Dragging teaching method and control system based on nonlinear kinetic model

Abstract

The invention provides a dragging teaching method and control system based on a nonlinear dynamic model, and the method comprises the steps: collecting a real-time robot state in a dragging teaching process, inputting the real-time robot state into a nonlinear dynamic model, the nonlinear dynamic model calculates real-time driving torques of all joints of the robot in the robot spraying process according to the real-time robot state, and a set of corresponding torque control instructions are generated according to the real-time driving torques of all the joints of the robot; establishing a corresponding relation between the torque control instruction and the path of the dragging teaching process; standard spraying path data are generated, and the obtained standard spraying path data are finally used for automatic spraying; through the nonlinear friction force item of the nonlinear kinetic model, the generated standard spraying path can solve the problem that point positions where the spraying directions of all joints of the robot change in the spraying process are paused due to sudden change of the torque direction, so that the accuracy of automatic spraying is higher, and the flexibility is better.

Description

Technical Field

[0001] The present invention relates to the field of automation control, and in particular to the technical field of a dragging teaching method and a control system based on a nonlinear dynamics model. Background Art

[0002] With the continuous development of science and technology, automated control is used more and more in industry. In the industrial production link of workpiece spraying, general programming teaching spraying is usually used to control the robot to automatically complete the workpiece spraying. However, it has the problem of complex programming. When the product changes, it requires another set of codes to adapt to the new product. Therefore, this method is not suitable for production environments where products change rapidly.

[0003] There is another automated control method for automated spraying robots - the drag teaching method. This method uses human-machine collaboration to assist the robot in quickly learning automatic movement in different scenarios to meet the automatic spraying requirements in different scenarios. However, this method is slightly insufficient for the control accuracy of the robot, which can easily lead to poor results in the robot spraying workpieces. Summary of the invention

[0004] In view of the problem that the existing drag-based teaching method has low accuracy of the spray path data generated in the sensorless state and there is a pause in the spray steering process, resulting in low accuracy of the automatic spraying effect, the present technical solution provides a drag teaching method and control system based on a nonlinear dynamic model; a nonlinear dynamic model is used to calculate the joint torque and friction force of the robot in the drag teaching process, thereby achieving zero-force spraying in the drag teaching process, so that the robot can maintain stability and accuracy in the automatic spraying process according to the generated spray path data, thereby achieving comprehensive and precise coating of the spray surface and reducing waste and environmental pollution during the spraying process.

[0005] The present invention is achieved through the following technical solutions: On the one hand, the present invention proposes a drag teaching method based on a dynamic model, comprising: S11: collecting the real-time robot state during the drag teaching process and inputting it into a nonlinear dynamic model; S12: the nonlinear dynamics model calculates the real-time driving torque of each joint of the robot during the spraying process according to the real-time state of the robot; The nonlinear dynamic model is expressed as follows:

[0006]

[0007] in, represents the driving torque of the robot joint; is the angular displacement of each joint of the robot, is the angular velocity of each joint of the robot, is the angular acceleration of each joint of the robot; represents the inertia moment term; represents the Coriolis moment term; represents the gravitational moment term; , , Composition of frictionless dynamics model; represents the friction torque model; is the Coulomb friction coefficient; is the static friction coefficient; is the Stribeck speed threshold; is the empirical constant of Stribeck curve attenuation; is the viscous friction coefficient; S13: Generate a set of corresponding torque control instructions according to the real-time driving torque of each joint of the robot; establish a corresponding relationship between the torque control instructions and the path of the dragging teaching process; and generate standard spraying path data.

[0008] Furthermore, the Coulomb friction coefficient and the viscous friction coefficient According to the high-speed dynamic parameter identification method, the Stribeck speed threshold and the static friction coefficient The real friction torque of the robot when running at low speed is solved according to the frictionless dynamics model, and then combined with and Substitution , and finally optimized by genetic algorithm.

[0009] Furthermore, the frictionless dynamics model and the Coulomb friction coefficient of the nonlinear dynamics model and viscous friction coefficient It is obtained through the following steps: S21: Initializing the initial dynamics model of the robot according to the parameters of the robot. At this time, the initial dynamics model is:

[0010]

[0011] in, is the angular displacement of each joint of the robot, is the angular velocity of each joint of the robot, is the angular acceleration of each joint of the robot; represents the inertia moment term; represents the Coriolis moment term; represents the gravitational moment term, is the Coulomb friction coefficient; is the viscous friction coefficient; S22: linearizing the initial dynamics model of the robot to obtain a linear friction dynamics model; S23: Under preset constraint conditions, calculating the excitation trajectory coefficient using the linear friction dynamics model; S24: Collect the robot motion state when the robot runs the high-speed excitation trajectory under the excitation trajectory coefficient, and estimate the dynamic parameters of the linear friction dynamics model offline by the least squares method to obtain the frictionless dynamics model and the Coulomb friction coefficient. and viscous friction coefficient .

