Method and robot system for achieving feeding and discharging of special-shaped parts on coating production line

Abstract

The invention relates to a method for achieving feeding and discharging of special-shaped parts of a coating production line and a robot system, and belongs to the technical field of industrial robots, and the method for achieving feeding and discharging of the special-shaped parts of the coating production line comprises the steps that scene point cloud of the coated special-shaped parts in three-dimensional point cloud of a material box is obtained; converting the constructed three-dimensional model of the coated special-shaped part to obtain a model point cloud of the coated special-shaped part; performing coarse registration and fine registration on the scene point cloud and the model point cloud in sequence, planning a feeding and discharging motion track of the robot by adopting a 3-5-3 combined piecewise polynomial interpolation function so as to construct a robot joint space track model, and taking motion time as an optimization variable so as to construct a robot joint space track model; the comprehensive optimization of the loading and unloading time, impact and dexterity of the robot is taken as a target function, and the particle swarm whale combinatorial optimization algorithm is adopted to optimize the trajectory parameters of the robot joint space trajectory model so as to obtain the optimal loading and unloading trajectory of the robot, so that the loading and unloading takt time of the robot is shortened, and the production efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial robots, and in particular to a method and a robot system for loading and unloading special-shaped parts on a coating production line. Background Art

[0002] In a coating production line with a multi-variety, small-batch production model, the coating line often needs to process special-shaped workpieces of different specifications, such as automobile chassis brackets, sheet metal parts, etc. Traditional loading and unloading mostly relies on manual operation or fixed programmed robots.

[0003] Traditional robots need to be reprogrammed for each workpiece, resulting in long switching times and poor adaptability to multiple varieties; most incoming materials are randomly stacked special-shaped parts. The random stacking of special-shaped parts with surface reflections and coating interference makes it difficult for the template matching-based vision system to accurately identify the workpiece posture, and the recognition success rate is low in disordered stacking scenarios; the irregular features of painted special-shaped parts can easily lead to inaccurate grasping posture. The existing system relies on offline planning and cannot handle dynamic scene changes in real time, resulting in long loading and unloading cycles.

[0004] Therefore, on the coating production line, traditional robots are used to load and unload special-shaped parts, resulting in a low success rate in identifying special-shaped parts and inaccurate grasping posture, which in turn leads to a long loading and unloading cycle for special-shaped parts. Summary of the Invention

[0005] In view of this, it is necessary to provide a method and a robot system for loading and unloading special-shaped parts on a coating production line, so as to solve the technical problem that the loading and unloading cycle time of traditional robots on the coating production line is long.

[0006] In order to solve the above problems, in a first aspect, the present invention provides a method for loading and unloading special-shaped parts on a coating production line, comprising: Obtain the scene point cloud of the painted special-shaped part in the three-dimensional point cloud of the material box, and transform the constructed three-dimensional model of the painted special-shaped part to obtain the model point cloud of the painted special-shaped part; The SVD-ICP algorithm is used to perform coarse registration and fine registration on the scene point cloud and the model point cloud in sequence to obtain the centroid pose of the painted special-shaped part; Based on the center of mass posture of the painted special-shaped parts, a 3-5-3 combined piecewise polynomial interpolation function is used to plan the initial acceleration segment, the intermediate uniform speed segment, and the terminal deceleration segment of the robot loading and unloading motion trajectory, so as to construct a robot joint space trajectory model. Based on the robot joint space trajectory model, the motion time of each joint of the robot is determined; Taking the motion time as an optimization variable and taking the time, impact and dexterity of the robot loading and unloading as a target function, a particle swarm whale combination optimization algorithm is used to optimize the trajectory parameters of the robot joint space trajectory model to obtain an optimal trajectory of the robot loading and unloading, and the robot is controlled to load and unload based on the optimal trajectory.

[0007] In a possible implementation, the obtaining of the scene point cloud of the coating special-shaped part in the three-dimensional point cloud of the material box comprises: The three-dimensional point cloud of the material box is obtained, and a straight-through filtering algorithm is used to screen the three-dimensional point cloud of the material box. An improved RANSAC plane segmentation algorithm based on normal constraint is used to segment the three-dimensional point cloud of the screened material box, wherein the normal of each point in the three-dimensional point cloud of the screened material box is determined, a normal consistency angle threshold is set, three-point combinations with a normal direction deviation less than the normal consistency angle threshold in the three-dimensional point cloud of the material box are selected to generate an initial plane, a geometric error and a normal angle error double threshold criterion is used to determine the point cloud of the initial plane, and the segmented three-dimensional point cloud of the material box is obtained based on the point cloud of the initial plane. A voxel downsampling method is used to downsample the segmented three-dimensional point cloud of the material box to obtain a coating line scene point cloud. A DBSCAN density clustering algorithm is used to cluster the coating line scene point cloud to extract the scene point cloud of the coating special-shaped part.

[0008] In a possible implementation, the SVD-ICP algorithm is used to sequentially perform coarse registration and fine registration on the scene point cloud and the model point cloud to obtain the centroid pose of the coating special-shaped part, comprising: Point pair feature matching is performed on the model point cloud of the coating special-shaped part and the scene point cloud of the coating special-shaped part to obtain a point pair set, wherein the point pair feature comprises an Euclidean distance, a normal, a normal angle and a rotation angle. A covariance matrix is constructed based on the point pair set, an SVD algorithm is used to decompose the covariance matrix to obtain an initial transformation matrix of the model point cloud of the coating special-shaped part to the scene point cloud of the coating special-shaped part to complete the coarse registration of the scene point cloud and the model point cloud. An ICP algorithm is used to fine register the initial transformation matrix to determine the centroid pose of the model point cloud of the coating special-shaped part after fine registration, and the centroid pose of the model point cloud of the coating special-shaped part is converted through a camera-to-robot hand-eye calibration matrix to obtain the centroid pose of the coating special-shaped part, wherein the centroid pose comprises a position and an Euler angle.

