Electronic assembly control method, system, device and storage medium based on teaching

Through the teaching-based electronic assembly control method, establishing a terminal stiffness model of human upper limbs and combining depth cameras and admission control, the problem of long deployment time and low accuracy of robotic arms in electronic assembly is solved, and assembly quality and stability are improved .

CN118617402BActive Publication Date: 2024-12-06WUHAN UNIV
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202410351621.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-03-26
Publication Date
2024-12-06
Estimated Expiration
2044-03-26

AI Technical Summary

Technical Problem

The existing automated electronic assembly methods have low accuracy and are vulnerable to damage when facing data lines. The robotic arms are deployed for a long time in industrial applications, have high technical thresholds, and lack rigidity control means.

Method used

Using the electronic assembly control method based on teaching, the teaching author is simulated by establishing a mathematical model of end stiffness estimation of human upper limbs, recording the teaching author's action data, generating a trajectory time mapping model and a trajectory stiffness mapping model, combining the depth camera and admission control to simulate the teaching author. Action to achieve variable stiffness control of the robot arm.

Benefits of technology

It reduces the cost of robotic arm deployment time, improves the working quality and stability of electronic assembly, reduces the loss rate during automated assembly, and achieves higher accuracy and safety.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118617402B_ABST
    Figure CN118617402B_ABST
Patent Text Reader

Abstract

The present invention provides an electronic assembly control method, system, device and storage medium based on teaching, including establishing a mathematical model for estimating the stiffness of the end of the upper limb of the human body, the instructor completes the target electronic assembly task, screens qualified assembly data, obtains the instructor's end motion trajectory, instructor's end contact force, instructor's end stiffness corresponding to the target electronic assembly task and establishes a data set, studies and trains the data set, obtains a trajectory time mapping model and a trajectory stiffness mapping model, and the robot arm sequentially realizes rough positioning, fitting of the wiring terminal with the motherboard socket, and insertion of the wiring terminal. The present invention learns human motion skills in electronic assembly to construct a trajectory time mapping model, generates the robot arm end motion trajectory required for the work task, and reduces the deployment time cost of the robot arm in industrial applications; at the same time, a trajectory stiffness mapping model in the electronic assembly process is established to extract the stiffness characteristics in the electronic assembly task process and improve the work quality of electronic assembly.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence and robotic arm technology, and in particular to an electronic assembly control method, system, device and storage medium based on teaching. Background Art

[0002] Electronic assembly is the last stage in the production line of microelectronic products, also known as tertiary assembly. In this step, the plug-in boards or flexible circuit boards of the secondary assembly are connected to the motherboard through interconnection sockets to form a three-dimensional assembly product with higher density, more functions and more complexity. The development of the microelectronics field requires higher and higher automation of electronic assembly: the faster and faster update and iteration of electronic products requires automated electronic assembly to quickly build the assembly process of new products; at the same time, the current automated electronic assembly control methods generally have low accuracy and easy damage to data cables, especially the connection of flat cables.

[0003] In related technologies, automated electronic assembly using robotic arms can speed up assembly. However, current robotic arms have problems in practical industrial applications, such as time-consuming offline programming, high technical barriers, and lack of stiffness control methods. Summary of the invention

[0004] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a teaching-based electronic assembly control method, system, equipment and storage medium. Through teaching learning technology, the human action skills in electronic assembly are learned, a trajectory time mapping model is constructed, and the motion trajectory of the end of the robot arm required for the work task is generated, thereby reducing the deployment time cost of the robot arm in industrial applications; at the same time, a trajectory stiffness mapping model in the electronic assembly process is established to extract the stiffness characteristics in the electronic assembly task process, thereby improving the work quality of the electronic assembly.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] The electronic assembly control method based on teaching includes the following steps:

[0007] Establish a mathematical model for estimating the stiffness of the human upper limb extremities;

[0008] The instructor completes the target electronic assembly task and records the real-time video and corresponding assembly data;

[0009] Filter qualified assembly data, obtain the motion trajectory of the end of the instructor and the contact force of the end of the instructor corresponding to the target electronic assembly operation, and obtain the stiffness of the end of the instructor corresponding to the target electronic assembly operation according to the upper limb end stiffness estimation mathematical model, and generate a data set including the motion trajectory of the end of the instructor, the stiffness of the end of the instructor, and the contact force of the end of the instructor;

[0010] The data set is studied and trained to obtain a trajectory time mapping model and a trajectory stiffness mapping model;

[0011] The depth camera identifies the motherboard socket and obtains the coordinates of the motherboard socket, and the robotic arm moves to the vicinity of the motherboard socket to achieve coarse positioning; according to the position and posture of the motherboard socket and the cable terminal plane transmitted by the depth camera, the robotic arm adjusts the end posture to align the posture and make the two planes fit together; the robotic arm inserts the cable terminal into the motherboard socket according to the trajectory time mapping model and the trajectory stiffness mapping model and combined with the motion adjustment strategy to complete the variable stiffness control of the target electronic assembly operation.

[0012] Furthermore, for the completed electronic assembly parts, assembly data whose quality reaches a preset threshold is selected as valid data in combination with real-time video playback;

[0013] Preprocess the raw electromyographic signals to obtain the muscle activation level sequence;

[0014] Calculate the upper limb Jacobian matrix of the instructor at each time point during the target electronic assembly operation, and output the upper limb Jacobian matrix sequence;

[0015] The muscle activation level sequence and upper limb motion Jacobian matrix sequence in the effective data are input into the human upper limb end stiffness estimation mathematical model, and the end stiffness sequence of the instructor in the target electronic assembly operation process is output;

[0016] The effective data is aligned with the motion trajectory sequence of the end of the instructor and the stiffness sequence of the end of the instructor using a time warping algorithm;

[0017] Each set of assembly data is divided into two sections: free space and contact space according to the magnitude of the end contact force of the instructor;

[0018] Generate a dataset of the motion of the instructor in free space and contact space, including the end motion trajectory sequence, the instructor end stiffness sequence, and the instructor end contact force sequence.

[0019] Furthermore, the motion trajectory of the end of the instructor and the stiffness of the end of the instructor in the target electronic assembly operation are aligned by a time warping algorithm. The specific method is as follows:

[0020] A teaching data in multiple sets of assembly data is used as a reference curve ,in, is the motion trajectory data of the end of the instructor in the reference curve, is the stiffness data of the end of the instructor in the reference curve;

[0021] The remaining teaching data is used as observation curves , is the motion trajectory data of the end of the instructor in the reference curve, is the stiffness data of the end of the instructor in the reference curve;

[0022] The track grid consisting of the reference points on the reference curve and the test points on the test curve is obtained, where the distance between the reference point and the test point in any track grid point is , search for a path with the minimum sum of distance values ​​from the starting track grid point to the ending track grid point, and rearrange with the index on the path to align the reference curve with the test curve.

