A robot autonomous engraving method and system for complex surfaces

Abstract

The present invention discloses a method and system for autonomous robot engraving operations on complex curved surfaces. The method includes: S1, designing a font segmentation method and constructing a letter segmentation information library; constructing a sub-skill primitive library; using a discrete dynamic motion primitive model to generate a planar reference trajectory; S2, using a 3D camera to acquire a point cloud of the complex curved surface of interest, and performing eye-out-of-hand calibration on the camera to obtain the transformation relationship from the camera coordinate system to the robotic arm base coordinate system, thereby obtaining the surface reference trajectory and surface normal vector; S3, performing parameter identification and gravity compensation on the robotic arm end-end engraving tool based on a six-dimensional force sensor; S4, using an adaptive admittance controller based on a radial basis function neural network to generate a desired engraving trajectory, so that the interaction force between the environment and the robotic arm tracks the desired interaction force, thereby achieving robotic arm engraving operations. The present invention enables a robot to perform engraving operations on complex curved surfaces with high engraving precision, good engraving effects, and strong practicality.

Description

Technical Field

[0001] The present invention belongs to the technical field of robots, and in particular relates to a robot autonomous engraving operation method and system for complex curved surfaces. Background Art

[0002] In recent years, robotics technology has developed rapidly. Industrial robots, due to their advantages such as reliable stability, high repeatability, and ability to operate in high-risk environments, are increasingly being used in industrial manufacturing applications such as engraving, welding, assembly, and polishing. Robotic engraving operations often require teaching to perform new tasks. However, traditional teaching techniques are inefficient, highly specialized, and can only perform specific tasks. They cannot meet the customization and efficiency requirements of engraving operations. Therefore, robots must possess more efficient skill generalization capabilities to quickly adapt to complex environments and diverse task requirements, thereby continuously improving their intelligence.

[0003] Carving on complex surfaces requires ensuring the robot's carving accuracy and effect, and places high demands on the robot's force controller. In the patent, engraving control method, device, engraving equipment and storage medium CN116330260A, a traditional PID controller is used to correct the current trajectory of the engraving robot to achieve constant force engraving. There are also many documents that use classic fixed-parameter admittance controllers to achieve constant force control of robot engraving or polishing. However, when the engraving robot encounters different surface shapes and materials with different stiffness, it is necessary to manually adjust the PID gain or admittance parameters according to changes in the working environment, which not only affects the efficiency of the engraving operation, but also the lack of adaptive parameter adjustment of the controller will greatly reduce the engraving accuracy on complex surfaces, resulting in poor engraving results. Summary of the Invention

[0004] The main purpose of the present invention is to overcome the shortcomings and deficiencies of the prior art and to provide a method and system for autonomous robot engraving of complex curved surfaces.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A robot autonomous engraving method for complex curved surfaces comprises the following steps:

[0007] S1. Design a glyph segmentation method for English letters and build a letter segmentation information database; construct a teaching sub-trajectory for each glyph category and build a sub-skill primitive library; use discrete dynamic motion primitive models (DMPs) to generate planar reference trajectories;

[0008] S2. Use a 3D camera to acquire point cloud information of the complex surface of interest, and perform eye-outside-hand calibration on the camera to obtain the transformation relationship between the camera coordinate system and the robot arm base coordinate system. Based on the transformation relationship, obtain the surface point cloud in the robot arm base coordinate system. Based on the surface trajectory planning method of the wrapping projection algorithm, project the generated plane reference trajectory onto the surface to obtain the surface reference trajectory and surface normal vector;

[0009] S3. Parameter identification of the engraving tool at the end of the robotic arm based on the six-dimensional force sensor, gravity compensation of the engraving tool at the end of the robotic arm, eliminating the influence of gravity of the engraving tool at the end of the robotic arm, and obtaining the interaction force between the external environment and the robotic arm;

[0010] S4. The robotic arm adopts an adaptive admittance controller based on radial basis function neural network (RBFNN) to generate the desired engraving trajectory, while making the interaction force between the environment and the robotic arm track the desired interaction force, thus realizing smooth engraving operation of the robotic arm.

[0011] The present invention also includes a robot autonomous engraving operation system for complex curved surfaces. The system adopts the robot autonomous engraving operation method provided by the present invention. The system includes a plane reference trajectory generation module, a surface reference trajectory and surface normal vector generation module, a parameter identification and gravity compensation module, and an adaptive admittance control module.