[0012] Furthermore, the Stribeck speed threshold and static friction coefficient In order to obtain the frictionless dynamics model, Coulomb friction coefficient and viscous friction coefficient Then obtained through the following steps: S25: collecting the motion state of each joint of the robot and the real torque data of each joint of the robot when the robot runs the low-speed excitation trajectory, and combining the frictionless dynamics model to obtain the real friction force in the low-speed state; S26: The actual friction force and the Coulomb friction coefficient and viscous friction coefficient Substituting into the nonlinear friction torque model, a plurality of groups of preliminary Stribeck speed thresholds and static friction coefficients are obtained, and the preliminary Stribeck speed thresholds and static friction coefficients are optimized by genetic algorithm to obtain the Stribeck speed thresholds and static friction coefficients; The nonlinear friction torque model is expressed as:

[0013] in, is the angular velocity of each joint of the robot, is the static friction coefficient, is the Coulomb friction coefficient, is the viscous friction coefficient, is the Stribeck velocity threshold, is the empirical constant of Stribeck curve attenuation.

[0014] Furthermore, after obtaining the frictionless dynamic model, Coulomb-viscous friction coefficient, Stribeck velocity threshold and static friction coefficient, a nonlinear dynamic model is generated and the error coefficient of the nonlinear dynamic model is calculated by the following formula: :

[0015]

[0016]

[0017] in, is the real friction force, is the true torque, It is a frictionless dynamics model; represents the moment of inertia; represents the Coriolis moment; represents the gravitational moment, represents the driving torque of the robot joint, i represents the robot state parameter; If the error coefficient is greater than the preset threshold, the Stribeck velocity threshold and the static friction coefficient are continuously optimized by genetic algorithm iteration until the error coefficient is less than the preset threshold.

[0018] On the other hand, the present invention also provides a drag teaching control system, which includes: A data acquisition module collects the real-time robot status during the drag teaching process and inputs it into a nonlinear dynamics model; A nonlinear dynamics model, used to calculate the real-time driving torque of each joint of the robot during the spraying process according to the nonlinear dynamics model and the real-time state of the robot; The nonlinear dynamic model is expressed as follows:

[0019]

[0020] in, represents the driving torque of the robot joint; is the angular displacement of each joint of the robot, is the angular velocity of each joint of the robot, is the angular acceleration of each joint of the robot; represents the inertia moment term; represents the Coriolis moment term; represents the gravitational moment term; , , Composition of frictionless dynamics model; represents the friction torque model; is the Coulomb friction coefficient; is the static friction coefficient; is the Stribeck speed threshold; is the empirical constant of Stribeck curve attenuation; is the viscous friction coefficient; The standard path data generation module generates a set of corresponding torque control instructions according to the real-time driving torque of each joint of the robot; establishes a mapping relationship between the torque control instructions and the path of the dragging teaching process; and generates standard spraying path data.

[0021] On the other hand, the present invention also provides an automatic spraying control device, which comprises: the above-mentioned drag teaching control system; Pose transformation matrix generation module: used to match the image data of the workpiece to be sprayed with a template point cloud to obtain a pose transformation matrix; Spraying path generation module: used to transform the standard spraying path data according to the posture transformation matrix to obtain real-time spraying path data; Spraying controller: controls the robot to complete automatic spraying according to real-time spraying path data.

[0022] On the other hand, the present invention also provides a model construction system for a nonlinear dynamic model, comprising: Robot dynamics model building module: used to initialize the robot's initial dynamics model according to the robot's parameters. At this time, the robot's initial dynamics model is:

[0023]

[0024] in, is the angular displacement of each joint of the robot, is the angular velocity of each joint of the robot, is the angular acceleration of each joint of the robot; represents the inertia moment term; represents the Coriolis moment term; represents the gravitational moment term, is the Coulomb friction coefficient; is the viscous friction coefficient; Linearization processing module: used for performing linearization processing on the initial dynamics model of the robot to obtain a linear friction dynamics model; Excitation trajectory coefficient solving module: under preset constraint conditions, the excitation trajectory coefficient is calculated by the linear friction dynamics model; High-speed parameter identification module: used to collect the robot motion state when the robot runs the high-speed excitation trajectory under the excitation trajectory coefficient, and estimate the dynamic parameters of the linear friction dynamics model offline through the least squares method to obtain the frictionless dynamics model and Coulomb friction coefficient. and viscous friction coefficient .

