Machine vision-based unmanned monorail hoist positioning system

Abstract

The application relates to the technical field of image processing, and discloses an unmanned monorail hoist positioning system based on machine vision. The system comprises a noise reduction module, a matching module, a filtering module, a tracking module, a sampling module and an evaluation module, and the modules are connected in series. The noise reduction module processes an image to obtain track feature data, the matching module generates three-dimensional feature data, the filtering module fuses multi-source data, the tracking module optimizes position data, the sampling module calculates motion parameters, and the evaluation module outputs correction data to a driving system. The application improves the efficiency and accuracy of unmanned monorail hoist positioning based on machine vision.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image processing, and particularly relates to a positioning system for unmanned monorail hoist based on machine vision. BACKGROUND

[0002] At present, the monorail hoist in coal mine has been widely applied in underground material transportation system. The traditional monorail hoist positioning mainly relies on manual operation or single sensor for navigation positioning, such as positioning system based on laser ranging, positioning system based on RFID, positioning system based on inertial navigation, etc. Among them, the laser ranging positioning system determines the position by measuring the distance between the hoist and the roadway wall, but it is easily affected by the environmental factors such as dust, water vapor in the mine; the RFID positioning system needs to arrange a large number of tags along the track, and the maintenance cost is high; the inertial navigation system is easy to produce cumulative error under the interference of strong magnetic field in the mine, and the positioning accuracy is significantly reduced after long time operation.

[0003] The main deficiency of the prior art is that the positioning method of single sensor has poor anti-interference ability, and it is difficult to ensure the reliability and accuracy of positioning in the complex environment of the mine. Especially in the harsh environment conditions such as weak light, dust, humidity and strong magnetic field in the mine, the single sensor is easily disturbed or fails, resulting in unstable positioning system. At the same time, the prior art lacks effective fusion mechanism for multi-source sensor data, cannot fully utilize the advantages of various sensors to improve the positioning accuracy, and also lacks adaptive processing ability for special environment in the mine. SUMMARY

[0004] The present application provides a positioning system for unmanned monorail hoist based on machine vision, which is used to solve how to realize high-precision and reliable positioning of monorail hoist in complex environment of the mine. The present application establishes a complete positioning solution by multi-sensor information fusion technology, combined with machine vision processing, deep learning and other advanced algorithms, which can effectively overcome the adverse effects of underground environment on the positioning system, and significantly improve the positioning accuracy and system reliability.

[0005] The present application provides a positioning system for unmanned monorail hoist based on machine vision, which comprises a noise reduction module, a matching module, a filtering module, a tracking module, a sampling module and an evaluation module.

[0006] The output end of the noise reduction module is connected with the input end of the matching module; the output end of the matching module is connected with the input end of the filtering module; the output end of the filtering module is connected with the input end of the tracking module; the output end of the tracking module is connected with the input end of the sampling module; the output end of the sampling module is connected with the input end of the evaluation module; and the output end of the evaluation module is connected with the input end of the driving system.

[0007] a denoising module, configured to denoise and enhance contrast of the collected downhole track image, and to calculate track contour data, node position data and feature marker data through feature extraction;

[0008] a matching module, configured to perform binocular matching calculation by using the track contour data, the node position data and the feature marker data, and to obtain three-dimensional feature data through disparity mapping conversion;

[0009] a filtering module, configured to perform data fusion and filtering processing on the three-dimensional feature data, inertial measurement data and distance measurement data, and to calculate initial position coordinate data;

[0010] a tracking module, configured to track track feature points according to the initial position coordinate data, and to generate continuous position data through local area optimization processing;

[0011] a sampling module, configured to perform track sampling calculation according to the continuous position data, and to obtain motion parameter data through constraint condition optimization;

[0012] an evaluation module, configured to perform real-time state evaluation calculation on the motion parameter data, to generate correction data through deviation analysis processing, and to send the correction data to a driving system.

[0013] In the technical scheme provided in the application, through the synergistic effect of the various modules, a high-precision and reliable positioning effect is achieved. The denoising module denoises and enhances the contrast of the downhole track image, effectively overcoming problems such as insufficient downhole environmental illumination and dust interference, and the extracted track contour data, node position data and feature marker data provide a reliable basis for subsequent processing. The matching module converts two-dimensional image features into three-dimensional feature data through binocular matching calculation and disparity mapping conversion, significantly improving the accuracy of spatial position perception. The filtering module fuses and filters the three-dimensional feature data, inertial measurement data and distance measurement data, improves the robustness of positioning through multi-source data fusion, and the obtained initial position coordinate data has high credibility. The tracking module continuously tracks track feature points, effectively eliminates cumulative errors through local area optimization, and generates continuous position data that ensures the continuity of positioning. The sampling module performs track sampling calculation based on the continuous position data, and the obtained motion parameter data through constraint condition optimization ensures the feasibility of motion planning. The evaluation module discovers and corrects positioning deviations in a timely manner through real-time state evaluation and deviation analysis, and the output correction data can ensure the continuous and stable operation of the system, achieving precise positioning of the monorail hoist in a complex downhole environment. BRIEF DESCRIPTION OF DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0015] Figure 1 An embodiment of the unmanned monorail hoist positioning system based on machine vision in the present application is shown in the figure.

[0016] Figure 2 The structure of the noise reduction module in the unmanned monorail hoist positioning system based on machine vision in the present application is shown in the figure. DETAILED DESCRIPTION

[0017] The present application provides an unmanned monorail hoist positioning system based on machine vision. The terms "first", "second", "third", "fourth" and the like (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily mean a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the term "comprising" or "having" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0018] For the sake of understanding, the specific process of the embodiments of the present application will be described below. Please refer to Figure 1 An embodiment of the unmanned monorail hoist positioning system based on machine vision in the present application includes:

[0019] The noise reduction module 101 is used to reduce noise and enhance contrast of the collected underground track image, and track contour data, node position data and feature marker data are obtained through feature extraction calculation;

[0020] The matching module 102 is used to perform binocular matching calculation by using the track contour data, node position data and feature marker data, and three-dimensional feature data is obtained by disparity mapping conversion;

[0021] The filtering module 103 is used to perform data fusion and filtering processing on the three-dimensional feature data, inertial measurement data and distance measurement data, and initial position coordinate data is calculated;

[0022] The tracking module 104 is configured to perform tracking calculation on the track feature points according to the initial position coordinate data, and generate continuous position data through local area optimization processing.