[0023] Furthermore, the trajectory time mapping model learning and training process specifically includes:

[0024] The first Gaussian mixture model is trained using the motion trajectory and time information in the training set:

[0025] Build the first Gaussian mixture model

[0026] in, , , is the probability density function of each first Gaussian component in the first Gaussian mixture model, is the number of the first Gaussian component in the first Gaussian mixture model in the teaching trajectory training, Indicates The prior probability of the first Gaussian component, Indicates The mean vector of the probability density function of the first Gaussian components, Indicates The covariance matrix of the first Gaussian components, , are the means of the trajectory variable and the time variable, respectively. The trajectory variables and time variables The variance of is the trajectory variable and time variables The covariance between

[0027] The log-likelihood function of the first Gaussian mixture model is:

[0028]

[0029] in, , which are the two parameters in the Gaussian mixture model, is the number of points of the teaching trajectory;

[0030] Use the maximum expectation algorithm to iteratively update the parameters of the first Gaussian mixture model and establish the first Gaussian mixture model of the trajectory:

[0031] The algorithm calculates each trajectory in step E By The posterior probability generated by the first Gaussian component ;

[0032] in, Indicates The prior probability of the first Gaussian component, Indicates The mean vector of the probability density function of the first Gaussian components, Indicates The covariance matrix of the first Gaussian components;

[0033] Update the first Gaussian mixture model parameters in N steps: , , ;

[0034] in, They are respectively The mean vector, covariance matrix and prior probability of the first Gaussian component, T represents the transpose;

[0035] Repeat steps E and N until convergence conditions are met. ;

[0036] in, is the latest model parameter iterated, is the previous set of model parameters, The thresholds set for the algorithm;

[0037] After obtaining the first Gaussian mixture model, Gaussian mixture regression is used to obtain the expected trajectory :

[0038] ;

[0039] in, It is the time series when the robot performs the teaching task. For a given time series , the weights of each first Gaussian component;

[0040] The trajectory stiffness mapping model learning and training process specifically includes:

[0041] The second Gaussian mixture model is trained using the trajectory and stiffness information in the training set:

[0042] Build the second Gaussian mixture model

[0043] in, , , is the probability density function of each first Gaussian component in the first Gaussian mixture model, is the number of the second Gaussian component in the second Gaussian mixture model in trajectory stiffness mapping model training, Indicates The prior probability of the second Gaussian component, Indicates The mean vector of the probability density function of the second Gaussian component, Indicates The covariance matrix of the second Gaussian component, , are the means of the stiffness variable and trajectory variable, The stiffness variables are and trajectory variables The variance of is the stiffness variable and trajectory variables The covariance between

[0044] Iteratively update the parameters in the Gaussian mixture model using the maximum expectation algorithm to establish a second Gaussian mixture model of trajectory stiffness;

[0045] After obtaining the second Gaussian mixture model, Gaussian mixture regression is used to obtain the mapping relationship between trajectory and stiffness :

[0046]

[0047] in, For a given time series , the weights of each second Gaussian component.

[0048] Furthermore, the entire robot motion control is implemented based on admittance control, the difference lies in the expected position, expected force and virtual stiffness at each stage:

[0049] In the free space rough positioning stage, the expected position of the robot arm is the trajectory generated by the free space trajectory time mapping model. The reference trajectory needs to be corrected by the target position input by the depth camera. The expected force is 0, and the virtual stiffness is the stiffness curve generated by the trajectory stiffness mapping model.

[0050] In the stage of aligning the motherboard socket with the flat terminal plane, the expected position of the robot arm is to make the flat terminal plane fit with the template socket plane, the expected force is the preset fixed force on the Z axis of the motherboard socket posture, and the virtual stiffness is the preset fixed value;

[0051] The expected position of the robot arm during the terminal insertion phase is the trajectory generated by the contact space trajectory time mapping model, the expected force is the taught contact force during the teaching process, and the virtual stiffness is the stiffness curve generated by the trajectory stiffness mapping model.

[0052] Furthermore, in the stage of aligning the motherboard socket and the cable terminal, the robot combines the depth camera visual information with the robot inverse kinematics to solve and fine-tune the cable terminal posture. The posture matrices of the motherboard socket, the cable terminal and the end of the robot in the world coordinate system are respectively , where the pose matrix of the motherboard socket and the cable terminal is obtained by the depth camera visual recognition, and the pose matrix of the end of the robot arm is obtained by the built-in sensor of the robot arm;

[0053] When the cable terminal is fitted with the motherboard socket, the cable terminal posture matrix should meet the following requirements:

[0054]

[0055] in, Indicates that Around The rotation matrix of the axis rotation 180°;

[0056] At this time, the end-arm pose matrix should be:

[0057]

[0058] in, All of them are unit vectors described by coordinates in the world coordinate system. The three are orthogonal to form a new right-handed coordinate system to represent the posture of the end of the robot arm in the world coordinate system. The position of the end of the robot arm at this time is They represent the six joint transformation matrices of the robot arm respectively. By solving the inverse kinematics of the robot arm, the coordinates of each joint of the robot arm when the cable terminal is fitted with the motherboard socket can be solved according to the end posture matrix.

[0059] Furthermore, in the terminal insertion stage, after finding the precise assembly position, the robot arm adopts admittance control, and inputs the motion trajectory obtained by the trajectory time mapping model and the stiffness curve obtained by the stiffness trajectory mapping model into the admittance controller as feedforward input, simulating the electronic assembly operation process of the instructor. During the automatic electronic assembly process of the robot arm, the difference between the force signal transmitted by the six-dimensional force / torque sensor and the contact force of the instructor during the teaching process is used as feedback input for motion adjustment, and trajectory planning and stiffness strategy planning are performed at the same time. The planning method is:

[0060] If the current time step coordinates The coordinates of the previous time step The distance between them is less than the preset motion threshold : , then the stiffness matrix of the next time step is Each element adds 1 stiffness adjustment step ;

[0061] Get the contact force in the X or Y axis direction or , respectively, with the expected force of the contact force in the X-axis or Y-axis direction or For comparison, if or , then the desired coordinates of the next time step are moved by 1 position in the direction of reducing the contact force to adjust the step size ;

[0062] Get the contact torque in the X-axis or Y-axis direction or , respectively, with the desired contact moment in the X-axis or Y-axis direction or For comparison, if or , then the stiffness matrix of the next time step is Each element reduces the stiffness adjustment step by 1 ;

[0063] If the contact force in the Z-axis direction Reaching the preset contact threshold , and the ratio of the contact force in the Z-axis direction to the arithmetic square root of the sum of the squares of the contact forces in the X-axis and Y-axis directions is greater than the contact force direction threshold , it is considered that the cable terminal has been assembled with the motherboard socket and the task is completed.