[0025] Furthermore, the model building system also includes: Low-speed friction calculation module: used to collect the motion state of each joint of the robot and the real torque data of each joint of the robot when the robot runs the low-speed excitation trajectory, and combine the frictionless dynamics model to obtain the real friction force in the low-speed state; Nonlinear parameter genetic optimization module: used to transform the real friction force, Coulomb friction coefficient and viscous friction coefficient Substituting into the nonlinear friction torque model, a plurality of groups of preliminary Stribeck speed thresholds and static friction coefficients are obtained, and the preliminary Stribeck speed thresholds and static friction coefficients are optimized by genetic algorithm to obtain the Stribeck speed thresholds and static friction coefficients; The nonlinear friction torque model is expressed as:

[0026] in, is the angular velocity of each joint of the robot, is the static friction coefficient, is the Coulomb friction coefficient, is the viscous friction coefficient, is the Stribeck velocity threshold, is the empirical constant of Stribeck curve attenuation.

[0027] Furthermore, the model building system also includes: Error coefficient calculation module: used to calculate the error coefficient of the nonlinear dynamics model using the following formula :

[0028]

[0029]

[0030] in, is the real friction force, is the true torque, It is a frictionless dynamics model; represents the moment of inertia; represents the Coriolis moment; represents the gravitational moment, represents the driving torque of the robot joint, i Indicates robot status parameters; Error coefficient determination module: if the error coefficient is greater than a preset threshold, the genetic algorithm is continued to be iteratively optimized for the Stribeck speed threshold and the static friction coefficient until the error coefficient is less than the preset threshold.

[0031] Based on the characteristic that the friction force of each joint of the robot is greatly affected by nonlinear influence during the dragging teaching process, the present invention uses a nonlinear friction model as the friction model of the robot dynamics model; and discovers the sensitive relationship between the coefficient of the nonlinear friction model and the operating speed of the robot. Based on the correlation between the coefficient of the friction model and the operating speed of the robot, the Coulomb-viscous friction coefficient of the robot under high-speed operation to the linear friction model and the static friction coefficient and Stribeck speed threshold of the robot under low-speed operation are solved step by step. Finally, the obtained linear dynamic model, Coulomb-viscous friction coefficient, static friction coefficient and Stribeck speed threshold are integrated to obtain a high-precision nonlinear dynamic model. The real-time driving torque of each joint of the robot during the robot dragging teaching process is calculated according to the obtained nonlinear dynamic model; and a set of corresponding torque control instructions are generated according to the real-time driving torque of each joint of the robot; then a mapping relationship between the torque control instruction and the path of the dragging teaching process is established; standard spraying path data is generated, and automatic spraying is performed through the standard spraying path data, thereby ensuring the stability and accuracy of the automatic spraying control system; finally, sensorless zero-force dragging teaching is realized, so that the generated spraying path data can improve the flexibility and accuracy of automatic spraying, and reduce waste and environmental pollution during the spraying process.

[0032] For better understanding and implementation, the present invention is described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 A flowchart of a dragging teaching method based on a dynamic model provided by the present invention; Figure 2 For execution Figure 1 The structural block diagram of the spray control system of the drag teaching method shown; Figure 3 A flowchart of a method for constructing a nonlinear dynamics model provided by the present invention; Figure 4 For execution Figure 3 A structural block diagram of a model building system of the model building method shown; Figure 5 A structural block diagram of an exemplary automatic spraying control device provided by the present invention. DETAILED DESCRIPTION

[0034] In order to solve the problems of low automation level of spraying workpieces in domestic small and medium-sized enterprises and complex programming of general programming teaching spraying robots, which are not suitable for production needs of rapid product changes, an automatic spraying system based on zero-force control dragging is proposed. The teaching path of the six-degree-of-freedom spraying robot is realized by dragging teaching based on zero-force control, and the robot spraying path is autonomously corrected in combination with the object posture information fed back by the depth camera. The collaborative robot zero-force spraying system can achieve comprehensive coating and precise painting of the spraying surface, reducing waste and environmental pollution during the spraying process.

[0035] The key point is that when the existing robot dynamics model is used to calculate the torque term, the robot's joints are greatly affected by nonlinear forces at low speeds, and the dragging teaching process cannot be well described. Therefore, the present invention introduces nonlinear terms in the friction force to construct a dynamics model, and uses the nonlinear dynamic model to calculate the real-time torque data of the robot during the dragging teaching process. The torque output is combined with the robot's terminal path to generate standard spraying path data, and the subsequent automatic spraying task is completed based on the standard spraying path data to ensure the state stability and continuity of the robot during the automatic spraying process and improve the spraying effect.