[0009] In a possible implementation, based on the centroid posture of the coated profiled part, a 3-5-3 combined piecewise polynomial interpolation function is used to plan a start acceleration section, a middle constant speed section and an end deceleration section of a robot feeding and discharging motion trajectory respectively to construct a robot joint space trajectory model, including: determining coordinates of a start point, a middle point and an end point of the robot based on the centroid posture of the coated profiled part; based on the coordinates of the start point, the middle point and the end point of the robot, a third-order polynomial, a fifth-order polynomial and a third-order polynomial of a 3-5-3 combined piecewise polynomial interpolation function are used to plan a start acceleration section, a middle constant speed section and an end deceleration section of a robot feeding and discharging motion trajectory respectively to construct a robot joint space trajectory model, wherein the robot feeding and discharging motion trajectory includes a motion trajectory of the robot in the start acceleration section using the third-order polynomial to realize zero impact start and stop, a motion trajectory of the robot in the middle constant speed section using the fifth-order polynomial to maintain constant linear speed, and a motion trajectory of the robot in the end deceleration section using the third-order polynomial to ensure that the end acceleration is zero.

[0010] In a possible implementation, the robot joint space trajectory model is: , wherein, , , are angles of a first joint of the robot in the start acceleration section, the middle constant speed section and the end deceleration section trajectory respectively, , , , are time variables of the first-order polynomial, the second-order polynomial and the third-order polynomial of the robot in the start acceleration section respectively, , , , are polynomial coefficients of the start acceleration section, , , , , , are polynomial coefficients of the middle constant speed section; , , , , are time variables of the fifth-order polynomial, the fourth-order polynomial, the third-order polynomial, the second-order polynomial and the first-order polynomial of the robot in the middle constant speed section respectively, , , , are polynomial coefficients of the end deceleration section, , , are time variables of the third-order polynomial, the second-order polynomial and the first-order polynomial of the terminal deceleration section, respectively.

[0011] In a possible implementation, the trajectory parameters of the robot joint space trajectory model include running times of the robot in the initial acceleration section, the intermediate constant speed section and the terminal deceleration section; the trajectory parameters of the robot joint space trajectory model are optimized by using a particle swarm whale combined optimization algorithm with the motion times as optimization variables and with a comprehensive optimal time, impact and dexterity of robot loading and unloading as a target function, so as to obtain an optimal trajectory of robot loading and unloading, including: constructing population particles of the particle swarm optimization algorithm based on the running times; initializing the population particles to obtain particle velocities and particle positions; updating the particle velocities and the particle positions by using the particle swarm optimization algorithm to obtain individual optimal solutions of the particles and a global optimal solution of the population particles; taking the individual optimal solutions of the particles and the global optimal solution of the population particles as reference targets of a whale optimization algorithm, updating whale positions by using a contraction and enclosure mechanism, a spiral swimming mechanism and a random search mechanism of the whale optimization algorithm to obtain optimal running times, and obtaining an optimal trajectory of robot loading and unloading based on the optimal running times.

[0012] In a possible implementation, the target function is: , , wherein, is a total time of robot loading and unloading, is a total impact of robot loading and unloading, is a global dexterity average of robot loading and unloading, is a trajectory running time of the initial acceleration section, , , are accelerations of the first, the second and the third joints of the robot in the initial acceleration section, the intermediate constant speed section and the terminal deceleration section, respectively. is a number of interpolation points of the robot loading and unloading trajectory, is a dexterity of the i th interpolation point, is a Jacobian matrix of the interpolation point, , , , are weight coefficients of the total time, the total impact and the global dexterity average of the robot loading and unloading trajectory. ​

[0013] In a possible implementation, the particle velocity is updated as: , wherein, is the velocity of a particle in the first iteration, is the velocity of a particle in the second iteration, is an inertia weight, is an individual learning factor, is a social learning factor, , is a random number, is an individual optimal solution of the current particle, is a global optimal solution of the population, , is a current position of the particle; The contraction enclosure mechanism updates the whale position as: , , , wherein, is a current iteration number, is a whale position, is an optimal whale position, , is a coefficient, is a distance between the current whale position and the optimal position; The spiral swimming mechanism updates the whale position as: , wherein, is a random number, is a distance between the current whale position and the optimal position under the spiral swimming mechanism, is a logarithmic spiral shape constant, is a random number; The random search mechanism updates the whale position as: , , wherein, is a randomly selected whale individual position in the population, is a distance from the randomly selected whale individual to the prey.

[0014] In a possible implementation, the robot feeding and unloading based on the optimal trajectory comprises: ​​determine a center of mass pose of the coating special-shaped part and a start time, a duration of the robot in the three-segment trajectory based on the optimal trajectory; determine a target position and pose of the robot based on the center of mass pose of the coating special-shaped part and the start time, the duration of the robot in the three-segment trajectory, and control the robot feeding and discharging by a TwinCAT continuous motion mode based on the target position and pose of the robot.

[0015] In a second aspect, the application further provides a robot system for feeding and discharging special-shaped parts in a coating production line, comprising: a special-shaped part point cloud obtaining module configured to obtain scene point cloud of the coating special-shaped part in three-dimensional point cloud of a magazine, and convert a constructed three-dimensional model of the coating special-shaped part to obtain model point cloud of the coating special-shaped part; a point cloud registration module configured to sequentially perform coarse registration and fine registration on the scene point cloud and the model point cloud by using an SVD-ICP algorithm, and obtain the center of mass pose of the coating special-shaped part; a trajectory model constructing module configured to plan start acceleration segment, middle uniform speed segment and end deceleration segment of a robot feeding and discharging motion trajectory by using a 3-5-3 combined segmented polynomial interpolation function based on the center of mass pose of the coating special-shaped part, construct a robot joint space trajectory model, and determine motion time of each joint of the robot based on the robot joint space trajectory model; a trajectory optimization module configured to take the motion time as an optimization variable, take time, impact and dexterity comprehensive optimization of the robot feeding and discharging as a target function, optimize trajectory parameters of the robot joint space trajectory model by using a particle swarm whale combined optimization algorithm, obtain optimal trajectory of the robot feeding and discharging, and control the robot feeding and discharging based on the optimal trajectory.

