Intelligent position control dynamic calibration method

Abstract

The invention relates to an intelligent position control dynamic calibration method, and relates to the technical field of precise motion control, and the intelligent position control dynamic calibration method comprises the steps: dividing a platform movement interval into a plurality of calibration motion segments according to a calibration motion path, and planning and generating a determinant scanning path; positioning a regional calibration point according to the determinant scanning path, extracting a calibration point image, determining theoretical coordinates of the calibration point, and collecting actual coordinates of the calibration point; the theoretical coordinates of the calibration points and the actual coordinates of the calibration points are aligned, positioning error values are calculated, and positioning calibration coefficients are solved; forming a spatial error data set according to a calibration motion segment aggregation positioning calibration coefficient, and constructing a position error compensation model; outputting a compensation coordinate difference value according to the position error compensation model in combination with the target position coordinates, generating a motion control instruction, and determining actual position coordinates until the actual position coordinates coincide with the target position coordinates; and through a position error compensation model, the adaptability, precision and robustness of space error compensation are improved.

Description

Technical Field

[0001] This application relates to the field of precision motion control technology, and in particular to an intelligent position control dynamic calibration method. Background Technology

[0002] With the rapid development of ultra-precision machining and measurement technologies, ultra-precision motion platforms are finding increasingly wider applications in direct-write lithography, leading to ever-higher requirements for their positional accuracy. For two-dimensional motion platforms, the main error stems from systematic errors, such as the motion accuracy of servo motors, the installation accuracy of guide rails, and the accuracy of feedback measurement components. Correcting these systematic errors can effectively improve the platform's accuracy. Currently, while high-precision laser interferometers can be used to calibrate and compensate for the positional accuracy of motion platforms, they are often significantly affected by external environmental factors, and the operation process is quite cumbersome.

[0003] Existing patents disclose a positioning error compensation system and method for an XY motion platform. This system utilizes multimodal data fusion and normalization, combined with a neural network hybrid model to extract spatiotemporal features, and employs forget gates, input gates, and output gates to dynamically manage temporal errors. Furthermore, it uses an online adaptive training mechanism based on error source classification to update compensation parameters. This effectively addresses problems in existing technologies, such as encoder accuracy degradation due to mechanical wear and thermal deformation, the difficulty of PID algorithms in handling nonlinear errors, the susceptibility to interference and complex calibration of pure visual positioning, and the lack of online learning mechanisms. It reduces the overall positioning error to the micrometer level, improves the system's anti-interference robustness and real-time compensation capability, and significantly enhances positioning accuracy and production efficiency.

[0004] The existing technical solutions described above have the following drawbacks: 1. Existing mobile platforms mostly use linear motors or air-bearing platforms, which results in problems such as large system size, high cost, and reliance on original manufacturer support for parameter adjustment. This leads to inconvenience and insufficient flexibility for equipment manufacturers during actual debugging and maintenance, as well as excessive deviation in motion trajectory and inaccurate compensation. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this application achieves its goals through the following technical solutions: Firstly, this application provides an intelligent position control dynamic calibration method, comprising: Based on the preset calibration motion path, the platform movement interval is divided into several calibration motion segments, and a matrix scanning path is planned and generated. Based on the matrix scanning path, locate the calibration point in the region, trigger and extract the calibration point image, determine the theoretical coordinates of the calibration point, and collect the actual coordinates of the calibration point; Align the theoretical coordinates of the calibration point with the actual coordinates of the calibration point, calculate the positioning error value, and solve the positioning calibration coefficient of the calibration motion segment; The positioning calibration coefficients are aggregated based on the calibrated motion segments to form a spatial error dataset, and a position error compensation model is constructed. Based on the position error compensation model and the target position coordinates, the compensation coordinate difference is output, motion control commands are generated, and the actual position coordinates are determined until they coincide with the target position coordinates.

[0006] By adopting the above technical solution, a calibration motion trajectory is generated through a deterministic path planning algorithm, the actual coordinates of the calibration points are extracted based on image recognition technology, the positioning calibration coefficients of each motion segment are solved using the least squares method, a spatial error compensation model is constructed through a surface fitting algorithm, and finally, a precise motion command is generated using a real-time compensation mechanism. This achieves high-precision adaptive positioning of the entire working space of the two-dimensional platform, effectively improving the absolute accuracy and automation level of motion control.

