Industrial robot teaching method and system

Abstract

The present invention provides an industrial robot teaching method and system, and relates to the technical field of industrial robots. The vibration offset evaluation index of the present invention can quantify the vibration offset of each teaching action, thereby providing an evaluation standard for the correction of the teaching path; improving the perception and processing capabilities of small vibrations in the teaching process; introducing the correlation influence coefficient, so that the relationship between the vibration offset and the muscle reaction of the instructor can be quantified and analyzed, providing data support for the dynamic adjustment and optimization of the teaching path; the generated vibration compensation strategy corrects the teaching path, and improves the teaching accuracy and stability.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial robots, and in particular to an industrial robot teaching method and system. Background Art

[0002] With the rapid development of industrial automation, industrial robots have been widely used in manufacturing, logistics, medical care, agriculture and other fields. The traditional robot teaching method mainly relies on the teach pendant or programmer, and the position and posture of each joint of the robot are recorded by manually operating the teach pendant. Although this method is mature, it has certain limitations, such as complex operation, low efficiency, and the teaching accuracy depends on the skill level of the operator. In recent years, the teaching method based on image recognition and motion capture has gradually become a research hotspot. By capturing and analyzing the hand movements of the instructor, a more natural, intuitive and efficient robot teaching process can be achieved.

[0003] In the prior art, the publication number is CN107160364B, and the name is an industrial robot teaching system based on machine vision, including an image sensor, a marker, a computer equipped with a robot teaching module, a robot controller, and a robot. The image sensor is connected to the computer equipped with the robot teaching module, and is used to obtain the image of a human hand during the robot teaching process; the marker is placed on the back of the instructor's hand; the computer obtains the position and posture of the marker and the human palm in the camera coordinate system through image processing and P4P posture estimation, estimates the angles of the three joints of the instructor's index finger, and obtains the posture relationship between the fingers and the palm; the computer controls the robot through Ethernet to repeat the path demonstrated by the human hand, thereby realizing the robot teaching reproduction.

[0004] At present, most of the existing teaching methods based on image recognition rely on markers or specific gloves to accurately identify the position and posture of the hand in the image; however, these methods have some shortcomings. First, the use of markers or specific gloves may affect the instructor's operational flexibility and comfort, making the teaching process less natural; second, the existing algorithms still lack accuracy and stability when processing complex hand movements, especially in the case of high-frequency vibrations or small movements. In addition, the existing technology often fails to fully consider or ignores the analysis and compensation of small vibrations and electromyographic signals in the teaching path, resulting in possible errors in the teaching results.

[0005] The above information disclosed in the above Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The purpose of the present invention is to provide an industrial robot teaching method and system to solve the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] An industrial robot teaching method, the specific steps include:

[0009] Step S1: collecting an image sequence of the instructor's hand in the robot workspace and extracting the teaching path from the image sequence;

[0010] Step S2: Divide the teaching path according to the action type to form different teaching action sequences {1, 2, ..., i, ..., n}, where i represents the index of the i-th teaching action, and n represents the total number of different teaching actions, and collect the micro-vibration data of each teaching action at the corresponding action holding time in real time;

[0011] Step S3: analyzing the micro-vibration data of each teaching action at the corresponding action holding time, and generating a vibration offset evaluation index for evaluating the vibration offset amount of each teaching action at the corresponding action holding time;

[0012] Step S4: collecting the electromyographic signal of each teaching action when the vibration offset occurs, and performing correlation analysis on the electromyographic signal and the vibration offset evaluation index to obtain the correlation influence coefficient, the electromyographic signal includes the electromyographic signal amplitude and the electromyographic frequency;

[0013] Step S5: Fitting and analyzing the teaching path, the vibration offset evaluation index and the associated influence coefficient to generate a vibration compensation strategy for correcting the teaching path, and defining the vibration compensation formula of the teaching path as follows:

[0014]

[0015] Among them, p t,i is the corrected path point of the i-th teaching action in the teaching path at time step t, is the original path point of the i-th teaching action at time step t, k i is the compensation coefficient of the i-th type of teaching action, Δd i is the vibration displacement of the i-th type of teaching action;

[0016] Step S6: Convert the posture relationship information of the fingers and the hand body to the robot base coordinate system, discretely record the posture of the teaching points in the continuous teaching path, and perform median filtering on the corrected path point data to obtain a smooth teaching path; transmit the smoothed teaching path data to the robot controller to realize the robot's teaching reproduction.