[0016] The application has the beneficial effects that: the SVD-ICP algorithm is used to sequentially perform coarse registration and fine registration on the scene point cloud and the model point cloud, and the center of mass pose of the coating special-shaped part is obtained, high-precision pose estimation of the coating special-shaped part is realized; the 3-5-3 combined segmented polynomial interpolation function is used to plan start acceleration segment, middle uniform speed segment and end deceleration segment of a robot feeding and discharging motion trajectory based on the center of mass pose of the coating special-shaped part, a robot joint space trajectory model is constructed, the motion time is taken as an optimization variable, time, impact and dexterity comprehensive optimization of the robot feeding and discharging are taken as a target function, the particle swarm whale combined optimization algorithm is used to optimize trajectory parameters of the robot joint space trajectory model, the optimal trajectory of the robot feeding and discharging is obtained, high-precision identification, pose updating of the coating line special-shaped part and dynamic adjustment and optimization control of the robot trajectory are realized, the overall feeding and discharging beat time of the robot is shortened, and the production efficiency and coating quality are improved. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description only constitute some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without any creative effort.

[0018] Figure 1 An embodiment flow chart of the method for realizing the loading and unloading of special-shaped parts in a coating production line provided by the present application; Figure 2 A robot loading and unloading experiment table schematic diagram of the method for realizing the loading and unloading of special-shaped parts in a coating production line provided by the present application; Figure 3 A coating special-shaped part scene point cloud processing result schematic diagram of the method for realizing the loading and unloading of special-shaped parts in a coating production line provided by the present application; Figure 4 A coating special-shaped part model point cloud processing result schematic diagram of the method for realizing the loading and unloading of special-shaped parts in a coating production line provided by the present application; Figure 5 A coating special-shaped part scene point cloud and coating special-shaped part model point cloud registration schematic diagram of the method for realizing the loading and unloading of special-shaped parts in a coating production line provided by the present application; Figure 6 A robot loading and unloading process schematic diagram of the method for realizing the loading and unloading of special-shaped parts in a coating production line provided by the present application; Figure 7 A structure schematic diagram of an embodiment of the robot system for realizing the loading and unloading of special-shaped parts in a coating production line provided by the present application. DETAILED DESCRIPTION

[0019] The preferred embodiments of the present application will be described in detail below with reference to the drawings, which form a part of this application. The drawings and the embodiments together illustrate the principles of the present application, but are not intended to limit the scope of the present application.

[0020] Reference herein to "embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive of one another. It is explicitly contemplated that embodiments described herein can be combined with other embodiments.

[0021] Before the embodiments are described, the following terms are explained.

[0022] SVD algorithm: (Singular Value Decomposition) is a mathematical method in linear algebra that decomposes any real or complex matrix.

[0023] ICP algorithm: (Iterative Closest Point) is a core method for point cloud registration in computer vision and three-dimensional reconstruction.

[0024] PSO optimization algorithm: (Particle Swarm Optimization) is a heuristic optimization algorithm based on swarm intelligence, which simulates the cooperative work of groups in searching for food.

[0025] WOA optimization algorithm: (Whale Optimization Algorithm) is a bionic heuristic algorithm derived from the simulation of the behavior of a group of humpback whales hunting small fish and shrimp, which can optimize the robot loading and unloading trajectory.

[0026] One embodiment of the present application discloses a method for implementing the loading and unloading of special-shaped parts in a coating production line. Figure 1 As shown in the figure, the method for implementing the loading and unloading of special-shaped parts in a coating production line comprises: S101, obtaining the scene point cloud of the coating special-shaped part in the three-dimensional point cloud of the magazine, and transforming the constructed three-dimensional model of the coating special-shaped part to obtain the model point cloud of the coating special-shaped part; It should be noted that the model point cloud of the coating special-shaped part obtained by three-dimensional modeling is used for template matching with the scene point cloud of the coating special-shaped part.

[0027] S102, using SVD-ICP algorithm to sequentially perform coarse registration and fine registration on the scene point cloud and the model point cloud, and obtaining the centroid pose of the coating special-shaped part; It should be noted that after using SVD and ICP algorithms to complete coarse registration and fine registration on the scene point cloud and the model point cloud of the coating special-shaped part, the centroid pose of the model point cloud after registration is calculated, and the centroid pose of the coating special-shaped part is obtained by converting the camera-to-robot hand-eye calibration matrix into the coordinates in the robot coordinate system, thereby realizing high-precision pose estimation of the coating special-shaped part.

[0028] S103, based on the centroid pose of the coating special-shaped part, using a 3-5-3 combined segmented polynomial interpolation function to plan the start acceleration segment, the middle constant speed segment and the end deceleration segment of the robot loading and unloading motion trajectory, respectively, to construct a robot joint space trajectory model, and determining the motion time of each joint of the robot based on the robot joint space trajectory model. It should be noted that the 3-5-3 combination piecewise polynomial interpolation function is used to plan the robot feeding and discharging motion trajectory to construct the robot joint space trajectory model, ensuring the continuity of the feeding and discharging motion.

[0029] In S104, the motion time is taken as an optimization variable, the time, impact and dexterity of the robot feeding and discharging are taken as a target function, the particle swarm whale combination optimization algorithm is used to optimize the trajectory parameters of the robot joint space trajectory model to obtain the optimal trajectory of the robot feeding and discharging, and the robot feeding and discharging is controlled based on the optimal trajectory. It should be noted that the PSO-WOA hybrid optimization algorithm is used to optimize the robot feeding and discharging trajectory, realizing high-precision identification, pose updating and dynamic adjustment and optimization control of the robot trajectory of the special-shaped part of the coating line, shortening the overall feeding and discharging beat time, and improving the production efficiency and coating quality.

