An intelligent control method and system for a sorting manipulator based on big data

Abstract

The present invention relates to the technical field of intelligent sorting, and discloses an intelligent control method and system for a sorting manipulator based on big data. The method includes the following steps: calculating the geometric data of a target item; constructing a data set based on the two-dimensional coordinates of the target item and the geometric data of the target item, and training the data set using a convolutional neural network to obtain the grasping force and the grasping coordinates; obtaining motion parameters based on the grasping force and the grasping coordinates; obtaining the three-dimensional coordinates of the target item according to the target item image; obtaining a motion deviation value according to the three-dimensional coordinates of the target item and the preset three-dimensional motion coordinates; constructing a data set based on the motion parameters and the motion deviation value, and performing optimization using a particle swarm optimization algorithm to obtain the optimal motion parameters; obtaining the optimal motion mode of the manipulator according to the matching of the optimal motion parameters and the preset data. This method can dynamically adjust the motion mode according to the real-time feedback during the grasping process of the manipulator.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent sorting, and in particular, to an intelligent control method and system for a sorting manipulator based on big data. Background Art

[0002] In an intelligent sorting system, the application of manipulators is becoming increasingly widespread, and they need to adjust the motion mode in real time according to the characteristics of the items to be sorted, such as shape, size, and material.

[0003] The existing manipulator sorting methods mainly rely on traditional motion planning. By obtaining the geometric shape and size characteristics of the target item and combining image processing algorithms, the motion mode of the manipulator is generated using a preset grasping strategy and motion planning algorithm, and the grasping strategy is adjusted through visual feedback in the image.

[0004] However, this manipulator sorting method can only adjust the grasping strategy through visual feedback in the image, and cannot dynamically adjust the motion mode according to the real-time feedback during the grasping process, thereby affecting the overall efficiency and reliability of the sorting system. Summary of the Invention

[0005] In view of the technical problems mentioned in the above background art, an intelligent control method and system for a sorting manipulator based on big data are provided. The present invention provides an intelligent control method and system for a sorting manipulator based on big data to solve the problem that the existing manipulator sorting methods can only adjust the grasping strategy through visual feedback in the image, and cannot dynamically adjust the motion mode according to the real-time feedback during the grasping process, thereby affecting the overall efficiency and reliability of the sorting system. The technical means adopted by the present invention are as follows:

[0006] An intelligent control method for a sorting manipulator based on big data, comprising the following steps:

[0007] Obtain the image of the target item, the two-dimensional coordinates of the target item, the shape of the target item, and the size of the target item;

[0008] Perform geometric calculations based on the shape and size of the target item to obtain the geometric data of the target item;

[0009] Construct a geometric data set based on the two-dimensional coordinates and geometric data of the target item, and train the geometric data set using a convolutional neural network to obtain the grasping force of the manipulator and the grasping coordinates of the manipulator;

[0010] Perform closed-loop adjustment based on the grasping force of the manipulator and the grasping coordinates to obtain the motion parameters of the manipulator;

[0011] Preprocess the image of the target object and calculate the three-dimensional coordinates of the preprocessed image to obtain the three-dimensional coordinates of the target object;

[0012] Calculate the deviation between the three-dimensional coordinates of the target object and the preset three-dimensional motion coordinates to obtain the motion deviation value;

[0013] Construct a motion data set based on the manipulator motion parameters and the motion deviation value, and use the particle swarm optimization algorithm to optimize the motion data set to obtain the optimal motion parameters of the manipulator;

[0014] Match the optimal motion parameters of the manipulator with the preset manipulator motion data to obtain the optimal motion mode of the manipulator, and control the motion of the manipulator according to the optimal motion mode of the manipulator.

[0015] Further, the shapes of the target objects include: circular objects, rectangular objects, rhombus objects, triangular objects, and trapezoidal objects;

[0016] The dimensions of the target objects include: object length, object width, object radius, and object angle;

[0017] The geometric data of the target objects include: object area, object perimeter, object curvature, and object centroid.

[0018] Further, the process of constructing a geometric data set based on the two-dimensional coordinates of the target object and the geometric data of the target object and training it using a convolutional neural network to obtain the grasping force of the manipulator and the grasping coordinates of the manipulator includes the following steps:

[0019] Perform data preprocessing on the geometric data set to obtain standard geometric data;

[0020] Construct a convolutional layer based on the standard geometric data to obtain a convolutional layer data set;

[0021] Based on the convolutional layer data set, and use a convolutional neural network for training to obtain the grasping force of the manipulator and the grasping coordinates of the manipulator;

[0022] The preprocessing includes normalization processing and standardization processing;

[0023] Among them, the normalization processing formula is:

[0024] ;

[0025] Among them, represents the normalized geometric data, represents the geometric data, represents the minimum value in the geometric data set, Represents the maximum value in the geometric dataset;

[0026] The normalization formula is:

[0027] ;

[0028] Where, Represents the normalized geometric data, Represents the mean value in the geometric dataset, Represents the standard deviation in the geometric dataset.

[0029] Furthermore, based on the convolutional layer dataset and using a convolutional neural network for training, the grasping force of the manipulator and the grasping coordinates of the manipulator are obtained, including the following steps:

[0030] Divide the convolutional layer dataset into convolutional layer training set data and convolutional layer test set data;

[0031] Perform average pooling operation on the convolutional layer training set data to obtain pooling layer data;

[0032] Iterate based on the pooling layer data, a preset loss function, and a preset activation function to obtain the grasping force of the initial manipulator and the grasping coordinates of the initial manipulator;

[0033] Calculate the root mean square error between the pooling layer data and the convolutional layer test set data to obtain the convolutional root mean square error;

[0034] When the convolutional root mean square error is less than the preset error threshold, perform backpropagation operation on the grasping force of the initial manipulator and the grasping coordinates of the initial manipulator to obtain the grasping force of the manipulator and the grasping coordinates of the manipulator.

[0035] Furthermore, based on the grasping force of the manipulator and the grasping coordinates of the manipulator, perform closed-loop adjustment to obtain the manipulator motion parameters, including the following steps:

[0036] Take the grasping force of the manipulator as the feedforward parameter and perform Laplace transform on the feedforward parameter to obtain the Laplace feedforward parameter;

[0037] Take the grasping coordinates of the manipulator as the negative feedback parameter and perform Laplace transform on the negative feedback parameter to obtain the Laplace negative feedback parameter;

[0038] Perform closed-loop calculation based on the Laplace feedforward parameter and the Laplace negative feedback parameter to obtain the manipulator motion parameters;

[0039] Where, the manipulator motion parameters are calculated through the following formula:

[0040] ;

[0041] Among them, represents the motion parameters of the manipulator, represents the proportionality coefficient, represents the Laplace feedforward parameter, represents the Laplace negative feedback parameter.

