculate the corrected target cruising speed. This operation uses a simplified variation of the proportional-integral adjustment concept, and the calculation formula is as follows:

[0117] in, This is the corrected target cruise rate vector; Weights are used to maintain inertia and stabilize the existing motion trend. The drift response weights ensure that the required displacement deviation can be adequately responded to; the weighting here is intended to smooth out command abrupt changes caused by sensor noise.

[0118] To convert this physical speed command into a motor control signal, the system utilizes the robot's inverse kinematics model. Decompose into the linear velocity of the left wheel and the linear velocity of the right wheel This is then further converted into the corresponding motor drive frequency signal. The conversion formula is:

[0119]

[0120] in, Substitute or Both are in m / s; This refers to the reduction ratio of the gearbox; The physical radius of the drive wheel; This represents the number of pole pairs of the brushless motor.

[0121] Finally, the calculated left and right wheel motor drive frequency signals are associated and packaged with the associated robot's device identification code, encapsulated according to the industrial-grade CAN-Open communication protocol, to generate differential drift control commands, which are then sent to the execution control unit via the wireless communication module to drive the robot chassis to perform precise differential motion, thereby achieving the expected meeting point drift.

[0122] It should be added that... Decompose into the linear velocity of the left wheel and the linear velocity of the right wheel The specific decomposition steps are as follows:

[0123] according to The required rotational angular velocity is calculated by determining the angular deviation between the current vehicle orientation and the current vehicle orientation. Then, combined with the robot's wheelbase parameters That is, the distance between the centers of the two drive wheels, which can be solved using the following set of kinematic equations:

[0124]

[0125] in, for The linear velocity component in the direction of travel; and These are the target linear velocities of the right and left wheels, respectively. This is achieved by superimposing an angular velocity onto the baseline linear velocity v. The generated differential component This allows the robot to generate a yaw torque while maintaining forward propulsion, thus allowing it to travel along a predetermined drift trajectory.

[0126] If the drift speed variable at the meeting point exceeds the preset allowable speed range, the diffusion elliptical domain model data is solved in reverse based on the preset safety threshold to calculate the extreme operation intensity value that satisfies the non-interference condition and generate a frequency conversion adjustment command. At the same time, virtual uncleaned area marker data is generated based on the extreme operation intensity value and written into the global task scheduling queue for subsequent path planning.

[0127] Preferably, the calculation of the limit operation intensity value that satisfies the condition of non-interference and the generation of frequency conversion adjustment commands include:

[0128] Numerical verification is performed on the drift velocity variable at the meeting point and the preset allowable speed change range. When it is confirmed that the range is exceeded, the expected spatiotemporal coordinate set of the meeting at the current moment is locked, and the preset safety threshold is used as the maximum allowable overlap boundary constraint for the diffusion elliptical domain model data.

[0129] While keeping the expected spatiotemporal coordinate set unchanged, the diffusion elliptical domain model data is geometrically contracted based on the maximum allowable overlap boundary constraint to obtain the safe geometric dimension parameters after contraction. Based on the preset correlation model between diffusion range and output intensity, the safe geometric dimension parameters are used as target indexes to reverse match the corresponding functional component driving parameters in the correlation model, and the functional component driving parameters are defined as the extreme operation intensity value.

[0130] The extreme work intensity value is converted into a corresponding hardware drive level signal, a frequency conversion adjustment command is generated, and then sent to the associated robot execution unit.

[0131] It should be added that the virtual unwashed area marker data generated based on the extreme operation intensity value includes:

[0132] Calculate the power difference between the extreme work intensity value and the preset rated full load intensity value. If the power difference indicates a lack of cleaning efficiency, extract the coordinate range of the current work section and combine it with the power difference to generate virtual uncleaned area marking data containing supplementary work weights.

[0133] The virtual unwashed area marking data is uploaded to the central control system and appended to the end of the global task scheduling queue as a remedial task node to be assigned for subsequent idle robots to use.

[0134] Preferably, the geometric shrinkage model data of the diffusion elliptical domain is derived based on the maximum permissible overlap boundary constraint to obtain the safe geometric dimension parameters after shrinkage. The specific operation logic is as follows:

[0135] Extract the relative position vector from the expected encounter spacetime coordinate set, input it as a fixed invariant into the geometric collision simulation space, and set the maximum allowable overlap boundary constraint as the critical judgment value for simulation termination;

[0136] Apply proportionally decreasing scaling to the major axis radius data and minor axis radius data in the diffusion elliptic domain model data to generate multiple sets of continuously changing intermediate geometric model data;

[0137] Calculate the theoretical intersection area of ​​each group of intermediate geometric model data under the relative position vector in turn. When the theoretical intersection area is found to be lower than the maximum allowable overlap boundary constraint, stop the decreasing scaling process immediately.