[0064] Electronic assembly control system based on teaching, including:

[0065] A module for establishing a mathematical model for estimating the stiffness of the distal end of the upper limbs of the human body, which is used to establish a mathematical model for estimating the stiffness of the distal end of the upper limbs of the human body;

[0066] The target electronic assembly operation teaching module is used for the instructor to complete the target electronic assembly operation and record the real-time video and corresponding assembly data;

[0067] A data set generation module is used to screen qualified assembly data, obtain the motion trajectory of the end of the instructor and the contact force of the end of the instructor corresponding to the target electronic assembly operation, and obtain the stiffness of the end of the instructor corresponding to the target electronic assembly operation according to the upper limb end stiffness estimation mathematical model, and generate a data set including the motion trajectory of the end of the instructor, the stiffness of the end of the instructor, and the contact force of the end of the instructor;

[0068] A trajectory time mapping model and a trajectory stiffness mapping model generation module is used to learn and train the data set to obtain a trajectory time mapping model and a trajectory stiffness mapping model;

[0069] The variable stiffness control module of the robot arm is used to identify the motherboard socket and move to the vicinity of the motherboard socket to complete coarse positioning. According to the position and posture of the motherboard socket and the cable terminal plane transmitted by the depth camera, the robot arm adjusts the end posture to align the posture and make the two planes fit together. The robot arm imitates the teaching process and combines the motion adjustment strategy to insert the cable terminal into the motherboard socket to complete the variable stiffness control of the target electronic assembly operation.

[0070] The electronic assembly control device based on teaching includes a processor and a memory for storing a computer program that can be run on the processor. When the processor is used to run the computer program, the electronic assembly control method based on teaching is executed.

[0071] A storage medium stores a computer program, and when the computer program is executed by a processor, the electronic assembly control method based on teaching is implemented.

[0072] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0073] The electronic assembly control method based on teaching provided by the present invention establishes a mathematical model for estimating the stiffness of the end of the human upper limb, and completes the target electronic assembly operation through an instructor, records real-time video and corresponding assembly data, and then converts the instructor's assembly data into the action behavior of the robot arm, generates a data set including the end motion trajectory, end stiffness, and end contact force, and generates a trajectory-time mapping model required for the work task by training the motion information data set, which is convenient for the robot arm to imitate the instructor to perform the target electronic assembly operation, can reduce the deployment time cost of the robot arm in industrial applications, and is an effective way to achieve rapid task deployment of the robot arm in an unstructured environment. At the same time, a trajectory stiffness mapping model of the target electronic assembly operation is established, and the stiffness characteristics of the target electronic assembly operation are extracted, so that the robot arm learns and imitates the stiffness changes of the human arm during the execution of tasks, especially the contact tasks. In the execution of electronic assembly operations, the contact force between the cable terminal and the motherboard socket is significantly reduced when performing complex force conditions such as plugging in the wires. Compared with the traditional plugging method, it has higher stability and safety, reduces the loss rate in the automated assembly process, and improves the work quality of the target electronic assembly operation. In addition, since the cable terminals are usually small, in order to facilitate the insertion of the cable terminals, the present invention finally identifies the motherboard socket through a depth camera and obtains its coordinates, and the robot arm moves to the vicinity of the motherboard socket to achieve coarse positioning; according to the position and posture of the motherboard socket and the cable terminal plane transmitted by the depth camera, the robot arm adjusts the end posture to align the posture and make the two planes fit together; the robot arm imitates the teaching process and combines the motion adjustment strategy to insert the cable terminal into the motherboard socket, completes the variable stiffness control of the target electronic assembly operation, and can accurately insert the cable terminal. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0075] Figure 1 The flowchart of the electronic assembly control method based on teaching of the present invention.

[0076] Figure 2 The present invention is a device diagram for realizing the electronic assembly control method based on teaching.

[0077] Figure 3 The present invention is a flow chart of a method for establishing a mathematical model for estimating the stiffness of the distal end of the upper limb of a human body.

[0078] Figure 4 Flowchart for generating data sets for the present invention.

[0079] Figure 5 It is a schematic diagram of the electronic assembly control system based on teaching of the present invention.

[0080] Among them, 1. Platform base; 2. Robotic arm; 3. Six-dimensional force / torque sensor; 4. Electronic assembly tools; 5. Depth camera; 6. Motion capture system; 7. EMG signal sensor; 8. Camera. DETAILED DESCRIPTION

[0081] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0082] Example 1

[0083] The present invention provides an electronic assembly control method based on teaching, such as Figure 1 As shown, the following steps are included:

[0084] Step S1, establishing a mathematical model for estimating the stiffness of the distal end of the human upper limb;

[0085] Step S2: The instructor completes the target electronic assembly operation and records the real-time video and corresponding assembly data;

[0086] Step S3, screening qualified assembly data, obtaining the motion trajectory of the teaching end and the contact force of the teaching end corresponding to the target electronic assembly operation, and obtaining the teaching end stiffness corresponding to the target electronic assembly operation according to the upper limb end stiffness estimation mathematical model, and generating a data set including the motion trajectory of the teaching end, the stiffness of the teaching end, and the contact force of the teaching end;

[0087] Step S4, learning and training the data set to obtain a trajectory time mapping model and a trajectory stiffness mapping model;

[0088] Step S5, the depth camera identifies the motherboard socket and obtains its coordinates, and the robotic arm moves to the vicinity of the motherboard socket to achieve rough positioning; according to the position and posture of the motherboard socket and the cable terminal plane transmitted by the depth camera, the robotic arm adjusts the end posture to align the posture and make the two planes fit together; the robotic arm imitates the teaching process and combines the motion adjustment strategy to insert the cable terminal into the motherboard socket to complete the variable stiffness control of the target electronic assembly operation.

