Concrete water chiseling robot mechanical arm control method and system

Abstract

The application provides a concrete water flushing and chiseling robot mechanical arm control method and system, relates to the technical field of mechanical arm motion control, and collects original point clouds of a concrete surface to be operated, converts and registers the original point clouds to a mechanical arm base coordinate system, and then constructs a global operation surface point cloud model; divides local grid units according to a single effective chiseling width, solves surface normal vectors, matches reference process parameters, generates a grid level process parameter mapping table, plans an operation path, executes S curve acceleration and deceleration planning, and generates an end speed sequence; in operation, a position and attitude double closed loop architecture is adopted to correct the end position and posture in real time, target distance deviation and incidence angle deviation are controlled within a preset threshold, a speed-jet parameter coupling model is used to synchronously adjust jet parameters, and after global operation is completed, secondary scanning is performed to account for chiseling quality, under-chiseling areas are automatically identified, and supplementary chiseling is performed. The application improves chiseling uniformity and automation level through grid parameter adaptation and double closed loop regulation.

Description

Technical Field

[0001] This invention relates to the field of robotic arm motion control technology, specifically to a control method and system for a concrete water jetting and roughening robot robotic arm. Background Technology

[0002] With the increasing demand for intelligent and refined construction in infrastructure projects, the use of robotic arms equipped with high-pressure water jets to roughen concrete surfaces has gradually replaced traditional manual roughening and mechanical milling methods, effectively improving the working environment and reducing labor intensity. Currently, conventional concrete water jet roughening robots mostly employ fixed parameter teaching and reproduction or globally unified path planning strategies, which can only achieve simple traversal and coverage of the work surface. When facing complex conditions with undulating concrete surface topography and significant differences in local flatness, existing control systems still have obvious technical deficiencies, making it difficult to meet the high-standard construction requirements for roughening uniformity, roughening accuracy, and interface bonding stability.

[0003] The shortcomings of existing technologies are specifically reflected in the following aspects: First, traditional chiseling control methods generally adopt water jet process parameters for full-area solidification, which cannot adapt to the local elevation fluctuations, surface normal differences and different concrete grade characteristics of the concrete working surface. They lack regional adaptive control capabilities, resulting in uneven chiseling effects in concave and convex areas. They are prone to local under-chiseling and over-chiseling, which greatly reduces the uniformity of chiseling depth and effective chiseling coverage, and the adaptability of chiseling construction is poor. Secondly, the existing chiseling operation system suffers from drawbacks such as disconnect between electromechanical control and lack of closed-loop correction for operational accuracy. On the one hand, the motion control of the robotic arm and the high-pressure water jet operation are independent of each other. The speed fluctuations of the robotic arm during start-up, stop, reversal, acceleration and deceleration cannot be dynamically matched with the jet parameters, which easily leads to an imbalance in jet impact energy and exacerbates the dispersion of chiseling quality. On the other hand, the pure open-loop trajectory control mode is adopted without real-time detection and closed-loop correction mechanisms for target distance and jet injection angle. This makes it impossible to compensate for concrete morphology deviations and robotic arm motion errors. During the operation, the nozzle position is prone to deviation and the jet operation posture is unstable, ultimately making it difficult to guarantee the overall chiseling forming accuracy and failing to meet the requirements of refined construction.

[0004] Therefore, it is necessary to provide a control method and system for the robotic arm of a concrete water jetting and scabbing robot to solve the above problems.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a control method and system for a concrete water jetting and roughening robot arm to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A method for controlling the robotic arm of a concrete water jetting and roughening robot, comprising the following steps: Step 1: Collect the original point cloud data of the concrete surface to be worked on, register and transform it to the robot arm base coordinate system, and construct a point cloud model of the entire working surface; Step 2: Divide the point cloud model of the entire operation surface into several local grid units according to the effective chiseling width of a single high-pressure water jet. Use principal component analysis to solve the surface normal vector of the center point of each grid unit. Combine the preset target chiseling depth and the grade of concrete to be operated with the reference process parameters of each grid unit to generate a grid-level process parameter mapping table. Step 3: Based on the grid-level process parameter mapping table, grid layout and width parameters, plan the robotic arm operation path, and perform S-curve acceleration and deceleration planning on the operation path, divide it into multiple nodes, and generate the end velocity sequence corresponding to each node; Step 4: The robotic arm performs chiseling operations using the work path and end-effector velocity sequence as the nominal trajectory. Using the current grid's reference target distance and surface normal vector as control targets, the actual target distance is collected in real time to calculate the target distance deviation value. At the same time, the incident angle deviation value between the jet axis and the concrete surface normal vector is calculated. Through a dual closed-loop control architecture of position and attitude, the position and attitude of the robotic arm's end-effector are corrected in real time to maintain the target distance deviation value and incident angle deviation value within the preset threshold. Step 5: Based on the real-time movement speed of the robotic arm, call the preset speed-jet parameter coupling model, adjust the jet parameters synchronously, perform a second scan after the full-area chiseling operation is completed, obtain the point cloud data after the operation, compare the point cloud data before and after the operation, calculate the chiseling quality, and if there are under-chiseled areas, control the robotic arm to perform supplementary chiseling operations until the chiseling quality meets the standard.