[0007] This application further specifies the following steps for dividing the platform movement interval into several calibration motion segments according to a preset calibration motion path, and planning and generating a matrix scanning path: Based on the safety distance value and the mechanical limitations of the equipment, the total range of the platform is limited to obtain the platform's movement range; Based on the calibration parameters within the platform's movement range, the preset calibration motion path is analyzed and discretized to obtain a sequence of node coordinates; Based on the preset movement length and the overlapping point mechanism, the node coordinate sequence is aggregated and connected to obtain several calibrated movement segments; Based on the starting and ending points of each calibrated motion segment and a preset path topology mechanism, the platform's movement range is scanned to calculate the path point sequence. The calibrated motion segments and the path point sequence are associated and verified, and a smooth connection is achieved by combining the segment connection strategy to generate a matrix scanning path.

[0008] By adopting the above technical solution, integrating safety constraints and kinematic models, using a path discretization algorithm based on an overlap mechanism to divide the platform's movement interval into calibration motion segments, and using a path topology optimization algorithm to generate a matrix scanning path, combined with a smooth connection strategy between segments to ensure trajectory continuity, efficient and blind-spot-free full-area coverage scanning is achieved, significantly improving the automation level of the calibration process and the integrity of data acquisition.

[0009] This application further specifies the following steps for locating calibration points in the region based on the matrix scanning path, triggering and extracting calibration point images, determining the theoretical coordinates of the calibration points, and acquiring the actual coordinates of the calibration points: The matrix scanning path is serialized and decomposed to determine and move to the region calibration point, generating a visual trigger signal; The calibration point image of the region is obtained by acquiring an image of the calibration point based on the visual trigger signal. Template matching and corner detection are performed on the calibration point image to extract the calibration points and determine the coordinates of the calibration point image; The image coordinates of the calibration point are distorted according to the preset camera intrinsic and extrinsic parameters, and the coordinate system is transformed by combining the perspective transformation matrix to obtain the theoretical coordinates of the calibration point. The calibration point is located using the position feedback device inside the mobile platform, and its actual coordinates are determined.

[0010] By adopting the above technical solution, the calibration point is located by combining the path serialization algorithm with the visual triggering mechanism, the image coordinates are extracted by template matching and corner detection technology, and the theoretical coordinates are obtained by perspective transformation and distortion correction based on the camera calibration parameters. At the same time, the actual coordinates are obtained by integrating platform position feedback, thus realizing high-precision and automated calibration point data acquisition, effectively improving the calibration accuracy and efficiency of the positioning system.

[0011] This application further specifies the following steps for aligning the theoretical coordinates of the calibration point with the actual coordinates of the calibration point, calculating the positioning error value, and solving the positioning calibration coefficient of the calibration motion segment: Based on the timestamp, the theoretical coordinates of the calibration point (X) are determined. t i ,Y t i ) and the actual coordinates of the calibration point (X a i ,Y a i Perform mapping alignment and calculate positioning error values. ; Based on the theoretical coordinates and actual coordinates of the calibration point, a system of error correction equations is constructed. Let be the residual coordinates, k be the scaling factor matrix, and b be the translation deviation matrix; Based on the residual coordinates and a preset direction weighting coefficient matrix, an objective function is constructed using the residual coordinates. The minimum coordinates are solved by combining the preset convergence parameters. And calculate the positioning calibration coefficients k and b for the current calibrated motion segment.

[0012] By adopting the above technical solution, the theoretical coordinates and actual coordinates are aligned by timestamp mapping to construct a set of error correction equations. The scaling matrix k and translation matrix b are solved iteratively by weighted least squares method with directional weight coefficients. This achieves rapid convergence and high-precision calculation of the positioning calibration coefficients of the calibration motion segment, significantly improving the absolute positioning accuracy and anti-interference capability of the motion platform.

