Photovoltaic cleaning robot positioning method and system based on UWB and RTK

Abstract

The invention provides a photovoltaic cleaning robot positioning method and system based on UWB and RTK, and the method comprises the steps: synchronously collecting ultra-wideband base station signal time difference data and real-time dynamic differential positioning data in the moving process of a robot, and generating initial three-dimensional positioning data through space-time matching; and calculating the three-dimensional offset compensation amount of the four corners relative to the photovoltaic panel according to the surface characteristics of the photovoltaic panel and the chassis posture of the robot. And in combination with ultra-wideband ranging errors, differential positioning deviation and inertial acceleration data, a dynamic weighted fusion algorithm is adopted to optimize a pose estimation value in real time. And for dynamic interference caused by a metal frame and a glass coating on the surface of the photovoltaic panel, establishing a signal propagation correction model to eliminate positioning drift, and finally outputting high-precision target three-dimensional positioning data adapted to a complex photovoltaic environment. The robot positioning precision and stability in the photovoltaic panel surface dynamic interference environment are improved.

Description

Technical Field

[0001] The present application relates to the field of UWB and RTK technology, and in particular to a photovoltaic cleaning robot positioning method and system based on UWB and RTK. Background Art

[0002] In photovoltaic power station maintenance, cleaning robots must achieve millimeter-level positioning accuracy across large photovoltaic panel arrays. The complex interference environment of photovoltaic panels, including reflections from metal frames, transmission from glass coverings, and sudden inclination angles between adjacent panels, requires positioning systems capable of dynamic obstacle recognition, real-time fusion of multi-source data, and 3D pose compensation. Traditional satellite positioning is severely affected by photovoltaic panel obstruction, and inertial navigation suffers from cumulative errors. Therefore, a collaborative positioning solution tailored to the unique structure of photovoltaic panels is urgently needed.

[0003] Existing solutions primarily utilize a vision-assisted inertial navigation system. This system utilizes a binocular camera and an inertial measurement unit (IMU) mounted on the robot chassis, and combines this with pre-defined photovoltaic panel edge feature points to calculate the robot's position and orientation. This system uses visual recognition to establish a relative coordinate system for the photovoltaic panel's frame, infers the robot's motion trajectory from inertial data, and finally corrects positioning errors through feature point matching.

[0004] This solution lacks stability in visual feature extraction under strong sunlight or reflective glass conditions. The uniform texture of the photovoltaic panel surface reduces the success rate of feature point matching, and it fails to effectively address the amplification of inertial navigation errors caused by sudden changes in the inclination angles of adjacent photovoltaic panels. The multipath effect caused by metal frames during dynamic cleaning further interferes with visual positioning accuracy, causing centimeter-level drift in the robot's path tracking. Summary of the Invention

[0005] The present application provides a photovoltaic cleaning robot positioning method and system based on UWB and RTK, which is used to solve the problems of low robot positioning accuracy and poor stability in the dynamic interference environment of the photovoltaic panel surface in the prior art.

[0006] In a first aspect, the present application provides a photovoltaic cleaning robot positioning method based on UWB and RTK, comprising: During the movement of the photovoltaic cleaning robot, the signal time difference data between ultra-wideband base stations, ultra-wideband ranging error, real-time dynamic differential positioning data, real-time dynamic differential positioning deviation and inertial acceleration data are obtained; Generating preliminary three-dimensional positioning data of the photovoltaic cleaning robot based on the signal time difference data and the real-time dynamic differential positioning data; Calculating the offset compensation of the four corners of the chassis of the photovoltaic cleaning robot relative to the surface of the photovoltaic panel based on the preliminary three-dimensional positioning data; Calculating a real-time pose estimation value of the photovoltaic cleaning robot by a dynamic weighting method based on the offset compensation amount, the ultra-wideband ranging error, the real-time dynamic differential positioning deviation, and the inertial acceleration data; Based on the real-time pose estimation value and combined with the real-time interference characteristics of dynamic obstacles on the surface of the photovoltaic panel, target three-dimensional positioning data of the photovoltaic cleaning robot is generated.

[0007] Optionally, the calculating of the real-time pose estimation value of the photovoltaic cleaning robot by a dynamic weighting method based on the offset compensation amount, the ultra-wideband ranging error, the real-time dynamic differential positioning deviation and the inertial acceleration data includes: Determining a dynamic error distribution parameter according to a real-time fluctuation range of the ultra-wideband ranging error; generating spatiotemporal correlation parameters based on the statistical characteristics of the real-time dynamic differential positioning deviation; Performing vector superposition of the offset compensation amount and the motion trend of the inertial acceleration data to generate a dynamic adjustment vector value; According to the dynamic error distribution parameter, the spatiotemporal correlation parameter and the dynamic adjustment vector value, the weight coefficients of the ultra-wideband ranging error, the real-time dynamic differential positioning deviation and the inertial acceleration data are distributed in real time to generate a real-time pose estimation value.

[0008] Optionally, the allocating in real time the weight coefficients of the ultra-wideband ranging error, the real-time dynamic differential positioning deviation, and the inertial acceleration data according to the dynamic error distribution parameter, the spatiotemporal correlation parameter, and the dynamic adjustment vector value to generate a real-time pose estimation value includes: Calculating a credibility weight factor of the ultra-wideband ranging error according to the dynamic error distribution parameter; Based on the spatiotemporal correlation parameters, generating a regional adaptation weight factor of the real-time dynamic differential positioning deviation; determining a motion consistency weight factor of the inertial acceleration data according to the dynamic adjustment vector value; Dynamically normalizing the credibility weight factor, the region adaptation weight factor, and the motion consistency weight factor to generate a weight coefficient set; Based on the weight coefficient set, the distance data after ultra-wideband ranging error correction, the coordinate data after real-time dynamic differential positioning deviation compensation, and the displacement data after inertial acceleration integration are weightedly fused to generate a real-time pose estimation value.

[0009] Optionally, based on the weight coefficient set, weighted fusion is performed on the distance data after ultra-wideband ranging error correction, the coordinate data after real-time dynamic differential positioning deviation compensation, and the displacement data after inertial acceleration integration to generate a real-time pose estimation value, including: Scaling the distance data after the ultra-wideband ranging error correction according to the ultra-wideband signal weight value in the weight coefficient set to generate a first weighted result; Scaling the coordinate data after the real-time dynamic differential positioning deviation compensation according to the real-time dynamic differential positioning weight value in the weight coefficient set to generate a second weighted result; Scaling the displacement data after the inertial acceleration integration according to the inertial acceleration weight value in the weight coefficient set to generate a third weighted result; Performing vector superposition on the first weighted result, the second weighted result, and the third weighted result to generate a fusion estimation value; Performing trajectory smoothing processing on the fused estimated value to generate a real-time pose estimated value.

[0010] Optionally, generating preliminary three-dimensional positioning data of the photovoltaic cleaning robot based on the signal time difference data and the real-time dynamic differential positioning data includes: Determining a dynamic distance difference between the photovoltaic cleaning robot and at least three ultra-wideband base stations based on the signal time difference data; Extracting the geographic coordinate reference point of the photovoltaic panel array where the photovoltaic cleaning robot is located from the real-time dynamic differential positioning data; Dynamically compensate and match the dynamic distance difference with the geographic coordinate reference point to establish a signal propagation path correction model related to the dynamic obstacle distribution on the photovoltaic panel surface; The signal propagation path correction model is used to eliminate signal reflection interference caused by the metal frame and glass covering on the surface of the photovoltaic panel, and generate preliminary three-dimensional positioning data.

[0011] Optionally, calculating the offset compensation amounts of the four corners of the chassis of the photovoltaic cleaning robot relative to the surface of the photovoltaic panel based on the preliminary three-dimensional positioning data includes: According to the preliminary three-dimensional positioning data, the real-time projection coordinates of the four corners of the chassis of the photovoltaic cleaning robot in the photovoltaic panel array coordinate system are obtained; Based on the dynamic change parameters of the photovoltaic panel installation inclination angle, the geometric projection relationship between the chassis plane of the photovoltaic cleaning robot and the photovoltaic panel surface is established; Calculate the offset distances of the four corners of the chassis of the photovoltaic cleaning robot in the normal direction of the photovoltaic panel surface based on the geometric projection relationship; Based on the offset distance, combined with the dynamic adjustment parameters of the movement direction of the photovoltaic cleaning robot and the rate of change of the photovoltaic panel inclination angle, an offset compensation amount is generated.

[0012] Optionally, generating target three-dimensional positioning data of the photovoltaic cleaning robot based on the real-time pose estimation value and in combination with real-time interference characteristics of dynamic obstacles on the surface of the photovoltaic panel includes: According to the real-time interference characteristics of dynamic obstacles on the surface of photovoltaic panels, the signal blocking area of ​​the edge of the dynamic obstacle in the movement direction of the photovoltaic cleaning robot is determined; Based on the real-time pose estimation value, a dynamic interference model between the motion trajectory of the photovoltaic cleaning robot and the signal blocking area is established; The positioning drift of the real-time pose estimation value in the signal blocking area is corrected by the dynamic interference model, and combined with the preset trajectory constraints to generate the target three-dimensional positioning data.

