Gluing path planning method based on industrial robot

Abstract

The invention relates to a gluing path planning method based on an industrial robot. According to the method, a real component gluing process is taken as an example, a simulation model is established by applying three-dimensional software, a gluing path is generated by utilizing an automatic path function in the simulation software, and the gluing path is optimized by adjusting and optimizing point position poses, an axis parameter configuration method, collision monitoring and track operation simulation technical means to generate a motion track. Compared with a traditional method for generating a track path through industrial robot field teaching point programming, the path path generated after automatic path function optimization in virtual simulation software is adopted is better in fitting degree with the geometric body curved surface contour, the gluing thickness is more uniform, and the gluing quality and efficiency are higher.

Description

A glue application path planning method based on industrial robots Technical Field

[0001] This invention relates to computer graphics, 3D reconstruction, point cloud registration, and adhesive application path planning, specifically a method for planning adhesive application paths by performing 3D reconstruction of the adhesive application surface. Background Technology

[0002] With the rise of China's comprehensive national strength, the aerospace industry is also developing rapidly. In the production process, adhesive coating on the surface of structural panels is a crucial step. Structural panels are used to construct load-bearing structures, manufacture thermal barrier systems, and ensure the normal operation of electronic equipment. However, the current production process of adhesive coating on structural panel surfaces still faces many challenges. For example, it mainly relies on manual coating, which has low automation, long cycles, and heavy workloads. Furthermore, manual coating suffers from low motion stability and accuracy, making it difficult to control coating quality and resulting in poor consistency. Additionally, it lacks versatility for workpieces with special shapes. Therefore, researching ways to improve the automation level and accuracy of workpiece adhesive coating is essential. Applying 3D reconstruction to automated surface coating robots to help them determine an optimal trajectory planning scheme has become a research hotspot both domestically and internationally. Although the application of 3D reconstruction technology to workpiece processing, surface coating, and adhesive coating production processes is becoming increasingly widespread, it also faces technical challenges such as improving adaptability to adhesive coating operations on different workpiece surfaces, increasing work efficiency, and improving the accuracy of related algorithms to ensure the quality of workpiece surface adhesive coating.

[0003] In the production process, applying adhesive to the surface of structural panel workpieces is a crucial step. By covering the surface with a special adhesive, effects such as heat insulation, bonding, and scratch resistance can be achieved. To improve the automation of the surface adhesive application process, equipment such as laser scanners can be used to collect point cloud data of the workpiece surface and convert it into a 3D model using relevant algorithms. Relying on 3D reconstruction technology, high-precision scanning and measurement of the workpiece surface can be achieved, leading to more accurate adhesive application plans. Furthermore, compared to traditional manual adhesive application, using automated equipment such as robots can improve application efficiency and consistency. Moreover, 3D reconstruction can provide accurate models for adhesive application of aerospace industry workpieces, helping automated adhesive application equipment plan application paths, predict application amounts, and check application quality. In addition, it can reduce human error during the adhesive application process.

[0004] In addition, regarding workpiece adhesive application, current surface adhesive application methods include contact and non-contact methods. Non-contact slit-type wide-nozzle applicators offer advantages such as high application speed, good film thickness consistency, wide viscosity range of the coating liquid, fewer coating defects (closed-loop system), and high coating liquid utilization. Traditional contact dispensing technology can no longer meet market demands. Furthermore, point cloud data obtained by scanning the workpiece using a laser point cloud scanning device is processed for workpiece positioning and parameter extraction. A precise surface point cloud model is then generated using 3D reconstruction technology to guide the trajectory planning and coating path optimization of the adhesive application robot. Improving the automation level and coating quality stability of adhesive application is a current research focus for automated surface adhesive application robots. Simultaneously, slit-type wide-nozzle applicators place more stringent requirements on trajectory planning and motion control accuracy, making the design of high-precision adhesive trajectory fitting algorithms a crucial step in designing automated adhesive application robots. Summary of the Invention