[0017] Furthermore, the generation of the vibration offset evaluation index specifically includes:

[0018] Set the vibration offset evaluation index to E i , which represents the vibration offset evaluation index of the i-th type of teaching action within its holding time Ti, and the calculation formula is as follows:

[0019]

[0020] Where m is the total number of micro-vibration data acquisitions for each teaching action at the corresponding action holding time Ti;

[0021] f i,j is the microseismic frequency collected at the jth time for the i-th type of teaching action; ZA i,j is the microseismic amplitude collected at the jth time for the i-th type of teaching action;

[0022] max(f i,j ) is the maximum microseismic frequency of the i-th type of teaching action in all acquisitions;

[0023] max(ZA i,j ) is the maximum microseismic amplitude of the i-th type of teaching action in all acquisitions;

[0024] w j is the weight function, defined represents the weight of the jth acquisition; α and β are attenuation coefficients, which are used to adjust the influence of frequency and amplitude on the vibration offset evaluation index respectively, and the value range of α and β are both in the range of (0,1).

[0025] Furthermore, the generation of the correlation influence coefficient specifically includes:

[0026] Assume the correlation coefficient is K i , represents the correlation between the electromyographic signal and the vibration offset evaluation index of the i-th type of teaching action when the vibration offset occurs; the calculation formula is as follows:

[0027]

[0028] Among them, A EMG,i,j is the amplitude of the electromyographic signal collected at the jth time for the i-th type of teaching action;

[0029] max(A EMG,i,j ) is the maximum EMG signal amplitude of the i-th type of teaching action in all acquisitions;

[0030] f EMG,i,j is the frequency of the electromyographic signal collected at the jth time for the i-th type of teaching action;

[0031] max(f EMG,i,j) is the maximum EMG signal frequency of the i-th type of teaching action in all acquisitions;

[0032] γ and δ are attenuation coefficients, which are used to adjust the influence of the amplitude and frequency of the electromyographic signal on the correlation coefficient respectively; the value range of γ and δ is within the range of (0,1); set K i The valid value range of is between (0,1);

[0033] When K i The closer it is to 1, the higher the correlation between the electromyographic signal and the vibration excursion evaluation index, and the greater the influence of muscle activity on vibration excursion during action execution;

[0034] When K i The closer it is to 0, the lower the correlation between the electromyographic signal and the vibration offset evaluation index, and the smaller the effect of muscle activity on vibration offset during action execution.

[0035] Furthermore, a vibration compensation strategy for correcting the teaching path is generated, specifically including:

[0036] When K i ≥0.65, and E i ≥0.75, for k i Make the following adjustments:

[0037]

[0038] When K i ≥0.65, and 0.46≤E i <0.75, for k i Make the following adjustments:

[0039]

[0040] When K i ≥0.65, and E i <0.46, for k i Make the following adjustments:

[0041]

[0042] When K i <0.65, and E i ≥0.75, for k i Make the following adjustments:

[0043]

[0044] When K i <0.65, and 0.46≤E i <0.75, for k i Make the following adjustments:

[0045]

[0046] When K i <0.65, and E i <0.46, for k i Make the following adjustments:

[0047]

[0048] Among them, k' is the adjusted compensation coefficient; η1 and η2 are both adjustment coefficients, and

[0049] An industrial robot teaching system, the system is used to execute the industrial robot teaching method, comprising:

[0050] Teaching path determination module: collects the image sequence of the instructor's hand in the robot workspace and extracts the teaching path from the image sequence;

[0051] Division module: used to divide the teaching path according to the action type to form different teaching action sequences {1, 2, ..., i, ..., n}, where i represents the index of the i-th teaching action, and n represents the total number of different teaching actions, and collects the micro-vibration data of each teaching action at the corresponding action holding time in real time;

[0052] Vibration offset evaluation index generation module: used to analyze the micro-vibration data of each teaching action at the corresponding action holding time, and generate a vibration offset evaluation index for evaluating the vibration offset amount of each teaching action at the corresponding action holding time;

[0053] Correlation influence coefficient generation module: used to collect the electromyographic signal of each action when the vibration offset occurs, and to correlate and analyze the electromyographic signal with the vibration offset evaluation index to obtain the correlation influence coefficient. The electromyographic signal includes the electromyographic signal amplitude and the electromyographic frequency.

