A kind of automobile parts processing robot and control method based on fixed-point navigation technology

Abstract

The present invention relates to the field of automobile production and processing technology, and in particular to an automobile parts processing robot and control method based on fixed-point navigation technology. If the detection result of the special-shaped processing surface of an automobile part is an ordinary special-shaped processing area, the regional boundary mapping matrix of the ordinary special-shaped processing area is mapped according to the established processing technology of the first preset processing shape, and a Voronoi diagram is constructed based on the regional boundary mapping matrix and the fixed-point navigation technology is introduced to plan the control route; if the detection result of the special-shaped processing surface of the automobile part is an extraordinary special-shaped processing area, the optimal working condition space field of the necessary processing mechanism for the easy-to-touch spline area is established, and the coverage status of the optimal working condition space field and the redundant space field is analyzed by the fixed-point navigation technology to plan the control route. The present invention is based on the precise control of the processing path of the robot for different types of special-shaped processing surfaces on automobile parts supported by the fixed-point navigation technology, which can improve the processing accuracy and quality of the special-shaped surfaces of automobile parts.

Description

Technical Field

[0001] The present invention relates to the technical field of automobile production and processing, and in particular to an automobile parts processing manipulator based on fixed-point navigation technology and a control method thereof. Background Art

[0002] Auto parts processing is a crucial link in the automobile manufacturing process. It usually involves processing various raw materials into parts that meet the design and performance requirements of the automobile. These parts can be divided into multiple categories such as engines, transmission systems, bodies, suspension systems or brake systems. Auto parts processing is a complex and delicate process. With the continuous development of the automobile industry, auto parts processing technology is also constantly improving and innovating. At present, the mainstream equipment for auto parts processing is still robot processing. Robots can significantly improve processing speed and production efficiency, shorten processing cycle time, and reduce human intervention errors, ensuring the controllability of processing accuracy. However, the existing auto parts processing robots still have large control deviations for the processing control of special-shaped surfaces of different complexities, and the selection and positioning accuracy of processing points is low, resulting in excessive processing, omission or missing of special-shaped surfaces when the robot controls the processing of auto parts on the processing mechanism under the support of fixed-point navigation technology. This makes it difficult for the processing effect of special-shaped surfaces of auto parts to meet the expected requirements, and at the same time negatively optimizes the processing route decision performance of fixed-point navigation technology, reducing the accuracy and efficiency of auto parts processing. Summary of the invention

[0003] The invention overcomes the shortcomings of the prior art and provides an automobile parts processing manipulator based on fixed-point navigation technology and a control method.

[0004] To achieve the above object, the technical solution adopted by the present invention is:

[0005] The first aspect of the present invention provides a method for controlling an automobile parts processing robot based on fixed-point navigation technology, comprising the following steps:

[0006] S102: Acquire global laser point cloud data of automobile parts, obtain special-shaped point clouds by sparsely calculating normal vector deviations of the global laser point cloud data, construct special-shaped fields based on smoothness learning constraints of the special-shaped point clouds, and generate smooth special-shaped field models of automobile parts;

[0007] S104: obtaining the designated processing area and processing range of the automobile part, extracting a number of low-frequency sparse subband signals of each designated processing area in the smooth special-shaped field model, performing function segmentation and VC dimension induction analysis on the low-frequency sparse subband signals using the processing range, and generating the detection result of the special-shaped processing surface of the automobile part;

[0008] S106: If the detection result of the special-shaped processing surface of the automobile part is a common special-shaped processing area, then the unprocessed three-dimensional model of the common special-shaped processing area is compared with the redundant model area block of the ideal three-dimensional model, and the redundant model area block is mapped according to the established processing technology of the first preset processing shape to obtain a region boundary mapping matrix, and a Voronoi diagram is constructed based on the region boundary mapping matrix and a fixed-point navigation technology is introduced to plan a control route to obtain a type of control strategy;

[0009] S108: If the detection result of the special-shaped processing surface of the automobile part is an unusual special-shaped processing area, the mean shift is used to cluster the easy-to-touch spline area of ​​the unusual special-shaped processing area, and the optimal working condition space field of the necessary processing mechanism for the easy-to-touch spline area is established. The fixed-point navigation technology is combined with the analysis of the coverage status of the optimal working condition space field and the redundant space field of the easy-to-touch spline area to plan the control route, and the second type of control strategy is obtained.

[0010] More specifically, the step S102 includes the following steps:

[0011] Obtain the specified processing task area of ​​the manipulator, and scan the automobile parts in the current specified processing task area through a laser scanner to obtain global laser point cloud data of the automobile parts;

[0012] Construct a sparse graph, fit the global laser point cloud data with topology in the sparse graph, generate a sparse topological composition of the automobile parts, and capture and extract the sparse nodes of the normal vectors of the automobile parts based on the sparse topological composition;

[0013] Obtain the adjacent normal vector values ​​of each normal vector sparse node and preset the critical deviation of the irregular normal vector. If the normal vector deviation between the adjacent normal vector values ​​is greater than the critical deviation of the irregular normal vector, calibrate the point cloud corresponding to the adjacent normal vector value as an irregular point cloud.

[0014] Obtain all the special-shaped point clouds contained in the global laser point cloud data of the automobile parts, construct a neighborhood undirected association space based on the global laser point cloud, define all the special-shaped point clouds as neighborhood graph nodes and map them into the neighborhood undirected association space to generate a neighborhood undirected graph of the special-shaped point clouds;

[0015] The Gibbs distribution method is introduced to estimate the conditional probability between each irregular point cloud in the Gibbs distribution method, and the joint probability distribution of the irregular point cloud is obtained. The potential function of each irregular point cloud is determined according to the joint probability distribution.

[0016] Based on the potential function, node pair items between each shaped point cloud are obtained in the undirected graph of the neighborhood of the shaped point cloud, a smoothness constraint threshold is preset based on the node pair items, and the data of each shaped point cloud is learned by maximum a posteriori estimation, and the smoothness index between each node pair item corresponding to the shaped point cloud in the learning process is output;

[0017] According to the potential function, the smoothness index between the corresponding irregular point clouds of each node pair is constrained during the learning process until the smoothness index meets the smoothness constraint threshold. The above constraint steps are repeated to traverse the learning data of all irregular point clouds and model them, and finally a smooth irregular field model of automobile parts is generated.

[0018] More specifically, the step S104 includes the following steps:

[0019] Obtaining processing requirements of automobile parts, obtaining designated processing areas of automobile parts and processing ranges of each designated processing area according to the processing requirements, and dividing the smooth profiled field model into a plurality of sub-model sample point sets based on the designated processing areas;

[0020] Obtain the processing log of the manipulator, and extract the laser signals scanned by the laser scanner for each sub-model sample point set in the smooth profile field model state of the automobile parts through the processing log, which are defined as smooth laser signals;

[0021] The wavelet transform algorithm is introduced, and the wavelet basis function is preset in the wavelet transform algorithm based on the smooth shaped field model. In the wavelet transform algorithm, several smooth laser signals of each sub-model sample point set are decomposed into N sub-band signals;

[0022] A low-pass filter is constructed according to the wavelet basis function, and each sub-band signal is subjected to filtering and sparse decomposition processing by using the low-pass filter to attenuate the high-frequency signal until a preset number of sparse decomposition layers is reached, thereby obtaining a number of low-frequency sparse sub-band signals of each sub-model sample point set;

[0023] Preset an expected partitioning function interval, determine multiple partitioning function criteria according to the processing range, and divide each sub-model sample point set one by one based on each partitioning function criterion to obtain multiple result partitioning functions;

[0024] When the expected partitioning function interval contains one or more result partitioning functions, the result partitioning functions are summarized and added to obtain the partitioning function summary addition amount, and the VC dimension value of each sub-model sample point set is determined according to the partitioning function summary addition amount to generate the special-shaped machining surface detection result of the automobile part;

[0025] If the VC dimension value is less than the preset VC dimension value, the designated processing area corresponding to the calibration sub-model sample point set is a common special-shaped processing area;

[0026] If the VC dimension value is greater than the preset VC dimension value, the designated processing area corresponding to the sub-model sample point set is calibrated as an unusual special-shaped processing area.