[0012] Generate a planar reference trajectory module, which is used to segment all English letters and build a letter segmentation information library containing all letter segmentation information; construct a teaching sub-trajectory for each letter category to build a sub-skill primitive library; and use discrete dynamic motion primitive models (DMPs) to generate planar reference trajectories.

[0013] The module generates surface reference trajectories and surface normal vectors. It uses a 3D camera to acquire point cloud information of the complex surface of interest and calibrates the camera with the eye outside the hand to obtain the transformation relationship between the camera coordinate system and the robot base coordinate system. Based on the transformation relationship, the surface point cloud in the robot base coordinate system is obtained. The surface trajectory planning method based on the wrapping projection algorithm is used to project the generated plane reference trajectory onto the surface to obtain the surface reference trajectory and surface normal vector.

[0014] The parameter identification and gravity compensation module is used to identify the parameters of the engraving tool at the end of the robotic arm based on the six-dimensional force sensor, perform gravity compensation on the engraving tool at the end of the robotic arm, and obtain the interaction force between the external environment and the robotic arm;

[0015] The adaptive admittance control module adopts an adaptive admittance controller based on radial basis neural network (RBFNN) to generate the desired engraving trajectory, while making the interaction force between the environment and the robotic arm track the desired interaction force, thus realizing smooth engraving operation of the robotic arm.

[0016] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0017] 1. The present invention adopts a method of generating a plane reference trajectory based on any given English letter. When generating a reference trajectory of different letters or different positions and sizes each time, it avoids the re-teaching and planning of the traditional teaching method, greatly reducing the workload and operation difficulty. At the same time, it gives the robot more efficient skill generalization ability to adapt to the changing needs of engraving operations.

[0018] 2. The present invention adopts the method of generating surface reference trajectory and surface normal vector to project the plane reference trajectory onto the complex curved surface, and expands the plane engraving operation to the curved surface, which greatly enriches the environment categories of robot engraving operation and adapts to complex engraving operation surfaces.

[0019] 3. The present invention adopts parameter identification and gravity compensation to obtain the identification parameters of the end-carving tool and the six-dimensional force sensor of the robotic arm, and eliminates the influence of the gravity of the end-carving tool, thereby obtaining the interaction force between the external environment and the robotic arm.

[0020] 4. The present invention adopts an adaptive admittance controller to complete the task of the robot to perform flexible engraving operations. At the same time, it not only ensures the safety of the interaction between the robot and the environment, but also obtains good engraving accuracy and engraving effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a flow chart of the method of the present invention;

[0022] Figure 2 It is a schematic diagram of the glyph segmentation method;

[0023] Figure 3 It is a flow chart of a surface trajectory planning method based on a wrapping projection algorithm;

[0024] Figure 4 is the damping parameter learning and convergence curve;

[0025] Figure 5 is the stiffness parameter learning and convergence curve;

[0026] Figure 6 It is the interaction force tracking expected interaction force curve between the environment and the robotic arm;

[0027] Figure 7 Schematic diagram of the system of the present invention. DETAILED DESCRIPTION

[0028] The present invention will be described in further detail below with reference to the embodiments and drawings, but the embodiments of the present invention are not limited thereto.

[0029] Example

[0030] like Figure 1 As shown, the present invention provides a robot autonomous engraving method for complex curved surfaces, comprising the following steps:

[0031] S1. Design a glyph segmentation method for English letters and build a glyph segmentation information library; construct a teaching sub-trajectory for each glyph category and build a sub-skill primitive library; use discrete dynamic motion primitive models (DMPs) to generate planar reference trajectories; specifically include:

[0032] S11. For any letter, it can be composed of a finite number of specific combinations of sub-shapes, so all English letters are segmented; a letter segmentation information library containing all letter segmentation information is constructed, and the information library contains the shape category and starting and ending points of each sub-track after each letter segmentation. The three-dimensional coordinates of the constructed starting and ending points are all on the OXY plane of the robot arm base coordinate system; in this embodiment, Figure 2 As shown, the sub-shapes after any letter segmentation are included in Figure 2 For a given letter to be engraved, the segmentation information of the given letter (including the glyph category and start and end points of each sub-track after segmentation) is read from the letter segmentation information library.

