Robot grinding and polishing processing self-adaptive compensation method based on multi-line laser

Abstract

The invention relates to a robot grinding and polishing self-adaptive compensation method based on multi-line laser, and relates to the field of robot positioning precision. According to the technical scheme, online high-resolution measurement of the workpiece is achieved through the multi-line laser scanning sensor, hand-eye parameters are dynamically updated in combination with multi-view calibration and data fusion, the tail end pose of the robot is corrected in real time through the mixed error model fusing MDH kinematics and machine learning, and the grinding and polishing precision and consistency are guaranteed. Wherein the MDH model is responsible for coarse positioning to guarantee interpretability of a kinematic structure, the machine learning model captures nonlinear dynamic errors such as thermal expansion and abrasion, and the MDH model and the machine learning model are combined to achieve self-adaptive real-time compensation of a complex error source. The error compensation precision is improved from the millimeter level to the submillimeter level, the cost of the line laser measurement system is only tens of thousands of yuan, the size is small, installation is easy, and compared with a laser tracker, the cost and the installation difficulty are greatly reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of robot positioning accuracy, and in particular to a multi-line laser-based robot grinding and polishing adaptive compensation method. Background Art

[0002] In the field of grinding and polishing, industrial robots are gradually replacing traditional grinding and polishing methods with their significant advantages such as low cost, high flexibility, large workspace and a wide range of processing objects. However, current robot grinding and polishing technology is mainly used in scenarios with light loads and moderate precision requirements, and it is difficult to achieve large-scale promotion in the field of precision machining. The root cause is that the robot itself has inherent defects such as low absolute positioning accuracy and insufficient rigidity. During long-term continuous operation, the interaction of multiple factors such as thermal expansion and zero-point drift of joint links caused by high loads, transmission clearance caused by wear of reducers, flexible deflection of mechanical components caused by external forces, as well as sensor errors, environmental interference and model deviations, causes the positioning error of the robot end effector to accumulate continuously, seriously affecting the material removal accuracy, surface quality consistency and trajectory planning rationality of the grinding and polishing of precision parts.

[0003] To address robot positioning errors, existing technologies typically use a laser tracker combined with an end-target sphere to compensate for position errors. However, this method has significant limitations: First, the equipment procurement and maintenance costs are high, significantly increasing production costs for enterprises. Second, it places stringent requirements on the installation environment, requiring a large, unobstructed workspace. Furthermore, equipment placement and relocation are extremely inconvenient when adjusting production line layouts or performing multi-station operations, making it difficult to meet the flexible and ever-changing production demands of modern manufacturing. Therefore, an efficient, low-cost, and highly adaptable error compensation technology solution is urgently needed to enable industrial robots to perform high-precision operations in the grinding and polishing of precision parts. Summary of the Invention

[0004] The purpose of the present invention is to provide a multi-line laser-based robot grinding and polishing adaptive compensation method to solve the problems existing in the above-mentioned prior art.

[0005] To achieve the above object, the technical solution adopted by the present invention is: In a first aspect, the present invention provides a multi-line laser-based robot grinding and polishing adaptive compensation method, the method comprising: S1. Build a multi-line laser measurement system to ensure that the robot end is within the coverage range of the multi-line laser measurement system; S2, scanning the calibration block at the end of the robot by the multi-line laser measurement system, and obtaining the coordinates of the feature points of different positions of the calibration block in the coordinate system of the multi-line laser measurement system; S3. Determine the rigid transformation matrix from the multi-line laser measurement system coordinate system to the robot end coordinate system using the Horn or Umeyama absolute orientation algorithm; S4. Construct the robot kinematic model based on the MDH parameter method to describe the transformation relationship between the robot base coordinate system and the robot end coordinate system; S5, in response to the robot under different running time, joint load, joint angle, and ambient temperature conditions, obtaining the actual end position and theoretical position of the robot, and constructing an error data set; S6. Build a multi-layer perceptron (MLP) residual neural network model, define the input layer, output layer, and mean square error loss function; S7, using the error data set to train the multi-layer perceptron (MLP) residual neural network model to fit the nonlinear error of the robot terminal posture; S8, based on the trained multi-layer perceptron MLP residual neural network model, combined with the LM algorithm, determine the posture parameter increment, update the posture parameter based on the posture parameter increment, and generate the compensated robot end posture based on the updated posture parameter combined with the multi-layer perceptron MLP residual neural network model; S9, periodically obtain the robot's actual end position and theoretical position. When the error between the two exceeds a preset threshold, re-execute steps S5 to S7 to correct the multi-layer perceptron MLP residual neural network model.