[0030] In some embodiments, in step S101, the scene point cloud of the coating special-shaped part in the bin three-dimensional point cloud is obtained by calling the video stream of the RGB-D camera to obtain the color image and the depth image of the coating scene, and the coating scene is a scene in which a plurality of special-shaped parts are placed in the bin. After obtaining the color image and the depth image of the coating scene, the bin three-dimensional point cloud is generated using the Open3D library, and the straight-through filtering algorithm is used to screen the bin three-dimensional point cloud. An improved RANSAC plane segmentation algorithm based on normal constraint is used to segment the screened bin three-dimensional point cloud. The process of the improved RANSAC plane segmentation algorithm based on normal constraint is to determine the normal vector of each point in the screened bin three-dimensional point cloud, set a normal consistency angle threshold, select three points in the bin three-dimensional point cloud whose normal direction deviation is less than the normal consistency angle threshold to generate an initial plane, determine the point cloud of the initial plane using a geometric error and a normal angle error double threshold criterion, obtain the segmented bin three-dimensional point cloud based on the point cloud of the initial plane, and downsample the segmented bin three-dimensional point cloud using a voxel downsampling method to obtain the coating line scene point cloud. The DBSCAN density clustering algorithm is used to cluster the coating line scene point cloud to extract the scene point cloud of the coating special-shaped part. The scene point cloud of the coating special-shaped part is preprocessed by extracting the region of interest, segmenting the bin plane, and downsampling the point cloud, and then the DBSCAN density clustering algorithm is used for point cloud segmentation. In the preprocessing stage, the space is first screened using the straight-through filtering algorithm, the scene point cloud is screened according to the possible spatial range of the coating special-shaped part, and the point cloud outside the range is removed, then the improved RANSAC plane segmentation algorithm based on normal constraint is used for plane segmentation to remove large-area planes in the coating scene, such as the ground, the wall, the conveyor belt, etc., and finally the point cloud is downsampled using the voxel downsampling method to reduce the number of point clouds and obtain the coating line scene point cloud, thereby improving the calculation efficiency. The DBSCAN density clustering algorithm is used to cluster the coating line scene point cloud obtained after preprocessing to extract the scene point cloud of the coating special-shaped part. The schematic diagram of the robot loading and unloading test bed is shown in Figure 2 As shown in Figure 2 , the robot loading and unloading test bed includes a robot, a depth camera, a suction cup, a coating special-shaped part, a bin, a hanger, and a hook. The robot places the coating special-shaped part in the bin on the hook or places the special-shaped part on the hook in the bin through the suction cup to realize the robot loading and unloading on the coating production line. The schematic diagram of the coating special-shaped part scene point cloud processing result is shown in Figure 3 As shown in Figure 3As shown, the length, width, and height of the ROI of the magazine area are extracted by straight-through filtering, the plane segmentation is performed by using the random sample consensus algorithm based on normal constraint, the conveyor belt plane noise point cloud is removed, the data density is reduced by voxel grid downsampling (resolution 2mm), the denoised sparse scene point cloud is formed, the DBSCAN density clustering algorithm (parameters eps=4, min_points=6) is used to realize the instance-level segmentation of the special-shaped workpiece under the complex working condition of the stacked workpiece, and the independent workpiece instance point cloud, i.e. the special-shaped workpiece scene point cloud, is output.

[0031] The model point cloud of the coated special-shaped workpiece is obtained by transforming the constructed three-dimensional model of the coated special-shaped workpiece, the three-dimensional model of the coated special-shaped workpiece is established by using Solidworks, the three-dimensional model of the coated special-shaped workpiece is transformed into the model point cloud of the coated special-shaped workpiece by using the Open3D library, the point pair features of the model point cloud of the coated special-shaped workpiece are calculated, and the point pair features are saved as a hash table, and a schematic diagram of the model point cloud processing result of the coated special-shaped workpiece is shown in Figure 4 The segmented multiple scene point clouds of the coated special-shaped workpiece correspond to the model point cloud of the coated special-shaped workpiece respectively.

[0032] In some embodiments, in step S102, the SVD-ICP algorithm is used to sequentially perform coarse registration and fine registration on the scene point cloud and the model point cloud to obtain the centroid pose of the painted special-shaped part, that is, the SVD-ICP algorithm is used to perform coarse registration and fine registration on the model point cloud and the scene point cloud, and the model point cloud of the painted special-shaped part and the scene point cloud of the painted special-shaped part are matched by point pairs, to obtain a set of point pairs, wherein the point pair features include Euclidean distance, normal, normal angle, rotation angle, the point pair features are a one-dimensional vector of 1x4, including the distance of the point pair, the angle between the two normals and the connecting line of the point pair, the angle between the two normals, and the rotation angle, the point pair features of the scene point cloud of the painted special-shaped part are calculated, the calculated point pair features of the scene point cloud of the painted special-shaped part are matched with the hash table of the model point cloud of the painted special-shaped part, for each point pair of the painted scene special-shaped part, the model point pair with the highest vote is selected, and the one-to-one correspondence between the painted scene special-shaped part point cloud and the model point cloud is completed, that is, the point pair feature matching of the model point cloud of the painted special-shaped part and the scene point cloud of the painted special-shaped part is completed, a covariance matrix is constructed based on the set of point pairs, the SVD algorithm is used to decompose the covariance matrix to obtain an initial transformation matrix of the model point cloud of the painted special-shaped part to the scene point cloud of the painted special-shaped part, the rotation matrix and the translation vector are obtained by decomposing the covariance matrix by SVD, so as to transform the scene point cloud to the coordinate system of the model point cloud, and preliminary registration is realized; the ICP algorithm is used to fine register the initial transformation matrix to determine the centroid pose of the model point cloud of the painted special-shaped part after fine registration, and the centroid pose of the model point cloud of the painted special-shaped part is converted by the camera-to-robot hand-eye calibration matrix to obtain the centroid pose of the painted special-shaped part, wherein the centroid pose includes position and Euler angle, the process of fine registering the initial transformation matrix by the ICP algorithm is that the ICP algorithm constantly finds the nearest point correspondence between the scene point cloud and the model point cloud, and calculates the optimal transformation matrix to minimize the distance between the point clouds until the convergence condition is met, in each iteration, the position of the scene point cloud is updated according to the current transformation matrix, and the nearest point correspondence is recalculated until the change in the distance between the point clouds is less than a set threshold or the maximum number of iterations is reached, finally, the matching degree of the registered point cloud and the model point cloud is evaluated, if the matching degree is high, it is considered that the identification and positioning of the painting line special-shaped part are successfully completed; if the matching degree is low, it may be necessary to adjust the parameters or check whether there is an error in the previous processing process, and a schematic diagram of the registration of the painted special-shaped part scene point cloud and the painted special-shaped part model point cloud is shown in Figure 5 .

[0033] After the SVD-ICP algorithm is used to complete coarse registration and fine registration on the painted special-shaped part scene point cloud and the model point cloud, the centroid pose (6DOF pose) of the model point cloud after registration is calculated, which is converted into the coordinates in the robot coordinate system through the camera-to-robot hand-eye calibration matrix, so as to obtain the centroid pose of the painted special-shaped part, and the centroid pose of the painted special-shaped part is used for subsequent trajectory planning.

