Window cleaning robot cleaning execution system and method based on dynamic control

Abstract

The invention relates to the technical field of cleaning control, and discloses a window-cleaning robot cleaning execution system and method based on dynamic control, and the method comprises the steps: calculating the stain density, and setting a target pressure and a target rotating speed according to the stain density; calculating a dynamic friction coefficient based on the stain density, and multiplying the dynamic friction coefficient by the window surface pressure and the radius of the cleaning cloth to obtain a friction torque; real pressure is obtained, and a pressure sliding mode surface is constructed according to the target pressure and the real pressure; adding the linear terms of the pressure sliding mode surface and the pressure sliding mode surface to obtain a pressure reaching law, and adding the linear terms of the rotation speed sliding mode surface and the rotation speed sliding mode surface to obtain a rotation speed reaching law; the air cylinder duty ratio output value is calculated according to the pressure reaching law and the centrifugal force compensation item, the motor voltage output value is calculated according to the rotating speed reaching law in combination with the friction torque, then synchronous and rapid tracking of the pressure and the rotating speed under the working condition that the stain density changes dramatically is achieved, and the stable and efficient cleaning execution capacity of the window cleaning robot is ensured.

Description

Technical Field

[0001] This application relates to the field of cleaning control technology, and in particular to a window cleaning robot cleaning execution system and method based on dynamic control. Background Technology

[0002] Window cleaning robots face complex conditions with uneven distribution of dirt density on window surfaces during high-rise building cleaning operations. Traditional cleaning methods use fixed pressure and rotation speed parameters, resulting in insufficient cleaning of heavily soiled areas and excessive pressure in lightly soiled areas, leading to energy waste and accelerated equipment wear. During cleaning, there is a coupled constraint relationship between the contact pressure of the cleaning cloth and the rotation speed. Increased pressure passively reduces the rotation speed due to increased frictional damping, while increased rotation speed weakens the actual pressure due to centrifugal effect. Pressure sensors are affected by factors such as temperature drift and cumulative wear of the cleaning cloth during long-term operation, resulting in measurement bias. This causes the control system to make adjustments based on erroneous feedback signals. Although traditional linear sliding mode control is robust, it has a slow convergence speed and suffers from high-frequency chattering, making it difficult to meet the dynamic response requirements under conditions of rapidly changing dirt density. Summary of the Invention

[0003] This application provides a window cleaning robot cleaning execution system and method based on dynamic control, which realizes synchronous and rapid tracking of pressure and rotation speed under conditions of drastic changes in dirt density, ensuring the stable and efficient cleaning execution capability of the window cleaning robot.

[0004] The first aspect of this application provides a window cleaning robot cleaning execution method based on dynamic control, the window cleaning robot cleaning execution method based on dynamic control includes: Calculate the stain density, and set the target pressure and target rotation speed based on the stain density; The dynamic friction coefficient is calculated based on the stain density, and the friction torque is obtained by multiplying the dynamic friction coefficient by the window surface pressure and the radius of the cleaning cloth. Obtain the actual pressure, and construct a pressure sliding surface based on the target pressure and the actual pressure; Adding the pressure sliding surface to the linear term of the pressure sliding surface yields the pressure approaching law; adding the speed sliding surface to the linear term of the speed sliding surface yields the speed approaching law. The cylinder duty cycle output value is calculated based on the pressure approach law and the centrifugal force compensation term, and the motor voltage output value is calculated based on the speed approach law and the friction torque.

[0005] In conjunction with the first aspect, in a first implementation of the first aspect of the present invention, the step of calculating the stain density and setting the target pressure and target rotation speed based on the stain density includes: The window image is divided into multiple grids, and the mean value of grayscale is calculated based on the red, green and blue channel values ​​of all pixels in a single grid. The stain density is obtained by subtracting the mean gray value from the maximum gray value and then dividing by the maximum gray value. The stain density is compared with a first stain density threshold and a second stain density threshold; When the stain density is greater than the second stain density threshold, the target pressure is set to the first target pressure value and the target rotation speed is set to the first target rotation speed value. When the stain density is less than the first stain density threshold, the target pressure is set to the second target pressure value and the target rotation speed is set to the second target rotation speed value. When the stain density is within the range of the first stain density threshold to the second stain density threshold, the target pressure is obtained by adding the second target pressure value to the first pressure interpolation coefficient and multiplying it by the stain density minus the first stain density threshold, and the target speed is obtained by adding the second target speed value to the first speed interpolation coefficient and multiplying it by the stain density minus the first stain density threshold.

[0006] In conjunction with the first aspect, in a second implementation of the first aspect of the present invention, the step of calculating the dynamic friction coefficient based on the stain density and multiplying the dynamic friction coefficient by the window surface pressure and the radius of the cleaning cloth to obtain the friction torque includes: Multiply the stain density by the first mapping coefficient and subtract the first bias parameter to obtain the density calculation result. Substitute the density calculation result into the hyperbolic tangent function to obtain the variable component of the friction coefficient. The dynamic friction coefficient is obtained by adding the product of the basic friction coefficient and the variable component of the friction coefficient multiplied by the second mapping coefficient. The friction torque is obtained by multiplying the dynamic friction coefficient by the window surface pressure and the radius of the cleaning cloth in sequence.

[0007] In conjunction with the first aspect, in a third implementation of the first aspect of the present invention, the step of obtaining the actual pressure and constructing a pressure sliding surface based on the target pressure and the actual pressure includes: The temperature drift compensation amount is obtained by subtracting the standard temperature value from the real-time temperature value and then multiplying the result by the temperature drift coefficient. The wear compensation amount is obtained by integrating the square of the rotational speed over time and multiplying it by the wear coefficient. The true pressure is obtained by successively subtracting the temperature drift compensation, the speed square time integral wear compensation, and the zero bias compensation from the sensor pressure measurement value; A pressure error is constructed based on the target pressure and the actual pressure, and a pressure sliding surface is created based on the pressure error.

[0008] In conjunction with the first aspect, in a fourth implementation of the first aspect of the present invention, the step of constructing a pressure error based on the target pressure and the actual pressure, and creating a pressure sliding surface based on the pressure error, includes: The pressure error is obtained by subtracting the actual pressure from the target pressure. The absolute value of the pressure error is subjected to a first power-law operation and a second power-law integration operation to obtain the pressure power term and the pressure integral term. The pressure error, the pressure power term, and the pressure integral term are added together to obtain the pressure sliding surface.