[0054] Vibration compensation strategy generation module: used to fit and analyze the teaching path, vibration offset evaluation index and associated influence coefficient to generate a vibration compensation strategy for correcting the teaching path, and define the vibration compensation formula of the teaching path as follows:

[0055]

[0056] Among them, p t,i is the corrected path point of the i-th teaching action in the teaching path at time step t, is the original path point of the i-th teaching action at time step t, k i is the compensation coefficient of the i-th type of teaching action, Δd iis the vibration displacement of the i-th type of teaching action;

[0057] Teaching and reproduction module: used to convert the posture relationship information of fingers and hand body into the robot base coordinate system, discretely record the posture of teaching points in the continuous teaching path, and perform median filtering on the corrected path point data to obtain a smooth teaching path; transmit the smoothed teaching path data to the robot controller to realize the robot's teaching reproduction.

[0058] Compared with the prior art, the beneficial effects of the present invention are: the vibration offset evaluation index can quantify the vibration offset of each teaching action, thereby providing an evaluation standard for the correction of the teaching path; it improves the perception and processing capabilities of tiny vibrations during the teaching process; the introduction of the correlation influence coefficient enables the relationship between the vibration offset and the muscle reaction of the instructor to be quantified and analyzed, providing data support for the dynamic adjustment and optimization of the teaching path; the generated vibration compensation strategy corrects the teaching path, improving the teaching accuracy and stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a schematic diagram of the overall method flow of the present invention;

[0060] Figure 2 It is a schematic diagram of the overall system module of the present invention. DETAILED DESCRIPTION

[0061] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0062] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0063] Embodiment 1:

[0064] See also Figure 1 , the present invention provides a technical solution:

[0065] An industrial robot teaching method, the specific steps include:

[0066] Step S1: collecting an image sequence of the instructor's hand in the robot workspace, and using a particle filter algorithm to track the position of the hand during the teaching process to extract the region of interest in the hand image, where the region of interest includes the hand body and fingers;

[0067] Step S2: Calculate the position and posture of the hand body in the camera coordinate system, and establish a three-dimensional model that can change according to the input finger joint parameters, and use a particle swarm optimization algorithm based on the three-dimensional model to estimate the finger joint angles, so as to obtain the posture relationship between the fingers and the hand body;

[0068] Step S3: Divide the teaching path according to the action type to form different teaching action sequences {1, 2, ..., i, ..., n}, where i represents the index of the i-th teaching action, and n represents the total number of different teaching actions, and collect the micro-vibration data of each teaching action at the corresponding action holding time in real time;

[0069] Step S4: analyzing the micro-vibration data of each teaching action at the corresponding action holding time, and generating a vibration offset evaluation index for evaluating the vibration offset amount of each teaching action at the corresponding action holding time;

[0070] Step S5: collecting the electromyographic signal of each action when the vibration offset occurs, and performing correlation analysis on the electromyographic signal and the vibration offset evaluation index to obtain the correlation influence coefficient, the electromyographic signal includes the electromyographic signal amplitude and the electromyographic frequency;

[0071] Step S6: Fitting and analyzing the teaching path, the vibration offset evaluation index and the associated influence coefficient to generate a vibration compensation strategy for correcting the teaching path, and defining the vibration compensation formula of the teaching path as follows:

[0072]

[0073] Among them, p t,i is the corrected path point of the i-th teaching action in the teaching path at time step t, is the original path point of the i-th teaching action at time step t, k i is the compensation coefficient of the i-th type of teaching action, Δd i is the vibration displacement of the i-th type of teaching action;

[0074] Step S7: Convert the posture relationship information of the fingers and the hand body to the robot base coordinate system, discretely record the posture of the teaching points in the continuous teaching path, and perform median filtering on the corrected path point data to obtain a smooth teaching path; transmit the smoothed teaching path data to the robot controller to realize the robot's teaching reproduction.