[0008] Furthermore, the method used to construct the global operation surface point cloud model is as follows: Using a 3D laser scanning unit mounted at the end of a robotic arm, the robotic arm drives the scanning unit to traverse the area to be worked along a preset scanning path to collect the original point cloud data of the concrete surface to be worked. based on The statistical filtering algorithm of the criterion is used to denoise the original point cloud data, and the hand-eye calibration method is used to solve the homogeneous transformation matrix between the coordinate system of the three-dimensional laser scanning unit and the coordinate system of the end effector of the robot arm. Combined with the end effector pose matrix obtained by solving the forward kinematics of the robot arm, the denoised point cloud data is mapped from the coordinate system of the scanning unit to the base coordinate system of the robot arm through matrix operations, and finally a global working surface point cloud model reflecting the shape of the working surface is constructed.

[0009] Furthermore, based on the effective chiseling width of a single high-pressure water jet, the point cloud model of the entire working surface is divided into several local grid cells. Principal component analysis is used to solve for the surface normal vector of the center point of each grid cell. The method used is as follows: The point cloud model of the entire working surface is divided into grids along the tangent plane of the working surface. The width of a single grid is set to be consistent with the effective chiseling width of a single high-pressure water jet. Several local grid units with uniform specifications, no overlap and full coverage are obtained. Each grid unit corresponds to a section of the area to be chiseled. For each local grid cell, extract all point cloud data within it, and determine the average coordinates of all point cloud data within the grid cell as the coordinates of the grid cell's center point. A covariance matrix is ​​constructed for the point cloud data within each local grid cell. The covariance matrix is ​​then decomposed into eigenvalues ​​using principal component analysis to obtain the eigenvector corresponding to the smallest eigenvalue of the covariance matrix. This eigenvector is then used as the surface normal vector of the center point of the corresponding grid cell.

[0010] Furthermore, by combining the target roughening depth with the grade of concrete to be processed, the baseline process parameters of each grid cell are matched to generate a grid-level process parameter mapping table. The method used is as follows: The baseline process parameters include baseline target distance, baseline moving speed, baseline jet water pressure, and baseline flow rate; Based on the jet impact dynamics model, a quantitative mapping relationship is established between the chiseling depth and the target distance, the robotic arm moving speed, the jet water pressure, and the jet flow rate. Combining the preset target chiseling depth and the grade of concrete to be processed, the corresponding benchmark target distance, benchmark moving speed, benchmark jet water pressure, and benchmark flow rate are matched for each local grid cell. The benchmark process parameters of each local grid cell and its corresponding benchmark process parameters are sorted out, and a grid-level process parameter mapping table with one-to-one correspondence between grid cells and process parameters is generated.

[0011] Furthermore, based on the grid-level process parameter mapping table, grid layout, and width parameters, the robotic arm's operating path is planned using the following method: The operation path type is determined to be a parallel strip reciprocating path, the strip extension direction is consistent with the row arrangement direction of the grid cells, and the center line spacing between adjacent strips is set to be equal to the difference between the effective chiseling width of a single high-pressure water jet and the preset overlap allowance. The preset overlap allowance is the overlap width of adjacent strips along the width direction. Within each strip, discrete path points are arranged sequentially along the strip extension direction. Each discrete path point corresponds one-to-one with the center point of the corresponding grid cell, and each discrete path point carries the reference moving speed parameter of the corresponding grid cell. The reversing transition section between adjacent strips is set outside the boundary of the work area, and the reversing transition section adopts an arc transition path.

[0012] Furthermore, S-curve acceleration / deceleration planning is performed on the work path, dividing it into multiple nodes and generating the end-velocity sequence for each node. The method used is as follows: Constraints are set for the rate of change of acceleration, maximum tangential acceleration, and limit speed. A seven-segment S-curve algorithm with continuously changing acceleration is used to perform seven-segment nodal velocity interpolation on the straight work strip. The uniform speed segment matches the grid reference speed. The circular arc reversal segment is speed-limited according to the normal acceleration constraint and smoothly transitions with the speed of the straight segment. Discrete motion sequence containing position, velocity, and acceleration is obtained by discrete interpolation of the entire path in Cartesian space according to a fixed control period.

[0013] Furthermore, the position and attitude dual closed-loop control architecture is used to correct the end effector pose of the robotic arm in real time. The method used is as follows: The robotic arm performs the chiseling operation with the working path and end velocity sequence as the nominal trajectory, and uses the reference target distance and surface normal vector of the grid cell corresponding to the current working position as the control target; Three laser rangefinders, evenly distributed around the end of the robotic arm, collect multi-point range data from the nozzle to the concrete working surface in real time. Based on this range data, the actual target distance is calculated. The difference between the actual target distance and the reference target distance of the corresponding grid cell is calculated to obtain the target distance deviation value. At the same time, the incident angle deviation value between the jet axis and the normal vector of the concrete surface is calculated. A dual closed-loop control architecture for position and attitude is constructed. The target distance deviation value is input into the position control loop, and a fuzzy adaptive PID algorithm is used for calculation to output the displacement correction value along the normal vector of the working surface. The incident angle deviation value is input into the attitude control loop, and a quaternion spherical linear interpolation algorithm is used to output the attitude correction value. The displacement and attitude correction values ​​output by the position control loop and attitude control loop are superimposed on the nominal trajectory of the robotic arm to correct the position and pose of the robotic arm end in real time, and maintain the target distance deviation and incident angle deviation values ​​within the corresponding preset thresholds.