[0089] The electronic assembly control method based on teaching provided by the present invention establishes a mathematical model for estimating the stiffness of the end of the human upper limb, and completes the target electronic assembly operation through an instructor, records real-time video and corresponding assembly data, and then converts the instructor's assembly data into the action behavior of the robot arm, generates a data set including the end motion trajectory, end stiffness, and end contact force, and generates a trajectory-time mapping model required for the work task by training the motion information data set, which is convenient for the robot arm to imitate the instructor to perform the target electronic assembly operation, can reduce the deployment time cost of the robot arm in industrial applications, and is an effective way to achieve rapid task deployment of the robot arm in an unstructured environment. At the same time, a trajectory stiffness mapping model of the target electronic assembly operation is established, and the stiffness characteristics of the target electronic assembly operation are extracted, so that the robot arm learns and imitates the stiffness changes of the human arm during the execution of tasks, especially the contact tasks. In the execution of electronic assembly operations, the contact force between the cable terminal and the motherboard socket is significantly reduced when performing complex force tasks such as plugging in the cable. Compared with the traditional plugging method, it has higher stability and safety, reduces the loss rate in the automated assembly process, and improves the work quality of the target electronic assembly operation. Because the data cable interface is usually small, in order to facilitate the insertion of the data cable, the present invention finally identifies the motherboard socket through a depth camera and obtains its coordinates, and the robot arm moves to the vicinity of the motherboard socket to achieve coarse positioning; according to the position and posture of the motherboard socket and the cable terminal plane transmitted by the depth camera, the robot arm adjusts the end posture to align the posture and make the two planes fit together; the robot arm imitates the teaching process and combines the motion adjustment strategy to insert the cable terminal into the motherboard socket, completes the variable stiffness control of the target electronic assembly operation, and can accurately insert the cable terminal.

[0090] In step S1, Figure 3 As shown in the figure, the method for establishing the mathematical model for estimating the stiffness of the upper limb end of the human body is:

[0091] In step S101, a mathematical model for estimating the stiffness of the upper limb end of the human body represented by the electromyographic signal is established;

[0092] In step S102, random motion information of the robot arm 2 controlled by the instructor is obtained;

[0093] In step S103, a kinematic model of the upper limbs of the instructor is established;

[0094] In step S104, the Jacobian matrix of the upper limb is obtained based on the upper limb kinematic model;

[0095] In step S105, the equivalent end stiffness of the movement of the robot arm 2 is obtained;

[0096] In step S106, a mathematical model for estimating the stiffness of the upper limb end of the human body is obtained based on the random motion information, the Jacobian matrix of the upper limb and the equivalent end stiffness fitting.

[0097] In step S101, the method for establishing a mathematical model for estimating the stiffness of the upper limb end of the human body represented by the electromyographic signal is as follows:

[0098] Construct the geometric change relationship between the stiffness of the upper limb end and the stiffness of multiple joints of the upper limb:

[0099]

[0100] in: represents the end stiffness matrix; represents the multi-joint stiffness matrix; Indicates the external force on the end; Represents the joint angles based on the upper limb kinematic model; represents the upper limb Jacobian matrix, represents the pseudo-inverse matrix of a matrix;

[0101] Construct the fitting relationship between the electromyographic signal and the stiffness of multiple joints of the human upper limbs:

[0102]

[0103] in: Indicates the level of muscle activation, , are two fitting parameters, Indicates the joint stiffness value under the state of minimum muscle activation;

[0104] The mathematical model for estimating the stiffness of the upper limb end of the human body represented by the electromyographic signal is:

[0105] .

[0106] In step S102, a mechanical perturbation method is used to obtain random motion information of the instructor controlling the movement of the robotic arm 2, specifically: the instructor sticks motion capture reflective points on his body and wears an electromyographic signal sensor 7, a six-dimensional force / torque sensor 3 is installed at the end of the robotic arm 2, the instructor controls the robotic arm 2 to perform random perturbations, the instructor holds the end of the robotic arm 2 and maintains a stable muscle activation level, and the instructor's motion information of this process is recorded, including the instructor's arm end displacement, the instructor's arm joint angle and upper limb Jacobian matrix, the instructor's end contact force and electromyographic signal. In this process, the instructor's hand holds the end of the robotic arm (the two are physically coupled), that is, the instructor's arm end displacement is the same as the robotic arm end displacement, and the instructor's end contact force is equal to the contact force received by the robotic arm end and opposite in direction.

[0107] In step S103, a kinematic model of the upper limbs of the instructor is established based on the DH method; specifically, the movement of the upper limbs of the human body is simplified into a 7-degree-of-freedom link kinematic model, and a transformation matrix of adjacent links is constructed based on the DH method.

[0108] In step S104, the instructor's upper limb Jacobian matrix is ​​calculated in combination with the instructor's arm joint angles in the process of establishing a mathematical model for estimating the end stiffness of the human body, which is determined by the transformation matrix from the end link of the upper limb model to each link.

[0109] In step S105, the stiffness of the upper limb end of the human body is identified; the upper limb of the instructor is simplified into a second-order mechanical system, the displacement of the instructor's arm end is regarded as the system input, and the contact force of the instructor's arm end is regarded as the system output, and the equivalent end stiffness of the instructor in the process of establishing the mathematical model for estimating the stiffness of the upper limb end of the human body is identified using the least squares method. .

[0110] In step S106, the parameters of the mathematical model for estimating the stiffness of the upper limb end of the human body are identified; the fitting parameters in the model are obtained by fitting the model using the least squares method. , , , and obtain the true model of the mathematical model for estimating the stiffness of the upper limb end of the human body of the instructor.

[0111] In step S2, in the process of manually completing the target electronic assembly operation, an artificial target electronic assembly operation plan is designed: the instructor needs to stick motion capture reflective points on his body and wear an electromyography signal sensor 7 to obtain assembly data of the instructor's motion trajectory changes, stiffness changes, and contact force changes in the target electronic assembly operation to maintain smooth and non-mutated assembly data, and at the same time perform video recording and various motion information data collection work.

[0112] like Figure 4 As shown, step S3 specifically includes:

[0113] Step S301: for the completed electronic assembly components, the target electronic assembly operation quality is evaluated from aspects such as component assembly effect, instructor motion trajectory, instructor stiffness, instructor displacement curve smoothness, etc., and assembly data with good quality (reaching a preset threshold) is selected as valid data in combination with video playback;

[0114] Step S302, preprocessing the original electromyographic signal to obtain a muscle activation level sequence;

[0115] Step S303, calculating the upper limb Jacobian matrix of the instructor at each time point during the target electronic assembly operation, and outputting an upper limb Jacobian matrix sequence;

[0116] Step S304, inputting the muscle activation level sequence and the upper limb movement Jacobian matrix sequence in the valid data into the human upper limb end stiffness estimation mathematical model, and outputting the end stiffness sequence in the target electronic assembly operation process;

[0117] Step S305, aligning the motion trajectory sequence of the end of the instructor and the stiffness sequence of the end of the instructor using a time warping algorithm for the valid data;

[0118] Step S306, dividing each set of assembly data into two segments of data, namely, free space and contact space, according to the magnitude of the contact force of the end of the instructor;

[0119] Step S307 , generating a data set of the motion of the instructor in free space and contact space, including a terminal motion trajectory sequence, an instructor terminal stiffness sequence, and an instructor terminal contact force sequence.