[0009] In conjunction with the first aspect, in a fifth implementation of the first aspect of the present invention, the step of performing a first power-law operation and a second power-law integration operation on the absolute value of the pressure error to obtain a pressure power term and a pressure integral term includes: The absolute value of the pressure error is subjected to a first power of convergence, multiplied by a first sign function, and then multiplied by the ratio of the first sliding mode gain to the first convergence exponent to obtain the pressure power term. The absolute value of the pressure error is multiplied by a second convergence exponent and then by a second sign function to obtain the pressure calculation result. The pressure calculation result is then integrated over time to obtain the intermediate integral. The pressure integral term is obtained by multiplying the intermediate integral by the ratio of the difference between the first sliding mode gain and the second convergence index minus 2.

[0010] In conjunction with the first aspect, in a sixth implementation of the first aspect of the present invention, the step of adding the pressure sliding surface to the linear term of the pressure sliding surface to obtain the pressure approaching law, and adding the speed sliding surface to the linear term of the speed sliding surface to obtain the speed approaching law, includes: The absolute value of the pressure sliding surface is raised to the power of 0.5, multiplied by the pressure sign function value, and then multiplied by the first switching gain to obtain the pressure power approximation term; the absolute value of the speed sliding surface is raised to the power of 0.5, multiplied by the speed sign function value, and then multiplied by the second switching gain to obtain the speed power approximation term. Multiplying the pressure sliding surface by a first linear coefficient yields the pressure linear approximation term, and multiplying the rotational speed sliding surface by a second linear coefficient yields the rotational speed linear approximation term. The pressure power-law approximation term is added to the pressure linear approximation term to obtain the pressure approximation law, and the speed power-law approximation term is added to the speed linear approximation term to obtain the speed approximation law.

[0011] In conjunction with the first aspect, in the seventh implementation of the first aspect of the present invention, the step of raising the absolute value of the pressure sliding surface to the power of 0.5, multiplying it by the pressure sign function value, and then multiplying it by the first switching gain to obtain the pressure power approximation term; and raising the absolute value of the speed sliding surface to the power of 0.5, multiplying it by the speed sign function value, and then multiplying it by the second switching gain to obtain the speed power approximation term, includes: The absolute value of the pressure sliding surface is taken and raised to the power of 0.5 to obtain the basic value of the pressure power; the absolute value of the rotational speed sliding surface is taken and raised to the power of 0.5 to obtain the basic value of the rotational speed power. The pressure sign function value is determined based on the sign of the pressure sliding surface, and the rotational speed sign function value is determined based on the sign of the rotational speed sliding surface. The pressure power base value is multiplied by the pressure sign function value and then by the first switching gain to obtain the pressure power approximation term. The speed power base value is multiplied by the speed sign function value and then by the second switching gain to obtain the speed power approximation term.

[0012] In conjunction with the first aspect, in the eighth implementation of the first aspect of the present invention, the step of calculating the cylinder duty cycle output value based on the pressure approaching law and the centrifugal force compensation term, and calculating the motor voltage output value based on the speed approaching law and the friction torque, includes: The current rotational speed of the window cleaning robot is subtracted from the rotational speed of the previous cycle, and then divided by the sampling period to obtain the rotational speed differential value. The product of the cleaning component mass, eccentricity, tilt angle cosine value, rotational speed value and the rotational speed differential value is divided by the maximum output force of the cylinder to obtain the centrifugal force compensation term. The cylinder duty cycle output value is obtained by sequentially adding the product of the suspended working point duty cycle value, the pressure proportional coefficient and the pressure approach law, and the centrifugal force compensation term. Divide the frictional torque by the motor torque constant to obtain the frictional torque compensation term. Then, add the voltage value at the suspended working point, the product of the speed proportional coefficient and the speed approach law, and the frictional torque compensation term in sequence to obtain the motor voltage output value.

[0013] A second aspect of this application provides a window cleaning robot cleaning execution system based on dynamic control, the window cleaning robot cleaning execution system based on dynamic control comprising: A setting module is used to calculate the stain density and set the target pressure and target rotation speed based on the stain density; The calculation module is used to calculate the dynamic friction coefficient based on the stain density, and multiply the dynamic friction coefficient by the window surface pressure and the radius of the cleaning cloth to obtain the friction torque; A construction module is used to obtain the actual pressure and construct a pressure sliding surface based on the target pressure and the actual pressure; The processing module is used to add the pressure sliding surface to the linear term of the pressure sliding surface to obtain the pressure approaching law, and to add the speed sliding surface to the linear term of the speed sliding surface to obtain the speed approaching law. The calculation module is used to calculate the cylinder duty cycle output value based on the pressure approach law and the centrifugal force compensation term, and to calculate the motor voltage output value based on the speed approach law and the friction torque.

[0014] Compared with existing technologies, this application has the following advantages: By constructing a dynamic friction coefficient mapping relationship based on dirt density and a pressure-speed coupled dynamic model, it reveals the bidirectional coupling mechanism of centrifugal force weakening pressure and friction torque reducing speed during the cleaning process. Compared with the independent control method of existing technologies, it solves the control mismatch problem caused by the mutual constraint of pressure and speed. By adopting a three-source compensation model, the influence of temperature drift, cumulative wear, and reference zero bias on sensor measurements is quantified separately, achieving accurate calculation of the real pressure and eliminating the control drift phenomenon caused by the accumulation of sensor bias error in existing technologies. By using a non-singular fast terminal sliding surface combined with the fractional power term and power integral term of pressure error, the singularity problem of traditional terminal sliding is avoided while ensuring global fast convergence. Compared with existing linear sliding mode control, it shortens the response time and reduces overshoot. By adopting an adaptive reaching law, the reaching speed is adaptively adjusted with the sliding surface deviation through the combined effect of the power reaching term and the linear reaching term, effectively suppressing the high-frequency chattering caused by the existing constant velocity reaching law and extending the actuator service life. The centrifugal force compensation term and friction torque compensation term introduced in the control law of this invention pre-counteract the coupling disturbance through a feedforward mechanism, thereby achieving synchronous and rapid tracking of pressure and rotation speed under conditions of drastic changes in dirt density, ensuring the stable and efficient cleaning performance of the window cleaning robot. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art 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.