[0036] See also Figure 1 and Figure 2 , Figure 1 A flowchart of a dragging teaching method based on a nonlinear dynamics model provided by the present invention, Figure 2 For execution Figure 1 The drag teaching control system of the drag teaching method shown in the figure comprises a data acquisition unit 11 , a nonlinear dynamics model 12 and a standard path generation module 13 .

[0037] The data acquisition unit 11 is used to perform step S11: collect the real-time robot state during the drag teaching process and input it into a nonlinear dynamic model 12; The nonlinear dynamics model 12 is used to execute step S12: calculating the real-time driving torque of each joint of the robot during the spraying process according to the real-time state of the robot; The torque drive controller 13 is used to execute step S13: generate a set of corresponding torque control instructions according to the real-time driving torque of each joint of the robot; establish a mapping relationship between the torque control instructions and the path of the drag teaching process; and generate standard spraying path data. Specifically, in step S11, it includes: First, create a point cloud template for drag teaching, fix the position of the workpiece to be sprayed, and place the workpiece in the center of the Meka-Mind Pro M Enhanced industrial camera. After the position is adjusted, perform the following steps: (1) Image acquisition and point cloud generation Before image acquisition, it is necessary to first frame the plane of the current workbench as a reference for the depth of the subsequent point cloud module. After framing, take a depth map of the workpiece to be sprayed and generate a point cloud template for the sprayed workpiece.

[0038] (2) Point cloud preprocessing Point cloud preprocessing mainly involves manually setting the scope of the target area of ​​the point cloud template, determining the region of interest (ROI) in the point cloud template, and then extracting the target in the 3D ROI. The software interface for setting the scope of the target area and the extracted target workpiece template.

[0039] After the workpiece template is established, the handheld robot terminal performs the dragging teaching task. During the dragging process, the states of the robot joints change. The data acquisition unit 11 collects real-time robot state data during the dragging teaching process, such as the angular displacement, angular velocity and angular acceleration of each joint, and the world coordinates of the robot terminal, and transmits it to the nonlinear dynamic model that has completed parameter identification. The nonlinear dynamic model is represented as follows: (1) (2) in, represents the driving torque of the robot joint; is the angular displacement of each joint of the robot, is the angular velocity of each joint of the robot, is the angular acceleration of each joint of the robot; represents the inertia moment term; represents the Coriolis moment term; represents the gravitational moment term; , , Composition of frictionless dynamics model; represents the friction torque model; is the Coulomb friction coefficient; is the static friction coefficient; is the Stribeck speed threshold; is the empirical constant of Stribeck curve attenuation; is the viscous friction coefficient.

[0040] in, and According to the high-speed dynamics parameter identification method, and The real friction torque of the robot when running at low speed is solved according to the frictionless dynamics model, and then combined with and Substitution , and finally obtained through genetic algorithm optimization. At this time, the real-time torque data of each joint of the robot is calculated through the above nonlinear dynamics model.

[0041] The present invention takes a six-degree-of-freedom industrial robot as a research object. There are two main methods for dynamic modeling of a six-degree-of-freedom industrial robot: Newton-Euler method and Lagrange method. Considering the convenience of programming, the present invention adopts Newton-Euler method to perform dynamic modeling of the robot, and uses Stribeck model as the friction torque model of the nonlinear dynamic model. The Stribeck model takes into account the discontinuous separation force and nonlinear Stribeck effect during stick-slip transition, and can better fit the robot joint friction at low speed.

[0042] The parameters of the frictionless dynamics model, the Coulomb-viscous friction coefficient, the static friction coefficient and the Stribeck speed threshold are obtained through the parameter identification method and substituted into formulas (1) and (2). The various parameters of the solved nonlinear dynamics model can accurately describe the change of torque output under different robot motion states. That is, as long as the motion state data of each joint of the robot is collected in real time, the corresponding torque data required to be output by each joint of the robot can be calculated.

[0043] For the nonlinear dynamic model, the more its coefficients match the dynamic parameters of the robot, the smaller the error of the robot's zero-force control will be. Taking the derivative of each coefficient of , we get: (3) (4) (5) (6) In equations (3) to (6), , The partial derivative of increases with the increase of and The partial derivative of decreases with the increase of , , The partial derivatives of Item, It decays exponentially with the increase of angular velocity, which makes the coefficient containing this term extremely sensitive to velocity. and Insensitive to low speed but sensitive to high speed, and On the contrary, exponential decay leads to and The influence on the friction force is approximately 0. Based on this, the present invention considers identifying different coefficients of the friction model through step-by-step identification processing to improve the accuracy of the model.