[0075] It is further explained that the image of the instructor's hand in the robot's workspace is acquired through an image sensor;

[0076] The particle filter algorithm is used to track the position of the instructor's hand in the continuous image sequence to extract the region of interest in the hand image; the particle filter algorithm further includes the following steps:

[0077] Initialize the particle collection;

[0078] 1.1) Particle filter algorithm initialization and tracking:

[0079] Initialize the particle set for the initial position estimation of the hand; each particle represents a possible hand position;

[0080] 1.2) Initialization of particle set:

[0081] Based on prior knowledge, particles are distributed in areas where the hand may appear;

[0082] Update the state of the particle through the prediction model;

[0083] 1.3) Status Update:

[0084] Use the prediction model to update the state of each particle. The specific formula is:

[0085]

[0086] in, is the state of the i′th particle, A is the state transfer matrix, B is the control input model, u t is the control input, w t is the process noise;

[0087] Calculate the weight of the particle according to the observation model;

[0088] 1.4) Weight calculation and resampling:

[0089] Calculate the weight of each particle according to the observation model:

[0090]

[0091] Among them, z t is the observed value, is the observation model, σ is the standard deviation of measurement noise;

[0092] Generate a new particle set by selecting high-weight particles through resampling;

[0093] According to the particle filtering results, the region of interest of the hand body and fingers is intercepted, and the region of interest is referred to as ROI;

[0094] Calculating the position and posture of the hand body in the camera coordinate system includes:

[0095] Several markers are set on the instructor's hand body, and key target feature points are selected from these markers. The corresponding relationship between the three-dimensional object coordinates and the two-dimensional image coordinates of these target feature points is determined by using the markers, and the position and posture of the hand body in the camera coordinate system are calculated by the P4P algorithm optimized by Dogleg; the connection between the hand body and the finger joints is used as the target feature point;

[0096] The marker is used to provide a reference for the instructor's hand feature points, and the P4P algorithm is used to recover the pose of the hand body in three-dimensional space from the two-dimensional image data; the P4P algorithm further includes the following steps:

[0097] Extracting landmark feature points in the image;

[0098] Establish the corresponding relationship between the feature points and the object coordinate system; specifically, use the perspective projection model to project the three-dimensional points onto the two-dimensional plane to connect the two-dimensional image coordinates with the three-dimensional object coordinates; the perspective projection model is specifically a pinhole camera model (PinholeCameraModel);

[0099] The Dogleg optimization method is used to solve the three-dimensional position and posture of the feature points; the definition formula is as follows:

[0100]

[0101] Among them, p is the pose parameter of the hand body, x i″ is the coordinate of the i″th feature point in the two-dimensional image, f is the projection function, X i″ is the three-dimensional coordinate of the i″th feature point;

[0102] The 3D model of the finger is built using OpenGL software, and the particle swarm optimization algorithm is used to estimate the angles of the finger joints to determine the accurate position and posture of the finger.

[0103] The established 3D finger model contains multiple joints and bone segments. The movement of each joint is controlled by the rotation angle. The specific steps are as follows:

[0104] 2.1) Finger joint parameter settings:

[0105] Define the rotation angle range of each joint. Each joint of the finger can rotate within the set range.

[0106] 2.2) Real-time rendering:

[0107] Use OpenGL for real-time rendering to display the current state of the fingers and dynamically adjust the joint angles based on input parameters;

[0108] 2.3) Initial model setting:

[0109] The initial joint angles of the fingers are set according to prior knowledge, so that the model is close to the shape of the actual fingers in the initial state;

[0110] For Particle Swarm Optimization:

[0111] 3.1) Particle initialization:

[0112] Initialize the particle swarm, where each particle i″′ represents a possible combination of finger joint angles;

[0113] Particle representation: Let each particle be p i″′ =[θ1,θ2,...,θ j′ ,...,θ n″′ ], where θ j′ is the rotation angle of the j′th joint;

[0114] Initial distribution: particles are evenly distributed within the set joint angle range;

[0115] 3.2)Fitness function definition:

[0116] Define the fitness function to evaluate the quality of particles;

[0117]

[0118] in, is the k′th feature point observed; is the current i″′ particle p i″′ The k′th feature point position of the lower model, m′ is the total number of feature points;

[0119] 3.3) Particle update:

[0120] Use the particle swarm optimization algorithm to update the particle position and gradually optimize the joint angle;

[0121] 3.4) Iteration stop condition:

[0122] Set the iteration stop condition, and stop when the change of the fitness function is less than the preset threshold or reaches the maximum number of iterations;

[0123] 3.5) Optimization result output:

[0124] Output the optimal particle joint angle combination as the final finger joint parameters.