[0023] The sampling module 105 is configured to perform trajectory sampling calculation according to the continuous position data, and obtain motion parameter data through constraint condition optimization.

[0024] The evaluation module 106 is configured to perform real-time state evaluation calculation on the motion parameter data, generate correction data through deviation analysis processing, and send the correction data to the driving system.

[0025] It can be understood that the execution subject of the present application can be an unmanned monorail hoist positioning system based on machine vision, and can also be a terminal or a server, which is not limited here. The server is taken as an example for description of the embodiments of the present application.

[0026] Specifically, the noise reduction module 101 first processes the collected underground track image, performs multi-scale decomposition through wavelet transform, decomposes the image into different frequency components, and performs hierarchical processing on the common Gaussian noise and salt and pepper noise in the underground environment. In actual application, for an original image with a resolution of 1920x1080, 16 sub-band images are obtained after 4-layer wavelet decomposition, each sub-band containing information of a specific frequency range. The threshold decision method is used to screen the coefficients of each sub-band, and a noise coefficient matrix is generated, which records the noise distribution characteristics in the image. After obtaining the noise coefficient matrix, the noise reduction module 101 enhances the image through adaptive histogram equalization technology. Adaptive histogram equalization is different from traditional global histogram equalization, but divides the image into multiple local areas, respectively calculates and optimizes the contrast of each area. For dark areas, the brightness value is improved and the detail information is maintained; for bright areas, it is appropriately suppressed to avoid overexposure. The enhanced image data obtained through this processing has more uniform gray level distribution and more obvious features.

[0027] Then the denoising module 101 uses the Canny edge detection operator to perform edge detection on the enhanced image to extract the contour information of the track. The Canny operator first uses a Gaussian filter to smooth the image, then calculates the gradient amplitude and direction of the image, and obtains complete edge information through a double-threshold method and edge connection. After edge detection, the initial data of the track contour is obtained, which is represented in the form of a binary image, where the value of the edge point is 255 and the value of the background point is 0. After obtaining the initial data of the track contour, the matching module 102 extracts the key node positions through curve fitting technology. First, the Douglas-Peucker algorithm is used to simplify the contour line and retain the key turning points, and then the corner detection algorithm is used to identify the connection and change points of the track, which constitute the initial data of the node position. The node position data is converted into actual physical coordinates through spatial coordinate transformation, which converts the image coordinate system into the actual physical coordinate system, facilitating subsequent three-dimensional reconstruction.

[0028] The initial data of the converted node positions is subjected to cluster analysis, and the matching module 102 uses the DBSCAN density clustering algorithm to combine similar feature points into feature point groups according to spatial distance and point density. The DBSCAN algorithm does not need to specify the number of categories in advance, can effectively identify point groups of any shape, and has good robustness to noise points. Finally, the feature point groups are labeled and paired, the correlation between points is calculated, and the final track contour data, node position data and feature label data are obtained through feature merging. In the matching module 102, first, the images collected by the binocular camera are subjected to epipolar rectification to eliminate distortion caused by camera installation errors, so that the corresponding points are on the same horizontal line, facilitating subsequent stereo matching. The corrected feature data is subjected to regional matching, and the SAD (Sum of Absolute Differences) algorithm is used to calculate the similarity of the corresponding regions of the left and right images to generate initial disparity data.

[0029] After depth constraint judgment and detail compensation processing, the filtering module 103 removes false matching points to obtain reliable disparity data. Based on the principle of triangulation, the disparity data is converted into three-dimensional spatial coordinates to establish point cloud data. The point cloud data is subjected to statistical filtering to remove outliers and noise points, improving data quality. Finally, spatial feature descriptors are constructed to complete feature matching and fusion, and accurate three-dimensional feature data is output. The filtering module 103 receives data from the inertial measurement unit, identifies the stationary state of the crane through the zero velocity detection algorithm, and compensates for the zero bias error of the gyroscope and accelerometer. At the same time, the three-dimensional feature data and inertial data are fused, and the Kalman filter framework is used to calculate the optimal estimate. In practice, the state vector of the filter contains position, velocity, attitude and other information, and through the prediction-update cycle, the positioning result is continuously optimized.

[0030] The tracking module 104 tracks the feature points using a pyramid optical flow algorithm, improves the robustness and accuracy of tracking through a multi-layer image pyramid. The sliding window optimization is performed on the feature point sequence obtained by tracking to eliminate the cumulative error. The Lie group Lie algebra theory is used for attitude interpolation to ensure the continuity and accuracy of the attitude estimation. The Bezier curve is used by the sampling module 105 to smooth the trajectory, ensuring the continuity and smoothness of the motion trajectory. The boundary conditions of velocity and acceleration are determined through dynamic constraints to ensure that the motion parameters meet the physical limitations of the crane. The main features of the trajectory are extracted using the recursive least squares method, and a polynomial curve family is established to describe the feasible motion path.

[0031] The evaluation module 106 monitors the system state in real time, extracts the intrinsic features of the motion parameters through variational decomposition, and establishes a Gaussian mixture model to evaluate the system state. The conditional entropy criterion is used for decision analysis, and the Lyapunov stability theory is used to ensure the convergence of the control system. Finally, the recursive least squares method is used for online identification and optimization of the control parameters.

[0032] For example, in the positioning process of a monorail crane in a certain mine, the denoising module 101 collects an original image with a resolution of 1920x1080, and obtains 16 sub-band images through 4-layer wavelet decomposition. After adaptive histogram equalization processing, the contrast of the image is improved by 40%, and the edge features are more clear. After Canny edge detection, 50 key nodes are extracted, and 15 feature point groups are formed after DBSCAN clustering. In the matching module 102, the matching window size of the SAD algorithm is set to 11x11 pixels, and the generated disparity map reaches sub-pixel level accuracy. The filter module 103 fuses IMU data and visual features, and the root mean square error of position estimation is controlled within ±5 centimeters. The tracking module 104 uses a 3-layer image pyramid for feature tracking, and the attitude estimation accuracy is better than 0.5 degrees. The sampling module 105 generates a motion trajectory with continuous curvature, the maximum speed is not more than 2 meters / second, and the acceleration is limited within 0.5 meters / second square. The control period of the evaluation module 106 is 20 milliseconds, and the real-time and stability are guaranteed.