[0027] More specifically, the step S106 includes the following steps:

[0028] If the detection result of the special-shaped processing surface of the automobile part shows that the designated processing area is a common special-shaped processing area, then the preset processing shapes of each common special-shaped processing area in the automobile part are obtained according to the processing requirements, which are defined as first preset processing shapes, and the established processing technology of the first preset processing shape is obtained;

[0029] Obtaining processing design drawings of common special-shaped processing areas, constructing an ideal three-dimensional model of the common special-shaped processing areas according to the processing design drawings through SolidWorks design software, and simultaneously establishing an unprocessed three-dimensional model of the common special-shaped processing areas in automobile parts;

[0030] The non-overlapping model area between the ideal three-dimensional model and the unprocessed three-dimensional model in the ordinary special-shaped processing area is stripped off and marked as a model area block to be redundant. At this time, the model boundary points of the ideal three-dimensional model are identified and obtained, which are defined as the first model boundary points, and the model boundary points of the unprocessed three-dimensional model are identified and obtained, which are defined as the second model boundary points;

[0031] Taking the established processing technology as a mapping specification, mapping the boundary points of the second model to the boundary points of the first model based on the mapping specification until the region block of the model to be redundant completely covers the mapped boundary points, thereby generating a region boundary mapping matrix of the region block of the model to be redundant;

[0032] The historical difficulty coefficient of a given processing technology when processing the region boundary surface matrix is ​​obtained, and a Voronoi diagram of an ordinary special-shaped processing area is constructed based on the historical difficulty coefficient and the region boundary mapping matrix. The Voronoi diagram is analyzed by fixed-point navigation technology to plan the processing control route of the robot, and a class of control strategies is obtained.

[0033] More specifically, in the step S106, the historical difficulty coefficient of the established processing technology when processing the regional boundary surface matrix is ​​obtained, the Voronoi diagram of the common special-shaped processing area is constructed based on the historical difficulty coefficient and the regional boundary mapping matrix, and the Voronoi diagram is analyzed by fixed-point navigation technology to plan the processing control route of the manipulator to obtain a type of control strategy, which specifically includes the following steps:

[0034] Based on the big data network, multiple historical processing cases associated with the common special-shaped processing area are obtained, and the VC dimension value and the established processing technology are comprehensively identified through the multiple historical processing cases to identify the historical difficulty coefficient of the established processing technology when processing the boundary surface matrix of the area;

[0035] The ordinary special-shaped processing area is divided into several sub-areas according to the area boundary mapping matrix, and a processing seed point set of each sub-area is generated based on the historical difficulty coefficient, and a series of perpendicular medians are formed by the combination of every two processing seed points, and the Voronoi adjacent edges of the sub-areas and the intersection points of each Voronoi adjacent edge are determined according to the series of perpendicular medians;

[0036] Construct a Voronoi diagram of the common special-shaped processing area according to the intersection of the Voronoi adjacent edges of the sub-area and each Voronoi adjacent edge, and obtain the point distance distribution pattern between each processing seed point in the Voronoi diagram and each boundary point of the redundant model area block;

[0037] Based on the preset maximum service distance threshold of the region boundary mapping matrix, if the distance between the processing seed point in the Voronoi diagram and the boundary point of the region block of the model to be redundant is greater than the maximum service distance threshold, the processing seed point is removed or ignored;

[0038] If the distance between the processing seed point and the boundary point of the redundant model area block in the Voronoi diagram is less than the maximum service distance threshold, the processing seed point is retained and marked as the optimal processing seed point, and all the optimal processing seed points are gathered to construct the optimal processing seed point network of the common special-shaped processing area;

[0039] Fixed-point navigation technology is introduced. The processing route is planned for the optimal processing seed point network of the ordinary special-shaped processing area through the fixed-point navigation technology, and the fixed-point navigation processing route of the ordinary special-shaped processing area is generated. Based on the fixed-point navigation processing route, the manipulator is controlled to control the processing of the ordinary special-shaped processing area, and a class of control strategies is obtained.

[0040] More specifically, the step S108 includes the following steps:

[0041] If the detection result of the special-shaped processing surface of the automobile part shows that the designated processing area is an unusual special-shaped processing area, a preset processing shape of the unusual special-shaped processing area in the automobile part is obtained according to the processing requirements, which is defined as a second preset processing shape;

[0042] Construct a spline model of the unusual special-shaped processing area, set the Gaussian kernel function and the kernel function bandwidth range based on the VC dimension value of the unusual special-shaped processing area, and calculate the mean shift vector of each spline in the spline model within the kernel function bandwidth range;

[0043] The mean shift clustering algorithm is introduced to move each spline in the spline model along the corresponding mean shift vector to obtain the current moving distance value. If the current moving distance value is less than the preset moving distance value, the movement is stopped and a cluster cluster is generated at the current moving position to obtain the VC dimension value-spline clustering result;

[0044] The spline model of the unusual special-shaped processing area is obtained through the VC dimension value-spline clustering result. There is a spline area with a VC dimension value greater than the preset VC dimension value, which is marked as an easy-to-touch spline area. The spline of the easy-to-touch spline area is converted into a tensor spline to obtain the tension control point array of the tensor spline.

[0045] Obtaining a necessary processing mechanism of an automobile part, extracting optimal historical working condition data of the necessary processing mechanism for processing the tension control point array through a processing log, constructing an optimal working condition space domain of the necessary processing mechanism based on the optimal historical working condition data, and establishing a redundant space domain of an easy-to-touch spline area based on a second preset processing shape;

[0046] Determine whether the optimal working condition space field completely covers the redundant space field. If not, obtain the uncovered space field volume between the optimal working condition space field and the redundant space field, and adjust the processing vector of the unusual special-shaped processing area according to the uncovered space field volume. After the adjustment is completed, locate the first batch of special-shaped processing points of the processing vector;

[0047] If it is completely covered or there is interference from the external space after it is completely covered, the external interference volume is calculated, and the processing vector of the unusual special-shaped processing area is adjusted based on the external interference volume to locate the second batch of special-shaped processing points of the processing vector;

[0048] Through the fixed-point navigation technology, the first batch of special-shaped processing points and the second batch of special-shaped processing points are comprehensively analyzed and planned, and the fixed-point navigation processing route of the unusual special-shaped processing area is generated. Based on the fixed-point navigation processing route, the robot is controlled to control the unusual special-shaped processing area to obtain the second type of control strategy.

[0049] The second aspect of the present invention provides an automobile parts processing robot based on fixed-point navigation technology, which is applied to any one of the above-mentioned automobile parts processing robot control methods based on fixed-point navigation technology, and specifically includes:

[0050] A robot body, which is used to control the transport and movement of the robot with any degree of freedom and position;

[0051] A manipulator clamping mechanism, wherein the manipulator clamping mechanism includes a clamping sub-joint, which is responsible for clamping the automobile parts to a designated processing area;

[0052] A point cloud data scanning module, wherein the point cloud data scanning module includes a laser scanner for scanning automobile parts within a currently specified processing task area;

[0053] A low-pass filtering module, wherein the low-pass filtering module includes a low-pass filter, which is responsible for filtering and sparsely decomposing each sub-band signal to attenuate the high-frequency signal to obtain a low-frequency sparse sub-band signal of each sub-model sample point set;

[0054] A fixed-point navigation module is used to locate special-shaped processing points and plan fixed-point navigation processing routes for ordinary special-shaped processing areas and extraordinary special-shaped processing areas based on the special-shaped processing points.