[0013] In a second aspect, the present application provides a photovoltaic cleaning robot positioning system based on UWB and RTK, comprising: The acquisition module is used to obtain the signal time difference data between ultra-wideband base stations, ultra-wideband ranging error, real-time dynamic differential positioning data, real-time dynamic differential positioning deviation and inertial acceleration data during the movement of the photovoltaic cleaning robot; A first generating module is used to generate preliminary three-dimensional positioning data of the photovoltaic cleaning robot based on the signal time difference data and the real-time dynamic differential positioning data; A first calculation module is used to calculate the offset compensation amount of the four corners of the chassis of the photovoltaic cleaning robot relative to the surface of the photovoltaic panel based on the preliminary three-dimensional positioning data; A second calculation module is configured to calculate a real-time pose estimation value of the photovoltaic cleaning robot by a dynamic weighted method based on the offset compensation amount, the ultra-wideband ranging error, the real-time dynamic differential positioning deviation, and the inertial acceleration data; The second generation module is used to generate target three-dimensional positioning data of the photovoltaic cleaning robot based on the real-time posture estimation value and the real-time interference characteristics of dynamic obstacles on the surface of the photovoltaic panel.

[0014] In a third aspect, the present application provides a computing device comprising a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute a photovoltaic cleaning robot positioning method based on UWB and RTK as described in any one of the first aspects.

[0015] In a fourth aspect, the present application provides a computer storage medium having computer program instructions stored thereon, which, when executed by a processor, implements a photovoltaic cleaning robot positioning method based on UWB and RTK as described in any one of the first aspects.

[0016] In the present application, a photovoltaic cleaning robot positioning method based on UWB and RTK is provided, which includes: during the movement of the photovoltaic cleaning robot, obtaining signal time difference data between ultra-wideband base stations, ultra-wideband ranging error, real-time dynamic differential positioning data, real-time dynamic differential positioning deviation and inertial acceleration data; generating preliminary three-dimensional positioning data of the photovoltaic cleaning robot based on the signal time difference data and the real-time dynamic differential positioning data; calculating the offset compensation amount of the four corners of the chassis of the photovoltaic cleaning robot relative to the surface of the photovoltaic panel based on the preliminary three-dimensional positioning data; calculating the real-time posture estimation value of the photovoltaic cleaning robot by a dynamic weighted method based on the offset compensation amount, the ultra-wideband ranging error, the real-time dynamic differential positioning deviation and the inertial acceleration data; based on the real-time posture estimation value, combined with the real-time interference characteristics of dynamic obstacles on the surface of the photovoltaic panel, generating the target three-dimensional positioning data of the photovoltaic cleaning robot.

[0017] The technical solution provided by this application has the following beneficial effects: This application provides heterogeneous measurement input for multi-source fusion positioning by synchronously collecting ultra-wideband time difference, differential positioning and inertial data, solving the problem that a single sensor is susceptible to interference on the surface of photovoltaic panels; combining ultra-wideband relative ranging and differential positioning absolute coordinates to establish a unified initial posture with a time and space reference, eliminating coordinate system jumps during large-scale positioning of photovoltaic panel arrays; dynamically adjusts the projection relationship of the four corners of the chassis based on the inclination of the photovoltaic panel to compensate for the robot posture measurement error caused by the tilt of the panel surface; through real-time weight distribution of multi-source data, suppresses ultra-wideband errors caused by metal frame reflections, differential positioning deviations caused by photovoltaic panel occlusions, and inertial navigation cumulative errors; and corrects the posture estimation value in combination with the interference characteristics of dynamic obstacles to ensure that the output positioning data is consistent with the physical space of the photovoltaic panel surface, avoiding clean path tracking deviations.

[0018] Furthermore, this application also determines the dynamic error distribution by analyzing the real-time fluctuations of ultra-wideband ranging errors, generates correlation parameters based on the spatiotemporal statistical characteristics of differential positioning deviations, superimposes the chassis offset compensation amount with the inertial motion trend vector to form a dynamic adjustment amount, and finally allocates multi-source data weights in real time based on the three types of parameters to achieve adaptive optimization of the posture estimation value of the photovoltaic panel in a complex environment.

[0019] In addition, the dynamic weight allocation mechanism effectively balances the reliability differences of ultra-wideband, differential positioning and inertial data in different interference areas, and achieves centimeter-level real-time pose estimation for specific interference sources such as metal reflection, glass transmission and sudden inclination angle changes on the photovoltaic panel surface, solving the problem of positioning drift in dynamic obstacle environments caused by existing solutions.

[0020] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0022] Figure 1 A flowchart of a photovoltaic cleaning robot positioning method based on UWB and RTK provided in an embodiment of the present application; Figure 2 A schematic diagram of a photovoltaic cleaning robot positioning system based on UWB and RTK provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0024] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0025] Existing positioning solutions for photovoltaic cleaning robots primarily rely on vision-assisted inertial navigation systems, which calculate their position and orientation by identifying feature points along the edges of photovoltaic panels. However, this solution suffers from visual feature extraction instability under strong sunlight or glass reflections, and the uniform surface texture of photovoltaic panels reduces the success rate of feature matching. When the inclination angles of adjacent photovoltaic panels change suddenly, inertial navigation errors accumulate due to the lack of external correction. Multipath effects caused by reflections from metal frames further cause centimeter-level drift in positioning data, ultimately rendering the robot's path tracking accuracy inadequate for efficient panel cleaning.

[0026] In response to the above-mentioned defects, this application proposes a positioning method for a photovoltaic cleaning robot based on UWB and RTK, which constructs a multi-source collaborative spatiotemporal alignment model by synchronously collecting ultra-wideband base station ranging data, differential positioning coordinates and inertial measurement data. Specifically, the anti-multipath interference characteristics of ultra-wideband signals are used to compensate for the reflection error of the metal frame, and the accumulated deviation of inertial navigation is corrected in combination with the absolute coordinates of differential positioning. The chassis offset compensation amount is then calculated in real time through the dynamic parameters of the photovoltaic panel inclination angle, and finally a high-precision three-dimensional pose is output based on the dynamic allocation of multi-source data weights. This method breaks through the failure bottleneck of the visual solution in strong light and low-texture environments. Through the spatiotemporal complementarity of ultra-wideband and differential positioning, the positioning accuracy under dynamic interference on the photovoltaic panel surface is improved to the millimeter level, effectively solving the problems of positioning drift and cleaning path deviation caused by environmental sensitivity in the existing technology.

[0027] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0028] Figure 1 A flowchart of a photovoltaic cleaning robot positioning method based on UWB and RTK is provided in an embodiment of the present application, as shown in FIG. Figure 1 As shown, the method includes: Step 101: During the movement of the photovoltaic cleaning robot, signal time difference data between ultra-wideband base stations, ultra-wideband ranging error, real-time dynamic differential positioning data, real-time dynamic differential positioning deviation, and inertial acceleration data are obtained.

[0029] In step 101, the ultra-wideband base stations are deployed in a distributed manner, with at least three base stations fixedly installed in a non-collinear manner on the peripheral brackets or frames of the photovoltaic panel array. The spacing between the base stations is dynamically adjusted according to the size of the photovoltaic panel and the signal coverage range, ensuring that the robot can simultaneously receive ultra-wideband signals from at least three base stations at any position on the photovoltaic panel surface. The signal time difference data represents the time difference between the ultra-wideband base station transmitted signal reaching the robot receiving end, reflecting the relative distance relationship between the robot and each base station. The ultra-wideband ranging error represents the deviation between the distance measurement value and the true value caused by multipath effects, clock synchronization deviation, etc. The real-time dynamic differential positioning data represents the positioning error correction data between the base station and the mobile station, providing an absolute geographic coordinate reference. The real-time dynamic differential positioning deviation represents the systematic offset between the differential positioning solution result and the true position. The inertial acceleration data represents the three-axis acceleration change collected by the robot's built-in inertial measurement unit.

[0030] In an embodiment of the present application, during the movement of the robot, the ultra-wideband base station periodically transmits a pulse signal, the robot receiving module records the arrival time difference of each base station signal and calculates the original ranging value, synchronously collects the latitude and longitude coordinates and correction data output by the differential positioning receiver, and the inertial measurement unit uploads the three-axis acceleration data in real time; the ranging module generates the current ultra-wideband ranging error estimate based on the historical error record, and the differential positioning system calculates the real-time positioning deviation by comparing the base station data, and finally packages all the data for subsequent processing.

[0031] For example, when a photovoltaic cleaning robot moves along the X-axis on a photovoltaic panel surface with a 10-degree inclination angle, the three ultra-wideband base stations measure the signal arrival time differences as T1, T2, and T3, respectively. After time-to-distance conversion, the original ranging values ​​L1, L2, and L3 are obtained. The historical error library shows that the average ranging error in the current environment is E; the differential positioning receiver outputs the coordinates (X1, Y1), and the deviation (ΔX, ΔY) is obtained by comparing with the base station data; the inertial unit measures the X-axis acceleration a, Y-axis acceleration b, and Z-axis acceleration c.