[0016] The structures, proportions, sizes, etc., shown in the accompanying drawings of this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed in the specification, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0017] Figure 1 This is a flowchart illustrating the cleaning execution method of a window cleaning robot based on dynamic control provided in an embodiment of the present invention. Figure 2 This is a schematic block diagram of the window cleaning robot cleaning execution system based on dynamic control provided in an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0020] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0021] It should also be further understood that the term "and / or" as used in this application specification and the appended claims refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes such combinations. See also Figure 1 One embodiment of the window cleaning robot cleaning execution method based on dynamic control in this application includes: Step 100: Calculate the stain density and set the target pressure and target rotation speed based on the stain density; Specifically, the window image acquired by the window cleaning robot in its current working area is spatially divided into several regular grid units according to a preset grid size. Each grid unit contains several pixels. The red, green, and blue channel brightness values ​​of all pixels in each grid are extracted, and the average value of each channel is calculated. Then, the RGB channel values ​​are converted to grayscale values ​​using a weighted formula. The average grayscale values ​​of all pixels within a single grid are then calculated to obtain the grayscale mean value for that grid. Using the grayscale mean value as input, the maximum grayscale value in all grids of the current image is subtracted from the grayscale mean value. The difference is then divided by the maximum grayscale value to obtain a standardized stain density index. The stain density index ranges from [0,1] and is used to quantitatively characterize the degree of contamination in the grid area. A higher stain density indicates weaker light reflection and more severe contamination in the current area. A first and a second stain density threshold are preset, representing the judgment boundaries for low and high contamination, respectively. When the stain density value of a grid is greater than the second stain density threshold, the grid area is identified as a heavily contaminated area. The target pressure will be set to the first target pressure value to enhance the adhesion of the cleaning mechanism, and the target rotation speed will be set to the first target rotation speed value to increase the cleaning frequency and enhance the removal of contaminants. When the stain density value is lower than the first stain density threshold, the grid area is identified as a low-contamination area. The target pressure will be set to the second target pressure value to reduce structural load and save energy. The target rotation speed will be set to the second target rotation speed. The values ​​are used to control wear and extend component life. If the stain density is between the first stain density threshold and the second stain density threshold, it indicates that the grid area is a medium-pollution level area. Then, the target pressure and target speed are calculated by linear interpolation. The second target pressure value is used as the interpolation starting point, the first pressure interpolation coefficient is used as the slope, and the difference between the current stain density and the first stain density threshold is multiplied and added to the second target pressure value to obtain the interpolated target pressure value. At the same time, the second target speed value is used as the base speed, and the product of the first speed interpolation coefficient and the stain density difference is added to the base speed to obtain the target speed, thus forming a control variable adjustment scheme.

[0022] Step 200: Calculate the dynamic friction coefficient based on the stain density, and multiply the dynamic friction coefficient by the window surface pressure and the radius of the cleaning cloth to obtain the friction torque; Specifically, the dirt density value is multiplied by a first mapping coefficient to enhance its sensitivity in the friction model, and then the first bias parameter is subtracted to adjust its mapping starting point, resulting in the density calculation result, which reflects the nonlinear response of the degree of contamination to the friction characteristics. The density calculation result is substituted as an independent variable into the hyperbolic tangent function, i.e., the tanh function is used to evaluate the density calculation result. The tanh function has an S-shaped output characteristic when its domain is the set of real numbers, and can map the input value to the (-1, 1) interval, obtaining a bounded, continuous, and smoothly changing variable component of the friction coefficient, reflecting the changing trend of friction behavior caused by differences in lubricity, adsorption, or adhesion in different contaminated areas. The variable component of the friction coefficient is multiplied by a second mapping coefficient to adjust its proportion in the total friction coefficient, and then linearly superimposed with a preset base friction coefficient to obtain the dynamic friction coefficient corresponding to the current grid area. The dynamic friction coefficient exhibits dynamic adjustment characteristics with different dirt densities, enabling the robot to identify stronger frictional reaction requirements in highly contaminated areas and maintain control over energy consumption and wear in low-contamination areas. The dynamic friction coefficient is multiplied sequentially by the actual normal pressure value applied to the window surface and the effective radius of the cleaning cloth to obtain the friction torque of the corresponding area.

[0023] Step 300: Obtain the actual pressure and construct a pressure sliding surface based on the target pressure and the actual pressure; It should be noted that the ambient temperature data during the operation of the window cleaning robot is acquired, and the current real-time temperature value minus a preset standard temperature value is used as the basis for temperature deviation. The temperature deviation is multiplied by a set of preset temperature drift coefficients to calculate the temperature drift compensation amount, which is used to correct the zero-point drift of the pressure sensor caused by temperature changes. The rotational speed signal of the cleaning motor is monitored in real time, and its square value is used as the integration object. The square value of the rotational speed is accumulated and integrated over time, and then multiplied by the corresponding wear coefficient to obtain the rotational speed square time integral wear compensation amount, which reflects the rigidity attenuation and force feedback transmission error caused by the wear of moving parts during long-term operation of the cleaning system. Considering the inherent zero-point offset of the sensor in the state of no load or not in contact with the glass surface, i.e., the suspension zero offset compensation amount, it is obtained through pre-calibration. Finally, the original pressure measurement value of the sensor is subtracted sequentially from the temperature drift compensation amount, the rotational speed square time integral wear compensation amount, and the suspension zero offset compensation amount to obtain the true pressure value after environmental adaptation and state adaptive correction. The difference between the actual pressure value and the target pressure set by the current control objective is calculated to form a pressure error term. The pressure error is then introduced into the sliding mode function construction method in sliding mode control theory to obtain the pressure sliding surface expression. The pressure sliding surface is represented as a real-time updated state function, and its changing trend reflects the magnitude and approach speed of the current deviation of the system in the force control dimension.