[0055] The present invention solves the technical defects existing in the background technology, and the beneficial technical effects of the present invention are:

[0056] If the detection result of the special-shaped processing surface of the automobile part is an ordinary special-shaped processing area, the redundant model area block of the ordinary special-shaped processing area is mapped according to the established processing technology of the first preset processing shape to obtain the area boundary mapping matrix, and the Voronoi diagram is constructed based on the area boundary mapping matrix and the fixed-point navigation technology is introduced to plan the control route to obtain a type of control strategy; if the detection result of the special-shaped processing surface of the automobile part is an unusual special-shaped processing area, the mean shift clustering is used to obtain the easy-to-touch spline area, and the optimal working condition space field of the necessary processing mechanism for the easy-to-touch spline area is established. The coverage of the optimal working condition space field and the redundant space field is analyzed by the fixed-point navigation technology to plan the control route to obtain the second type of control strategy. The present invention can accurately control the processing path of the manipulator for different types of special-shaped processing surfaces on automobile parts based on the support of fixed-point navigation technology, thereby avoiding the increase of rework rate caused by the processing error of the special-shaped surface and improving the processing precision and quality of automobile parts. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, drawings of other embodiments can be obtained based on these drawings without paying creative work.

[0058] Figure 1 The process of a control method for an automobile parts processing robot based on fixed-point navigation technology is shown. Figure 1 ;

[0059] Figure 2 The process of a control method for an automobile parts processing robot based on fixed-point navigation technology is shown. Figure 2 ;

[0060] Figure 3 The overall structure diagram of an automobile parts processing robot based on fixed-point navigation technology is shown. DETAILED DESCRIPTION

[0061] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0062] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0063] The first aspect of the present invention provides a method for controlling an automobile parts processing robot based on fixed-point navigation technology, such as Figure 1 As shown, the following steps are included:

[0064] S102: Acquire global laser point cloud data of automobile parts, obtain special-shaped point clouds by sparsely calculating normal vector deviations of the global laser point cloud data, construct special-shaped fields based on smoothness learning constraints of the special-shaped point clouds, and generate smooth special-shaped field models of automobile parts;

[0065] S104: obtaining the designated processing area and processing range of the automobile part, extracting a number of low-frequency sparse subband signals of each designated processing area in the smooth special-shaped field model, performing function segmentation and VC dimension induction analysis on the low-frequency sparse subband signals using the processing range, and generating the detection result of the special-shaped processing surface of the automobile part;

[0066] S106: If the detection result of the special-shaped processing surface of the automobile part is a common special-shaped processing area, then the unprocessed three-dimensional model of the common special-shaped processing area is compared with the redundant model area block of the ideal three-dimensional model, and the redundant model area block is mapped according to the established processing technology of the first preset processing shape to obtain a region boundary mapping matrix, and a Voronoi diagram is constructed based on the region boundary mapping matrix and a fixed-point navigation technology is introduced to plan a control route to obtain a type of control strategy;

[0067] S108: If the detection result of the special-shaped processing surface of the automobile part is an unusual special-shaped processing area, the mean shift is used to cluster the easy-to-touch spline area of ​​the unusual special-shaped processing area, and the optimal working condition space field of the necessary processing mechanism for the easy-to-touch spline area is established. The fixed-point navigation technology is combined with the analysis of the coverage status of the optimal working condition space field and the redundant space field of the easy-to-touch spline area to plan the control route, and the second type of control strategy is obtained.

[0068] It should be noted that the present invention is based on the relatively static processing mechanism and the relatively dynamic processing motion state of the robot gripping the automobile parts on the processing mechanism, that is, the dynamic robot gripping the automobile parts performs processing and transportation on the static processing mechanism.

[0069] More specifically, the step S102 includes the following steps:

[0070] Obtain the specified processing task area of ​​the manipulator, and scan the automobile parts in the current specified processing task area through a laser scanner to obtain global laser point cloud data of the automobile parts;

[0071] Construct a sparse graph, fit the global laser point cloud data with topology in the sparse graph, generate a sparse topological composition of the automobile parts, and capture and extract the sparse nodes of the normal vectors of the automobile parts based on the sparse topological composition;

[0072] Obtain the adjacent normal vector values ​​of each normal vector sparse node and preset the critical deviation of the irregular normal vector. If the normal vector deviation between the adjacent normal vector values ​​is greater than the critical deviation of the irregular normal vector, calibrate the point cloud corresponding to the adjacent normal vector value as an irregular point cloud.

[0073] Obtain all the special-shaped point clouds contained in the global laser point cloud data of the automobile parts, construct a neighborhood undirected association space based on the global laser point cloud, define all the special-shaped point clouds as neighborhood graph nodes and map them into the neighborhood undirected association space to generate a neighborhood undirected graph of the special-shaped point clouds;

[0074] The Gibbs distribution method is introduced to estimate the conditional probability between each irregular point cloud in the Gibbs distribution method, and the joint probability distribution of the irregular point cloud is obtained. The potential function of each irregular point cloud is determined according to the joint probability distribution.

[0075] Based on the potential function, node pair items between each shaped point cloud are obtained in the undirected graph of the neighborhood of the shaped point cloud, a smoothness constraint threshold is preset based on the node pair items, and the data of each shaped point cloud is learned by maximum a posteriori estimation, and the smoothness index between each node pair item corresponding to the shaped point cloud in the learning process is output;

[0076] According to the potential function, the smoothness index between the corresponding irregular point clouds of each node pair is constrained during the learning process until the smoothness index meets the smoothness constraint threshold. The above constraint steps are repeated to traverse the learning data of all irregular point clouds and model them, and finally a smooth irregular field model of automobile parts is generated.

[0077] It should be noted that there are many types of auto parts. Some auto parts have strange and complex shapes due to their functional requirements, so they may have more special-shaped surfaces. The stringent requirements for automobile performance also lead to strict processing requirements for the processing accuracy and quality of special-shaped surfaces on auto parts during production and processing. Some special-shaped surfaces on auto parts have low processing complexity, while some have high processing complexity. However, the processing control of existing manipulators is difficult to adapt to the different complexities of special-shaped surfaces, resulting in increased errors in the processing of special-shaped surfaces. When the existing manipulators control the processing of special-shaped surfaces on the processing mechanism, over-processing or processing omissions often occur, thus deviating from the expected processing shape. Therefore, if the manipulator is to accurately control the processing of special-shaped surfaces of auto parts with different complexities, it is necessary to know the smoothness of the special-shaped surfaces on the auto parts. This method obtains global laser point cloud data of automobile parts through a laser scanner. Since these global laser point cloud data contain the shape data of all special-shaped processing surfaces and non-special-shaped processing surfaces of automobile parts, these global laser point cloud data can be sparsely processed to obtain the normal vector change of the special-shaped surface. If the normal vector deviation between adjacent normal vector values ​​is greater than the special-shaped normal vector critical deviation, it means that the normal vector of the two adjacent normal vector sparse nodes changes dramatically, usually a special-shaped normal vector change, so the point cloud is a special-shaped point cloud; the spatial arrangement of these special-shaped point clouds shows the smoothness of the lines of the special-shaped surface of the automobile part, so the special-shaped point cloud neighborhood undirected graph can be constructed by mapping the special-shaped point cloud to the neighborhood undirected association space, and the special-shaped point cloud neighborhood undirected graph expresses the potential smoothness between the lines of the special-shaped surface; then the Gibbs distribution method is used to estimate the joint probability distribution between each special-shaped point cloud to determine the potential function, and finally the smooth special-shaped field model learned according to the smooth constraint of the potential function is the visual smooth display of the special-shaped surface of the automobile part. Among them, the potential function is an energy function that constrains the smoothness between nodes, which can further make the alien field learning smoother, so that the alien surface displayed by the smooth alien field has as high smoothness as possible, and restores the smoothness of the alien surface lines to the greatest extent.