[0006] In a possible implementation, in step S1, the multi-line laser measurement system includes: Two line laser sensors are mounted on a workbench at a 90° angle.

[0007] In a possible implementation, in step S3, the rigid transformation matrix from the multi-line laser measurement system coordinate system to the robot end coordinate system includes: , and satisfies , in, is the homogeneous transformation matrix from the multi-line laser measurement system coordinate system to the robot end coordinate system; is a 3×3 rotation matrix; is a 3×1 translation vector; is the coordinate of the feature point in the robot terminal coordinate system; are the coordinates of the feature points in the coordinate system of the multi-line laser measurement system.

[0008] In a possible implementation, the adjacent coordinate system transformation matrix involved in the robot kinematic model in step S4 includes: ; in, is the coordinate transformation matrix between two adjacent links; For connecting rod i joint angles; For connecting rod i connecting rod offset; For connecting rod i The joint torsion angle; For connecting rod i Length of connecting rod; For connecting rod i Y-axis parallelism parameter; c represents the cosine function cos, s Represents the sine function sin.

[0009] In a possible implementation, the transformation matrix of the robot end coordinate system {6} relative to the robot base coordinate system {0} involved in the robot kinematic model in step S4 includes: ; in, is the direction vector of the X-axis in the robot's end coordinate system in the robot's base coordinate system; is the direction vector of the Y axis in the robot's end coordinate system in the robot's base coordinate system; is the direction vector of the Z axis in the robot's end coordinate system in the robot's base coordinate system; is the position vector of the origin of the robot's end coordinate system in the robot's base coordinate system.

[0010] In a possible implementation, in step S5, the error of the error data set is defined as: ; in, is the residual between the actual end position of the robot and the theoretical position; is the actual end position of the robot; is the theoretical end position of the robot; For connecting rod i joint angles; T is the ambient temperature; t is the running time; F The joint load.

[0011] In a possible implementation, in step S6: The input layer of the multi-layer perceptron MLP residual neural network model includes: Theoretical end position of the robot ,link i Joint angle , ambient temperatureT , running time t , and joint loads F ; The output layer of the multi-layer perceptron MLP residual neural network model includes: The residual between the actual end position of the robot and the theoretical position ; The mean square error loss function of the multi-layer perceptron MLP residual neural network model includes: ; in, Loss is the mean square error loss value; is the residual between the actual end position of the robot and the theoretical position; is the residual value predicted by the multi-layer perceptron MLP residual neural network model; N is the number of training samples.

[0012] In a possible implementation, in step S8, determining a pose parameter increment based on the trained multi-layer perceptron (MLP) residual neural network model in combination with the LM algorithm, and updating the pose parameter based on the pose parameter increment include: ; ; ; in, is the residual value predicted by the multi-layer perceptron MLP residual neural network model; is the connecting rod offset; is the joint torsion angle; is the connecting rod length; p is the current pose parameter vector; is the pose parameter increment; λ is the damping factor; I is the identity matrix; J is the Jacobian matrix; T is the matrix transpose; r is the residual function; are the updated pose parameters.

[0013] In one possible implementation, in step S9, the actual end position and theoretical position of the robot are periodically obtained. When the error between the two exceeds a preset threshold, steps S5 to S7 are re-executed to correct the multi-layer perceptron (MLP) residual neural network model, including: ; in, is the residual between the actual end position of the robot and the theoretical position; is the preset error threshold coefficient.