[0033] S12. Construct a teaching sub-trajectory for each glyph category. Each teaching sub-trajectory describes the three-dimensional coordinate point set of the glyph category. All teaching sub-trajectories constitute a sub-skill primitive library. The three-dimensional coordinate point sets of each teaching sub-trajectory are all located on the OXY plane of the robot arm's base coordinate system. In this embodiment, a teaching sub-trajectory is constructed for each of the nine categories, and these nine teaching sub-trajectories constitute the sub-skill primitive library.

[0034] According to the glyph category of the sub-trajectory of a given letter to be engraved read from the letter information library, the corresponding teaching sub-trajectory is extracted from the sub-skill primitive library for subsequent trajectory generalization.

[0035] S13, using discrete dynamic motion primitive models DMPs to generate a planar reference trajectory, specifically:

[0036] For a given letter to be engraved and a specified position and size, a discrete dynamic motion primitive model DMPs is used to generate a planar reference trajectory. The model is expressed as:

[0037]

[0038] Among them, x is the system state (such as position coordinates or joint angles), x0 and x g They represent the initial value and target value of the motion trajectory respectively; α and β represent the system stiffness and damping respectively, and α = β is usually set 2 / 4 to ensure the stability of the system and make x converge to xg ; τ is the time scaling constant, and by changing this parameter, the trajectory can be generalized in the time dimension; s is a phase variable that decays exponentially, and γ and μ are defined positive constants; the nonlinear term f(s) is specifically:

[0039]

[0040] ψ i (s)=exp(-h i (sc i ) 2 ), i=1,2,…N

[0041] Among them, ψ i (s) represents a set of basis functions with Gaussian distribution, ω i is the corresponding weight, N is the number of basis functions, c i and h i are ψ i (s) center point and width value;

[0042] The glyph category and start and end points of each sub-trajectory after the given letter to be engraved is read in step S11, and the teaching sub-trajectory extracted in step S12 are input into the discrete DMPs model to generalize and generate a planar reference sub-trajectory. The planar reference trajectory of the given letter is obtained by splicing and combining all the reference sub-trajectories and performing translation and scaling operations according to the specified position and size.

[0043] In this embodiment, the parameters of the discrete DMPs model can be selected as shown in Table 1 below.

[0044]

[0045] Table 1

[0046] S2. Use a 3D camera to acquire point cloud information of the complex surface of interest, and perform eye-outside-hand calibration on the camera to obtain the transformation relationship between the camera coordinate system and the robot arm base coordinate system. Based on the transformation relationship, obtain the surface point cloud in the robot arm base coordinate system. Based on the surface trajectory planning method of the wrapping projection algorithm, project the generated plane reference trajectory onto the surface to obtain the surface reference trajectory and surface normal vector;

[0047] In this embodiment, it specifically includes:

[0048] S21. Select a high-precision RGBD depth camera and an 8×6×25mm checkerboard calibration plate. Fix the camera directly above the engraving platform and take a bird's-eye view of the platform. Fix the calibration plate directly below the camera. Use the eye-in-hand robot hand-eye calibration method to read the camera's built-in internal parameters, take a frontal and clear checkerboard image, and call the OpenCV library for processing to obtain the external parameter matrix, that is, the transformation relationship from the calibration plate coordinate system to the camera coordinate system.

[0049] S22: Control the robot's end-mounted engraving tool to move to the origin of the calibration plate, a point on the X-axis, and a point on the Y-axis, and record the coordinates of these three points in the base coordinate system. Based on these three coordinates, calculate the position of the calibration plate coordinate system in the robot arm base coordinate system. Then calculate the transformation relationship from the camera coordinate system to the base coordinate system:

[0050]

[0051] S23. Clamp the material with a complex surface to be engraved using a bench vise. The vise is then secured to the engraving platform and positioned directly below the camera. The camera's depth map function is enabled to obtain point cloud data for the complex surface of interest. This data is then converted from the camera coordinate system to the base coordinate system to obtain the surface point cloud in the base coordinate system.

[0052] S24. Using a surface trajectory planning method based on a wrapping projection algorithm, the generated plane reference trajectory is projected onto the surface to obtain the surface reference trajectory and the surface normal vector.

