Posture control method based on photovoltaic cleaning robot

Abstract

The invention discloses a posture control method based on a photovoltaic cleaning robot, and belongs to the technical field of photovoltaic cleaning robots, and the method comprises the steps: obtaining a difference value between a to-be-cleaned photovoltaic panel and a current photovoltaic panel, setting a preliminary adjustment method, and calculating a preliminary torsion angle; a preset cleaning path and a robot kinematic model are obtained, an analog simulation algorithm is set, the preliminary torsion angle is optimized, and an actual torsion angle applied to actual cleaning is generated; the photovoltaic cleaning robot is driven to be adjusted to the actual torsion angle, cleaning operation is conducted on the photovoltaic panel to be cleaned, after cleaning is completed, a blind area compensation method is set, a blind area compensation angle is calculated, blind area dirt is judged, the specific cleaning posture is determined, and compensation cleaning is conducted; the cleaning data of the to-be-cleaned photovoltaic panel are recorded, the adjustment optimization method is set, the adjustment coefficient used for calculating the torsion angle is optimized, the method is applied to calculation of the next torsion angle, and the overall cleaning effect is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of photovoltaic cleaning robots and relates to a posture control method based on photovoltaic cleaning robots. Background Art

[0002] Under the background of actively promoting the development of clean energy globally, as an important form of renewable energy utilization, photovoltaic power generation has developed rapidly in recent years. The scale of photovoltaic power stations continues to expand, and the number of both large-scale ground-mounted centralized photovoltaic power stations and widely distributed rooftop photovoltaic systems in cities is increasing continuously. However, the presence of dust, dirt, and obstacles on the surface of photovoltaic panels seriously affects the power generation efficiency of photovoltaic power stations.

[0003] The patent application with the publication number CN117289696A discloses a system and method for balancing adjustment of a photovoltaic cleaning robot based on attitude information. The method includes the following steps: providing a torsion-type photovoltaic cleaning robot, including a robot main body, a cleaning device, and an attitude adjustment system; in the attitude adjustment system, using a 9-axis attitude sensor to detect the attitude and motion state of the robot; receiving and processing the data of the attitude sensor through a control circuit to perform the calculation of balance adjustment; according to the calculation result, performing balance adjustment on the robot main body through an actuator to maintain the stability of the robot during the cleaning process. This technical solution uses a 9-axis attitude sensor to monitor the attitude and motion state of the robot in real time, realizes balance adjustment, and ensures the stability and safety of the robot during the cleaning process.

[0004] Although the existing technology enables the robot to adapt to different terrains and inclinations through autonomous balance adjustment, improving the cleaning efficiency and coverage, when adjusting the cleaning posture, there is a lack of precise control over the torsion angle, speed, and force, and the cleaning feedback after the torsion operation is not considered, resulting in limitations in the cleaning effect and the stability of the robot. In particular, it is impossible to evaluate the cleaning effect after the torsion adjustment in real time, and it is impossible to determine whether the cleaning device is completely attached to the surface of the photovoltaic panel or whether there are cleaning blind spots. For example, during the cleaning process, if there are local depressions or protrusions on the surface of the photovoltaic panel, the robot may not be able to perceive and adjust the cleaning posture in time, resulting in these areas not being thoroughly cleaned. Therefore, the present application provides a posture control method based on photovoltaic cleaning robots. Summary of the Invention

[0005] Aiming at the deficiencies of the existing technology, the purpose of the present invention is to provide an attitude control method for a photovoltaic cleaning robot. By real-time sensing the terrain changes on the surface of the photovoltaic panel, dynamically adjusting the torsion angle, speed and force, ensuring the complete fit of the cleaning device with the surface of the photovoltaic panel, and based on the fusion of vision, pressure and temperature data, comprehensively evaluating the cleaning effect, avoiding cleaning blind spots, establishing a dynamic feedback mechanism, and real-time adjusting the cleaning strategy to ensure the optimization of the cleaning effect.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] An attitude control method for a photovoltaic cleaning robot, which is applied to a photovoltaic cleaning robot. An upper end component and a lower end component are provided on the photovoltaic cleaning robot, and a rotating shaft is provided on the upper end component and the lower end component. The attitude control method for the photovoltaic cleaning robot includes:

[0008] Obtain the difference between the photovoltaic panel to be cleaned and the current photovoltaic panel, set a preliminary adjustment method, and calculate the preliminary torsion angle;

[0009] Obtain the preset cleaning path and the robot kinematic model, set a simulation algorithm, optimize the preliminary torsion angle, and generate an actual torsion angle applied to actual cleaning;

[0010] Drive the photovoltaic cleaning robot to adjust to the actual torsion angle, perform a cleaning operation on the photovoltaic panel to be cleaned. After the cleaning is completed, set a blind spot compensation method, calculate the blind spot compensation angle, and judge the dirt in the blind spot, determine the cleaning attitude, and perform compensatory cleaning;

[0011] Record the cleaning data of the photovoltaic panel to be cleaned, set an adjustment and optimization method, optimize the adjustment coefficient used to calculate the torsion angle, and apply the optimized adjustment coefficient to the calculation of the next torsion angle.

[0012] Furthermore, when calculating the preliminary torsion angle, the preliminary adjustment method includes:

[0013] Obtain the relative angle difference and height difference between the photovoltaic panel to be cleaned and the current photovoltaic panel, and calculate the angle to be adjusted of the lower end component ;

[0014] Set the theoretical adjustable threshold of the lower end component to , and judge whether the angle to be adjusted is executable;

[0015] If , the angle to be adjusted is executable, and set the torsion angle of the lower end component to the angle to be adjusted; if , the angle to be adjusted is not executable, and set the torsion angle of the lower end component to the theoretical adjustable threshold;

[0016] Obtain the direction vector of the upper end component and the expected direction vector of the lower end component, and calculate the relative torsion angle between the upper and lower end components; wherein, the expected direction vector of the lower end component is the direction vector after the torsion angle of the lower end component is adjusted.

[0017] Generate a preliminary torsion angle, including the torsion angle of the lower end component and the relative torsion angle between the upper and lower end components.