[0120] In step S302, the linear noise component in the original low-frequency electromyographic signal is first removed, and then the amplitude is corrected to adjust it to a positive value; then the electromyographic signal is processed using a moving average algorithm, and finally a low-pass filter is performed to obtain the amplitude envelope of the electromyographic signal, and the ratio of the amplitude envelope to the maximum value of the subject's electromyographic signal is the muscle activation level.

[0121] In step S305, the specific method of aligning the teaching data with the time warping algorithm for the valid data is:

[0122] A teaching data in multiple sets of assembly data is used as a reference curve ,in, is the motion trajectory data of the end of the instructor in the reference curve, is the stiffness data of the end of the instructor in the reference curve;

[0123] The remaining teaching data is used as observation curves , is the motion trajectory data of the end of the instructor in the reference curve, is the stiffness data of the end of the instructor in the reference curve;

[0124] The track grid consisting of the reference points on the reference curve and the test points on the test curve is obtained, where the distance between the reference point and the test point in any track grid point is , search for a path with the minimum sum of distance values ​​from the starting track grid point to the ending track grid point, and rearrange with the index on the path to align the reference curve with the test curve.

[0125] In step S4, the trajectory time mapping model learning and training process specifically includes:

[0126] In step S401, the first Gaussian mixture model is trained using the motion trajectory and time information in the training set:

[0127] Build the first Gaussian mixture model

[0128] in, , , is the probability density function of each first Gaussian component in the first Gaussian mixture model, is the number of the first Gaussian component in the first Gaussian mixture model in the teaching trajectory training, Indicates The prior probability of the first Gaussian component, Indicates The mean vector of the probability density function of the first Gaussian components, Indicates The covariance matrix of the first Gaussian components, , are the means of the trajectory variable and the time variable, respectively. The trajectory variables and time variables The variance of is the trajectory variable and time variables The covariance between

[0129] The log-likelihood function of the first Gaussian mixture model is:

[0130]

[0131] in, , which are the two parameters in the Gaussian mixture model, is the number of points of the teaching trajectory;

[0132] In step S402, the maximum expectation algorithm (EM algorithm) is used to iteratively update the parameters of the first Gaussian mixture model to establish the first Gaussian mixture model of the trajectory:

[0133] The algorithm calculates each trajectory in step E By The posterior probability generated by the first Gaussian component ;

[0134] in, Indicates The prior probability of the first Gaussian component, Indicates The mean vector of the probability density function of the first Gaussian components, Indicates The covariance matrix of the first Gaussian components;

[0135] Update the first Gaussian mixture model parameters in N steps: , , ;

[0136] in, They are respectively The mean vector, covariance matrix and prior probability of the first Gaussian components, T is the transpose.

[0137] Repeat steps E and N until convergence conditions are met. ;

[0138] in, is the latest model parameter iterated, is the previous set of model parameters, The thresholds set for the algorithm;

[0139] In step S403, after obtaining the first Gaussian mixture model, the expected trajectory is obtained by Gaussian mixture regression :

[0140]

[0141] in, It is the time series when the robot performs the teaching task. For a given time series , the weights of each first Gaussian component;

[0142] In step S4, the trajectory stiffness mapping model learning and training process specifically includes:

[0143] Step S411: train the second Gaussian mixture model using the trajectory and stiffness information in the training set:

[0144] Build the second Gaussian mixture model

[0145] in, , , is the probability density function of each second Gaussian component in the second Gaussian mixture model, is the number of the second Gaussian component in the second Gaussian mixture model in trajectory stiffness mapping model training, Indicates The prior probability of the second Gaussian component, Indicates The mean vector of the probability density function of the second Gaussian component, Indicates The covariance matrix of the second Gaussian component, , are the means of the stiffness variable and trajectory variable, The stiffness variables are and trajectory variables The variance of is the stiffness variable and trajectory variables The covariance between

[0146] Step S412: Use the maximum expectation algorithm to iteratively update the parameters in the Gaussian mixture model to establish a second Gaussian mixture model of trajectory stiffness;

[0147] Step S413: After obtaining the second Gaussian mixture model, the mapping relationship between trajectory and stiffness is obtained by using Gaussian mixture regression. :

[0148]

[0149] in, For a given time series , the weights of each second Gaussian component.

[0150] Gaussian mixture model and Gaussian mixture regression are commonly used robot trajectory planning methods, but in the past they were only used for planning robot position motion. This embodiment extends them to the learning and planning of robot stiffness. Based on this system, variable stiffness motion control can be easily achieved.

[0151] In step S5, the variable stiffness control of the target electronic assembly operation is specifically divided into three stages: the depth camera identifies the motherboard socket and obtains its coordinates, and the robot moves to the vicinity of the motherboard socket to achieve rough positioning; the depth camera identifies the position and posture of the motherboard socket and the cable terminal plane, and the robot adjusts the end posture to align the posture and make the two planes fit together; the robot imitates the teaching process and combines the motion adjustment strategy to insert the cable terminal into the motherboard socket. The entire robot motion control is implemented based on admittance control. The difference lies in the different expected positions, expected forces and virtual stiffness in each stage:

[0152] Step S501, establish the kinematic and dynamic models of the robot 2 and the environment, study the motion and force analysis of the robot 2 during the task operation, simulate the environment into a spring damping system, and then realize the admittance control of the robot 2. The actual external contact force / torque measured by the six-dimensional force / torque sensor 3 at the end of the robot 2 is used to generate a new position control quantity through a second-order admittance model, and the pre-set position trajectory is corrected and sent to the inner loop of the position control to realize the admittance control of the robot 2.

[0153] Design the admittance controller:

[0154]

[0155] in, are position control quantity, reference trajectory and actual trajectory respectively. are the virtual mass, damping and stiffness matrices respectively, is the external force on the end of robot arm 2, is the desired force at the end of robot arm 2, is the control cycle. In the automatic electronic assembly process, the reference trajectory This is the GMM-GMR teaching trajectory after adjustment of the actual jack position. , virtual mass, damping matrix Set to a constant value, virtual stiffness matrix It is the teaching stiffness calculated by the trajectory stiffness mapping model.

