Laser radar and bim-based automatic obstacle avoidance method for external wall spraying robot

Abstract

This invention discloses an automatic obstacle avoidance method for exterior wall spraying robots based on lidar and BIM, belonging to the field of exterior wall spraying robot technology. It solves the problem that existing exterior wall spraying robots cannot automatically avoid obstacles such as windows or balconies. This invention obtains a point cloud map of the exterior wall through a BIM model and projects the point cloud map into a grid map for planning operations. Based on a two-step point cloud registration method using feature extraction, the lidar point cloud is registered with the point cloud map. This not only realizes the localization of the exterior wall spraying robot, but also obtains information on obstacles such as windows and balconies. This not only improves the accuracy and efficiency of spraying operations, but also reduces human intervention and lowers operational risks, especially in high-altitude operations, greatly improving safety.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of external wall spraying robots, and particularly relates to an automatic obstacle avoidance method for an external wall spraying robot based on a laser radar and BIM. BACKGROUND

[0002] Manual operation is the main construction method for external wall spraying, and high-altitude operation is mainly realized through a liftable basket. The traditional construction method is not only low in efficiency, but also prone to quality problems such as uneven spraying, and has great safety hazards. The volatile toxic and harmful substances in the spraying process also harm the occupational health and safety of construction personnel.

[0003] In recent years, external wall spraying machines as new construction equipment with automation and intelligence have gradually been valued and applied to actual project construction. The application of the external wall spraying robot can avoid the risk of high-altitude operation from the root and improve the quality and efficiency of spraying. However, the external wall spraying robot still has deficiencies in actual application, and cannot actively avoid obstacles such as external wall windows and balconies, which needs manual operation intervention. This deficiency limits the application range and use efficiency of the external wall spraying robot, and it is difficult to automatically spray and construct on a relatively complex wall surface. SUMMARY

[0004] In view of the above problems in the prior art, the present application provides an automatic obstacle avoidance method for an external wall spraying robot based on a laser radar and BIM, which solves the problem that the existing external wall spraying robot cannot automatically avoid obstacles such as windows or balconies.

[0005] In order to achieve the above purpose, the technical scheme adopted by the present application is as follows:

[0006] An automatic obstacle avoidance method for an external wall spraying robot based on a laser radar and BIM is provided, comprising the following steps:

[0007] S1, a BIM model of a construction wall surface is established, the BIM model is converted into a point cloud map, and the point cloud map is converted into a grid map, and the point cloud map and the grid map are imported into the external wall spraying robot. By establishing a detailed BIM model, the specific conditions of the construction wall surface, including the positions of windows and balconies, can be known in advance, which helps to improve the accuracy of subsequent spraying operations. The grid map can clearly show the planar layout of the construction wall surface, especially the positions of obstacles such as windows and balconies, providing effective path planning basis for the robot. At the same time, the use of the grid map simplifies the path planning process, so that the robot can quickly plan the optimal spraying path to avoid the windows and balcony obstacles.

[0008] S2, the laser radar on the outer wall spraying robot is used to obtain real-time laser point cloud of the construction wall surface, main plane points and corner points of the real-time laser point cloud are extracted, and the main plane points and the corner points are real-time registered with the main plane points and the corner points of the point cloud map.

[0009] S3, the Boustrophedon algorithm is used to generate a work path of the outer wall spraying robot avoiding obstacles on the grid map. The Boustrophedon algorithm can effectively plan a continuous and non-repeated spraying path, so that the uniformity and integrity of spraying are ensured.

[0010] S4, the outer wall spraying robot sprays along the work path.