[0018] Furthermore, when optimizing the preliminary torsion angle, the simulation algorithm includes:

[0019] Construct a comprehensive simulation model, and input the preliminary torsion angle and the cleaning parameter values into the comprehensive simulation model;

[0020] Set initial parameters, and monitor the running state of the photovoltaic cleaning robot in real time for problem detection;

[0021] Set the adjustment step size to , once the determination result of the problem detection is that there is a problem, immediately update the torsion angle;

[0022] Input the updated torsion angle into the comprehensive simulation model for iterative calculation until stable operation;

[0023] Output the optimized torsion angle and define it as the actual torsion angle.

[0024] Furthermore, the cleaning parameters include cleaning speed and cleaning intensity;

[0025] Set the basic cleaning intensity, basic cleaning speed and basic temperature, and obtain the real-time cleaning intensity, real-time cleaning speed of the photovoltaic cleaning robot and the real-time temperature of the current environment;

[0026] Calculate the cleaning intensity and cleaning speed based on the obtained real-time data and adjustment coefficients.

[0027] Furthermore, when performing compensation cleaning, the blind area compensation method includes:

[0028] Obtain the round-trip time from the laser radar emitting a laser beam to receiving it, and calculate the distance between the laser radar and each point on the surface of the photovoltaic panel;

[0029] Repeat the process of emitting and receiving laser beams, obtain distance data, and construct the three-dimensional terrain information of the surface of the photovoltaic panel;

[0030] Synchronously photograph the photovoltaic panel, obtain an image and perform preprocessing;

[0031] Analyze the preprocessed image, identify the features on the surface of the photovoltaic panel, and mark the abnormal areas;

[0032] Register the three-dimensional terrain information with the abnormal areas, and perform terrain analysis on the abnormal areas to identify the cleaning blind spots;

[0033] Extract the blind spot depth of the cleaning blind spot from the three-dimensional terrain information, and calculate the blind spot compensation angle based on the actual torsion angle.

[0034] Furthermore, the blind spot compensation method further includes:

[0035] Adjust the cleaning posture according to the blind spot compensation angle and monitor it in real time;

[0036] Compare the dirt features in the cleaning blind spot with a pre-established dirt sample library to obtain the sample similarity of the image, and judge the dirt condition of the cleaning blind spot;

[0037] If the sample similarity exceeds the preset similarity threshold and the dirt sample corresponding to the sample similarity is a heavy dirt sample, it is determined that there is more dirt in the cleaning blind spot, and the cleaning intensity is increased;

[0038] Otherwise, it is determined that the dirt level in the cleaning blind spot is relatively light, and the original cleaning intensity and speed are maintained;

[0039] Clean the cleaning blind spot according to the blind spot compensation angle and cleaning intensity.

[0040] Furthermore, when optimizing the adjustment coefficient for calculating the torsion angle, the adjustment and optimization method includes:

[0041] Set a target cleaning score and calculate the score difference between the actual cleaning score and the target cleaning score ;

[0042] Set the acceptable threshold of the score difference , and judge whether the score difference is within the acceptable range;

[0043] If , the score difference is within the acceptable range, and the next cleaning is directly performed;

[0044] If , the score difference is not within the acceptable range, then optimize the adjustment coefficient;

[0045] Collect historical cleaning data and generate a training data set for training a machine learning model;

[0046] Set up the functional expression of the adjustment coefficient and the actual cleaning score, define the score difference as the objective function, calculate the gradient of the objective function with respect to the adjustment coefficient, and update the adjustment coefficient according to the gradient descent algorithm;

[0047] Repeat the process of calculating the gradient and updating the adjustment coefficient, perform multiple iterations until the objective function converges, and stop the iteration;

[0048] Apply the optimized adjustment coefficient to the parameter calculation of the next cleaning task.

[0049] Furthermore, before obtaining the difference between the photovoltaic panel to be cleaned and the current photovoltaic panel, various sensors arranged on the surface of the photovoltaic cleaning robot frame work together to continuously collect data;

[0050] And during the cleaning process, introduce the adaptive optimization of the sensor layout, dynamically adjust the position and angle of the sensors according to the terrain and dirt distribution on the surface of the photovoltaic panel, and set the time interval of the adaptive optimization to Every Perform an adaptive optimization.

[0051] Furthermore, the specific steps of the adaptive optimization include:

[0052] Based on the collected historical environmental data, obtain the terrain, dirt, and temperature distribution characteristics, and generate environmental information;

[0053] Define the environmental information as the state, the adjustment of the sensor position and angle as the action, and the data acquisition quality as the reward;

[0054] Learn based on the Q-learning algorithm, continuously iterate and update The value until the algorithm converges, and output the converged Value table;

[0055] When the algorithm converges, select the corresponding action in the Value table according to the current environmental information, and perform the position and angle adjustment operations.

[0056] Furthermore, before performing the compensation cleaning, use the visual camera, pressure sensor, and infrared sensor to collect data in real time and evaluate the cleaning effect, including:

[0057] Obtain the multi-dimensional real-time data on the surface of the photovoltaic panel;

[0058] Preprocess the image on the surface of the photovoltaic panel, divide the image into multiple regions, identify the dirt residue regions by comparing the pixel characteristics of each region before and after cleaning, and calculate the visual image score;

[0059] Calculate the pressure distribution score based on the average value and standard deviation of the pressure;

[0060] Divide the surface of the photovoltaic panel into multiple regions, and calculate the temperature uniformity score based on the average temperature and temperature range of each region;

[0061] Calculate the actual cleaning score based on the visual image score, pressure distribution score and temperature uniformity score.