[0042] Furthermore, preprocess the image of the target item and calculate the three-dimensional coordinates to obtain the three-dimensional coordinates of the target item, including the following steps:

[0043] Remove noise from the image of the target item to obtain a noise-free item image;

[0044] Remove distortion from the noise-free item image to obtain an undistorted item image;

[0045] Calculate the three-dimensional coordinates based on the undistorted item image to obtain the three-dimensional coordinates of the target item;

[0046] Among them, the three-dimensional coordinates of the target item are calculated by the following formula:

[0047] ;

[0048] ;

[0049] ;

[0050] Among them, represents the abscissa of the target item, represents the ordinate of the target item, represents the vertical coordinate of the target item, represents the image length, represents the image width, represents the focal length of the undistorted item image on the axis, represents the focal length of the undistorted item image on the axis.

[0051] Furthermore, calculate the motion deviation value based on the three-dimensional coordinates of the target item and the preset three-dimensional motion coordinates, including the following steps:

[0052] Calculate the translational offset based on the three-dimensional coordinates of the target item and the preset three-dimensional motion coordinates to obtain the translational deviation value;

[0053] Calculate the rotational deviation value by performing a rotation calculation on the translational deviation value and the preset rotation matrix;

[0054] Calculate the motion deviation based on the translation deviation value and the rotation deviation value to obtain the motion deviation value.

[0055] Among them, the translation deviation value is obtained through the following calculation formula:

[0056] ;

[0057] ;

[0058] ;

[0059] Among them, represents the translation deviation value, represents the three-dimensional coordinates of the target object, represents the preset three-dimensional motion coordinates;

[0060] The rotation deviation value is obtained through the following calculation formula:

[0061] ;

[0062] Among them, represents the rotation deviation value, represents the preset rotation matrix, represents the transpose of the translation deviation value;

[0063] The motion deviation value is obtained through the following calculation formula:

[0064] ;

[0065] Among them, represents the motion deviation value, represents the transpose of the rotation deviation value.

[0066] Furthermore, construct a motion data set according to the manipulator motion parameters and the motion deviation value, and use the particle swarm optimization algorithm for optimization to obtain the optimal motion parameters of the manipulator, including the following steps:

[0067] According to the motion data set, perform data cleaning and preprocessing to obtain motion training set data;

[0068] Establish a speed set and a position set according to the motion training set data, and initialize the data of the speed set and the position set to obtain initial speed data and initial position data;

[0069] According to the initial speed data and the initial position data, use the particle swarm optimization algorithm for parameter optimization to obtain the optimal motion parameters of the manipulator.

[0070] Further, match the optimal motion parameters of the manipulator with the preset manipulator motion data to obtain the optimal motion mode of the manipulator, and control the motion of the manipulator according to the optimal motion mode of the manipulator, including:

[0071] When the optimal motion parameter is greater than or equal to the preset first motion parameter threshold, the optimal motion mode of the manipulator is the preset first motion mode;

[0072] When the optimal motion parameter is greater than or equal to the preset second motion parameter threshold and less than the preset first motion parameter threshold, the optimal motion mode of the manipulator is the preset second motion mode;

[0073] When the optimal motion parameter is greater than or equal to the preset third motion parameter threshold and less than the preset second motion parameter threshold, the optimal motion mode of the manipulator is the preset third motion mode;

[0074] When the optimal motion parameter is greater than or equal to the preset fourth motion parameter threshold and less than the preset third motion parameter threshold, the optimal motion mode of the manipulator is the preset fourth motion mode;

[0075] When the optimal motion parameter is less than the preset fourth motion parameter threshold, the optimal motion mode of the manipulator is the preset fifth motion mode.

[0076] The present invention further includes an intelligent control system for a sorting manipulator based on big data, including:

[0077] A data acquisition module, a geometric data calculation module, a convolutional network training module, a closed-loop motion adjustment module, a three-dimensional coordinate calculation module, a motion deviation calculation module, a motion parameter optimization module, and an optimal mode matching module;

[0078] The data acquisition module is used to acquire the image of the target item, the two-dimensional coordinates of the target item, the shape of the target item, and the size of the target item;

[0079] The geometric data calculation module is used to perform geometric calculations based on the shape of the target item and the size of the target item to obtain target item geometric data;

[0080] The convolutional network training module is used to construct a geometric data set based on the two-dimensional coordinates of the target item and the geometric data of the target item, and train it using a convolutional neural network to obtain the grasping force of the manipulator and the grasping coordinates of the manipulator;

[0081] The closed-loop motion adjustment module is used to perform closed-loop adjustment based on the grasping force of the manipulator and the grasping coordinates of the manipulator to obtain the motion parameters of the manipulator;

[0082] The three-dimensional coordinate calculation module is used to preprocess the image of the target item and calculate the three-dimensional coordinates to obtain the three-dimensional coordinates of the target item.

[0083] The motion deviation calculation module is used to calculate the deviation between the three-dimensional coordinates of the target item and the preset three-dimensional motion coordinates to obtain the motion deviation value.

[0084] The motion parameter optimization module is used to construct a motion data set based on the manipulator motion parameters and the motion deviation value, and use the particle swarm optimization algorithm for optimization to obtain the optimal motion parameters of the manipulator.

[0085] The optimal mode matching module is used to match the optimal motion parameters of the manipulator with the preset manipulator motion data to obtain the optimal motion mode of the manipulator, so as to control the motion of the manipulator according to the optimal motion mode of the manipulator.

[0086] Compared with the prior art, the present invention has the following advantages:

[0087] The present invention discloses an intelligent control method for a sorting manipulator based on big data. First, by acquiring the image, two-dimensional coordinates, shape and size of the target item and calculating its geometric data, then constructing a data set based on the two-dimensional coordinates and geometric data, and obtaining the grasping force and grasping coordinates through training with a convolutional neural network, thereby determining the motion parameters. Then, calculate the three-dimensional coordinates of the target item according to the target item image, and compare with the preset three-dimensional motion coordinates to obtain the motion deviation value. Finally, construct a data set in combination with the motion parameters and the deviation value, use the particle swarm optimization algorithm to obtain the optimal motion parameters and match with the preset data to obtain the optimal motion mode of the manipulator.

