A method and system for determining the grasping posture of a robotic arm

Abstract

The present invention discloses a method and system for determining the gripping posture of a robotic arm, which relates to the field of robotic arms, including: collecting multi-angle images of an object to be gripped, inputting them into a neural network to extract feature lines of the object to be gripped; selecting the optimal feature lines to construct the overall structural contour line of the object to be gripped; performing morphological judgment on the plane figures formed by the overall structural contour lines to obtain an initial gripping surface set; performing roughness judgment on the plane figures within the initial gripping surface set, selecting the optimal gripping plane and the corresponding gripping center point; determining the robotic arm wrist posture and the coordinates of the wrist swing axis center point based on the gripping center point and the normal vector of the optimal gripping plane; and determining the gripping posture of the robotic arm through inverse calculation based on the coordinates of the wrist swing axis center point and the coordinates of the mechanical arm rotation axis center point. The present invention can analyze the gripping surface to select the optimal gripping surface, thereby determining the gripping posture of the robotic arm and reducing the phenomenon of the robotic arm falling off during gripping.

Description

Technical Field

[0001] The present invention relates to the technical field of robotic arms, and more particularly to a method and system for determining a gripping posture of a robotic arm. Background Art

[0002] With the development and application of robotics technology, robotic arms are widely used in industrial production, logistics and transportation, resource exploration, surgical medicine, aerospace and military fields. Among them, the six-axis industrial robotic arm has six degrees of freedom and can move in multiple directions in space to complete complex tasks. It has good applicability in assembly, grasping and handling, grinding and polishing and other operations.

[0003] Common six-axis robotic arms typically use claw-type or suction cup structures to grab and carry objects. Claw-type robotic arms typically require design adjustments to the shape, size, and number of their claws based on the size and shape of the object being grabbed, resulting in claw-type robotic arms having limited adaptability to the objects being grabbed. Suction cup robotic arms can adapt to objects of various structural shapes, and compared to claw-type gripping structures, suction cup structures have the advantages of being simple and quick to operate, and are less likely to damage the object being grabbed. However, the gripping force of a suction cup robotic arm is easily affected by the surface flatness of the object being grabbed. The contact surface flatness of the suction cup and the object being grabbed also varies in different gripping positions. When the flatness is insufficient, the object is prone to falling off.

[0004] Therefore, how to determine the gripping posture of the robotic arm to reduce the occurrence of falling off is an urgent problem that technicians in this field need to solve. Summary of the Invention

[0005] In view of this, the present invention provides a method and system for determining the grasping posture of a robotic arm, which determines the optimal grasping surface by analyzing the roughness of each surface of the object to be grasped, and then determines the grasping posture of the robotic arm based on the optimal grasping surface, thereby reducing the occurrence of the grasped object falling off.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] The present invention discloses a method for determining a gripping posture of a robotic arm, and the specific steps are as follows:

[0008] Collect multi-angle images of the object to be grasped and input them into a neural network to extract the feature lines of the object to be grasped;

[0009] According to the line feature optimization model, the optimal feature line is selected to construct the overall structural contour line of the object to be grasped;

[0010] Performing morphological judgment on the plane figure formed by the overall structural contour line to obtain an initial grasping surface set; performing roughness judgment on the plane figures within the initial grasping surface set to select the optimal grasping plane and the corresponding grasping center point;

[0011] According to the grasping center point and the normal vector of the optimal grasping plane, the wrist posture of the robot arm and the coordinates of the center point of the wrist swing axis are determined; according to the coordinates of the center point of the wrist swing axis and the coordinates of the center point of the mechanical arm rotation axis, the grasping posture of the robot arm is determined by inverse solution.

[0012] Furthermore, the feature optimization model is expressed as:

[0013] E(L b )=E data (L b )+λE smooth (L b );

[0014] Among them, L b represents the candidate feature line set, E data (L b ) is a data item that represents the length and direction of the candidate feature line set; E smooth (L b ) is a smoothing term that characterizes the relationship between line segments in the candidate feature line set; λ is a weight coefficient that defines the importance of the data term and the smoothing term; b is the feature label of the feature line, 1 represents the overall structural contour line, and 0 represents the unclassified feature line.