[0156] The free space coarse positioning stage:

[0157] Step S502: Obtain the coordinates of the target socket; specifically, obtain the coordinates of the socket under the depth camera 5 through the depth camera 5 , through the coordinate transformation between the depth camera 5 and the world coordinate system Get the position and posture of the socket in the world coordinate system ;

[0158] Step S503, the robot arm 2 clamps the cable and moves to the motherboard socket;

[0159] The free space approach adjusts the teaching trajectory and trajectory-stiffness mapping model according to the coordinates of the motherboard socket and transmits it to the admittance controller for motion control; for the motherboard socket position, the electronic assembly motion trajectory in free space ,in is the initial teaching trajectory The end point coordinates of ; the expected force and trajectory stiffness mapping model are:

[0160] ;

[0161] During this motion, set the desired end force is 0, The input to the admittance controller As a reference trajectory, As a virtual stiffness matrix in the admittance controller to regulate motion control, For the The weight of the second Gaussian component.

[0162] Step S504, aligning the cable terminal and the motherboard socket plane so that the two fit together;

[0163] Since the position of the cable clamped by the robot arm is not strictly consistent every time it performs an electronic assembly task, the motherboard socket and the cable terminal cannot be completely fitted. When the motherboard socket and the cable terminal are not aligned and the position is deviated, the docking assembly will be blocked. At this time, it is necessary to combine the visual information of the depth camera 5 with the inverse kinematics of the robot arm 2 to fine-tune the position of the cable terminal. The position matrices of the template socket, cable terminal and the end of the robot arm 2 in the world coordinate system are respectively , where the motherboard socket and cable terminal pose matrix is ​​obtained by visual recognition of the depth camera 5, and the robot arm end pose matrix is ​​obtained by the robot arm built-in sensor.

[0164] When the cable terminal is fitted with the motherboard socket, its posture matrix should meet the following requirements:

[0165]

[0166] in, Indicates that Around The rotation matrix for rotating the axis by 180°.

[0167] At this time, the end pose matrix of robot arm 2 should be:

[0168]

[0169] in, All of them are unit vectors described by coordinates in the world coordinate system. The three are orthogonal to form a new right-handed coordinate system to represent the end posture of the robot arm 2 in the world coordinate system. This is the position of the end of the robot arm 2 at this time, They represent the six joint transformation matrices of the robot 2 respectively. By solving the inverse kinematics of the robot, the coordinates of each joint of the robot 2 when the cable terminal is fitted with the motherboard socket can be solved according to the end posture matrix.

[0170] Step S505, the robot arm 2 inserts the cable terminal into the motherboard socket;

[0171] After finding the precise assembly position during the terminal insertion stage, the robot 2 uses admittance control, and inputs the motion trajectory obtained by the trajectory time mapping model and the stiffness curve obtained by the stiffness trajectory mapping model into the admittance controller as feedforward input to simulate the electronic assembly process of the instructor. The joining process between the terminal and the motherboard socket is complicated. During the automatic electronic assembly process of the robot arm, the difference between the force signal transmitted by the six-dimensional force / torque sensor 3 and the contact force of the instructor during the teaching process is used as feedback input for motion adjustment, and trajectory planning and stiffness strategy planning are performed at the same time. The planning method is:

[0172] If the current time step coordinates The coordinates of the previous time step The distance between them is less than the preset motion threshold : , then the stiffness matrix of the next time step is Each element adds 1 stiffness adjustment step ;

[0173] Get the contact force in the X or Y axis direction or , respectively, with the expected force of the contact force in the X-axis or Y-axis direction or For comparison, if or , then the desired coordinates of the next time step are moved by 1 position in the direction of reducing the contact force to adjust the step size ;

[0174] Get the contact torque in the X-axis or Y-axis direction or , respectively, with the desired contact moment in the X-axis or Y-axis direction or For comparison, if or , then the stiffness matrix of the next time step is Each element reduces the stiffness adjustment step by 1 ;

[0175] If the contact force in the Z-axis direction Reaching the preset contact threshold , and the ratio of the contact force in the Z-axis direction to the arithmetic square root of the sum of the squares of the contact forces in the X-axis and Y-axis directions is greater than the contact force direction threshold , it is considered that the cable terminal has been assembled with the motherboard socket and the task is completed.

[0176] In order to realize the electronic assembly control method based on teaching, the present invention provides an electronic assembly control device based on teaching, such as Figure 2As shown, it includes a platform base 1, a robotic arm 2, a six-dimensional force / torque sensor 3, an electronic assembly tool 4, a depth camera 5, a motion capture system 6, an electromyographic signal sensor 7 and a camera 8. The robotic arm 2 is arranged on the platform base 1, the electronic assembly tool 4 and the six-dimensional force / torque sensor 3 are arranged at the end of the robotic arm 2, the electronic assembly tool 4 is used to perform electronic assembly operations on microelectronic devices, the six-dimensional force / torque sensor 3 is used to obtain the end contact force of the robotic arm 2, the depth camera 5 is arranged on the platform base 1 to obtain the coordinates of the task end point, the motion capture system 6 is used to obtain the coordinates of the special position (such as each joint) of the instructor through a group of infrared cameras 8 systems to obtain the angles of each joint of the instructor, the electromyographic signal sensor 7 is worn on the instructor's arm to obtain electromyographic signals, and the camera 8 records the teaching process.

[0177] Example 2

[0178] The present invention provides an electronic assembly control system based on teaching, such as Figure 5 As shown, including:

[0179] A module for establishing a mathematical model for estimating the stiffness of the distal end of the upper limbs of the human body, which is used to establish a mathematical model for estimating the stiffness of the distal end of the upper limbs of the human body;

[0180] The target electronic assembly operation teaching module is used for the instructor to complete the target electronic assembly operation and record the real-time video and corresponding assembly data;

[0181] A data set generation module is used to screen qualified assembly data, obtain the motion trajectory of the end of the instructor and the contact force of the end of the instructor corresponding to the target electronic assembly operation, and obtain the stiffness of the end of the instructor corresponding to the target electronic assembly operation according to the upper limb end stiffness estimation mathematical model, and generate a data set including the motion trajectory of the end of the instructor, the stiffness of the end of the instructor, and the contact force of the end of the instructor;

[0182] A trajectory time mapping model and a trajectory stiffness mapping model generation module is used to learn and train the data set to obtain a trajectory time mapping model and a trajectory stiffness mapping model;

[0183] The variable stiffness control module of the robot arm 2 is used to identify the motherboard socket and obtain its coordinates. The robot arm moves to the vicinity of the motherboard socket to achieve rough positioning; according to the position and posture of the motherboard socket and the cable terminal plane transmitted by the depth camera, the robot arm adjusts the end posture to align the posture and make the two planes fit together; the robot arm imitates the teaching process and combines the motion adjustment strategy to insert the cable terminal into the motherboard socket to complete the variable stiffness control of the target electronic assembly operation.

[0184] Example 3

[0185] The present invention provides a teaching-based electronic assembly control device, including a processor and a memory for storing a computer program that can be run on the processor. When the processor is used to run the computer program, it executes the teaching-based electronic assembly control method described in Example 1.