[0013] This application further specifies the following steps for: aggregating the positioning calibration coefficients based on the calibrated motion segment to form a spatial error dataset, and constructing a position error compensation model: Based on the calibration points of the calibration motion segment, the theoretical coordinates and residual coordinates of the calibration points are matched to construct error coordinate tuples; Based on the motion segment identifier, the positioning calibration coefficients and the error coordinate tuples are serialized and aggregated to form a spatial error dataset; Benchmark analysis is performed on the error coordinate tuples in the spatial error dataset to obtain the inter-segment coordinate distribution type and determine the fitting primitives; Based on the fitting primitives, the positioning calibration coefficients are fitted to construct a set of error compensation functions; Based on the first extraction branch of the feature extraction layer, the error coordinate tuples are cross-extracted to obtain the error cross matrix; The second extraction branch performs validation extraction on the spatial error dataset to obtain the spatial calibration matrix; Based on the first training branch of the iterative training layer and the spatial attention mechanism, the error cross matrix is ​​iteratively trained several times to generate the error weight matrix. The second training branch combines a temporal attention mechanism to perform several iterations of training on the spatial calibration matrix to generate a calibration weight matrix. Based on the mapping extraction branch of the data fusion layer, the error compensation function group is mapped and extracted to obtain the error compensation matrix; The global fusion branch combines the global attention mechanism to fuse the error weight matrix and the calibration weight matrix to obtain a comprehensive compensation weight matrix; Based on the fitting output branch of the verification output layer, the comprehensive compensation weight matrix is ​​used to perform fitting calculations on the theoretical coordinates of the calibration point and output the error compensation value. The twin verification branch performs simulation verification on the theoretical coordinates of the calibration point based on the error compensation matrix to obtain the simulation compensation amount; The error compensation value and the simulated compensation amount are verified and judged based on the update and optimization layer; If the absolute difference between the error compensation value and the simulated compensation amount is within a preset difference range, then the comprehensive compensation weight matrix is ​​determined to be the optimal solution, and the position error compensation model is obtained. Otherwise, the comprehensive compensation weight matrix is ​​optimized based on the error compensation matrix and the cross-entropy loss function until the comprehensive compensation weight matrix is ​​the optimal solution.

[0014] By adopting the above technical solution, an error compensation function set is constructed through a fitting primitive algorithm, and the error cross matrix and spatial calibration matrix are obtained based on a two-branch feature extraction mechanism. The weight matrix is ​​generated through iterative training using spatial and temporal attention mechanisms, and a comprehensive compensation weight matrix is ​​obtained through a global fusion algorithm. The model is then continuously optimized using the cross-entropy loss function until it converges, ultimately forming a high-precision position error compensation model, which significantly improves the adaptability, accuracy, and robustness of spatial error compensation.

[0015] This application further includes: performing benchmark analysis on the error coordinate tuples in the spatial error dataset to obtain the inter-segment coordinate distribution type and determine the fitting primitives, including: If the coordinate distribution between segments follows a global distribution pattern without local abrupt changes, then data fitting analysis is performed on the calibration points to construct a geometric figure; When the geometric figures are coplanar, the fitting primitive is determined to be a first-order polynomial function; when the geometric figures are non-coplanar, the fitting primitive is determined to be a second-order polynomial function or a higher-order polynomial function. If the inter-segment coordinate distribution type is a regular grid distribution, then based on the preset surface smoothness, the fitting primitive is determined to be a bilinear interpolation function or a bicubic interpolation function. If the inter-segment coordinate distribution type is a globally irregular local mutation, then the fitting primitive is determined to be a radial basis function.

[0016] By adopting the above technical solution, the inter-segment coordinate distribution type in the spatial error dataset is intelligently identified through benchmark analysis algorithm. For global regular distribution, a multinomial function is used for fitting, a bilinear or bicubic interpolation function is selected for regular grid distribution, and a radial basis function is applied for irregular local mutations. In this way, the optimal fitting primitive is adaptively determined, achieving high-precision modeling and compensation for complex errors, and significantly improving the adaptability and robustness of the system.