[0015] Furthermore, the data item is represented as:

[0016]

[0017]

[0018] Among them, D(l c ) represents the characteristic line l c data items, Characteristic line l c The length, l max To select the longest line segment in the feature line set, m c Indicates that c The number of characteristic lines with the same direction, m represents L b The number of subsets of the same-direction characteristic line set M, α and β represent the weights of the two items respectively.

[0019] Furthermore, the smoothing term is expressed as:

[0020]

[0021] Among them, N represents the set of adjacent feature lines, V c,g (l c ,l g ) is used to measure the two adjacent characteristic lines l c , l g The smoothness of , the smaller the value, the smoother it is, and is defined as follows:

[0022]

[0023] Among them, (distlc,lg) represents the characteristic line l c With l g The closest geometric distance, f(c,g), is an indicator that if the feature line l c With l g The value is 0 if there is a direct intersection, and 1 otherwise.

[0024] Furthermore, the morphology judgment includes:

[0025] Sampling the plane figure at equal angle intervals to obtain a set of sampling points of the plane figure;

[0026] Determining the initial center coordinates of the maximum inscribed circle of the plane figure according to the sampling point set;

[0027] Determining the center range of the actual maximum inscribed circle based on the coordinates of the initial center of the maximum inscribed circle, determining the radius of the inscribed circle of each point within the center range with each point within the center range as the center, and using the coordinates of the point with the largest inscribed circle radius and the inscribed circle radius as the actual center and actual radius of the maximum inscribed circle of the plane figure;

[0028] The initial grasping surface set is obtained by taking the plane figure whose actual radius is larger than the outer edge radius of the robot arm suction cup as the initial grasping surface.

[0029] Furthermore, the calculation formula for the initial coordinates (x0, y0) of the maximum inscribed circle is:

[0030]

[0031]

[0032] Where n is the total number of sampling points in the sampling point set P; r i ,θ i are the midpoints p in the sampling point set P i Sampling radius and sampling angle;

[0033] The center range is a circular area with point (x0, y0) as the center and r0 as the radius;

[0034] The radius r of the inscribed circle of the point within the center rangej for:

[0035] r j =MINd ij (p j ,p i ), p i ∈P;

[0036]

[0037] Among them, MINd ij (p j ,p i ) represents a point p within the center of the actual maximum inscribed circle j The minimum distance from all points in the sampling point set P; (x i ,y i ), (x j ,y j ) are points p i , p j 's coordinates.

[0038] Furthermore, the roughness determination includes:

[0039] Acquire point cloud data of each plane figure within the initial grasping surface set, and obtain plane equations of each plane figure by fitting;

[0040] The annular area where the suction cup actually contacts each plane figure is determined by the actual center of each plane figure, the outer edge radius and the inner edge radius of the robot arm suction cup. The roughness is represented by the average distance between each point in the annular area and the plane equation. The plane figure with the smallest roughness is the optimal grasping plane, and the corresponding actual center of the circle is the grasping center point.

[0041] Furthermore, the plane equations of the fitted plane figures include:

[0042] Assume that the plane equation is: a1x′+a2y′+a3z′+1=0;

[0043] The objective function is: Indicates that the distance between each point on the plane figure and the plane equation is the shortest; where n′ is the total number of plane figure point clouds, x′ k ,y′ k ,z′ k The three-axis coordinates of each point on the plane figure, a1, a2, a3 are the constant terms of the plane equation; the matrix form of the plane equation is expressed as:

[0044]

[0045] After solving the matrix, we get the plane equation of the plane figure.

[0046] Furthermore, the inverse solution includes:

[0047] Determining the rotation angle of the rotation axis according to the coordinates of the center point of the wrist swing axis and the coordinates of the center point of the rotation axis;

[0048] Determine the coordinates of the center point of the lower arm axis according to the coordinates of the center point of the rotation axis and the length of the support arm; determine the swing angle of the lower arm axis according to the angle between the line connecting the coordinates of the center point of the lower arm axis and the coordinates of the center point of the wrist swing axis and the horizontal plane;

[0049] Determine the swing angle of the upper arm axis according to the distance between the coordinates of the center point of the lower arm axis and the coordinates of the center point of the wrist swing axis, as well as the length of the lower arm and the length of the upper arm;

[0050] Determine the coordinates of the center point of the upper arm axis based on the coordinates of the center point of the lower arm axis and the swing angle of the lower arm axis; determine the direction vector of the upper arm based on the coordinates of the center point of the upper arm axis and the coordinates of the center point of the wrist swing axis; determine the rotation angle of the wrist rotation axis and the swing angle of the wrist swing axis based on the direction vector of the upper arm and the direction vector of the wrist.