[0014] Furthermore, based on the real-time movement speed of the robotic arm, a preset velocity-jet parameter coupling model is invoked to synchronously adjust the jet parameters. The method used is as follows: The jet parameters include jet water pressure and jet flow rate; A velocity-pressure-flow coupling model based on the principle of constant jet energy per unit area is preset. During the roughening operation of the robotic arm, the actual movement speed of the robotic arm end is acquired in real time and input into the coupling model. The high-pressure water jet pressure and flow rate are synchronously matched through the coupling model. Specifically, for the uniform speed section of the operation path, the reference jet water pressure and reference jet flow rate are kept constant. For the acceleration / deceleration section and the reversal transition section, the jet water pressure and jet flow rate are linearly and synchronously adjusted according to the rate of change of the robotic arm speed, so that the jet impact energy received per unit operation area remains consistent throughout the entire operation.

[0015] Furthermore, the point cloud data before and after the operation were compared to calculate the chiseling quality. The method used was as follows: After the full-area roughening operation is completed, the robotic arm is controlled to drive the three-dimensional laser scanning unit mounted on its end to perform a full scan of the concrete working surface along a preset secondary scanning path to obtain the post-operation point cloud model. The post-operation point cloud model is registered and aligned with the constructed global operation surface point cloud model. The global chiseling depth distribution data is obtained through point cloud difference calculation. The average chiseling depth, chiseling depth uniformity coefficient, and effective chiseling coverage rate are used as chiseling quality evaluation indicators. The global average chiseling depth, the standard deviation of chiseling depth in each region, and the proportion of effective chiseling area are statistically calculated to obtain the chiseling quality. Based on the chiseling depth distribution data, under-chiseled areas with chiseling depths below the preset target chiseling depth threshold are identified. The location, range, and area of ​​the under-chiseled areas are determined. According to the location, range, and grid division of the under-chiseled areas, an appropriate supplementary chiseling operation path is automatically generated. The robotic arm is controlled to drive the jet nozzle and perform a second supplementary chiseling operation on the under-chiseled areas according to the supplementary chiseling operation path and the optimized baseline process parameters. After the supplementary chiseling is completed, point cloud scanning and quality calculation are performed again until the average chiseling depth, chiseling depth uniformity coefficient, and effective chiseling coverage all meet the preset chiseling quality threshold requirements, and the chiseling quality is judged to be up to standard.

[0016] The present invention also provides a control system for a concrete water jetting and scabbing robot arm, the system being used to execute the concrete water jetting and scabbing robot arm control method described in any one of the above claims, comprising: The data acquisition and registration module is used to acquire the original point cloud data of the concrete surface to be worked on, register and convert it to the robot arm base coordinate system, and build a point cloud model of the entire working surface. The parameter mapping module is used to divide the point cloud model of the entire operation surface into several local grid units according to the effective chiseling width of a single high-pressure water jet. The principal component analysis method is used to solve the surface normal vector of the center point of each grid unit. Combined with the preset target chiseling depth and the grade of concrete to be operated, the reference process parameters of each grid unit are matched to generate a grid-level process parameter mapping table. The speed planning module is used to plan the robot arm's working path based on the grid-level process parameter mapping table, grid layout and width parameters, and to perform S-curve acceleration and deceleration planning on the working path, dividing it into multiple nodes and generating the end velocity sequence corresponding to each node. The pose control module is used for the robotic arm to perform chiseling operations with the working path and end velocity sequence as the nominal trajectory. It uses the current grid's reference target distance and surface normal vector as the control targets, collects the actual target distance in real time to calculate the target distance deviation value, and solves the incident angle deviation value between the jet axis and the concrete surface normal vector. Through the position and attitude dual closed-loop control architecture, it corrects the robotic arm's end pose in real time and maintains the target distance deviation value and incident angle deviation value within the preset threshold. The collaborative quality closed-loop module is used to call the preset speed-jet parameter coupling model according to the real-time movement speed of the robotic arm, synchronously adjust the jet parameters, perform a second scan after the whole-domain chiseling operation is completed, obtain the point cloud data after the operation, compare the point cloud data before and after the operation, calculate the chiseling quality, and if there are under-chiseled areas, control the robotic arm to perform supplementary chiseling operations until the chiseling quality meets the standard.

[0017] Compared with the prior art, the beneficial effects of the present invention are: This invention utilizes point cloud modeling and mesh cell division, combined with the surface normal vectors, concrete grade, and target chiseling depth of each mesh region, to generate a mesh-level process parameter mapping table. This enables adaptive control of the chiseling process in local areas, effectively solving the problems of under-chiseling and over-chiseling in uneven terrain, and significantly improving the uniformity of chiseling depth and effective chiseling coverage across the entire area. Secondly, this invention implements S-curve acceleration and deceleration smoothing planning for the work path, generating a node-based end-effector velocity sequence, and constructing a velocity-jet parameter coupling model. This achieves dynamic coordination and matching between the robotic arm's movement speed and the high-pressure water jet parameters, eliminating the jet energy imbalance caused by speed fluctuations during start-up, shutdown, and reversal phases, and significantly reducing the dispersion of chiseling quality. Simultaneously, this invention establishes a dual closed-loop control architecture for position and attitude. Based on real-time detected target distance deviation and incident angle deviation values, it dynamically corrects the robotic arm's end-effector pose, accurately compensating for errors in the work surface shape and robotic arm movement, ensuring that the water jet always operates with the optimal target distance and best incident angle, greatly improving the stability and forming accuracy of the chiseling operation. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 This is a schematic diagram of the system module flow of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0020] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0021] Example: Please see Figure 1 A method for controlling the robotic arm of a concrete water jetting and roughening robot, the specific steps of which include: Step 1: Collect the original point cloud data of the concrete surface to be worked on, register and transform it to the robot arm base coordinate system, and construct a point cloud model of the entire working surface.