[0032] Step 102: Generate preliminary three-dimensional positioning data of the photovoltaic cleaning robot based on the signal time difference data and the real-time dynamic differential positioning data.

[0033] In step 102 , the preliminary three-dimensional positioning data represents an estimated value of the robot's initial pose that is a fusion of ultra-wideband relative distance and differential positioning absolute coordinates.

[0034] In an embodiment of the present application, the time difference data of the ultra-wideband signal is converted into the geometric distance relationship between the robot and each base station, and the distance equation group is constructed in combination with the base station deployment coordinates to solve the initial relative position of the robot; at the same time, the elevation information and plane coordinates in the differential positioning data are extracted, and the absolute position is mapped to the local coordinate system of the photovoltaic panel array through coordinate transformation; finally, the ultra-wideband relative positioning result and the differential positioning absolute coordinate are weighted averaged to eliminate the systematic deviation of a single data source, and output the three-dimensional position estimate in the photovoltaic panel surface coordinate system.

[0035] For example, based on the ranging values ​​L1, L2, L3 and the base station coordinates (Xa, Ya, Za), (Xb, Yb, Zb), (Xc, Yc, Zc), the equation group √[(X-Xa)²+(Y-Ya)²+(Z-Za)²]=L1+E is established to solve the robot's initial position (X, Y, Z), where E is the average ranging error of the current environment; at the same time, the differential positioning coordinates (X1, Y1) are converted to the photovoltaic panel coordinate system to obtain (X', Y'). The final fusion result is (X×0.6+X'×0.4) on the X axis and (Y×0.6+Y'×0.4) on the Y axis, and the Z axis takes the differential positioning elevation value Z1.

[0036] Step 103: Calculate the offset compensation of the four corners of the chassis of the photovoltaic cleaning robot relative to the surface of the photovoltaic panel based on the preliminary three-dimensional positioning data.

[0037] In step 103, the offset compensation amount represents the normal direction position correction amount of the four corners of the robot chassis caused by the change in the inclination angle of the photovoltaic panel.

[0038] In an embodiment of the present application, the angle between the robot chassis plane and the photovoltaic panel surface is calculated based on the height value Z in the preliminary three-dimensional positioning data and combined with the real-time data of the photovoltaic panel inclination sensor; the projection deviation of the four corners of the chassis in the normal direction of the photovoltaic panel is solved respectively through the geometric projection relationship; then, according to the current movement direction of the robot and the historical inclination change trend, the compensation amount is dynamically smoothed and filtered, and the three-dimensional compensation vectors of the four corners are output.

[0039] For example, the robot's current position Z = 1.2 meters, the photovoltaic panel inclination angle is 15 degrees, and the chassis size is 0.8×0.5 meters; the front left corner projection deviation ΔZ = 0.8×sin15° = 0.21 meters, combined with the X-axis component of the motion direction 0.9, the final compensation amount is (0, 0, 0.21×0.9) = (0, 0, 0.19).

[0040] Step 104: Calculate a real-time pose estimation value of the photovoltaic cleaning robot by a dynamic weighting method based on the offset compensation, the ultra-wideband ranging error, the real-time dynamic differential positioning deviation, and the inertial acceleration data.

[0041] In step 104, the dynamic weighting approach assigns fusion weights based on the real-time reliability of each data source. The real-time pose estimate is the robot's current three-dimensional position and pose in the photovoltaic panel surface coordinate system, output through the dynamic weighted fusion of ultra-wideband corrected distance, differentially compensated coordinates, and inertial displacement data. This estimate combines the real-time measurement characteristics of multiple sensor sources with historical error patterns to represent the robot's optimal pose in complex interference environments.

[0042] In an embodiment of the present application, the current credibility weight is calculated based on the statistical distribution characteristics of the ultra-wideband ranging error, the spatiotemporal correlation of the differential positioning deviation is analyzed to generate the regional adaptability weight, and the vector consistency test of the integral result of the inertial acceleration data and the offset compensation amount is performed to obtain the motion weight; after normalizing the three types of weights, the ultra-wideband corrected distance, differential compensation coordinates and inertial displacement are weightedly summed respectively, the robot posture state equation is iteratively updated, and the optimal posture estimate is output.

[0043] For example, the ultra-wideband weight is 0.5 (small error E), the differential weight is 0.3 (little occlusion in the current area), and the inertia weight is 0.2 (motion trend matching); after weighted fusion, the pose X = 2.1 × 0.5 + 2.3 × 0.3 + 2.0 × 0.2 = 2.14 meters.

[0044] Step 105: Based on the real-time pose estimation value and in combination with the real-time interference characteristics of dynamic obstacles on the surface of the photovoltaic panel, target three-dimensional positioning data of the photovoltaic cleaning robot is generated.

[0045] In step 105, the real-time interference characteristics of dynamic obstacles refer to the spatiotemporal distribution of interference caused by reflections, occlusions, and other interferences to the positioning signal from objects such as metal frames and dust accumulation areas on the photovoltaic panel surface. These characteristics are obtained by establishing a mapping relationship library between obstacle positions and error values ​​based on historical positioning data, matching abnormal drift components in the current pose estimate with characteristic patterns in the library in real time, and combining ultra-wideband signal strength mutation detection with differential positioning lock status to determine the boundaries of the dynamic interference area. The target's three-dimensional positioning data represents the final high-precision positioning result after correcting for dynamic obstacle interference.

[0046] In an embodiment of the present application, abnormal drift components caused by dynamic obstacles such as metal frames and dust accumulation areas on the surface of photovoltaic panels in real-time pose estimation are identified; a mapping relationship library between obstacle distribution and positioning errors is established, and regional compensation is performed on the pose estimation value; finally, the correction result is checked for consistency with the spatial constraints of the preset cleaning path, and positioning data that conforms to the physical boundaries of the photovoltaic panel is output.

[0047] For example, it is detected that the X-axis of the posture suddenly increases by 0.15 meters near the metal frame. The error database is queried to obtain the typical compensation amount in this area -0.12 meters. After correction, X=2.14-0.12=2.02 meters. After path verification, it does not exceed the effective width of the photovoltaic panel of 2.5 meters. The final output is (2.02,1.33,1.19).

[0048] This method effectively suppresses positioning errors caused by metal reflection, glass transmission, and sudden inclination angle changes on the surface of photovoltaic panels by complementing the spatiotemporal data of ultra-wideband and differential positioning. The dynamic weight allocation mechanism adapts to the interference characteristics of different areas, ensuring that the robot's three-dimensional positioning results are always consistent with the physical space of the photovoltaic panel. The ultimately achieved millimeter-level positioning accuracy ensures the accuracy of clean path tracking and solves the positioning instability problem of existing solutions in complex photovoltaic environments.

[0049] To address the positioning accuracy issue of a photovoltaic cleaning robot in a complex photovoltaic panel surface environment, in some embodiments, step 104: calculating a real-time pose estimate of the photovoltaic cleaning robot by a dynamic weighting method based on the offset compensation, the ultra-wideband ranging error, the real-time dynamic differential positioning deviation, and the inertial acceleration data, includes: Step 201: Determine a dynamic error distribution parameter according to the real-time fluctuation range of the ultra-wideband ranging error.

[0050] In step 201, the real-time fluctuation range of the UWB ranging error refers to the variation range between the UWB signal ranging value and the actual distance in the current environment. This range is obtained by real-time monitoring of the raw ranging data between the UWB base station and the robot tag, establishing an error distribution model based on historical calibration data, dynamically adjusting the error boundary based on the distribution characteristics of the metal structure on the photovoltaic panel surface, and ultimately outputting an error fluctuation range parameter reflecting the signal interference intensity in the current area. The dynamic error distribution parameter is a quantitative indicator reflecting the fluctuation characteristics of the UWB ranging error in different photovoltaic panel areas and is used to assess the reliability of the UWB signal in the current environment.

[0051] In this embodiment, by analyzing the temporal variation of UWB ranging errors and combining it with prior knowledge of the distribution of metal structures on the photovoltaic panel surface, a probability distribution model for the variation of errors with location was established, thereby determining the dynamic error distribution parameters for the current region. This parameter is used to assess the credibility of the UWB data during subsequent weighting.

[0052] Step 202: Generate spatiotemporal correlation parameters based on the statistical characteristics of the real-time dynamic differential positioning deviation.

[0053] In step 202, the statistical characteristics of the real-time dynamic differential positioning deviation refer to the spatiotemporal distribution of positioning deviations generated by the differential positioning system in different areas of the photovoltaic panel. These characteristics are generated by continuously recording the deviations between the differential positioning coordinates and the reference station calibration data, calculating the mean and variance of the deviations by grid area on the photovoltaic panel surface, and analyzing the correlation between the deviations and obstacle positions based on the robot's motion trajectory. Ultimately, spatiotemporal characteristic parameters reflecting the reliability of regional positioning are established. The spatiotemporal correlation parameter is a measure of the spatiotemporal correlation between the real-time dynamic differential positioning deviations and the dynamic obstacle distribution on the photovoltaic panel surface, and is used to characterize the adaptability of the differential positioning system in the current environment.