[0044] Based on the above analysis, the present invention also proposes a method for constructing a nonlinear dynamic model and a construction system thereof, please refer to Figure 3 and Figure 4 , Figure 3 A flowchart of a method for constructing a nonlinear dynamics model provided by the present invention, Figure 4 For execution Figure 3 The structural block diagram of the model building system of the model building method shown in the figure, wherein the model building system includes: a robot dynamics model building module 21, a linearization processing module 22, an excitation trajectory coefficient solving module 23, a high-speed parameter identification module 24, a low-speed friction force calculation module 25 and a nonlinear parameter genetic optimization module 26. The model building system is used to execute the construction method of the nonlinear dynamics model, and the functions of each module are as follows: The robot dynamics model building module 21 is used to execute step S21: to initialize the initial dynamics model of the robot according to the parameters of the robot; The various parts of the robot used for spraying are equivalently simplified, and then the dynamics modeling of each joint is carried out to obtain the equivalent simplified robot dynamics model. The parameter description at each joint includes the torque term and the friction term. In order to facilitate parameter identification, the linear term of the friction torque is solved first, so the nonlinear coefficient of the Stribeck model is ignored.

[0045] The linearization processing module 22 is used to execute step S22: linearize the initial dynamics model of the robot to obtain a linear friction dynamics model; According to the above analysis, when the robot runs at high speed and The effect on friction is approximately 0, so the initialized dynamic model is equivalent to replacing Approximately: (7) When the When it is expressed in formula (7), it is linearized and obtained (8) in, for The unknown parameter set of the connecting rod is n, and n is the number of connecting rods. The unknown parameter set of the connecting rod is mainly composed of 12 parameters including inertia tensor, first-order mass moment, connecting rod mass, Coulomb friction coefficient and viscous friction coefficient. is the observation matrix corresponding to the unknown parameter set.

[0046] After completing the linearization process, the unknown parameter set in equation (8) is reconstructed by the inertial parameter reorganization method. Processing to obtain the minimum parameter set , and find and The transformation relationship matrix C between them makes: (9) Substituting formula (9) into formula (8), we get: (10) in, .

[0047] The excitation trajectory coefficient solving module 23 is used to execute step S23: under preset constraint conditions, the excitation trajectory coefficient is calculated using the linear friction dynamics model.

[0048] Through linearization processing The linear terms of The remaining items of are transformed by the fifth-order Fourier transform, and the excitation trajectory design is carried out according to the transformation results. The excitation trajectory design is a special trajectory designed to identify the dynamic characteristics of the system. Its main principle is to collect data when the system runs a specific trajectory, ensuring that the collected data has rich system characteristics, so that the identified parameters are more accurate and reasonable.

[0049] The high-speed parameter identification module 24 is used to execute step S24: collect the robot motion state when the robot runs the high-speed excitation trajectory under the excitation trajectory coefficient, and estimate the dynamic parameters of the linear friction dynamic model offline by the least square method to obtain the frictionless dynamic model and the Coulomb friction coefficient. and viscous friction coefficient .

[0050] The robot runs the excitation trajectory and collects the angular displacement, angular velocity, angular acceleration and true torque of each joint at the same time. The minimum parameter set Xb is identified offline based on the collected data to obtain the data that meets the conditions in the minimum parameter set. These data that meet the conditions are substituted into formula (10) to obtain the frictionless dynamics model and Coulomb friction coefficient. and viscous friction coefficient .

[0051] The low-speed friction calculation module 25 is used to execute step S25: collecting the motion state of each joint of the robot and the real torque data of each joint of the robot when the robot runs the low-speed excitation trajectory, and combining the frictionless dynamics model to obtain the real friction under the low-speed state.

[0052] According to formula (1), the robot dynamics model includes a frictionless dynamics model and a friction item. In step S24, the robot parameters under high-speed operation have been identified and a frictionless dynamics model with good performance has been solved. Therefore, it is only necessary to subtract the non-friction torque value solved by the frictionless dynamics model from the measured real torque data, and the remaining torque value is the real friction.

[0053] The nonlinear parameter genetic optimization module 26 is used to execute step S26: convert the real friction force and the Coulomb friction coefficient and viscous friction coefficient Substituting into the nonlinear friction torque model, a plurality of groups of preliminary Stribeck speed thresholds and static friction coefficients are obtained, and the preliminary Stribeck speed thresholds and static friction coefficients are optimized by genetic algorithm to obtain the Stribeck speed thresholds and static friction coefficients.