[0011] In the scheme, considering that the construction wall surface belongs to a degenerated scene, there are few features in the scanning range of the laser radar, and the corner feature is provided by the protrusions and depressions with depth changes such as windows and balconies. When real-time positioning is performed, constraints are lacked, so that the pose of the outer wall spraying robot cannot be calculated. Based on a large number of actual observations, it is found that the features of the outer wall surface can be divided into two parts of the main plane and the corner point. Most of the points are concentrated on the main plane after screening, and the corner points are concentrated on the windows, balconies and other parts, and most of them are linear features. Therefore, a two-step point cloud registration method based on feature extraction is used to register the laser point cloud and the point cloud map. The positioning of the outer wall spraying robot is realized, and the information of the window and balcony obstacles is also obtained. The accuracy and efficiency of the spraying work are improved, the manual intervention is reduced, the operation risk is reduced, and the safety is greatly improved, especially in high-altitude operation.

[0012] Further, the method for converting the BIM model into the point cloud map is:

[0013] S1.1, a BIM model of the construction outer wall data in the rvt format is established by using the Revit software. The BIM model includes the window and balcony information on the construction wall surface.

[0014] S1.2, the BIM model is exported as an FBX format file, the geometric information in the FBX file is extracted, and the geometric information is written into a point cloud file to obtain an initial point cloud map.

[0015] S1.3, the initial point cloud map is down-sampled and denoised to obtain an optimized point cloud map. The sampling and denoising of the point cloud map improve the accuracy and reliability of the data.

[0016] Further, the registration method of the real-time laser point cloud and the point cloud map includes the following steps:

[0017] S2.1, identify the main plane points and corner points in the real-time laser point cloud and the point cloud map using the Harris 3D corner point detection algorithm; wherein the main plane point sets in the real-time laser point cloud and the point cloud map are M1 and M respectively, and the corner point sets of the real-time laser point cloud and the point cloud map are F1 and F respectively. The Harris 3D corner point detection algorithm can effectively identify the key feature points in the point cloud, including the main plane points and the corner points

[0018] S2.2, using SVD algorithm to fit the main plane points in each frame of real-time acquired real-time laser point cloud, obtaining the normalized plane equation of the real-time scanning laser main plane:

[0019] A1x+B1y+C1z+D1=0

[0020] wherein A1, B1 and C1 are constants, the normal vector of the real-time scanning laser plane is n 1 =(A1, B1, C1); x, y and z are respectively the three coordinate axes of the real-time scanning laser main plane, and D1 is the offset of the real-time scanning laser plane. Through the SVD singular value decomposition algorithm for plane fitting, the main plane equation of the real-time laser point cloud can be accurately calculated.

[0021] S2.3, obtain the point cloud map main plane normal vector n, and calculate the preliminary transformation matrix T of the real-time scanning laser main plane to the point cloud map main plane.

[0022] S2.4, align the points in M1 to the point cloud map main plane through the preliminary transformation matrix T, and obtain the preliminary registration plane. The application of the preliminary transformation matrix T makes the points in the real-time laser point cloud preliminary aligned with the points in the point cloud map, forming the preliminary registration plane, which simplifies the subsequent fine registration process and improves the speed and accuracy of registration.

[0023] S2.5, project the corner points in F1 to the preliminary registration plane, calculate the covariance matrix H between the points on the preliminary registration plane and the point cloud on the point cloud map by least square method, obtain the rotation matrix R and the translation vector t by singular value decomposition of the covariance matrix H, and synthesize the final transformation matrix T f Through the least square method and singular value decomposition, the accurate rotation matrix and translation vector can be obtained, and then the final transformation matrix is synthesized. It ensures the high accuracy of point cloud registration and provides reliable data support for the determination of real-time pose of the robot.

[0024] S2.6, combined with the initial pose of the external wall spraying robot, determine the real-time pose of the external wall spraying robot through the final transformation matrix T f The final transformation matrix T fThe application of this technology enables exterior wall spraying robots to accurately determine their real-time pose on the wall surface being painted. This ensures high-precision positioning of the robot in dynamic environments, providing important references for subsequent path planning and obstacle avoidance, and improving the efficiency and safety of spraying operations.

[0025] Furthermore, in step S2.3, n = (A, B, C), Where A, B, and C are all constants, A 2 +B 2 +C 2 =1,T y (θ) and T x (σ) represents the transformation matrix for the real-time scanning laser principal plane rotating about its y-axis, rotating about its z-axis, and translating along its x-axis, respectively; θ, σ and σ' are the rotation angles about the y-axis and about the z-axis, respectively. And the offset along the x-axis.