[0078] It should be noted that this method can use the neighborhood undirected relationship of the special-shaped point cloud of automobile parts in the global point cloud to calculate and learn the smoothness of the constrained special-shaped surface, thereby building a special-shaped surface model that can maximize the smoothness of its lines, providing a reliable and highly credible analysis basis for the complexity of subsequent processing surfaces or lines, and effectively improving the control accuracy of the robot for special-shaped surface processing on automobile parts. Among them, the Chinese name of the Gibbs distribution method is the Gibbs step method, which is used to describe the probability distribution of a data point in an equilibrium state. The Gibbs distribution method can accurately calculate the domain space probability relationship of the special-shaped point cloud, which can greatly improve the rationality and stability of the smooth constraints of the special-shaped surface and ensure the expression accuracy of the smooth special-shaped field.

[0079] More specifically, the step S104 includes the following steps:

[0080] Obtaining processing requirements of automobile parts, obtaining designated processing areas of automobile parts and processing ranges of each designated processing area according to the processing requirements, and dividing the smooth profiled field model into a plurality of sub-model sample point sets based on the designated processing areas;

[0081] Obtain the processing log of the manipulator, and extract the laser signals scanned by the laser scanner for each sub-model sample point set in the smooth profile field model state of the automobile parts through the processing log, which are defined as smooth laser signals;

[0082] The wavelet transform algorithm is introduced, and the wavelet basis function is preset in the wavelet transform algorithm based on the smooth shaped field model. In the wavelet transform algorithm, several smooth laser signals of each sub-model sample point set are decomposed into N sub-band signals;

[0083] A low-pass filter is constructed according to the wavelet basis function, and each sub-band signal is subjected to filtering and sparse decomposition processing by using the low-pass filter to attenuate the high-frequency signal until a preset number of sparse decomposition layers is reached, thereby obtaining a number of low-frequency sparse sub-band signals of each sub-model sample point set;

[0084] Preset an expected partitioning function interval, determine multiple partitioning function criteria according to the processing range, and divide each sub-model sample point set one by one based on each partitioning function criterion to obtain multiple result partitioning functions;

[0085] When the expected partitioning function interval contains one or more result partitioning functions, the result partitioning functions are summarized and added to obtain the partitioning function summary addition amount, and the VC dimension value of each sub-model sample point set is determined according to the partitioning function summary addition amount to generate the special-shaped machining surface detection result of the automobile part;

[0086] If the VC dimension value is less than the preset VC dimension value, the designated processing area corresponding to the calibration sub-model sample point set is a common special-shaped processing area;

[0087] If the VC dimension value is greater than the preset VC dimension value, the designated processing area corresponding to the sub-model sample point set is calibrated as an unusual special-shaped processing area.

[0088] It should be noted that after obtaining the smooth profiled field model of the automobile parts, the processing complexity of the profiled surface can be analyzed through the model; however, since the smooth profiled field model is in the form of nonlinear parameters, if the existing manipulator polygonal space model is used to disassemble the calculation method one by one, the complexity analysis steps of the smooth profiled field model will become extremely cumbersome, the amount of calculation will increase, and the final disassembly calculation result may have a large error, and the calculation efficiency and quality are unsatisfactory. In view of this, in order to ensure the calculation accuracy of the complexity, this method uses the wavelet transform algorithm to sparsely decompose the laser signals scanned by each sub-model sample point set when the automobile parts are in the smooth profiled field model state, so as to extract the low-frequency sparse sub-band signals contained therein, which retain the overall trend and main characteristics of the signal, can ensure that the model only focuses on the low-frequency part to improve the smoothness, and helps to improve the calculation accuracy of the complexity of the processing area. Then, according to the hypothesis class of the processing range decision model, that is, the partition function criterion, each criterion in the partition function criterion can make different decisions according to the training sample points, so that each sub-model sample point set can be accurately and perfectly divided and summarized, and finally the maximum number of different sub-model sample points that can be correctly classified by the calculation model is the partition function summary addition amount. The partition function summary addition amount expresses the complexity of a VC dimension of the smooth special-shaped field model; if the VC dimension value is less than the preset VC dimension value, it means that the processing complexity of the special-shaped surface or line in the processing area is low, and they are some common regular special-shaped surfaces that can be seen everywhere. Therefore, the designated processing area corresponding to the calibration sub-model sample point set is the ordinary special-shaped processing area; otherwise, it means that the processing complexity of the special-shaped surface or line in the processing area is high, and they are some unconventional special-shaped surfaces that are difficult to process and difficult for the robot to accurately identify. Therefore, the designated processing area corresponding to the calibration sub-model sample point set is the extraordinary special-shaped processing area.

[0089] It should be noted that this method can perform accurate computational analysis on the complexity of special-shaped machining surfaces in an all-round manner on the constructed smooth special-shaped field model, and then locate the difficulty of machining the required machining area by the robot. It can effectively improve the accuracy of the subsequent fixed-point navigation technology in planning the robot's machining control route, reduce the robot's machining deviation for different types of special-shaped surfaces, reduce the phenomenon of frequent machining rework, and at the same time simplify cumbersome calculation steps and improve the decision-making rate of the robot's machining control.

[0090] More specifically, the step S106 includes the following steps:

[0091] If the detection result of the special-shaped processing surface of the automobile part shows that the designated processing area is a common special-shaped processing area, then the preset processing shapes of each common special-shaped processing area in the automobile part are obtained according to the processing requirements, which are defined as first preset processing shapes, and the established processing technology of the first preset processing shape is obtained;

[0092] Obtaining processing design drawings of common special-shaped processing areas, constructing an ideal three-dimensional model of the common special-shaped processing areas according to the processing design drawings through SolidWorks design software, and simultaneously establishing an unprocessed three-dimensional model of the common special-shaped processing areas in automobile parts;

[0093] The non-overlapping model area between the ideal three-dimensional model and the unprocessed three-dimensional model in the ordinary special-shaped processing area is stripped off and marked as a model area block to be redundant. At this time, the model boundary points of the ideal three-dimensional model are identified and obtained, which are defined as the first model boundary points, and the model boundary points of the unprocessed three-dimensional model are identified and obtained, which are defined as the second model boundary points;

[0094] Taking the established processing technology as a mapping specification, mapping the boundary points of the second model to the boundary points of the first model based on the mapping specification until the region block of the model to be redundant completely covers the mapped boundary points, thereby generating a region boundary mapping matrix of the region block of the model to be redundant;

[0095] The historical difficulty coefficient of a given processing technology when processing the region boundary surface matrix is ​​obtained, and a Voronoi diagram of an ordinary special-shaped processing area is constructed based on the historical difficulty coefficient and the region boundary mapping matrix. The Voronoi diagram is analyzed by fixed-point navigation technology to plan the processing control route of the robot, and a class of control strategies is obtained.