[0051] The present invention also discloses a system for determining a gripping posture of a robotic arm, comprising:

[0052] Feature line extraction module: collects multi-angle images of the object to be grasped, inputs them into the neural network to extract the feature lines of the object to be grasped;

[0053] Contour line generation module: selects the optimal feature line according to the line feature optimization model to construct the overall structural contour line of the object to be grasped;

[0054] Grasping surface generation module: performs morphological judgment on the plane figure formed by the overall structure contour line to obtain an initial grasping surface set; performs roughness judgment on the plane figures in the initial grasping surface set, and selects the optimal grasping plane and the corresponding grasping center point;

[0055] Posture determination module: Determine the posture of the robot arm wrist and the coordinates of the center point of the wrist swing axis based on the grasping center point and the normal vector of the optimal grasping plane; determine the grasping posture of the robot arm through inverse solution based on the coordinates of the center point of the wrist swing axis and the coordinates of the center point of the rotation axis of the robot arm.

[0056] Through the above technical solutions, it can be seen that compared with the prior art, the present invention discloses a method and system for determining the grasping posture of a robotic arm, which can accurately and quickly extract the feature lines of the object to be grasped through images and neural network models; by preferentially selecting the extracted feature lines, a simplified overall contour representing the object to be grasped can be obtained, thereby reducing the amount of calculation required for grasping surface judgment; by judging the morphology of the initial grasping surface, the amount of calculation required for roughness analysis can be reduced, thereby improving the system operation speed; finally, the grasping surface with the best flatness is preferentially selected, and the posture of the robotic arm wrist is determined based on the grasping point and the normal vector of the grasping surface, and the final grasping posture of the robotic arm is determined by further inverse calculation. The present invention can analyze the grasped surface and select the optimal grasping surface, thereby determining the grasping posture of the robotic arm, reducing the phenomenon of the robotic arm grasping and falling off, and improving the adaptability of the robotic arm to different objects to be grasped. 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 following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0058] Figure 1 Schematic diagram of the overall process of an embodiment of the present invention.

[0059] Figure 2 Schematic diagram of a robotic arm according to an embodiment of the present invention.

[0060] In the figure, 1, rotation axis; 2, support arm; 3, lower arm axis; 4, lower arm; 5, upper arm axis; 6, upper arm; 7, wrist rotation axis; 8, wrist swing axis; 9, wrist; 10, wrist rotation axis; 11, suction cup. DETAILED DESCRIPTION

[0061] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0062] The embodiment of the present invention discloses a method for determining the grasping posture of a robotic arm, such as Figure 1 The specific steps are as follows:

[0063] Collect multi-angle images of the object to be grasped and input them into the neural network to extract the feature lines of the object to be grasped;

[0064] According to the line feature optimization model, the optimal feature line is selected to construct the overall structural contour line of the object to be grasped;

[0065] Perform morphological judgment on the plane figures formed by the overall structural contour lines to obtain the initial grasping surface set; perform roughness judgment on the plane figures within the initial grasping surface set to select the optimal grasping plane and the corresponding grasping center point;

[0066] According to the grasping center point and the normal vector of the optimal grasping plane, the wrist posture of the robot arm and the coordinates of the center point of the wrist swing axis are determined; according to the coordinates of the center point of the wrist swing axis and the coordinates of the center point of the mechanical arm rotation axis 1, the grasping posture of the robot arm is determined by inverse solution.

[0067] In a specific embodiment, the neural network is an improved three-branch YOLOv5 neural network, in which the head structure of the YOLOv5 neural network is divided into multiple branches, each branch is responsible for predicting a set of target boxes of different scales; the CBAM attention module is introduced into the backbone and feature fusion layer of the YOLOv5 neural network; and the MBConv convolution module is used to replace the ordinary convolution module in the YOLOv5 neural network.