[0017] This application further specifies the following steps for: based on the position error compensation model and the target position coordinates, outputting the compensation coordinate difference, generating motion control commands, and determining the actual position coordinates until they coincide with the target position coordinates: Based on the position error compensation model, the target position coordinates are calculated to compensate for errors, and the compensated coordinate difference is output. The motion control command is generated by calculating the compensation coordinate difference and combining it with the target position coordinates. The motion control command is parsed and interpolated based on the orientation of the motion axis to generate an axis displacement pulse sequence; The moving platform is controlled according to the axial displacement pulse sequence, and the actual position coordinates are determined in conjunction with the internal position feedback device. The actual position coordinates are compared with the target position coordinates. If they are different, a data fine-tuning instruction is generated based on the difference between them. The mobile platform is displaced according to the data fine-tuning instructions until the two overlap.

[0018] By adopting the above technical solution, the compensation difference of the target position is calculated through the position error compensation model. Combined with the motion control algorithm, instructions are generated and parsed into an axis displacement pulse sequence to drive the platform to move. At the same time, the actual position is monitored in real time using a position feedback device. The difference between the position and the target position is compared in a closed loop and iteratively fine-tuned until precise overlap is achieved, thereby achieving the advantages of high-precision positioning and adaptive control.

[0019] Secondly, this application also provides an intelligent position control dynamic calibration system, which adopts the following technical solution: An intelligent position control dynamic calibration system, comprising: The main control module is bidirectionally connected to the machine vision module and the mobile platform. It is used to divide the platform's movement range into several calibration motion segments, plan and generate a matrix scanning path, calculate the positioning error value, solve the positioning calibration coefficient, build a position error compensation model, and generate and distribute motion control commands. The machine vision module is used to locate the calibration point in the region according to the matrix scanning path, acquire and extract the calibration point image, and obtain the theoretical coordinates of the calibration point. The mobile platform is used to locate the actual coordinates of the calibration point according to the built-in position feedback device, and to feed back the actual position coordinates until they coincide with the target position coordinates according to the motion control command.

[0020] By adopting the above technical solution, the calibration path is planned and an error compensation model is constructed through the matrix scanning algorithm and surface fitting technology of the main control module. The theoretical coordinates are obtained by combining the template matching and perspective transformation of the machine vision module. The actual position is iteratively approximated by the closed-loop control algorithm of the mobile platform. Finally, fully automatic and high-precision motion platform calibration and positioning control are realized.

[0021] Thirdly, this application also provides a computer storage medium, which adopts the following technical solution: A computer storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.

[0022] In summary, the beneficial technical effects of this application are as follows: By accurately compensating for motion trajectory through a position error compensation model, it is possible to reduce system size and cost while achieving convenient parameter debugging and high-precision positioning control, effectively solving the dilemma faced by traditional solutions. An error compensation function set is constructed by fitting primitive algorithm, and the error cross matrix and spatial calibration matrix are obtained based on the bi-branch feature extraction mechanism. The weight matrix is ​​generated by iterative training combined with spatial and temporal attention mechanism, and then the comprehensive compensation weight matrix is ​​obtained by global fusion. The model is continuously optimized until it converges, and finally a high-precision position error compensation model is formed, which significantly improves the adaptability, accuracy and robustness of spatial error compensation. Attached Figure Description

[0023] Figure 1 This is a flowchart of an intelligent position control dynamic calibration method according to one embodiment of this application; Figure 2 This is a flowchart illustrating step S4 of one embodiment of this application; Figure 3 This is a flowchart illustrating step C of one embodiment of this application; Figure 4 This is a structural diagram of an intelligent position control dynamic calibration system according to one embodiment of this application. Detailed Implementation

[0024] If the absolute difference between the error compensation value and the simulated compensation amount is within a preset difference range, then the comprehensive compensation weight matrix is ​​determined to be the optimal solution, and the position error compensation model is obtained. Otherwise, the comprehensive compensation weight matrix is ​​optimized based on the error compensation matrix and the cross-entropy loss function until the comprehensive compensation weight matrix is ​​the optimal solution.