[0096] It should be noted that when the detection results of the special-shaped processing surface of automobile parts show that the designated processing area is an ordinary special-shaped processing area, it means that the processing complexity of the robot for the designated processing area is not high, and the possibility of errors is relatively small. Therefore, when the processing error rate is small, the selection of processing points can be further optimized to improve the processing quality and stability of the robot processing the special-shaped surface in the designated processing area. This method constructs an ideal three-dimensional model and an unprocessed three-dimensional model of a common special-shaped processing area. Since some processing techniques of automobile parts are the process of processing the unprocessed three-dimensional model into an ideal three-dimensional model, the model area that does not overlap between the two is the part that needs to be processed, that is, the redundant model area block; and some processing techniques are gradual and cannot be completed instantly for the processing of the part that needs to be processed. Therefore, this method uses the established processing technique as the mapping specification to map the second model boundary point of the unprocessed three-dimensional model to the first model boundary point of the ideal three-dimensional model, which reflects the processing process of the common special-shaped processing area from the unprocessed state to the ideal formed state, and reflects the step-by-step special-shaped surface processing condition of the redundant model area block by the established processing technique; and since the established processing technique is the elimination processing of the redundant model area block, it is necessary to ensure that the mapped boundary points are completely covered in the redundant model area, and generate a regional boundary mapping matrix of the redundant model area block, which expresses the special-shaped surface variables that are gradually processed to achieve ideal forming under the established processing technique. Finally, the Voronoi diagram of the ordinary special-shaped processing area is constructed through the regional boundary mapping matrix to provide a point selection basis for the processing control path planning of the fixed-point navigation technology. This method can screen and locate the best processing points for the robot processing control of ordinary special-shaped surfaces, so that the accuracy of the robot processing ordinary special-shaped surfaces under the support of fixed-point navigation technology is more stable and reliable, and the probability of error possibility is reduced or avoided as much as possible, thereby improving the processing quality and efficiency of ordinary special-shaped surfaces.

[0097] More specifically, in the step S106, the historical difficulty coefficient of the given processing technology when processing the regional boundary surface matrix is ​​obtained, a Voronoi diagram of the common special-shaped processing area is constructed based on the historical difficulty coefficient and the regional boundary mapping matrix, and the Voronoi diagram is analyzed by fixed-point navigation technology to plan the processing control route of the manipulator, and a type of control strategy is obtained, such as Figure 2 As shown, the specific steps include:

[0098] S202: acquiring multiple historical processing cases associated with the common special-shaped processing area based on the big data network, and comprehensively identifying the VC dimension value and the established processing technology through the multiple historical processing cases to identify the historical difficulty coefficient of the established processing technology when processing the boundary surface matrix of the area;

[0099] S204: dividing the ordinary special-shaped processing area into a plurality of sub-areas according to the area boundary mapping matrix, generating a processing seed point set for each sub-area based on the historical difficulty coefficient, and forming a series of perpendicular midlines from the combination of every two processing seed points, and determining the Voronoi adjacent edges of the sub-areas and the intersection points of each Voronoi adjacent edge according to the series of perpendicular midlines;

[0100] S206: constructing a Voronoi diagram of the common special-shaped processing area according to the intersection points of the Voronoi adjacent edges of the sub-area and each Voronoi adjacent edge, and obtaining a point distance distribution pattern between each processing seed point in the Voronoi diagram and each boundary point of the redundant model area block;

[0101] S208: Preset a maximum service distance threshold based on the region boundary mapping matrix, and if the distance between the processing seed point in the Voronoi diagram and the boundary point of the region block of the model to be redundant is greater than the maximum service distance threshold, remove or ignore the processing seed point;

[0102] S210: If the distance between the processing seed point and the boundary point of the redundant model area block in the Voronoi diagram is less than the maximum service distance threshold, the processing seed point is retained and marked as the optimal processing seed point, and all the optimal processing seed points are gathered to construct the optimal processing seed point network of the ordinary special-shaped processing area;

[0103] S212: Introduce fixed-point navigation technology, use the fixed-point navigation technology to plan the processing route of the optimal processing seed point network of the ordinary special-shaped processing area, generate a fixed-point navigation processing route for the ordinary special-shaped processing area, control the manipulator to control the ordinary special-shaped processing area based on the fixed-point navigation processing route, and obtain a type of control strategy.

[0104] It should be noted that the existence of the VC dimension value makes the established processing technology have a certain complexity when processing the redundant model area block, that is, different difficulties will be generated when processing different boundary surfaces in the regional boundary surface matrix. In order to enable the manipulator to perform the established processing technology more stably and reliably based on the processing points selected on the redundant model area block, it is necessary to identify the difficulty coefficient of the established processing technology when processing the regional boundary surface matrix; secondly, due to the irregularity of the special-shaped surface on the redundant model area block, there are many boundary points on it, and there are also many processing points that can be screened. Therefore, this method uses the method of constructing a Voronoi diagram to screen the processing points of the redundant model area block. The Chinese name of the Voronoi diagram is the Thiessen polygon diagram, which can associate each area with a given processing seed point. For ordinary special-shaped processing areas, it can better fit the boundary point distribution of the irregular special-shaped surface on the redundant model area block, and provide a point screening visualization tool with small error and high adaptability for the screening of processing points in ordinary special-shaped processing areas, effectively improving the accurate positioning performance of processing points. If the distance between the processing seed point in the Voronoi diagram and the boundary point of the model area block to be redundant is greater than the maximum service distance threshold, it means that the processing seed point cannot meet the processing service requirements of the area boundary mapping matrix, and the processing seed point is not considered, so it is removed or ignored; if the distance between the processing seed point in the Voronoi diagram and the boundary point of the model area block to be redundant is less than the maximum service distance threshold, it means that the processing seed point is closer to the processing service requirements of the area boundary mapping matrix, and it is a better processing point, so the processing seed point is retained and marked. Through this method, the irregular boundary point distribution of the required processing part in the ordinary special-shaped processing area can be matched in the form of constructing a Voronoi diagram, so that the optimal processing point selection when the established processing technology processes this part is more accurately matched, and the intelligent processing point screening effect of ordinary special-shaped surface processing is realized, replacing the steps of manual intervention adjustment, avoiding the robot control error caused by manual adjustment errors, and having high credibility; improving the processing point decision-making ability of fixed-point navigation technology for ordinary special-shaped surfaces and the processing control accuracy and reliability of the robot.

[0105] More specifically, the step S108 includes the following steps:

[0106] If the detection result of the special-shaped processing surface of the automobile part shows that the designated processing area is an unusual special-shaped processing area, a preset processing shape of the unusual special-shaped processing area in the automobile part is obtained according to the processing requirements, which is defined as a second preset processing shape;

[0107] Construct a spline model of the unusual special-shaped processing area, set the Gaussian kernel function and the kernel function bandwidth range based on the VC dimension value of the unusual special-shaped processing area, and calculate the mean shift vector of each spline in the spline model within the kernel function bandwidth range;

[0108] The mean shift clustering algorithm is introduced to move each spline in the spline model along the corresponding mean shift vector to obtain the current moving distance value. If the current moving distance value is less than the preset moving distance value, the movement is stopped and a cluster cluster is generated at the current moving position to obtain the VC dimension value-spline clustering result;

[0109] The spline model of the unusual special-shaped processing area is obtained through the VC dimension value-spline clustering result. There is a spline area with a VC dimension value greater than the preset VC dimension value, which is marked as an easy-to-touch spline area. The spline of the easy-to-touch spline area is converted into a tensor spline to obtain the tension control point array of the tensor spline.