[0014] In a possible implementation, the calibration block in step S2 is a trapezoidal sheet metal calibration block, which serves as a measurement standard. The positions of its feature points are known and can be easily identified by the multi-line laser measurement system.

[0015] In a second aspect, the present invention provides a computer device, comprising a processor and a memory, wherein the memory stores at least one instruction, at least one program, code set or instruction set, and the processor can load and execute at least one instruction, at least one program, code set or instruction set to implement the adaptive compensation method for multi-line laser-based robot grinding and polishing provided above.

[0016] In a third aspect, the present invention provides a computer-readable storage medium, which stores at least one instruction, at least one program, code set or instruction set. The processor can load and execute at least one instruction, at least one program, code set or instruction set to implement the adaptive compensation method for multi-line laser-based robot grinding and polishing provided above.

[0017] In a fourth aspect, the present invention provides a computer program product or computer program, comprising computer program instructions stored in a computer-readable storage medium. A processor reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing a computer device to perform the adaptive compensation method for multi-line laser-based robotic grinding and polishing as provided above.

[0018] The beneficial effects brought about by the technical solution provided by the present invention include at least: This technical solution uses multi-line laser scanning sensors to achieve high-resolution online measurement of workpieces. It combines multi-view calibration with data fusion to dynamically update hand-eye parameters. It utilizes a hybrid error model that integrates MDH kinematics and machine learning to correct the robot's end-user position in real time, ensuring grinding and polishing accuracy and consistency. The MDH model is responsible for coarse positioning to ensure the interpretability of the kinematic structure, while the machine learning model captures nonlinear dynamic errors such as thermal expansion and wear. Together, the two enable adaptive real-time compensation of complex error sources. Error compensation accuracy is improved from millimeter to submillimeter levels, and the line laser measurement system costs only tens of thousands of yuan. Its compact size and ease of installation significantly reduce the cost and installation difficulty of laser trackers. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0020] Figure 1 A schematic flow chart of an adaptive compensation method for multi-line laser-based robot grinding and polishing provided by an exemplary embodiment of the present invention is shown.

[0021] Figure 2 A schematic diagram of a connecting rod coordinate system of a six-axis robot provided by an exemplary embodiment of the present invention is shown.

[0022] Figure 3 A schematic diagram of a multi-layer perceptron (MLP) residual neural network model provided by an exemplary embodiment of the present invention is shown.

[0023] Figure 4 A schematic structural diagram of a computer device for executing an adaptive compensation method for multi-line laser-based robot grinding and polishing provided by an exemplary embodiment of the present invention is shown. DETAILED DESCRIPTION

[0024] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0025] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0026] Figure 1 A schematic flow chart of an adaptive compensation method for multi-line laser robot grinding and polishing provided by an exemplary embodiment of the present invention is shown. The adaptive compensation method for multi-line laser robot grinding and polishing includes: Step S1: Build a multi-line laser measurement system to ensure that the robot end is within the coverage of the multi-line laser measurement system.

[0027] In detail, the multi-line laser measurement system includes two line laser sensors installed on a workbench at a 90° angle.

[0028] In the embodiments of this application, a multi-line laser measurement system is used to establish three-dimensional spatial coordinate perception capabilities. Two line laser sensors are mounted on a workbench at a 90° angle. Based on the principle of laser triangulation, the offset of the light spot formed by the laser beam projected onto the object's surface is used in conjunction with the sensor imaging principle to calculate the coordinates of spatial points. This orthogonal installation method can obtain three-dimensional data of an object from different perspectives through the binocular parallax effect. Multi-line laser scanning forms a dense point cloud that covers the robot's end-of-line motion space. This configuration effectively eliminates blind spots in single-sensor measurement and improves spatial point positioning accuracy by leveraging geometric projection relationships.

[0029] Step S2: Scan the calibration block at the end of the robot through the multi-line laser measurement system to obtain the coordinates of the feature points of the calibration block in different positions in the coordinate system of the multi-line laser measurement system.