[0054] In this embodiment, the distribution characteristics of real-time dynamic differential positioning deviations in different areas of the photovoltaic panel are statistically analyzed. Combined with the spatiotemporal relationship between the robot's motion trajectory and the location of obstacles, a spatiotemporal correlation parameter is generated that reflects the correlation between positioning deviations and obstacle interference. This parameter serves as an important basis for weighting the differential positioning data.

[0055] Step 203: Perform vector superposition on the offset compensation amount and the motion trend of the inertial acceleration data to generate a dynamic adjustment vector value.

[0056] In step 203, the offset compensation and the motion trend of the inertial acceleration data are superimposed using vectors. The three-dimensional offset compensation at the four corners of the chassis is converted into displacement vectors. The inertial acceleration data is then time-integrated to obtain a motion trend vector. These vectors are then synthesized in the local coordinate system of the photovoltaic panel surface. During superposition, a dynamic scaling factor is applied to the vertical component based on the current inclination angle of the photovoltaic panel. In a specific embodiment, when a photovoltaic cleaning robot moves along the X-axis on a photovoltaic panel surface with a 15° inclination, the front of the chassis generates a Z-axis offset compensation of 0.12 meters due to the inclination. This is converted into a displacement vector (X: 0, Y: 0, Z: 0.12). The inertial acceleration is then integrated to obtain a motion trend vector (X: 0.8 meters, Y: 0, Z: 0.2 meters). A scaling factor of 0.6 is applied to the Z-axis component based on the inclination angle, resulting in a superposition result of (X: 0.8 meters, Y: 0 meters, Z: 0.12 + 0.2 × 0.6 = 0.24 meters). The dynamic adjustment vector value refers to the posture correction vector obtained by fusing the chassis offset compensation and the inertial motion trend, which is used to compensate for the positioning deviation caused by the change in the inclination angle of the photovoltaic panel.

[0057] In this embodiment, the offset compensation values ​​for the chassis' four corners are converted into three-dimensional correction vectors. Simultaneously, the inertial acceleration data is integrated to generate a motion trend vector. These vectors are then superimposed in the photovoltaic panel surface coordinate system to generate a dynamic adjustment vector value reflecting the robot's actual motion state. This vector value is used for motion consistency verification during subsequent multi-source data fusion.

[0058] Step 204: Based on the dynamic error distribution parameter, the spatiotemporal correlation parameter, and the dynamic adjustment vector value, the weight coefficients of the ultra-wideband ranging error, the real-time dynamic differential positioning deviation, and the inertial acceleration data are distributed in real time to generate a real-time pose estimation value.

[0059] In step 204, real-time allocation refers to dynamically adjusting the contribution of each positioning data source in pose estimation according to its real-time reliability.

[0060] In an embodiment of the present application, the ultra-wideband data weight is calculated based on the dynamic error distribution parameters, the differential positioning data weight is determined according to the spatiotemporal correlation parameters, the inertial data weight is evaluated with reference to the dynamic adjustment vector value, and the fusion weight coefficient of the three is obtained through normalization processing. Finally, all types of data are weightedly fused to output the optimal pose estimation value.

[0061] Here's a specific example: When the photovoltaic cleaning robot moves along the X-axis on the surface of a photovoltaic panel with a 10-degree inclination, the three ultra-wideband base stations measure the signal arrival time difference, which is converted into the ranging value L1 of 3.2 meters, L2 of 3.5 meters, and L3 of 3.8 meters. The historical error library shows that the average ranging error E in the current environment is 0.05 meters. The differential positioning receiver outputs coordinates X1 equal to 5.1 meters and Y1 equal to 2.4 meters. Compared with the base station data, the deviation ΔX is 0.02 meters and ΔY is 0.03 meters. The inertial unit measures the acceleration of the X axis of 0.3 meters per square second, the acceleration of the Y axis of 0.1 meters per square second, and the acceleration of the Z axis of 0.2 meters per square second. Based on the ranging values ​​and the base station coordinates, the equation group is established. The initial position is obtained by solving X equal to 2.1 meters, Y equal to 1.2 meters, and Z equal to 1.5 meters. At the same time, the differential positioning coordinates are converted to the photovoltaic panel coordinate system, and X' equals 2.08 meters and Y' equals 1.18 meters. The final fusion result is , , the Z axis takes the differential positioning elevation value of 1.5 meters. Based on the current position Z equal to 1.5 meters and the photovoltaic panel inclination angle of 15 degrees, calculate the projection deviation of the robot chassis front left corner , combined with the X-axis component of the motion direction 0.9, generate the Z-axis compensation . The analysis shows that the fluctuation range of the ultra-wideband ranging error is between 0.03 meters and 0.07 meters, and the dynamic error distribution parameter is determined to be 0.5; the standard deviation of the statistical differential positioning deviation near the current position is 0.04 meters, and the spatiotemporal correlation parameter is generated as 0.3; the compensation amount is superimposed on the inertial data integration result to obtain a dynamic adjustment vector value of 0.2. After assigning weights according to the parameters, the ultra-wideband correction value of 2.092 meters is given a weight of 0.5, the differential compensation value of 2.08 meters is given a weight of 0.3, and the inertial estimation value of 2.05 meters is given a weight of 0.2. The weighted fusion is used to obtain the real-time posture. .

[0062] In an embodiment of the present application, adaptive fusion of multi-source positioning data is achieved through the collaborative calculation of dynamic error distribution parameters, spatiotemporal correlation parameters and dynamic adjustment vector values, so that the pose estimation value can adapt to the interference characteristics of different areas on the surface of the photovoltaic panel in real time, thereby improving the positioning stability and accuracy in complex environments.

[0063] In order to solve the weight distribution problem of multi-source positioning data fusion of the photovoltaic cleaning robot in a complex environment, in some embodiments, step 204: the weight coefficients of the ultra-wideband ranging error, the real-time dynamic differential positioning deviation, and the inertial acceleration data are distributed in real time based on the dynamic error distribution parameter, the spatiotemporal correlation parameter, and the dynamic adjustment vector value to generate a real-time pose estimation value, including: Step 301: Calculate the credibility weight factor of the ultra-wideband ranging error according to the dynamic error distribution parameter.

[0064] In step 301, the credibility weight factor refers to a quantitative indicator reflecting the reliability of ultra-wideband ranging data in the current environment.

[0065] In this embodiment, the reliability of the UWB ranging data in the current region is calculated by combining the error fluctuation characteristics of the UWB signal, as characterized by the dynamic error distribution parameters, with prior knowledge of the distribution of metal structures on the photovoltaic panel surface. The higher the reliability, the larger the weight factor. This factor serves as a measure of the importance of the UWB data in the fusion process.

[0066] Step 302: Generate a regional adaptation weight factor of the real-time dynamic differential positioning deviation based on the spatiotemporal correlation parameters.

[0067] In step 302, the regional adaptation weight factor refers to an evaluation parameter reflecting the applicability of the real-time dynamic differential positioning data in the current region.

[0068] In this embodiment, the differential positioning data quality of the robot's current area is analyzed based on the relationship between the differential positioning deviation described by the spatiotemporal correlation parameters and the distribution of obstacles on the photovoltaic panel surface. The smaller the obstacle interference, the larger the regional adaptation weight factor. This factor ensures that the differential positioning data is appropriately weighted in different areas.

[0069] Step 303: Determine a motion consistency weight factor of the inertial acceleration data according to the dynamic adjustment vector value.

[0070] In step 303, the motion consistency weight factor refers to a parameter for evaluating the degree of consistency between the inertial measurement data and the actual motion state of the robot.

[0071] In this embodiment, the consistency of the inertial acceleration data integration results is verified by dynamically adjusting the chassis posture change trend reflected by the vector value. The higher the matching degree, the larger the motion consistency weight factor. This factor is used to ensure the reliability of the inertial data when the motion state suddenly changes.

[0072] Step 304: Dynamically normalize the credibility weight factor, the region adaptation weight factor, and the motion consistency weight factor to generate a weight coefficient set.

[0073] In step 304, the weight coefficient set refers to the final fusion weight combination of the three types of data sources after normalization.

[0074] In this embodiment, the credibility weight factor, regional adaptation weight factor, and motion consistency weight factor are normalized so that their sum is a fixed value. This generates a weight distribution scheme for ultra-wideband data, differential positioning data, and inertial data. This set ensures a balanced and reasonable contribution of each type of data during fusion.

[0075] Step 305: Based on the weight coefficient set, the distance data after ultra-wideband ranging error correction, the coordinate data after real-time dynamic differential positioning deviation compensation, and the displacement data after inertial acceleration integration are weightedly fused to generate a real-time pose estimation value.