[0033] In the embodiments of the present application, through the synergistic effect of various modules, a high-precision and reliable positioning effect is achieved. The noise reduction module 101 performs noise reduction and contrast enhancement on the underground track image, effectively overcoming problems such as insufficient underground environmental illumination, dust interference, etc., and the extracted track contour data, node position data and feature marker data provide a reliable basis for subsequent processing. The matching module 102 converts two-dimensional image features into three-dimensional feature data through binocular matching calculation and disparity mapping conversion, significantly improving the accuracy of spatial position perception. The filtering module 103 fuses and filters three-dimensional feature data with inertial measurement data and distance measurement data, improving the robustness of positioning through multi-source data fusion, and the obtained initial position coordinate data has high reliability. The tracking module 104 continuously tracks the track feature points, effectively eliminates cumulative errors through local area optimization, and generates continuous position data to ensure the continuity of positioning. The sampling module 105 performs trajectory sampling calculation based on the continuous position data, and the motion parameter data obtained by combining constraint condition optimization ensures the feasibility of motion planning. The evaluation module 106 discovers and corrects positioning deviations in time through real-time state evaluation and deviation analysis, and the output correction data can ensure the continuous and stable operation of the system, realizing the precise positioning of the monorail crane in the complex underground environment.

[0034] In a specific embodiment, as shown in Figure 2 The noise reduction module 101 specifically includes:

[0035] The decomposition unit 1011 is configured to perform multi-scale decomposition on the underground track image through wavelet transform, and obtain a noise coefficient matrix in combination with threshold decision;

[0036] The calculation unit 1012 is configured to generate enhanced image data through adaptive histogram equalization calculation according to the noise coefficient matrix;

[0037] The tracking unit 1013 is configured to perform edge detection and contour tracking using the enhanced image data, and obtain track contour initial data through gradient operator operation;

[0038] The input unit 1014 is configured to input the track contour initial data into a feature analyzer, and obtain node position initial data through curve fitting and feature point extraction;

[0039] The screening unit 1015 is configured to perform spatial coordinate transformation and feature screening on the node position initial data, and obtain feature point group data through clustering analysis;

[0040] The merging unit 1016 is configured to perform marker pairing and correlation calculation on the feature point group data, and output track contour data, node position data and feature marker data through feature merging operation.

[0041] Specifically, the decomposition unit 1011 first performs a wavelet transform to multi-scale decompose the downhole track image. The wavelet transform is a time-frequency analysis method that realizes multi-resolution analysis of image details by decomposing the image into sub-bands of different scales. Specifically, for the input downhole track image, a two-dimensional discrete wavelet transform is used for decomposition, and the image is sequentially decomposed into a low-frequency approximation component and a high-frequency detail component. The high-frequency component mainly contains noise and edge information, while the low-frequency component retains the main structural information of the image.

[0042] The mathematical expression of the wavelet transform decomposition process is:

[0043] W j,k (x,y)=∑ m ∑ n f(m,n)ψ j,k (m-x,n-y)

[0044] where: W j,k (x,y) represents the wavelet coefficient at scale j and direction k; f(m,n) represents the gray value of the original image at point (m,n); ψ j,k represents the wavelet basis function; (x,y) represents the image coordinate position.

[0045] After wavelet decomposition, the calculation unit 1012 performs threshold decision on the coefficients of each scale to obtain a noise coefficient matrix. The threshold decision uses a soft threshold method, and the threshold is adaptively set according to the noise level of each scale. Next, based on the obtained noise coefficient matrix, adaptive histogram equalization calculation is performed to generate enhanced image data. Adaptive histogram equalization is different from traditional global histogram equalization, but divides the image into multiple local regions, and calculates and optimizes the contrast of each region.

[0046] The calculation formula of adaptive histogram equalization is:

[0047]

[0048] where: g(i,j) represents the gray value of the image at position (i,j) after equalization; L represents the number of gray levels; MxN represents the size of the local region; h(x,y) represents the gray value of the original image at position (x,y).

[0049] After generating the enhanced image data, the tracking unit 1013 processes the image using edge detection and contour tracking technology. The edge detection uses an improved Canny operator, which includes four steps of Gaussian filtering, gradient calculation, non-maximum suppression and double threshold detection. The initial data of the track contour is obtained by gradient operator operation. The specific calculation of the gradient operator uses the following formula:

[0050]

[0051] wherein: G x ,G y denote the gradient in the x and y direction, respectively; I(x,y) denotes the gray value of the image at point (x,y); * denotes the convolution operation.

[0052] The gradient magnitude is calculated by .

[0053] For the initial data of the track contour, the input unit 1014 performs curve fitting and feature point extraction through the feature analyzer to obtain the initial data of the node position. The curve fitting adopts a cubic spline interpolation method to ensure the smoothness and continuity of the fitted curve. The feature point extraction is based on curvature calculation to extract key points at positions with larger curvature values. The curvature calculation formula for feature point extraction is:

[0054]

[0055] wherein: κ(t) denotes the curvature of the curve at parameter t; x . (t),y . (t) denote the first-order derivative of the curve at t; x : (t),y : (t) denote the second-order derivative of the curve at t.

[0056] Subsequently, the screening unit 1015 performs spatial coordinate transformation and feature screening on the initial data of the node position. The spatial coordinate transformation converts the two-dimensional image coordinates into three-dimensional physical coordinates, and the following transformation matrix is adopted:

[0057]

[0058] wherein: (X3,Y3,Z3) denotes the point coordinates in the world coordinate system; (x : ,y : ,z : ) denotes the point coordinates in the image coordinate system; β ij (i,j=1,2,3) denotes the rotation matrix elements; γ k (k=1,2,3) denotes the translation vector components.

[0059] The merging unit 1016 obtains the feature point group data through clustering analysis, adopts the density-based clustering algorithm DBSCAN, which automatically determines the number and range of clusters by calculating the distance relationship and density distribution characteristics between points. Finally, the feature point group data is labeled, paired and associated, the pairing process adopts the Hungarian algorithm, the association degree calculation is based on the distance and direction information between feature points, and finally the track contour data, node position data and feature label data are output through feature merging operation.