[0030] In detail, the calibration block is a trapezoidal sheet metal calibration block, which serves as a measurement standard. The positions of its feature points are known and easy to be identified by the multi-line laser measurement system.

[0031] In an embodiment of the present application, when a multi-line laser beam is projected onto the surface of a trapezoidal sheet metal calibration block, its regular geometric features (such as edges and corners) form a specific light spot pattern on the sensor imaging plane. Utilizing the principle of laser triangulation, two orthogonally mounted line laser sensors synchronously capture the light spot offset. Combined with the known three-dimensional coordinates of the calibration block's feature points (preset physical dimensions converted to spatial coordinates), a mapping relationship between the measurement system coordinate system and the calibration block's coordinate system can be established. By driving the robot to transform its end-point posture and collecting multiple sets of data, the least squares method in statistics is used to fit the spatial transformation parameters, thereby eliminating random errors in single measurements.

[0032] Step S3: Use the Horn or Umeyama absolute orientation algorithm to determine the rigid transformation matrix from the multi-line laser measurement system coordinate system to the robot end coordinate system.

[0033] In detail, the rigid transformation matrix from the multi-line laser measurement system coordinate system to the robot end coordinate system includes: , and satisfies , in, is the homogeneous transformation matrix from the multi-line laser measurement system coordinate system to the robot end coordinate system; is a 3×3 rotation matrix; is a 3×1 translation vector; is the coordinate of the feature point in the robot terminal coordinate system; are the coordinates of the feature points in the coordinate system of the multi-line laser measurement system.

[0034] In the embodiment of the present application, the optimal rotation and translation parameters between the two coordinate systems are solved by the Horn or Umeyama algorithm, with the goal of minimizing the Euclidean distance error of the feature point pairs. First, the center of gravity of the two sets of coordinate points is calculated, and the coordinate system is converted to the center of gravity coordinate system. The rotation matrix is ​​solved by constructing the covariance matrix. , and then use the translation vector Complete the coordinate transformation. The Horn algorithm solves the rotation matrix based on quaternion optimization and is applicable to rigid transformations without scaling. The Umeyama algorithm extends this to affine transformations with scaling factors. Both achieve closed-form solutions through singular value decomposition (SVD) or eigenvalue decomposition. This process maps the 3D point cloud coordinates acquired by the multi-line laser measurement system to the robot's end-user coordinate system, providing a unified coordinate reference for subsequent kinematic modeling and error compensation.

[0035] Step S4: Construct the robot kinematic model based on the MDH parameter method to describe the transformation relationship between the robot base coordinate system and the robot end coordinate system. (See Figure 2 , showing a schematic diagram of the link coordinate system of the six-axis robot) In detail, the adjacent coordinate system transformation matrices involved in the robot kinematic model include: ; in, is the coordinate transformation matrix between two adjacent links; For connecting rod i joint angles; For connecting rod i connecting rod offset; For connecting rod i The joint torsion angle; For connecting rod i Length of connecting rod; For connecting rod i Y-axis parallelism parameter; c represents the cosine function cos, s Represents the sine function sin.

[0036] Furthermore, the transformation matrix of the robot end coordinate system {6} relative to the robot base coordinate system {0} involved in the robot kinematic model includes: ; ; in, is the direction vector of the X-axis in the robot's end coordinate system in the robot's base coordinate system; is the direction vector of the Y axis in the robot's end coordinate system in the robot's base coordinate system; is the direction vector of the Z axis in the robot's end coordinate system in the robot's base coordinate system; is the position vector of the origin of the robot's end coordinate system in the robot's base coordinate system.