[0024] Step 400: Add the linear terms of the pressure sliding surface to obtain the pressure approaching law; add the linear terms of the speed sliding surface to obtain the speed approaching law. Specifically, nonlinear enhancement processing is applied to the pressure sliding surface and the speed sliding surface to improve the convergence speed and disturbance rejection capability of the system in the near-zero error region. A hybrid reaching law model is constructed by combining power-law and linear reaching terms. The absolute value of the pressure sliding surface is raised to the power of 0.5 (square root operation) to form a nonlinear dynamic response characteristic that decays slowly as the error decreases and responds rapidly as the error increases. The result of the power operation is then multiplied by the sign function value of the pressure sliding surface to restore its original sign direction, resulting in a directional nonlinear pressure error adjustment. This adjustment is then multiplied by the first switching gain coefficient to obtain the pressure power-law reaching term. The same processing method is applied to the speed sliding surface: its absolute value is raised to the power of 0.5, multiplied by its sign function value, and then multiplied by the second switching gain to obtain the speed power-law reaching term, which is used to suppress and quickly adjust dynamic disturbances in speed control. Multiplying the pressure sliding surface by the first linear coefficient yields the pressure linear reaching term, and multiplying the speed sliding surface by the second linear coefficient yields the speed linear reaching term. These two linear terms are used to construct the controller's fine-tuning capability under small error conditions, ensuring the reaching law exhibits strong error-preserving characteristics and response consistency in the region near zero. Summing the pressure power-law reaching term and the pressure linear reaching term forms the pressure reaching law function, and summing the speed power-law reaching term and the speed linear reaching term forms the speed reaching law, which is used by the motor drive controller to calculate the target voltage or current.

[0025] Step 500: Calculate the cylinder duty cycle output value based on the pressure approach law and centrifugal force compensation term, and calculate the motor voltage output value based on the speed approach law and friction torque.

[0026] Specifically, the actual rotational speed of the window cleaning robot within the current control cycle is obtained. The difference between the current rotational speed and the speed recorded in the previous control cycle is calculated and then divided by the sampling cycle of the control system to obtain the time derivative of the rotational speed, i.e., the instantaneous acceleration index, reflecting the trend of angular acceleration change of the cleaning component within the current cycle. Combining the physical parameters of the cleaning component, including component mass, eccentricity, and the cosine value of the tilt angle under the actual working posture, these are multiplied with the current rotational speed and the rotational speed derivative to obtain a product term reflecting the change in centrifugal inertial force. This product term is then divided by the maximum output force of the cylinder and standardized to obtain a centrifugal force compensation term, used to correct the normal force offset problem caused by structural eccentricity and rotational inertia when the window cleaning robot operates on inclined glass surfaces or curved trajectories. The preset duty cycle value of the suspended operating point is used as the basic output. The pressure approach law is multiplied by the pressure proportional coefficient to form the control term strength. This product term is then added sequentially to the basic duty cycle value and the centrifugal force compensation term to obtain the cylinder duty cycle output value. This value is used to drive the solenoid valve to output the corresponding PWM control signal, achieving continuous adjustment of the clamping force and environmental adaptive control. In the motor control path, the friction torque value is divided by the motor torque constant to obtain the voltage compensation term required by the motor under load conditions. The torque constant is one of the motor's physical parameters, representing the torque output capability per unit voltage, thus performing an equivalent transformation on the friction load. The set suspended operating point voltage value is added to the control quantity obtained by multiplying the speed approach law by the speed proportional coefficient, and then sequentially accumulated with the friction torque compensation term to form the motor voltage output value.

[0027] In one specific embodiment, the process of performing step 100 may specifically include the following steps: The window image is divided into multiple grids, and the mean value of grayscale is calculated based on the red, green and blue channel values ​​of all pixels in a single grid. The stain density is obtained by subtracting the mean gray value from the maximum gray value and then dividing by the maximum gray value. The stain density is compared with a first stain density threshold and a second stain density threshold; When the stain density is greater than the second stain density threshold, the target pressure is set to the first target pressure value and the target rotation speed is set to the first target rotation speed value. When the stain density is less than the first stain density threshold, the target pressure is set to the second target pressure value and the target rotation speed is set to the second target rotation speed value. When the stain density is between the first stain density threshold and the second stain density threshold, the target pressure is obtained by adding the second target pressure value to the first pressure interpolation coefficient and multiplying it by the stain density minus the first stain density threshold. The target speed is obtained by adding the second target speed value to the first speed interpolation coefficient and multiplying it by the stain density minus the first stain density threshold.

[0028] Specifically, the window image undergoes regularization processing, dividing the entire image into multiple grid units of a preset size. Each grid region corresponds to a local area on the robot's cleaning path. Within each grid unit, all pixels are decomposed at the channel level, extracting the brightness values ​​of the red, green, and blue channels respectively. The brightness values ​​of all pixels within each channel are accumulated and averaged to form the three-channel mean. The three-channel mean values ​​are then weighted and synthesized using standard grayscale conversion weights to obtain the grayscale mean of the current grid region. The grayscale value serves as a quantitative indicator of the optical reflectivity of the local area; a lower grayscale value indicates greater darkness, meaning more severe coverage by contaminants such as dust, oil film, and watermarks. The maximum grayscale value among all grids in the current image is selected as the normalization benchmark. The difference between the maximum grayscale value and the grayscale mean of the current grid is divided by the maximum grayscale value to obtain the stain density corresponding to the current grid. The stain density ranges from [0,1], where a value closer to 1 indicates a higher degree of contamination. The stain density is compared with a set first stain density threshold and a second stain density threshold. For example, the first stain density threshold ρ1 = 0.25 and the second stain density threshold ρ2 = 0.70 are set, where ρ < 0.25 is defined as a low-pollution area, ρ > 0.70 is defined as a heavily polluted area, and the area between the two is a medium-pollution area. If ρ > 0.70 for a given grid, it indicates significant contaminant coverage in the area. The cleaning system is set to a target pressure of P1 = 12.0 N and a target rotation speed of ω1 = 320 rpm to maximize local cleaning intensity. If ρ < 0.25, it indicates almost no obvious stains in the area. The target pressure is set to P2 = 5.0 N and the target rotation speed to ω2 = 180 rpm to reduce the load on the mechanism and extend the life of the cleaning elements. If ρ ∈ [0.25, 0.70], linear interpolation is used to continuously adjust the pressure and rotation speed. The first pressure interpolation coefficient α1 = (P1 - P2) / (ρ2 - ρ1) = (12.0 - 5.0) / (0.70 - 0.25) ≈ 15.56 N, and the first rotation speed interpolation coefficient β1 = (ω1 - ω2) / (ρ2 - ρ1) = (320 - 180) / (0.70 - 0.25) ≈ 311.11 rpm.

[0029] In one specific embodiment, the process of performing step 200 may specifically include the following steps: Multiply the stain density by the first mapping coefficient and subtract the first bias parameter to obtain the density calculation result. Substitute the density calculation result into the hyperbolic tangent function to obtain the variable component of the friction coefficient. The dynamic friction coefficient is obtained by adding the product of the basic friction coefficient and the variable component of the friction coefficient multiplied by the second mapping coefficient. The friction torque is obtained by multiplying the dynamic friction coefficient by the window surface pressure and the radius of the cleaning cloth in sequence.