[0060] For example, in the process of positioning the monorail hoist in the well, a 1920x1080 resolution original image is decomposed into 16 subbands by four-layer wavelet decomposition. After threshold decision processing, the size of the noise coefficient matrix is 960x540. The adaptive histogram equalization uses a 16x16 local window for processing, and the enhanced image clarity is obviously improved. The low threshold of the Canny edge detection is set to 50, and the high threshold is set to 150, and the complete track contour is successfully extracted. The feature analyzer extracts about 200 initial feature points, and after curvature screening, 50 key nodes are retained. The accuracy of spatial coordinate transformation reaches millimeter level, and the final feature point group contains 15 stable feature clusters, each feature cluster contains 3-5 key points, and the spatial distribution of these feature points can accurately reflect the geometric characteristics of the track.

[0061] In an embodiment, the matching module 102 is specifically configured to:

[0062] (1) performing binocular matching calculation by using the track contour data, the node position data and the feature marker data, and obtaining three-dimensional feature data by disparity mapping conversion, including:

[0063] (2) performing correction calculation on the track contour data and the node position data by polar line constraint optimization, to obtain corrected feature data;

[0064] (3) performing similarity calculation on the corrected feature data and the feature marker data based on regional cost aggregation, to output initial disparity data;

[0065] (4) performing detail compensation processing on the initial disparity data by depth constraint determination and screening, to generate effective disparity data;

[0066] (5) mapping the effective disparity data to a spatial coordinate system by triangulation, to obtain initial point cloud data;

[0067] (6) removing outliers from the initial point cloud data according to a density clustering criterion, to generate optimized point cloud data after noise reduction processing;

[0068] (7) constructing a spatial feature descriptor for the optimized point cloud data, and outputting three-dimensional feature data after completing feature matching and fusion.

[0069] Specifically, the matching module 102 processing process involves several key links, first receive the track profile data, node position data and feature marker data output by the noise reduction module 101, these data contain the geometric shape of the track, the key node position and the feature marker information for matching. The epipolar constraint optimization is an important step in binocular stereo vision, through the epipolar constraint optimization to correct the track profile data and node position data, ensure that the corresponding points in the left and right images are on the same horizontal line, reduce the search range of stereo matching. The epipolar constraint optimization uses the basis matrix calculation and the epipolar equation, in the calculation process, first estimate the basis matrix, then calculate the epipolar geometry relationship according to the basis matrix, correct the epipolar deviation caused by image distortion and camera installation error. For the corrected feature data, the similarity is calculated by using the regional cost aggregation method. Regional cost aggregation is the core step in stereo matching, by calculating the gray difference, gradient difference and other features of the corresponding regions of left and right images, a matching cost matrix is constructed. In the calculation process, the sliding window method is used, the window size is usually set to 11x11 or 15x15 pixels, the cost value is calculated for each candidate matching point. The cost calculation uses a combination of SAD (Sum of Absolute Differences), Census transformation and other measurement methods to improve the robustness of matching. When the cost is aggregated, the spatial position relationship and color similarity of the pixels are considered, and an adaptive weight method is used to assign different weights to different pixels in the window.

[0070] After obtaining the initial disparity data, filtering is performed through depth constraint judgment. Depth constraint judgment is based on the geometric features of the scene, using prior information such as track installation height and roadway width to set a reasonable depth range. For disparity values that exceed the depth range or violate geometric constraints, detail compensation processing is used to correct them. Detail compensation processing includes left-right consistency check, disparity continuity check and occlusion region processing steps, through iterative optimization to eliminate abnormal disparity values, fill in disparity holes, and generate continuous and reliable effective disparity data. Triangulation is a key step in converting disparity data into three-dimensional space coordinates. Based on the calibrated camera parameters, the mapping relationship between two-dimensional image coordinates and three-dimensional space coordinates is established. In the triangulation process, the depth information is calculated through the disparity value, combined with the internal and external parameters of the camera, each feature point is mapped to the world coordinate system, and the initial three-dimensional point cloud data is generated. Each point in the point cloud data contains spatial coordinates and corresponding feature attribute information.

[0071] For the acquired point cloud initial data, outlier rejection is performed using a density clustering criterion. The density clustering is based on the DBSCAN algorithm, which identifies and rejects noise points and outliers by setting a neighborhood radius and a minimum point number threshold. In the noise reduction process, statistical filtering and radius filtering are combined to remove points with abnormal spatial distribution and retain representative point cloud data. In this way, the quality and reliability of the point cloud data are significantly improved. After obtaining the optimized point cloud data, a spatial feature descriptor needs to be constructed to achieve feature matching and fusion. The spatial feature descriptor uses a local feature description method based on normal vectors and curvature to calculate the geometric features and topological relationships of each point. Feature matching uses nearest neighbor search and feature similarity measurement to register and fuse point cloud data acquired at different times or from different perspectives, and finally outputs complete three-dimensional feature data.

[0072] For example, in an actual monorail hoist positioning application, when a pair of 1920x1080 resolution stereo image pairs are captured by the binocular camera, the images are first corrected by epipolar constraint optimization. During the correction process, the fundamental matrix is estimated by the RANSAC algorithm, more than 100 pairs of feature points are selected for calculation, the number of iterations is set to 1000 times, and the inlier proportion threshold is set to 0.8. The registration error of the corrected image pair is controlled within 1 pixel. The regional cost aggregation uses a window size of 15x15, combined with the hybrid cost measurement of SAD and Census transform, to generate an initial disparity map. The effective depth range is set to 0.5 meters to 5 meters in the depth constraint determination, and the disparity values outside the range are marked as invalid. Through detail compensation processing, the effective pixel rate of the disparity map is improved from the initial 85% to 95%. In the triangulation process, based on the camera focal length 2000 pixels, baseline length 120 mm and other parameters obtained by calibration, the disparity data is converted into spatial coordinates. The neighborhood radius is set to 0.05 meters and the minimum point number is set to 10 in the density clustering, and about 5% of the outliers are successfully removed. The final generated three-dimensional feature data contains more than 1000 valid feature points, with uniform spatial distribution, accurately reflecting the three-dimensional geometric structure of the track.