[0125] It is further explained that the micro-vibration data is obtained by installing high-precision micro-vibration sensors on the hand body and fingers of the instructor for measurement; the micro-vibration data includes the micro-vibration frequency and amplitude, and the frequency and amplitude are recorded as f and ZA respectively;

[0126] Based on the teaching action sequence {1, 2, ..., i, ..., n}, the holding time of the i-th teaching action is marked as Ti, forming a holding time sequence {T1, T2, ..., Ti, ..., Tn}, where Tn represents the holding time of the n-th teaching action;

[0127] The total number of micro-vibration data collection for each teaching action in the corresponding action holding time Ti is set to m, forming a collection number sequence {1, 2, ..., j, ..., m} for each action holding time, where j represents the index of the jth collection in the current action holding time.

[0128] Further explanation: the generation of the vibration offset evaluation index specifically includes:

[0129] Set the vibration offset evaluation index to E i , which represents the vibration offset evaluation index of the i-th type of teaching action within its holding time Ti, and the calculation formula is as follows:

[0130]

[0131] Where m is the total number of micro-vibration data acquisitions for each teaching action at the corresponding action holding time Ti;

[0132] f i,j is the microseismic frequency collected at the jth time for the i-th type of teaching action; ZA i,j is the microseismic amplitude collected at the jth time for the i-th type of teaching action; f i,j and ZA i,j All the dimensions have been normalized and will not be described in detail.

[0133] max(f i,j ) is the maximum microseismic frequency of the i-th type of teaching action in all acquisitions;

[0134] max(ZA i,j ) is the maximum microseismic amplitude of the i-th type of teaching action in all acquisitions;

[0135] w j is the weight function, defined represents the weight of the jth acquisition; α and β are attenuation coefficients, which are used to adjust the influence of frequency and amplitude on the vibration offset evaluation index respectively, and the value range of α and β are both in the range of (0,1).

[0136] To further illustrate, let E i The valid value range of is between (0,1);

[0137] When E i The closer it is to 0, the smaller the micro-vibration of the i-th teaching action during the holding time, the more stable the teaching action, and the lower the offset; specifically:

[0138] If f i,j and ZA i,j are close to 0, the expression Also approaches 0, and Approaching 0, by adjusting the weight w j and attenuation coefficients α, β, so that E i Keep the trend close to 0;

[0139] When E i The closer it is to 1, the greater the micro-vibration of the i-th type of teaching action during the holding time, the more unstable the teaching action, and the higher the offset.

[0140] Further explanation: the generation of the correlation influence coefficient includes:

[0141] Assume the correlation coefficient is K i , represents the correlation between the electromyographic signal and the vibration offset evaluation index of the i-th type of teaching action when the vibration offset occurs; the calculation formula is as follows:

[0142]

[0143] Among them, A EMG,i,j is the amplitude of the electromyographic signal collected at the jth time for the i-th type of teaching action;

[0144] max(A EMG,i,j ) is the maximum EMG signal amplitude of the i-th type of teaching action in all acquisitions;

[0145] f EMG,i,j is the frequency of the electromyographic signal collected at the jth time for the i-th type of teaching action;

[0146] max(f EMG,i,j ) is the maximum EMG signal frequency of the i-th type of teaching action in all acquisitions;

[0147] γ and δ are attenuation coefficients, which are used to adjust the influence of the amplitude and frequency of the electromyographic signal on the correlation coefficient respectively; the value range of γ and δ is within the range of (0,1); set K iThe valid value range of is between (0,1);

[0148] When K i The closer it is to 1, the higher the correlation between the electromyographic signal and the vibration excursion evaluation index, and the greater the influence of muscle activity on vibration excursion during action execution;

[0149] When K i The closer it is to 0, the lower the correlation between the electromyographic signal and the vibration offset evaluation index, and the smaller the effect of muscle activity on vibration offset during action execution.