[0026] Furthermore, the expression for the covariance matrix H in step S2.5 is as follows: f′ 1i =T·f i ;

[0027] Among them, f i For points (x) in the real-time laser point cloud corner point set F1 i ,y i ,z i ), f i =(x i ,y i ,z i )∈F1, (x i ,y i ,z i () represents the i-th point in F1;

[0028] f′ 1i f i The transformed point, f′ 1i =(x′) i ,y′ i ,z′ i ), (x′ i ,y′ i ,z′ i ) is (x i ,y i ,z i The transformation points, where i is a constant and n is the total number of corner points in F1; Let F be the centroid of the set of corner points F in the point cloud map. For the initial registration of plane F′1i The center of mass, F′ 1i f′ 1i A set of.

[0029] Furthermore, in step S2.5, H = UΣV T R = VU T ,

[0030] Where Σ is a diagonal matrix, and the elements on the diagonal are singular values ​​arranged in descending order; U and V are both obtained through singular value decomposition, where U is the principal direction orthogonal matrix of the corner point set F in the point cloud map; and V is the initial registration planar corner point set F′. 1i The principal direction orthogonal matrix.

[0031] Furthermore, to prevent the exterior wall painting robot from spraying onto windows and balconies, and to facilitate obstacle avoidance, windows and balconies are designated as impassable areas. Specifically, considering that windows typically recess inwards from the outer surface of the wall, if the robot's spraying path does not account for this depth, it may accidentally spray onto the inner side of the window, leading to unnecessary waste and damage. Balconies typically protrude outwards from the outer surface of the wall; if the robot's spraying path does not account for this protrusion distance, it may accidentally spray onto the outer side of the balcony. Therefore, the method for projecting the point cloud map into a raster map in step S1 is as follows:

[0032] S1.4. Using the thickness direction of the construction wall as the Y-axis, the length direction of the construction wall as the X-axis, and the height direction of the construction wall as the Z-axis, determine the axis position Y0 of the wall thickness direction, and obtain the depth of the window [Y0,Y1] and the depth of the balcony [Y0,Y2]. Y1 and Y2 are the recessed and protruding positions of the window and balcony on the Y-axis, respectively.

[0033] S1.5 Set the projection intervals U1 and U2 for the window and balcony, where U1∈[Y0,Y1-h2], U2∈[Y0+h1,Y2], and h1 and h2 are both set projection thresholds.

[0034] S1.6. Sequentially determine whether the Y-axis coordinate of each point in the point cloud map falls within the projection intervals U1 and U2. If so, project the X-axis and Z-axis coordinates of that point onto the XZ plane along the Y-axis direction; otherwise, ignore the point. This projection rule creates empty areas on the grid map, representing window and balcony obstacles. The exterior wall painting robot will not pass through these empty areas, thus achieving obstacle avoidance for the exterior wall painting robot.

[0035] Furthermore, the exterior wall painting robot uses the Boustrophedon algorithm to plan a "bow"-shaped path to cover the entire grid map, avoiding empty areas. The path generated by the Boustrophedon algorithm is similar to the "bow"-shaped trajectory in farmland cultivation. This path planning method ensures that the robot can efficiently and evenly cover the entire construction wall surface, avoiding repeated spraying and missed areas. This path planning method guarantees the continuity and smoothness of the robot's movement, reduces unnecessary turning and stopping, and improves the efficiency of the spraying operation.

[0036] Furthermore, considering that failure to control the nozzles of the exterior wall spraying robot could lead to repeated spraying of already sprayed areas or spraying of areas that should not be sprayed when avoiding windows, the spraying method of the exterior wall spraying robot during the spraying process is as follows:

[0037] S4.1 Divide the construction wall surface into multiple construction units along the work path. Each construction unit is a rectangular area with length and width of a and b respectively.