[0037] In the embodiment of the present application, based on the Denavit-Hartenberg (DH) kinematic modeling theory, the robot kinematic model is constructed by defining the link coordinate system and joint parameters. The traditional DH model uses four parameters: joint angle, link offset, link length and joint torsion angle to describe the transformation of adjacent coordinate systems. The ModifiedDH (MDH) model of the present application adds a Y-axis parallelism parameter and realizes the homogeneous transformation of adjacent link coordinate systems through a matrix, where cθ and sθ are the cosine and sine values ​​of the joint angles, respectively. In addition, the transformation matrix of the robot end coordinate system {6} relative to the base coordinate system {0} is obtained by concatenating the transformation matrices of each link, and its direction vector Describe the end posture, position vector Determine the end space coordinates to achieve the forward kinematics solution from joint angles to end pose.

[0038] Step S5: In response to the robot under different running time, joint load, joint angle, and ambient temperature conditions, the actual end position and theoretical position of the robot are obtained, and an error data set is constructed.

[0039] In detail, the error of the error dataset is defined as: ; in, is the residual between the actual end position of the robot and the theoretical position; is the actual end position of the robot; is the theoretical end position of the robot; For connecting rod i joint angles; T is the ambient temperature; t is the running time; F The joint load.

[0040] In an embodiment of the present application, an error mapping model is constructed by collecting the posture data of the robot under dynamic working conditions. When the robot is running, factors such as mechanical deformation caused by changes in joint angles, transmission clearance caused by loads, thermal expansion caused by temperature fluctuations, and accumulated wear during operation will cause the end posture to deviate from the theoretical value. This step uses sensors to obtain the true end posture of the robot under different joint angles, load conditions, ambient temperatures, and operating times in real time, and at the same time calculates the theoretical posture based on the kinematic model. The difference between the two constitutes an error data set. The error is defined as the residual between the true posture and the theoretical posture. This data set covers the error characteristics under the coupling of multiple physical fields.

[0041] Step S6: Construct a multi-layer perceptron (MLP) residual neural network model, define the input layer, output layer, and mean square error loss function. (See Figure 3 , showing a schematic diagram of the multi-layer perceptron MLP residual neural network model) In detail, the input layer of the multi-layer perceptron MLP residual neural network model includes: Theoretical end position of the robot ,link i Joint angle , ambient temperature T , running time t , and joint loads F ; The output layer of the multi-layer perceptron MLP residual neural network model includes: The residual between the actual end position of the robot and the theoretical position ; The mean square error loss function of the multi-layer perceptron MLP residual neural network model includes: ; in, Loss is the mean square error loss value; is the residual between the actual end position of the robot and the theoretical position; is the residual value predicted by the multi-layer perceptron MLP residual neural network model; N is the number of training samples.

[0042] In an embodiment of the present application, a dynamic prediction model of the robot posture error is constructed by a multi-layer perceptron based on the nonlinear mapping capability of a neural network. The model uses the theoretical end posture, joint angle, ambient temperature, running time and joint load as input layer variables. These parameters cover the error inducements of the robot kinematics theoretical value and the multi-physical field coupling. The output layer directly corresponds to the residual between the actual posture and the theoretical posture, and the complex nonlinear error is mapped into a computable correction value through the residual learning mechanism. The mean square error loss function minimizes the sum of the squares of the predicted residual and the actual residual, and optimizes the network weights using the gradient descent algorithm, so that the model can adaptively fit the nonlinear distribution characteristics of dynamic errors such as thermal expansion and wear clearance.

[0043] Step S7: Use the error data set to train the multi-layer perceptron (MLP) residual neural network model to fit the nonlinear error of the robot's terminal posture.

[0044] In an embodiment of the present application, an MLP neural network is iteratively optimized using an error dataset. This process uses the error dataset constructed in step S5 as input, with parameters such as theoretical pose and joint angles serving as input features and pose residuals serving as labels. The network weights are adjusted using a backpropagation algorithm. The MLP uses multiple layers of nonlinear activation functions (such as ReLU) to extract the implicit associations between input parameters and errors layer by layer, capturing the nonlinear distribution of dynamic errors such as thermal expansion and wear clearance. During training, a mean squared error loss function guides the model to continuously narrow the deviation between the predicted and actual residuals, enabling the network to map multi-physics coupling parameters to terminal errors, ultimately forming a nonlinear error prediction model that can adapt to changing operating conditions.