[0005] To address the aforementioned shortcomings of existing technologies, this invention provides a glue application path planning method based on industrial robots. It applies 3D reconstruction technology to automated surface glue application robots to improve the automation level and control precision of surface glue application equipment. This invention employs a slotted, wide-nozzle glue application method for surface glue application on large workpieces with irregular, small-curvature, locally convex surfaces due to lack of processing or deformation, such as structural panels. Laser point cloud equipment is used for workpiece positioning, position parameter extraction, and 3D reconstruction. Glue application trajectory planning improves the robot's efficiency and accuracy in surface glue application on such workpieces. This has practical engineering significance for improving the automation level of surface glue application equipment and enhancing the control precision of the glue application process in the aerospace industry.

[0006] The technical solution adopted by this invention to achieve the above objectives is: a glue application path planning method based on an industrial robot, which optimizes the glue application trajectory by performing the following steps, the method including:

[0007] Step 1: Use a laser scanner to scan the structural plate workpiece in segments multiple times, perform point cloud processing, and obtain the actual point cloud data of the structural plate workpiece; make the position and coordinate system of the structural plate workpiece DXF drawing file correspond to the coordinate system of the actual point cloud data, and obtain the transformation matrix dTp;

[0008] Step 2: By selecting the glue application area on the DXF drawing, generate the corresponding three-dimensional vertex information, and use the transformation matrix dTp to transform it to the actual point cloud data coordinate system, which is used to extract the glue application working area from the actual point cloud data using the crophull filter.

[0009] Step 3: For the point cloud data of the obtained glue application area, use point cloud equidistant segmentation to obtain glue application trajectory points and filter and segment them; perform NURBS spatial curve fitting on the glue application trajectory points.

[0010] Step 4: Optimize the adhesive application trajectory during the adhesive application process.

[0011] The point cloud processing includes data preprocessing, point cloud stitching, filling point cloud gaps, point cloud smoothing and densification, and three-dimensional surface reconstruction.

[0012] The data preprocessing includes point cloud data preprocessing using statistical filtering, pass-through filtering, and voxel filtering.

[0013] The point cloud stitching method is point cloud registration, which includes obtaining an initial transformation matrix using 3D-NDT coarse registration, and performing fine point cloud registration using the ICP algorithm to obtain a complete three-dimensional surface reconstruction model and model parameters in the three-dimensional reconstruction process.

[0014] Mitigating point cloud vulnerabilities involves generating reliable intercept boundaries through point cloud interpolation optimization.

[0015] The DXF drawing is read and displayed by the host computer, and the transformation matrix dTp between the drawing coordinate system and the actual point cloud coordinate system is obtained, which is used to realize the workpiece positioning.

[0016] The segmentation involves using a Euclidean clustering algorithm to segment the point cloud and extract each trajectory point; the filtering involves using a spherical search algorithm to discard points whose Euclidean distance to neighboring points of each trajectory point is less than a threshold, considering them invalid points.

[0017] In the NURBS curve implementation process, the points on the curve are obtained by calculating the basis function values ​​of each point and performing a weighted summation according to the weights.

[0018] The adhesive application process is optimized, including optimizing the spacing on the y-axis of the adhesive application trajectory.

[0019] The present invention has the following beneficial effects and advantages:

[0020] 1. This invention defines a design scheme for a component gluing system. Addressing the shortcomings of current manual gluing methods, it utilizes a laser scanner to design a point cloud data acquisition device and applies relevant algorithms to process the point cloud data, achieving automated improvement of the satellite component gluing system. Furthermore, it performs gluing surface positioning extraction, 3D reconstruction of the workpiece, and gluing trajectory extraction on the workpiece to be glued, and evaluates the gluing effect generated by the gluing robot.