[0038] S4.2 Control the exterior wall spraying robot to spray the construction units sequentially along the work path, and accumulate and calculate the X-axis displacement and Z-axis displacement of the spraying robot in each construction unit; after the spraying of a construction unit is completed, clear the X-axis displacement and Z-axis displacement, and start accumulating and calculating the displacement information again to ensure that the displacement caused by windows in each construction unit can be identified and judged until the wall construction is completed.

[0039] S4.3 When the exterior wall spraying robot sprays each construction unit, it determines whether the Z-axis displacement generated by the X-axis displacement of the exterior wall spraying robot within the step size a is less than b. If so, it is determined that the exterior wall spraying robot has made a displacement to avoid obstacles and stops spraying; otherwise, it continues spraying to complete the current construction unit.

[0040] Furthermore, the process includes the following step: S2.7, using a drone to periodically patrol the painting robot. The drone uses visual SLAM to locate itself and then uses this location to inversely resolve the position information of the exterior wall painting robot. The exterior wall painting robot then receives the position information inversely resolved by the drone and performs pose correction. Through the drone's periodic patrol and visual SLAM technology, the position and attitude of the painting robot can be monitored in real time, and pose deviations can be detected and corrected in a timely manner. This not only improves the accuracy of the painting operation but also enhances the robustness and reliability of the system, ensuring the smooth progress of the painting operation.

[0041] This invention discloses an automatic obstacle avoidance method for exterior wall spraying robots based on lidar and BIM, the beneficial effects of which are:

[0042] 1. This invention obtains point cloud maps of the exterior walls through BIM models and projects these point cloud maps into grid maps for planning operations. A two-step point cloud registration method based on feature extraction registers the laser point cloud with the point cloud map, which not only enables the positioning of the exterior wall spraying robot but also obtains information on obstacles such as windows and balconies. This not only improves the accuracy and efficiency of spraying operations but also reduces manual intervention and operational risks, especially in high-altitude operations, greatly enhancing safety.

[0043] 2. This invention realizes the operation route planning of the exterior wall spraying robot to avoid windows through the preset construction unit Boustrophedon algorithm, uses a two-step point cloud registration and positioning algorithm to obtain the operation position of the exterior wall spraying robot, and realizes timed correction through drone patrol, thereby realizing the automatic window avoidance and automatic obstacle avoidance of the exterior wall spraying robot and expanding the application scope of the exterior wall spraying robot. Attached Figure Description

[0044] Figure 1 This is a flowchart illustrating an automatic obstacle avoidance method for exterior wall spraying robots based on lidar and BIM.

[0045] Figure 2 This is a schematic diagram of the projection directions of point cloud maps and raster maps;

[0046] Figure 3 This is a schematic diagram of the construction path planning for an exterior wall spraying robot. Detailed Implementation

[0047] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0048] refer to Figure 1 This embodiment provides an automatic obstacle avoidance method for an exterior wall spraying robot based on lidar and BIM, including the following steps:

[0049] S1. Create a BIM model of the construction wall surface, convert the BIM model into a point cloud map, and then convert the point cloud map into a raster map. Import the point cloud map and the raster map into the exterior wall spraying robot. By creating a detailed BIM model, the specific conditions of the construction wall surface can be understood in advance, including the location of windows and balconies, which helps to improve the accuracy of subsequent spraying operations.

[0050] Specifically, the method for converting a BIM model into a point cloud map is as follows:

[0051] S1.1. Create a BIM model of the construction exterior wall data using Revit software. The BIM model format is rvt, and the BIM model includes information on windows and balconies on the construction wall.

[0052] S1.2 Export the BIM model as an FBX format file, traverse the FBX file to extract the geometric information, and write the geometric information into the point cloud file to obtain the initial point cloud map.

[0053] S1.3. The initial point cloud map is downsampled and denoised to obtain an optimized point cloud map. The sampling and denoising of the point cloud map improves the accuracy and reliability of the data.