[0068] RGB-D cameras positioned directly in front of, behind, to the left, right, and above the object to be grasped capture images with depth information, which are then fed into an improved three-branch YOLOv5 neural network to identify the object's feature lines. The improved three-branch YOLOv5 neural network has three branches, enabling better identification of feature lines of varying scales. An attention module is also introduced to improve the accuracy of feature line recognition. The MBConv convolutional module is a lightweight convolutional module that effectively reduces the number of parameters used in the convolution process compared to traditional convolutional modules. Replacing the original YOLOv5 neural network convolutional module with the improved MBConv convolutional module speeds up model training and recognition. After identifying feature lines using the improved three-branch YOLOv5 neural network, the camera's 3D coordinates of each pixel on the feature line are determined based on the image's depth information. These coordinates are then converted to 3D coordinates with the center of the robotic arm's rotation axis as the origin.

[0069] In a specific embodiment, the feature optimization model is expressed as:

[0070] E(L b )=E data (L b )+λE smooth (L b );

[0071] in, The candidate feature line set h is the total number of feature lines, E data (L b) is a data item that represents the length and direction of the candidate feature line set; E smooth (L b ) is a smoothing term that characterizes the relationship between line segments in the candidate feature line set; λ is a weight coefficient that defines the importance of the data term and the smoothing term; b is the feature label of the feature line, 1 represents the overall structural contour line, and 0 represents the unclassified feature line.

[0072] Specifically, when using the feature optimization model to select the optimal feature line, first set the initial feature flags of all feature lines to 0. The feature flags of the feature lines are only used to mark whether the feature lines have been classified and do not affect the calculation of the optimization value. Then, determine whether the feature lines are the optimal feature lines one by one. Specifically, calculate the optimization value E1 of the current candidate feature line set and the optimization value E′1 of the candidate feature line set when it does not contain the feature line to be classified. If E1 is less than or equal to E′1, the feature line is removed from the candidate feature line set; if E1 is greater than E′1, the feature flag of the feature line is set to 1, and the classification judgment of the next feature line is continued. Finally, when the feature flags of all feature lines in the candidate feature line set are 1, the feature line optimization ends, and the feature lines of the selected feature line set at this time are used as the overall structural contour lines.

[0073] In a specific embodiment, the data items are represented as:

[0074]

[0075]

[0076] Among them, D(l c ) represents the characteristic line l c data items, Characteristic line l c The length, l max To select the longest line segment in the feature line set, m c Indicates that c The number of characteristic lines with the same direction, m represents L b The number of subsets of the same-direction characteristic line set M, α and β represent the weights of the two items respectively.

[0077] The smoothing term is expressed as:

[0078]

[0079] Among them, N represents the set of adjacent feature lines, V c,g (l c ,l g ) is used to measure the two adjacent characteristic lines l c , l g The smoothness of , the smaller the value, the smoother it is, and is defined as follows:

[0080]

[0081] Among them, (distlc,lg) represents the characteristic line l c With l g The closest geometric distance, f(c,g), is an indicator that if the feature line l c With l g The value is 0 if there is a direct intersection, and 1 otherwise.

[0082] In a specific embodiment, morphology determination includes:

[0083] Performing equal-angle sampling on the plane figure to obtain a set of sampling points of the plane figure; determining the coordinates of the initial center of the maximum inscribed circle of the plane figure based on the set of sampling points;

[0084] Determine the center range of the actual maximum inscribed circle based on the coordinates of the initial center of the maximum inscribed circle, take each point within the center range as the center, determine the radius of the inscribed circle of each point within the center range, and take the coordinates of the point with the largest inscribed circle radius and the inscribed circle radius as the actual center and actual radius of the maximum inscribed circle of the plane figure;

[0085] The initial grasping surface set is obtained by taking a plane figure whose actual radius is larger than the outer edge radius of the robot arm suction cup as the initial grasping surface.

[0086] Specifically, for the convenience of calculation, after obtaining multiple plane figures based on the overall structural contour line, all the plane figures are placed in a plane coordinate system of uniform size, and sampling is performed at equal angles with the origin of the coordinate system. Then, the maximum inscribed circle radius is calculated, and by judging the size of the inscribed circle and the outer edge of the robotic arm suction cup, a grasping plane that meets the suction cup grasping requirements in terms of morphology is selected.