[0021] 2. Point cloud data preprocessing is performed using statistical filtering, pass-through filtering, and voxel filtering. Point cloud stitching is then performed using point cloud registration. Coarse registration with 3D-NDT is used to obtain the initial transformation matrix for fine registration. Fine registration is then performed using the ICP algorithm, ensuring that the z-axis error of the stitched point cloud is within 0.026 mm. This yields a complete 3D surface reconstruction model and useful parameters for the 3D reconstruction process, providing support for the subsequent glue application operation of the automated glue application robot for satellite components.

[0022] 3. The adhesive application area is segmented, adhesive trajectory points are extracted, and the adhesive trajectory fitting and optimization are performed. Due to the fluidity of the adhesive itself and the mechanical structure of the adhesive application end effector, the spacing on the y-axis of the adhesive trajectory is optimized during the adhesive application process to ensure the adhesive application effect.

[0023] 4. Experiments were conducted to verify the results in three aspects: point cloud stitching, 3D surface reconstruction, and adhesive application trajectory planning. The test results met the adhesive application quality indicators. It was ensured that the fitted adhesive application trajectory met the adhesive application quality requirements. Attached Figure Description

[0024] Figure 1 is a flowchart of the method of the present invention;

[0025] Figure 2 shows the interface for loading a DXF file on the host computer.

[0026] Figure 3 shows the retrieved DXF drawing;

[0027] Figure 4 shows the 3D-Harris corner detection map;

[0028] Figure 5 illustrates the point cloud vulnerability generation process;

[0029] Figure 6 shows the adhesive application trajectory points;

[0030] Figure 7 shows the glue application trajectory points segmented by Euclidean clustering;

[0031] Figure 8 shows the adhesive application trajectory fitted by the NURBS curve. Detailed Implementation

[0032] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the specific implementation methods of the present invention will be further described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of the present invention; however, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the invention. Therefore, the present invention is not limited to the specific implementations disclosed below.

[0033] Industrial robot technology is increasingly widely used in intelligent manufacturing. In the field of adhesive application, it can effectively solve problems such as low efficiency, poor quality, and harmful gases posing a risk to human health during manual adhesive application. This invention provides an industrial robot-based adhesive application path planning method for applying adhesive to structural panels. The adhesive trajectory planning of this invention involves the selection of the adhesive application area, extraction of adhesive trajectory points, and optimization of the adhesive trajectory fitting. Due to the special workpiece structure of the structural panel and the characteristics of its adhesive application surface, as shown in Figure 1, the adhesive trajectory planning of this invention mainly consists of four steps:

[0034] Step 1: Using the DXF drawing, a manual intervention method is employed. On the host computer's human-computer interface, the adhesive application area on the structural board DXF drawing is selected, and the position of the structural board workpiece and the transformation matrix dTp between the drawing coordinate system and the actual point cloud data coordinate system are obtained.

[0035] The actual point cloud data was collected by scanning satellite components multiple times in segments using a 3D laser scanner. Before stitching together the point cloud data obtained from these multiple scans, preprocessing was required. Multiple segments of point cloud data were stitched together using point cloud registration, ensuring that the z-axis error met the adhesive application accuracy. Finally, the processed point cloud data was smoothed and densified before 3D surface reconstruction.

[0036] As shown in Figure 2, the interface for loading the DXF file by the host computer is as follows: before reading the file into the scene member variable QGraphicsScene*m_Scene of the drawing board, the read primitive information is divided into five categories. The primitive information in the original DXF file is drawn using the QPen class. The above data structure is used to generate primitives for manual selection of drawings in actual glue application operations.

[0037] Step 2: By selecting the glue application area on the DXF drawing, as shown in Figure 3, the corresponding three-dimensional vertex information is generated and transformed to the actual point cloud data coordinate system using the transformation matrix dTp. In this way, the glue application working area is extracted from the actual point cloud data using the crophull filter.