[0034] In some embodiments, in step S103, based on the centroid posture of the coated special-shaped part, a 3-5-3 combined segmented polynomial interpolation function is used to plan the starting acceleration segment, the middle constant speed segment and the terminal deceleration segment of the robot feeding and discharging motion trajectory respectively to construct a robot joint space trajectory model, the coordinates of the starting point, the middle point and the terminal point of the robot are determined based on the centroid posture of the coated special-shaped part, the third order polynomial, the fifth order polynomial and the third order polynomial of the 3-5-3 combined segmented polynomial interpolation function are used to plan the starting acceleration segment, the middle constant speed segment and the terminal deceleration segment of the robot feeding and discharging motion trajectory respectively based on the coordinates of the starting point, the middle point and the terminal point of the robot to construct a robot joint space trajectory model, a three-segment motion trajectory of the robot feeding and discharging is determined based on the robot joint space trajectory model, and the motion time of each joint of the robot in the three-segment motion trajectory is determined; wherein the robot feeding and discharging motion trajectory (three-segment motion trajectory) includes the motion trajectory of the robot in the starting acceleration segment (0-T1) using the 3rd order polynomial to realize zero impact start-stop, the motion trajectory in the middle constant speed segment (T1-T2) using the 5th order polynomial to maintain constant linear speed, and the motion trajectory in the terminal deceleration segment (T2-T) using the 3rd order polynomial to ensure that the terminal acceleration is zero, and the robot joint space trajectory model is: , wherein, 、 、 are the angles of the first, joint of the robot in the starting acceleration segment, the middle constant speed segment and the terminal deceleration segment trajectory respectively, 、 、 are the time variables of the first order polynomial, the second order polynomial and the third order polynomial of the robot in the starting acceleration segment, 、 、 、 are the polynomial coefficients of the starting acceleration segment, 、 、 、 、 、 are the polynomial coefficients of the middle constant speed segment; 、 、 、 、 are the time variables of the fifth order polynomial, the fourth order polynomial, the third order polynomial, the second order polynomial and the first order polynomial of the robot in the middle constant speed segment, 、 、 、 a polynomial coefficient of the end deceleration section, 、 、 are time variables of the third-order polynomial, the second-order polynomial and the first-order polynomial of the end deceleration section, respectively; The three-section trajectory is a 3-5-3 combined segmented polynomial interpolation trajectory. When constructing the 3-5-3 combined segmented polynomial interpolation trajectory curve, the joint angles of the robot at the four interpolation points are obtained through inverse kinematics according to the space coordinates of the starting point, the intermediate point, the intermediate point and the terminal point of the robot in the Cartesian coordinate system, and the robot joint space trajectory model is constructed through the joint angles. The 3-5-3 combined segmented polynomial interpolation method ensures the continuity of the feeding and unloading motion. The 3-5-3 combined segmented polynomial interpolation trajectory planning needs four interpolation nodes, i.e., the starting point, the intermediate point and the terminal point. After completing the pose estimation of the coated special-shaped workpiece in the box, the grasping pose of the coated special-shaped workpiece is taken as the first interpolation node of the robot feeding and unloading trajectory, the fixed image acquisition point above the box is taken as the second interpolation node, the point in front of the hanger is taken as the third interpolation node, and the workpiece placement point is taken as the fourth interpolation node. The robot feeding and unloading process diagram is shown in Figure 6 .

[0035] In some embodiments, in step S104, the motion time is taken as the optimization variable, the time, impact and dexterity of the robot feeding and unloading are taken as the objective function, the particle swarm whale combination optimization algorithm is adopted to optimize the trajectory parameters of the robot joint space trajectory model, and the optimal trajectory of the robot feeding and unloading is obtained. Based on the optimal trajectory, the robot feeding and unloading is controlled. As can be seen from the robot joint space trajectory model, the robot feeding and unloading trajectory polynomial coefficients are affected by the running time of the robot in the starting acceleration section, the intermediate uniform speed section and the end deceleration section 、 and Therefore, the running time of the robot in the starting acceleration section, the intermediate uniform speed section and the end deceleration section is taken as the optimization variable; the time, impact and dexterity of the robot feeding and unloading are taken as the objective function, and the objective function is: , , wherein, is the total time of the robot feeding and unloading, which measures the efficiency of the robot feeding and unloading, is the total impact of the robot feeding and unloading, which measures the stability and joint wear of the robot feeding and unloading, The global dexterity average of the robot loading and unloading is used to measure the flexibility of the robot loading and unloading trajectory, and the optimization of the trajectory dexterity can avoid the motion uncertainty caused by the joint limit and singularity in the motion process, and enhance the effectiveness of the loading and unloading working time, For Segment trajectory running time, the first 、 、 The first Joint acceleration of the robot at the start of the acceleration segment, the middle constant speed segment, and the end of the deceleration segment, The number of interpolation points of the robot loading and unloading trajectory, The dexterity of the first Interpolation point, The Jacobian matrix of the interpolation point, 、 、 The total time, total impact, and global dexterity average weight coefficient of the robot loading and unloading trajectory; The particle swarm whale combination optimization algorithm is adopted for optimization, which includes the PSO optimization algorithm and the WOA optimization algorithm, and the WOA optimization algorithm includes the shrinkage surrounding mechanism, the spiral swimming mechanism and the random search mechanism. The process of using the PSO-WOA hybrid optimization algorithm to optimize the trajectory parameters of the robot joint space trajectory model is as follows: based on the running time, the population particles of the PSO optimization algorithm are constructed, and the population particles are initialized to obtain the particle velocity and the particle position. The particle velocity and the particle position are updated using the PSO optimization algorithm to obtain the individual optimal solution of the particle and the global optimal solution of the population particle. The individual optimal solution of the particle and the global optimal solution of the population particle are used as the reference target of the WOA optimization algorithm, and the whale position is updated through the shrinkage surrounding mechanism, the spiral swimming mechanism and the random search mechanism to obtain the optimal running time. Based on the optimal running time, the optimal trajectory of the robot loading and unloading is obtained. The particle swarm optimization algorithm (PSO) optimizes the robot loading and unloading trajectory by simulating the cooperative work of the group in the search for food. In the optimization process, the position of each particle is the solution of the trajectory optimization problem. The particles exchange information in the solution space, continuously adjust their positions, and gradually approach the optimal solution to realize the search for the optimal solution in the trajectory optimization space. The particle velocity update in the particle swarm optimization algorithm (PSO) is as follows: , Wherein, The velocity of the particle In the first Iteration, The velocity of the particle In the first Iteration, The inertia weight, is the individual learning factor, is the social learning factor, 、 for A random number in the range, is the individual optimal solution of the current particle, is the global optimal solution of the population, is the current position of the particle; The particle position is updated as: , in, For particles In the The position in the iteration For particles In the The position in the iteration; The optimal solution is obtained by updating the particle position and velocity. When the optimal solution is iterated for five consecutive times and the improvement in the position and velocity of the optimal solution is less than 1%, the spiral swimming mechanism in the WOA optimization algorithm is switched to. The individual optimal solution of the particle and the global optimal solution of the swarm of particles are used as the reference targets of the WOA optimization algorithm to dynamically adjust the position update direction of the whale. WOA (Whale Optimization Algorithm) is a bionic metaheuristic algorithm derived from the simulation of the behavior of humpback whale groups preying on small fish and shrimp. It can optimize the robot loading and unloading trajectory. During the optimization process, the position of each whale is the solution to the trajectory optimization problem. The whale continuously adjusts its position and preys on prey to search for the optimal solution to the trajectory optimization problem. In the shrinking and encircling mechanism stage, the whale moves towards the global optimal solution. The shrinking and encircling mechanism updates the whale's position as follows: , , , , , in, is the current iteration number, For the whale position, For the optimal whale position, 、 is the coefficient, is the distance between the current whale position and the optimal position, The value range is A random number, is the maximum number of iterations; In the spiral mechanism stage, the whale can update its position by contracting the surrounding and spiral mechanism with a probability of 50%, and the spiral mechanism updates the position of the whale as: , wherein, is a random number with a value range of between is the distance between the current position of the whale and the optimal position under the spiral mechanism, is a constant of logarithmic spiral shape, is a random number; In the random search mechanism stage, the random search mechanism updates the position of the whale as: , , wherein, is the position of a randomly selected whale individual in the population, is the distance from the randomly selected whale individual to the prey; The PSO-WOA hybrid optimization algorithm outputs the optimal solution set after 200 iterations, i.e., obtains the optimal running time of the robot in the starting acceleration segment, the intermediate uniform speed segment, and the terminal deceleration segment. The optimal running time is brought into the robot joint space trajectory model to obtain the position, speed, acceleration, and jerk of each joint of the robot, which is used for robot motion control to complete the trajectory optimization of the robot loading and unloading and obtain the optimal trajectory of the robot loading and unloading.