[0054] To prevent the exterior wall painting robot from spraying onto windows and balconies, and to facilitate obstacle avoidance, windows and balconies are designated as impassable areas. Specifically, considering that windows typically recess inwards from the exterior of the wall, if the robot's spraying path does not account for this depth, it may accidentally spray onto the inner side of the window, leading to unnecessary waste and damage. Balconies typically protrude outwards from the exterior of the wall; if the robot's spraying path does not account for this protrusion, it may accidentally spray onto the outer side of the balcony. Therefore, the method for projecting the point cloud map into a raster map in step S1 is as follows:

[0055] S1.4, Reference Figure 2 Using the thickness direction of the construction wall as the Y-axis, the length direction of the construction wall as the X-axis, and the height direction of the construction wall as the Z-axis, determine the axis position Y0 of the wall thickness direction, and obtain the depth of the window [Y0,Y1] and the depth of the balcony [Y0,Y2]. Y1 and Y2 are the recessed and protruding positions of the window and balcony on the Y-axis, respectively.

[0056] S1.5 Set the projection intervals U1 and U2 for the window and balcony, where U1∈[Y0,Y1-h2], U2∈[Y0+h1,Y2], and h1 and h2 are both set projection thresholds.

[0057] S1.6. Sequentially determine whether the Y-axis coordinate of each point in the point cloud map falls within the projection intervals U1 and U2. If so, project the X-axis and Z-axis coordinates of that point onto the XZ plane along the Y-axis direction; otherwise, ignore the point. This projection rule creates empty areas on the grid map, representing windows and balcony obstacles. The exterior wall painting robot will not pass through these empty areas, thus achieving obstacle avoidance for the exterior wall painting robot. S2. Use the lidar installed on the exterior wall painting robot to acquire real-time laser point clouds of the construction wall surface, extract the main plane points and corner points of the real-time laser point clouds, and perform real-time registration with the main plane points and corner points of the point cloud map.

[0058] S3. The Boustrophedon algorithm is used to generate the working path of the exterior wall spraying robot to avoid obstacles on the grid map. The Boustrophedon algorithm can effectively plan a continuous and non-repeating spraying path, ensuring the uniformity and integrity of the spraying.

[0059] In this embodiment, reference Figure 3 The exterior wall painting robot uses the Boustrophedon algorithm to plan a "bow"-shaped path to cover the entire grid map, avoiding empty areas. The path generated by the Boustrophedon algorithm is similar to the "bow"-shaped trajectory in farmland cultivation. This path planning method ensures that the robot can efficiently and evenly cover the entire construction wall surface, avoiding repeated spraying and missed areas. This path planning method guarantees the continuity and smoothness of the robot's movement, reduces unnecessary turning and stopping, and improves the efficiency of the spraying operation.

[0060] S4. The exterior wall spraying robot performs spraying operations along the work path.

[0061] Specifically, the registration method between real-time laser point clouds and point cloud maps includes the following steps:

[0062] S2.1. Use the Harris 3D corner detection algorithm to identify principal plane points and corner points in real-time laser point clouds and point cloud maps; where the sets of principal plane points in the real-time laser point cloud and point cloud map are M1 and M, respectively, and the sets of corner points in the real-time laser point cloud and point cloud map are F1 and F, respectively. The Harris 3D corner detection algorithm can effectively identify key feature points in the point cloud, including principal plane points and corner points.

[0063] S2.2. Using the SVD algorithm, perform plane fitting on the principal plane points in the real-time laser point cloud acquired in each frame to obtain the normalized plane equation of the real-time scanning laser principal plane:

[0064] A1x+B1y+C1z+D1=0

[0065] Where A1, B1, and C1 are all constants. The normal vector of the laser plane in real-time scanning is n 1 = (A1, B1, C1); x, y, and z are the three coordinate axes of the real-time scanning laser principal plane, and D1 is the offset of the real-time scanning laser plane. The principal plane equation of the real-time laser point cloud can be accurately calculated by performing plane fitting using the SVD singular value decomposition algorithm.

[0066] S2.3 Obtain the normal vector n of the main plane of the point cloud map, and calculate the preliminary transformation matrix T from the real-time scanning laser main plane to the main plane of the point cloud map.