[0053] like Figure 3 As shown, step S24 is specifically as follows:

[0054] The target surface point cloud after transformation is recorded as S, and the plane trajectory point set is recorded as C = {c1, c2, ..., c m};

[0055] Select a starting reference projection point c0 in the plane and specify a reference projection vector pointing to the target surface. Project c0 along the reference vector Projection on the target surface to obtain d0;

[0056] Search for k neighboring points of d0 in S, and use the least squares method to fit the tangent plane H0 and normal vector at d0 according to and Calculate the rotation matrix R0 of the two;

[0057] Then project the m plane trajectory points separately to a certain point c i , according to the preset projection step l, the line segment c0ci Divided into several segments, when c0c i ≤l, there is no intermediate point; when c0c i When >l, the set of intermediate points is recorded as:

[0058] {c i,j |j=1,…,q i}

[0059]

[0060] When c0c i >l, first the first segment c0c i,1 Projection, use the above rotation matrix R0 to transform the plane line segment c0c i,1 Transform to the tangent plane H0 and get the middle projection point Then search in S k neighboring points of the tangent plane H and fit them together to form a new tangent plane H i,1 and normal vector And according to and Calculate the new rotation matrix R i,1 .Pick On the tangent plane H i,1 The projection point d on i,1 As c i,1 The projection point on the target surface; then iteratively calculate the projection point of all intermediate points and the normal vector at that point until c is obtained i Projection point d on the target surface i and the normal vector at that point

[0061] When c0c i When ≤l, directly c0c i Projection point d is obtained by the above method i and the normal vector at that point

[0062] Take off a little c i+1 Repeat the above operation and finally get the surface trajectory point set D = {d1, d2, ..., d m} and the corresponding normal vector set

[0063] S3. Parameter identification of the engraving tool at the end of the robotic arm based on the six-dimensional force sensor is performed, gravity compensation is performed on the engraving tool mounted at the end of the robotic arm, the influence of gravity of the end engraving tool is eliminated, and the interaction force between the external environment and the robotic arm is obtained;

[0064] In this embodiment, a six-axis Elite robotic arm and an ATI-MINI45 six-axis force sensor are used. The force sensor is mounted at the end of the robotic arm, and then the engraving tool is mounted on the force sensor. The engraving tool includes a flange, an engraving motor, and an engraving tool head. Ensure that the Z axis of the robotic arm end, the Z axis of the force sensor, and the center axis of the engraving tool are aligned. Step S3 specifically includes:

[0065] S31, change the posture of the engraving tool at the end of the robot arm, and obtain N (N≥3) groups of α under different postures i ,β i ,γ i Six-axis force sensor reading F xi, F yi ,F zi ,M xi ,M yi ,M zi , i=1,2,…,N, the force and torque have the following relationship:

[0066]

[0067] Among them, F x0 ,F y0 ,F z0 ,M x0 ,M y0 ,M z0 is the sensor zero value, F x ,F y ,F z ,M x ,M y ,M z is the sensor measurement value. After simplification, the least square method can be used to obtain the center of mass coordinates (cx, cy, cz) and three constant values ​​​​k1, k2, k3.

[0068] Let's denote the world coordinate system as O0X0Y0Z0 and the base coordinate system as O1X1Y1Z1. Assuming the robotic arm is installed at an inclination angle, O0X0Y0Z0 is first rotated by angle U around the X-axis and then by angle V around the Y-axis to obtain O1X1Y1Z1. The force sensor coordinate system, O2X2Y2Z2, can be obtained by rotating the base coordinate system O1X1Y1Z1 by angle α around the Z1 axis, angle β around the Y2 axis, and angle γ around the X2 axis. Based on the N sets of posture data above, the load gravity, force sensor force, and zero point have the following relationship:

[0069]

[0070] Where G is the load gravity and I is the 3×3 unit matrix. After simplification, the force zero point F of the force sensor can be obtained by the least squares method. x0 ,F y0 ,F z0 and three constant values ​​Lx ,L y ,L z .

[0071] According to the above results, calculate the end engraving tool gravity G, installation inclination angle U, V and sensor torque zero point M x0 ,M y0 ,M z0 , completing the entire parameter identification process.

[0072] In this embodiment, N (N=17) groups of six-axis force sensor readings in different postures are shown in Table 2 below.

[0073]

[0074]

[0075] Table 2

[0076] S32, performing gravity compensation on the engraving tool at the end of the robotic arm, specifically:

[0077] Get the current robot arm end posture matrix and force sensor reading F x ,F y ,F z ,M x ,M y ,M z , according to the identification parameters obtained in step S31, the real-time gravity compensation of the end engraving tool is realized, and the interaction force and torque between the external environment and the robotic arm are calculated:

[0078]

[0079] Among them, F ex 、F ey 、F ez are the interaction forces of the sensor X, Y, and Z axes, M ex 、M ex 、M ez They are the interaction torques of the sensor X, Y, and Z axes, G x , G y , G z They are the end engraving tool gravity of the sensor X, Y, and Z axes respectively.