[0062] Advantages of the present invention:

[0063] By adaptively optimizing the sensor layout, according to the complex terrain and dirt distribution on the surface of the photovoltaic panel, at specific time intervals, dynamically adjust the position and angle of the sensors to ensure stable and high-quality data collection in complex environments, improving the data collection efficiency and stability; at the same time, comprehensively consider multiple key factors to calculate the torsion angle and cleaning parameters, optimize through a comprehensive simulation model, monitor the motion state, correct the torsion angle according to the adjustment coefficient and step size, and generate the actual torsion angle, enabling the robot to flexibly adapt to the conditions of the photovoltaic panel, avoiding incomplete cleaning, and effectively improving the comprehensiveness and thoroughness of cleaning; through multi-dimensional calculation of the actual cleaning score and detection of cleaning blind spots, achieve precise and efficient compensatory cleaning, further improving the overall cleaning effect; after each cleaning is completed, optimize the adjustment coefficient based on the machine learning algorithm and apply it cyclically to subsequent cleaning tasks, continuously improving the cleaning performance of the robot to ensure efficient and high-quality cleaning working conditions in different photovoltaic panel environments. Description of the drawings

[0064] Figure 1 It is the overall structure diagram of the photovoltaic cleaning robot of the present invention;

[0065] Figure 2 It is the structure diagram of the upper end component of the present invention;

[0066] Figure 3 It is the flow chart of the attitude control method based on the photovoltaic cleaning robot;

[0067] Figure 4 It is the flow chart of generating the actual torsion angle of the present invention;

[0068] Figure 5 It is the flow chart of the blind area compensation method of the present invention;

[0069] Figure 6 It is the flow chart of the adjustment and optimization method of the present invention.

[0070] Reference numerals: 1, cleaning component; 2, lower end component; 3, upper end component; 4, rotating shaft. Detailed implementation manners

[0071] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. The embodiments of the present invention and the technical features in the embodiments may be combined with each other unless there is a conflict.

[0072] The photovoltaic cleaning robot used in the present invention refers to a photovoltaic cleaning device in a photovoltaic cleaning device supporting large slope walking disclosed in Chinese patent with authorization announcement number CN222515284U, and the structure of the photovoltaic cleaning device itself is not described in detail.

[0073] Example 1

[0074] refer to Figures 1 to 4 As shown, this embodiment introduces a posture control method based on a photovoltaic cleaning robot, which is applied to a photovoltaic cleaning robot. The photovoltaic cleaning robot is provided with an upper component 3 and a lower component 2, which are used to twist and adjust the angle of the cleaning component 1 on the photovoltaic cleaning robot. The upper component 3 and the lower component 2 are provided with a rotating shaft 4, and the method includes the following steps:

[0075] The photovoltaic cleaning robot is driven to clean the photovoltaic panels. Various sensors arranged on the surface of the robot frame work together to continuously collect data and pre-process the collected data. The sampling interval is set to , in order to capture the dynamic changes of the photovoltaic panel status in real time, the sensors include high-precision laser radar, high-resolution visual camera, sensitive infrared sensor and pressure sensor. The laser radar is used to accurately measure the position, boundary and surface micro-topographic changes of the photovoltaic panel. The visual camera shoots the surface of the photovoltaic panel to obtain image data for identifying the location, shape and distribution range of dirt. The infrared sensor detects abnormal surface temperature of the photovoltaic panel. The pressure sensor senses the pressure change between the cleaning component 1 and the photovoltaic panel in real time. The preprocessing includes denoising, filtering and normalization.

[0076] Combined with the preset cleaning path and the robot kinematic model, the difference between the photovoltaic panel to be cleaned and the current photovoltaic panel is obtained, a preliminary adjustment method is set, and a preliminary torsion angle is calculated, including the torsion angle of the lower component 2, the relative torsion angle between the upper component 3 and the lower component 2, and the cleaning parameter value, the cleaning parameter including the cleaning force and the cleaning speed; and the mechanical properties and motion range of each component of the robot are comprehensively considered to ensure that the torsion angle meets the cleaning requirements and does not exceed the robot's bearing capacity, a simulation algorithm is set, the preliminary torsion angle is optimized, and an actual torsion angle applied to actual cleaning is generated; wherein, the upper component 3 and the lower component 2 of the photovoltaic cleaning robot are connected by a connecting plate and can be relatively twisted to meet the torsion requirements between the connecting bridges on the photovoltaic panels with large slopes;

[0077] Drive the photovoltaic cleaning robot to adjust to the actual torsion angle, perform cleaning operations on the photovoltaic panel to be cleaned, and after the cleaning is completed, use visual cameras, pressure sensors and infrared sensors to collect data in real time to evaluate the cleaning effect. At the same time, use lidar and visual cameras to cooperate to detect the cleaning blind spots on the surface of the photovoltaic panel, set the blind spot compensation method, calculate the blind spot compensation angle, and judge the dirt in the blind spot to determine the specific cleaning posture for compensation cleaning;

[0078] After each cleaning is completed, record the cleaning data of the photovoltaic panel to be cleaned. According to the difference between the actual cleaning score and the target cleaning score, judge whether the adjustment coefficient is applicable, and set the adjustment and optimization method to optimize the adjustment coefficient used to calculate the torsion angle, and apply the optimized adjustment coefficient to the calculation of the next torsion angle to provide better parameters for the next cleaning until the cleaning task of the entire photovoltaic panel area is completed.

[0079] Furthermore, during the operation of the photovoltaic cleaning robot, the quality of the data obtained by the sensors directly affects the robot's judgment of the surface condition of the photovoltaic panel and the execution effect of the cleaning task. The layout of traditional sensors is fixed and cannot adapt to the complex and changeable surface environment of the photovoltaic panel, such as the difference in the tilt angle of the photovoltaic panel under different terrains, the randomness of the surface dirt distribution, and the possible local occlusion, thus unable to ensure stable and high-quality data collection;

[0080] By introducing the adaptive optimization of the sensor layout, dynamically adjust the position and angle of the sensors according to the complex terrain and dirt distribution on the surface of the photovoltaic panel to maximize the data collection efficiency. At the same time, set the time interval of the adaptive optimization to and perform the adaptive optimization every to ensure that the sensor layout can periodically adapt to environmental changes. Even if the environmental changes are not obvious, the high efficiency of data collection can be maintained through regular optimization. For some relatively stable photovoltaic panel areas, regular optimization can timely adjust the layout deviation of the sensors caused by vibration and slight displacement to ensure the stability of the data collection quality;

[0081] The specific steps of the adaptive optimization are as follows:

[0082] For lidar data, the Kalman filtering algorithm is used to remove noise interference, and the terrain distribution characteristics of the photovoltaic panels, such as the average height and terrain standard, are extracted based on the clustering algorithm. For visual image data, grayscale processing is first performed, and then the Canny edge detection algorithm is used to extract the edge contour of the dirt. Through image segmentation techniques (such as the Otsu algorithm), the dirt distribution characteristics, such as the dirt area ratio and the centroid position, are determined. The data of the infrared sensor and the temperature sensor are normalized to unify the dimension and range, and the temperature distribution characteristics, such as the maximum temperature and the temperature gradient, are generated. The processed terrain, dirt, and temperature distribution characteristics are integrated to construct the environmental information required for the reinforcement learning algorithm;

[0083] Define the environmental information as the state, and the adjustment of the sensor position and angle as the action, including the position translation amount of each sensor in the three-dimensional space and the rotation angles around different axes. The data acquisition quality is the reward. For example, when the clarity of the image collected by the sensor is improved (measured by the image information entropy), the data coverage range is increased (calculating the coverage rate of the collected data points on the surface of the photovoltaic panel), or the data consistency is enhanced (analyzing the difference degree of the data collected multiple times), a higher reward is given, otherwise a lower reward is given;

[0084] The Q-learning algorithm is used for learning. First, a two-dimensional table is created, where the rows represent different states and the columns represent different actions, and all values are initialized to 0 or a small random value to initialize the value table; where is the expected cumulative reward after performing action in state . The purpose of initializing the value table is to provide a starting point for the algorithm, allowing the algorithm to gradually update and optimize these values in the subsequent learning process;

[0085] At each time step , according to the current state , the greedy strategy is used to select action ; with probability a random action is selected, and with probability the action with the maximum value in the current state is selected; as the learning progresses, gradually decreases, and , where is the initial value;

[0086] Execute action to adjust the position and angle of the sensor and perform the data acquisition operation. After the acquisition is completed, observe the new state , and update the value, and calculate the maximum change amount of all elements in the value table , the expression is as follows:

[0087] ;

[0088] ;

[0089] In the formula, and are the values before and after update respectively, is the learning rate, which controls the influence degree of new information, is the immediate reward obtained after taking the action , is the discount factor, which controls the current value of future rewards, is the maximum value of all possible actions in the new state , representing the maximum expectation of future rewards; is the expected cumulative reward after executing the action in the state ; is the combination of all states and actions , calculating the absolute value of the difference between the values before and after update, and taking the maximum value among these absolute values;

[0090] Continuously repeat the above process of selecting actions, executing actions, observing results and updating the value until the algorithm converges; among them, the judgment condition for the algorithm to converge is that the maximum change amount is less than the preset convergence threshold;

[0091] After the algorithm converges, obtain the environmental information on the current photovoltaic panel surface, look up the corresponding row in the value table that has converged according to the state vector, and select the action corresponding to the column with the largest value. This action is the optimal action for sensor layout adjustment in the current environment, and includes the position translation and angle rotation operations that each sensor should perform;

[0092] Analyze the adjustment information of each sensor in the action, and drive the moving and rotating mechanisms of the sensors according to their respective adjustment instructions to perform precise position and angle adjustments;

[0093] After the sensor layout adjustment is completed, obtain the current position and angle information of each sensor, encode this information in a certain format, for example, represent the position by coordinate values and the rotation angle by angle values, and store it in the local storage device of the robot.

[0094] Further, the specific steps of the preliminary adjustment method include:

[0095] In a photovoltaic power station, there are generally differences in tilt angles, irregular arrangements, and uneven distributions of obstacles among different photovoltaic panels. These problems make it impossible for the photovoltaic cleaning robot to effectively contact the surface of the photovoltaic panel during the cleaning process if it cannot flexibly adjust its own posture, resulting in incomplete cleaning or missed areas, seriously affecting the cleaning effect and the power generation efficiency of the photovoltaic panel; According to the preset cleaning path, the robot starts to clean the photovoltaic panel. During the cleaning process, the tilt angle and height change information of the photovoltaic panel surface are obtained through sensors, and the relative angle difference and height difference between the photovoltaic panel at the next moment and the photovoltaic panel at the current moment are collected in real time using lidar and vision cameras to reflect the real-time state of the photovoltaic panel, and the angle to be adjusted of the lower component 2 is calculated. Taking into account both the height difference and the angle difference of the photovoltaic panel, the robot can accurately adjust the posture of the lower component 2 according to the actual situation to better adapt to the surface condition of the photovoltaic panel; The expression of the angle to be adjusted is as follows:

[0096] ;

[0097] In the formula, is the axle distance between the lower component 2 and the upper component 3, is the adjustment coefficient, which is used to finely adjust the angle to be adjusted according to the relative angle difference of the photovoltaic panel, and the height difference is the position difference of the photovoltaic panel in the vertical direction, caused by uneven installation of the photovoltaic panel, terrain undulation, or inclination of the connecting bridge;

[0098] Set the theoretical adjustable threshold of the upper component 3 to be and the theoretical adjustable threshold of the lower component 2 to be ; Among them, the theoretical adjustable threshold is related to the corresponding arc and preset angle of the connecting chute in the upper component 3 and the lower component 2. If the arc of the chute is , then the theoretical adjustable threshold is ;

[0099] Compare the calculated angle to be adjusted with the theoretical adjustable angle to determine whether the angle to be adjusted is executable, ensuring that the cleaning component 1 will not exceed the structural limit during the adjustment process, avoiding equipment damage, and at the same time maintaining the parallel posture of the cleaning roller brush to ensure the cleaning effect;

[0100] If , indicating that the angle to be adjusted does not exceed the maximum adjustable angle of the lower end component 2, then the angle to be adjusted can be executed. Let , at this time, the torsion angle of the lower end component 2 is ;

[0101] If , indicating that the angle to be adjusted exceeds the maximum adjustable angle of the lower end component 2, then the angle to be adjusted cannot be executed. Let , at this time, the torsion angle of the lower end component 2 is ; Among them, is the torsion angle of the lower end component 2;