[0022] In this embodiment, the steps of collecting the original point cloud data of the concrete surface to be worked on, registering and converting it to the robot arm's base coordinate system, and constructing a point cloud model of the entire working surface are implemented as follows: The 3D laser scanning unit is fixedly mounted on the flange at the end of the robotic arm using an "eye-on-hand" installation method, maintaining a fixed relative position with the high-pressure water jet nozzle. Before operation, a grid-style scanning path is planned: the outer rectangle of the area to be worked on is used as the scanning range, and reciprocating scanning rows are set along the long side of the rectangle. The spacing between adjacent scanning rows is set to 85% of the effective scanning width of a single pass of the 3D laser scanning unit, ensuring that there is an overlapping area between adjacent row point clouds and avoiding missed scans at the edge of the working surface. The robotic arm is controlled to move the scanning unit along the preset scanning path at a constant speed of 0.3-0.5 m / s, traversing the entire area to be worked on. Simultaneously, the 3D laser scanning unit is activated to collect data, with the scanning sampling frequency matched to the robotic arm's moving speed. The spacing between adjacent sampling points in the point cloud is controlled to be no more than 0.5 mm to ensure the accuracy of the concrete surface morphology restoration. After the data acquisition is completed, the point cloud is filtered and denoised, and then registered to the robotic arm's base coordinate system through coordinate transformation. A point cloud model of the entire working surface is then generated through fitting.

[0023] The collected raw point cloud data is denoised using a statistical filtering algorithm based on the three-times-standard-deviation criterion. The average distance between each data point and its nearest neighbor is calculated. The effective data interval is defined by the mean and standard deviation of the overall distance distribution. Outliers outside the interval are identified as noise points and removed. The effective point cloud data that reflects the true shape of the working surface is retained, and data interference caused by environmental dust, surface reflection and equipment vibration is eliminated. After noise reduction, the hand-eye calibration method is used to complete the registration transformation from the scanning unit coordinate system to the robot arm base coordinate system. The specific implementation process is as follows: First, hand-eye calibration is completed in advance. A standard checkerboard calibration board is used as the calibration reference. The robotic arm is controlled to drive the three-dimensional laser scanning unit to collect feature points of the calibration board under nine different end-effector poses. The homogeneous transformation matrix of the coordinate system of the three-dimensional laser scanning unit relative to the coordinate system of the robotic arm end-effector is obtained by solving the least squares method. This matrix contains a three-dimensional translation vector and a three-dimensional rotation matrix, which are used to characterize the fixed relative pose relationship between the scanning unit and the end flange of the robotic arm. Secondly, during the point cloud acquisition process, the real-time angle values ​​of each joint of the robotic arm are read synchronously, and the homogeneous transformation matrix of the end-effector coordinate system of the robotic arm relative to the base coordinate system of the robotic arm is obtained by solving the forward kinematic equation of the robotic arm at the current moment. Finally, the point cloud coordinate mapping is completed through homogeneous coordinate transformation: For any point in the denoised point cloud, its coordinates in the scanning unit coordinate system are transformed by two levels of homogeneous transformation matrix, first mapped to the coordinate system of the robotic arm end effector, and then further mapped to the base coordinate system of the robotic arm, finally obtaining the corresponding coordinates of the point in the base coordinate system. The above coordinate transformation is performed point by point on all the denoised valid point clouds to obtain a set of three-dimensional point clouds that are all unified in the robot arm base coordinate system. The point cloud set is then processed by voxel mesh downsampling, with the voxel side length set to 2mm. The data volume is reduced while preserving the shape features of the working surface, and finally a global working surface point cloud model that reflects the true shape of the working surface is obtained.

[0024] In this embodiment, step 1 serves as the foundation and core spatial reference of the full-process control method of this invention. By uniformly mapping point cloud data to the robotic arm's base coordinate system, a unified spatial reference for the entire process is constructed. High-fidelity concrete working surface morphology data is obtained based on standardized scanning strategies and point cloud denoising processing. Then, hand-eye calibration is used to establish the coordinate mapping association between the visual perception system and the robotic arm's actuator. Simultaneously, the initial original morphology reference is retained for later quality comparison. The above work provides a precise and consistent spatial reference system and original data basis for subsequent mesh-differentiated process parameter matching, motion trajectory planning, end-effector pose dual-loop real-time correction, and chiseling quality closed-loop re-inspection, avoiding coordinate accumulation errors and spatial pose misalignment caused by multiple coordinate system conversions from the source. Step 2: According to the effective chiseling width of a single high-pressure water jet, the point cloud model of the entire working surface is divided into several local grid units. Principal component analysis is used to solve the surface normal vector of the center point of each grid unit. Combined with the preset target chiseling depth and the grade of concrete to be worked, the reference process parameters of each grid unit are matched to generate a grid-level process parameter mapping table.