[0080] S4. The robotic arm uses an adaptive admittance controller based on radial basis function neural network (RBFNN) to generate the desired engraving trajectory. At the same time, the interaction force between the environment and the robotic arm tracks the desired interaction force, achieving smooth engraving operation of the robotic arm.

[0081] In this embodiment, the robotic arm adopts an adaptive admittance controller based on RBFNN, which is expressed as:

[0082]

[0083] Among them, B and K are damping parameters and stiffness parameters respectively, F d represents the expected interaction force between the environment and the robotic arm, F e Represents the actual interaction force between the environment and the robotic arm, e=ΔX e Indicates the compensation amount of the manipulator's posture error caused by the force of the environment on the manipulator. It represents the rate of change of the posture error compensation amount, X e =X d -X r represents the pose error, X r represents the reference pose, i.e., the surface reference trajectory of step S24, X d represents the desired pose generated.

[0084] By optimizing the cost function Adaptively adjust the damping and stiffness parameters so that the actual interaction force between the environment and the manipulator tracks the expected interaction force; the damping and stiffness parameters are specifically expressed as:

[0085] B=w B T h(z)+ξ B

[0086] K=w K T h(z)+ξ K

[0087] Where z = F d -F e is the RBFNN input, w B ,w K represents the weight vector, n is the number of radial basis functions, ξ B ,ξ K Represents the approximate error of the neural network, h(z) represents the vector composed of Gaussian functions;

[0088] Use the gradient descent method to get the weight update rate of the cost function J

[0089]

[0090] in, Denote the weighted estimates of damping and stiffness, η B ,η K denote the learning rates of damping and stiffness, respectively, and are defined as positive constants;

[0091] Calculate the damping and stiffness parameters of the adaptive admittance controller, and the weight w after iterative trainingB ,w K The learning is obtained, and finally the damping and stiffness parameters B, K converge. In this embodiment, the initial weight is set to w B =w K =0, weight learning rate η B =1,η K =100, the initial impedance parameters are set to B = 10N·s / mm, K = 400N / mm. The desired interaction force is selected as a constant value F d =5N. Figure 4 and Figure 5 As shown, after iterative training, the weight W B ,W K After learning, the damping and stiffness parameters B and K converge to appropriate values.

[0092] The surface reference trajectory obtained in step S24 and the interaction force between the external environment and the robotic arm in step S32 are fed back to the RBFNN-based adaptive admittance controller to generate the desired engraving trajectory; the end pose trajectory is converted to the joint space through the robot inverse kinematics, and the robotic arm is controlled by the joint controller of the robotic arm, such as Figure 6 As shown, the interaction force between the environment and the robotic arm is made to track the expected interaction force, achieving smooth engraving operation of the robotic arm.

[0093] In another embodiment, a robot autonomous engraving system for complex curved surfaces is provided. The system adopts the robot autonomous engraving method of the above embodiment. Figure 7 As shown, the system includes a plane reference trajectory generation module, a surface reference trajectory and surface normal vector generation module, a parameter identification and gravity compensation module, and an adaptive admittance control module;

[0094] Generate a planar reference trajectory module, which is used to segment all English letters and build a letter segmentation information library containing all letter segmentation information; construct a teaching sub-trajectory for each letter category to build a sub-skill primitive library; and use discrete dynamic motion primitive models (DMPs) to generate planar reference trajectories.

[0095] The module generates surface reference trajectories and surface normal vectors. It uses a 3D camera to acquire point cloud information of the complex surface of interest and calibrates the camera with the eye outside the hand to obtain the transformation relationship between the camera coordinate system and the robot base coordinate system. Based on the transformation relationship, the surface point cloud in the robot base coordinate system is obtained. The surface trajectory planning method based on the wrapping projection algorithm is used to project the generated plane reference trajectory onto the surface to obtain the surface reference trajectory and surface normal vector.

[0096] The parameter identification and gravity compensation module is used to identify the parameters of the engraving tool at the end of the robotic arm based on the six-dimensional force sensor, perform gravity compensation on the engraving tool at the end of the robotic arm, eliminate the influence of gravity on the end engraving tool, and obtain the interaction force between the external environment and the robotic arm;

[0097] The adaptive admittance control module adopts an adaptive admittance controller based on radial basis neural network (RBFNN) to generate the desired engraving trajectory, while making the interaction force between the environment and the robotic arm track the desired interaction force, thus realizing smooth engraving operation of the robotic arm.