[0088] The method of the present invention can dynamically adjust the motion mode according to the real-time feedback during the grasping process, thereby improving the overall efficiency and reliability of the sorting system. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0090] Figure 1 FIG. is a schematic flow chart of an intelligent control method for a sorting manipulator based on big data provided by the first embodiment of the present invention;

[0091] Figure 2 FIG. is a schematic structural diagram of an intelligent control system for a sorting manipulator based on big data provided by the second embodiment of the present invention. Detailed implementation mode

[0092] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

[0093] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0094] Refer to Figure 1 , the first embodiment of the present invention provides an intelligent control method for a sorting manipulator based on big data, including the following steps:

[0095] S11, obtaining an image of the target item, two-dimensional coordinates of the target item, the shape of the target item, and the size of the target item;

[0096] S12, performing geometric calculations based on the shape and size of the target item to obtain geometric data of the target item;

[0097] S13, constructing a geometric data set based on the two-dimensional coordinates and geometric data of the target item, and training the geometric data set using a convolutional neural network to obtain the grasping force of the manipulator and the grasping coordinates of the manipulator;

[0098] S14, performing closed-loop adjustment based on the grasping force of the manipulator and the grasping coordinates of the manipulator to obtain the motion parameters of the manipulator;

[0099] S15, preprocessing the image of the target item and performing three-dimensional coordinate calculation on the preprocessed image to obtain the three-dimensional coordinates of the target item;

[0100] S16, calculating the deviation between the three-dimensional coordinates of the target item and the preset three-dimensional motion coordinates to obtain a motion deviation value;

[0101] S17. Construct a motion data set based on the robot arm motion parameters and the motion deviation value, and use the particle swarm optimization algorithm to optimize the motion data set to obtain the optimal motion parameters of the robot arm.

[0102] S18. Match the optimal motion parameters of the robot arm with the preset robot arm motion data to obtain the optimal motion mode of the robot arm, and control the motion of the robot arm according to the optimal motion mode of the robot arm.

[0103] In step S11, obtain the target item image, the two-dimensional coordinates of the target item, the shape of the target item, and the size of the target item.

[0104] It should be noted that the image of the target item is obtained through 3D scanning technology. The 3D scanner can provide high-precision 3D point cloud data, and can accurately capture the shape, size, and surface features of the target item. For example, when detecting mechanical parts with complex shapes, the high-precision scanning function of the 3D scanner can clearly display the contour, curvature, and surface details of the parts, which is convenient for accurately identifying the features of the parts in subsequent image processing, and can even detect minor manufacturing defects or wear conditions. The automatic scanning function of the 3D scanner further improves the efficiency and quality of data acquisition. Since the surface of the target item has complex geometries, manual measurement will consume a lot of time and effort, and it is difficult to ensure the accuracy of each measurement. The automatic scanning function can quickly obtain complete 3D data to ensure that every part can be accurately recorded, thereby obtaining high-quality images.

[0105] In one implementation, relying solely on the target item image from a single perspective has limitations when dealing with complex scenarios. To overcome this limitation, a multi-view imaging system is combined. The system consists of multiple components, including multiple high-resolution cameras, a synchronous trigger controller, and image fusion software. These components work together to achieve the all-round imaging of the target item. Specifically, multiple high-resolution cameras simultaneously capture the target item from different angles, and the synchronous trigger controller ensures that all cameras capture at the same moment, thus avoiding image inconsistencies caused by time differences. The image fusion software then aligns and fuses the images captured from multiple perspectives to generate a complete 3D reconstruction image. For example, when detecting a target item with complex surface textures and occlusions, the multi-view imaging system can capture detailed information from different angles. These local images are processed by the image fusion software to form a complete 3D image containing specific details, thereby providing more comprehensive data support for subsequent analysis and processing.

[0106] It should be noted that the two-dimensional coordinates refer to the position of the target object in the image, which is achieved through target detection techniques such as edge detection and contour extraction. By locating the bounding box of the target object, the two-dimensional coordinates can be obtained. The upper-left coordinates and the width and height of the bounding box are the two-dimensional coordinates of the target object. The shape of the target object can be determined by analyzing its contour features. First, use an edge detection algorithm (such as the Canny algorithm) to extract the edge information in the image, and then obtain the contour of the target object through contour detection. According to the geometric features of the contour (such as area, perimeter, shape factor, etc.), the shape of the target object can be judged. For example, calculate the shape factor through the area and perimeter of the contour, and then distinguish between circular, rectangular or irregular shapes. After obtaining the contour of the target object, calculate its circumscribed rectangle or minimum circumscribed circle to obtain the width and height of the target object. For regular shapes (such as circles and rectangles), directly obtain the dimensions through the width and height of the bounding box; for irregular shapes, approximate its dimensions through the geometric features of the contour (such as the maximum width and height).

[0107] In step S12, geometric calculations are performed based on the shape and dimensions of the target object to obtain geometric data of the target object.

[0108] The shapes of the target object include: circular objects, rectangular objects, rhombus objects, triangular objects, and trapezoidal objects;

[0109] The dimensions of the target object include: object length, object width, object radius, and object angle;

[0110] The geometric data of the target object includes: object area, object perimeter, object curvature, and object centroid.

[0111] It should be noted that for a circular object, first measure its radius . Based on geometric formulas, calculate the area of the circular object as , and the perimeter as . Since the edge of the circular object is a continuous curve, its curvature can be calculated by the formula and remains consistent at each point on the circumference. The object centroid is located at the center of the circle, that is, the coordinates (0, 0). For example, when the radius of a circular object is measured to be 5 cm, its area is 78.5 square cm, its perimeter is 31.4 cm, its curvature is 0.2 , and the centroid is located at the center of the circle.

[0112] It should be noted that for a rectangular object, measure its length and width . The area of the rectangular object can be calculated by the formula , and the perimeter is 2 . Since the edges of a rectangular object are straight lines, its curvature is zero. The centroid of the object is located at the geometric center of the rectangle, i.e., the coordinates . For example, for a rectangular object with a length of 10 cm and a width of 5 cm, its area is 50 square centimeters, its perimeter is 30 cm, its curvature is zero, and its centroid is located at (5 2.5).