[0087] In a specific embodiment, the calculation formula for the initial coordinates (x0, y0) of the maximum inscribed circle is:

[0088]

[0089]

[0090] Where n is the total number of sampling points in the sampling point set P; r i ,θ i are the midpoints p in the sampling point set P i Sampling radius and sampling angle;

[0091] The circle center range is a circular area with point (x0, y0) as the center and r0 as the radius;

[0092] The radius r of the inscribed circle of the point within the center of the circle j for:

[0093] r j =MINd ij (p j ,p i ), p i ∈P;

[0094]

[0095] Among them, MINd ij (p j ,p i ) represents a point p within the center of the actual maximum inscribed circle j The minimum distance from all points in the sampling point set P; (x i ,y i ), (x j ,y j ) are points p i , p j 's coordinates.

[0096] In a specific embodiment, the roughness determination includes:

[0097] Obtain the point cloud data of each plane figure within the initial grasping surface set, and obtain the plane equation of each plane figure by fitting;

[0098] The actual center of each plane figure, the outer radius and inner radius of the robot arm suction cup are used to determine the annular area where the suction cup actually contacts each plane figure. The average distance between each point in the annular area and the plane equation is used to represent the roughness. The plane figure with the smallest roughness is the optimal grasping plane, and the corresponding actual center of the circle is the grasping center point.

[0099] Specifically, the point cloud data of each plane figure can be obtained through the laser radar, ultrasonic radar, etc. installed on the robotic arm. By fitting the plane equation and calculating the roughness of the actual contact annular area, the plane with the best flatness in the initial grasping plane set can be selected, thereby ensuring that the final grasping posture can maximize the grasping force of the suction cup and reduce falling.

[0100] In a specific embodiment, the plane equations of the plane figures obtained by fitting include:

[0101] Assume that the plane equation is: a1x′+a2y′+a3z′+1=0;

[0102] The objective function is: Indicates that the distance from each point on the plane figure to the plane equation is the shortest, where n′ is the total number of plane figure point clouds, x′ k ,y′ k ,z′ kThe three-axis coordinates of each point on the plane figure, a1, a2, a3 are the constant terms of the plane equation; then:

[0103]

[0104]

[0105]

[0106] The matrix form of the plane equation is expressed as:

[0107]

[0108] After solving the matrix, we get the plane equation of the plane figure.

[0109] In a specific embodiment, Figure 2 As shown, the robotic arm includes a rotating shaft 1, a supporting arm 2, a lower arm shaft 3, a lower arm 4, an upper arm shaft 5, an upper arm 6, a wrist swing shaft 8, a wrist 9, a wrist rotation shaft 10, and a suction cup 11 connected in sequence, wherein the upper arm shaft 5 is also provided with a wrist rotating shaft 7; the rotating shaft 1 is used to control the circumferential rotation of the robotic arm in the horizontal plane, the lower arm shaft 3 and the upper arm shaft 5 are used to control the lower arm 4 and the upper arm 6 to swing up and down in the vertical plane respectively, the wrist rotating shaft 7 and the wrist swing shaft 8 are used to control the wrist 9 to realize two-axis spherical motion, and the wrist rotation shaft 10 is used to control the circumferential rotation of the suction cup 11.

[0110] The inverse solution includes:

[0111] Determine the rotation angle of the rotation axis 1 according to the coordinates of the center point of the wrist swing axis 8 and the coordinates of the center point of the rotation axis 1;

[0112] The coordinates of the center point of the lower arm axis 3 are determined according to the coordinates of the center point of the rotation axis 1 and the length of the support arm 2; the swing angle of the lower arm axis 3 is determined according to the angle between the line connecting the coordinates of the center point of the lower arm axis 3 and the coordinates of the center point of the wrist swing axis 8 and the horizontal plane;

[0113] The swing angle of the upper arm axis 5 is determined based on the distance between the center coordinates of the lower arm axis 3 and the center coordinates of the wrist swing axis 8, as well as the lengths of the lower arm 4 and the upper arm 6;

[0114] Determine the center point coordinates of the upper arm axis 5 based on the center point coordinates of the lower arm axis 3 and the swing angle of the lower arm axis 3; determine the direction vector of the upper arm 6 based on the center point coordinates of the upper arm axis 5 and the center point coordinates of the wrist swing axis 8; determine the rotation angle of the wrist rotation axis 7 and the swing angle of the wrist swing axis 8 based on the direction vector of the upper arm 6 and the direction vector of the wrist 9.