[0067] In step S2.3, n = (A, B, C) Where A, B, and C are all constants, A 2 +B 2 +C 2 =1,T y (θ) and T x (σ) represents the transformation matrix for the real-time scanning laser principal plane rotating about its y-axis, rotating about its z-axis, and translating along its x-axis, respectively; θ, σ and σ' are the rotation angles about the y-axis and about the z-axis, respectively. And the offset along the x-axis.

[0068] S2.4. Align the points in M1 to the main plane of the point cloud map using the preliminary transformation matrix T to obtain the preliminary registration plane. The application of the preliminary transformation matrix T enables the points in the real-time laser point cloud to be initially aligned with the points in the point cloud map, forming a preliminary registration plane, which simplifies the subsequent fine registration process and improves the speed and accuracy of registration.

[0069] S2.5 Project the corner points in F1 onto the preliminary registration plane. Calculate the covariance matrix H between the point cloud on the preliminary registration plane and the point cloud map using the least squares method. Perform singular value decomposition on the covariance matrix H to obtain the rotation matrix R and translation vector t. Then, synthesize the final transformation matrix T using the rotation matrix R and the translation vector t. f By using the least squares method and singular value decomposition, accurate rotation matrices and translation vectors can be obtained, which can then be used to synthesize the final transformation matrix. This ensures high accuracy in point cloud registration and provides reliable data support for determining the robot's real-time pose.

[0070] Specifically, the expression for the covariance matrix H in step S2.5 is as follows: f′ 1i =T·f i Among them, f i For points (x) in the real-time laser point cloud corner point set F1 i ,y i ,z i ), f i =(x i ,y i ,z i )∈F1, (x i ,y i ,z i Let f' be the i-th point in F1. 1i f i The transformed point, f′ 1i =(x′) i ,y′ i ,z′i ), (x′ i ,y′ i ,z′ i ) is (x i ,y i ,z i The transformation points, where i is a constant and n is the total number of corner points in F1; Let F be the centroid of the set of corner points F in the point cloud map. For the initial registration of plane F′ 1i The center of mass, F′ 1i f′ 1i A set of.

[0071] In step S2.5, H = UΣV T R = VU T , Where Σ is a diagonal matrix, and the elements on the diagonal are singular values ​​arranged in descending order; U and V are both obtained through singular value decomposition, where U is the principal direction orthogonal matrix of the corner point set F in the point cloud map; and V is the initial registration planar corner point set F′. 1i The principal direction orthogonal matrix.

[0072] S2.6. Based on the initial pose of the exterior wall spraying robot, the final transformation matrix T is used to... f Determine the real-time pose of the exterior wall painting robot. Final transformation matrix T f The application of this technology enables exterior wall spraying robots to accurately determine their real-time pose on the wall surface being painted. This ensures high-precision positioning of the robot in dynamic environments, providing important references for subsequent path planning and obstacle avoidance, and improving the efficiency and safety of spraying operations.

[0073] As a further aspect of this embodiment, in step S2.7, a drone is used to periodically patrol the painting robot. The drone uses visual SLAM to locate itself and then uses this location to inversely resolve the position information of the exterior wall painting robot. The exterior wall painting robot then receives the position information resolved by the drone and performs pose correction. Through the timed patrol of the drone and visual SLAM technology, the position and attitude of the painting robot can be monitored in real time, and pose deviations can be detected and corrected in a timely manner. This not only improves the accuracy of the painting operation but also enhances the robustness and reliability of the system, ensuring the smooth progress of the painting operation.

[0074] As a further solution in this embodiment, considering that failure to control the nozzles of the exterior wall spraying robot would lead to repeated spraying of already sprayed areas or spraying of areas that should not be sprayed when avoiding windows, the spraying method of the exterior wall spraying robot during the spraying process is as follows:

[0075] S4.1 Divide the construction wall surface into multiple construction units along the work path. Each construction unit is a rectangular area with length and width of a and b respectively.