[0038] For two-dimensional pixel images, commonly used feature points include SIFT, SURF, ORB, and Harris corner detection. Similarly, for three-dimensional point cloud data, there is 3D-Harris corner detection, an extension of Harris corner detection. Since 3D-Harris corner detection uses the normal vectors of the point cloud data for detection and extraction, it may not necessarily obtain the point clouds at the four corners of the required point cloud data. In the application of 3D-Harris corner detection to the obtained point cloud data, it was found that the extracted feature point cloud, i.e., the red point cloud, is not the point cloud data required for engineering, as shown in Figure 4.

[0039] Step 3: The point cloud data of the obtained glue application area is processed by equidistant segmentation of the point cloud to obtain glue application trajectory points, and NURBS spatial curve fitting is performed on the glue application trajectory points.

[0040] When a 3D laser scanner scans and collects data from a structural panel, point cloud gaps occur due to the panel's structure. As shown in Figure 5, the green laser beam emitted by the scanner's laser emitter is blocked by a protruding part of the structural panel, preventing the reflected red laser beam from being collected by the laser receiver. This results in missing point cloud data within the dotted line area, i.e., point cloud gaps. When the gaps do not involve the adhesive application area, the impact on adhesive application quality is minimal. However, if gaps occur in the adhesive application area where data needs to be extracted, it severely affects the extraction of adhesive trajectory points, leading to a decrease in adhesive application quality. Since point cloud interpolation is only effective for dense point clouds in small areas, it cannot repair point cloud gaps over larger areas. Therefore, this invention employs a geometric completion algorithm based on controllable point cloud shape to compensate for these gaps. Step 3 specifically includes:

[0041] Step 1: Coarsely extract the slice index value. Calculate minp and maxp of the point cloud dataset P in the x, y, and z dimensions. The coarsely extracted slice index value id(i) is obtained using the following formula:

[0042]

[0043] Where n is the slice spacing, foror is defined as the traversal and inspection slice index, and p i Defined as point cloud data with slice index i, p min Defined as the minimum coordinate on this slice.

[0044] Step 2: Calculate the slice coordinate interval. Obtain the starting coordinates p of the slice. slice(i) :

[0045] p slice(i) =p min +id(i)×n

[0046] By adding the manually set slice thickness δ to slice(i), the slice coordinate interval can be obtained: (slice(i), slice(i)+δ), which corresponds to the slice index value id(i).

[0047] Step 3: Calculate the point cloud index within the slice thickness. All point cloud data p in the input point cloud dataset P are placed into a container vector, and the corresponding slice indices are initialized to -1. If P contains m point cloud data p(j), then id(i) is used.

[0048] Step 4: Check the index values. Iterate through the slice index values ​​id(i) corresponding to the container vector to see if they are all greater than -1. If so, push the slice index values ​​id(i) of points p(e) that are greater than -1 into the newly created container vector_1; otherwise, discard them. This process is not only used to check whether the slice index value id(i) corresponding to each point cloud data set P is valid, but also serves to remove point cloud data sets p(r) that are not within the slice thickness.

[0049] Step 5: A point cloud data set PP(r) is created within a newly established container vector_1, denoted as P′, representing the point cloud data within each slice. Since the data in each loop has no sequential relationship, this invention utilizes OpenMP to perform parallel multi-threaded optimization on the for loop to improve the slicing speed. The resulting adhesive application trajectory points are shown in Figure 6.

[0050] Step 6: After extracting the point cloud dataset P′ of the adhesive application trajectory points, all the obtained trajectory points are within a single point cloud dataset. If trajectory curve fitting is performed on this point cloud dataset P′ at this point, the resulting adhesive application trajectory will not be what is needed in the project. Therefore, before performing trajectory curve fitting, a point cloud segmentation algorithm is required to segment each trajectory point. Euclidean clustering is used to perform point cloud clustering segmentation on the point cloud dataset P′. Finally, 117 trajectory points are obtained, as shown in Figure 7.

[0051] Since invalid points cannot be guaranteed to be removed during equidistant segmentation, potentially affecting the accuracy of the adhesive application trajectory fitting, a spherical domain search algorithm is used to discard points with an Euclidean distance of less than 2.5 mm from each point ip as invalid points. NURBS is also used to fit the obtained three-dimensional spatial curve with small curvature.