[0102] Through the robot kinematic model, path planning information, and perception of the photovoltaic panel surface, the direction vector of the upper end component 3 and the expected direction vector of the lower end component 2 are determined. The relative torsion angle between the upper end component 3 and the lower end component 2 is calculated using the vector angle formula. Among them, in order to make the cleaning component in good contact with the photovoltaic panel surface, the expected direction vector of the lower end component 2 is the direction vector after the torsion angle adjustment of the chute of the lower end component 2. The expression is as follows:

[0103] ;

[0104] In the formula, is the formula for calculating the modulus value of the direction vector;

[0105] Considering the uneven dirt distribution on the photovoltaic panel surface and the differences in cleaning requirements of the robot at different positions and motion states, it is necessary to dynamically adjust the cleaning force and cleaning speed; set the basic cleaning force as , the basic cleaning speed as and the basic temperature as , and use the pressure sensor to obtain the real-time cleaning force during current cleaning, the infrared sensor to obtain the real-time temperature of the current environment, and at the same time obtain the real-time cleaning speed of the robot. Based on the obtained data, calculate the cleaning force and the cleaning speed to achieve dynamic adjustment of the cleaning force. The expression is as follows:

[0106] ;

[0107] ;

[0108] In the formula, , , , is the adjustment factor, The target pressure value is determined according to the dirt situation. The adjustment of the cleaning speed is closely related to the cleaning force. At the same time, the robot's power performance, energy consumption and the bearing capacity of the photovoltaic panel are comprehensively considered. When the cleaning force increases, the cleaning speed needs to be appropriately reduced to ensure the cleaning effect and the stability of the robot. On the contrary, in areas with less dirt, the cleaning speed can be increased to improve work efficiency.

[0109] Output the cleaning parameter value and the preliminary twisting angle, including the twisting angle of the lower end component 2 and the relative twisting angle between the upper end component 3 and the lower end component 2.

[0110] Furthermore, the specific steps of the simulation algorithm include:

[0111] In order to reproduce the motion state of the robot in the actual cleaning process in a virtual environment, various actual parameters are input to comprehensively and intuitively observe the performance of the robot under different working conditions. Computer simulation technology is used to integrate the robot kinematic model, mechanical model and photovoltaic panel environment model to construct a comprehensive simulation model. The expected cleaning path, calculated torsion angle and cleaning parameter adjustment amount are input into the comprehensive simulation model. In the kinematic model, the robot is divided into an upper component 3, a lower component 2 and a cleaning component. The motion relationship between the various components of the robot is determined according to the structural parameters of the robot (such as wheelbase, size of each component, range of motion of the joints). The mechanical model analyzes the stress and strain of each component under stress according to the material properties and structural design. The photovoltaic panel environment model covers the layout, tilt angle, row and column spacing of the photovoltaic panels and possible obstacle information.

[0112] The torsion angle when a problem occurs in the current simulation (such as unstable center of gravity, excessive force on components, or collision with the frame of the photovoltaic panel) is used as the initial parameter, and the motion state of the robot is monitored in real time. With the help of the robot kinematic model and mechanical model, key indicators such as the center of gravity position (calculated through the mass distribution and position coordinates of each component), the force on the component (such as the torque of the rotating shaft 4, the tension of the connecting component), and the distance from the frame of the photovoltaic panel are calculated to detect the problem; the problem detection monitors the key indicators of the photovoltaic cleaning robot. When the value of any of the above indicators exceeds the preset threshold, it is determined that the key indicator is abnormal (such as unstable center of gravity, excessive force on the component, too close distance to the frame of the photovoltaic panel), and the motion state of the photovoltaic cleaning robot has a problem.

[0113] Set the adjustment step size to , once a problem occurs, the torsion angle is updated immediately. The expression is as follows:

[0114] ;

[0115] ;

[0116] In the formula, is an adjustment coefficient, which is determined according to the robot structure and the center of gravity offset, and can be obtained through multiple simulation tests; is the torsion angle of the adjusted lower end component 2, is the relative torsion angle between the adjusted upper end component 3 and the lower end component 2;

[0117] Input the adjusted torsion angles and into the comprehensive simulation model, re-simulate the motion state of the robot during the entire cleaning process, and perform iterative calculations. When the simulation results meet the requirements that the robot operates stably during the entire cleaning process, and the mechanical properties and motion ranges of all components are within the safe range, that is, the center of gravity is within the support range, the force on the components is within the allowable range, and the distance from the photovoltaic panel frame is greater than the safety threshold, then stop the iteration; among them, when the adjustment effect is not good for several consecutive times, reduce the adjustment step size, let , continue to adjust the torsion angle parameters, and re-perform the simulation;

[0118] Output the optimized torsion angles, including the torsion angle of the lower end component 2 that meets the requirements and the relative torsion angle between the upper end component 3 and the lower end component 2 , and define them as the actual torsion angles, which are applied to subsequent actual cleaning.

[0119] Furthermore, the specific steps for evaluating the cleaning effect include:

[0120] Use visual cameras, pressure sensors, and infrared sensors to start real-time data collection, obtain multi-dimensional real-time data on the cleaning state of the photovoltaic panel surface, and provide rich data support for the cleaning effect evaluation model to more accurately evaluate the cleaning effect; among them, the visual camera continuously takes pictures of the photovoltaic panel surface and records the state changes of the photovoltaic panel surface before and after cleaning; the pressure sensor measures the pressure distribution when the cleaning roller brush contacts the photovoltaic panel surface and obtains the pressure data in different areas; the infrared sensor monitors the temperature change on the photovoltaic panel surface and collects temperature data;

[0121] Fuse the data collected by the visual camera, pressure sensor, and infrared sensor, and calculate the actual cleaning score , and the expression is as follows:

[0122] ;

[0123] In the formula, is the visual image score, which analyzes the pictures of the photovoltaic panel surface taken by the image recognition algorithm, identifies indicators such as dirt residue degree and cleaning coverage rate, and performs quantitative scoring, It is the pressure distribution score. According to the pressure data collected by the pressure sensor, the uniformity and stability of the pressure distribution are analyzed for scoring. It is the temperature uniformity score. Based on the temperature data obtained by the infrared sensor, the uniformity of the surface temperature of the photovoltaic panel is evaluated for scoring. , , is the weight coefficient, and ;