[0150] Further, generating a vibration compensation strategy for correcting the teaching path specifically includes:

[0151] When K i ≥0.65, and E i ≥0.75, for k i Make the following adjustments:

[0152]

[0153] When K i ≥0.65, and 0.46≤E i <0.75, for k i Make the following adjustments:

[0154]

[0155] When K i ≥0.65, and E i <0.46, for k i Make the following adjustments:

[0156]

[0157] When K i <0.65, and E i ≥0.75, for k i Make the following adjustments:

[0158]

[0159] When K i <0.65, and 0.46≤E i <0.75, for k i Make the following adjustments:

[0160]

[0161] When K i <0.65, and E i <0.46, for k i Make the following adjustments:

[0162]

[0163] Among them, k′ i is the compensation coefficient of the i-th teaching action after adjustment; η1 and η2 are both adjustment coefficients, and The specific values ​​of η1 and η2 are determined by the expert group based on experimental data.

[0164] Further explanation: the smoothed teaching path data is transmitted to the robot controller to realize the teaching reproduction of the robot, which specifically includes:

[0165] Use the homogeneous coordinate transformation matrix to transform the collected position and posture relationship information of the finger and hand body into the robot base coordinate system; the homogeneous coordinate transformation matrix is ​​obtained from the CSDN author hitgavin "Robot Forward Kinematics---Three Interpretations of Homogeneous Transformation Matrix" or related official documents, which will not be elaborated on;

[0166] Discretize the teaching path into a series of teaching points and record the position and posture of each point;

[0167] Obtain the path point data corrected by the vibration compensation strategy and apply a median filter to remove noise from the collected data and smooth the path;

[0168] The smoothed teaching path data is transmitted to the robot controller through the communication interface;

[0169] The robot controller calls the received teaching path data and executes teaching playback.

[0170] Embodiment 2:

[0171] See also Figure 2 , an industrial robot teaching system, the system is used to execute the industrial robot teaching method, comprising:

[0172] Teaching path determination module: collects the image sequence of the instructor's hand in the robot workspace and extracts the teaching path from the image sequence;

[0173] Division module: used to divide the teaching path according to the action type to form different teaching action sequences {1, 2, ..., i, ..., n}, where i represents the index of the i-th teaching action, and n represents the total number of different teaching actions, and collects the micro-vibration data of each teaching action at the corresponding action holding time in real time;

[0174] Vibration offset evaluation index generation module: used to analyze the micro-vibration data of each teaching action at the corresponding action holding time, and generate a vibration offset evaluation index for evaluating the vibration offset amount of each teaching action at the corresponding action holding time;

[0175] Correlation influence coefficient generation module: used to collect the electromyographic signal of each action when the vibration offset occurs, and to correlate and analyze the electromyographic signal with the vibration offset evaluation index to obtain the correlation influence coefficient. The electromyographic signal includes the electromyographic signal amplitude and the electromyographic frequency.

[0176] Vibration compensation strategy generation module: used to fit and analyze the teaching path, vibration offset evaluation index and associated influence coefficient to generate a vibration compensation strategy for correcting the teaching path, and define the vibration compensation formula of the teaching path as follows:

[0177]

[0178] Among them, p t,i is the corrected path point of the i-th teaching action in the teaching path at time step t, is the original path point of the i-th teaching action at time step t, k i is the compensation coefficient of the i-th type of teaching action, Δd i is the vibration displacement of the i-th type of teaching action;

[0179] Teaching and reproduction module: used to convert the posture relationship information of fingers and hand body into the robot base coordinate system, discretely record the posture of teaching points in the continuous teaching path, and perform median filtering on the corrected path point data to obtain a smooth teaching path; transmit the smoothed teaching path data to the robot controller to realize the robot's teaching reproduction.