[0025] In this embodiment, the steps of dividing the local mesh, solving the surface normal vector, and generating the mesh-level process parameter mapping table are implemented as follows: Based on the point cloud model of the entire working surface obtained in step 1, the overall principal tangent plane of the working surface is first obtained through plane fitting. The working surface is then meshed along the two orthogonal directions of this principal tangent plane. The side length of a single mesh is set to be equal to the effective chiseling width of a single high-pressure water jet, where the effective chiseling width refers to the lateral width of the effective working area where the jet chiseling depth meets the target requirements under the benchmark working condition. After meshing, several local mesh units with uniform specifications, no overlap, and full coverage are obtained. Each mesh unit corresponds to the processing area of ​​one jet operation, ensuring that the subsequent operation path corresponds one-to-one with the mesh layout. For each local mesh unit, all the contents contained within it are extracted. For point cloud data, the arithmetic mean of the three-dimensional coordinates of all points is taken as the spatial coordinates of the center point of the grid cell. Based on this, a three-dimensional covariance matrix is ​​constructed for the point cloud data in each grid cell. The covariance matrix is ​​decomposed into eigenvalues ​​using principal component analysis to obtain three eigenvalues ​​and their corresponding eigenvectors. The two larger eigenvalues ​​correspond to the principal distribution direction of the point cloud in the local tangent plane, and the smallest eigenvalue corresponds to the normal distribution direction of the point cloud. Therefore, the eigenvector corresponding to the smallest eigenvalue is determined as the surface normal vector of the center point of the grid cell, and the direction of the normal vector is uniformly adjusted to point outward from the working surface to ensure the consistency of the reference for subsequent angle calculations. After completing the mesh generation and normal vector solution, the baseline process parameters are matched grid by grid and a mesh-level process parameter mapping table is generated. The baseline process parameters include four items: baseline target distance, baseline moving speed, baseline jet water pressure, and baseline jet flow rate. Based on the principle of jet impact dynamics and combined with the erosion damage strength threshold of different grades of concrete, a quantitative mapping relationship between the chiseling depth and the target distance, robotic arm moving speed, jet water pressure, and jet flow rate is established through calibration tests. The higher the grade of concrete, the stronger its erosion resistance, and the higher the jet energy required to achieve the same chiseling depth. During parameter matching, the preset target roughening depth and the grade of concrete to be worked on are used as inputs. The surface normal vector of the center point of each grid cell is used as the reference direction of the jet injection. First, the reference target distance of the corresponding grid is determined. Then, the corresponding reference moving speed, reference jet water pressure and reference jet flow rate are obtained by matching through the pre-established quantitative mapping relationship. Finally, the center point coordinates, surface normal vector and the corresponding four reference process parameters of each grid are stored one by one using the grid cell number as an index, generating a complete grid-level process parameter mapping table, which provides a parameter query basis for subsequent operation path planning and real-time control.

[0026] In this embodiment, step 2 is the core connecting link for achieving localized precise roughening in this invention. By dividing the local grid units according to the effective roughening width of a single jet, the process control unit and the jet operation unit are precisely matched. The grid surface normal vector obtained by principal component analysis is used to obtain the precise orientation reference of the local morphology. Then, based on the jet impact dynamics mapping relationship, the reference process parameters that are adapted to the surface morphology and concrete grade are matched grid by grid to form a grid-level parameter mapping table. This not only breaks through the limitation of traditional globally unified process parameters that cannot adapt to the local undulations and material differences of the working surface, but also ensures the uniformity of the roughening depth of the entire area from the planning level. It also provides a standardized parameter index and process control target for subsequent operation path planning and real-time closed-loop control of end pose, realizing the precise implementation of morphology perception data into process control commands.

[0027] Step 3: Based on the grid-level process parameter mapping table, grid layout and width parameters, plan the robotic arm operation path, and perform S-curve acceleration and deceleration planning on the operation path, dividing it into multiple nodes and generating the end velocity sequence corresponding to each node.