[0030] Specifically, the stain density value ρ is multiplied by a preset first mapping coefficient a1. This first mapping coefficient is used to adjust the slope sensitivity of the stain density in the function mapping process, so that the response amplitude of different levels of contamination shows a non-linear amplification trend. The first bias parameter b1 is subtracted from the product result. The first bias value is used to adjust the position of the zero point of the function, so that ρ = 0.25 corresponds to the middle of the negative half axis and ρ = 0.70 corresponds to the middle of the positive half axis, realizing the differentiation mapping between medium and high contamination states, and obtaining the density calculation result x = a1×ρ - b1. Substituting the calculation result x into the hyperbolic tangent function, i.e., tanh(x), the density change is mapped to a bounded interval (-1,1) through its S-shaped non-linear output characteristics, obtaining the variable component of the friction coefficient. The variable component of the friction coefficient is approximately negative in the low contamination region and approaches positive in the high contamination region, with a consistent response and smooth slope, effectively characterizing the changing trend of friction surface lubrication state, surface adhesion and material coupling effect caused by stain density. A preset baseline friction coefficient μ0 is used as the friction performance benchmark for the cleaning system under clean window surfaces or standard conditions. The variable component of the friction coefficient is multiplied by a second mapping coefficient a2, which controls the proportion of friction variation in the total friction coefficient. A linear scaling operation can be used to set an upper limit trend where the additional friction increases with heavier contamination. The product result is then added to μ0 to obtain the dynamic friction coefficient. The dynamic friction coefficient changes smoothly from 0 to 1 with the dirt density, allowing the robot to maintain a relatively high speed and avoid idling in low-friction areas, while improving drive compensation to overcome load resistance in high-friction areas. The dynamic friction coefficient is then multiplied sequentially by the normal pressure P currently applied to the window surface and the effective radius r of the cleaning cloth to obtain the friction torque corresponding to the current area.

[0031] In one specific embodiment, the process of performing step 300 may specifically include the following steps: The temperature drift compensation amount is obtained by subtracting the standard temperature value from the real-time temperature value and then multiplying the result by the temperature drift coefficient. The wear compensation amount is obtained by integrating the square of the rotational speed over time and multiplying it by the wear coefficient. The true pressure is obtained by successively subtracting the temperature drift compensation, the speed square time integral wear compensation, and the suspension zero bias compensation from the sensor pressure measurement value; The pressure error is constructed based on the target pressure and the actual pressure, and a pressure sliding surface is created based on the pressure error.

[0032] Specifically, the temperature value of the current working environment is collected in real time by a temperature sensor integrated on the robot body or control module. This temperature value is then compared with the standard temperature value set during factory calibration to obtain the current ambient temperature difference. This ambient temperature difference has a non-linear impact on factors such as the sensitivity of the pressure sensor, zero-point voltage, and bridge imbalance. Therefore, the temperature difference is multiplied by a preset temperature drift coefficient to obtain the temperature drift compensation amount. During long-term operation, the high-speed rotating components in the window cleaning robot's cleaning assembly will experience slight changes in the actual contact force transmission path due to mechanical wear, changes in bearing clearance, and decreased lubrication efficiency. This indirectly affects the output stability of the sensor. Therefore, a historical compensation mechanism based on rotational speed wear is constructed. The current rotational speed value ω(t) is acquired in real time, and its square value ω is calculated. 2 (t) Perform cumulative integration, that is, calculate ∫ω within a certain time window. 2 The integral of (t)dt is multiplied by the wear coefficient to obtain the wear compensation amount based on the square of the rotational speed over time. A static zero-bias compensation amount is introduced, which is obtained by averaging multiple measurements taken by the system in a static state. This reflects the inherent zero-point offset error of the sensor under no-load and no-temperature-change conditions. The temperature drift compensation amount, rotational speed wear compensation amount, and zero-bias compensation amount are successively subtracted from the original pressure measurement value output by the sensor to obtain the true pressure value at the current moment. Based on the true pressure value, the difference between it and the target pressure currently desired by the control system is calculated to form the pressure error e_P. A pressure sliding mode surface function is constructed based on the pressure error.

[0033] In one specific embodiment, the process of constructing a pressure error based on the target pressure and the actual pressure, and creating a pressure sliding surface based on the pressure error, may specifically include the following steps: The pressure error is obtained by subtracting the actual pressure from the target pressure. The absolute value of the pressure error is subjected to the first power of convergence and the second power of convergence integration to obtain the pressure power term and the pressure integral term. The pressure error, the pressure power term, and the pressure integral term are added together to obtain the pressure sliding surface.

[0034] Specifically, within each control cycle, the real-time pressure value output by the force sensor is collected and compared with the target pressure value set by the control algorithm. The difference between the target pressure and the real pressure is calculated to obtain the current pressure error, reflecting the contact deviation of the cleaning mechanism during execution. A positive pressure error indicates that the pressure applied by the robot is lower than the expected value, requiring an increase in output air pressure; a negative pressure error indicates that the actual pressure is too high, requiring reverse adjustment to avoid damage to the glass. Different forms of nonlinear processing are applied to the absolute value of the error. The absolute value of the pressure error is raised to the power of the first convergence exponent. For example, when the first convergence exponent is 0.7, it exhibits higher response sensitivity to smaller errors while providing a more moderate adjustment for larger errors, effectively avoiding oscillations. This yields the pressure exponent term, which is used to improve the local nonlinear response capability of the control system. The absolute value of the pressure error is then raised to the power of the second convergence exponent, followed by time integration. For example, when the second convergence exponent is set to 1.2, the pressure integral term is obtained through integration. This term has a memory effect in the control system, accumulating the influence of small errors over a long period. It is used to correct persistent deviations and system hysteresis effects, thereby improving steady-state accuracy. The pressure error, the pressure exponent term, and the pressure integral term are added together to construct the pressure sliding surface. Mathematically, this describes the comprehensive deviation relationship between the system state and the desired state, serving as the basis for the sliding mode controller to determine the control direction and approach rate.