[0036] The PSO-WOA hybrid optimization algorithm combines the local search ability of PSO with the global search ability of WOA, ensuring global search while improving the accuracy of local search, and converging faster and more accurately in solving optimization problems. The PSO optimization algorithm is applied to the development stage (spiral mechanism stage) of WOA, improving the ability of the algorithm to search for global optimal solutions. The parameters of PSO include a population size N of 50, a maximum number of iterations of 100, an individual learning factor of 1.5, a social learning factor of 1.5, and an inertia weight of 0.8. The parameters of WOA include a population size M of 50 and a maximum number of iterations of 100. After obtaining the optimal trajectory of the robot loading and unloading, the robot loading and unloading is controlled based on the optimal trajectory of the robot loading and unloading, the centroid posture of the coated special-shaped part and the starting time and duration of the robot in the three-section trajectory are determined based on the optimal trajectory, the target position and posture of the robot are determined based on the centroid posture of the coated special-shaped part and the starting time and duration of the robot in the three-section trajectory, the robot loading and unloading is controlled through the TwinCAT continuous motion mode based on the target position and posture of the robot, the ADS communication protocol is configured on the TwinCAT 3.1 platform, the connection parameters of the server and the client are set to ensure stable communication with the trajectory planning module, according to the actual network environment, the IP address, port number and AMS address and other key information are specified to establish an efficient data transmission channel, the received robot grasping pose and three-section trajectory running time are further analyzed to extract the starting time, duration and 6DOF pose position (x, y, z) and Euler angle (roll, pitch, yaw) information of each section of the trajectory, the pose matrix is converted into the target position and posture of the robot joint space, and preparation is made for subsequent trajectory generation and motion control.

[0037] During the movement of the robot end effector, the position information is monitored in real time, when it moves to the intermediate node 1, the preset hardware interrupt signal is triggered through the logical judgment function of TwinCAT 3.1, the interrupt signal is sent to the RGB-D camera through the EtherCAT master station, the point cloud collection process of the workpiece pose estimation module is started, the updated pose of the next coated special-shaped part to be grasped is fed back to the robot in real time, after receiving the updated pose of the coated special-shaped part, the next trajectory pre-planning process based on PSO-WOA is triggered immediately, according to the new pose information, the trajectory of the next loading and unloading is calculated in advance, and the calculation result is sent back to the TwinCAT 3.1 platform, realizing high-precision identification, pose updating, dynamic adjustment and optimization control of the robot trajectory of the coated special-shaped part, shortening the overall loading and unloading cycle, improving the production efficiency and coating quality.

[0038] The high-precision pose estimation of the coated special-shaped metal part is realized by the Obi mid-light Gemini Pro RGB-D camera and the improved SVD-ICP optimization algorithm, the optimal motion trajectory is generated by combining the 3-5-3 segmented polynomial trajectory model and the PSO-WOA hybrid optimization algorithm, the six-axis industrial robot loading and unloading time is shortened from 4.5s to 3.2s (efficiency is improved by 28%), the dynamic re-planning system is built based on the TwinCAT 3.1 platform, the real-time pose update and the trajectory parameter pre-transmission of the ADS protocol are triggered by the EtherCAT hardware interrupt in the uniform speed segment, the seamless trajectory switching and the second optimization of the beat (from 3.8s to 3.2s) are realized, the robot loading and unloading time is shortened, and the robot working efficiency is improved.