[0025] Reference Figure 3 Step C includes: If the coordinate distribution between segments follows a global distribution pattern without local abrupt changes, then data fitting analysis is performed on the calibration points to construct a geometric figure; When the geometric figures are coplanar, the fitting primitive is determined to be a first-order polynomial function; when the geometric figures are non-coplanar, the fitting primitive is determined to be a second-order polynomial function or a higher-order polynomial function. If the inter-segment coordinate distribution type is a regular grid distribution, then based on the preset surface smoothness, the fitting primitive is determined to be a bilinear interpolation function or a bicubic interpolation function. If the inter-segment coordinate distribution type is a globally irregular local mutation, then the fitting primitive is determined to be a radial basis function.

[0026] The specific steps of step S5 include: Based on the position error compensation model, the target position coordinates are calculated to compensate for errors, and the compensated coordinate difference is output. The motion control command is generated by calculating the compensation coordinate difference and combining it with the target position coordinates. The motion control command is parsed and interpolated based on the orientation of the motion axis to generate an axis displacement pulse sequence; The moving platform is controlled according to the axial displacement pulse sequence, and the actual position coordinates are determined in conjunction with the internal position feedback device. The actual position coordinates are compared with the target position coordinates. If they are different, a data fine-tuning instruction is generated based on the difference between them. The mobile platform is displaced according to the data fine-tuning instructions until the two overlap.

[0027] The implementation principle of this embodiment is as follows: the platform's movement range is divided by a safe distance and mechanical constraints, and a matrix scanning path is generated using path topology and overlap point mechanisms; theoretical coordinates of calibration points are extracted based on visual triggering and image processing technology, and actual coordinates are obtained in conjunction with a position feedback device; the positioning error is calculated after aligning the coordinates with timestamps, and the calibration coefficient is solved using the weighted least squares method; all motion segment data are aggregated to form a spatial error dataset, and after determining the fitting primitives (such as polynomials or radial basis functions) through benchmark analysis, an error compensation function set is constructed, and a comprehensive compensation weight matrix is ​​generated through iterative training using multi-branch feature extraction and a spatial-temporal attention mechanism; finally, the position error compensation model outputs the compensation difference, and the actual position and the target position are precisely coincident through motion control commands and closed-loop iteration, achieving high-precision adaptive motion control.

[0028] An intelligent position control dynamic calibration system, comprising: The main control module is bidirectionally connected to the machine vision module and the mobile platform. It is used to divide the platform's movement range into several calibration motion segments, plan and generate a matrix scanning path, calculate the positioning error value, solve the positioning calibration coefficient, build a position error compensation model, and generate and distribute motion control commands. The machine vision module is used to locate the calibration point in the region according to the matrix scanning path, acquire and extract the calibration point image, and obtain the theoretical coordinates of the calibration point. The mobile platform is used to locate the actual coordinates of the calibration point according to the built-in position feedback device, and to feed back the actual position coordinates until they coincide with the target position coordinates according to the motion control command.

[0029] In this embodiment, the machine vision module can be an industrial camera, and the main control module can be a control system. Example

[0030] The calibration motion path of the platform is pre-planned, and based on the calibration tools (such as calibration boards) used by the machine vision system (industrial camera), the motion range of the platform is divided into several regular calibration areas (e.g., 20 segments each on the X and Y axes). A matrix scanning path that can traverse the nodes of each area is planned for subsequent automatic point finding.

[0031] Once the platform moves to the preset calibration position, the vision system is triggered to acquire images and automatically identify and extract the sub-pixel-level image coordinates of standard markers (such as feature points on the calibration board) within the field of view. These coordinates are then converted into high-precision spatial coordinates in the platform's coordinate system using camera calibration data. Simultaneously, the vision system records the position coordinate values ​​returned by the platform's built-in position feedback device (such as a grating ruler or encoder) at the moment it identifies the calibration point.

[0032] For each movement segment, establish the correspondence between its theoretical target position and the actual position value collected, calculate the coordinate deviation between the two, and use the deviation value directly as the positioning error, or further calculate the calibration coefficient of the position based on it.

[0033] The calibration coefficients of all motion segments are integrated to form a complete spatial error dataset. Based on this dataset, a surface fitting algorithm is used to construct a continuous mathematical function that takes the platform's theoretical coordinates as input and the intelligent position control dynamic calibration value as output, thus creating a position error compensation model covering the entire workspace of the platform.