[0138] Extract the specific major axis and minor axis values ​​corresponding to the intermediate state geometric model data at the stopping moment, and combine and encapsulate them into safe geometric dimension parameters.

[0139] In one specific embodiment, for Modulus and velocity range Numerical verification is performed. When an out-of-bounds error is confirmed, the system maintains the set of predicted meeting time-space coordinates derived at the current moment (i.e., assumes a future time). Two robots in position The encounter is inevitable. At this point, the control objective is changed from "changing position" to "reducing the scope of influence," and the preset safety threshold is set. As a geometric constraint.

[0140] To determine the safe operating range, the system constructs a geometric collision simulation space in memory, storing the relative position vectors of the two machines at the expected encounter. As fixed parameters, the major axis L and minor axis W of the original diffusion elliptic domain model data are then subjected to a proportionally decreasing scaling process. That is, a scaling factor k is set to generate intermediate geometric model data: where the scaling factor k is initially 1.0 and the step size is -0.01;

[0141]

[0142]

[0143] For each value of k, calculate the distance between the two reduced ellipses at a fixed interval. Theoretical intersection area When identified When the maximum allowable overlap boundary constraint is satisfied, immediately stop the iteration and record the current scaling factor. And extract the corresponding specific major axis values. With a specific minor axis value They are combined and encapsulated into safe geometric dimensional parameters.

[0144] Based on the pre-defined correlation model between diffusion range and output intensity, the safety geometric parameters are... Using the target index as the reverse calculation, the corresponding functional component driving parameters are obtained, which is the extreme workload value. To ensure the uniqueness of the solution, the major axis constraint is usually matched first, and its reverse calculation formula is as follows:

[0145] in, This is the maximum rated strength of the equipment. This formula ensures that the output working intensity does not produce a physical diffusion range that does not exceed the safe geometric dimensions. The system will calculate... The signal is converted into a corresponding PWM duty cycle signal and sent directly to the execution unit to forcibly reduce the water pump pressure, thereby physically shrinking the pollution cloud and ensuring that the two machines do not interfere with each other when they meet.

[0146] Meanwhile, to compensate for the decrease in cleaning quality caused by reduced power operation, the system executes remedial marking logic. The power difference is calculated. ,in This represents the operating power under extreme intensity. If... If the performance exceeds the preset performance tolerance threshold, it indicates that the cleaning of the current road section is incomplete. The system extracts the coordinate range of the current working road section and combines it with... Generate virtual unwashed area marking data. This data includes the road segment ID, estimated amount of remaining stains, and rework weight. The estimated amount of remaining stains should be consistent with... Proportional;

[0147] Homework weighting The calculation uses the following logic:

[0148]

[0149] in Based on weights, This represents the urgency coefficient. This data is uploaded to the central control system and appended to the end of the global task scheduling queue. Thus, when an idle robot passes nearby or enters the next scheduling round, the scheduling algorithm will adjust accordingly. The secondary operation of the "virtual uncleaned area" is arranged according to the priority of high and low, sacrificing the efficiency of individual machines for safety during congested periods, and improving the quality of the closed loop of remedial tasks during idle periods.

[0150] It should be added that the preset logic of the correlation model between the preset diffusion range and the output intensity mentioned above is as follows:

[0151] The pre-defined correlation model between diffusion range and output intensity is essentially a mathematical tool for establishing a two-way mapping relationship between physical operational parameters (such as pump pressure) and spatial geometric parameters (diffusion distance). During the system initialization phase, simulations are used to collect data at different operational intensities. (Independent variable) The actual diffusion major axis L and minor axis W generated under different wind speed conditions (dependent variable) are fitted with a positive function relationship using a polynomial regression algorithm. .

[0152] Since the working intensity I and the diffusion axis L exhibit a monotonically increasing relationship in physics—that is, the greater the pressure, the farther the spray—this function is reversible. The system will use the safety geometric axis... As a known target value, the real-time wind speed As an environmental parameter, a unique work intensity value is calculated through numerical iteration or direct table lookup interpolation. This process ensures that when the system requires limiting the diffusion range... Within a certain range, it is possible to accurately deduce the power output of the device, i.e., the driving parameters of the functional components.