[0054] According to formula (3)-formula (6), when running at low speed, the main influencing items in the Stribeck friction model are and , combined with formula (2):

[0055] Then the , Substituting the above formula, there are multiple sets of nonlinear equations according to the actual friction force under different excitation trajectories. Since there are exponential terms in the equations, they cannot be solved. Therefore, the genetic algorithm is used to solve the equations. and Perform iterative optimization to find the optimal static friction coefficient and Stribeck velocity threshold .

[0056] In another preferred embodiment, in order to ensure the identification results of the nonlinear dynamic model parameters, the present invention also uses the residual root mean square (RMS) of the theoretical value and the actual value as a criterion to judge whether the identification effect of the Stribeck friction model meets expectations and controls the end of the genetic algorithm optimization based on the residual root mean square (RMS) as the error coefficient. Specifically, the error coefficient is solved by the following formula: : (13) in, is the true torque, is the theoretical torque, is the residual root mean square.

[0057] If the error coefficient is greater than the preset threshold, the genetic algorithm will continue to iteratively optimize the Stribeck speed threshold and the static friction coefficient until the error coefficient is less than the preset threshold. The reliability of the model parameters is ensured by parameter verification and repeated genetic algorithm screening process. The real-time torque data of each joint of the robot obtained by the nonlinear dynamic model after the above construction process is more accurate.

[0058] After obtaining the real-time torque data of each joint of the robot, the torque drive controller can control each joint of the robot to generate a torque of corresponding magnitude according to the real-time torque data of each joint of the robot, balance the internal torque of each joint of the robot, so as to achieve zero-force control of the robot; and record the real-time torque data of each joint of the robot during the drag teaching process. Therefore, according to the change of the world coordinates of the robot end combined with the real-time torque data of each joint of the robot, the standard spraying path of the workpiece to be sprayed can be obtained. The robot can restore the spraying path during the drag teaching by performing the corresponding end movement task. Due to the existence of nonlinear terms, the operation of each joint of the robot is also flexible when turning or stopping, and the output torque value will not change suddenly like when only linear terms exist, so that when performing the spraying task, the workpiece to be sprayed can be sprayed evenly.

[0059] Based on the above nonlinear dynamic model drag teaching control system, the present invention also provides an automatic spraying control device, please refer to Figure 5 , which includes the above-mentioned drag teaching control system, and a posture transformation matrix generation module, a spraying path generation module and a spraying controller.

[0060] Pose transformation matrix generation module: used to match the image data of the workpiece to be sprayed with a template point cloud to obtain a pose transformation matrix; Specifically, the posture transformation matrix is ​​obtained through the following steps: The image data of the workpiece to be sprayed is matched with a template point cloud to obtain a pose transformation matrix.

[0061] The posture transformation matrix is ​​obtained by the following steps: An image of the sprayed workpiece is acquired, and a three-dimensional point cloud is generated according to the image of the sprayed workpiece.

[0062] The three-dimensional point cloud data is preprocessed to extract target point cloud features.

[0063] The target point cloud feature is roughly matched with a template point cloud to obtain a first matching feature.

[0064] The first matching feature is precisely matched with the template point cloud to obtain a three-dimensional point cloud, and a posture transformation matrix is ​​obtained according to the matching result.

[0065] Spraying path generation module: used to transform the standard spraying path data according to the posture transformation matrix to obtain real-time spraying path data; Spraying controller: controls the robot to complete automatic spraying according to real-time spraying path data.

[0066] The workpiece offset is corrected through the posture transformation matrix, and the standard spray path data is adjusted so that the robot can start the spraying task from the correct starting point. Combined with the torque data of the standard spray path data, each joint of the robot outputs the corresponding torque and starts spraying according to the preset spray path. Through the above processing, even if the corresponding position of the workpiece is offset, the automatic spray control system can perform self-correction to ensure the accuracy of spraying.

[0067] Based on the characteristic that the friction force of each joint of the robot is greatly affected by nonlinear influence during the dragging teaching process, the present invention uses a nonlinear friction model as the friction model of the robot dynamics model; and discovers the sensitive relationship between the coefficient of the nonlinear friction model and the operating speed of the robot. Based on the correlation between the coefficient of the friction model and the operating speed of the robot, the Coulomb-viscous friction coefficient of the robot under high-speed operation to the linear friction model and the static friction coefficient and Stribeck speed threshold of the robot under low-speed operation are solved step by step. Finally, the obtained linear dynamic model, Coulomb-viscous friction coefficient, static friction coefficient and Stribeck speed threshold are integrated to obtain a high-precision nonlinear dynamic model. The real-time driving torque of each joint of the robot during the robot dragging teaching process is calculated according to the obtained nonlinear dynamic model; and a set of corresponding torque control instructions are generated according to the real-time driving torque of each joint of the robot; then a mapping relationship between the torque control instruction and the path of the dragging teaching process is established; standard spraying path data is generated, and automatic spraying is performed through the standard spraying path data, thereby ensuring the stability and accuracy of the automatic spraying control system; finally, sensorless zero-force dragging teaching is realized, so that the generated spraying path data can improve the flexibility and accuracy of automatic spraying, and reduce waste and environmental pollution during the spraying process.