[0113] It should be noted that for a rhombus object, measure the lengths of its diagonals and . The area of a rhombus object can be calculated by the formula , and the perimeter is . The curvature of a rhombus object is zero because its edges are straight lines. The centroid of the object is located at the intersection of the diagonals, i.e., the coordinates . For example, for a rhombus object with diagonal lengths of 8 cm and 6 cm respectively, its area is 24 square centimeters, its perimeter is 20 cm, its curvature is zero, and its centroid is located at (4 3).

[0114] It should be noted that for a triangular object, measure the lengths of its three sides , and . The area of a triangular object can be calculated by Heron's formula , where . The perimeter is . The curvature of a triangular object is zero because its edges are straight lines. The centroid of the object is located at the geometric center of the triangle and can be calculated by the formula , where , and are the coordinates of the three vertices of the triangle. For example, for a right triangular object with side lengths of 3 cm, 4 cm, and 5 cm respectively, its area is 6 square centimeters, its perimeter is 12 cm, its curvature is zero, and its centroid is located at (1 1).

[0115] It should be noted that for a trapezoidal object, measure its upper base , lower base and height . The area of a trapezoidal object can be calculated by the formula , and the perimeter is , where and are the lengths of the two waists of the trapezoid. The curvature of a trapezoidal object is zero because its edges are straight lines. The centroid of the object is located at the geometric center of the trapezoid and can be calculated by the formula Calculation. For example, for a trapezoidal item with an upper base of 2 cm, a lower base of 4 cm, and a height of 3 cm, its area is 9 square centimeters, its perimeter is 12 centimeters, its curvature is zero, and its centroid is located at (3 1).

[0116] In step S13, a geometric data set is constructed based on the two-dimensional coordinates of the target item and the geometric data of the target item, and a convolutional neural network is used for training to obtain the grasping force of the manipulator and the grasping coordinates of the manipulator.

[0117] Data preprocessing is performed on the geometric data set to obtain standard geometric data;

[0118] A convolutional layer is constructed based on the standard geometric data to obtain a convolutional layer data set;

[0119] Based on the convolutional layer data set, a convolutional neural network is used for training to obtain the grasping force of the manipulator and the grasping coordinates of the manipulator;

[0120] The preprocessing includes normalization processing and standardization processing;

[0121] Among them, the normalization processing formula is:

[0122] ;

[0123] Among them, represents the normalized geometric data, represents the geometric data, represents the minimum value in the geometric data set, represents the maximum value in the geometric data set;

[0124] The standardization processing formula is:

[0125] ;

[0126] Among them, represents the standardized geometric data, represents the mean value in the geometric data set, represents the standard deviation in the geometric data set;

[0127] The convolutional layer data set is divided into convolutional layer training set data and convolutional layer test set data;

[0128] An average pooling operation is performed on the convolutional layer training set data to obtain pooling layer data;

[0129] Iteration is performed based on the pooling layer data, a preset loss function, and a preset activation function to obtain the grasping force of the initial manipulator and the grasping coordinates of the initial manipulator;

[0130] Calculate the root mean square error based on the pooled layer data and the convolutional layer test set data to obtain the convolutional root mean square error.

[0131] When the convolutional root mean square error is less than the preset error threshold, perform backpropagation operations on the grasping force of the initial manipulator and the grasping coordinates of the initial manipulator to obtain the grasping force of the manipulator and the grasping coordinates of the manipulator.

[0132] It should be noted that first, data preprocessing is performed on the geometric dataset, including normalization and standardization processing, to ensure data consistency and trainability. Through these preprocessing steps, the geometric data is converted into standard geometric data suitable for convolutional neural network training. Next, a convolutional layer is constructed based on the standard geometric data to obtain the convolutional layer dataset. The convolutional layer slides a convolutional kernel over the input data to extract local features. For example, assume that the input data is the two-dimensional coordinates and geometric data of a preprocessed target object, the convolutional kernel size is 3×3, and the stride is 1. The feature map generated after the convolutional operation will be used for subsequent feature extraction. The convolutional layer dataset is then divided into a convolutional layer training set and a convolutional layer test set. The convolutional layer training set is used to train the convolutional neural network, while the convolutional layer test set is used to evaluate the performance of the model. During the training process, average pooling operations are performed on the convolutional layer training set data to reduce the spatial size of the feature map and extract key features. The pooled layer data is then combined with a preset loss function (such as mean squared error) and activation function (such as ReLU) for multiple iterative trainings. Through iterative optimization, the network gradually learns the mapping relationship between the input data and the output targets (i.e., the grasping force and grasping coordinates of the manipulator), and finally obtains the initial grasping force and grasping coordinates of the manipulator. To verify the performance of the model, the root mean square error (RMSE) is calculated between the pooled layer data and the convolutional layer test set data to obtain the convolutional root mean square error. When the convolutional root mean square error is less than the preset error threshold, it indicates that the model performance has met the requirements. At this time, backpropagation operations are performed on the initial grasping force and grasping coordinates of the manipulator to further optimize the network parameters, and finally obtain the accurate grasping force and grasping coordinates of the manipulator. For example, assume that the preset error threshold is 0.01. After multiple iterative trainings, the convolutional root mean square error drops to 0.008. At this time, the prediction accuracy of the model meets the actual application requirements.

[0133] Among them, the root mean square error calculation formula is:

[0134] ;

[0135] Among them, represents the root mean square error, represents the number of samples, represents the th predicted value data, represents the One actual value data.

[0136] In step S14, a closed-loop adjustment is performed according to the grasping force of the manipulator and the grasping coordinates of the manipulator to obtain the manipulator motion parameters.

[0137] Take the grasping force of the manipulator as the feedforward parameter, and perform Laplace transform on the feedforward parameter to obtain the Laplace feedforward parameter;

[0138] Take the grasping coordinates of the manipulator as the negative feedback parameter, and perform Laplace transform on the negative feedback parameter to obtain the Laplace negative feedback parameter;

[0139] Perform closed-loop calculation according to the Laplace feedforward parameter and the Laplace negative feedback parameter to obtain the manipulator motion parameters;

[0140] Among them, the manipulator motion parameters are calculated by the following formula:

[0141] ;

[0142] Among them, represents the manipulator motion parameters, represents the proportionality coefficient, represents the Laplace feedforward parameter, represents the Laplace negative feedback parameter.