[0098] The present invention provides a method and system for autonomous robot engraving operations on complex surfaces. When reference trajectories of different letters or different positions and sizes need to be generated each time, the re-teaching and planning of traditional teaching methods are avoided, which greatly reduces the workload and operation difficulty. At the same time, the robot is given a more efficient skill generalization ability to adapt to the changing engraving operation requirements; the present invention projects the planar reference trajectory onto the complex curved surface, expands the planar engraving operation to the curved surface, greatly enriches the environmental categories of the robot engraving operation, and adapts to complex engraving operation surfaces; the present invention obtains the identification parameters of the end load of the robotic arm and the six-dimensional force sensor, and eliminates the influence of the gravity of the end tool, thereby obtaining the interaction force between the external environment and the robotic arm; the present invention completes the task of the robot to perform flexible engraving operations, and at the same time not only ensures the interaction safety between the robot and the environment, but also obtains good engraving accuracy and engraving effects.

[0099] It should also be noted that, in this specification, terms such as "comprises", "includes" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or apparatus comprising the element.

[0100] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A robot autonomous engraving method for complex curved surfaces, characterized in that: The following steps are involved: S1. Design a glyph segmentation method for English letters and build a letter segmentation information database; construct a teaching sub-trajectory for each glyph category and build a sub-skill primitive library; use discrete dynamic motion primitive models (DMPs) to generate planar reference trajectories; S2. Use a 3D camera to obtain point cloud information of the complex surface of interest, and calibrate the camera with the eye outside the hand to obtain the transformation relationship between the camera coordinate system and the robot arm base coordinate system. According to the transformation relationship, the surface point cloud in the robot arm base coordinate system is obtained; the surface trajectory planning method based on the wrapping projection algorithm is used to project the generated plane reference trajectory onto the surface to obtain the surface reference trajectory and surface normal vector, specifically: The target surface point cloud after transformation is recorded as S, and the plane trajectory point set is recorded as C = {c1, c2, ..., c m }; Select a starting reference projection point c0 in the plane and specify a reference projection vector pointing to the target surface. Project c0 along the reference vector Projection onto the target surface yields d0; Search for k neighboring points of d0 in S, and use the least squares method to fit the tangent plane H0 and normal vector at d0 according to and Calculate the rotation matrix R0 of the two; Project the m plane trajectory points separately, and project them to a certain point c i , according to the preset projection step l, the line segment c0c i Divided into several segments, when c0c i ≤l, there is no intermediate point; when c0c i When >l, the set of intermediate points is recorded as: When c0c i >l, first the first segment c0c i,1 Perform projection and use the rotation matrix R0 to transform the plane segment c0c i,1 Transform to the tangent plane H0 and get the middle projection point Search in S k neighboring points of the tangent plane H and fit them together to form a new tangent plane H i,1 and normal vector And according to and Calculate the new rotation matrix R i,1 ;Pick On the tangent plane H i,1 The projection point d on i,1 As c i,1 The projection point on the target surface; then iteratively calculate the projection point of all intermediate points and the normal vector at that point until c is obtained i Projection point d on the target surface i and the normal vector at that point When c0c i When ≤l, directly c0c i Projection point d is obtained by the above method i and the normal vector at that point Take off a little c i+1 Repeat the above operation and finally get the surface trajectory point set D = {d1, d2, ..., d m } and the corresponding normal vector set S3. Parameter identification of the engraving tool at the end of the robotic arm based on the six-dimensional force sensor, gravity compensation of the engraving tool at the end of the robotic arm, eliminating the influence of gravity of the engraving tool at the end of the robotic arm, and obtaining the interaction force between the external environment and the robotic arm; S4. The robotic arm uses an adaptive admittance controller based on a radial basis function neural network (RBFNN) to generate the desired engraving trajectory. At the same time, the interaction force between the environment and the robotic arm tracks the desired interaction force, achieving a smooth engraving operation. Specifically: The obtained surface reference trajectory and the interaction force between the external environment and the robotic arm are fed back into the adaptive admittance controller based on the radial basis function neural network (RBFNN) to generate the desired engraving trajectory. The end-position trajectory is converted to the joint space through robot inverse kinematics, and the robotic arm is controlled through the joint controller of the robotic arm, so that the interaction force between the environment and the robotic arm tracks the desired interaction force, achieving a smooth engraving operation by the robotic arm. The adaptive admittance controller based on radial basis function neural network RBFNN is expressed as: Among them, B and K are damping parameters and stiffness parameters respectively, F d represents the expected interaction force between the environment and the robotic arm, F e Represents the actual interaction force between the environment and the robotic arm, e=ΔX e Indicates the compensation amount of the manipulator's posture error caused by the force of the environment on the manipulator. It represents the rate of change of the posture error compensation amount, X e =X d -X r represents the pose error, X r represents the reference pose, i.e. the surface reference trajectory of step S2, X d represents the desired pose generated; By optimizing the cost function Adaptively adjust the damping and stiffness parameters so that the actual interaction force between the environment and the manipulator tracks the expected interaction force; the damping and stiffness parameters are specifically expressed as: B=w B T h(z)+ξ B K=w K T h(z)+ξ K Where z = F d -F e is the RBFNN input, w B ,w K represents the weight vector, n is the number of radial basis functions, ξ B ,ξ K Represents the approximate error of the neural network, h(z) represents the vector composed of Gaussian functions; Use the gradient descent method to get the weight update rate of the cost function J in, Denote the weighted estimates of damping and stiffness, η B ,η K denote the learning rates of damping and stiffness, respectively, and are defined as positive constants; Calculate the damping and stiffness parameters of the adaptive admittance controller, and the weight w after iterative training B ,w K After learning, the damping and stiffness parameters B and K finally converge.