[0028] In this embodiment, the step of planning the robotic arm's working path and generating the end-effector velocity sequence is implemented as follows: Based on the grid-level process parameter mapping table generated in step 2, a parallel strip reciprocating operation path is planned with the row arrangement direction of the grid cells as the reference. This path form is suitable for large-area surface coverage roughening operations, which can minimize idle travel and improve operation efficiency. The centerline spacing between adjacent strips is set as the difference between the effective roughening width of a single high-pressure water jet and a preset overlap allowance. The preset overlap allowance is the overlap width of adjacent strips along the width direction, and is taken as 5%-10% of the effective roughening width of a single jet. It is used to compensate for the positioning error of the robotic arm path and avoid gaps at the junction of adjacent strips. For example, when the effective roughening width of a single jet is 50mm, the overlap allowance is 3mm, corresponding to a centerline spacing of 47mm. Discrete path points are sequentially set along the extension direction of each working strip. The center point of each grid cell in the corresponding row is aligned with the center point of each discrete path point, and each discrete path point is directly bound to the reference moving speed of the corresponding grid cell in step 2, realizing a one-to-one mapping between spatial position and process parameters. This ensures that when the robotic arm moves to the corresponding point, the matching reference process parameters can be directly called. The reversal transition section between adjacent strips is set outside the boundary of the working area, and the distance from the working boundary is not less than the minimum stroke required for the robotic arm to complete deceleration-reversal-acceleration, ensuring that the acceleration and deceleration process does not fall within the effective working area. The reversal transition section adopts an arc transition path, and the radius of the arc is determined according to the maximum movement speed of the robotic arm and the allowable normal acceleration threshold, with a value not less than 0.2m. It also satisfies that the normal acceleration during the reversal process is not greater than 80% of the maximum tangential acceleration, avoiding speed changes and mechanical shocks caused by right-angle reversal. After completing the operation path planning, the S-curve acceleration and deceleration planning is executed and the end velocity sequence is generated: First, the motion constraint boundary is set. Combining the rated load of the robotic arm and the operation stability requirements, the upper limit of the acceleration change rate is set to 500mm / s³, the maximum tangential acceleration is 200mm / s², and the maximum moving speed does not exceed the maximum value of the moving speed of all grid references, to ensure that the planning result does not exceed the hardware capability and process allowable range of the robotic arm. A seven-segment S-curve algorithm with continuously varying acceleration is adopted. Velocity interpolation is performed on each straight work strip, which is divided into seven nodes in sequence: acceleration segment, uniform acceleration segment, deceleration segment, uniform speed segment, acceleration-deceleration segment, uniform deceleration segment, and deceleration-deceleration segment. The speed of the uniform speed segment matches the reference movement speed of the corresponding grid. The acceleration of the entire segment is smoothly transitioned without abrupt changes, which can effectively suppress the impact of the robotic arm joints and the jitter of the end jet. Speed ​​adaptation is performed separately for the arc reversal transition segment: the maximum allowable running speed of the reversal segment is calculated based on the arc radius and the preset upper limit of normal acceleration. The speed of the reversal segment is smoothly connected with the end speed of the straight segments before and after, so as to achieve continuous speed without abrupt changes in the entire work path. Finally, using the standard control cycle of 10ms for the robotic arm controller as the unified time step, the entire path is discretized in the Cartesian space at the end of the robotic arm to generate the end position coordinates, velocity vector and acceleration vector at each moment, forming a complete discrete motion sequence. This sequence is directly used as the nominal control quantity of the robotic arm motion and input to the joint controller. After being converted into angle and angular velocity commands for each joint through inverse kinematics solution, it drives the servo motor to execute the corresponding motion.

[0029] In this embodiment, step 3 serves as a core intermediate link connecting process parameter planning and robotic arm motion execution. By generating parallel strip reciprocating operation paths based on grid layout, discrete path points are bound one by one to the corresponding grid reference process parameters, achieving precise mapping between spatial operation positions and process parameters. Combined with overlap allowance design and the arc reversal scheme on the outer side of the operation area, this effectively avoids missed chiseling gaps at the joints of adjacent strips, ensuring the integrity of the entire operation coverage, and moves the acceleration / deceleration reversal process out of the effective operation area, reducing the interference of speed fluctuations on chiseling quality. At the same time, a seven-segment S-curve acceleration / deceleration algorithm is adopted and adapted to the normal acceleration constraint of the reversal segment to generate a discrete motion sequence with continuous and smooth acceleration. This can effectively suppress joint impact of the robotic arm and end-effector jet jitter, providing a stable nominal trajectory reference for subsequent dual-closed-loop position and posture control, and providing a continuously controllable speed input for the coordinated adjustment of motion and jet parameters. From the motion planning level, this ensures the operating efficiency, motion stability, and quality consistency of the chiseling operation.

[0030] Step 4: The robotic arm performs chiseling operations using the work path and end-effector velocity sequence as the nominal trajectory. Using the current grid's reference target distance and surface normal vector as control targets, the actual target distance is collected in real time to calculate the target distance deviation value. At the same time, the incident angle deviation value between the jet axis and the concrete surface normal vector is calculated. Through a dual closed-loop control architecture of position and attitude, the position and attitude of the robotic arm's end-effector are corrected in real time to maintain the target distance deviation value and incident angle deviation value within the preset threshold.

[0031] In this embodiment, the step of real-time correction of the robotic arm's end-effector pose using a dual closed-loop control architecture of position and attitude is implemented as follows: The robotic arm uses the work path and end-effector velocity sequence generated in step 3 as the nominal trajectory and performs motion control cycle by cycle according to a fixed control cycle. In this embodiment, the control cycle is set to 10ms, which is consistent with the discrete time step of the end-effector velocity sequence. During the operation, the corresponding local grid cell is matched according to the real-time position of the robotic arm end, and the reference target distance and reference surface normal vector corresponding to the grid are called from the grid-level process parameter mapping table as the target value of the closed-loop control in this cycle. Three laser rangefinders are evenly arranged at 120° intervals around the mounting flange of the high-pressure water jet nozzle. The measuring optical axes of the three sensors are parallel to the central axis of the jet, and the measuring end faces of all sensors are in the same plane perpendicular to the jet axis. The sensor sampling cycle is synchronized with the robotic arm control cycle. In each control cycle, the distance values ​​of the three laser rangefinders are collected synchronously. Combined with the fixed installation coordinates of the three sensors, the local concrete surface plane of the area currently facing the nozzle is obtained. The vertical distance from the center point of the nozzle to this local plane is calculated, which is the current actual target distance. At the same time, the normal vector corresponding to this local plane is solved as the actual surface normal vector of the current working position. The spatial angle between this and the reference surface normal vector of the corresponding grid is calculated to obtain the incident angle deviation value between the jet axis and the reference normal. The difference between the actual target distance and the reference target distance is used to obtain the target distance deviation value. The angle between the actual normal vector and the reference normal vector is used as the incident angle deviation value, which is used as the input of the dual closed-loop control.