[0180] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0181] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0182] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0183] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. An industrial robot teaching method, characterized in that: The specific steps include: Step S1: collecting an image sequence of the instructor's hand in the robot workspace and extracting the teaching path from the image sequence; Step S2: Divide the teaching path according to the action type to form different teaching action sequences {1, 2, ..., i, ..., n}, where i represents the index of the i-th teaching action, and n represents the total number of different teaching actions, and collect the micro-vibration data of each teaching action at the corresponding action holding time in real time; Step S3: analyzing the micro-vibration data of each teaching action at the corresponding action holding time, and generating a vibration offset evaluation index for evaluating the vibration offset amount of each teaching action at the corresponding action holding time; Step S4: collecting the electromyographic signal of each teaching action when the vibration offset occurs, and performing correlation analysis on the electromyographic signal and the vibration offset evaluation index to obtain the correlation influence coefficient, the electromyographic signal includes the electromyographic signal amplitude and the electromyographic frequency; Step S5: Fitting and analyzing the teaching path, the vibration offset evaluation index and the associated influence coefficient to generate a vibration compensation strategy for correcting the teaching path, and defining the vibration compensation formula of the teaching path as follows: Among them, p t,i is the corrected path point of the i-th teaching action in the teaching path at time step t, is the original path point of the i-th teaching action at time step t, k i is the compensation coefficient of the i-th type of teaching action, Δd i is the vibration displacement of the i-th type of teaching action; Step S6: Convert the posture relationship information of the fingers and the hand body to the robot base coordinate system, discretely record the posture of the teaching points in the continuous teaching path, and perform median filtering on the corrected path point data to obtain a smooth teaching path; transmit the smoothed teaching path data to the robot controller to realize the robot's teaching reproduction.

2. An industrial robot teaching method according to claim 1, characterized in that: The acquisition of the teaching path includes: Several markers are set on the instructor's hand body, and key target feature points are selected from these markers. The corresponding relationship between the three-dimensional object coordinates and the two-dimensional image coordinates of these target feature points is determined by using the markers. The position and posture of the hand body in the camera coordinate system are calculated by the P4P algorithm optimized by Dogleg; the connection between the hand body and the finger joints is used as the target feature point; The P4P algorithm is used to recover the pose of the hand subject in three-dimensional space from two-dimensional image data; The three-dimensional model of the finger is built using OpenGL software, and the particle swarm optimization algorithm is used to estimate the angles of the finger joints to determine the position and posture of the finger and obtain the teaching path.

3. An industrial robot teaching method according to claim 2, characterized in that: The micro-vibration data is obtained by installing high-precision micro-vibration sensors on the hand body and fingers of the instructor. The micro-vibration data includes the micro-vibration frequency and amplitude, and the frequency and amplitude are recorded as f and ZA respectively. Based on the teaching action sequence {1, 2, ..., i, ..., n}, the holding time of the i-th teaching action is marked as Ti, forming a holding time sequence {T1, T2, ..., Ti, ..., Tn}, where Tn represents the holding time of the n-th teaching action; The total number of micro-vibration data collection for each teaching action in the corresponding action holding time Ti is set to m, forming a collection number sequence {1, 2, ..., j, ..., m} for each action holding time, where j represents the index of the jth collection in the current action holding time.

4. An industrial robot teaching method according to claim 3, characterized in that: The generation of the vibration offset evaluation index includes: Set the vibration offset evaluation index to E i , which represents the vibration offset evaluation index of the i-th type of teaching action within its holding time Ti, and the calculation formula is as follows: Where m is the total number of micro-vibration data acquisitions for each teaching action at the corresponding action holding time Ti; f i,j is the microseismic frequency collected at the jth time for the i-th type of teaching action; ZA i,j is the microseismic amplitude collected at the jth time for the i-th type of teaching action; max(f i,j ) is the maximum microseismic frequency of the i-th type of teaching action in all acquisitions; max(ZA i,j ) is the maximum microseismic amplitude of the i-th type of teaching action in all acquisitions; w j is the weight function, defined represents the weight of the jth acquisition; α and β are attenuation coefficients, which are used to adjust the influence of frequency and amplitude on the vibration offset evaluation index respectively, and the value range of α and β are both in the range of (0,1).

5. An industrial robot teaching method according to claim 4, characterized in that: Setting E i The valid value range of is between (0,1); When E i The closer it is to 0, the smaller the micro-vibration of the i-th teaching action during the holding time, the more stable the teaching action, and the lower the offset; When E i The closer it is to 1, the greater the micro-vibration of the i-th type of teaching action during the holding time, the more unstable the teaching action, and the higher the offset.