[0110] Obtaining a necessary processing mechanism of an automobile part, extracting optimal historical working condition data of the necessary processing mechanism for processing the tension control point array through a processing log, constructing an optimal working condition space domain of the necessary processing mechanism based on the optimal historical working condition data, and establishing a redundant space domain of an easy-to-touch spline area based on a second preset processing shape;

[0111] Determine whether the optimal working condition space field completely covers the redundant space field. If not, obtain the uncovered space field volume between the optimal working condition space field and the redundant space field, and adjust the processing vector of the unusual special-shaped processing area according to the uncovered space field volume. After the adjustment is completed, locate the first batch of special-shaped processing points of the processing vector;

[0112] If it is completely covered or there is interference from the external space after it is completely covered, the external interference volume is calculated, and the processing vector of the unusual special-shaped processing area is adjusted based on the external interference volume to locate the second batch of special-shaped processing points of the processing vector;

[0113] Through the fixed-point navigation technology, the first batch of special-shaped processing points and the second batch of special-shaped processing points are comprehensively analyzed and planned, and the fixed-point navigation processing route of the unusual special-shaped processing area is generated. Based on the fixed-point navigation processing route, the robot is controlled to control the unusual special-shaped processing area to obtain the second type of control strategy.

[0114] It should be noted that when the detection results of the special-shaped processing surface of automobile parts show that the designated processing area is an unusual special-shaped processing area, it means that the processing complexity of the designated processing area is high. If there is a deviation in the processing path control accuracy of the robot, it may cause the processing mechanism to over-process, miss or omit different special-shaped surfaces in the unusual special-shaped processing area, so that the final processed special-shaped surface does not meet the expected processing requirements. It can be seen that the higher the complexity of the special-shaped surface, the higher the robot's processing path control accuracy requirements. And accurate processing path control is also based on reasonable and accurate processing points. Reasonable and accurate processing points can ensure that the processing path planned by the fixed-point navigation technology is more stable and accurate. To this end, the method first performs complexity mean shift clustering on each spline in the spline model of the unusual special-shaped processing area according to the VC dimension value, so as to determine the splines and easy-to-touch spline areas that cause the unusual special-shaped processing area to produce corresponding VC dimension values, and then ensure that the processing point selection for the unusual special-shaped processing area is based on the easy-to-touch spline area where the high-complexity spline is located, so as to minimize the error phenomenon when processing these special-shaped surface splines; the processing of the easy-to-touch spline area by the necessary processing mechanism of automobile parts is a gradual contact elimination process, so the optimal working condition space field of the necessary processing mechanism will produce a gradual covering process state for the redundant space field of the easy-to-touch spline area; therefore, by judging whether the optimal working condition space field completely covers the redundant space field, it can be further determined whether the necessary processing mechanism has over-processing or processing omissions in the easy-to-touch spline area.

[0115] It should be noted that if it is not completely covered, it means that the necessary processing mechanism has omitted processing when processing the easy-to-touch spline area. At this time, the processing vector of the unusual special-shaped processing area of ​​the automobile part can be adjusted to make the necessary processing mechanism compensate for the area of ​​the uncovered volume; if it is completely covered or there is interference from the external space field after complete coverage, it means that the necessary processing mechanism has over-processed the easy-to-touch spline area. At this time, it is necessary to adjust the processing vector according to the over-processed volume to avoid the necessary processing mechanism from contacting the over-processed area. After the adjustment is completed, the processing point of the necessary processing mechanism for the unusual special-shaped processing area can be accurately located. This method can accurately and reliably select processing points for the processing route control of the fixed-point navigation technology decision manipulator when processing the unusual special-shaped processing area based on the analysis method of spatial field coverage interference, improve the processing accuracy of the manipulator on the processing mechanism to control the unusual special-shaped surface of the automobile part, reduce the probability of omission, over-processing or missing, reduce the processing rework rate of automobile parts, and optimize the control quality and efficiency of the manipulator under the support of fixed-point navigation technology.

[0116] In addition, the automobile parts processing robot control method based on fixed-point navigation technology also includes the following steps:

[0117] Obtaining a preset material loading and transportation trajectory of the automobile parts to be processed, and determining a fixed-point processing range of the manipulator according to the preset material loading and transportation trajectory;

[0118] Obtaining the layout sites of all manipulators within the fixed-point processing range of the manipulator, and at the same time obtaining a spatial schematic diagram of the fixed-point processing range of the manipulator, marking the layout site of each manipulator in the plane schematic diagram, and obtaining a schematic diagram of the manipulator space pattern;

[0119] Obtain the preset production and processing strategy of the current batch of automobile parts, locate the automobile parts to be processed on the current preset loading and transportation track in the robot space pattern diagram based on the preset production and processing strategy, mark them as automobile parts processing objects, and obtain the real-time spatial coordinate parameters of the automobile parts processing objects;

[0120] Acquire the spatial center coordinate parameters of the manipulator spatial pattern diagram, calculate the Manhattan distance between the real-time spatial coordinate parameters and the spatial center coordinate parameters, determine the coordinate eccentricity of the automotive parts processing object compared to the spatial center of the manipulator spatial pattern diagram according to the Manhattan distance, and set the processing scheduling planning threshold based on the coordinate eccentricity;

[0121] Taking the current automobile parts processing object as the search benchmark, the maximum scheduling range of the manipulator is planned in the manipulator space pattern diagram until the processing scheduling planning threshold is reached to obtain the maximum scheduling range;

[0122] Obtain the processing task amount of each manipulator within the current maximum scheduling range, calculate the distance between each manipulator and the current automobile part processing object, obtain multiple spatial coordinate distances, bundle the processing task amount of each manipulator and the corresponding spatial coordinate distance into a gain data packet, and calculate the information gain value of each gain data packet;

[0123] A scheduling algorithm is introduced. Based on the information gain value, the scheduling index of the robot processing the current automotive parts processing object corresponding to each gain data packet is calculated in the scheduling priority algorithm. All robots within the maximum scheduling range are prioritized for processing scheduling control according to the descending order of the scheduling index.

[0124] It should be noted that for the processing of large quantities of automobile parts, multiple manipulators are usually equipped to perform processing tasks. However, if multiple manipulators do not have an efficient scheduling control strategy for scheduling processing, some manipulators will be busy with processing tasks, while some manipulators will be idle, making it difficult to achieve efficient processing and production of automobile parts, reducing the processing rate of automobile parts. Therefore, this method can reasonably decide the scheduling priority allocation based on the spatial coordinate distance between the manipulators within the range and the automobile parts to be processed and the processing task volume of each manipulator. On the one hand, the automobile parts to be processed can achieve fast processing effects and greatly improve processing efficiency; on the other hand, all manipulators can be in a balanced task processing state, avoiding the phenomenon of busy processing of manipulators and unreasonable space allocation, and ensuring the rationality and uniformity of manipulator scheduling.

[0125] In addition, the automobile parts processing robot control method based on fixed-point navigation technology also includes the following steps:

[0126] The transport protection mechanism of the manipulator is activated to transport the automobile parts in a protected manner, and after the protected transportation, the image information of the actual holding posture of the manipulator at the preset transportation node is obtained;

[0127] Obtain the image information of incorrect holding postures of several automobile parts that are subjected to different vibration states during the transportation process when the robot opens the transportation protection mechanism to protect the transportation through the processing log of the robot, and construct an image comparison data warehouse based on the image information of several holding postures;

[0128] Calculating the hash misalignment function between the actual holding posture image information and each of the erroneous holding posture image information one by one in the image comparison data warehouse, determining the degree of fit between the actual holding posture image information and each of the erroneous holding posture image information according to the hash misalignment function, and obtaining a degree of fit data set;

[0129] An abnormal fit interval is set. If at least one fit in the fit data set is displayed within the abnormal fit interval, the manipulator is urgently controlled to stop the protection transport task and obtain the actual vibration sensor data packet of the transport protection mechanism at the preset transport node;

[0130] Extracting the actual vibration waveform signal of the transport protection mechanism based on the actual vibration sensing data packet, obtaining the standard vibration waveform signal of the transport protection mechanism for executing the preset protection strategy, and calculating the curvature overlap between the actual vibration waveform signal and the standard vibration waveform signal;

[0131] If the curvature overlap is less than the preset curvature overlap, the proliferation or missing signal between the actual vibration waveform signal and the standard vibration waveform signal is calculated, and the transportation protection mechanism is reconfigured based on the proliferation or missing signal to eliminate or reduce the transportation vibration phenomenon of the manipulator.