[0045] Step S8: Based on the trained multi-layer perceptron MLP residual neural network model, combined with the LM algorithm, determine the posture parameter increment, update the posture parameter based on the posture parameter increment, and generate the compensated robot end posture based on the updated posture parameter combined with the multi-layer perceptron MLP residual neural network model.

[0046] In detail, based on the trained multi-layer perceptron MLP residual neural network model and combined with the LM algorithm, the pose parameter increment is determined, and the pose parameters are updated based on the pose parameter increment, including: ; ; ; in, is the residual value predicted by the multi-layer perceptron MLP residual neural network model; is the connecting rod offset; is the joint torsion angle; is the connecting rod length; p is the current pose parameter vector; is the pose parameter increment; λ is the damping factor; I is the identity matrix; J is the Jacobian matrix; T is the matrix transpose; r is the residual function; are the updated pose parameters.

[0047] In the embodiment of the present application, based on nonlinear optimization and iterative correction, the LM algorithm is used to solve the optimal correction amount of the posture parameters. The algorithm takes the posture residual predicted by the MLP model as the optimization target and constructs the residual function r（p） And calculate its Jacobian matrix J , through the damping factor λBalance the local convergence of the Gauss-Newton method and the global search capability of the gradient descent method. During the iteration, the formula Solve for the pose parameter increment, where JᵀJ is the Hessian matrix approximation, λI Prevent matrix singularity; each round of iteration updates the pose parameters , gradually reducing the deviation between the predicted residual and the actual error, and finally generating the compensated terminal pose through the multi-layer perceptron model.

[0048] Step S9: Periodically obtain the robot's actual end position and theoretical position. When the error between the two exceeds a preset threshold, re-execute steps S5 to S7 to correct the multi-layer perceptron (MLP) residual neural network model.

[0049] Specifically, the actual end-position and theoretical position of the robot are periodically obtained. When the error between the two exceeds a preset threshold, steps S5 to S7 are re-executed to correct the multi-layer perceptron (MLP) residual neural network model, including: ; in, is the residual between the actual end position of the robot and the theoretical position; is the preset error threshold coefficient.

[0050] In an embodiment of the present application, based on adaptive learning and online optimization, the dynamic update of the error model is achieved through a threshold trigger mechanism. During the long-term operation of the robot, time-varying factors such as mechanical wear, lubricant attenuation, and ambient temperature changes will cause the error characteristics to drift, and the prediction accuracy of the originally trained MLP model will gradually decrease. This step periodically collects the residuals between the actual posture and the theoretical posture, and compares them with the preset threshold coefficient. When the error exceeds the threshold, it indicates that the current model can no longer accurately describe the real-time error characteristics. At this time, the data collection and model training process is re-executed, and the MLP model is incrementally learned using the latest error data set to correct the model deviation caused by time-varying factors and ensure that the error compensation model always matches the current actual error distribution characteristics of the robot.

[0051] Figure 4 A schematic diagram of the structure of a computer device for executing an adaptive compensation method for multi-line laser-based robot grinding and polishing provided by an exemplary embodiment of the present invention is shown. The computer device includes: The processor 401 includes one or more processing cores. The processor 401 executes various functional applications and data processing by running software programs and modules.

[0052] Receiver 402 and transmitter 403 can be implemented as a communication component, which can be a communication chip. Optionally, the communication component can include signal transmission functionality. That is, transmitter 403 can be used to transmit control signals to the image acquisition device and scanning device, and receiver 402 can be used to receive corresponding feedback instructions.

[0053] The memory 404 is connected to the processor 401 via a bus 405 .

[0054] The memory 404 may be used to store at least one instruction, and the processor 401 may be used to execute the at least one instruction to implement each step in the above method embodiment.