[0032] Furthermore, this embodiment adopts a dual-closed-loop control architecture with parallel operations of the position loop and attitude loop. The two control loops complete the calculations synchronously and jointly output correction values ​​that are superimposed on the nominal trajectory. The specific control logic is as follows: The position control loop is set up as follows: Using the target distance deviation as input, a fuzzy adaptive PID control algorithm is employed. Initial PID control parameters are pre-defined based on the response characteristics of the chiseling operation. The absolute value of the current target distance deviation and its rate of change are then used as two fuzzy inputs, divided into three levels of fuzzy subsets: large, medium, and small. The proportional, integral, and derivative control parameters are adjusted online in real-time according to preset fuzzy control rules: when the target distance deviation is large, the proportional coefficient is increased to accelerate the response and quickly reduce the deviation; when the target distance deviation is small, the proportional coefficient is decreased and the integral action is enhanced to eliminate steady-state error. Simultaneously, the derivative stage suppresses deviation fluctuations and avoids overshoot. The algorithm ultimately outputs a displacement correction along the direction of the current actual surface normal vector. The correction direction aims to reduce the target distance deviation; if the actual target distance is greater than the reference value, the correction moves closer to the working surface, and vice versa.

[0033] The attitude control loop is set up as follows: Using the incident angle deviation as input, a quaternion spherical linear interpolation algorithm is employed to output the attitude correction. First, the attitude deviation between the reference surface normal vector and the actual surface normal vector is converted into a unit quaternion. Starting from the unit quaternion corresponding to the nominal attitude of the current cycle, and ending with the target attitude quaternion after eliminating the incident angle deviation, a spherical linear interpolation algorithm is used to calculate the smooth transition intermediate attitude within a single control cycle, outputting the attitude correction quaternion for the current cycle. This method ensures continuous angular velocity without abrupt changes during attitude adjustment, avoiding the gimbal lock-up problem that easily occurs in traditional Euler angle adjustment. Simultaneously, it ensures the jet axis smoothly aligns with the reference normal vector direction, suppressing end-jet jitter caused by attitude adjustment.

[0034] Finally, the end-effector pose correction is performed. The displacement correction amount output by the position control loop is superimposed along the current actual surface normal vector direction onto the nominal end-effector position of this cycle to obtain the corrected end-effector target position. The attitude correction amount output by the attitude control loop is superimposed onto the nominal end-effector attitude of this cycle to obtain the corrected end-effector target attitude. The corrected complete end-effector pose command is input into the robot arm inverse kinematics solving module, converted into angle and angular velocity control commands for each joint, and sent to the joint servo driver for execution, completing the single-cycle pose closed-loop correction.

[0035] It should be noted that in this embodiment, the preset target distance deviation threshold is ±2mm and the incident angle deviation threshold is ±2°. Through continuous dynamic control of the dual closed loop, the two deviation values ​​can be stably maintained within the preset threshold range, effectively offsetting the positional disturbances caused by the robotic arm positioning error, surface morphology undulations, and mechanical vibration, and ensuring the stability and controllability of the jet target distance and incident angle.

[0036] In this embodiment, step 4 is the core dynamic control link that ensures the chiseling accuracy and quality stability during the operation execution phase of this invention. Three laser rangefinders arranged circumferentially collect multi-point ranging data in real time, calculate the actual target distance and incident angle deviation, and construct a dual-closed-loop control architecture that combines position and attitude control using the corresponding grid's baseline process parameters as the control target. The position control loop uses a fuzzy adaptive PID algorithm to output a normal displacement correction, balancing deviation response speed and steady-state control accuracy. The attitude control loop uses a quaternion spherical linear interpolation algorithm to output a smooth attitude correction, avoiding sudden changes in angular velocity and end-effector jitter during attitude adjustment. Finally, the dual-loop corrections are superimposed on the nominal trajectory in real time to correct the end-effector posture. This effectively offsets the posture disturbances caused by robotic arm positioning errors, surface morphology fluctuations, and mechanical vibrations, maintaining the target distance deviation and incident angle deviation within preset thresholds. This ensures the continuous stability of the jet impact posture and energy, guaranteeing the uniformity of chiseling depth and the reliability of operational accuracy across the entire area from a dynamic execution perspective.

[0037] Step 5: Based on the real-time movement speed of the robotic arm, call the preset speed-jet parameter coupling model, adjust the jet parameters synchronously, perform a second scan after the full-area chiseling operation is completed, obtain the point cloud data after the operation, compare the point cloud data before and after the operation, calculate the chiseling quality, and if there are under-chiseled areas, control the robotic arm to perform supplementary chiseling operations until the chiseling quality meets the standard.