[0186] Example 4

[0187] The present invention provides a storage medium, in which a computer program is stored. When the computer program is executed by a processor, the electronic assembly control method based on teaching as described in Example 1 is implemented.

[0188] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. An electronic assembly control method based on teaching, characterized in that: The following steps are involved: Establish a mathematical model for estimating the stiffness of the human upper limb extremities; The instructor completes the target electronic assembly task and records the real-time video and corresponding assembly data; Filter qualified assembly data, obtain the motion trajectory of the end of the instructor and the contact force of the end of the instructor corresponding to the target electronic assembly operation, and obtain the stiffness of the end of the instructor corresponding to the target electronic assembly operation according to the upper limb end stiffness estimation mathematical model, and generate a data set including the motion trajectory of the end of the instructor, the stiffness of the end of the instructor, and the contact force of the end of the instructor; The data set is studied and trained to obtain a trajectory time mapping model and a trajectory stiffness mapping model; The learning and training process of the trajectory time mapping model specifically includes: The first Gaussian mixture model is trained using the motion trajectory and time information in the training set: Build the first Gaussian mixture model in, is the probability density function of each first Gaussian component in the first Gaussian mixture model, E is the number of first Gaussian components in the first Gaussian mixture model in the teaching trajectory training, π e represents the prior probability of the e-th first Gaussian component, μ e represents the mean vector of the probability density function of the e-th first Gaussian component, Σ e represents the covariance matrix of the e-th first Gaussian component, μ ξ,e , μ t,e are the means of the trajectory variable and the time variable, Σ ξξ,e ,Σ tt,e are the variances of the trajectory variable ξ and the time variable t, Σ tξ,e ,Σ ξt,e is the covariance between the trajectory variable ξ and the time variable t; The log-likelihood function of the first Gaussian mixture model is: Where θ = {μ e ,Σ e }, which are two parameters in the Gaussian mixture model, and N is the number of teaching trajectory points; Use the maximum expectation algorithm to iteratively update the parameters of the first Gaussian mixture model and establish the first Gaussian mixture model of the trajectory: The algorithm calculates each trajectory ξ in step E i The posterior probability generated by the e-th first Gaussian component Among them, π j represents the prior probability of the jth first Gaussian component, μ j represents the mean vector of the probability density function of the jth first Gaussian component, Σ j represents the covariance matrix of the jth first Gaussian component; Update the first Gaussian mixture model parameters in N steps: in, are the mean vector, covariance matrix and prior probability of the e-th first Gaussian component after N steps of updating, and T represents the transpose; Repeat step E and step N until the convergence condition ln L(θ is met new )-ln L(θ)<∈; in, is the latest model parameter iterated, θ={μ e ,Σ e } is the previous set of model parameters, ∈ is the threshold set by the algorithm; After obtaining the first Gaussian mixture model, Gaussian mixture regression is used to obtain the expected trajectory ξ′: Among them, t′ is the time series when the robot performs the teaching task, h e (t′) is the weight of each first Gaussian component for a given time series t′; The trajectory stiffness mapping model learning and training process specifically includes: The second Gaussian mixture model is trained using the trajectory and stiffness information in the training set: Build the second Gaussian mixture model in, is the probability density function of each first Gaussian component in the first Gaussian mixture model, C is the number of second Gaussian components in the second Gaussian mixture model in the trajectory stiffness mapping model training, π c represents the prior probability of the cth second Gaussian component, μ c represents the mean vector of the probability density function of the cth second Gaussian component, Σ c represents the covariance matrix of the cth second Gaussian component, μ K,c , μ ξ,c are the means of the stiffness variable and trajectory variable, Σ KK,c ,Σ ξξ,c are the variances of the stiffness variable K and trajectory variable ξ, Σ ξK,c ,Σ Kξ,c is the covariance between the stiffness variable K and the trajectory variable ξ; Iteratively update the parameters in the Gaussian mixture model using the maximum expectation algorithm to establish a second Gaussian mixture model of trajectory stiffness; After obtaining the second Gaussian mixture model, Gaussian mixture regression is used to obtain the mapping relationship K′ between trajectory and stiffness: Among them, h c (ξ′) is the weight of each second Gaussian component for a given time series ξ′; The depth camera identifies the motherboard socket and obtains the coordinates of the motherboard socket, and the robotic arm moves to the vicinity of the motherboard socket to achieve coarse positioning; according to the position and posture of the motherboard socket and the cable terminal plane transmitted by the depth camera, the robotic arm adjusts the end posture to align the posture and make the two planes fit together; the robotic arm inserts the cable terminal into the motherboard socket according to the trajectory time mapping model and the trajectory stiffness mapping model and combined with the motion adjustment strategy to complete the variable stiffness control of the target electronic assembly operation.

2. The electronic assembly control method based on teaching according to claim 1, characterized in that: For the completed electronic assembly parts, the assembly data whose quality reaches the preset threshold is selected as valid data by combining the real-time video playback; Preprocess the raw electromyographic signals to obtain the muscle activation level sequence; Calculate the upper limb Jacobian matrix of the instructor at each time point during the target electronic assembly operation, and output the upper limb Jacobian matrix sequence; The muscle activation level sequence and upper limb motion Jacobian matrix sequence in the effective data are input into the human upper limb end stiffness estimation mathematical model, and the end stiffness sequence of the instructor in the target electronic assembly operation process is output; The effective data is aligned with the motion trajectory sequence of the end of the instructor and the stiffness sequence of the end of the instructor using a time warping algorithm; Each set of assembly data is divided into two sections: free space and contact space according to the magnitude of the end contact force of the instructor; Generate a dataset of the motion of the instructor in free space and contact space, including the end motion trajectory sequence, the instructor end stiffness sequence, and the instructor end contact force sequence.

3. The electronic assembly control method based on teaching according to claim 1, characterized in that: The motion trajectory of the end pendant and the stiffness of the end pendant in the target electronic assembly operation are aligned by the time warping algorithm. The specific method is as follows: A certain teaching data in multiple sets of assembly data is used as the reference curve ξ r ={x,K end }, where x is the motion trajectory data of the end of the instructor in the reference curve, K end is the stiffness data of the end of the instructor in the reference curve; The remaining teaching data is used as the observation curve ξ t ={x′,K e ' nd }, x′ is the motion trajectory data of the end of the instructor in the reference curve, K e ' nd is the stiffness data of the end of the instructor in the reference curve; A trajectory grid consisting of reference points on the reference curve and test points on the test curve is obtained, wherein the distance between the reference point and the test point in any trajectory grid point is d, and a path with the minimum sum of distance values ​​is searched from the starting trajectory grid point to the ending trajectory grid point, and the indexes on the path are rearranged to align the reference curve with the test curve.