[0068] An embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, it implements a drag teaching method based on a nonlinear dynamic model and a method for constructing the nonlinear model as described in any of the above embodiments.

[0069] The present invention may take the form of a computer program product implemented on one or more storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing program code. Computer-readable storage media include permanent and non-permanent, removable and non-removable media, and information storage can be achieved by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include but are not limited to: phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission medium that can be used to store information that can be accessed by a computing device.

[0070] The above-mentioned embodiments only express several implementation methods of the present invention, and the description is relatively specific and detailed, but it cannot be understood as limiting the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, and the present invention is also intended to include these modifications and modifications.

Claims

1. A drag teaching method based on a nonlinear dynamics model, characterized in that: include: S11: collecting the real-time robot state during the drag teaching process and inputting it into a nonlinear dynamic model; S12: the nonlinear dynamics model calculates the real-time driving torque of each joint of the robot during the spraying process according to the real-time state of the robot; The nonlinear dynamic model is expressed as follows: in, represents the driving torque of the robot joint; is the angular displacement of each joint of the robot, is the angular velocity of each joint of the robot, is the angular acceleration of each joint of the robot; represents the inertia moment term; represents the Coriolis moment term; represents the gravitational moment term; , , Composition of frictionless dynamics model; represents the friction torque model; is the Coulomb friction coefficient; is the static friction coefficient; is the Stribeck speed threshold; is the empirical constant of Stribeck curve attenuation; is the viscous friction coefficient; S13: Generate a set of corresponding torque control instructions according to the real-time driving torque of each joint of the robot; establish a corresponding relationship between the torque control instructions and the path of the dragging teaching process; and generate standard spraying path data.

2. The drag teaching method according to claim 1, characterized in that: The Coulomb friction coefficient and the viscous friction coefficient According to the high-speed dynamic parameter identification method, the Stribeck speed threshold and the static friction coefficient The real friction torque of the robot when running at low speed is solved according to the frictionless dynamics model, and then combined with and Substitution , and finally optimized by genetic algorithm.

3. The drag teaching method according to claim 2, characterized in that: The frictionless dynamic model and Coulomb friction coefficient of the nonlinear dynamic model and viscous friction coefficient It is obtained through the following steps: S21: Initializing the initial dynamics model of the robot according to the parameters of the robot. At this time, the initial dynamics model is: in, is the angular displacement of each joint of the robot, is the angular velocity of each joint of the robot, is the angular acceleration of each joint of the robot; represents the inertia moment term; represents the Coriolis moment term; represents the gravitational moment term, is the Coulomb friction coefficient; is the viscous friction coefficient; S22: linearizing the initial dynamics model of the robot to obtain a linear friction dynamics model; S23: Under preset constraint conditions, calculating the excitation trajectory coefficient using the linear friction dynamics model; S24: Collect the robot motion state when the robot runs the high-speed excitation trajectory under the excitation trajectory coefficient, and estimate the dynamic parameters of the linear friction dynamics model offline by the least squares method to obtain the frictionless dynamics model and the Coulomb friction coefficient. and viscous friction coefficient .

4. The drag teaching method according to claim 3, characterized in that: The Stribeck speed threshold and static friction coefficient In order to obtain the frictionless dynamics model, Coulomb friction coefficient and viscous friction coefficient Then obtained through the following steps: S25: collecting the motion state of each joint of the robot and the real torque data of each joint of the robot when the robot runs the low-speed excitation trajectory, and combining the frictionless dynamics model to obtain the real friction force in the low-speed state; S26: The actual friction force and the Coulomb friction coefficient and viscous friction coefficient Substituting into the nonlinear friction torque model, a plurality of groups of preliminary Stribeck speed thresholds and static friction coefficients are obtained, and the preliminary Stribeck speed thresholds and static friction coefficients are optimized by genetic algorithm to obtain the Stribeck speed thresholds and static friction coefficients; The nonlinear friction torque model is expressed as: in, is the angular velocity of each joint of the robot, is the static friction coefficient, is the Coulomb friction coefficient, is the viscous friction coefficient, is the Stribeck velocity threshold, is the empirical constant of Stribeck curve attenuation.