[0115] The coordinates of the center point of the rotation axis 1 are the intersection of the rotation axis of the rotation axis 1 and the horizontal plane; the center points of the lower arm axis 3, the upper arm axis 5, and the wrist swing axis 8 are the intersections of their respective rotation axis center lines with the center lines of the lower arm 4, upper arm 6, and wrist 9 in the length direction.

[0116] In a specific embodiment, the method further includes performing collision judgment on the determined robot arm posture. When it is determined that a collision occurs, the determined optimal grasping plane is removed from the initial grasping plane set, and a new optimal grasping plane is selected.

[0117] The embodiment of the present invention further discloses a system for determining a gripping posture of a robotic arm, comprising:

[0118] Feature line extraction module: collects multi-angle images of the object to be grasped and inputs them into the neural network to extract the feature lines of the object to be grasped;

[0119] Contour line generation module: Based on the line feature optimization model, the optimal feature line is selected to construct the overall structural contour line of the object to be grasped;

[0120] Grasping surface generation module: performs morphological judgment on the plane figures formed by the overall structural contour lines to obtain the initial grasping surface set; performs roughness judgment on the plane figures within the initial grasping surface set, and selects the optimal grasping plane and the corresponding grasping center point;

[0121] Posture determination module: Determine the posture of the robot arm wrist and the coordinates of the center point of the wrist swing axis based on the grasping center point and the normal vector of the optimal grasping plane; determine the grasping posture of the robot arm through inverse solution based on the coordinates of the center point of the wrist swing axis and the coordinates of the center point of the rotation axis of the robot arm.

[0122] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0123] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for determining a grasping posture of a robotic arm, characterized in that: The specific steps are as follows: Collect multi-angle images of the object to be grasped and input them into a neural network to extract the feature lines of the object to be grasped; According to the line feature optimization model, the optimal feature line is selected to construct the overall structural contour line of the object to be grasped; Performing morphological judgment on the plane figure formed by the overall structural contour line to obtain an initial grasping surface set; performing roughness judgment on the plane figures within the initial grasping surface set to select the optimal grasping plane and the corresponding grasping center point; According to the grasping center point and the normal vector of the optimal grasping plane, the wrist posture of the robot arm and the coordinates of the center point of the wrist swing axis are determined; according to the coordinates of the center point of the wrist swing axis and the coordinates of the center point of the mechanical arm rotation axis, the grasping posture of the robot arm is determined by inverse solution.

2. A method for determining a gripping posture of a robotic arm according to claim 1, characterized in that: The feature optimization model is expressed as: E(L b )=E data (L b )+λE smooth (L b ); Among them, L b represents the candidate feature line set, E data (L b ) is a data item that represents the length and direction of the candidate feature line set; E smooth (L b ) is a smoothing term that characterizes the relationship between line segments in the candidate feature line set; λ is a weight coefficient that defines the importance of the data term and the smoothing term; b is the feature label of the feature line, 1 represents the overall structural contour line, and 0 represents the unclassified feature line.

3. A method for determining a gripping posture of a robotic arm according to claim 2, characterized in that: The data items are represented as: Among them, D(l c ) represents the characteristic line l c data items, Characteristic line l c The length, l max To select the longest line segment in the feature line set, m c Indicates that c The number of characteristic lines with the same direction, m represents L b The number of subsets of the same-direction characteristic line set M, α and β represent the weights of the two items respectively.

4. A method for determining a gripping posture of a robotic arm according to claim 2, characterized in that: The smoothing term is expressed as: Among them, N represents the set of adjacent feature lines, V c,g (l c ,l g ) is used to measure the two adjacent characteristic lines l c , l g The smoothness of , the smaller the value, the smoother it is, and is defined as follows: Among them, (distlc,lg) represents the characteristic line l c With l g The closest geometric distance, f(c,g), is an indicator that if the feature line l c With l g The value is 0 if there is a direct intersection, and 1 otherwise.