[0052] Step 7: In the NURBS curve implementation process, the points on the curve are obtained by calculating the basis function values ​​of each point and performing a weighted sum according to the weights. The position and weight of the control points can be adjusted to control the curve shape. This paper calculates the control points by inputting point cloud data (i.e., input point cloud data).

[0053] The coefficient matrix is ​​composed of the values ​​of the B-spline basis functions. One set of adhesive application trajectories obtained using NURBS curve fitting is shown in Figure 8.

[0054] Step 4: Optimize the adhesive application trajectory during the application process. Due to the fluidity of the adhesive itself and the mechanical structure of the end effector, the spacing along the y-axis of the application trajectory is optimized to ensure effective adhesive application.

[0055] The embodiments described above will help those skilled in the art to further understand the present invention, but do not limit the present invention in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

Claims

1. A method for adhesive application path planning based on industrial robots, characterized in that, The following steps are performed to optimize the adhesive application trajectory: Step 1: Use a laser scanner to scan the structural plate workpiece multiple times in segments, perform point cloud processing, and obtain the actual point cloud data of the structural plate workpiece; ensure that the position and coordinate system of the DXF drawing file of the structural plate workpiece correspond to the coordinate system of the actual point cloud data, and obtain the transformation matrix dTp; Step 2: Select the adhesive application area on the DXF drawing, generate the corresponding three-dimensional vertex information, and use the transformation matrix dTp to transform it to the coordinate system of the actual point cloud data, which is then used to extract the adhesive application working area from the actual point cloud data using a crophull filter; Step 3: For the extracted adhesive application working area point cloud data, use point cloud equidistant segmentation processing to obtain adhesive application trajectory points and filter and segment them; perform NURBS spatial curve fitting on the adhesive application trajectory points; Step 4: Optimize the adhesive application trajectory during the adhesive application process.

2. The glue application path planning method based on an industrial robot according to claim 1, characterized in that, The point cloud processing includes data preprocessing, point cloud stitching, filling point cloud gaps, point cloud smoothing and densification, and three-dimensional surface reconstruction.

3. The adhesive application path planning method based on an industrial robot according to claim 2, characterized in that, The data preprocessing includes point cloud data preprocessing using statistical filtering, pass-through filtering, and voxel filtering.

4. The glue application path planning method based on an industrial robot according to claim 2, characterized in that, The point cloud stitching method is point cloud registration, which includes obtaining an initial transformation matrix using 3D-NDT coarse registration, and performing fine point cloud registration using the ICP algorithm to obtain a complete three-dimensional surface reconstruction model and model parameters in the three-dimensional reconstruction process.

5. The adhesive application path planning method based on an industrial robot according to claim 2, characterized in that, Mitigating point cloud vulnerabilities includes generating truncation boundaries through point cloud interpolation optimization.

6. The glue application path planning method based on an industrial robot according to claim 1, characterized in that, The DXF drawing is read and displayed by the host computer, and the transformation matrix dTp between the drawing coordinate system and the actual point cloud coordinate system is obtained, which is used to realize the workpiece positioning.

7. The glue application path planning method based on an industrial robot according to claim 1, characterized in that, The segmentation involves using a Euclidean clustering algorithm to segment the point cloud and extract each trajectory point; the filtering involves using a spherical search algorithm to discard points whose Euclidean distance to neighboring points of each trajectory point is less than a threshold, considering them invalid points.

8. The glue application path planning method based on an industrial robot according to claim 1, characterized in that, In the NURBS curve implementation process, the points on the curve are obtained by calculating the basis function values ​​of each point and performing a weighted summation according to the weights.

9. The adhesive application path planning method based on an industrial robot according to claim 1, characterized in that, Optimizing the adhesive application trajectory during the application process includes optimizing the spacing on the y-axis of the adhesive application trajectory.

Images