[0076] In step 305, the process for generating distance data after ultra-wideband ranging error correction involves performing reflection path compensation and noise filtering on the original ranging values ​​based on the signal time difference data between the ultra-wideband base stations, using the signal strength attenuation pattern recorded in the dynamic error distribution parameters and the interference characteristics of the metal structure on the photovoltaic panel surface, and outputting error-corrected actual distance data from the robot to each base station. The process for generating coordinate data after real-time dynamic differential positioning deviation compensation involves interpolating and compensating for positioning points obscured by the photovoltaic panel frame based on the longitude and latitude coordinates provided by the real-time dynamic differential positioning data, combined with the mapping relationship between the photovoltaic panel array layout and the RTK signal obstruction area in the spatiotemporal correlation parameters, and outputting deviation-corrected absolute coordinate data for the robot. The process for generating displacement data after inertial acceleration integration involves performing attitude decoupling calculations using the three-axis acceleration values ​​of the inertial acceleration data and the correlation parameters between the chassis inclination angle and the motion direction in the dynamic adjustment vector. The acceleration data is then integrated in the time domain to eliminate accumulated errors and output the robot's relative displacement data relative to the previous moment. Weighted fusion is the process of comprehensively calculating multi-source positioning data based on weight coefficients.

[0077] In an embodiment of the present application, the distribution ratio in the weight coefficient set is used to perform weighted calculations on the distance data after ultra-wideband correction, the coordinate data after differential compensation, and the displacement data after inertial integration, and the optimal real-time posture estimation result of the robot is output through iterative optimization through the state estimation method.

[0078] Here's a specific example: When the photovoltaic cleaning robot moves along the X-axis on the photovoltaic panel surface with an inclination angle of 10 degrees, based on the dynamic error distribution parameter 0.5, the spatiotemporal correlation parameter 0.3 and the dynamic adjustment vector value 0.2 obtained in the previous step, the credibility weight factor of the ultra-wideband ranging error is first calculated according to the dynamic error distribution parameter 0.5, and the credibility weight factor is obtained using the parameter square relationship. Then, the regional adaptation weight factor of the real-time dynamic differential positioning deviation is generated based on the spatiotemporal correlation parameter 0.3. ; Then determine the motion consistency weight factor of the inertial acceleration data based on the dynamic adjustment vector value 0.2, The three weight factors 0.25, 0.36 and 0.3 are dynamically normalized so that the sum is 1. , , Based on this weight coefficient set, the distance data of 2.092 meters after ultra-wideband correction is assigned a weight of 0.275, the coordinate data of 2.08 meters after differential compensation is assigned a weight of 0.396, and the displacement data of 2.05 meters after inertial integration is assigned a weight of 0.330, and weighted fusion calculation is performed: , , , the three are added together to get the real-time pose estimation value , the final output is rounded to 2.075 meters.

[0079] In an embodiment of the present application, adaptive weight distribution of multi-source positioning data is achieved through comprehensive analysis of dynamic error distribution parameters, spatiotemporal correlation parameters and dynamic adjustment vector values, so that the pose estimation results can intelligently adapt to the interference characteristics of different areas on the surface of the photovoltaic panel, thereby improving the robustness and accuracy of the positioning system in complex environments.

[0080] In order to solve the problem of accuracy of multi-source data fusion of photovoltaic cleaning robots in dynamic environments, in some embodiments, step 305: based on the weight coefficient set, weighted fusion is performed on the distance data after ultra-wideband ranging error correction, the coordinate data after real-time dynamic differential positioning deviation compensation, and the displacement data after inertial acceleration integration to generate a real-time pose estimate, including: Step 401: scaling the distance data after the ultra-wideband ranging error correction according to the ultra-wideband signal weight value in the weight coefficient set to generate a first weighted result.

[0081] In step 401, the UWB signal weight value refers to the proportion of UWB ranging data in the fusion positioning. It is calculated by calculating the credibility weight factor using the dynamic error distribution parameter, and then normalized with the regional adaptation weight factor and the motion consistency weight factor. It reflects the reliability of the UWB signal in the current environment. The smaller the error, the higher the weight. The first weighted result is the positioning component of the UWB ranging data after error correction, adjusted by the weight ratio.

[0082] In this embodiment, the three-dimensional distance information from the ultra-wideband base station ranging data, after multipath error compensation, is proportionally scaled according to the ultra-wideband signal weight value determined in the weight coefficient set to generate a position component that reflects the ultra-wideband positioning contribution. This result preserves the high accuracy of the ultra-wideband signal in interference-free areas.

[0083] Step 402: scaling the coordinate data after the real-time dynamic differential positioning deviation compensation according to the real-time dynamic differential positioning weight value in the weight coefficient set to generate a second weighted result.

[0084] In step 402, the real-time dynamic differential positioning weight value refers to the contribution ratio of the differential positioning data to the fusion positioning. It is determined by normalizing the regional adaptation weight factor generated based on the spatiotemporal correlation parameters, along with the credibility weight factor and the motion consistency weight factor. It reflects the applicability of differential positioning in the current area. The lower the environmental interference, the higher the weight. The second weighted result refers to the positioning component of the differential positioning coordinate data after bias compensation, adjusted according to the weight ratio.

[0085] In this embodiment, the absolute coordinate data from differential positioning, which has undergone reference station error correction, is proportionally adjusted according to the differential positioning weights assigned in the weight coefficient set to generate position components that reflect the reliability of the differential positioning system. This result ensures the reasonable contribution of the absolute coordinate reference in the fusion.

[0086] Step 403: scaling the displacement data after the inertial acceleration integration according to the inertial acceleration weight value in the weight coefficient set to generate a third weighted result.

[0087] In step 403, the inertial acceleration weight refers to the proportion of inertial measurement data in fused positioning. This weight is calculated based on the dynamic adjustment vector value, and then normalized with the credibility weight and regional adaptation weight. It represents the degree of match between the inertial data and the robot's actual motion state. Higher consistency results in a greater weight. The third weighted result is the motion component adjusted according to the weight ratio after integrating the inertial measurement data.

[0088] In this embodiment, the acceleration data collected by the inertial measurement unit is time-integrated to obtain a displacement estimate, which is then scaled according to the inertial data weights in the weight coefficient set to generate a position component that reflects the robot's instantaneous motion state. This result maintains the continuity of the motion process.

[0089] Step 404: Perform vector superposition on the first weighted result, the second weighted result, and the third weighted result to generate a fusion estimation value.

[0090] In step 404, the fusion estimation value refers to the synthesis value of the weighted results of multi-source positioning data.

[0091] In this embodiment, the ultra-wideband component of the first weighted result, the differential positioning component of the second weighted result, and the inertial component of the third weighted result are vector-superimposed in the photovoltaic panel surface coordinate system to generate a preliminary fused position estimate. This step achieves unified integration of positioning data in different coordinate systems.

[0092] Step 405: Perform trajectory smoothing on the fused estimated value to generate a real-time pose estimated value.

[0093] In step 405 , trajectory smoothing refers to the process of performing time sequence optimization on the fusion result.

[0094] In an embodiment of the present application, based on the robot kinematic model, the current fusion estimation value and the historical posture data are optimized by sliding window to eliminate instantaneous jumps and ensure trajectory continuity, and finally output a real-time posture estimation value that conforms to the physical motion laws of the photovoltaic panel surface.

[0095] Here's a specific example: When the photovoltaic cleaning robot moves along the X-axis on the photovoltaic panel surface with an inclination angle of 10 degrees, based on the ultra-wideband weight coefficient 0.275, the difference weight coefficient 0.396 and the inertia weight coefficient 0.330 calculated in the previous steps, the distance data of 2.092 meters after the ultra-wideband ranging error correction is first scaled by the weight 0.275. The calculation process is as follows: ; Then the coordinate data 2.08 meters after real-time dynamic differential positioning deviation compensation is scaled by weight 0.396. The calculation process is: ; Then the displacement data of 2.05 meters after the inertial acceleration integration is scaled by the weight 0.330. The calculation process is The three weighted results are vector superimposed, and the calculation process is: Finally, the trajectory of the fusion estimate is smoothed, and the first three moments of the pose data 2.076 meters, 2.077 meters and 2.075 meters are taken for sliding average. The calculation process is: .

[0096] In the embodiment of the present application, through the step-by-step weighting and vector fusion of multi-source data, the relative positioning accuracy of the ultra-wideband signal is retained, the absolute coordinate advantage of differential positioning is integrated, and the motion continuity of inertial measurement is taken into account. The final generated pose estimation value can accurately reflect the actual motion state of the robot on the surface of the photovoltaic panel, effectively overcoming the limitations of a single sensor in a complex environment.

[0097] In order to solve the problem of insufficient initial positioning accuracy of the photovoltaic cleaning robot in complex environments, in some embodiments, step 102: generating preliminary three-dimensional positioning data of the photovoltaic cleaning robot based on the signal time difference data and the real-time dynamic differential positioning data, includes: Step 501: Based on the signal time difference data, determine the dynamic distance difference of the photovoltaic cleaning robot relative to at least three ultra-wideband base stations.

[0098] In step 501, the dynamic distance difference refers to the real-time relative distance change between the photovoltaic cleaning robot and each ultra-wideband base station.