[0076] S4.2 Control the exterior wall spraying robot to spray the construction units sequentially along the planned path, and accumulate and calculate the X-axis displacement and Z-axis displacement of the spraying robot in each construction unit; after the spraying of a construction unit is completed, clear the X-axis displacement and Z-axis displacement, and start accumulating and calculating the displacement information again to ensure that the displacement caused by windows in each construction unit can be identified and judged until the wall construction is completed.

[0077] S4.3 When the exterior wall spraying robot sprays each construction unit, it determines whether the Z-axis displacement generated by the X-axis displacement of the exterior wall spraying robot within the step size a is less than b. If so, it is determined that the exterior wall spraying robot has made a displacement to avoid obstacles and stops spraying. Otherwise, it continues to spray and completes the construction of the current construction unit.

[0078] In summary, this solution considers that the construction wall surface is a degraded environment with few features within the LiDAR scan range. Corner features are provided by protrusions and depressions with varying depths, such as windows and balconies, which lack constraints during real-time localization, making it impossible to calculate the pose of the exterior wall painting robot. Based on extensive practical observations, the features of the exterior wall surface can be divided into two parts: the main plane and corner points. Most points, after filtering, are concentrated on the main plane, while corner points are concentrated in areas such as windows and balconies, and the vast majority are linear features. Therefore, this solution adopts a two-step point cloud registration method based on feature extraction to register the LiDAR point cloud with the point cloud map. This achieves both localization of the exterior wall painting robot and acquisition of obstacle information such as windows and balconies. This not only improves the accuracy and efficiency of the painting operation but also reduces human intervention and operational risks, especially significantly enhancing safety in high-altitude operations.

[0079] Although specific embodiments of the invention have been described in detail with reference to the accompanying drawings, this should not be construed as limiting the scope of protection of this patent. Various modifications and variations that can be made by a person skilled in the art without inventive effort within the scope described in the claims still fall within the scope of protection of this patent.

Claims

1. An automatic obstacle avoidance method for exterior wall spraying robots based on lidar and BIM, characterized in that, Including the following steps: S1. Establish a BIM model of the construction wall surface, convert the BIM model into a point cloud map, and then convert the point cloud map into a raster map. Import the point cloud map and the raster map into the exterior wall spraying robot. The method for projecting the point cloud map into a raster map in step S1 is as follows: S1.

4. Using the thickness direction of the construction wall as the Y-axis, the length direction of the construction wall as the X-axis, and the height direction of the construction wall as the Z-axis, determine the position of the axis in the thickness direction of the wall. Get the depth of the window Depth of the balcony , and These represent the recessed and protruding positions of the window and balcony on the Y-axis, respectively. S1.5, Set the projection area for windows and balconies. and ,in, , , and All are set projection thresholds; S1.

6. Determine sequentially whether the Y-axis coordinate of each point in the point cloud map falls within the projection range. and If so, then project the X-axis and Z-axis coordinates of the point onto the XZ plane along the Y-axis direction; otherwise, ignore the point. S2. The laser radar on the exterior wall spraying robot acquires real-time laser point clouds of the construction wall surface, extracts the main plane points and corner points of the real-time laser point clouds, and performs real-time registration with the main plane points and corner points of the point cloud map; the registration method between the real-time laser point cloud and the point cloud map includes the following steps: S2.

1. Use the Harris 3D corner detection algorithm to identify principal plane points and corner points in real-time laser point clouds and point cloud maps; wherein, the sets of principal plane points in real-time laser point clouds and point cloud maps are respectively... and The corner point sets of the real-time laser point cloud and the point cloud map are respectively and ; S2.

2. Using the SVD algorithm, perform plane fitting on the principal plane points in the real-time laser point cloud acquired in each frame to obtain the normalized plane equation of the real-time scanning laser principal plane: in, , and All are constants. The normal vector of the laser plane in real-time scanning is ; , and These are the three coordinate axes of the real-time scanning laser master plane. This is the offset of the laser plane in real-time scanning; S2.3 Obtain the normal vector of the principal plane of the point cloud map. The preliminary transformation matrix from the real-time scanning laser principal plane to the point cloud map principal plane is calculated. ; S2.4, through preliminary matrix transformation Will Align the points in the point cloud map to the main plane of the point cloud map to obtain the initial registration plane; S2.5, will The corner points in the map are projected onto the initial registration plane, and the covariance matrix between the point cloud on the initial registration plane and the point cloud map is calculated using the least squares method. For the covariance matrix Perform singular value decomposition to obtain the rotation matrix Translation vector And through rotation matrix Translation vector Synthesize the final transformation matrix ; S2.