[0132] It should be noted that the existing manipulator has multiple movable joints, and usually a special transportation protection mechanism is set on each movable joint to protect the processing and transportation behavior of the manipulator and reduce vibration interference; however, if the parameters of the transportation protection mechanism are not configured correctly, the transportation protection mechanism will not be able to accurately protect the transportation of the manipulator, which will cause the manipulator to have a deviation in the holding posture when transporting automobile parts for processing, thereby causing a deviation in the processing synchronization of automobile parts. In this regard, the actual holding posture of the manipulator is compared with the wrong holding posture under different vibration states through this method to determine whether the manipulator is affected by vibration. If there is a vibration effect, the curvature overlap between the actual vibration waveform signal and the standard signal is further analyzed to reconfigure the transportation protection mechanism, thereby eliminating or reducing the transportation vibration phenomenon of the manipulator, improving the stability and safety factor of the manipulator's processing and transportation, and avoiding the risk of automobile parts falling and being damaged due to the vibration of the manipulator.

[0133] The second aspect of the present invention provides an automobile parts processing robot based on fixed-point navigation technology, which is applied to any one of the above-mentioned automobile parts processing robot control methods based on fixed-point navigation technology, such as Figure 3 As shown, specifically including:

[0134] A robot body 1011, which is used to control the transportation and movement of the robot with any degree of freedom and position;

[0135] A manipulator clamping mechanism 1012, which includes a clamping sub-joint 1013, responsible for clamping the automobile parts to a designated processing area;

[0136] A point cloud data scanning module, the point cloud data scanning module includes a laser scanner 1014, which is used to scan the automobile parts in the current specified processing task area;

[0137] A low-pass filtering module, the low-pass filtering module includes a low-pass filter 1015, which is responsible for filtering and sparsely decomposing each sub-band signal to attenuate the high-frequency signal to obtain a low-frequency sparse sub-band signal of each sub-model sample point set;

[0138] A fixed-point navigation module is used to locate special-shaped processing points and plan fixed-point navigation processing routes for ordinary special-shaped processing areas and extraordinary special-shaped processing areas based on the special-shaped processing points.

[0139] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A control method for an automobile parts processing robot based on fixed-point navigation technology, characterized in that: The following steps are involved: S102: Acquire global laser point cloud data of automobile parts, obtain special-shaped point clouds by sparsely calculating normal vector deviations of the global laser point cloud data, construct special-shaped fields based on smoothness learning constraints of the special-shaped point clouds, and generate smooth special-shaped field models of automobile parts; S104: obtaining the designated processing area and processing range of the automobile part, extracting a number of low-frequency sparse subband signals of each designated processing area in the smooth special-shaped field model, performing function segmentation and VC dimension induction analysis on the low-frequency sparse subband signals using the processing range, and generating the detection result of the special-shaped processing surface of the automobile part; S106: If the detection result of the special-shaped processing surface of the automobile part is a common special-shaped processing area, then the unprocessed three-dimensional model of the common special-shaped processing area is compared with the redundant model area block of the ideal three-dimensional model, and the redundant model area block is mapped according to the established processing technology of the first preset processing shape to obtain a region boundary mapping matrix, and a Voronoi diagram is constructed based on the region boundary mapping matrix and a fixed-point navigation technology is introduced to plan a control route to obtain a type of control strategy; S108: If the detection result of the special-shaped processing surface of the automobile part is an unusual special-shaped processing area, the easy-to-touch spline area of ​​the unusual special-shaped processing area is clustered by mean shift, and the optimal working condition space field of the necessary processing mechanism for the easy-to-touch spline area is established. The coverage of the optimal working condition space field and the redundant space field of the easy-to-touch spline area is analyzed by the fixed-point navigation technology to plan the control route, and obtain the second type of control strategy; The step S104 specifically includes the following steps: Obtaining processing requirements of automobile parts, obtaining designated processing areas of automobile parts and processing ranges of each designated processing area according to the processing requirements, and dividing the smooth profiled field model into a plurality of sub-model sample point sets based on the designated processing areas; Obtain the processing log of the manipulator, and extract the laser signals scanned by the laser scanner for each sub-model sample point set in the smooth profile field model state of the automobile parts through the processing log, which are defined as smooth laser signals; The wavelet transform algorithm is introduced, and the wavelet basis function is preset in the wavelet transform algorithm based on the smooth shaped field model. In the wavelet transform algorithm, several smooth laser signals of each sub-model sample point set are decomposed into N sub-band signals; A low-pass filter is constructed according to the wavelet basis function, and each sub-band signal is subjected to filtering and sparse decomposition processing by using the low-pass filter to attenuate the high-frequency signal until a preset number of sparse decomposition layers is reached, thereby obtaining a number of low-frequency sparse sub-band signals of each sub-model sample point set; Preset an expected partitioning function interval, determine multiple partitioning function criteria according to the processing range, and divide each sub-model sample point set one by one based on each partitioning function criterion to obtain multiple result partitioning functions; When the expected partitioning function interval contains one or more result partitioning functions, the result partitioning functions are summarized and added to obtain the partitioning function summary addition amount, and the VC dimension value of each sub-model sample point set is determined according to the partitioning function summary addition amount to generate the special-shaped machining surface detection result of the automobile part; If the VC dimension value is less than the preset VC dimension value, the designated processing area corresponding to the calibration sub-model sample point set is a common special-shaped processing area; If the VC dimension value is greater than the preset VC dimension value, the designated processing area corresponding to the sub-model sample point set is calibrated as an unusual special-shaped processing area; The step S108 specifically includes the following steps: If the detection result of the special-shaped processing surface of the automobile part shows that the designated processing area is an unusual special-shaped processing area, a preset processing shape of the unusual special-shaped processing area in the automobile part is obtained according to the processing requirements, which is defined as a second preset processing shape; Construct a spline model of the unusual special-shaped processing area, set the Gaussian kernel function and the kernel function bandwidth range based on the VC dimension value of the unusual special-shaped processing area, and calculate the mean shift vector of each spline in the spline model within the kernel function bandwidth range; The mean shift clustering algorithm is introduced to move each spline in the spline model along the corresponding mean shift vector to obtain the current moving distance value. If the current moving distance value is less than the preset moving distance value, the movement is stopped and a cluster cluster is generated at the current moving position to obtain the VC dimension value-spline clustering result; The spline model of the unusual special-shaped processing area is obtained through the VC dimension value-spline clustering result. There is a spline area with a VC dimension value greater than the preset VC dimension value, which is marked as an easy-to-touch spline area. The spline of the easy-to-touch spline area is converted into a tensor spline to obtain the tension control point array of the tensor spline. Obtaining a necessary processing mechanism of an automobile part, extracting optimal historical working condition data of the necessary processing mechanism for processing the tension control point array through a processing log, constructing an optimal working condition space domain of the necessary processing mechanism based on the optimal historical working condition data, and establishing a redundant space domain of an easy-to-touch spline area based on a second preset processing shape; Determine whether the optimal working condition space field completely covers the redundant space field. If not, obtain the uncovered space field volume between the optimal working condition space field and the redundant space field, and adjust the processing vector of the unusual special-shaped processing area according to the uncovered space field volume. After the adjustment is completed, locate the first batch of special-shaped processing points of the processing vector; If it is completely covered or there is interference from the external space after it is completely covered, the interference volume of the external space is calculated, and the processing vector of the unusual special-shaped processing area is adjusted based on the interference volume of the external space, and the second batch of special-shaped processing points of the processing vector are located; Through the fixed-point navigation technology, the first batch of special-shaped processing points and the second batch of special-shaped processing points are comprehensively analyzed and planned, and the fixed-point navigation processing route of the unusual special-shaped processing area is generated. Based on the fixed-point navigation processing route, the robot is controlled to control the unusual special-shaped processing area to obtain the second type of control strategy.