[0073] In an embodiment, the filtering module 103 is specifically configured to:

[0074] (1) compensate for zero bias error based on zero velocity detection on inertial measurement data, generate initial inertial data through quaternion transformation and integral operation;

[0075] (2) establish a spatial geometric constraint for three-dimensional feature data, calculate and output three-dimensional calibration data from feature point projection and depth calibration;

[0076] (3) calculate weight coefficients for the initial inertial data and the three-dimensional calibration data according to the Kalman filter framework, and generate a fusion weight matrix through optimization iteration;

[0077] (4) The distance measurement data is probabilistically resampled with the fusion weight matrix according to the covariance propagation criterion, and the optimal estimation data is extracted;

[0078] (5) The optimal estimation data is analyzed for time series correlation by Bayesian inference, and the position estimation data is obtained by screening through the particle swarm optimization algorithm;

[0079] (6) The position estimation data is quantified and compensated for uncertainty through ellipsoid error distribution, and the initial position coordinate data is output in combination with the multi-sensor consistency constraint.

[0080] Specifically, the filtering module 103 performs multi-sensor data fusion and filtering processing, and first processes data from an inertial measurement unit (IMU). Zero velocity detection is a key step in processing IMU data, which determines whether the crane is in a stationary state by analyzing the output signal characteristics of the accelerometer and gyroscope. When a stationary state is detected, the measurement value at this moment is used to estimate and compensate the zero bias of the IMU online. The data after zero bias compensation is updated in attitude through quaternion transformation, and then integrated to obtain inertial initial data, which contains the position, velocity and attitude information of the crane. When processing three-dimensional feature data, strict spatial geometric constraints need to be established. These constraints include the spatial layout characteristics of the track (such as the standard cross-sectional size of the track, the installation height, etc.) and the kinematic constraints (such as the freedom degree limit of the crane). By projecting the three-dimensional feature points onto these constraint planes and performing depth calibration, the spatial positions of the feature points are ensured to conform to the actual scene features. In the depth calibration process, the depth information measured by binocular vision is used as a reference to optimize the spatial coordinates of the feature points, and the calibrated three-dimensional data is output.

[0081] The Kalman filter framework is the core algorithm for realizing multi-source data fusion. Under this framework, the inertial initial data is taken as the system state prediction value, and the three-dimensional calibrated data is taken as the observation value, and the optimal estimation is calculated through recursive method. The calculation of weight coefficients is based on the measurement noise characteristics and dynamic performance of each sensor, which reflects the credibility of different data sources in the fusion process. Through multiple iterations of optimization, a fusion weight matrix reflecting the reliability of each data source is generated. When the distance measurement data is obtained, it is processed according to the covariance propagation criterion. Covariance propagation describes the transmission law of measurement error in coordinate transformation and data fusion process. The distance measurement data is combined with the aforementioned fusion weight matrix to perform probabilistic resampling, thereby obtaining the optimal estimation data. The probabilistic resampling process is based on the Monte Carlo method, which retains high-weight data samples and eliminates unreliable measurement values through random sampling.

[0082] Bayesian inference introduces prior knowledge to the optimal estimation data for time series analysis. In the process of monorail crane movement, the positions of adjacent time have strong correlation. By establishing a state transition probability model, the correlation between the current state and the historical state is calculated. Particle swarm optimization algorithm is used to search for the optimal solution in high-dimensional state space. Each particle represents a possible state estimation. By iteratively updating the position and velocity of the particle, the optimal solution is finally converged to obtain reliable position estimation data. Finally, the ellipsoid error distribution model is used to analyze the uncertainty of the position estimation data. Ellipsoid error distribution can more accurately describe the position error characteristics in three-dimensional space, and is more consistent with the actual situation than spherical error distribution. By calculating the principal axis direction and size of the error ellipsoid, the uncertainty of the positioning result is quantified. On this basis, combined with multi-source data from vision, IMU and distance sensors, multi-sensor consistency constraints are established, and the final initial position coordinate data is output by weighted fusion.

[0083] For example, in practical application scenarios, when the speed of the crane is less than 0.01 meters per second for more than 1 second, the zero speed detection module determines that it is in a stationary state. At this time, 100 groups of IMU data are collected for zero bias estimation, and the typical value of the gyroscope zero bias is about 0.02 degrees per second, and the accelerometer zero bias is about 0.01 meters per square second. The initial trajectory is obtained by integrating the compensated inertial data. The three-dimensional feature data provided by the vision system contains about 200 feature points, and after spatial geometric constraint processing, 150 effective points are retained. The prediction period of Kalman filter is 10 milliseconds, and after 20 iterations of optimization, the diagonal element values of the fusion weight matrix are distributed between 0.6 and 0.9, representing the credibility of different sensors. Laser ranging data is input at a frequency of 50 Hz, with a measurement range of 0.1 meters to 10 meters and an accuracy of ±2 millimeters. When probability resampling is performed, 1000 particles are maintained, and convergence results are obtained after 50 iterations. The standard deviation of the final output position coordinate data is controlled within ±1 centimeter in the horizontal direction and within ±0.5 centimeter in the vertical direction, fully meeting the accuracy requirements of crane positioning.

[0084] In an embodiment, the tracking module 104 is specifically configured to:

[0085] (1) extracting feature descriptors from the initial position coordinate data via pyramid layering, and generating feature matching data by means of an optical flow tracking algorithm;

[0086] (2) segmenting and processing the feature matching data according to a sliding window criterion, and outputting segmented trajectory data by local optimization calculation;

[0087] (3) performing pose interpolation operation on the segmented trajectory data according to Lie algebra space, and obtaining a key frame sequence after key frame selection;

[0088] (4) Construct local covariance matrix for keyframe sequence, eliminate cumulative error through sparse optimization algorithm, output optimized trajectory data;

[0089] (5) Calculate state transition probability of optimized trajectory data by hidden Markov chain, obtain position prediction data through maximum likelihood estimation;

[0090] (6) According to the graph optimization framework, the position prediction data is globally constrained with the historical observations, and the continuous position data is generated through iterative optimization.