[0153] Example 2:

[0154] Please see Figure 2 As shown, a multi-robot collaborative cleaning scheduling system suitable for photovoltaic power plants is disclosed. The system includes a diffusion model construction module, an interference coefficient calculation module, a drift speed generation module, and a drift speed change judgment module.

[0155] The various modules are connected via wired and / or wireless connections to enable data transmission between them;

[0156] Diffusion model construction module: Collect real-time operating status data and environmental flow field vector data of the robot in the whole field of photovoltaic power station, extract the operation jet vector from the real-time operating status data and perform vector superposition operation with the environmental flow field vector data to construct diffusion elliptical domain model data with the coordinates of each robot as the focus;

[0157] Interference coefficient calculation module: Based on the travel speed vector in the real-time running status data, the module performs spatiotemporal trajectory extrapolation of multiple robots, calculates the expected encounter spatiotemporal coordinate set, maps the diffusion elliptical domain model data to the expected encounter spatiotemporal coordinate set, and generates the interference coupling coefficient by calculating the overlapping area of ​​the elliptical domains of the robots moving towards each other.

[0158] Drift speed generation module: compares the interference coupling coefficient with the preset safety threshold. When the interference coupling coefficient exceeds the preset safety threshold, it retrieves the safety node coordinate data in the pre-stored photovoltaic module layout map and calculates the drift speed variable of the meeting point corresponding to the expected encounter time-space coordinate set translated to the safety node coordinate data.

[0159] Drift speed change judgment module: Determines whether the drift speed variable at the meeting point is within the preset allowable speed change range. If so, it outputs a differential drift control command containing the drift speed variable at the meeting point to the associated robot. If the drift speed variable at the meeting point exceeds the preset allowable speed change range, it performs inverse solution on the diffusion elliptical domain model data based on the preset safety threshold, calculates the extreme operation intensity value that satisfies the non-interference condition, and generates a frequency conversion adjustment command. At the same time, it generates virtual uncleaned area marker data based on the extreme operation intensity value and writes the virtual uncleaned area marker data into the global task scheduling queue for subsequent path planning calls.

[0160] It should be noted that the interval and threshold sizes are set for ease of comparison. The size of the threshold depends on the amount of sample data and the base number set by those skilled in the art for each set of sample data, as long as it does not affect the proportional relationship between the parameter and the quantized value. Furthermore, the above formulas are all dimensionless calculations, and the formulas are derived from software simulations using a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0161] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0162] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0163] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-robot collaborative cleaning scheduling method suitable for photovoltaic power plants, characterized in that, include: Real-time operating status data and environmental flow field vector data of the full-domain robot of the photovoltaic power station are collected. The operation jet vector in the real-time operating status data and the environmental flow field vector data are extracted and vector superimposed to construct a diffusion elliptical domain model data with the coordinates of each robot as the focus. Based on the travel velocity vector in the real-time operating status data, the spatiotemporal trajectory of multiple robots is extrapolated, the expected meeting spatiotemporal coordinate set is calculated, and the diffusion elliptical domain model data is mapped to the expected meeting spatiotemporal coordinate set. The interference coupling coefficient is generated by calculating the overlapping area of ​​the elliptical domains of the robots moving towards each other. The interference coupling coefficient is compared with the preset safety threshold. When the interference coupling coefficient exceeds the preset safety threshold, the safety node coordinate data in the pre-stored photovoltaic module layout map is retrieved, and the drift speed variable of the meeting point corresponding to the expected encounter time-space coordinate set is calculated. Determine whether the drift speed variable at the meeting point is within the preset allowable speed change range. If so, output a differential drift control command containing the drift speed variable at the meeting point to the associated robot. If the drift speed variable at the meeting point exceeds the preset allowable speed range, the diffusion elliptical domain model data is solved in reverse based on the preset safety threshold to calculate the extreme operation intensity value that satisfies the non-interference condition and generate a frequency conversion adjustment command. At the same time, virtual uncleaned area marker data is generated based on the extreme operation intensity value and written into the global task scheduling queue for subsequent path planning.

2. The multi-robot collaborative cleaning scheduling method for photovoltaic power plants according to claim 1, characterized in that, Construct a diffusion elliptical domain model data with each robot's coordinates as the focus, including: Real-time location coordinates, cleaning intensity, and initial spray direction vector are extracted from real-time operation status data, and environmental wind speed scalar and environmental wind direction vector are separated from environmental flow field vector data. The initial jet direction vector and the ambient wind direction vector are vector synthesized to generate a composite diffusion direction vector. Based on the combination relationship between the cleaning operation intensity value and the environmental wind speed scalar, the corresponding diffusion distance parameter is matched in the preset influence range mapping relationship, and the diffusion distance parameter is defined as the major axis data and minor axis data of the geometric model; Using real-time location coordinate data as spatial anchor points, the spatial orientation angle is determined based on the synthetic diffusion direction vector, and a closed geometric boundary is generated by combining the major axis data and the minor axis data, thereby constructing the diffusion elliptical domain model data.