5. The drag teaching method according to claim 4, characterized in that: After obtaining the frictionless dynamic model, Coulomb-viscous friction coefficient, Stribeck velocity threshold and static friction coefficient, a nonlinear dynamic model is generated and the error coefficient of the nonlinear dynamic model is calculated by the following formula: : in, is the real friction force, is the true torque, It is a frictionless dynamics model; represents the moment of inertia; represents the Coriolis moment; represents the gravitational moment, represents the driving torque of the robot joint, i Indicates robot status parameters; If the error coefficient is greater than the preset threshold, the Stribeck velocity threshold and the static friction coefficient are continuously optimized by genetic algorithm iteration until the error coefficient is less than the preset threshold.

6. A drag teaching control system, characterized in that: include: A data acquisition module collects the real-time robot status during the drag teaching process and inputs it into a nonlinear dynamics model; A nonlinear dynamics model, used to calculate the real-time driving torque of each joint of the robot during the spraying process according to the nonlinear dynamics model and the real-time state of the robot; The nonlinear dynamic model is expressed as follows: in, represents the driving torque of the robot joint; is the angular displacement of each joint of the robot, is the angular velocity of each joint of the robot, is the angular acceleration of each joint of the robot; represents the inertia moment term; represents the Coriolis moment term; represents the gravitational moment term; , , Composition of frictionless dynamics model; represents the friction torque model; is the Coulomb friction coefficient; is the static friction coefficient; is the Stribeck speed threshold; is the empirical constant of Stribeck curve attenuation; is the viscous friction coefficient; The standard path data generation module generates a set of corresponding torque control instructions according to the real-time driving torque of each joint of the robot; establishes the corresponding relationship between the torque control instructions and the path of the dragging teaching process; and generates standard spraying path data.

7. An automatic spraying control device, characterized in that: include: The drag teaching control system as claimed in claim 6; Pose transformation matrix generation module: used to match the image data of the workpiece to be sprayed with a template point cloud to obtain a pose transformation matrix; Spraying path generation module: used to transform the standard spraying path data according to the posture transformation matrix to obtain real-time spraying path data; Spraying controller: controls the robot to complete automatic spraying according to real-time spraying path data.

8. A model building system for a nonlinear dynamic model in a drag teaching control system according to claim 6, comprising: Robot dynamics model building module: used to initialize the robot's initial dynamics model according to the robot's parameters. At this time, the robot's initial dynamics model is: in, is the angular displacement of each joint of the robot, is the angular velocity of each joint of the robot, is the angular acceleration of each joint of the robot; represents the inertia moment term; represents the Coriolis moment term; represents the gravitational moment term, is the Coulomb friction coefficient; is the viscous friction coefficient; Linearization processing module: used for performing linearization processing on the initial dynamics model of the robot to obtain a linear friction dynamics model; Excitation trajectory coefficient solving module: under preset constraint conditions, the excitation trajectory coefficient is calculated by the linear friction dynamics model; High-speed parameter identification module: used to collect the robot motion state when the robot runs the high-speed excitation trajectory under the excitation trajectory coefficient, and estimate the dynamic parameters of the linear friction dynamics model offline by the least squares method to obtain the frictionless dynamics model and the Coulomb friction coefficient. and viscous friction coefficient .

9. The model building system according to claim 8, characterized in that: Also includes: Low-speed friction calculation module: used to collect the motion state of each joint of the robot and the real torque data of each joint of the robot when the robot runs the low-speed excitation trajectory, and combine the frictionless dynamics model to obtain the real friction force in the low-speed state; Nonlinear parameter genetic optimization module: used to transform the real friction force, Coulomb friction coefficient and viscous friction coefficient Substituting into the nonlinear friction torque model, a plurality of groups of preliminary Stribeck speed thresholds and static friction coefficients are obtained, and the preliminary Stribeck speed thresholds and static friction coefficients are optimized by genetic algorithm to obtain the Stribeck speed thresholds and static friction coefficients; The nonlinear friction torque model is expressed as: in, is the angular velocity of each joint of the robot, is the static friction coefficient, is the Coulomb friction coefficient, is the viscous friction coefficient, is the Stribeck velocity threshold, is the empirical constant of Stribeck curve attenuation.

10. The model building system according to claim 9, characterized in that: Also includes: Error coefficient calculation module: used to calculate the error coefficient of the nonlinear dynamics model using the following formula : in, is the real friction force, is the true torque, It is a frictionless dynamics model; represents the moment of inertia; represents the Coriolis moment; represents the gravitational moment, represents the driving torque of the robot joint, i Indicates robot status parameters; Error coefficient determination module: if the error coefficient is greater than a preset threshold, the genetic algorithm is continued to be iteratively optimized for the Stribeck speed threshold and the static friction coefficient until the error coefficient is less than the preset threshold.

Images