2. The method for autonomous robot engraving of complex curved surfaces according to claim 1, characterized in that: In step S1, a method for segmenting the English letters is designed and a letter segmentation information database is constructed, specifically: Perform glyph segmentation on all English letters and build a letter segmentation information library containing all letter segmentation information. The information library contains the glyph category and starting and ending points of each sub-track after each letter segmentation; for a given letter to be engraved, read the segmentation information of the given letter from the letter segmentation information library.

3. The method for autonomous robot engraving of complex curved surfaces according to claim 2, characterized in that: In step S1, a teaching sub-track is constructed for each glyph category to form a sub-skill primitive library, specifically: Construct a teaching sub-track for each glyph category. The teaching sub-track describes the three-dimensional coordinate point set of the shape of the glyph category to which it belongs. All teaching sub-tracks constitute a sub-skill primitive library. According to the glyph category of the sub-track of a given letter to be engraved read from the letter information library, the corresponding teaching sub-track is extracted from the sub-skill primitive library.

4. The method for autonomous robot engraving of complex curved surfaces according to claim 3, characterized in that: In step S1, the planar reference trajectory is generated using discrete dynamic motion primitive models DMPs, specifically: For a given letter to be engraved and a specified position and size, a discrete dynamic motion primitive model DMPs is used to generate a planar reference trajectory. The model is expressed as: Among them, x is the system state, x0 and x g denote the initial value and target value of the motion trajectory respectively; α and β denote the system stiffness and damping respectively; τ is the time scaling constant; s is a phase variable that decays exponentially, γ and μ are defined positive constants; the nonlinear term f(s) is specifically: ψ i (s)=exp(-h i (s-c i ) 2 ),i=1,2,…N Among them, ψ i (s) represents a set of basis functions with Gaussian distribution, ω i is the corresponding weight, N is the number of basis functions, c i and h i are ψ i (s) center point and width value; The glyph category and starting and ending points of each sub-track after the given letter to be engraved is segmented, as well as the extracted teaching sub-track, are input into the discrete DMPs model to generalize and generate a planar reference sub-track. The planar reference track of the given letter is obtained by splicing and combining all the reference sub-tracks and performing translation and scaling operations according to the specified position and size.

5. The method for autonomous robot engraving of complex curved surfaces according to claim 1, characterized in that: Step S2 specifically includes: S21. Use the eye-on-hand robot camera installation method and the robot hand-eye calibration method to calibrate the camera's extrinsic parameters and obtain the extrinsic parameter matrix, that is, the transformation relationship from the calibration plate coordinate system to the camera coordinate system. S22, according to the position and posture of the calibration plate coordinate system in the robot base coordinate system Calculate the transformation relationship from the camera coordinate system to the robotic arm base coordinate system: S23, converting the point cloud data of the complex surface of interest obtained from the 3D camera from the camera coordinate system to the robot arm base coordinate system to obtain the surface point cloud in the robot arm base coordinate system.