[0055] An embodiment of the present invention also provides a computer-readable storage medium, which stores at least one instruction, at least one program, code set or instruction set, which is loaded and executed by a processor to implement the above-mentioned multi-line laser-based robot grinding and polishing adaptive compensation method.

[0056] The present invention also provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the adaptive compensation method for multi-line laser-based robotic grinding and polishing described in any of the above embodiments.

[0057] Optionally, the computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), solid-state drive (SSD), or optical disk. Among them, random access memory may include resistance random access memory (ReRAM) and dynamic random access memory (DRAM). The serial numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0058] It should be understood that the specific examples herein are only intended to help those skilled in the art to better understand the present disclosure, rather than to limit the scope of the present invention.

[0059] It can be understood that in the various implementations of this specification, the size of the sequence number of each process does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of this disclosure.

[0060] It can be understood that the various embodiments described in this specification can be implemented individually or in combination, and the present disclosure is not limited thereto.

[0061] Unless otherwise indicated, all technical and scientific terms used in this disclosure have the same meaning as commonly understood by those skilled in the art in the technical field of this specification. The terms used in this specification are only for the purpose of describing specific embodiments and are not intended to limit the scope of this specification. The term "and / or" used in this specification includes any and all combinations of one or more related listed items. The singular forms "a", "above", and "the" used in this disclosure and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0062] It is understood that the processor of the present disclosure can be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above-mentioned method implementation can be completed by hardware integrated logic circuits in the processor or software instructions. The above-mentioned processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The various methods, steps, and logic block diagrams disclosed in this disclosure can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this disclosure can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium mature in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above-mentioned method in combination with its hardware.

[0063] It will be understood that the memory in the present disclosure may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. Among them, the non-volatile memory may be read-only memory (ROM), programmable read-only memory (programmable ROM, PROM), erasable programmable read-only memory (erasable PROM, EPROM), electrically erasable programmable read-only memory (EEPROM) or flash memory. The volatile memory may be random access memory (RAM). It should be noted that the memory of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0064] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this specification.

[0065] The above description is merely a specific embodiment of this specification, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this specification should be included in the scope of protection of this specification. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A multi-line laser-based robot grinding and polishing adaptive compensation method, characterized in that: The method comprises: S1. Build a multi-line laser measurement system to ensure that the robot end is within the coverage range of the multi-line laser measurement system; S2, scanning the calibration block at the end of the robot by the multi-line laser measurement system, and obtaining the coordinates of the feature points of different positions of the calibration block in the coordinate system of the multi-line laser measurement system; S3. Determine the rigid transformation matrix from the multi-line laser measurement system coordinate system to the robot end coordinate system using the Horn or Umeyama absolute orientation algorithm; S4. Construct the robot kinematic model based on the MDH parameter method to describe the transformation relationship between the robot base coordinate system and the robot end coordinate system; S5, in response to the robot under different running time, joint load, joint angle, and ambient temperature conditions, obtaining the actual end position and theoretical position of the robot, and constructing an error data set; S6. Build a multi-layer perceptron (MLP) residual neural network model, define the input layer, output layer, and mean square error loss function; S7, using the error data set to train the multi-layer perceptron (MLP) residual neural network model to fit the nonlinear error of the robot terminal posture; S8, based on the trained multi-layer perceptron MLP residual neural network model, combined with the LM algorithm, determine the posture parameter increment, update the posture parameter based on the posture parameter increment, and generate the compensated robot end posture based on the updated posture parameter combined with the multi-layer perceptron MLP residual neural network model; S9, periodically obtain the robot's actual end position and theoretical position. When the error between the two exceeds a preset threshold, re-execute steps S5 to S7 to correct the multi-layer perceptron MLP residual neural network model.

2. The adaptive compensation method for robot grinding and polishing based on multi-line laser according to claim 1 is characterized in that: In step S1, the multi-line laser measurement system includes: Two line laser sensors are mounted on a workbench at a 90° angle.