[0039] In summary, the method for realizing the loading and unloading of the special-shaped part of the coating production line provided by the present application obtains the scene point cloud of the coated special-shaped part in the three-dimensional point cloud of the material box, and converts the three-dimensional model of the coated special-shaped part to obtain the model point cloud of the coated special-shaped part; the scene point cloud and the model point cloud are sequentially coarsely registered and finely registered by using the SVD-ICP algorithm to obtain the centroid attitude of the coated special-shaped part; based on the centroid attitude of the coated special-shaped part, a 3-5-3 combined segmented polynomial interpolation function is used to plan the starting acceleration segment, the middle uniform speed segment and the terminal deceleration segment of the robot loading and unloading motion trajectory respectively to construct a robot joint space trajectory model, and the motion time of each joint of the robot is determined based on the robot joint space trajectory model; the motion time is used as an optimization variable, the time, impact and dexterity of the robot loading and unloading are comprehensively optimized as a target function, and the particle swarm whale combined optimization algorithm is used to optimize the trajectory parameters of the robot joint space trajectory model to obtain the optimal trajectory of the robot loading and unloading, so that the robot loading and unloading time is shortened and the production efficiency is improved.

[0040] In order to better implement the method for realizing the loading and unloading of the special-shaped part of the coating production line in the embodiment of the present application, on the basis of the method for realizing the loading and unloading of the special-shaped part of the coating production line, as shown in Figure 7 The robot system for realizing the loading and unloading of the special-shaped part of the coating production line provided by the present application embodiment comprises: The special-shaped part point cloud obtaining module 701 is used to obtain the scene point cloud of the coated special-shaped part in the three-dimensional point cloud of the material box, and convert the three-dimensional model of the coated special-shaped part to obtain the model point cloud of the coated special-shaped part; The point cloud registration module 702 is used to sequentially coarsely register and finely register the scene point cloud and the model point cloud by using the SVD-ICP algorithm to obtain the centroid attitude of the coated special-shaped part. The trajectory model construction module 703 is configured to plan a start acceleration section, a middle uniform speed section and a terminal deceleration section of a robot feeding and discharging motion trajectory respectively by using a 3-5-3 combined segmented polynomial interpolation function based on the centroid posture of the coated special-shaped part, to construct a robot joint space trajectory model, and to determine the motion time of each joint of the robot based on the robot joint space trajectory model; The trajectory optimization module 704 is configured to take the motion time as an optimization variable, take the time, impact and dexterity of the robot feeding and discharging as a target function, and optimize the trajectory parameters of the robot joint space trajectory model by using a particle swarm whale combined optimization algorithm, to obtain an optimal trajectory of the robot feeding and discharging, and control the robot feeding and discharging based on the optimal trajectory.

[0041] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for loading and unloading special-shaped parts on a coating production line, characterized in that: include: Obtain the scene point cloud of the painted special-shaped part in the three-dimensional point cloud of the material box, and transform the constructed three-dimensional model of the painted special-shaped part to obtain the model point cloud of the painted special-shaped part; The SVD-ICP algorithm is used to perform coarse registration and fine registration on the scene point cloud and the model point cloud in sequence to obtain the centroid pose of the painted special-shaped part; Based on the center of mass posture of the painted special-shaped parts, a 3-5-3 combined piecewise polynomial interpolation function is used to plan the initial acceleration segment, the intermediate uniform speed segment, and the terminal deceleration segment of the robot loading and unloading motion trajectory, so as to construct a robot joint space trajectory model. Based on the robot joint space trajectory model, the motion time of each joint of the robot is determined; Taking the motion time as the optimization variable and the comprehensive optimization of the time, impact and dexterity of the robot loading and unloading as the objective function, the particle swarm whale combination optimization algorithm is adopted to optimize the trajectory parameters of the robot joint space trajectory model to obtain the optimal trajectory of the robot loading and unloading, and the robot loading and unloading is controlled based on the optimal trajectory.

2. The method for loading and unloading special-shaped parts on a coating production line according to claim 1, characterized in that: The step of obtaining a scene point cloud of a special-shaped part to be painted in a three-dimensional point cloud of a material box includes: Obtaining a three-dimensional point cloud of a material box, and filtering the three-dimensional point cloud of the material box using a straight-through filtering algorithm; An improved RANSAC plane segmentation algorithm based on normal constraints is used to perform plane segmentation on the screened three-dimensional point cloud of the material box, wherein the normal vector of each point in the screened three-dimensional point cloud of the material box is determined, a normal consistency angle threshold is set, and a three-point combination in the three-dimensional point cloud of the material box is selected whose normal direction deviation is less than the normal consistency angle threshold to generate an initial plane. The point cloud of the initial plane is determined using a dual threshold judgment criterion of geometric error and normal angle error, and the segmented three-dimensional point cloud of the material box is obtained based on the point cloud of the initial plane; The voxel downsampling method is used to downsample the segmented 3D point cloud of the material box to obtain the painting line scene point cloud; The DBSCAN density clustering algorithm is used to cluster the painting line scene point cloud to extract the scene point cloud of the painting special-shaped parts.

3. The method for loading and unloading special-shaped parts on a coating production line according to claim 2, characterized in that: The method of using the SVD-ICP algorithm to sequentially perform coarse registration and fine registration on the scene point cloud and the model point cloud to obtain the centroid posture of the painted special-shaped part includes: Performing point pair feature matching on the model point cloud of the painted special-shaped part and the scene point cloud of the painted special-shaped part to obtain a point pair set, wherein the point pair features include Euclidean distance, normal, normal angle, and rotation angle; Constructing a covariance matrix based on the point pair set, decomposing the covariance matrix using an SVD algorithm, and obtaining an initial conversion matrix from the painted special-shaped part model point cloud to the painted special-shaped part scene point cloud, so as to complete a coarse registration between the scene point cloud and the model point cloud; The ICP algorithm is used to precisely align the initial transformation matrix to determine the centroid pose of the model point cloud of the painted special-shaped part after precise alignment. The centroid pose of the model point cloud of the painted special-shaped part is converted through the camera-to-robot hand-eye calibration matrix to obtain the centroid pose of the painted special-shaped part, wherein the centroid pose includes position and Euler angle.