6. The method for autonomous robot engraving of complex curved surfaces according to claim 5, characterized in that: In step S3, the parameters of the engraving tool at the end of the robotic arm based on the six-dimensional force sensor are identified, specifically: Change the posture of the engraving tool at the end of the robotic arm to obtain N groups of α under different postures i ,β i ,γ i Six-axis force sensor reading F xi ,F yi ,F zi ,M xi ,M yi ,M zi , i=1,2,…,N, N≥3, the relationship between force and torque is: Among them, F x0 ,F y0 ,F z0 ,M x0 ,M y0 ,M z0 is the sensor zero value, F x ,F y ,F z ,M x ,M y ,M z is the sensor measurement value; after simplification, the least square method is used to obtain the center of mass coordinates of the end engraving tool (cx, cy, cz) and the three constant values ​​​​k1, k2, k3; The world coordinate system is O0X0Y0Z0, and the base coordinate system is O1X1Y1Z1. Assuming that the robot arm has an installation inclination angle, O0X0Y0Z0 is first rotated around the X axis by angle U, and then around the Y axis by angle V to obtain O1X1Y1Z1. The force sensor coordinate system is O2X2Y2Z2, which is obtained by rotating the base coordinate system O1X1Y1Z1 around the Z1 axis by angle α, around the Y2 axis by angle β, and around the X2 axis by angle γ. According to N sets of posture data, the relationship between the load gravity, the force of the force sensor and the zero point is: Where G is the load gravity, I is the 3×3 identity matrix, and L x ,L y ,L z is a constant; after simplification, the force zero point F of the force sensor is obtained by the least square method. x0 ,F y0 ,F z0 and three constant values ​​L x ,L y ,L z ; According to the above results, calculate the end engraving tool gravity G, installation inclination angle U, V and sensor torque zero point M x0 ,M y0 ,M z0 , completing the parameter identification process.

7. The method for autonomous robot engraving of complex curved surfaces according to claim 6, characterized in that: In step S3, gravity compensation is performed on the engraving tool at the end of the robotic arm, specifically: Get the current robot arm end posture matrix and force sensor reading F x ,F y ,F z ,M x ,M y ,M z , based on the identification parameters obtained in the parameter identification step, the real-time gravity compensation of the end engraving tool is realized, and the interaction force and torque between the external environment and the robotic arm are calculated: Among them, F ex 、F ey 、F ez They are the interaction forces of the sensor X, Y, and Z axes, M ex 、M ex 、M ez They are the interaction torques of the sensor X, Y, and Z axes, G x , G y , G z are the end-tool gravity of the sensor's X, Y, and Z axes, respectively.

8. A robot autonomous engraving system for complex curved surfaces, characterized by: The system adopts the robot autonomous engraving operation method according to any one of claims 1 to 7, and the system includes a plane reference trajectory generation module, a surface reference trajectory and surface normal vector generation module, a parameter identification and gravity compensation module, and an adaptive admittance control module; Generate a planar reference trajectory module, which is used to segment all English letters and build a letter segmentation information library containing all letter segmentation information; construct a teaching sub-trajectory for each letter category to build a sub-skill primitive library; and use discrete dynamic motion primitive models (DMPs) to generate planar reference trajectories. The module generates surface reference trajectories and surface normal vectors. It uses a 3D camera to acquire point cloud information of the complex surface of interest and calibrates the camera with the eye outside the hand to obtain the transformation relationship between the camera coordinate system and the robot base coordinate system. Based on the transformation relationship, the surface point cloud in the robot base coordinate system is obtained. The surface trajectory planning method based on the wrapping projection algorithm is used to project the generated plane reference trajectory onto the surface to obtain the surface reference trajectory and surface normal vector. The parameter identification and gravity compensation module is used to identify the parameters of the engraving tool at the end of the robotic arm based on the six-dimensional force sensor, perform gravity compensation on the engraving tool at the end of the robotic arm, and obtain the interaction force between the external environment and the robotic arm; The adaptive admittance control module adopts an adaptive admittance controller based on radial basis neural network (RBFNN) to generate the desired engraving trajectory, while making the interaction force between the environment and the robotic arm track the desired interaction force, thus realizing smooth engraving operation of the robotic arm.

Images