[0035] In one specific embodiment, the process of performing a first power-law operation and a second power-law integration operation on the absolute value of the pressure error to obtain the pressure power term and the pressure integral term can specifically include the following steps: The absolute value of the pressure error is multiplied by the first convergence exponent after being raised to the first sign function, and then multiplied by the ratio of the first sliding mode gain to the first convergence exponent to obtain the pressure exponent term. The absolute value of the pressure error is multiplied by the second convergence exponent and then by the second sign function to obtain the pressure calculation result. The pressure calculation result is then integrated over time to obtain the intermediate integral. The pressure integral term is obtained by multiplying the intermediate integral by the ratio of the difference between the first sliding mode gain and the second convergence index minus 2.

[0036] Specifically, the pressure error e_P at the current moment is obtained, and the absolute value of the pressure error is substituted into the power control structure for nonlinear transformation. |e_P| is calculated with the first convergence exponent n1 as the power, resulting in |e_P|^n1. This amplifies the control response in the small error range and mitigates the oscillation trend in the large error range, forming a power amplification term with convergence capability. e_P is then substituted into the sign function sign(e_P) to obtain the first sign function value. The sign function is used to restore the power term to the same positive and negative sign range as the original error. The power term is multiplied by the sign function result, and then multiplied by the ratio of the first sliding mode gain K1 to the first convergence exponent n1, i.e., K1 / n1, to obtain the pressure power term F1 = K1 / n1 × |e_P|^n1 × sign(e_P). The pressure power term has strong nonlinear adjustment capability and achieves smooth and continuous control transition within the error variation range. The absolute value of the pressure error is introduced into the second convergence exponent n2 for exponentiation, resulting in |e_P|^n2. This is then multiplied by the sign function sign(e_P) of e_P to obtain the signed nonlinear error enhancement term F2 = |e_P|^n2 × sign(e_P). This term constitutes the integrable function on the pressure integration path. Integrating F2 over time forms the intermediate integral I_F2 = ∫F2dt. The integration process effectively accumulates the persistent pressure deviation that occurs in the system during cleaning operations, compensating for steady-state shifts caused by mechanical hysteresis, feedback lag, or errors in flexible structures. The intermediate integral I_F2 is multiplied by the ratio of K1 divided by (n2-2), i.e., K1 / (n2-2), to construct the pressure integration term F3 = K1 / (n2-2) × ∫|e_P|^n2 × sign(e_P)dt.

[0037] In one specific embodiment, the process of performing step 400 may specifically include the following steps: The absolute value of the pressure sliding surface is raised to the power of 0.5, multiplied by the pressure sign function value, and then multiplied by the first switching gain to obtain the pressure power-approaching term; the absolute value of the speed sliding surface is raised to the power of 0.5, multiplied by the speed sign function value, and then multiplied by the second switching gain to obtain the speed power-approaching term. Multiplying the pressure sliding surface by the first linear coefficient yields the pressure linear approximation term, and multiplying the rotational speed sliding surface by the second linear coefficient yields the rotational speed linear approximation term. Adding the power-law approach term of pressure to the linear approach term of pressure yields the pressure approach law, and adding the power-law approach term of rotational speed to the linear approach term of rotational speed yields the rotational speed approach law.

[0038] Specifically, the absolute value of the pressure sliding surface is raised to the power of 0.5 to construct a nonlinear adjustment basis term. This nonlinear adjustment basis term exhibits high sensitivity when the error is close to zero, and its response curve tends to flatten when the error is large. The power result is multiplied by the sign function of the pressure sliding surface to preserve the positive and negative directional characteristics of the sliding surface in the nonlinear approximation term. The product is then multiplied by a preset first switching gain parameter of 5.0 to obtain the pressure power approximation term. This serves as a nonlinear absorption control mechanism to suppress the divergence trend of the system under structural disturbances, environmental disturbances, or sudden load changes. Similarly, for the speed control channel, the absolute value of the speed sliding surface is raised to the power of 0.5, multiplied by its sign function to preserve directionality, and then multiplied by a second switching gain parameter of 4.5 to obtain the speed power approximation term. Its mechanism is consistent with the corresponding structure of the pressure channel, both used to enhance the control convergence capability of the system within the nonlinear error range. Multiplying the pressure sliding surface by the first linear coefficient of 3.2 directly constructs the pressure linear reaching term, which constrains the convergence speed of the reaching trajectory near the zero error point, thereby reducing oscillations and over-control. Multiplying the speed sliding surface by the second linear coefficient of 2.8 yields the speed linear reaching term, ensuring the controller maintains a stable and adjustable output during the speed stabilization phase. Adding the nonlinear reaching term and the linear reaching term of the pressure channel yields the composite reaching law of the pressure control loop; adding the nonlinear reaching term and the linear reaching term of the speed channel yields the composite reaching law of the speed control loop.

[0039] In one specific embodiment, the process of raising the absolute value of the pressure sliding surface to the power of 0.5, multiplying it by the pressure sign function value, and then multiplying it by the first switching gain to obtain the pressure power approximation term; and raising the absolute value of the speed sliding surface to the power of 0.5, multiplying it by the speed sign function value, and then multiplying it by the second switching gain to obtain the speed power approximation term, can specifically include the following steps: The absolute value of the pressure sliding surface is taken and raised to the power of 0.5 to obtain the basic value of the pressure power; the absolute value of the speed sliding surface is taken and raised to the power of 0.5 to obtain the basic value of the speed power. The sign function value of the pressure is determined based on the sign of the pressure sliding surface, and the sign function value of the rotational speed is determined based on the sign of the rotational speed sliding surface. Multiplying the pressure power base value by the pressure sign function value and then by the first switching gain yields the pressure power approximation term. Multiplying the speed power base value by the speed sign function value and then by the second switching gain yields the speed power approximation term.

[0040] Specifically, after obtaining the current pressure sliding surface value, its absolute value is taken and then raised to the power of 0.5 to obtain the pressure power base value. The pressure power base value exhibits non-linear amplification characteristics, providing higher sensitivity when the error is small and showing a gentle convergence trend when the error is large. The same process is performed on the speed sliding surface, i.e., its absolute value is taken and raised to the power of 0.5 to obtain the speed power base value, which is used to construct the dynamic absorption and adjustment capability in the speed control loop. The positive and negative signs of the pressure sliding surface and the speed sliding surface are determined respectively, i.e., based on the logical judgment that their current values ​​are greater than zero, less than zero, or equal to zero, a sign function value of 1, -1, or 0 is assigned respectively, constructing the pressure sign function value and the speed sign function value. The pressure power base value is multiplied by the pressure sign function value and then multiplied by the first switching gain, which is set to 5.0 to enhance the dominant role of the nonlinear term in error adjustment, thus obtaining the pressure power approximation term. Similarly, the speed power base value is multiplied by the speed sign function value and then multiplied by the second switching gain, which is set to 4.5, thus obtaining the speed power approximation term.