4. The method for loading and unloading special-shaped parts on a coating production line according to claim 3 is characterized in that: Based on the center of mass posture of the painted special-shaped parts, a 3-5-3 combined piecewise polynomial interpolation function is used to respectively plan the initial acceleration segment, the intermediate uniform speed segment, and the terminal deceleration segment of the robot loading and unloading motion trajectory to construct a robot joint space trajectory model, including: Determining the coordinates of the robot's starting point, intermediate point, and end point based on the center of mass posture of the painted special-shaped part; Based on the coordinates of the starting point, middle point and end point of the robot, the third-order polynomial, fifth-order polynomial and third-order polynomial of the 3-5-3 combined piecewise polynomial interpolation function are used to respectively plan the initial acceleration segment, the intermediate uniform speed segment and the terminal deceleration segment of the robot loading and unloading motion trajectory to construct a robot joint space trajectory model, wherein the robot loading and unloading motion trajectory includes the robot using a third-order polynomial in the initial acceleration segment to achieve zero-impact start and stop, the robot using a fifth-order polynomial in the intermediate uniform speed segment to maintain a constant linear velocity, and the robot using a third-order polynomial in the terminal deceleration segment to ensure that the terminal acceleration returns to zero.

5. The method for loading and unloading special-shaped parts on a coating production line according to claim 4 is characterized in that: The robot joint space trajectory model is: ， in, 、 、 The robot The angles of each joint in the trajectory of the initial acceleration segment, the middle uniform speed segment, and the final deceleration segment, 、 、 are the time variables of the first-order polynomial, second-order polynomial, and third-order polynomial of the robot in the initial acceleration section, 、 、 、 are the polynomial coefficients of the initial acceleration segment, 、 、 、 、 、 is the polynomial coefficient of the middle uniform speed segment; 、 、 、 、 are the time variables of the fifth-order polynomial, fourth-order polynomial, third-order polynomial, second-order polynomial, and first-order polynomial of the robot in the middle uniform speed section, 、 、 、 is the polynomial coefficient of the terminal deceleration section, 、 、 are the time variables of the third-order polynomial, second-order polynomial, and first-order polynomial of the terminal deceleration section, respectively.

6. The method for loading and unloading special-shaped parts on a coating production line according to claim 4, characterized in that: The trajectory parameters of the robot joint space trajectory model include the running time of the robot in the initial acceleration section, the intermediate uniform speed section, and the terminal deceleration section; the trajectory parameters of the robot joint space trajectory model are optimized using the particle swarm whale combination optimization algorithm with the movement time as the optimization variable and the comprehensive optimization of the robot loading and unloading time, impact, and dexterity as the objective function to obtain the optimal trajectory of the robot loading and unloading, including: Constructing a swarm of particles of a particle swarm optimization algorithm based on the running time; Initializing the population particles to obtain particle velocity and particle position; The particle swarm optimization algorithm is used to update the particle velocity and particle position to obtain the individual optimal solution of the particle and the global optimal solution of the swarm particles; The individual optimal solution of the particles and the global optimal solution of the swarm particles are used as reference targets of the whale optimization algorithm. The whale position is updated through the shrinking and surrounding mechanism, spiral swimming mechanism and random search mechanism of the whale optimization algorithm to obtain the optimal running time. Based on the optimal running time, the optimal trajectory of the robot loading and unloading is obtained.

7. The method for loading and unloading special-shaped parts on a coating production line according to claim 6, characterized in that: The objective function is: ， ， in, is the total time for loading and unloading the robot, is the total impact of robot loading and unloading, is the global average dexterity of the robot loading and unloading, for Segment trajectory running time, 、 、 The robot The acceleration of each joint in the initial acceleration section, the middle uniform speed section, and the final deceleration section, The interpolation points for the robot loading and unloading trajectory, For the The dexterity of the interpolation points, is the Jacobian matrix of the interpolation point, 、 、 are the weight coefficients of the total time, total impact, and global dexterity mean of the robot's loading and unloading trajectory.

8. The method for loading and unloading special-shaped parts on a coating production line according to claim 6, characterized in that: The particle velocity is updated as: ， in, For particles In the The speed in iterations, For particles In the The speed in iterations, is the inertia weight, is the individual learning factor, is the social learning factor, 、 is a random number, is the individual optimal solution of the current particle, is the global optimal solution of the population, is the current position of the particle; The shrinking and encircling mechanism updates the whale's position as follows: ， ， in, is the current iteration number, For the whale position, For the optimal whale position, 、 is the coefficient, is the distance between the current whale position and the optimal position; The spiral swimming mechanism updates the whale's position as follows: ， in, is a random number, is the distance between the current whale position and the optimal position under the spiral swimming mechanism, is the logarithmic spiral shape constant, is a random number; The random search mechanism updates the whale position as follows: ， ， in, is the position of a randomly selected individual whale in the population, is the distance from a randomly selected whale to its prey.

9. The method for loading and unloading special-shaped parts on a coating production line according to claim 6, characterized in that: The controlling the robot to load and unload materials based on the optimal trajectory includes: Determine the center of mass posture of the painted special-shaped part and the start time and duration of the robot in the three trajectories based on the optimal trajectory; The target position and posture of the robot are determined based on the center of mass posture of the painted special-shaped parts and the start time and duration of the robot in the three trajectories. Based on the target position and posture of the robot, the robot is controlled to load and unload materials through the TwinCAT continuous motion mode.

10. A robot system for loading and unloading special-shaped parts on a coating production line, characterized in that: include: The special-shaped parts point cloud acquisition module is used to obtain the scene point cloud of the painted special-shaped parts in the three-dimensional point cloud of the material box, and convert the constructed three-dimensional model of the painted special-shaped parts to obtain the model point cloud of the painted special-shaped parts; A point cloud registration module is used to perform coarse registration and fine registration on the scene point cloud and the model point cloud in sequence using the SVD-ICP algorithm to obtain the centroid posture of the painted special-shaped part; A trajectory model construction module is used to plan the initial acceleration segment, the intermediate uniform speed segment, and the terminal deceleration segment of the robot loading and unloading motion trajectory based on the center of mass posture of the painted special-shaped part using a 3-5-3 combined piecewise polynomial interpolation function to construct a robot joint space trajectory model, and determine the motion time of each joint of the robot based on the robot joint space trajectory model; The trajectory optimization module is used to optimize the trajectory parameters of the robot joint space trajectory model using the movement time as the optimization variable and the comprehensive optimization of the time, impact, and dexterity of the robot loading and unloading as the objective function, using the particle swarm whale combination optimization algorithm to obtain the optimal trajectory of the robot loading and unloading, and control the robot loading and unloading based on the optimal trajectory.

Images