5. The method for determining the grasping posture of a robotic arm according to claim 1, wherein: The morphology judgment includes: Sampling the plane figure at equal angle intervals to obtain a set of sampling points of the plane figure; Determining the initial center coordinates of the maximum inscribed circle of the plane figure according to the sampling point set; Determining the center range of the actual maximum inscribed circle based on the coordinates of the initial center of the maximum inscribed circle, determining the radius of the inscribed circle of each point within the center range with each point within the center range as the center, and using the coordinates of the point with the largest inscribed circle radius and the inscribed circle radius as the actual center and actual radius of the maximum inscribed circle of the plane figure; The initial grasping surface set is obtained by taking the plane figure whose actual radius is larger than the outer edge radius of the robot arm suction cup as the initial grasping surface.

6. A method for determining a gripping posture of a robotic arm according to claim 5, characterized in that: The calculation formula for the initial coordinates (x0, y0) of the maximum inscribed circle is: Where n is the total number of sampling points in the sampling point set P; r i ,θ i are the midpoints p in the sampling point set P i Sampling radius and sampling angle; The center range is a circular area with point (x0, y0) as the center and r0 as the radius; The radius r of the inscribed circle of the point within the center range j for: r j =MINd ij (p j ,p i ),p i ∈P; Among them, MINd ij (p j ,p i ) represents a point p within the center of the actual maximum inscribed circle j The minimum distance from all points in the sampling point set P; (x i ,y i ), (x j ,y j ) are points p i , p j 's coordinates.

7. The method for determining the grasping posture of a robotic arm according to claim 5, wherein: The roughness judgment includes: Acquire point cloud data of each plane figure within the initial grasping surface set, and obtain plane equations of each plane figure by fitting; The annular area where the suction cup actually contacts each plane figure is determined by the actual center of each plane figure, the outer edge radius and the inner edge radius of the robot arm suction cup. The roughness is represented by the average distance between each point in the annular area and the plane equation. The plane figure with the smallest roughness is the optimal grasping plane, and the corresponding actual center of the circle is the grasping center point.

8. A method for determining a gripping posture of a robotic arm according to claim 7, characterized in that: The plane equations of the fitted plane figures include: Assume that the plane equation is: a1x′+a2y′+a3z′+1=0; The objective function is: Indicates that the distance between each point on the plane figure and the plane equation is the shortest; where n′ is the total number of plane figure point clouds, x′ k ,y′ k ,z′ k The three-axis coordinates of each point on the plane figure, a1, a2, a3 are the constant terms of the plane equation; the matrix form of the plane equation is expressed as: After solving the matrix, we get the plane equation of the plane figure.

9. The method for determining the grasping posture of a robotic arm according to claim 1, wherein: The inverse solution includes: Determining the rotation angle of the rotation axis according to the coordinates of the center point of the wrist swing axis and the coordinates of the center point of the rotation axis; Determine the coordinates of the center point of the lower arm axis according to the coordinates of the center point of the rotation axis and the length of the support arm; determine the swing angle of the lower arm axis according to the angle between the line connecting the coordinates of the center point of the lower arm axis and the coordinates of the center point of the wrist swing axis and the horizontal plane; Determine the swing angle of the upper arm axis according to the distance between the coordinates of the center point of the lower arm axis and the coordinates of the center point of the wrist swing axis, as well as the length of the lower arm and the length of the upper arm; Determine the coordinates of the center point of the upper arm axis based on the coordinates of the center point of the lower arm axis and the swing angle of the lower arm axis; determine the direction vector of the upper arm based on the coordinates of the center point of the upper arm axis and the coordinates of the center point of the wrist swing axis; determine the rotation angle of the wrist rotation axis and the swing angle of the wrist swing axis based on the direction vector of the upper arm and the direction vector of the wrist.

10. A robotic arm grasping posture determination system, characterized in that: include: Feature line extraction module: collects multi-angle images of the object to be grasped, inputs them into the neural network to extract the feature lines of the object to be grasped; Contour line generation module: selects the optimal feature line according to the line feature optimization model to construct the overall structural contour line of the object to be grasped; Grasping surface generation module: performs morphological judgment on the plane figure formed by the overall structure contour line to obtain an initial grasping surface set; performs roughness judgment on the plane figures in the initial grasping surface set, and selects the optimal grasping plane and the corresponding grasping center point; Posture determination module: Determine the posture of the robot arm wrist and the coordinates of the center point of the wrist swing axis based on the grasping center point and the normal vector of the optimal grasping plane; determine the grasping posture of the robot arm through inverse solution based on the coordinates of the center point of the wrist swing axis and the coordinates of the center point of the rotation axis of the robot arm.

Images