[0091] Specifically, the tracking module 104 processes the initial position coordinate data based on a pyramid layering mechanism. Pyramid layering is a multi-scale image representation method that constructs an image pyramid by layer-wise downsampling, with the bottom layer maintaining the original resolution and each upper layer reducing the image size by half. When extracting feature descriptors from the pyramid structure, the FAST (Features from Accelerated Segment Test) corner detection algorithm is used to detect features in each layer of the image. The Lucas-Kanade algorithm is used for optical flow tracking, which is based on the assumption of constant image brightness to calculate the motion vector of feature points between adjacent frames. The sliding window is an efficient data processing mechanism. When processing feature matching data, a fixed-size time window is set, and only the data within the window is optimized. The choice of window size needs to balance the calculation efficiency and optimization accuracy, and usually contains 3-5 frames of data. When performing local optimization within the window, a graph optimization-based method is used to construct the observation constraint and motion model constraint of the feature points into an optimization objective function, and the least squares solution is used to obtain the segmented trajectory data.

[0092] The Lie algebra space provides a theoretical basis for processing continuous pose transformation. The SO(3) Lie group represents a three-dimensional rotation, and its corresponding Lie algebra SO(3) is in the Euclidean space, which is convenient for interpolation calculation of the pose. When interpolating the segmented trajectory data, first map the pose data to the Lie algebra space, use linear interpolation to obtain the intermediate pose, and then map it back to the Lie group space. The selection of key frames is based on the pose change and feature point disparity. When the pose change exceeds the threshold or the average disparity is large, the current frame is marked as a key frame.

[0093] The local covariance matrix reflects the uncertainty of the trajectory estimation. For the keyframe sequence, the covariance matrix of each keyframe pose is calculated to construct an error model containing the uncertainty of position and attitude. Sparse optimization utilizes the geometric constraint relationship between keyframes to construct an edge-based graph structure, and uses the g2o (General Graph Optimization) framework to solve the optimization, effectively eliminating the cumulative error and outputting the coherent trajectory data. The hidden Markov chain model describes the time-dependent relationship of the trajectory. The state transition probability matrix represents the transition relationship of the trajectory state at adjacent time, and the transition probability parameters are estimated by statistical learning method. Combined with maximum likelihood estimation, the most likely state sequence is calculated under the condition of given observation sequence, and the future position of the crane is predicted.

[0094] Finally, in the graph optimization framework, the position prediction data is added to the global pose graph as a new node, and the constraint relationship with the historical observation data is established. The nodes in the graph represent the pose state, and the edges represent the observation constraint and motion constraint. Through the iterative optimization method, the error of all constraints is minimized to generate smooth and continuous position data.

[0095] For example, when processing a sequence of images with a resolution of 1920x1080, a 4-layer image pyramid is constructed, with the top layer having a resolution of 120x68. FAST corner features are extracted at each layer, with a corner response threshold of 40, resulting in approximately 1000 feature points. The optical flow tracking uses a 21x21 search window with a brightness difference threshold of 0.3, achieving a tracking success rate of 95%. The sliding window size is set to 5 frames, and the number of feature points within the window is maintained at more than 800. When performing Lie algebra interpolation, the rotation threshold for keyframe selection is set to 5 degrees, and the translation threshold is set to 0.1 meters. The eigenvalues of the local covariance matrix indicate that the standard deviation of the position estimation is controlled within ±2 cm in the horizontal direction and ±1 cm in the vertical direction. The number of states of the hidden Markov model is set to 10, and the transition probability matrix is estimated from 1000 groups of training data. The final position tracking accuracy is better than ±5 cm, and the attitude accuracy is better than ±0.5 degrees.

[0096] In an embodiment, the sampling module 105 is specifically configured to:

[0097] (1) fitting and interpolating the continuous position data based on the Bezier curve to obtain smooth trajectory data through curvature constraint optimization;

[0098] (2) calculating the speed-acceleration boundary of the smooth trajectory data through dynamic constraint, and generating dynamic reference data combined with a safety distance threshold;

[0099] (3) decomposing the dynamic reference data into principal component features according to the recursive least squares criterion, and outputting the trajectory principal component data through feature reconstruction;

[0100] (4) Construct polynomial curve family based on trajectory principal component data, and obtain optimal path data by nonlinear programming solver;

[0101] (5) Apply kinematic constraints to optimal path data via Lagrange multiplier method, and generate trajectory control data by iterative optimization;

[0102] (6) Convert trajectory control data into velocity, acceleration and angular velocity sequence in variational inference framework, and output motion parameter data combined with dynamic boundary constraints.

[0103] Specifically, the sampling module 105 processes the continuous position data by a Bezier curve, which has good smoothness and continuity and is an ideal tool in trajectory planning. For n control points, an n-1 order Bezier curve is constructed, and the selection of control points is based on the key points of position data. Curvature constraint optimization ensures that the trajectory meets the turning ability of the crane. By setting a maximum curvature threshold, the curve is locally adjusted so that the curvatures of all points are within the safe range, thereby obtaining smooth trajectory data that meets the motion constraints. The dynamic constraints consider the physical characteristics of the crane, including maximum speed limit, acceleration limit and deceleration limit. Based on the smooth trajectory data, the velocity and acceleration boundaries of each trajectory segment are calculated. In the calculation process, the slope of the track, the curve and the load condition are considered, and the upper limit of the speed is dynamically adjusted. The safety distance threshold is determined based on the braking performance of the crane and the situation of the obstacle in front. By integrating these constraint conditions, dynamic reference data containing time information are generated.

[0104] The recursive least squares method is used to extract the main features of the trajectory. The dynamic reference data is regarded as a high-dimensional time series, and the parameter estimates are updated by recursive calculation to reduce the redundancy of the data. The feature extraction process focuses on the key changes of the trajectory, such as the points of speed change and acceleration change, and through principal component analysis, the most representative feature components are retained to reconstruct the simplified trajectory principal component data. Based on the trajectory principal component data, a polynomial curve family is constructed, and different order polynomials are used to describe possible motion paths. The polynomial coefficients are optimized by a nonlinear programming solver, and the optimization objectives include path smoothness, motion continuity and energy consumption, etc. The solution process uses the interior point method to find the optimal solution that meets all the constraints through iterative calculation.