[0034] Surface fitting is the process of approximating a three-dimensional data point (x, y, z) from discrete XY data points in a plane using a mathematical model, in order to construct a continuous and smooth surface. Here, x and y are the actual coordinates of the points, and z is the positioning error coefficient. Commonly used surface fitting algorithms based on discrete XY points include: Assuming the surface satisfies a polynomial function form, in scenarios where the global distribution of calibration points is relatively clear and there are no significant local abrupt changes (such as smooth experimental data or simple geometric modeling), the sum of squared errors between discrete points and the polynomial surface is minimized using the least squares method to solve for the polynomial coefficients. Common polynomial orders are as follows: Linear polynomial (plane fitting): z = a_0 + a_1x + a_2y, suitable for scenarios where the data is approximately planar; The quadratic polynomial z = a_0 + a_1x + a_2y + a_3x^2 + a_4xy + a_5y^2 can fit a slightly curved surface. Higher-order polynomials: order n≥3, It can fit more complex surfaces, but it is prone to overfitting.

[0035] When the calibration points are distributed in a regular rectangular grid (such as pixel images or uniformly sampled experimental data), the XY plane is divided into a rectangular grid. Within each grid, a bilinear interpolation function z = a_0 + a_1x + a_2y + a_3xy is constructed to fit the local surface using the Z values ​​of the four vertices. When dealing with regular mesh data requiring high surface smoothness (such as medical image reconstruction and high-precision graphics scaling), in a rectangular mesh, each mesh depends on the Z values ​​of its 16 surrounding vertices, and is processed using a bicubic interpolation function. Fitting, while ensuring the continuity of the first and second derivatives of the surface at the grid boundary; When dealing with irregularly distributed discrete points and scenarios where data exhibits localized complex variations (such as modeling the surface of mechanical parts or geological exploration data), it is assumed that the fitted surface is a linear combination of radial basis functions: Where Φ(r) is the radial basis function, P(x,y) is a low-order polynomial (such as a first-order polynomial) used to eliminate global trend error; λi is the coefficient to be determined, which is solved by the least squares method or interpolation conditions.

[0036] During actual positioning, the desired theoretical target position is input into the established spatial error model. The compensated coordinate values ​​required by the platform's internal feedback device to achieve the theoretical position are calculated, and these compensated coordinate values ​​are sent to the platform as new motion commands to offset system errors.

[0037] The implementation principle of this embodiment is as follows: the main control module divides the platform's movement range into calibration motion segments based on safety distance and mechanical limitations, and plans a matrix scanning path through path topology and overlap point mechanism; the machine vision module serializes and decomposes the path and triggers image acquisition, extracts the image coordinates of calibration points using template matching and corner detection, and then obtains the theoretical coordinates of the calibration points through distortion correction and perspective transformation using camera calibration parameters; at the same time, the moving platform records the actual coordinates through a position feedback device; The positioning error value is calculated by aligning theoretical coordinates and actual coordinates with timestamps, and the positioning calibration coefficient is solved by weighted least squares method combined with direction weight matrix. After aggregating all motion segment data to form a spatial error dataset, an error compensation function set is constructed by surface fitting algorithm according to the type of inter-segment coordinate distribution (multinomial function is selected for global regular distribution, bilinear or bicubic interpolation function is selected for regular grid distribution, and radial basis function is selected for local abrupt change). Then, a comprehensive compensation weight matrix is ​​generated by multi-branch feature extraction and spatial-temporal attention mechanism iterative training. After optimization, the position error compensation model is obtained. Ultimately, during actual positioning, the model outputs a compensation difference for the target position coordinates, generates motion control commands, and parses them into an axis displacement pulse sequence to drive the platform to move. Through a position feedback device, the actual position is compared with the target position in real time for closed-loop iterative fine-tuning until they are precisely aligned, thus achieving high-precision adaptive motion control.

[0038] A computer storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.