[0143] It should be noted that the Laplace transform is a mathematical tool that converts a time-domain signal into a complex-frequency domain signal, which can simplify the analysis and design of system dynamic characteristics. For example, assume that the grasping force of the manipulator is a function that changes with time , through the Laplace transform, it can be converted into an expression in the complex-frequency domain , where is a complex frequency variable. Through this calculation method, the expression of the motion parameters of the manipulator in the complex frequency domain can be obtained. Then, through the inverse Laplace transform, it can be converted back to the time domain to obtain the actual motion parameters of the manipulator, which are used to control the precise motion of the manipulator. During the motion of the manipulator, a feedforward control and a negative feedback mechanism are introduced, combined with the Laplace transform, to achieve dynamic adjustment of the motion parameters of the manipulator to optimize the motion effect. The Laplace feedforward parameter directly transfers a part or all of the input signal to the output to improve the response speed of the system, and the Laplace negative feedback parameter is used to reduce errors and improve the stability of the system. The selection of the proportional coefficient is very important. If the proportional coefficient is too large, the system may exhibit oscillation phenomena; if the proportional coefficient is too small, the system may not be able to respond to errors quickly; the proportional coefficient can be determined through experiments. For example, first select a suitable and as short as possible sampling time to let the system work; then add a proportional link and adjust the proportional coefficient until the output of the system shows critical oscillation; if the single proportional link cannot meet the design requirements, then add an integral link at this time, reduce the adjusted proportional coefficient to 0.8 of the original, and then adjust the integral time parameter so that the system can maintain a small steady-state error and a small oscillation time. At this time, the proportional coefficient and the integral time constant can be adjusted simultaneously.

[0144] Exemplarily, when the proportional coefficient is , the Laplace feedforward parameter is , and the Laplace negative feedback parameter is , then through the Laplace transform, is obtained. When is large enough, is obtained through the inverse Laplace transform, and the feedback cutting parameter

[0145] In step S15, preprocessing is performed on the target item image and three-dimensional coordinate calculation is carried out to obtain the three-dimensional coordinates of the target item.

[0146] Noise is removed from the image of the target item to obtain a noise-free item image;

[0147] Distortion is removed from the noise-free item image to obtain an undistorted item image;

[0148] Three-dimensional coordinate calculation is carried out on the undistorted item image to obtain the three-dimensional coordinates of the target item;

[0149] Among them, the three-dimensional coordinates of the target item are calculated through the following formula:

[0150] ;

[0151] ;

[0152]

[0153] In the formula, is the abscissa of the target object, is the ordinate of the target object, is the vertical coordinate of the target object, is the image length, is the image width, is the focal length of the undistorted object image on the axis, is the focal length of the undistorted object image on the axis.

[0154] It should be noted that first, preprocessing is performed on the target object image, and then three-dimensional coordinate calculation is performed based on the preprocessed image. The first step of preprocessing is to remove the noise in the image to improve the quality of the image and the accuracy of subsequent processing. For example, assume that the target object image is affected by ambient light or sensor noise during the acquisition process, resulting in random pixel interference in the image. By applying denoising algorithms such as median filtering or Gaussian filtering, these random noises can be effectively eliminated to obtain a noise-free object image. After obtaining the noise-free object image, further perform undistortion processing on it to correct the image distortion caused by the camera lens distortion. Lens distortion appears as barrel distortion or pincushion distortion, which will affect the true shape and size of the objects in the image. By pre-calibrating the distortion parameters of the camera and using these parameters to correct the image, an undistorted object image can be obtained. For example, assume that the measured distortion parameters during the camera calibration process include radial distortion coefficients and tangential distortion coefficients. Using these parameters to perform distortion correction on the noise-free image can restore the straight lines in the image to straight lines, ensuring that the shape and size of the target object are accurately presented in the image. Finally, three-dimensional coordinate calculation is performed based on the undistorted object image to obtain the precise position of the target object. Three-dimensional coordinate calculation is to convert the image coordinates into actual space coordinates, and this process depends on the focal length of the camera and the position of the target object in the image.

[0155] Exemplarily, when the image length , the image width , the focal length of the undistorted object image on the axis , the focal length of the undistorted object image on the axis , the calculated three-dimensional coordinates of the target object are .

[0156] In step S16, a deviation calculation is performed based on the three-dimensional coordinates of the target object and the preset three-dimensional motion coordinates to obtain a motion deviation value.

[0157] Perform a translation offset calculation based on the three-dimensional coordinates of the target object and the preset three-dimensional motion coordinates to obtain a translation deviation value;

[0158] Perform a rotation calculation based on the translation deviation value and the preset rotation matrix to obtain a rotation deviation value;

[0159] Perform a motion deviation calculation based on the translation deviation value and the rotation deviation value to obtain a motion deviation value,

[0160] wherein, the translation deviation value is obtained through the following calculation formula:

[0161] ;

[0162] ;

[0163] ;

[0164] wherein, represents the translation deviation value, represents the three-dimensional coordinates of the target object, represents the preset three-dimensional motion coordinates;

[0165] The rotation deviation value is obtained through the following calculation formula:

[0166] ;

[0167] wherein, represents the rotation deviation value, represents the preset rotation matrix, represents the transpose of the translation deviation value;

[0168] The motion deviation value is obtained through the following calculation formula:

[0169] ;

[0170] wherein, represents the motion deviation value, represents the transpose of the rotation deviation value.

[0171] It should be noted that in order to ensure that the manipulator can accurately grasp the target object, it is necessary to calculate the deviation between the three-dimensional coordinates of the target object and the preset three-dimensional motion coordinates, so as to obtain the motion deviation value. First, the translational offset is calculated, that is, the translational deviation value is obtained by comparing the three-dimensional coordinates of the target object with the preset three-dimensional motion coordinates. Then, the rotational calculation is performed according to the translational deviation value and the preset rotation matrix to obtain the rotational deviation value. The rotation matrix is used to describe the rotation state of the target object in space. For example, when the rotation matrix is an identity matrix, the rotational deviation value is the same as the translational deviation value. Finally, the transpose of the rotational deviation value is added to the translational deviation value to obtain the motion deviation value.

[0172] Exemplarily, when the three-dimensional coordinates of the target object , the preset three-dimensional motion coordinates , and the preset rotation matrix , the calculated translational deviation value is , the rotational deviation value is , and the motion deviation value is .