[0124] Furthermore, using image recognition technology, the surface image of the photovoltaic panel captured by the vision camera is preprocessed. The image is divided into multiple small regions. By comparing the pixel features of each region before and after cleaning, the dirt residue regions are identified and quantitatively scored. Let the area of the dirt residue region in the image be , the area of the entire photovoltaic panel image be , and the dirt type influence coefficient be . Calculate the visual image score . The expression is as follows:

[0125] ;

[0126] In the formula, represents the dirt coverage ratio, is used to adjust the influence weight of different types of dirt, The scoring range of is between 0 and 1, where 1 represents completely clean and 0 represents completely unclean;

[0127] Obtain the pressure data of the pressure sensor at different positions, and calculate the average value of the pressure and the standard deviation . The better the pressure distribution uniformity, the smaller the standard deviation, indicating that the cleaning roller brush contacts the surface of the photovoltaic panel more evenly, and the cleaning effect may be better; and calculate the pressure distribution score . The expression is as follows:

[0128] ;

[0129] In the formula, reflects the dispersion degree of the pressure distribution, the smaller, the closer to 1, indicating that the pressure distribution is more uniform and the cleaning effect is better; conversely, the larger, the closer to 0, indicating that the pressure distribution is more uneven and there are areas where cleaning is insufficient;

[0130] Use the infrared sensor to obtain the temperature data of the surface of the photovoltaic panel. Divide the surface of the photovoltaic panel into multiple regions, and calculate the average temperature and the temperature range , is the difference between the highest temperature and the lowest temperature. The temperature uniformity score mainly measures the consistency of the surface temperature of the photovoltaic panel. The smaller the temperature range, the better the temperature uniformity; and calculate the temperature uniformity score , and the expression is as follows:

[0131] ;

[0132] In the formula, is the relative change degree of temperature, the smaller, the closer to 1, indicating that the surface temperature uniformity of the photovoltaic panel is better, and there is no temperature anomaly caused by excessive local friction or other abnormal conditions during the cleaning process. On the contrary, the larger, the closer to 0, indicating that the temperature uniformity is poor and there are cleaning problems.

[0133] Furthermore, referring to Figure 5 , the specific steps of the blind area compensation method include:

[0134] During the cleaning process of the photovoltaic cleaning robot, the lidar continuously works and emits laser beams at a preset frequency. After the laser beams encounter the surface of the photovoltaic panel, they are reflected back, and the lidar receives the reflected signals; since the propagation speed of light in the air is constant, by accurately measuring the round-trip time of the laser from emission to reception , calculate the distances between the lidar and each point on the surface of the photovoltaic panel , and the expression is as follows:

[0135] ;

[0136] In the formula, is the speed of light;

[0137] Continuously repeat the process of emitting and receiving laser beams. As the number of measurements increases, a large amount of distance data is obtained. Taking the position of the lidar as the origin, a global coordinate system is constructed. Through the positioning sensor equipped on the lidar, the position and attitude information of the robot in the global coordinate system are obtained, and the relative coordinates (distances) measured by the lidar are converted into absolute coordinates. These absolute coordinates are integrated to form point cloud data, and interpolation and fitting are performed on the point cloud data. Using 3D modeling software, the point cloud data is converted into a 3D model, and the 3D terrain information of the surface of the photovoltaic panel is constructed and presented in the form of point cloud data;

[0138] Use the vision camera to synchronously take pictures of the surface of the photovoltaic panel to obtain a series of images and preprocess the images;

[0139] Use a deep learning-based object detection algorithm to analyze the preprocessed image, identify various features on the surface of the photovoltaic panel, compare each region in the image with the pre-set normal photovoltaic panel surface features, and mark the regions that do not conform to the normal photovoltaic panel surface features as abnormal regions, such as parts with significant differences in color and texture from the normal regions; among them, the object detection algorithm learns through a large number of labeled photovoltaic panel surface images (including normal regions and various abnormal regions), so as to have the ability to identify the features of different regions; the visual camera can capture rich image details on the surface of the photovoltaic panel, and image recognition technology can use these details to quickly detect abnormal conditions on the surface. Compared with other detection methods, it can intuitively judge the abnormalities on the surface and provide more intuitive information for the detection of cleaning blind spots;

[0140] In order to effectively fuse the three-dimensional terrain information obtained by the lidar and the abnormal regions identified by the visual camera, register the point cloud data of the lidar with the visual image. By finding common feature points in the point cloud data and the visual image (such as the edges of the photovoltaic panel, specific marker points), use a spatial transformation algorithm (such as the ICP algorithm) to make them correspond in spatial position;

[0141] After completing the data registration, combine the information of the lidar and the visual camera to perform terrain analysis on the abnormal regions marked in the visual image. When there are local depressions or protrusions in the abnormal regions in the visual image and the depth or height exceeds a certain threshold, it is determined that the abnormal region is a cleaning blind spot;

[0142] Obtain the actual torsion angle, and extract the blind spot depth of the cleaning blind spot from the point cloud data measured by the lidar , and calculate the blind spot compensation angle. The expression is as follows:

[0143] ;

[0144] ;

[0145] In the formula, , are compensation coefficients, which are determined by comprehensively analyzing the data of simulation tests and actual tests; the blind spot depth refers to the height difference between the blind spot calculated from the point cloud data and the normal photovoltaic panel surface, and is obtained by analyzing and calculating the point cloud coordinates of the blind spot and the normal region; is the compensation angle of the lower component 2, is the relative compensation angle between the upper component 3 and the lower component 2;

[0146] According to the calculated blind area compensation angle, send instructions to adjust its own cleaning posture. During the adjustment process, monitor the posture change of the robot in real time to ensure that the cleaning roller brush can accurately move towards the cleaning blind area, making it closer to or covering the cleaning blind area;

[0147] When compensating and cleaning the cleaning blind area, compare the dirt characteristics in the cleaning blind area with the pre-established dirt sample library. By calculating the similarity between the current image and the images in the sample library, sort the calculated series of similarity values from largest to smallest, define the largest similarity value as the sample similarity of the image, and judge the dirt situation in the cleaning blind area; among them, the dirt sample library contains image data of different types and degrees of dirt and their corresponding characteristic parameters;

[0148] If the sample similarity exceeds the preset similarity threshold and the dirt sample corresponding to the sample similarity is a severe dirt sample, it is determined that there is more dirt in the cleaning blind area and the cleaning intensity is increased;

[0149] Otherwise, it is determined that the dirt degree in the cleaning blind area is relatively light, and the original cleaning intensity and speed are maintained;

[0150] Clean the cleaning blind area according to the adjusted cleaning posture, blind area compensation angle and cleaning intensity.