[0039] The embodiments described in this specific implementation are preferred embodiments of this application and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for intelligent position control dynamic calibration, characterized in that, include: Based on the preset calibration motion path, the platform movement interval is divided into several calibration motion segments, and a matrix scanning path is planned and generated. Based on the matrix scanning path, locate the calibration point in the region, trigger and extract the calibration point image, determine the theoretical coordinates of the calibration point, and collect the actual coordinates of the calibration point; Align the theoretical coordinates of the calibration point with the actual coordinates of the calibration point, calculate the positioning error value, and solve the positioning calibration coefficient of the calibration motion segment; The positioning calibration coefficients are aggregated based on the calibrated motion segments to form a spatial error dataset, and a position error compensation model is constructed. Based on the position error compensation model and the target position coordinates, the compensation coordinate difference is output, motion control commands are generated, and the actual position coordinates are determined until they coincide with the target position coordinates.

2. The intelligent position control dynamic calibration method according to claim 1, characterized in that, The specific steps of dividing the platform movement interval into several calibration movement segments according to the preset calibration movement path, and planning and generating a matrix scanning path include: Based on the safety distance value and the mechanical limitations of the equipment, the total range of the platform is limited to obtain the platform's movement range; Based on the calibration parameters within the platform's movement range, the preset calibration motion path is analyzed and discretized to obtain a sequence of node coordinates; Based on the preset movement length and the overlapping point mechanism, the node coordinate sequence is aggregated and connected to obtain several calibrated movement segments; Based on the starting and ending points of each calibrated motion segment and a preset path topology mechanism, the platform's movement range is scanned to calculate the path point sequence. The calibrated motion segments and the path point sequence are associated and verified, and a smooth connection is achieved by combining the segment connection strategy to generate a matrix scanning path.

3. The intelligent position control dynamic calibration method according to claim 1, characterized in that: The specific steps of locating the calibration point in the region according to the matrix scanning path, triggering and extracting the calibration point image, determining the theoretical coordinates of the calibration point, and acquiring the actual coordinates of the calibration point include: The matrix scanning path is serialized and decomposed to determine and move to the region calibration point, generating a visual trigger signal; The calibration point image of the region is obtained by acquiring an image of the calibration point based on the visual trigger signal. Template matching and corner detection are performed on the calibration point image to extract the calibration points and determine the coordinates of the calibration point image; The image coordinates of the calibration point are distorted according to the preset camera intrinsic and extrinsic parameters, and the coordinate system is transformed by combining the perspective transformation matrix to obtain the theoretical coordinates of the calibration point. The calibration point is located using the position feedback device inside the mobile platform, and its actual coordinates are determined.

4. The intelligent position control dynamic calibration method according to claim 1, characterized in that, The specific steps for aligning the theoretical coordinates and actual coordinates of the calibration point, calculating the positioning error value, and solving the positioning calibration coefficient of the calibration motion segment include: Based on the timestamp, the theoretical coordinates of the calibration point (X) are determined. t i ,Y t i ) and the actual coordinates of the calibration point (X a i ,Y a i Perform mapping alignment and calculate positioning error values. ; Based on the theoretical coordinates and actual coordinates of the calibration point, a system of error correction equations is constructed. Let be the residual coordinates, k be the scaling factor matrix, and b be the translation deviation matrix; Based on the residual coordinates and a preset direction weighting coefficient matrix, an objective function is constructed using the residual coordinates. The minimum coordinates are solved by combining the preset convergence parameters. And calculate the positioning calibration coefficients k and b for the current calibrated motion segment.

5. The intelligent position control dynamic calibration method according to claim 1, characterized in that, The specific steps for aggregating the positioning calibration coefficients based on the calibrated motion segments to form a spatial error dataset and constructing a position error compensation model include: Based on the calibration points of the calibration motion segment, the theoretical coordinates and residual coordinates of the calibration points are matched to construct error coordinate tuples; Based on the motion segment identifier, the positioning calibration coefficients and the error coordinate tuples are serialized and aggregated to form a spatial error dataset; Benchmark analysis is performed on the error coordinate tuples in the spatial error dataset to obtain the inter-segment coordinate distribution type and determine the fitting primitives; The positioning calibration coefficients are fitted using the fitted primitives to construct a set of error compensation functions.