[0173] In step S17, a motion data set is constructed based on the manipulator motion parameters and the motion deviation value, and the particle swarm optimization algorithm is used for optimization to obtain the optimal motion parameters of the manipulator.

[0174] According to the motion data set, data cleaning and preprocessing are performed to obtain motion training set data;

[0175] A velocity set and a position set are established based on the motion training set data, and the data of the velocity set and the position set are initialized to obtain initial velocity data and initial position data;

[0176] According to the initial velocity data and the initial position data, the particle swarm optimization algorithm is used for parameter optimization to obtain the optimal motion parameters of the manipulator.

[0177] It should be noted that, in order to achieve the optimal motion control of the manipulator, the motion data set is first subjected to data cleaning and preprocessing to remove outliers, missing values and duplicate values in the data, ensuring the accuracy and consistency of the data. For example, if the motion deviation value at a certain moment shows an abnormal jump, it can be corrected by interpolation or referring to the values of adjacent data points; for missing data points, linear interpolation or mean filling methods are used to complete them. In addition, the data is standardized or normalized so that motion parameters with different dimensions are under the same standard, and the values of all motion parameters are converted to the range of 0-1 for subsequent optimization calculations. Then, based on the preprocessed motion training set data, a speed set and a position set are established, and the data of the speed set and the position set are initialized to obtain initial speed data and initial position data. In the particle swarm optimization algorithm, each particle represents a set of manipulator motion parameters. The position set of the particle represents the initial values of this set of parameters, and the speed set represents the moving speed of the particle in the solution space. The speed set and the position set are initialized by random generation, that is, each particle is given a set of random initial speed and position data. These initial data will be used as the starting point of the particle swarm optimization algorithm for subsequent iterative optimization processes. Subsequently, parameter optimization is carried out using the particle swarm optimization algorithm according to the initial speed data and the initial position data. In the optimization process, the particle swarm algorithm iteratively updates the position and speed of each particle, gradually searches the solution space, and finds the optimal manipulator motion parameters. Specifically, each particle adjusts its flight direction and speed according to the individual optimal solution and the global optimal solution. After multiple iterations, when the convergence condition of the particle swarm algorithm meets the preset accuracy range, the optimal motion parameters of the manipulator can be obtained.

[0178] Exemplarily, in the process of optimizing the motion parameters of the manipulator, it is assumed that there are three particles in the particle swarm, representing three different combinations of motion parameters and deviation values. The initial position sets are [0.5, 0.3], [0.6, 0.4] and [0.4, 0.2] respectively, representing three different initial motion speeds and deviation values. The initial speed sets are [0.1, 0.1], [0.1, 0.1] and [0.1, 0.1] respectively, representing the initial adjustment speed of each parameter. Set the inertia weight , learning factor , maximum number of iterations . After 100 iterations, the global optimal solution is 0.7, that is, the optimal motion parameter of the manipulator is 0.7.

[0179] In step S18, the optimal motion mode of the manipulator is obtained by matching the optimal motion parameters of the manipulator with the preset manipulator motion data, so as to control the motion of the manipulator according to the optimal motion mode of the manipulator.

[0180] When the optimal motion parameter is greater than or equal to a preset first motion parameter threshold, the optimal motion mode of the manipulator is a preset first motion mode;

[0181] When the optimal motion parameter is greater than or equal to a preset second motion parameter threshold and less than the preset first motion parameter threshold, the optimal motion mode of the manipulator is a preset second motion mode;

[0182] When the optimal motion parameter is greater than or equal to a preset third motion parameter threshold and less than the preset second motion parameter threshold, the optimal motion mode of the manipulator is a preset third motion mode;

[0183] When the optimal motion parameter is greater than or equal to a preset fourth motion parameter threshold and less than the preset third motion parameter threshold, the optimal motion mode of the manipulator is a preset fourth motion mode;

[0184] When the optimal motion parameter is less than the preset fourth motion parameter threshold, the optimal motion mode of the manipulator is a preset fifth motion mode.

[0185] It should be noted that each motion mode is different. For example, the first motion mode corresponds to the manipulator performing fast and large-amplitude motion tasks, such as fast material handling on a production line. At this time, the manipulator needs to complete the task with a relatively high speed and a large acceleration to meet the requirements of high-efficiency production. The second motion mode is suitable for tasks with medium speed and accuracy requirements, such as grasping and installing components on an assembly line. The third motion mode is used for tasks that require high precision but not high speed, such as precise operation of tiny electronic components in electronic manufacturing. The fourth motion mode is suitable for tasks with extremely high precision requirements and slow speed, such as the assembly of precision components in medical device manufacturing. The fifth motion mode is used for low-speed and high-precision micro-operation tasks, such as the processing of biological samples in a laboratory. Each motion mode has been optimized according to different task requirements to ensure that the manipulator can complete tasks efficiently and accurately under various complex working conditions. Through this hierarchical matching strategy based on the optimal motion parameter, the manipulator can intelligently select the most suitable motion mode for the current task requirements, thereby achieving highly adaptive motion control.

[0186] Exemplarily, when the preset first motion parameter threshold is 0.8, the preset second motion parameter threshold is 0.6, the preset third motion parameter threshold is 0.4, and the preset fourth motion parameter threshold is 0.2, and the optimal motion parameter of the manipulator is 0.7, at this time the optimal motion parameter is greater than the preset second motion parameter threshold and less than the preset first motion parameter threshold, so the optimal motion mode of the manipulator is the preset second motion mode. In the embodiments of the present invention, the setting of the threshold is not limited.