6. An industrial robot teaching method according to claim 5, characterized in that: The generation of the correlation influence coefficient includes: Assume the correlation coefficient is K i , represents the correlation between the electromyographic signal and the vibration offset evaluation index of the i-th type of teaching action when the vibration offset occurs; the calculation formula is as follows: Among them, A EMG,i,j is the amplitude of the electromyographic signal collected at the jth time for the i-th type of teaching action; max(A EMG,i,j ) is the maximum EMG signal amplitude of the i-th type of teaching action in all acquisitions; f EMG,i,j is the frequency of the electromyographic signal collected at the jth time for the i-th type of teaching action; max(f EMG,i,j ) is the maximum EMG signal frequency of the i-th type of teaching action in all acquisitions; γ and δ are attenuation coefficients, which are used to adjust the influence of the amplitude and frequency of the electromyographic signal on the correlation coefficient respectively; the value range of γ and δ is within the range of (0,1); set K i The valid value range of is between (0,1); When K i The closer it is to 1, the higher the correlation between the electromyographic signal and the vibration excursion evaluation index, and the greater the influence of muscle activity on vibration excursion during action execution; When K i The closer it is to 0, the lower the correlation between the electromyographic signal and the vibration offset evaluation index, and the smaller the effect of muscle activity on vibration offset during action execution.

7. An industrial robot teaching method according to claim 6, characterized in that: Generate a vibration compensation strategy for correcting the taught path, including: When K i ≥0.65, and E i ≥0.75, for k i Make the following adjustments: When K i ≥0.65, and 0.46≤E i <0.75, for k i Make the following adjustments: When K i ≥0.65, and E i <0.46, for k i Make the following adjustments: When K i <0.65, and E i ≥0.75, for k i Make the following adjustments: When K i <0.65, and 0.46≤E i <0.75, for k i Make the following adjustments: When K i <0.65, and E i <0.46, for k i Make the following adjustments: Among them, k′ i is the compensation coefficient of the i-th teaching action after adjustment; η1 and η2 are both adjustment coefficients, and 8. An industrial robot teaching method according to claim 7, characterized in that: The smoothed teaching path data is transmitted to the robot controller to realize the robot's teaching reproduction, which includes: The collected position and posture relationship information of the fingers and the hand body is converted into the robot base coordinate system using a homogeneous coordinate transformation matrix; Discretize the teaching path into a series of teaching points and record the position and posture of each point; Obtain the path point data corrected by the vibration compensation strategy and apply a median filter to remove noise from the collected data and smooth the path; The smoothed teaching path data is transmitted to the robot controller through the communication interface; The robot controller calls the received teaching path data and executes teaching playback.

9. An industrial robot teaching system, characterized in that: The system is used to execute the industrial robot teaching method according to any one of claims 1 to 8, comprising: Teaching path determination module: collects the image sequence of the instructor's hand in the robot workspace and extracts the teaching path from the image sequence; Division module: used to divide the teaching path according to the action type to form different teaching action sequences {1, 2, ..., i, ..., n}, where i represents the index of the i-th teaching action, and n represents the total number of different teaching actions, and collects the micro-vibration data of each teaching action at the corresponding action holding time in real time; Vibration offset evaluation index generation module: used to analyze the micro-vibration data of each teaching action at the corresponding action holding time, and generate a vibration offset evaluation index for evaluating the vibration offset amount of each teaching action at the corresponding action holding time; Correlation influence coefficient generation module: used to collect the electromyographic signal of each action when the vibration offset occurs, and to correlate and analyze the electromyographic signal with the vibration offset evaluation index to obtain the correlation influence coefficient. The electromyographic signal includes the electromyographic signal amplitude and the electromyographic frequency. Vibration compensation strategy generation module: used to fit and analyze the teaching path, vibration offset evaluation index and associated influence coefficient to generate a vibration compensation strategy for correcting the teaching path, and define the vibration compensation formula of the teaching path as follows: Among them, p t,i is the corrected path point of the i-th teaching action in the teaching path at time step t, is the original path point of the i-th teaching action at time step t, k i is the compensation coefficient of the i-th type of teaching action, Δd i is the vibration displacement of the i-th type of teaching action; Teaching and reproduction module: used to convert the posture relationship information of fingers and hand body into the robot base coordinate system, discretely record the posture of teaching points in the continuous teaching path, and perform median filtering on the corrected path point data to obtain a smooth teaching path; transmit the smoothed teaching path data to the robot controller to realize the robot's teaching reproduction.

Images