[0099] In this embodiment, the dynamic distance difference between the robot and each base station is calculated by measuring the propagation time difference of the ultra-wideband signal from each base station to the robot and combining it with the signal propagation speed. These differences reflect the relative position of the robot to the fixed base station as it moves across the photovoltaic panel surface.

[0100] Step 502: extracting the geographic coordinate reference point of the photovoltaic panel array where the photovoltaic cleaning robot is located from the real-time dynamic differential positioning data.

[0101] In step 502 , the geographic coordinate reference point refers to the absolute position reference point of the photovoltaic panel array in the earth coordinate system.

[0102] In an embodiment of the present application, the latitude, longitude and elevation information of the photovoltaic panel array corresponding to the current position of the robot is extracted from the data output by the real-time dynamic differential positioning system, and converted into the reference coordinate point in the local coordinate system of the photovoltaic panel array to provide an absolute position reference for subsequent data fusion.

[0103] Step 503: Dynamically compensate and match the dynamic distance difference with the geographic coordinate reference point to establish a signal propagation path correction model related to the dynamic obstacle distribution on the photovoltaic panel surface.

[0104] In step 503, dynamic compensation matching refers to the process of spatially and temporally aligning relative distance data with absolute coordinate data. The photovoltaic panel array is the overall structural framework to which the photovoltaic panel surface is attached. The photovoltaic panel surface is the specific working plane within the array where the cleaning robot needs to be positioned. The signal propagation path correction model is a positioning compensation model that eliminates multipath interference.

[0105] In an embodiment of the present application, based on the distribution of dynamic obstacles on the surface of the photovoltaic panel, a spatial mapping relationship between the ultra-wideband signal propagation path and the differential positioning coordinates is established, and the two types of data are unified into the photovoltaic panel surface coordinate system through coordinate transformation, providing a benchmark for signal correction.

[0106] Step 504: Eliminate signal reflection interference caused by the metal frame and glass covering on the surface of the photovoltaic panel through the signal propagation path correction model to generate preliminary three-dimensional positioning data.

[0107] In step 504, the elimination process specifically involves establishing a signal reflection parameter library based on the known physical dimensions of the metal frame on the photovoltaic panel surface and the dielectric properties of the glass covering. Using the signal propagation path correction model, the multipath reflection components in the ultra-wideband signal are identified. The direct signal and the reflected signal are separated using the spatial geometric relationship between the dynamic distance difference and the geographic coordinate reference point. The reflected signal components are then subjected to reverse phase compensation and removed. Finally, the output is three-dimensional positioning data containing only the corrected direct signal. In an embodiment, when the robot is 1.5 meters from the photovoltaic panel frame, the model detects a 25cm reflected signal deviation in the X-axis direction. Reverse compensation is performed based on the 12ns phase delay corresponding to this position in the reflection parameter library. After elimination, the positioning deviation is reduced to 3cm.

[0108] In an embodiment of the present application, based on the physical properties of the metal frame and glass covering on the surface of the photovoltaic panel, a feature library of signal reflection interference is established. By matching the real-time signal characteristics with the patterns in the interference library, the multipath interference components are identified and eliminated, and three-dimensional positioning data compensated for the environmental characteristics is output.

[0109] Here's a specific example: When the photovoltaic cleaning robot moves along the X-axis on the surface of a photovoltaic panel with a 10-degree inclination angle, the three ultra-wideband base stations measure the signal arrival time difference, which is converted into the ranging value L1 of 3.2 meters, L2 of 3.5 meters, and L3 of 3.8 meters. The historical error library shows that the average ranging error E in the current environment is 0.05 meters; the differential positioning receiver outputs coordinates X1 equal to 5.1 meters and Y1 equal to 2.4 meters. Compared with the base station data, the deviation ΔX is 0.02 meters and ΔY is 0.03 meters. Based on the ranging value and the base station coordinates, the equation is established. 、 、 , the robot's initial position is obtained by solving X = 2.1 meters, Y = 1.2 meters, and Z = 1.5 meters. At the same time, the differential positioning coordinates are converted to the photovoltaic panel coordinate system, and X' is equal to 2.08 meters and Y' is equal to 1.18 meters. The final fusion result is , , The metal frame area features were detected through the signal propagation path correction model. After compensating the X-axis data by negative 0.12 meters, preliminary three-dimensional positioning data of 1.972 meters on the X axis, 1.192 meters on the Y axis, and 1.5 meters on the Z axis were obtained. This result served as the reference input for subsequent pose calculations.

[0110] In the embodiment of the present application, through the coordinated processing of ultra-wideband relative ranging and differential absolute positioning, the high-precision ranging characteristics of ultra-wideband are utilized, and the absolute coordinate advantages of differential positioning are combined, effectively overcoming the interference of the complex environment on the surface of the photovoltaic panel on the single positioning system, and laying a reliable basic data for subsequent precise positioning.

[0111] To solve the posture compensation problem of the photovoltaic cleaning robot on the inclined photovoltaic panel surface, in some embodiments, step 103: calculating the offset compensation amount of the four corners of the chassis of the photovoltaic cleaning robot relative to the photovoltaic panel surface based on the preliminary three-dimensional positioning data, includes: Step 601: Based on the preliminary three-dimensional positioning data, obtain the real-time projection coordinates of the four corners of the chassis of the photovoltaic cleaning robot in the photovoltaic panel array coordinate system.

[0112] In step 601, the photovoltaic array coordinate system is a three-dimensional coordinate system established based on the physical structure of the photovoltaic array to describe the robot's position. The real-time projected coordinates refer to the mapped positions of the four corners of the chassis in the photovoltaic array coordinate system.

[0113] In an embodiment of the present application, the robot center position is determined based on preliminary 3D positioning data, combined with the known chassis structure dimensions, to calculate the 3D coordinate projections of the four corners in the photovoltaic array coordinate system, providing a reference point for subsequent offset compensation. The specific process is as follows: Using the robot center coordinates (Xc, Yc, Zc) given by the preliminary 3D positioning data as a reference, based on the known dimensions of the chassis length L and width W, and in accordance with the correspondence between the chassis coordinate system and the photovoltaic array coordinate system, the front left corner coordinates (Xc+L / 2, Yc+W / 2, Zc), front right corner coordinates (Xc+L / 2, Yc-W / 2, Zc), rear left corner coordinates (Xc-L / 2, Yc+W / 2, Zc), and rear right corner coordinates (Xc-L / 2, Yc-W / 2, Zc) are calculated, forming a set of 3D coordinate projections of the four corners in the photovoltaic array coordinate system.

[0114] Step 602: Based on the dynamic change parameters of the photovoltaic panel installation inclination angle, a geometric projection relationship between the chassis plane of the photovoltaic cleaning robot and the photovoltaic panel surface is established.

[0115] In step 602 , the geometric projection relationship refers to a spatial position relationship model between the chassis plane and the photovoltaic panel surface.

[0116] In an embodiment of the present application, the tilt angle measured in real time by the photovoltaic panel tilt sensor is used to establish a transformation relationship between the chassis plane coordinate system and the photovoltaic panel surface coordinate system, including a rotation matrix and a translation vector, to accurately describe the relative posture between the two.

[0117] Step 603: Calculate the offset distances of the four corners of the chassis of the photovoltaic cleaning robot in the normal direction of the photovoltaic panel surface based on the geometric projection relationship.

[0118] In step 603 , the normal direction offset distance refers to the deviation of the chassis corner in the direction perpendicular to the photovoltaic panel surface.

[0119] In an embodiment of the present application, based on the established geometric projection relationship, the distance component of each corner position in the normal direction of the photovoltaic panel surface is calculated to reflect the actual gap between the chassis and the photovoltaic panel surface caused by the change in inclination angle.

[0120] Step 604: Based on the offset distance, combined with the dynamic adjustment parameters of the movement direction of the photovoltaic cleaning robot and the rate of change of the photovoltaic panel inclination angle, an offset compensation amount is generated.

[0121] In step 604, combining the motion direction of the photovoltaic cleaning robot with the dynamic adjustment parameter of the photovoltaic panel inclination change rate refers to coupling the motion direction vector and the inclination change rate parameter to perform calculations, wherein the motion direction provides a vector reference, and the inclination change rate parameter provides a dynamic scaling factor. The two generate a direction-sensitive compensation coefficient through vector dot product operation, and then multiply it with the offset distance to obtain the final offset compensation amount. The dynamic change parameter of the photovoltaic panel installation inclination refers to the real-time inclination angle change of the photovoltaic panel relative to the horizontal plane, which is derived from the installation data of the photovoltaic bracket or the real-time measurement of the inclination sensor. The dynamic adjustment parameter of the photovoltaic panel inclination change rate refers to the speed at which the photovoltaic panel inclination changes over time, which is obtained by calculating the real-time derivative of the inclination change parameter. The two are related but different concepts: the former describes the inclination state, and the latter describes the inclination change trend. The dynamic change parameter of the photovoltaic panel installation inclination describes the instantaneous value of the inclination, while the dynamic adjustment parameter of the photovoltaic panel inclination change rate is its time derivative; the former is used to establish a static geometric projection relationship, and the latter is used to predict the dynamic compensation requirements during movement. The two are related through a time difference equation: the inclination change rate parameter at the current moment is equal to the quotient of the difference value of the dynamic change parameter of two adjacent frames and the time interval.