[0187] Refer toFigure 2 , the second embodiment of the present invention provides an intelligent control system for a sorting manipulator based on big data, including: a data acquisition module, a geometric data calculation module, a convolutional network training module, a closed-loop motion adjustment module, a three-dimensional coordinate calculation module, a motion deviation calculation module, a motion parameter optimization module, and an optimal mode matching module;

[0188] The data acquisition module is used to acquire the image of the target item, the two-dimensional coordinates of the target item, the shape of the target item, and the size of the target item;

[0189] The geometric data calculation module is used to perform geometric calculations based on the shape and size of the target item to obtain the geometric data of the target item;

[0190] The convolutional network training module is used to construct a geometric data set based on the two-dimensional coordinates of the target item and the geometric data of the target item, and train it using a convolutional neural network to obtain the grasping force of the manipulator and the grasping coordinates of the manipulator;

[0191] The closed-loop motion adjustment module is used to perform closed-loop adjustment based on the grasping force of the manipulator and the grasping coordinates of the manipulator to obtain the motion parameters of the manipulator;

[0192] The three-dimensional coordinate calculation module is used to preprocess the image of the target item and perform three-dimensional coordinate calculation to obtain the three-dimensional coordinates of the target item;

[0193] The motion deviation calculation module is used to calculate the deviation between the three-dimensional coordinates of the target item and the preset three-dimensional motion coordinates to obtain the motion deviation value;

[0194] The motion parameter optimization module is used to construct a motion data set based on the motion parameters of the manipulator and the motion deviation value, and optimize it using a particle swarm optimization algorithm to obtain the optimal motion parameters of the manipulator;

[0195] The optimal mode matching module is used to match the optimal motion parameters of the manipulator with the preset manipulator motion data to obtain the optimal motion mode of the manipulator, so as to control the motion of the manipulator according to the optimal motion mode of the manipulator.

[0196] It should be noted that the intelligent control system for a sorting manipulator based on big data provided by the embodiment of the present invention is used to execute all the process steps of the intelligent control method for a sorting manipulator based on big data in the above embodiment, and the working principles and beneficial effects of the two correspond one by one, so they will not be repeated here.

[0197] An embodiment of the present invention also provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as an intelligent control program for a sorting manipulator based on big data. When the processor executes the computer program, the steps in each of the above embodiments of the intelligent control method for a sorting manipulator based on big data are implemented, such as Figure 1 step S11 shown. Alternatively, when the processor executes the computer program, the functions of each module / unit in each of the above device embodiments are implemented, such as a data acquisition module.

[0198] Exemplarily, the computer program may be divided into one or more modules / units, and the one or more modules / units are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in the electronic device.

[0199] The electronic device may be a computing device such as a desktop computer, a notebook, a palm computer, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the above components are only examples of the electronic device and do not constitute a limitation on the electronic device. It may include more or fewer components than the above, or combine certain components, or different components. For example, the electronic device may further include input / output devices, network access devices, a bus, etc.

[0200] The so-called processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor is the control center of the electronic device and connects various parts of the entire electronic device through various interfaces and lines.

[0201] The memory can be used to store the computer programs and / or modules. By running or executing the computer programs and / or modules stored in the memory, and by invoking the data stored in the memory, the processor implements various functions of the electronic device. The memory mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function (such as a sound playback function, an image playback function, etc.); the data storage area can store data created according to the use of the mobile phone (such as audio data, phone book, etc.). In addition, the memory can include high-speed random access memory, and can also include non-volatile memory, such as a hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one magnetic disk storage device, flash device, or other volatile solid-state storage devices.

[0202] Among them, if the modules / units integrated in the electronic device are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above-mentioned embodiment methods of the present invention, it can also be completed by a computer program instructing relevant hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, it can implement the steps of the above-mentioned various method embodiments. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the content included in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0203] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0204] The specific embodiments described above further elaborate on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. In particular, for those skilled in the art, any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An intelligent control method for a sorting manipulator based on big data, characterized in that, Including the following steps: Obtain the image of the target object, the two-dimensional coordinates of the target object, the shape of the target object, and the size of the target object; Perform geometric calculations based on the shape and size of the target object to obtain the geometric data of the target object; Construct a geometric dataset based on the two-dimensional coordinates and geometric data of the target object, and use a convolutional neural network to train the geometric dataset to obtain the grasping force of the manipulator and the grasping coordinates of the manipulator; Perform closed-loop adjustment based on the grasping force of the manipulator and the grasping coordinates of the manipulator to obtain the motion parameters of the manipulator; Preprocess the image of the target object and perform three-dimensional coordinate calculation on the preprocessed image to obtain the three-dimensional coordinates of the target object; Calculate the deviation based on the three-dimensional coordinates of the target object and the preset three-dimensional motion coordinates to obtain the motion deviation value; Construct a motion dataset based on the motion parameters of the manipulator and the motion deviation value, and use the particle swarm optimization algorithm to optimize the motion dataset to obtain the optimal motion parameters of the manipulator; Match the optimal motion parameters of the manipulator with the preset manipulator motion data to obtain the optimal motion mode of the manipulator, and control the motion of the manipulator according to the optimal motion mode of the manipulator; Perform closed-loop adjustment based on the grasping force of the manipulator and the grasping coordinates of the manipulator to obtain the motion parameters of the manipulator, including the following steps: Take the grasping force of the manipulator as the feedforward parameter, and perform Laplace transform on the feedforward parameter to obtain the Laplace feedforward parameter; Take the grasping coordinates of the manipulator as the negative feedback parameter, and perform Laplace transform on the negative feedback parameter to obtain the Laplace negative feedback parameter; Perform closed-loop calculation based on the Laplace feedforward parameter and the Laplace negative feedback parameter to obtain the motion parameters of the manipulator; Among them, the motion parameters of the manipulator are calculated by the following formula: ； Among them, represents the motion parameters of the manipulator, represents the proportional coefficient, represents the Laplace feedforward parameter, represents the Laplace negative feedback parameter.

2. The intelligent control method for a sorting manipulator based on big data according to claim 1, characterized in that The shapes of the target objects include: circular objects, rectangular objects, rhombus objects, triangular objects, and trapezoidal objects; The sizes of the target objects include: object length, object width, object radius, and object angle; The geometric data of the target object includes: object area, object perimeter, object curvature, and object centroid.

3. The intelligent control method for a sorting manipulator based on big data according to claim 1, characterized in that The process of constructing a geometric dataset based on the two-dimensional coordinates and geometric data of the target object and using a convolutional neural network for training to obtain the grasping force of the manipulator and the grasping coordinates of the manipulator includes the following steps: Perform data preprocessing on the geometric dataset to obtain standard geometric data; Construct a convolutional layer based on the standard geometric data to obtain a convolutional layer dataset; Use a convolutional neural network to train according to the convolutional layer dataset to obtain the grasping force of the manipulator and the grasping coordinates of the manipulator; The preprocessing includes normalization processing and standardization processing; Among them, the normalization processing formula is: ； Among them, represents the normalized geometric data, represents the geometric data, represents the minimum value in the geometric data set, represents the maximum value in the geometric data set; The standardization processing formula is: ； Among them, represents the standardized geometric data, represents the mean value in the geometric dataset, represents the standard deviation in the geometric dataset.