[0151] Furthermore, referring to Figure 6 , the specific steps of the adjustment and optimization method include:

[0152] Taking into comprehensive consideration the power generation efficiency requirements of the photovoltaic panel, the cleaning industry standards and past cleaning experience, set the target cleaning score as , which reflects the cleaning effect standard that is expected to be achieved by the photovoltaic cleaning robot after each cleaning, and calculate the score difference between the actual cleaning score and the target cleaning score , which reflects the deviation degree between the actual cleaning effect and the expected effect. The expression is as follows:

[0153] ;

[0154] Set the acceptable threshold of the score difference as , which is used to judge whether the gap between the current cleaning effect and the target effect is within the acceptable range;

[0155] Compare the calculated score difference with the acceptable threshold ; if , the current cleaning effect is close to the target cleaning effect, and the currently used adjustment coefficient can meet the cleaning requirements to a certain extent and is more applicable, and directly proceed to the next cleaning; if , it indicates that there is a large gap between the cleaning effect and the target, and the adjustment coefficient needs to be optimized and proceed to the next step to optimize the adjustment coefficient; among them, the adjustment coefficient includes , , , , , ; represents the absolute value of the scoring difference ;

[0156] Collect historical cleaning data, including the actual cleaning score, the target cleaning score, and the corresponding adjustment coefficient for each cleaning, and organize the historical cleaning data into a training dataset , for training a machine learning model; among them, , is the cleaning dataset for the th cleaning in the historical cleaning data, , are the actual and target cleaning scores for the th cleaning respectively, is the th coefficient in the th cleaning, , ;

[0157] For the adjustment coefficient , set the function expression of the adjustment coefficient and the actual cleaning score as , and the specific form of the function expression is determined by analyzing the historical data and training the machine learning model. Define the scoring difference as the objective function, and calculate the gradient of the objective function with respect to the adjustment coefficient according to the calculus principle. According to the gradient descent algorithm, update the adjustment coefficient, and the expression is as follows:

[0158] ;

[0159] ;

[0160] In the formula, is the learning rate; is the updated adjustment coefficient, is the adjustment coefficient before update;

[0161] Repeat the process of calculating the gradient and updating the adjustment coefficient, perform multiple iterations, and adjust the adjustment coefficient in the direction of reducing the objective function for each iteration until the scoring difference is less than the set convergence threshold, and stop the iteration;

[0162] Apply the optimized adjustment coefficient to the parameter calculation of the next cleaning task.

[0163] In summary, in the embodiments of the present invention, the robot is driven to clean, multiple sensors cooperate to collect and preprocess data, the data quality is improved by adaptive optimization layout, combined with the cleaning path and the kinematic model, the torsion angle and cleaning parameters are calculated, the torsion angle is iteratively optimized through simulation, and the actual torsion angle is generated; during cleaning, the data of multiple sensors are collected in real time, the actual cleaning score is evaluated through multi-dimensional calculation, the blind area is detected and compensated for cleaning by the cooperation of the lidar and the vision camera. After each cleaning, the actual cleaning score is compared with the target cleaning score, the adjustment coefficient is dynamically optimized based on machine learning, a closed-loop feedback mechanism is formed, and through continuous cycling and iterative operations, the cleaning performance of the robot is continuously improved, and the cleaning task of the entire photovoltaic panel area is completed, ensuring that the robot can accurately and efficiently adapt to the complex photovoltaic panel environment and complete the high-quality cleaning task.

[0164] The above are only the preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the idea of the present invention belong to the protection scope of the present invention. It should be pointed out that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A posture control method for a photovoltaic cleaning robot, which is applied to the photovoltaic cleaning robot. An upper end assembly (3) and a lower end assembly (2) are provided on the photovoltaic cleaning robot, and a rotating shaft (4) is provided on the upper end assembly (3) and the lower end assembly (2), characterized in that, Including: Obtain the difference between the photovoltaic panel to be cleaned and the current photovoltaic panel, set a preliminary adjustment method, and calculate the preliminary torsion angle; Obtain the preset cleaning path and the robot kinematic model, set a simulation algorithm, optimize the preliminary torsion angle, and generate the actual torsion angle applied to actual cleaning; Drive the photovoltaic cleaning robot to adjust to the actual torsion angle, perform a cleaning operation on the photovoltaic panel to be cleaned. After the cleaning is completed, set a blind area compensation method, calculate the blind area compensation angle, judge the dirt in the blind area, determine the cleaning posture, and perform compensation cleaning; Record the cleaning data of the photovoltaic panel to be cleaned, set an adjustment and optimization method, optimize the adjustment coefficient used to calculate the torsion angle, and apply the optimized adjustment coefficient to the calculation of the next torsion angle.

2. The attitude control method based on a photovoltaic cleaning robot according to claim 1, characterized in that: When calculating the preliminary torsion angle, the preliminary adjustment method includes: Obtain the relative angular difference and height difference between the photovoltaic panel to be cleaned and the current photovoltaic panel, and calculate the angle to be adjusted for the lower component (2) ; Set the theoretical adjustable threshold of the lower end component (2) to be , and determine whether the angle to be adjusted is executable; If , if the angle to be adjusted is executable, set the torsion angle of the lower end component (2) to the angle to be adjusted; if , if the angle to be adjusted is not executable, set the torsion angle of the lower end component (2) to the theoretical adjustable threshold; Obtain the direction vector of the upper end component (3) and the expected direction vector of the lower end component (2), and calculate the relative torsion angle between the upper end component (3) and the lower end component (2); wherein, the expected direction vector of the lower end component (2) is the direction vector after the torsion angle adjustment of the lower end component (2); Generate a preliminary torsion angle, including the torsion angle of the lower end component (2), and the relative torsion angle between the upper end component (3) and the lower end component (2).