[0105] The Lagrange multiplier method adds kinematic constraints as hard constraints to the optimization problem. These constraints include maximum steering angle, maximum angular velocity, and other limitations. By constructing a Lagrange function, the constrained optimization problem is converted into an unconstrained optimization problem. During the iterative optimization process, the original variable and the Lagrange multiplier are updated simultaneously until a solution that satisfies all constraints is obtained, resulting in executable trajectory control data. The variational inference framework converts trajectory control data into specific motion instructions. Through variational calculation, the speed, acceleration, and angular velocity values at each time point are obtained. These motion parameters must satisfy the dynamic boundary constraints of the crane, including motor torque limits, mechanical strength limits, and other constraints. The final output motion parameter data contains complete timing information, ensuring the smoothness and safety of the motion process.

[0106] For example, for a 50-meter long track motion planning, a 5th order Bezier curve is used for trajectory fitting, and the control point interval is set to 5 meters. The curvature constraint is set to a maximum value of 0.2 / m, ensuring that the turning radius is not less than 5 meters. In the dynamic constraint, the maximum speed limit is 2 meters / second, the maximum acceleration is 0.5 meters / second, the maximum deceleration is 0.8 meters / second, and the safety distance threshold is set to 3 meters. After recursive least squares processing, the first 5 principal components are retained, accounting for more than 95% of the total variance. The polynomial curve family uses a 7th order polynomial, and the nonlinear programming solver iterates 50 times to reach convergence. The final generated motion parameter data shows that during the 50-meter motion process, the maximum speed is 1.8 meters / second, the average acceleration is 0.3 meters / second, and the maximum angular velocity is 0.15 radians / second. The entire process is smooth and controllable, meeting the safety operation requirements.

[0107] In a specific embodiment, the evaluation module 106 is specifically configured to:

[0108] (1) Perform intrinsic mode extraction on the motion parameter data through variational decomposition, and obtain state feature data through singular spectrum analysis;

[0109] (2) Construct a probability density function for the state feature data based on a Gaussian mixture distribution, and generate state clustering data through density clustering operations;

[0110] (3) Map the state clustering data to the decision space according to the conditional entropy criterion, and output state evaluation data through information gain calculation;

[0111] (4) Construct a robust controller for the state evaluation data, and obtain deviation compensation data through Lyapunov stability analysis;

[0112] (5) Perform error propagation analysis on the deviation compensation data based on adaptive iterative learning, and obtain correction control data through feedback correction calculation;

[0113] (6) Parameter identification and optimization of the modified control data via recursive least squares, combined with system dynamics constraints to output correction data.

[0114] Specifically, the evaluation module 106 first processes the motion parameter data using a variational decomposition method. Variational decomposition is a signal processing method that can decompose complex signals into components of different frequency characteristics. In the process of intrinsic mode extraction, a series of intrinsic mode functions are obtained by iterative screening decomposition, and each mode function represents a characteristic component of motion. Singular spectrum analysis further analyzes the spectrum of these modes to extract the main frequency components and energy distribution, generating state feature data reflecting the dynamic characteristics of the system. Gaussian mixture model is used to probabilistically model the state feature data. By combining multiple Gaussian distributions, the probability density function of the state space is constructed. Each Gaussian component represents a motion state, and the weight coefficient reflects the probability of the occurrence of each state. The density clustering operation is based on the EM (Expectation Maximization) algorithm, which iteratively optimizes the parameters of the Gaussian components until the optimal state clustering result is obtained.

[0115] The conditional entropy criterion evaluates the uncertainty of the state clustering by calculating the prediction ability of the current state for the future state, quantifying the degree of certainty of state transition. When mapping the state clustering data to the decision space, information gain is used as the evaluation index to select the most discriminative features for mapping. Information gain calculation is based on the reduction of information entropy, and the data containing the state evaluation results are output. Robust controller design uses H∞ control theory, considering the uncertainty of the system model and external disturbances. The design goal of the controller is to maintain the stability and performance of the system under the worst conditions. Lyapunov stability analysis proves the stability of the system under disturbance by constructing a suitable Lyapunov function, and calculates the required bias compensation.

[0116] Adaptive iterative learning control optimizes for repetitive tasks. Based on historical running data, the error propagation law of the bias compensation data is analyzed, and an error model is established. In the feedback correction process, the control parameters are adjusted according to the real-time error, and the control accuracy is improved through iterative optimization to generate modified control data.

[0117] Recursive least squares method updates the system parameter estimation in real time. Online parameter identification is performed on the modified control data, and the weight of historical data is processed using a forgetting factor to ensure that the model can adapt to changes in system parameters. Finally, the control parameters are optimized in combination with the dynamics constraints of the crane, such as maximum load, speed limit, etc., to output the final correction data.

[0118] For example, when processing motion parameter data with a sampling frequency of 100 Hz, the variational decomposition obtains 5 main eigenmodes, and the energy ratio is more than 98%. The Gaussian mixture model selects 3 Gaussian components, and converges after 50 EM iterations. The eigenvalues of the covariance matrix of each component are between 0.01 and 0.1. The conditional entropy calculation shows that the uncertainty of state prediction is reduced by 60%, and the information gain is 0.8. The H∞ norm of the robust controller is less than 1.5, and the derivative of the Lyapunov function is always negative, proving the stability of the system. After 10 iterations of adaptive iterative learning, the tracking error is reduced from the initial 5 cm to within 1 cm. The recursive least squares uses a forgetting factor of 0.98, and the parameter error estimated in real time is maintained below 2%. The final output correction data makes the system control accuracy reach ±0.5 cm.