6. The intelligent position control dynamic calibration method according to claim 5, characterized in that, The specific steps of aggregating the positioning calibration coefficients based on the calibrated motion segments to form a spatial error dataset and constructing a position error compensation model further include: Based on the first extraction branch of the feature extraction layer, the error coordinate tuples are cross-extracted to obtain the error cross matrix; The second extraction branch performs validation extraction on the spatial error dataset to obtain the spatial calibration matrix; Based on the first training branch of the iterative training layer and the spatial attention mechanism, the error cross matrix is ​​iteratively trained several times to generate the error weight matrix. The second training branch combines a temporal attention mechanism to perform several iterations of training on the spatial calibration matrix to generate a calibration weight matrix. Based on the mapping extraction branch of the data fusion layer, the error compensation function group is mapped and extracted to obtain the error compensation matrix; The global fusion branch combines the global attention mechanism to fuse the error weight matrix and the calibration weight matrix to obtain a comprehensive compensation weight matrix; Based on the fitting output branch of the verification output layer, the comprehensive compensation weight matrix is ​​used to perform fitting calculations on the theoretical coordinates of the calibration point and output the error compensation value. The twin verification branch performs simulation verification on the theoretical coordinates of the calibration point based on the error compensation matrix to obtain the simulation compensation amount; The error compensation value and the simulated compensation amount are verified and judged based on the update and optimization layer; If the absolute difference between the error compensation value and the simulated compensation amount is within a preset difference range, then the comprehensive compensation weight matrix is ​​determined to be the optimal solution, and the position error compensation model is obtained. Otherwise, the comprehensive compensation weight matrix is ​​optimized based on the error compensation matrix and the cross-entropy loss function until the comprehensive compensation weight matrix is ​​the optimal solution.

7. The intelligent position control dynamic calibration method according to claim 5, characterized in that, The benchmark analysis of the error coordinate tuples in the spatial error dataset to obtain the inter-segment coordinate distribution type and determine the fitting primitives includes: If the coordinate distribution between segments follows a global distribution pattern without local abrupt changes, then data fitting analysis is performed on the calibration points to construct a geometric figure; When the geometric figures are coplanar, the fitting primitive is determined to be a first-order polynomial function; when the geometric figures are non-coplanar, the fitting primitive is determined to be a second-order polynomial function or a higher-order polynomial function. If the inter-segment coordinate distribution type is a regular grid distribution, then based on the preset surface smoothness, the fitting primitive is determined to be a bilinear interpolation function or a bicubic interpolation function. If the inter-segment coordinate distribution type is a globally irregular local mutation, then the fitting primitive is determined to be a radial basis function.

8. The intelligent position control dynamic calibration method according to claim 1, characterized in that, The specific steps of outputting the compensation coordinate difference based on the position error compensation model and the target position coordinates, generating motion control commands, determining the actual position coordinates, and continuing until they coincide with the target position coordinates include: Based on the position error compensation model, the target position coordinates are calculated to compensate for errors, and the compensated coordinate difference is output. The motion control command is generated by calculating the compensation coordinate difference and combining it with the target position coordinates. The motion control command is parsed and interpolated based on the orientation of the motion axis to generate an axis displacement pulse sequence; The moving platform is controlled according to the axial displacement pulse sequence, and the actual position coordinates are determined in conjunction with the internal position feedback device. The actual position coordinates are compared with the target position coordinates. If they are different, a data fine-tuning instruction is generated based on the difference between them. The mobile platform is displaced according to the data fine-tuning instructions until the two overlap.

9. An intelligent position control dynamic calibration system for implementing the method as described in any one of claims 1-8, characterized in that, include: The main control module is bidirectionally connected to the machine vision module and the mobile platform. It is used to divide the platform's movement range into several calibration motion segments, plan and generate a matrix scanning path, calculate the positioning error value, solve the positioning calibration coefficient, build a position error compensation model, and generate and distribute motion control commands. The machine vision module is used to locate the calibration point in the region according to the matrix scanning path, acquire and extract the calibration point image, and obtain the theoretical coordinates of the calibration point. The mobile platform is used to locate the actual coordinates of the calibration point according to the built-in position feedback device, and to feed back the actual position coordinates until they coincide with the target position coordinates according to the motion control command.

10. A computer storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-8.

Images