[0041] In one specific embodiment, the process of performing step 500 may specifically include the following steps: The differential value of the rotation speed is obtained by subtracting the rotation speed value of the previous cycle from the current rotation speed value of the window cleaning robot and dividing it by the sampling period. The centrifugal force compensation term is obtained by dividing the product of the mass of the cleaning component, the eccentricity, the cosine value of the tilt angle, the rotation speed value and the differential value of the rotation speed by the maximum output force of the cylinder. The cylinder duty cycle output value is obtained by sequentially adding the duty cycle value of the suspended working point, the product of the pressure proportional coefficient and the pressure approach law, and the centrifugal force compensation term. Divide the friction torque by the motor torque constant to obtain the friction torque compensation term. Then, add the voltage value at the suspended operating point, the product of the speed proportional coefficient and the speed approach law, and the friction torque compensation term in sequence to obtain the motor voltage output value.

[0042] Specifically, the robot's rotational speed is acquired within the current control cycle, and the historical rotational speed value from the previous control cycle is read. The current rotational speed value is subtracted from the previous cycle's value, and then divided by the set sampling period to obtain the time derivative of the rotational speed, i.e., the change in rotational speed per unit time, approximated as angular acceleration, used to dynamically describe the rotational trend of the cleaning component. This time derivative is multiplied by the cleaning component's mass, structural eccentricity, the current actual rotational speed, and the cosine of the tilt angle corresponding to the working angle, forming a physical quantity expression reflecting the change in the system's inertial centrifugal load. This product term reflects the influence of the eccentric structure on the cylinder's normal force distribution under different postures and rotational states. Dividing the product term by the cylinder's maximum output force value completes the normalization process, yielding a centrifugal force compensation term. This centrifugal force compensation term serves as a disturbance suppression quantity in the cylinder control channel, used to correct the error caused by inertial deviation between the actual air pressure output and the target contact force. The preset duty cycle value of the suspended operating point, i.e., the steady-state control value of the cylinder when it is not in contact with the glass surface, is used as the basic offset. The currently calculated pressure approach law is multiplied by the preset pressure proportional coefficient to obtain the adjustment term. The pressure proportional coefficient is used to determine the weight contribution of pressure sliding mode control in PWM regulation. The product result is added to the suspended operating point duty cycle value and the centrifugal force compensation term in sequence to obtain the cylinder duty cycle output value. In the motor control channel, the currently calculated friction torque value is divided by the motor torque constant to obtain the friction torque compensation term. The friction torque compensation term reflects the additional voltage compensation required for the motor to maintain the desired speed under the current friction load. The set suspended operating point voltage value is used as the static bias reference, and the current speed approach law is multiplied by the speed proportional coefficient to obtain the dynamic speed adjustment term. The dynamic speed adjustment term is then added to the suspended voltage value and the friction torque compensation term in sequence to obtain the motor voltage output value.

[0043] The above describes the window cleaning robot cleaning method based on dynamic control in the embodiments of this application. The following describes the window cleaning robot cleaning system based on dynamic control in the embodiments of this application. Please refer to [link / reference]. Figure 2 One embodiment of the window cleaning robot cleaning execution system based on dynamic control in this application includes: The setting module 11 is used to calculate the stain density and set the target pressure and target rotation speed based on the stain density; Calculation module 12 is used to calculate the dynamic friction coefficient based on the stain density, and multiply the dynamic friction coefficient by the window surface pressure and the radius of the cleaning cloth to obtain the friction torque; Module 13 is used to obtain the actual pressure and construct a pressure sliding surface based on the target pressure and the actual pressure. Processing module 14 is used to add the linear terms of the pressure sliding surface to obtain the pressure approaching law, and to add the linear terms of the speed sliding surface to obtain the speed approaching law. The calculation module 15 is used to calculate the cylinder duty cycle output value based on the pressure approach law and the centrifugal force compensation term, and to calculate the motor voltage output value based on the speed approach law and the friction torque.

[0044] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0045] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0046] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A window cleaning robot cleaning execution method based on dynamic control, characterized in that, include: Calculate the stain density, and set the target pressure and target rotation speed based on the stain density; The dynamic friction coefficient is calculated based on the stain density, and the friction torque is obtained by multiplying the dynamic friction coefficient by the window surface pressure and the radius of the cleaning cloth. Obtain the actual pressure, and construct a pressure sliding surface based on the target pressure and the actual pressure; Adding the pressure sliding surface to the linear term of the pressure sliding surface yields the pressure approaching law; adding the speed sliding surface to the linear term of the speed sliding surface yields the speed approaching law. The cylinder duty cycle output value is calculated based on the pressure approach law and the centrifugal force compensation term, and the motor voltage output value is calculated based on the speed approach law and the friction torque.

2. The window cleaning robot cleaning execution method based on dynamic control according to claim 1, characterized in that, The calculation of stain density, and the setting of target pressure and target rotation speed based on the stain density, includes: The window image is divided into multiple grids, and the mean value of grayscale is calculated based on the red, green and blue channel values ​​of all pixels in a single grid. The stain density is obtained by subtracting the mean gray value from the maximum gray value and then dividing by the maximum gray value. The stain density is compared with a first stain density threshold and a second stain density threshold; When the stain density is greater than the second stain density threshold, the target pressure is set to the first target pressure value and the target rotation speed is set to the first target rotation speed value. When the stain density is less than the first stain density threshold, the target pressure is set to the second target pressure value and the target rotation speed is set to the second target rotation speed value. When the stain density is within the range of the first stain density threshold to the second stain density threshold, the target pressure is obtained by adding the second target pressure value to the first pressure interpolation coefficient and multiplying it by the stain density minus the first stain density threshold, and the target speed is obtained by adding the second target speed value to the first speed interpolation coefficient and multiplying it by the stain density minus the first stain density threshold.