6. Based on the initial pose of the exterior wall spraying robot, the final transformation matrix is ​​used... Determine the real-time pose of the exterior wall spraying robot; S3. Generate the operation path of the exterior wall spraying robot to avoid obstacles on the grid map using the Boustrophedon algorithm; S4. The exterior wall spraying robot performs spraying operations along the work path.

2. The automatic obstacle avoidance method for exterior wall spraying robots according to claim 1, characterized in that, The method for converting a BIM model into a point cloud map is as follows: S1.

1. Create a BIM model of the construction exterior wall data in Revit software with the format type rvt. The BIM model includes information on windows and balconies on the construction wall. S1.2 Export the BIM model as an FBX format file, traverse the FBX file to extract the geometric information, and write the geometric information into the point cloud file to obtain the initial point cloud map. S1.

3. The initial point cloud map is downsampled and denoised to obtain an optimized point cloud map.

3. The automatic obstacle avoidance method for an exterior wall spraying robot according to claim 1, characterized in that, In step S2.3, , ; in, , and All are constants. , , and These are the transformation matrices for rotating the real-time scanning laser principal plane around its y-axis, rotating it around the z-axis, and translating it along the x-axis, respectively. , and These are the rotation angles about the y-axis and the rotation angles about the z-axis, respectively. And the offset along the x-axis.

4. The automatic obstacle avoidance method for an exterior wall spraying robot according to claim 1, characterized in that, In step S2.5, the covariance matrix The expression is, ; ; ; in, Real-time laser point cloud corner set Points in , , for The first in One point, for The transformed points, , for The point of transformation It is a constant; for The total number of corner points in the; The set of corner points in a point cloud map The center of mass, For preliminary registration on the plane The center of mass, for A set of.

5. The automatic obstacle avoidance method for an exterior wall spraying robot according to claim 1, characterized in that, In step S2.5, , , , ; in, It is a diagonal matrix, and the elements on the diagonal are singular values, arranged in descending order; Both V and V are obtained through singular value decomposition. The set of corner points in a point cloud map The principal direction orthogonal matrix; To initially register the set of plane corner points The principal direction orthogonal matrix.

6. The automatic obstacle avoidance method for an exterior wall spraying robot according to claim 1, characterized in that, The exterior wall spraying robot uses the Boustrophedon algorithm to plan a "bow"-shaped path to cover the entire grid map and avoid empty areas on the grid map.

7. The automatic obstacle avoidance method for an exterior wall spraying robot according to claim 6, characterized in that, The spraying method used by the exterior wall spraying robot during the spraying process is as follows: S4.1 Divide the construction wall surface into multiple construction units along the work path. Each construction unit has a length and width of [missing information]. and A rectangular area; S4.2 Control the exterior wall spraying robot to spray the construction units sequentially along the work path, and accumulate and calculate the X-axis displacement and Z-axis displacement of the spraying robot in each construction unit; S4.3 When the exterior wall spraying robot sprays each construction unit, it determines whether the exterior wall spraying robot is in the process of spraying. Is the Z-axis displacement generated within the step size range less than If yes, stop spraying; otherwise, continue spraying to complete the current construction unit.

8. The automatic obstacle avoidance method for an exterior wall spraying robot according to claim 1, characterized in that, It also includes the following steps: S2.

7. Use drones to patrol the painting robot at regular intervals. The drones use visual SLAM to locate themselves and then use their own location to solve for the position information of the exterior wall painting robot. The exterior wall painting robot uses the position information solved by the drones to perform pose correction.

Images