3. The adaptive compensation method for robot grinding and polishing based on multi-line laser according to claim 1, characterized in that: In step S3, the rigid transformation matrix from the multi-line laser measurement system coordinate system to the robot end coordinate system includes: , and satisfies , in, is the homogeneous transformation matrix from the multi-line laser measurement system coordinate system to the robot end coordinate system; is a 3×3 rotation matrix; is a 3×1 translation vector; is the coordinate of the feature point in the robot terminal coordinate system; are the coordinates of the feature points in the coordinate system of the multi-line laser measurement system.

4. The adaptive compensation method for robot grinding and polishing based on multi-line laser according to claim 1, characterized in that: The adjacent coordinate system transformation matrices involved in the robot kinematic model in step S4 include: ； in, is the coordinate transformation matrix between two adjacent links; For connecting rod i joint angles; For connecting rod i connecting rod offset; For connecting rod i The joint torsion angle; For connecting rod i Length of connecting rod; For connecting rod i Y-axis parallelism parameter; c represents the cosine function cos, s Represents the sine function sin.

5. The adaptive compensation method for robot grinding and polishing based on multi-line laser according to claim 1, characterized in that: The robot kinematic model in step S4 involves the transformation matrix of the robot end coordinate system {6} relative to the robot base coordinate system {0} include: ； in, is the direction vector of the X-axis in the robot's end coordinate system in the robot's base coordinate system; is the direction vector of the Y axis in the robot's end coordinate system in the robot's base coordinate system; is the direction vector of the Z axis in the robot's end coordinate system in the robot's base coordinate system; is the position vector of the origin of the robot's end coordinate system in the robot's base coordinate system.

6. The adaptive compensation method for robot grinding and polishing based on multi-line laser according to claim 1, characterized in that: In step S5, the error of the error data set is defined as: ； in, is the residual between the actual end position of the robot and the theoretical position; is the actual end position of the robot; is the theoretical end position of the robot; For connecting rod i joint angles; T is the ambient temperature; t is the running time; F The joint load.

7. The adaptive compensation method for robot grinding and polishing based on multi-line laser according to claim 1, characterized in that: In the step S6: The input layer of the multi-layer perceptron MLP residual neural network model includes: Theoretical end position of the robot ,link i Joint angle , ambient temperature T , running time t , and joint loads F ; The output layer of the multi-layer perceptron MLP residual neural network model includes: The residual between the actual end position of the robot and the theoretical position ; The mean square error loss function of the multi-layer perceptron MLP residual neural network model includes: ； in, Loss is the mean square error loss value; is the residual between the actual end position of the robot and the theoretical position; is the residual value predicted by the multi-layer perceptron MLP residual neural network model; N is the number of training samples.

8. The adaptive compensation method for robot grinding and polishing based on multi-line laser according to claim 1, characterized in that: In step S8, determining the pose parameter increment based on the trained multi-layer perceptron MLP residual neural network model in combination with the LM algorithm, and updating the pose parameter based on the pose parameter increment includes: ； ； ； in, is the residual value predicted by the multi-layer perceptron MLP residual neural network model; is the connecting rod offset; is the joint torsion angle; is the connecting rod length; p is the current pose parameter vector; is the pose parameter increment; λ is the damping factor; I is the identity matrix; J is the Jacobian matrix; T is the matrix transpose; r is the residual function; are the updated pose parameters.

9. The adaptive compensation method for robot grinding and polishing based on multi-line laser according to claim 1, characterized in that: In step S9, the actual end position and theoretical position of the robot are periodically obtained. When the error between the two exceeds a preset threshold, steps S5 to S7 are re-executed to correct the multi-layer perceptron (MLP) residual neural network model, including: ； in, is the residual between the actual end position of the robot and the theoretical position; is the preset error threshold coefficient.

10. The adaptive compensation method for robot grinding and polishing based on multi-line laser according to claim 1, characterized in that: The calibration block in step S2 is a trapezoidal sheet metal calibration block, which serves as a measurement standard. The positions of its characteristic points are known and can be easily identified by the multi-line laser measurement system.

Images