3. The window cleaning robot cleaning execution method based on dynamic control according to claim 1, characterized in that, The calculation of the dynamic friction coefficient based on the stain density, and the multiplication of the dynamic friction coefficient by the window surface pressure and the cleaning cloth radius to obtain the friction torque, includes: Multiply the stain density by the first mapping coefficient and subtract the first bias parameter to obtain the density calculation result. Substitute the density calculation result into the hyperbolic tangent function to obtain the variable component of the friction coefficient. The dynamic friction coefficient is obtained by adding the product of the basic friction coefficient and the variable component of the friction coefficient multiplied by the second mapping coefficient. The friction torque is obtained by multiplying the dynamic friction coefficient by the window surface pressure and the radius of the cleaning cloth in sequence.

4. The window cleaning robot cleaning execution method based on dynamic control according to claim 1, characterized in that, The step of obtaining the actual pressure and constructing a pressure sliding surface based on the target pressure and the actual pressure includes: The temperature drift compensation amount is obtained by subtracting the standard temperature value from the real-time temperature value and then multiplying the result by the temperature drift coefficient. The wear compensation amount is obtained by integrating the square of the rotational speed over time and multiplying it by the wear coefficient. The true pressure is obtained by successively subtracting the temperature drift compensation, the speed square time integral wear compensation, and the zero bias compensation from the sensor pressure measurement value; A pressure error is constructed based on the target pressure and the actual pressure, and a pressure sliding surface is created based on the pressure error.

5. The window cleaning robot cleaning execution method based on dynamic control according to claim 4, characterized in that, The step of constructing a pressure error based on the target pressure and the actual pressure, and creating a pressure sliding surface based on the pressure error, includes: The pressure error is obtained by subtracting the actual pressure from the target pressure. The absolute value of the pressure error is subjected to a first power-law operation and a second power-law integration operation to obtain the pressure power term and the pressure integral term. The pressure error, the pressure power term, and the pressure integral term are added together to obtain the pressure sliding surface.

6. The window cleaning robot cleaning execution method based on dynamic control according to claim 5, characterized in that, The absolute value of the pressure error is subjected to a first power-law operation and a second power-law integration operation to obtain a pressure power term and a pressure integral term, including: The absolute value of the pressure error is multiplied by a first convergence exponent after being raised to the first power, and then by a first sign function, and then by the ratio of the first sliding mode gain to the first convergence exponent to obtain the pressure exponent term. The absolute value of the pressure error is multiplied by a second convergence exponent and then by a second sign function to obtain the pressure calculation result. The pressure calculation result is then integrated over time to obtain the intermediate integral. The pressure integral term is obtained by multiplying the intermediate integral by the ratio of the difference between the first sliding mode gain and the second convergence index minus 2.

7. The window cleaning robot cleaning execution method based on dynamic control according to claim 1, characterized in that, The step of adding the pressure sliding surface to the linear term of the pressure sliding surface to obtain the pressure approaching law, and adding the speed sliding surface to the linear term of the speed sliding surface to obtain the speed approaching law, includes: The absolute value of the pressure sliding surface is raised to the power of 0.5, multiplied by the pressure sign function value, and then multiplied by the first switching gain to obtain the pressure power approximation term; the absolute value of the speed sliding surface is raised to the power of 0.5, multiplied by the speed sign function value, and then multiplied by the second switching gain to obtain the speed power approximation term. Multiplying the pressure sliding surface by a first linear coefficient yields the pressure linear approximation term, and multiplying the rotational speed sliding surface by a second linear coefficient yields the rotational speed linear approximation term. The pressure power-law approximation term is added to the pressure linear approximation term to obtain the pressure approximation law, and the speed power-law approximation term is added to the speed linear approximation term to obtain the speed approximation law.

8. The window cleaning robot cleaning execution method based on dynamic control according to claim 7, characterized in that, The step of raising the absolute value of the pressure sliding surface to the power of 0.5, multiplying it by the pressure sign function value, and then multiplying it by the first switching gain, yields the pressure power approximation term; raising the absolute value of the speed sliding surface to the power of 0.5, multiplying it by the speed sign function value, and then multiplying it by the second switching gain, yields the speed power approximation term, including: The absolute value of the pressure sliding surface is taken and raised to the power of 0.5 to obtain the basic value of the pressure power; the absolute value of the rotational speed sliding surface is taken and raised to the power of 0.5 to obtain the basic value of the rotational speed power. The pressure sign function value is determined based on the sign of the pressure sliding surface, and the rotational speed sign function value is determined based on the sign of the rotational speed sliding surface. The pressure power base value is multiplied by the pressure sign function value and then by the first switching gain to obtain the pressure power approximation term. The speed power base value is multiplied by the speed sign function value and then by the second switching gain to obtain the speed power approximation term.

9. The window cleaning robot cleaning execution method based on dynamic control according to claim 1, characterized in that, The calculation of the cylinder duty cycle output value based on the pressure approach law and centrifugal force compensation term, and the calculation of the motor voltage output value based on the speed approach law and friction torque, include: The current rotational speed of the window cleaning robot is subtracted from the rotational speed of the previous cycle, and then divided by the sampling period to obtain the rotational speed differential value. The product of the cleaning component mass, eccentricity, tilt angle cosine value, rotational speed value and the rotational speed differential value is divided by the maximum output force of the cylinder to obtain the centrifugal force compensation term. The cylinder duty cycle output value is obtained by sequentially adding the product of the suspended working point duty cycle value, the pressure proportional coefficient and the pressure approach law, and the centrifugal force compensation term. Divide the frictional torque by the motor torque constant to obtain the frictional torque compensation term. Then, add the voltage value at the suspended working point, the product of the speed proportional coefficient and the speed approach law, and the frictional torque compensation term in sequence to obtain the motor voltage output value.

10. A window cleaning robot cleaning execution system based on dynamic control, characterized in that, For performing the window cleaning robot cleaning method based on dynamic control as described in any one of claims 1-9, 9 includes: A setting module is used to calculate the stain density and set the target pressure and target rotation speed based on the stain density; The calculation module is used to calculate the dynamic friction coefficient based on the stain density, and multiply the dynamic friction coefficient by the window surface pressure and the radius of the cleaning cloth to obtain the friction torque; A construction module is used to obtain the actual pressure and construct a pressure sliding surface based on the target pressure and the actual pressure; The processing module is used to add the pressure sliding surface to the linear term of the pressure sliding surface to obtain the pressure approaching law, and to add the speed sliding surface to the linear term of the speed sliding surface to obtain the speed approaching law. The calculation module is used to calculate the cylinder duty cycle output value based on the pressure approach law and the centrifugal force compensation term, and to calculate the motor voltage output value based on the speed approach law and the friction torque.

Images