[0122] In the embodiments of the present application, by analyzing the vector relationship between the photovoltaic panel tilt angle change trend and the robot's current motion direction, dynamic adjustment parameters are generated, and the normal offset distance is adaptively corrected to the motion state to ensure that the compensation amount meets the actual motion requirements. The specific process is: the photovoltaic cleaning robot's real-time motion direction unit vector is dot-producted with the photovoltaic panel tilt angle change rate parameter to obtain the direction-rate coupling coefficient; this coefficient is multiplied by the base value of the offset distance, and then superimposed with the sinusoidal compensation component of the photovoltaic panel's current tilt angle to generate a three-dimensional offset compensation vector with direction sensitivity. Specific embodiment: When the robot moves in a 30° direction on a photovoltaic panel with a 25° inclination angle, the basic offset distance in the Y-axis direction is measured to be 0.15 meters, and the inclination change rate is 0.5° / s; the dot product of the motion direction vector (cos30°, sin30°) and the rate parameter obtains a coupling coefficient of 0.43, and the superimposed inclination sinusoidal compensation amount of 0.15×sin25°=0.063 meters finally generates a Y-axis offset compensation amount of 0.15×0.43+0.063=0.128 meters.

[0123] Here's a specific example: When the photovoltaic cleaning robot moves along the X-axis on the photovoltaic panel surface with an inclination angle of 10 degrees, based on the preliminary three-dimensional positioning data obtained in the previous step, the X-axis is 2.092 meters, the Y-axis is 1.192 meters, and the Z-axis is 1.5 meters. First, the coordinates of the four corners are calculated according to the chassis size of 0.8 meters by 0.5 meters: the front left corner is ; The front right corner is ; The rear left corner is ; The rear right corner is Then, based on the 15-degree inclination angle of the photovoltaic panel, a geometric projection relationship is established to calculate the offset distance of each corner in the normal direction of the photovoltaic panel: the offset distance of the front left corner is , the front right angle is also 0.207 meters, the rear left angle and rear right angle are Then, combined with the robot's current X-axis speed of 0.5 meters per second and the inclination change rate of 0.2 degrees per second, the dynamic adjustment parameter is calculated to be 0.9, and the offset compensation of the four corners is finally generated: the front left corner , front right corner , rear left corner , rear right corner .

[0124] In the embodiment of the present application, by accurately calculating the normal offset of the four corners of the chassis on the surface of the photovoltaic panel and combining it with dynamic adjustment of the motion state, the positioning error caused by the change in the inclination angle of the photovoltaic panel is effectively compensated, ensuring that the robot maintains stable contact on the inclined surface.

[0125] To address the final positioning accuracy issue of the photovoltaic cleaning robot in a dynamic obstacle environment, in some embodiments, step 105: generating target three-dimensional positioning data of the photovoltaic cleaning robot based on the real-time pose estimation value and in combination with the real-time interference characteristics of dynamic obstacles on the photovoltaic panel surface, includes: Step 701: According to the real-time interference characteristics of the dynamic obstacles on the surface of the photovoltaic panel, the signal blocking area of ​​the edge of the dynamic obstacle in the movement direction of the photovoltaic cleaning robot is determined.

[0126] In step 701, the signal blocking area refers to the effective influence range of the dynamic obstacles causing interference to the positioning signal.

[0127] In an embodiment of the present application, by analyzing the physical characteristics of dynamic obstacles such as the metal frame on the surface of the photovoltaic panel and the dust accumulation area, combined with the current movement direction of the robot, the occlusion range of the obstacle edge to the ultra-wideband and differential positioning signals is determined, providing a spatial reference for interference compensation.

[0128] Step 702: Based on the real-time pose estimation value, a dynamic interference model between the motion trajectory of the photovoltaic cleaning robot and the signal blocking area is established.

[0129] In step 702, the dynamic interference model refers to a spatial correlation model of pose estimation error and obstacle distribution.

[0130] In an embodiment of the present application, a spatial mapping relationship between the trajectory points and the signal blocking area is established based on the robot motion trajectory described by the real-time pose estimation value, and the degree of influence of obstacles in different positions on the positioning accuracy is quantified through historical error data analysis.

[0131] Step 703: Correct the positioning drift of the real-time pose estimation value in the signal blocking area through the dynamic interference model, and generate target three-dimensional positioning data in combination with preset trajectory constraints.

[0132] In step 703 , the preset trajectory constraint refers to the spatial boundary condition of the photovoltaic panel surface cleaning path.

[0133] In an embodiment of the present application, the posture correction output by the dynamic interference model is compared and verified with the pre-set allowable range of the cleaning path to ensure that the final positioning data not only eliminates obstacle interference but also complies with the physical boundary restrictions of the photovoltaic panel. The cleaning path of the photovoltaic panel surface = the motion trajectory of the photovoltaic cleaning robot. The specific process is: in the process of the photovoltaic cleaning robot moving along the preset cleaning path on the surface of the photovoltaic panel, when the robot approaches the metal frame at the edge of the photovoltaic panel array, the ultra-wideband signal is interfered by the metal reflection, resulting in a positioning drift of 12 cm in the X-axis direction of the real-time posture estimation value. The drift is calculated by the signal attenuation coefficient of 0.8 and the metal frame distance parameter of 1.2 meters in the dynamic interference model. The calculation formula is: Positioning drift = signal attenuation coefficient × metal frame distance × 0.1. The dynamic interference model first identifies that the signal blocking area formed by the metal frame is within a range of plus or minus 0.5 meters along the Y-axis. Then, based on the trajectory constraint that the robot should maintain parallel movement with the edge of the array in the preset cleaning path, the X-axis drift of 12 cm is compensated and corrected according to the path constraint weight of 0.6. Finally, the target three-dimensional positioning data with an X-axis correction value of 7.2 cm is output, while the Y-axis and Z-axis data are kept unchanged to ensure that the robot continues to move along the preset trajectory under the interference of the metal frame.

[0134] Here's a specific example: When the photovoltaic cleaning robot moves along the X-axis on the surface of a photovoltaic panel with a 10-degree inclination angle, when the real-time pose estimation value shows that the X-axis position reaches 2.14 meters, the ultra-wideband signal strength mutation detection and differential positioning loss state are used to identify the presence of a metal frame obstacle 0.3 meters in front, and determine that its signal blocking area ranges from 2.2 meters to 2.5 meters on the X-axis. Based on the robot's current movement speed of 0.5 meters per second and the pose estimation value of 2.14 meters, a dynamic interference model is established to predict that the X-axis positioning will produce positive drift after entering the blocking area. The drift calculation formula is: , when the robot is located at 2.14 meters on the X axis, the calculated potential drift is , indicating that it has not entered the main interference area; when the robot continues to move to 2.25 meters on the X axis, the model detects that the actual posture suddenly increases by 0.15 meters, exceeding the expected value. The error database is queried to obtain the typical compensation amount for this position, which is negative 0.12 meters. The real-time posture estimate of 2.25 meters is corrected to It has been verified that this value is within the X-axis range of 2.0 meters to 2.5 meters allowed by the preset cleaning path, while the Y-axis 1.33 meters and the Z-axis 1.19 meters remain unchanged. The final output is the target three-dimensional positioning data of 2.13 meters, 1.33 meters, and 1.19 meters that conforms to the physical boundaries of the photovoltaic panel, ensuring that the robot can still accurately clean along the preset path in the metal frame interference area.

[0135] In the embodiment of the present application, accurate compensation of positioning errors is achieved through dynamic obstacle interference feature identification and modeling, so that the final positioning result not only overcomes the influence of environmental interference, but also strictly follows the physical constraints of photovoltaic panel surface cleaning, thereby ensuring the accuracy and safety of the cleaning operation.

[0136] Figure 2 A schematic diagram of a photovoltaic cleaning robot positioning system based on UWB and RTK is provided in an embodiment of the present application, as shown in FIG. Figure 2 As shown, the system includes: The acquisition module 21 is used to obtain signal time difference data between ultra-wideband base stations, ultra-wideband ranging error, real-time dynamic differential positioning data, real-time dynamic differential positioning deviation and inertial acceleration data during the movement of the photovoltaic cleaning robot.

[0137] The first generating module 22 is used to generate preliminary three-dimensional positioning data of the photovoltaic cleaning robot based on the signal time difference data and the real-time dynamic differential positioning data.

[0138] The first calculation module 23 is used to calculate the offset compensation amount of the four corners of the chassis of the photovoltaic cleaning robot relative to the surface of the photovoltaic panel based on the preliminary three-dimensional positioning data.

[0139] The second calculation module 24 is used to calculate the real-time pose estimation value of the photovoltaic cleaning robot through a dynamic weighted method based on the offset compensation, the ultra-wideband ranging error, the real-time dynamic differential positioning deviation and the inertial acceleration data.