[0119] The integrated unit, if implemented in the form of a software functional unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or say the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0120] The above-described embodiments are merely used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A machine vision-based positioning system for an unmanned monorail crane, characterized in that, The machine vision-based unmanned monorail crane positioning system includes: The system includes a noise reduction module, a matching module, a filtering module, a tracking module, a sampling module, and an evaluation module. The output of the noise reduction module is connected to the input of the matching module. The output of the matching module is connected to the input of the filtering module. The output of the filtering module is connected to the input of the tracking module. The output of the tracking module is connected to the input of the sampling module. The output of the sampling module is connected to the input of the evaluation module. The output of the evaluation module is connected to the input of the driving system. The noise reduction module is used to reduce noise and enhance contrast in the acquired underground track images. After feature extraction calculation, track contour data, node position data, and feature marker data are obtained. The matching module is used to perform binocular matching calculation using the track contour data, node position data, and feature marker data, and obtain three-dimensional feature data through disparity mapping conversion. The filtering module is used to fuse and filter the three-dimensional feature data with inertial measurement data and distance measurement data to calculate the initial position coordinate data. Specifically, it is used for: compensating for zero bias error in the inertial measurement data based on zero velocity detection, generating initial inertial data through quaternion transformation and integral operation; establishing spatial geometric constraints for the three-dimensional feature data, and outputting three-dimensional calibration data by feature point projection and depth calibration calculation; calculating weight coefficients for the initial inertial data and three-dimensional calibration data according to the Kalman filter framework, and generating a fusion weight matrix through optimization iteration; performing probability resampling on the distance measurement data and the fusion weight matrix according to the covariance propagation criterion to extract the optimal estimated data; performing temporal correlation analysis on the optimal estimated data using Bayesian inference, and obtaining position estimated data through particle swarm optimization algorithm; quantifying and compensating for uncertainty in the position estimated data through ellipsoidal error distribution, and outputting the initial position coordinate data in combination with multi-sensor consistency constraints. The tracking module is used to track and calculate the track feature points based on the initial position coordinate data, and generate continuous position data through local region optimization processing; The sampling module is used to perform trajectory sampling calculation based on the continuous position data, and obtain motion parameter data through constraint optimization. The evaluation module is used to perform real-time state evaluation calculations on the motion parameter data, generate correction data through deviation analysis, and send the correction data to the drive system.

2. The machine vision-based unmanned monorail crane positioning system according to claim 1, characterized in that, The noise reduction module includes: The decomposition unit is used to perform multi-scale decomposition of the downhole trajectory image through wavelet transform and obtain the noise coefficient matrix by combining threshold decision. The calculation unit is used to generate enhanced image data by adaptive histogram equalization based on the noise coefficient matrix. The tracking unit is used to perform edge detection and contour tracking using the enhanced image data, and to obtain initial orbital contour data through gradient operator operations. The input unit is used to input the initial data of the track contour into the feature analyzer, and obtain the initial data of the node position through curve fitting and feature point extraction. The filtering unit is used to perform spatial coordinate transformation and feature filtering on the initial data of the node positions, and to obtain feature point group data through cluster analysis. The merging unit is used to label and pair the feature point group data and calculate the correlation, and output the track contour data, node position data and feature label data through feature merging operation.

3. The machine vision-based unmanned monorail crane positioning system according to claim 1, characterized in that, The matching module is specifically used for: performing binocular matching calculations using the track contour data, node position data, and feature marker data, and obtaining three-dimensional feature data through disparity mapping transformation, including: The trajectory contour data and node position data are calculated and corrected through epipolar constraint optimization to obtain corrected feature data; Similarity calculation is performed on the corrected feature data and feature label data based on region cost aggregation, and the initial disparity data is output. The initial disparity data is filtered by depth constraint determination, and effective disparity data is generated by detail compensation processing. The effective disparity data is mapped to a spatial coordinate system using triangulation to obtain initial point cloud data; Outliers in the initial point cloud data are removed according to density clustering criteria, and optimized point cloud data is generated after noise reduction processing. A spatial feature descriptor is constructed for the optimized point cloud data, and three-dimensional feature data is output after feature matching and fusion.

4. The machine vision-based unmanned monorail crane positioning system according to claim 1, characterized in that, The tracking module is specifically used for: Feature descriptors are extracted from the initial position coordinate data through pyramid layering, and feature matching data is generated using an optical flow tracing algorithm. The feature matching data is segmented using a sliding window criterion, and segmented trajectory data is output by local optimization calculation. After performing attitude interpolation on the segmented trajectory data based on the Lie algebra space, a keyframe sequence is obtained after keyframe selection. For the keyframe sequence, a local covariance matrix is ​​constructed, and the accumulated error is eliminated by a sparse optimization algorithm to output optimized trajectory data; The state transition probability of the optimized trajectory data is calculated using a hidden Markov chain, and the position prediction data is obtained through maximum likelihood estimation. The location prediction data is globally constrained by the graph optimization framework and historical observations, and continuous location data is generated through iterative optimization.

5. The machine vision-based unmanned monorail crane positioning system according to claim 1, characterized in that, The sampling module is specifically used for: Based on Bézier curves, trajectory fitting and interpolation are performed on the continuous position data, and smooth trajectory data is obtained through curvature constraint optimization. The velocity-acceleration boundary is calculated on the smooth trajectory data using dynamic constraints, and dynamic reference data is generated by combining it with a safe distance threshold. The dynamic reference data is decomposed into principal component features using the recursive least squares criterion, and the trajectory principal metadata is output through feature reconstruction. A family of polynomial curves is constructed based on the trajectory master metadata, and the optimal path data is calculated by a nonlinear programming solver. Kinematic constraints are applied to the optimal path data using the Lagrange multiplier method, and trajectory control data is generated through iterative optimization. The trajectory control data is transformed into a sequence of velocity, acceleration and angular velocity using a variational inference framework, and the motion parameter data is output in combination with dynamic boundary constraints.

6. The machine vision-based unmanned monorail crane positioning system according to claim 1, characterized in that, The evaluation module is specifically used for: The motion parameter data is subjected to intrinsic mode extraction by variational decomposition, and state feature data is obtained by singular spectrum analysis. A probability density function is constructed based on the Gaussian mixture distribution for the state feature data, and state clustering data is generated by density clustering operation; The state clustering data is mapped to the decision space using the conditional entropy criterion, and the state evaluation data is output through information gain calculation. A robust controller is constructed based on the state assessment data, and deviation compensation data is obtained through Lyapunov stability analysis; Based on adaptive iterative learning, error propagation analysis is performed on the deviation compensation data, and corrected control data is obtained by feedback correction calculation. The corrected control data is then analyzed and optimized using recursive least squares, and the corrected data is output in conjunction with system dynamic constraints.

Images