4. The electronic assembly control method based on teaching according to claim 1, characterized in that: The entire robot motion control is implemented based on admittance control, the difference lies in the expected position, expected force and virtual stiffness at each stage: In the free space coarse positioning stage, the expected position of the robot arm is the trajectory generated by the free space trajectory time mapping model. The trajectory generated by the free space trajectory time mapping model needs to be corrected by the target position input by the depth camera. The expected force is 0, and the virtual stiffness is the stiffness curve generated by the trajectory stiffness mapping model. In the stage of aligning the motherboard socket with the flat terminal plane, the expected position of the robot arm is to make the flat terminal plane fit with the template socket plane, the expected force is the preset fixed force on the Z axis of the motherboard socket posture, and the virtual stiffness is the preset fixed value; The expected position of the robot arm during the terminal insertion phase is the trajectory generated by the contact space trajectory time mapping model, the expected force is the taught contact force during the teaching process, and the virtual stiffness is the stiffness curve generated by the trajectory stiffness mapping model.

5. The electronic assembly control method based on teaching according to claim 1, characterized in that: In the stage of aligning the motherboard socket and the cable terminal, the robot arm combines the depth camera visual information with the inverse kinematics of the robot arm to fine-tune the position of the cable terminal. The position matrices of the motherboard socket, the cable terminal and the end of the robot arm in the world coordinate system are T g ,T o ,T e , where the pose matrix of the motherboard socket and the cable terminal is obtained by the depth camera visual recognition, and the pose matrix of the end of the robot arm is obtained by the built-in sensor of the robot arm; When the cable terminal is fitted with the motherboard socket, the cable terminal posture matrix should meet the following requirements: Among them, Rot(x,π) means that T g Rotation matrix for 180° rotation around the x-axis; At this time, the end-arm pose matrix should be: Among them, [n x ,n y ,n z ]、[o x ,o y ,o z ]、[a x ,a y ,a z ] are all unit vectors described by coordinates in the world coordinate system. The three are orthogonal to form a new right-handed coordinate system to represent the end posture of the robot arm in the world coordinate system. x ,p y ,p z ] is the position of the end of the robot arm at this time, They represent the six joint transformation matrices of the robot arm respectively. By solving the inverse kinematics of the robot arm, the coordinates of each joint of the robot arm when the cable terminal is fitted with the motherboard socket can be solved according to the end posture matrix.

6. The electronic assembly control method based on teaching according to claim 1, characterized in that: After finding the precise assembly position in the terminal insertion stage, the robot adopts admittance control, and inputs the motion trajectory obtained by the trajectory time mapping model and the stiffness curve obtained by the stiffness trajectory mapping model into the admittance controller as feedforward input, simulating the electronic assembly operation process of the instructor. During the automatic electronic assembly process of the robot arm, the difference between the force signal transmitted by the six-dimensional force / torque sensor and the contact force of the instructor during the teaching process is used as feedback input for motion adjustment, and trajectory planning and stiffness strategy planning are performed at the same time. The planning method is: If the current time step coordinate ξ step With the previous time step coordinate ξ step-1 The distance between them is less than the preset motion threshold d th :||ξ step -ξ step-1 || <d th , then the stiffness matrix K of the next time step step+1 Each element adds 1 stiffness adjustment step ΔK; Get the contact force in the X-axis or Y-axis direction |F x |or|F y |, respectively, the expected force taught for the contact force in the X-axis or Y-axis direction |F xth |or|F yth |Compare, if |F x |>|F xth |or|F y |>|F yth |, then the desired coordinates of the next time step are moved by one position in the direction of reducing the contact force by the adjustment step length Δx; Get the contact torque in the X-axis or Y-axis direction |M x |or|M y |, respectively, the expected torque of the contact torque in the X-axis or Y-axis direction|M xth |or|M yth |Compare, if |M x |>|M xth |or|M y |>|M yth |, then the stiffness matrix K of the next time step step+1 Each element reduces the stiffness adjustment step size ΔK by 1; If the contact force F in the Z-axis direction z Reaching the preset contact threshold F th , and the ratio of the contact force in the Z-axis direction to the arithmetic square root of the sum of the squares of the contact forces in the X-axis and Y-axis directions is greater than the contact force direction threshold It is considered that the cable terminal has been assembled with the motherboard socket and the electronic assembly task is completed.

7. A teaching-based electronic assembly control system, applied to the teaching-based electronic assembly control method according to any one of claims 1 to 6, characterized in that: include: A module for establishing a mathematical model for estimating the stiffness of the distal end of the upper limbs of the human body, which is used to establish a mathematical model for estimating the stiffness of the distal end of the upper limbs of the human body; The target electronic assembly operation teaching module is used for the instructor to complete the target electronic assembly operation and record the real-time video and corresponding assembly data; A data set generation module is used to screen qualified assembly data, obtain the motion trajectory of the end of the instructor and the contact force of the end of the instructor corresponding to the target electronic assembly operation, and obtain the stiffness of the end of the instructor corresponding to the target electronic assembly operation according to the upper limb end stiffness estimation mathematical model, and generate a data set including the motion trajectory of the end of the instructor, the stiffness of the end of the instructor, and the contact force of the end of the instructor; A trajectory time mapping model and a trajectory stiffness mapping model generation module is used to learn and train the data set to obtain a trajectory time mapping model and a trajectory stiffness mapping model; The variable stiffness control module of the robot arm is used to identify the motherboard socket and move to the vicinity of the motherboard socket to complete coarse positioning. According to the position and posture of the motherboard socket and the cable terminal plane transmitted by the depth camera, the robot arm adjusts the end posture to align the posture and make the two planes fit together. The robot arm imitates the teaching process and combines the motion adjustment strategy to insert the cable terminal into the motherboard socket to complete the variable stiffness control of the target electronic assembly operation.

8. Electronic assembly control equipment based on teaching, characterized in that: It comprises a processor and a memory for storing a computer program that can be run on the processor. When the processor is used to run the computer program, it executes the electronic assembly control method based on teaching as described in any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the teaching-based electronic assembly control method according to any one of claims 1 to 6 is implemented.

Citation Information

Patent Citations

  • Robot shaft hole assembly system and method based on teaching learning

    CN109382828A

  • Robot polishing system and method based on demonstration teaching and learning

    CN109483556A

  • Track profiling flaw detection method based on mechanical arm

    CN114839264A

  • Control method of high-speed rail passenger station inspection robot mechanical arm operating system

    CN116533252A

  • Mechanical arm learning method and equipment for humanoid arm rigidity and motion skills

    CN116652938A