[0140] The second generating module 25 is used to generate target three-dimensional positioning data of the photovoltaic cleaning robot based on the real-time pose estimation value and in combination with the real-time interference characteristics of dynamic obstacles on the surface of the photovoltaic panel.

[0141] Figure 2 The photovoltaic cleaning robot positioning system based on UWB and RTK can be performed Figure 1 The implementation principle and technical effects of the UWB and RTK-based photovoltaic cleaning robot positioning method described in the illustrated embodiment will not be elaborated on here. The specific manner in which each module and unit performs operations in the UWB and RTK-based photovoltaic cleaning robot positioning system in the above embodiment has been described in detail in the embodiments of the method and will not be elaborated on here.

[0142] In one possible design, Figure 2 The photovoltaic cleaning robot positioning system based on UWB and RTK of the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32; The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0143] The processing component 32 is used to perform the above Figure 1 The embodiment provides a photovoltaic cleaning robot positioning method based on UWB and RTK.

[0144] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0145] The storage component 31 is configured to store various types of data to support operations on the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random-access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0146] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0147] The input / output interface provides an interface between the processing component and the peripheral interface module, which can be an output device, an input device, etc.

[0148] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0149] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0150] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The embodiment shown is a photovoltaic cleaning robot positioning method based on UWB and RTK.

[0151] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0152] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0153] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0154] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A photovoltaic cleaning robot positioning method based on UWB and RTK, characterized in that: include: During the movement of the photovoltaic cleaning robot, the signal time difference data between ultra-wideband base stations, ultra-wideband ranging error, real-time dynamic differential positioning data, real-time dynamic differential positioning deviation and inertial acceleration data are obtained; Generating preliminary three-dimensional positioning data of the photovoltaic cleaning robot based on the signal time difference data and the real-time dynamic differential positioning data; Calculating the offset compensation of the four corners of the chassis of the photovoltaic cleaning robot relative to the surface of the photovoltaic panel based on the preliminary three-dimensional positioning data; Calculating a real-time pose estimation value of the photovoltaic cleaning robot by a dynamic weighting method based on the offset compensation amount, the ultra-wideband ranging error, the real-time dynamic differential positioning deviation, and the inertial acceleration data; Based on the real-time pose estimation value and combined with the real-time interference characteristics of dynamic obstacles on the surface of the photovoltaic panel, target three-dimensional positioning data of the photovoltaic cleaning robot is generated.

2. The method according to claim 1, characterized in that The method of calculating the real-time pose estimation value of the photovoltaic cleaning robot by a dynamic weighting method based on the offset compensation amount, the ultra-wideband ranging error, the real-time dynamic differential positioning deviation, and the inertial acceleration data includes: Determining a dynamic error distribution parameter according to a real-time fluctuation range of the ultra-wideband ranging error; generating spatiotemporal correlation parameters based on the statistical characteristics of the real-time dynamic differential positioning deviation; Performing vector superposition of the offset compensation amount and the motion trend of the inertial acceleration data to generate a dynamic adjustment vector value; According to the dynamic error distribution parameter, the spatiotemporal correlation parameter and the dynamic adjustment vector value, the weight coefficients of the ultra-wideband ranging error, the real-time dynamic differential positioning deviation and the inertial acceleration data are distributed in real time to generate a real-time pose estimation value.

3. The method according to claim 2, characterized in that The method further comprises: allocating, in real time, the weight coefficients of the ultra-wideband ranging error, the real-time dynamic differential positioning deviation, and the inertial acceleration data according to the dynamic error distribution parameter, the spatiotemporal correlation parameter, and the dynamic adjustment vector value to generate a real-time pose estimation value, including: Calculating a credibility weight factor of the ultra-wideband ranging error according to the dynamic error distribution parameter; Based on the spatiotemporal correlation parameters, generating a regional adaptation weight factor of the real-time dynamic differential positioning deviation; determining a motion consistency weight factor of the inertial acceleration data according to the dynamic adjustment vector value; Dynamically normalizing the credibility weight factor, the region adaptation weight factor, and the motion consistency weight factor to generate a weight coefficient set; Based on the weight coefficient set, the distance data after ultra-wideband ranging error correction, the coordinate data after real-time dynamic differential positioning deviation compensation, and the displacement data after inertial acceleration integration are weightedly fused to generate a real-time pose estimation value.

4. The method according to claim 3, characterized in that The method performs weighted fusion based on the weight coefficient set on the distance data after ultra-wideband ranging error correction, the coordinate data after real-time dynamic differential positioning deviation compensation, and the displacement data after inertial acceleration integration to generate a real-time pose estimation value, including: Scaling the distance data after the ultra-wideband ranging error correction according to the ultra-wideband signal weight value in the weight coefficient set to generate a first weighted result; Scaling the coordinate data after the real-time dynamic differential positioning deviation compensation according to the real-time dynamic differential positioning weight value in the weight coefficient set to generate a second weighted result; Scaling the displacement data after the inertial acceleration integration according to the inertial acceleration weight value in the weight coefficient set to generate a third weighted result; Performing vector superposition on the first weighted result, the second weighted result, and the third weighted result to generate a fusion estimation value; Performing trajectory smoothing processing on the fused estimated value to generate a real-time pose estimated value.

5. The method according to claim 1, wherein Generating preliminary three-dimensional positioning data of the photovoltaic cleaning robot based on the signal time difference data and the real-time dynamic differential positioning data includes: Determining a dynamic distance difference between the photovoltaic cleaning robot and at least three ultra-wideband base stations based on the signal time difference data; Extracting the geographic coordinate reference point of the photovoltaic panel array where the photovoltaic cleaning robot is located from the real-time dynamic differential positioning data; Dynamically compensate and match the dynamic distance difference with the geographic coordinate reference point to establish a signal propagation path correction model related to the dynamic obstacle distribution on the photovoltaic panel surface; The signal propagation path correction model is used to eliminate signal reflection interference caused by the metal frame and glass covering on the surface of the photovoltaic panel, and generate preliminary three-dimensional positioning data.

6. The method according to claim 1, characterized in that Calculating the offset compensation of the four corners of the chassis of the photovoltaic cleaning robot relative to the surface of the photovoltaic panel based on the preliminary three-dimensional positioning data includes: According to the preliminary three-dimensional positioning data, the real-time projection coordinates of the four corners of the chassis of the photovoltaic cleaning robot in the photovoltaic panel array coordinate system are obtained; Based on the dynamic change parameters of the photovoltaic panel installation inclination angle, the geometric projection relationship between the chassis plane of the photovoltaic cleaning robot and the photovoltaic panel surface is established; Calculate the offset distances of the four corners of the chassis of the photovoltaic cleaning robot in the normal direction of the photovoltaic panel surface based on the geometric projection relationship; Based on the offset distance, combined with the dynamic adjustment parameters of the movement direction of the photovoltaic cleaning robot and the rate of change of the photovoltaic panel inclination angle, an offset compensation amount is generated.

7. The method according to claim 1, characterized in that The method of generating target three-dimensional positioning data of the photovoltaic cleaning robot based on the real-time pose estimation value and combining the real-time interference characteristics of dynamic obstacles on the surface of the photovoltaic panel includes: According to the real-time interference characteristics of dynamic obstacles on the surface of photovoltaic panels, the signal blocking area of ​​the edge of the dynamic obstacle in the movement direction of the photovoltaic cleaning robot is determined; Based on the real-time pose estimation value, a dynamic interference model between the motion trajectory of the photovoltaic cleaning robot and the signal blocking area is established; The positioning drift of the real-time pose estimation value in the signal blocking area is corrected by the dynamic interference model, and combined with the preset trajectory constraints to generate the target three-dimensional positioning data.

8. A photovoltaic cleaning robot positioning system based on UWB and RTK, characterized in that: include: The acquisition module is used to obtain the signal time difference data between ultra-wideband base stations, ultra-wideband ranging error, real-time dynamic differential positioning data, real-time dynamic differential positioning deviation and inertial acceleration data during the movement of the photovoltaic cleaning robot; A first generating module is used to generate preliminary three-dimensional positioning data of the photovoltaic cleaning robot based on the signal time difference data and the real-time dynamic differential positioning data; A first calculation module is used to calculate the offset compensation amount of the four corners of the chassis of the photovoltaic cleaning robot relative to the surface of the photovoltaic panel based on the preliminary three-dimensional positioning data; A second calculation module is configured to calculate a real-time pose estimation value of the photovoltaic cleaning robot by a dynamic weighted method based on the offset compensation amount, the ultra-wideband ranging error, the real-time dynamic differential positioning deviation, and the inertial acceleration data; The second generation module is used to generate target three-dimensional positioning data of the photovoltaic cleaning robot based on the real-time posture estimation value and the real-time interference characteristics of dynamic obstacles on the surface of the photovoltaic panel.

9. A computing device, characterized in that It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a photovoltaic cleaning robot positioning method based on UWB and RTK as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, a photovoltaic cleaning robot positioning method based on UWB and RTK is implemented as described in any one of claims 1 to 7.

Images