4. The intelligent control method of a sorting manipulator based on big data according to claim 3, characterized in that, Based on the convolutional layer dataset and using a convolutional neural network for training, the grasping force of the manipulator and the grasping coordinates of the manipulator are obtained, including the following steps: Divide the convolutional layer dataset into convolutional layer training set data and convolutional layer test set data; Perform average pooling operation according to the convolutional layer training set data to obtain pooling layer data; Iterate according to the pooling layer data, a preset loss function, and a preset activation function to obtain the grasping force of the initial manipulator and the grasping coordinates of the initial manipulator; Calculate the root mean square error between the pooling layer data and the convolutional layer test set data to obtain the convolutional root mean square error; When the convolutional root mean square error is less than a preset error threshold, perform backpropagation operation on the grasping force of the initial manipulator and the grasping coordinates of the initial manipulator to obtain the grasping force of the manipulator and the grasping coordinates of the manipulator.

5. The intelligent control method of a sorting manipulator based on big data according to claim 1, characterized in that, Preprocess the image of the target object and calculate the three-dimensional coordinates to obtain the three-dimensional coordinates of the target object, including the following steps: Remove noise from the image of the target object to obtain a noise-free object image; Remove distortion from the noise-free object image to obtain an undistorted object image; Calculate the three-dimensional coordinates according to the undistorted object image to obtain the three-dimensional coordinates of the target object; Among them, the three-dimensional coordinates of the target object are calculated by the following formula: ； ； ； Among them, represents the abscissa of the target object, represents the ordinate of the target object, represents the vertical coordinate of the target object, represents the image length, represents the image width, represents the focal length of the undistorted object image on the axis, represents the focal length of the undistorted object image on the axis.

6. The intelligent control method of a sorting manipulator based on big data according to claim 1, characterized in that, Calculate the motion deviation value according to the three-dimensional coordinates of the target object and the preset three-dimensional motion coordinates, including the following steps: Calculate the translational offset according to the three-dimensional coordinates of the target object and the preset three-dimensional motion coordinates to obtain the translational deviation value; Perform rotation calculation according to the translational deviation value and the preset rotation matrix to obtain the rotational deviation value; Calculate the motion deviation according to the translational deviation value and the rotational deviation value to obtain the motion deviation value, Among them, the translational deviation value is obtained by the following calculation formula: ； ； ； Among them, represents the translation deviation value, represents the three-dimensional coordinates of the target object, represents the preset three-dimensional motion coordinates; The rotational deviation value is obtained by the following calculation formula: ； Among them, represents the rotation deviation value, represents the preset rotation matrix, represents the transpose of the translation deviation value; The motion deviation value is obtained by the following calculation formula: ； Among them, represents the motion deviation value, represents the transpose of the rotation deviation value.

7. The intelligent control method for a sorting manipulator based on big data according to claim 1, wherein Construct a motion dataset according to the manipulator motion parameters and the motion deviation value, and use the particle swarm optimization algorithm for optimization to obtain the optimal motion parameters of the manipulator, including the following steps: Perform data cleaning and preprocessing on the motion dataset to obtain motion training set data; Establish a velocity set and a position set according to the motion training set data, and initialize the data of the velocity set and the position set to obtain initial velocity data and initial position data; Use the particle swarm optimization algorithm for parameter optimization according to the initial velocity data and the initial position data to obtain the optimal motion parameters of the manipulator.

8. The intelligent control method of a sorting manipulator based on big data according to claim 1, characterized in that Match the optimal motion parameters of the manipulator with the preset manipulator motion data to obtain the optimal motion mode of the manipulator, so as to control the motion of the manipulator according to the optimal motion mode of the manipulator, including: When the optimal motion parameter is greater than or equal to a preset first motion parameter threshold, the optimal motion mode of the manipulator is the preset first motion mode; When the optimal motion parameter is greater than or equal to a preset second motion parameter threshold and less than a preset first motion parameter threshold, the optimal motion mode of the manipulator is a preset second motion mode; When the optimal motion parameter is greater than or equal to a preset third motion parameter threshold and less than the preset second motion parameter threshold, the optimal motion mode of the manipulator is a preset third motion mode; When the optimal motion parameter is greater than or equal to a preset fourth motion parameter threshold and less than the preset third motion parameter threshold, the optimal motion mode of the manipulator is a preset fourth motion mode; When the optimal motion parameter is less than the preset fourth motion parameter threshold, the optimal motion mode of the manipulator is a preset fifth motion mode.

9. An intelligent control system for a sorting manipulator based on big data, which applies the method described in any one of claims 1-8, is characterized in that, Including: A data acquisition module, a geometric data calculation module, a convolutional network training module, a closed-loop motion adjustment module, a three-dimensional coordinate calculation module, a motion deviation calculation module, a motion parameter optimization module, and an optimal mode matching module; The data acquisition module is used to acquire the image of the target item, the two-dimensional coordinates of the target item, the shape of the target item, and the size of the target item; The geometric data calculation module is used to perform geometric calculations based on the shape and size of the target item to obtain the geometric data of the target item; The convolutional network training module is used to construct a geometric data set based on the two-dimensional coordinates and geometric data of the target item, and train it using a convolutional neural network to obtain the grasping force of the manipulator and the grasping coordinates of the manipulator; The closed-loop motion adjustment module is used to perform closed-loop adjustment based on the grasping force of the manipulator and the grasping coordinates of the manipulator to obtain the motion parameters of the manipulator; The three-dimensional coordinate calculation module is used to preprocess the image of the target item and perform three-dimensional coordinate calculation to obtain the three-dimensional coordinates of the target item; The motion deviation calculation module is used to calculate the deviation between the three-dimensional coordinates of the target item and the preset three-dimensional motion coordinates to obtain a motion deviation value; The motion parameter optimization module is used to construct a motion data set based on the motion parameters of the manipulator and the motion deviation value, and perform optimization using a particle swarm optimization algorithm to obtain the optimal motion parameters of the manipulator; The optimal mode matching module is used to match the optimal motion parameters of the manipulator with the preset manipulator motion data to obtain the optimal motion mode of the manipulator, so as to control the motion of the manipulator according to the optimal motion mode of the manipulator.

Images