3. The attitude control method based on a photovoltaic cleaning robot according to claim 2, characterized in that: When optimizing the preliminary torsion angle, the simulation algorithm includes: Construct a comprehensive simulation model, and input the preliminary torsion angle and the cleaning parameter values into the comprehensive simulation model; Set initial parameters, and real-time monitor the operating state of the photovoltaic cleaning robot to perform problem detection; Set the adjustment step size to , and once the determination result of problem detection is that a problem occurs, immediately update the torsion angle; Input the updated torsion angle into the comprehensive simulation model for iterative operation until stable operation; Output the optimized torsion angle and define it as the actual torsion angle.

4. The attitude control method based on a photovoltaic cleaning robot according to claim 3, characterized in that: The cleaning parameters include cleaning speed and cleaning intensity; Set a basic cleaning intensity, a basic cleaning speed, and a basic temperature, and obtain the real-time cleaning intensity, the real-time cleaning speed of the photovoltaic cleaning robot, and the real-time temperature of the current environment; Based on the obtained real-time data and the adjustment coefficient, calculate the cleaning intensity and the cleaning speed.

5. The attitude control method based on a photovoltaic cleaning robot according to claim 4, characterized in that: When performing compensation cleaning, the blind area compensation method includes: Obtain the round-trip time from the laser radar emitting a laser beam to receiving it, and calculate the distance between the laser radar and each point on the surface of the photovoltaic panel; Repeat the process of emitting and receiving the laser beam, obtain distance data, and construct the three-dimensional terrain information of the surface of the photovoltaic panel; Synchronously photograph the photovoltaic panel, obtain an image and perform preprocessing; Analyze the preprocessed image, identify the features on the surface of the photovoltaic panel, and mark the abnormal areas; Register the three-dimensional terrain information with the abnormal area, perform terrain analysis on the abnormal area, and identify cleaning blind spots; Extract the blind spot depth of the cleaning blind spot from the three-dimensional terrain information, and calculate the blind spot compensation angle based on the actual torsion angle.

6. The attitude control method of the photovoltaic cleaning robot according to claim 5, characterized in that The blind spot compensation method further includes: Adjust the cleaning posture according to the blind spot compensation angle and monitor in real time; Compare the dirt characteristics in the cleaning blind spot with a pre-established dirt sample library to obtain the sample similarity of the image, and judge the dirt condition in the cleaning blind spot; If the sample similarity exceeds a preset similarity threshold and the dirt sample corresponding to the sample similarity is a heavy dirt sample, it is determined that there is more dirt in the cleaning blind spot, and the cleaning intensity is increased; Otherwise, it is determined that the dirt degree in the cleaning blind spot is relatively light, and the original cleaning intensity and speed are maintained; Clean the cleaning blind spot according to the blind spot compensation angle and cleaning intensity.

7. The attitude control method based on a photovoltaic cleaning robot according to claim 6, wherein: When optimizing the adjustment coefficient for calculating the torsion angle, the adjustment and optimization method includes: Set a target cleaning score and calculate the score difference between the actual cleaning score and the target cleaning score ; Set an acceptable threshold for the scoring difference , and determine whether the scoring difference is within the acceptable range; If and the scoring difference is within an acceptable range, directly proceed to the next cleaning; If , and the scoring difference is not within the acceptable range, then optimize the adjustment coefficient; Collect historical cleaning data and generate a training data set for training a machine learning model; Set the functional expression of the adjustment coefficient and the actual cleaning score, define the score difference as the objective function, calculate the gradient of the objective function with respect to the adjustment coefficient, and update the adjustment coefficient according to the gradient descent algorithm; Repeat the process of calculating the gradient and updating the adjustment coefficient, perform multiple iterations until the objective function converges, and stop the iteration; Apply the optimized adjustment coefficient to the parameter calculation of the next cleaning task.

8. The attitude control method based on a photovoltaic cleaning robot according to claim 7, wherein: Before obtaining the difference between the photovoltaic panel to be cleaned and the current photovoltaic panel, various sensors arranged on the surface of the photovoltaic cleaning robot frame work together to continuously collect data; During the cleaning process, introduce adaptive optimization of the sensor layout, dynamically adjust the position and angle of the sensors according to the terrain and dirt distribution on the surface of the photovoltaic panel, and set the time interval for adaptive optimization to be , every Perform an adaptive optimization once.

9. The attitude control method for a photovoltaic cleaning robot according to claim 8, characterized in that The specific steps of adaptive optimization include: Based on the collected historical environmental data, obtain the terrain, dirt, and temperature distribution characteristics, and generate environmental information; Define the environmental information as the state, the sensor position and angle adjustment as the action, and the data acquisition quality as the reward; Learn based on the Q - learning algorithm, continuously iterate and update the value until the algorithm converges, and output the converged value table; After the algorithm converges, select the corresponding action in the value table according to the current environmental information, and perform operations to adjust the position and angle.

10. The attitude control method based on a photovoltaic cleaning robot according to claim 9, wherein: Before performing compensatory cleaning, use a vision camera, a pressure sensor, and an infrared sensor to collect data in real time and evaluate the cleaning effect, including: Obtain the multi-dimensional real-time data on the surface of the photovoltaic panel; Preprocess the image on the surface of the photovoltaic panel, divide the image into multiple regions, identify the dirt residue regions by comparing the pixel characteristics of each region before and after cleaning, and calculate the visual image score; Calculate the pressure distribution score based on the average value and standard deviation of the pressure; Divide the surface of the photovoltaic panel into multiple regions, and calculate the temperature uniformity score based on the average temperature value and temperature range of each region; Calculate the actual cleaning score based on the visual image score, the pressure distribution score, and the temperature uniformity score.

Images