3. The multi-robot collaborative cleaning scheduling method for photovoltaic power plants according to claim 2, characterized in that, Match the corresponding diffusion distance parameter from the preset influence range mapping relationship, specifically including: A weighted correlation calculation is performed between the cleaning operation intensity value and the environmental wind speed scalar to generate a longitudinal drift factor, and the cleaning operation intensity value is extracted separately as a lateral dispersion factor. The longitudinal drift factor and the lateral diffusion factor are used as index keys, respectively. The data are retrieved from the preset distance mapping database to extract the physical length values ​​corresponding to each factor value, and the longitudinal extension distance data and the lateral coverage distance data are generated respectively. Assign the longitudinal extension distance data to the major axis data of the geometric model, and assign the lateral coverage distance data to the minor axis data of the geometric model.

4. The multi-robot collaborative cleaning scheduling method for photovoltaic power plants according to claim 1, characterized in that, Calculating the interference coupling coefficient by the overlapping area of ​​the elliptical domains of opposing moving robots includes: Extract the travel speed vector and real-time position coordinate data from the real-time operating status data, calculate the relative speed and relative distance values ​​of the robots traveling in opposite directions, estimate the expected meeting time based on the ratio of the relative distance value to the relative speed value, and determine the corresponding expected meeting position point in combination with the expected meeting time point. Combine the expected meeting time point and the expected meeting position point into a set of expected meeting spatiotemporal coordinates. Read the diffusion elliptic domain model data and virtually translate its geometric center from the real-time position coordinates of the current moment to the expected meeting point, thereby constructing virtual interference field distribution data for future moments; Geometric intersection calculations are performed on the virtual interference field distribution data of the opposing robots to identify the common coverage area of ​​the two robots on the same plane, and the area value of the common coverage area is calculated. The area value of the public coverage area is compared with the preset benchmark area threshold and normalized to output the quantized interference coupling coefficient.

5. A multi-robot collaborative cleaning scheduling method for photovoltaic power plants according to claim 4, characterized in that, The virtual interference field distribution data for future moments is constructed, specifically including: Analyze the diffusion elliptic domain model data to separate the geometric morphological feature data and the diffusion direction vector; The coordinate difference between the real-time location coordinate data and the expected meeting point is calculated to generate a spatial displacement vector connecting the current point and the future point. The spatial displacement vector is superimposed onto the original center coordinates of the diffusion elliptical domain model data to obtain the updated virtual center coordinates. The virtual center coordinates are recombined with the geometric features of the diffusion elliptical domain model data and the diffusion direction vector to generate virtual disturbance field distribution data that characterizes the extent of pollution impact at future times.

6. The multi-robot collaborative cleaning scheduling method for photovoltaic power plants according to claim 1, characterized in that, Calculating the drift velocity variable of the meeting point corresponding to the expected spatiotemporal coordinate set translated to the safe node coordinate data includes: The interference coupling coefficient is compared with the preset safety threshold. If the interference coupling coefficient is determined to be greater than the preset safety threshold, the path replanning logic is activated and the pre-stored photovoltaic module layout map is read. Using the location coordinates in the expected encounter time-space coordinate set as the search center, the coordinate points of the non-power generation area are traversed in the photovoltaic module layout map, and the point with the smallest Euclidean distance from the location coordinates is selected and determined as the safe node coordinate data. Calculate the spatial displacement vector between the coordinate data of the safe node and its position coordinates, and extract the time parameter from the set of expected meeting time spatial coordinates; Based on the ratio of spatial displacement vector to time parameter, inverse deduction is performed to calculate the additional velocity component required to achieve position offset, and the additional velocity component is output as the drift velocity variable at the meeting point.

7. A multi-robot collaborative cleaning scheduling method for photovoltaic power plants according to claim 1, characterized in that, The output includes differential drift control commands containing drift speed variables at the meeting point, including: The rated power parameters and minimum stable operating speed parameters of the robot chassis drive system are retrieved, and the feasible physical boundaries for speed adjustment are defined based on the rated power parameters and minimum stable operating speed parameters, thereby constructing a preset allowable speed range. The numerical amplitude of the drift speed variable at the meeting point is verified to be inclusive of the preset allowable speed change range. When the verification result confirms the inclusive relationship, the associated robot in the current task is locked and its real-time travel speed data is read. The real-time travel speed data and the drift speed variable at the meeting point are algebraically weighted to calculate the corrected target cruise rate that meets the preset safety node arrival requirements, and the corrected target cruise rate is converted into the corresponding motor drive frequency signal. The motor drive frequency signal is associated with and packaged with the device identification code of the associated robot to generate differential drift control commands and send them to the execution control unit.