2. The method for controlling an automobile parts processing robot based on fixed-point navigation technology according to claim 1 is characterized in that: The step S102 specifically includes the following steps: Obtain the specified processing task area of ​​the manipulator, and scan the automobile parts in the current specified processing task area through a laser scanner to obtain global laser point cloud data of the automobile parts; Construct a sparse graph, fit the global laser point cloud data with topology in the sparse graph, generate a sparse topological composition of the automobile parts, and capture and extract the sparse nodes of the normal vectors of the automobile parts based on the sparse topological composition; Obtain the adjacent normal vector values ​​of each normal vector sparse node and preset the critical deviation of the irregular normal vector. If the normal vector deviation between the adjacent normal vector values ​​is greater than the critical deviation of the irregular normal vector, calibrate the point cloud corresponding to the adjacent normal vector value as an irregular point cloud. Obtain all the special-shaped point clouds contained in the global laser point cloud data of the automobile parts, construct a neighborhood undirected association space based on the global laser point cloud, define all the special-shaped point clouds as neighborhood graph nodes and map them into the neighborhood undirected association space to generate a neighborhood undirected graph of the special-shaped point clouds; The Gibbs distribution method is introduced to estimate the conditional probability between each irregular point cloud in the Gibbs distribution method, and the joint probability distribution of the irregular point cloud is obtained. The potential function of each irregular point cloud is determined according to the joint probability distribution. Based on the potential function, node pair items between each shaped point cloud are obtained in the undirected graph of the neighborhood of the shaped point cloud, a smoothness constraint threshold is preset based on the node pair items, and the data of each shaped point cloud is learned by maximum a posteriori estimation, and the smoothness index between each node pair item corresponding to the shaped point cloud in the learning process is output; According to the potential function, the smoothness index between the corresponding irregular point clouds of each node pair is constrained during the learning process until the smoothness index meets the smoothness constraint threshold. The above constraint steps are repeated to traverse the learning data of all irregular point clouds and model them, and finally a smooth irregular field model of automobile parts is generated.

3. The method for controlling an automobile parts processing robot based on fixed-point navigation technology according to claim 1 is characterized in that: The step S106 specifically includes the following steps: If the detection result of the special-shaped processing surface of the automobile part shows that the designated processing area is a common special-shaped processing area, then the preset processing shapes of each common special-shaped processing area in the automobile part are obtained according to the processing requirements, which are defined as first preset processing shapes, and the established processing technology of the first preset processing shape is obtained; Obtaining processing design drawings of common special-shaped processing areas, constructing an ideal three-dimensional model of the common special-shaped processing areas according to the processing design drawings through SolidWorks design software, and simultaneously establishing an unprocessed three-dimensional model of the common special-shaped processing areas in automobile parts; The non-overlapping model area between the ideal three-dimensional model and the unprocessed three-dimensional model in the ordinary special-shaped processing area is stripped off and marked as a model area block to be redundant. At this time, the model boundary points of the ideal three-dimensional model are identified and obtained, which are defined as the first model boundary points, and the model boundary points of the unprocessed three-dimensional model are identified and obtained, which are defined as the second model boundary points; Taking the established processing technology as a mapping specification, mapping the boundary points of the second model to the boundary points of the first model based on the mapping specification until the region block of the model to be redundant completely covers the mapped boundary points, thereby generating a region boundary mapping matrix of the region block of the model to be redundant; The historical difficulty coefficient of a given processing technology when processing the region boundary surface matrix is ​​obtained, and a Voronoi diagram of an ordinary special-shaped processing area is constructed based on the historical difficulty coefficient and the region boundary mapping matrix. The Voronoi diagram is analyzed by fixed-point navigation technology to plan the processing control route of the robot, and a class of control strategies is obtained.

4. The method for controlling an automobile parts processing robot based on fixed-point navigation technology according to claim 3 is characterized in that: In the step S106, the historical difficulty coefficient of the established processing technology when processing the regional boundary surface matrix is ​​obtained, the Voronoi diagram of the common special-shaped processing area is constructed based on the historical difficulty coefficient and the regional boundary mapping matrix, and the Voronoi diagram is analyzed by fixed-point navigation technology to plan the processing control route of the manipulator to obtain a type of control strategy, which specifically includes the following steps: Based on the big data network, multiple historical processing cases associated with the common special-shaped processing area are obtained, and the VC dimension value and the established processing technology are comprehensively identified through the multiple historical processing cases to identify the historical difficulty coefficient of the established processing technology when processing the boundary surface matrix of the area; The ordinary special-shaped processing area is divided into several sub-areas according to the area boundary mapping matrix, and a processing seed point set of each sub-area is generated based on the historical difficulty coefficient, and a series of perpendicular medians are formed by the combination of every two processing seed points, and the Voronoi adjacent edges of the sub-areas and the intersection points of each Voronoi adjacent edge are determined according to the series of perpendicular medians; Construct a Voronoi diagram of the common special-shaped processing area according to the intersection of the Voronoi adjacent edges of the sub-area and each Voronoi adjacent edge, and obtain the point distance distribution pattern between each processing seed point in the Voronoi diagram and each boundary point of the redundant model area block; Based on the preset maximum service distance threshold of the region boundary mapping matrix, if the distance between the processing seed point in the Voronoi diagram and the boundary point of the region block of the model to be redundant is greater than the maximum service distance threshold, the processing seed point is removed or ignored; If the distance between the processing seed point and the boundary point of the redundant model area block in the Voronoi diagram is less than the maximum service distance threshold, the processing seed point is retained and marked as the optimal processing seed point, and all the optimal processing seed points are gathered to construct the optimal processing seed point network of the common special-shaped processing area; Fixed-point navigation technology is introduced. The processing route is planned for the optimal processing seed point network of the ordinary special-shaped processing area through the fixed-point navigation technology, and the fixed-point navigation processing route of the ordinary special-shaped processing area is generated. Based on the fixed-point navigation processing route, the manipulator is controlled to control the processing of the ordinary special-shaped processing area, and a class of control strategies is obtained.

5. An automobile parts processing robot based on fixed-point navigation technology, applied to an automobile parts processing robot control method based on fixed-point navigation technology as claimed in any one of claims 1 to 4, characterized in that: Specifically include: A robot body, which is used to control the transport and movement of the robot with any degree of freedom and position; A manipulator clamping mechanism, wherein the manipulator clamping mechanism includes a clamping sub-joint, which is responsible for clamping the automobile parts to a designated processing area; A point cloud data scanning module, wherein the point cloud data scanning module includes a laser scanner for scanning automobile parts within a currently specified processing task area; A low-pass filtering module, wherein the low-pass filtering module includes a low-pass filter, which is responsible for filtering and sparsely decomposing each sub-band signal to attenuate the high-frequency signal to obtain a low-frequency sparse sub-band signal of each sub-model sample point set; A fixed-point navigation module is used to locate special-shaped processing points and plan fixed-point navigation processing routes for ordinary special-shaped processing areas and extraordinary special-shaped processing areas based on the special-shaped processing points.

Images