8. A multi-robot collaborative cleaning scheduling method for photovoltaic power plants according to claim 1, characterized in that, Calculate the limit operation intensity value that satisfies the condition of non-interference and generate frequency converter control commands, including: Numerical verification is performed on the drift velocity variable at the meeting point and the preset allowable speed change range. When it is confirmed that the range is exceeded, the expected spatiotemporal coordinate set of the meeting at the current moment is locked, and the preset safety threshold is used as the maximum allowable overlap boundary constraint for the diffusion elliptical domain model data. While keeping the expected spatiotemporal coordinate set unchanged, the diffusion elliptical domain model data is geometrically contracted based on the maximum allowable overlap boundary constraint to obtain the safe geometric dimension parameters after contraction. Based on the preset correlation model between diffusion range and output intensity, the safe geometric dimension parameters are used as target indexes to reverse match the corresponding functional component driving parameters in the correlation model, and the functional component driving parameters are defined as the extreme operation intensity value. The extreme work intensity value is converted into a corresponding hardware drive level signal, a frequency conversion adjustment command is generated, and then sent to the associated robot execution unit.

9. A multi-robot collaborative cleaning scheduling method for photovoltaic power plants according to claim 8, characterized in that, Based on the maximum permissible overlap boundary constraint, a geometric shrinkage deduction is performed on the diffusion elliptic domain model data to obtain the safe geometric dimension parameters after shrinkage. The specific operation logic is as follows: Extract the relative position vector from the expected encounter spacetime coordinate set, input it as a fixed invariant into the geometric collision simulation space, and set the maximum allowable overlap boundary constraint as the critical decision value for simulation termination; Apply proportionally decreasing scaling to the major axis radius data and minor axis radius data in the diffusion elliptic domain model data to generate multiple sets of continuously changing intermediate geometric model data; The theoretical intersection area of ​​each group of intermediate geometric model data under the relative position vector is calculated sequentially. When the theoretical intersection area is found to be lower than the maximum allowable overlap boundary constraint, the decreasing scaling process is stopped immediately. Extract the specific major axis and minor axis values ​​corresponding to the intermediate state geometric model data at the stopping moment, and combine and encapsulate them into safe geometric dimension parameters.

10. A multi-robot collaborative cleaning scheduling system suitable for photovoltaic power plants, characterized in that, A multi-robot collaborative cleaning scheduling method for photovoltaic power plants, as described in any one of claims 1-9, includes: Diffusion model construction module: Collect real-time operating status data and environmental flow field vector data of the robot in the whole field of photovoltaic power station, extract the operation jet vector from the real-time operating status data and perform vector superposition operation with the environmental flow field vector data to construct diffusion elliptical domain model data with the coordinates of each robot as the focus; Interference coefficient calculation module: Based on the travel speed vector in the real-time running status data, the module performs spatiotemporal trajectory extrapolation of multiple robots, calculates the expected encounter spatiotemporal coordinate set, maps the diffusion elliptical domain model data to the expected encounter spatiotemporal coordinate set, and generates the interference coupling coefficient by calculating the overlapping area of ​​the elliptical domains of the robots moving towards each other. Drift speed generation module: compares the interference coupling coefficient with the preset safety threshold. When the interference coupling coefficient exceeds the preset safety threshold, it retrieves the safety node coordinate data in the pre-stored photovoltaic module layout map and calculates the drift speed variable of the meeting point corresponding to the expected encounter time-space coordinate set translated to the safety node coordinate data. Drift speed change judgment module: Determines whether the drift speed variable at the meeting point is within the preset allowable speed change range. If so, it outputs a differential drift control command containing the drift speed variable at the meeting point to the associated robot. If the drift speed variable at the meeting point exceeds the preset allowable speed change range, it performs inverse solution on the diffusion elliptical domain model data based on the preset safety threshold, calculates the extreme operation intensity value that satisfies the non-interference condition, and generates a frequency conversion adjustment command. At the same time, it generates virtual uncleaned area marker data based on the extreme operation intensity value and writes the virtual uncleaned area marker data into the global task scheduling queue for subsequent path planning calls.

Images