[0038] In this embodiment, the steps of synchronously adjusting the jet parameters, performing secondary scanning to calculate the roughening quality, and executing the additional roughening operation are implemented as follows: Furthermore, based on the principle that the jet erosion energy per unit area is constant, a velocity-jet parameter coupling model is constructed. The core logic is: the jet erosion energy received per unit working area is directly proportional to the jet water pressure and jet flow rate, and inversely proportional to the moving speed of the robotic arm. When the moving speed decreases, the time the unit area is affected by the jet increases, and the total energy received increases. The water pressure and flow rate need to be reduced simultaneously to maintain constant energy. Conversely, the water pressure and flow rate need to be increased simultaneously to ensure consistent chiseling depth under different movement speeds. The model is specifically established through calibration experiments: For commonly used concrete grades, multiple sets of chiseling control tests are conducted under standard test conditions. First, a set of benchmark parameters is determined, namely benchmark target distance, benchmark moving speed, benchmark jet water pressure, and benchmark jet flow rate. This verifies that the chiseling depth under these parameters meets the preset target value. Then, with other conditions unchanged, multiple sets of different moving speeds are set, and the water pressure and flow rate are adjusted accordingly. The correction coefficient is derived by back-calculating the measured chiseling depth to eliminate the influence of nonlinear factors such as pipeline response delay and jet energy loss. Finally, a one-to-one linear mapping relationship between moving speed and jet water pressure and jet flow rate is fitted, solidified into a velocity-jet parameter coupled model, and stored in the control system. Different concrete grades correspond to independent coupled model parameter sets, which are called and matched according to actual working conditions during operation.

[0039] Furthermore, the jet parameters are adjusted synchronously in real time, specifically as follows: During the roughening operation of the robotic arm, the actual movement speed of the robotic arm's end effector is read in real time during each control cycle and input into a preset coupled model to complete parameter calculation. When the robotic arm is in a straight, uniform speed operation segment, the actual speed is consistent with the reference movement speed of the corresponding grid. The model outputs a reference jet water pressure and a reference jet flow rate, controlling the high-pressure water pump and proportional regulating valve to maintain parameter stability. When the robotic arm is in an acceleration / deceleration segment or a circular arc reversal transition segment, the actual speed deviates from the reference value. The model synchronously and linearly adjusts the output water pressure and flow rate commands according to the speed change ratio and a preset correction coefficient. If the speed increases, the water pressure and flow rate are increased proportionally; if the speed decreases, the water pressure and flow rate are decreased proportionally, ensuring that the jet erosion energy received per unit working area remains constant. At the same time, a lower limit threshold for jet parameters is set. When the speed in the reversal segment is too low, the water pressure and flow rate do not fall below the minimum working threshold to avoid jet morphology collapse and failure to operate normally.

[0040] Furthermore, point clouds were compared before and after the operation, and the roughening quality was calculated. The specific method is as follows: After the full-area chiseling operation is completed, the robotic arm drives the 3D laser scanning unit mounted on its end effector to perform a second scan of the work surface using the same path, speed, and sampling parameters as the initial topography scan, acquiring the original point cloud data after the operation. Using pre-done hand-eye calibration parameters, the post-operation point cloud is uniformly transformed to the robotic arm's base coordinate system. Then, the optimal rotation and translation rigid body transformation matrices are solved using an iterative nearest-point algorithm to correct the spatial pose of the two scanned point clouds, achieving high-precision alignment of the two sets of 3D point clouds in the reference coordinate system and eliminating positioning deviations caused by repeated scanning. After the scanned point clouds are aligned, for each valid data point in the post-operation point cloud, the nearest corresponding point is found in the pre-operation point cloud model, and the distance difference between the two points along the normal to the work surface is calculated. This difference is the chiseling depth at the corresponding location. After traversing all points within the work area, the full-area chiseling depth distribution data is obtained. Based on this, three chiseling quality evaluation indicators are calculated: First, the average chiseling depth, which is the arithmetic mean of the chiseling depth of all effective work points, reflecting the overall chiseling depth level; second, the chiseling depth uniformity coefficient, which is the ratio of the standard deviation of the chiseling depth of all points to the average chiseling depth, with 1 minus the standard deviation as the uniformity coefficient. The closer the value is to 1, the better the uniformity of the chiseling depth across the entire area; third, the effective chiseling coverage rate, which is the number of points with a chiseling depth not lower than 90% of the preset target chiseling depth, and the proportion of these points to all effective work points, reflecting the proportion of qualified chiseling areas.

[0041] Furthermore, the under-drilled areas are identified, and supplementary drilling operations are controlled. The specific method is as follows: Using 90% of the preset target chiseling depth as the qualified judgment threshold, points with chiseling depths below the threshold are marked as under-chiseling points; a region clustering algorithm is used to merge continuously distributed under-chiseling points to obtain several independent under-chiseling regions, and the spatial location, boundary range, area and average under-chiseling depth of each under-chiseling region are recorded one by one. For each under-carved area, a suitable reciprocating chiseling operation path is generated based on its shape and size, combined with the width and overlap rules of the original mesh division. Simultaneously, the chiseling process parameters are adjusted according to the average under-carved depth of the area: when the under-carved depth is small, the baseline process parameters are directly applied; when the under-carved depth is large, the moving speed is appropriately reduced or the jet water pressure is slightly increased to ensure that the depth meets the requirements after chiseling. After the chiseling path is generated, the robotic arm is controlled to perform the chiseling operation according to the aforementioned dual closed-loop pose control and speed-jet coupling rules. After chiseling is completed, point cloud scanning and quality calculation are performed again. If all three indicators meet the preset quality threshold, the chiseling quality is deemed satisfactory; otherwise, the process of under-carved identification-chiseling operation-quality verification is repeated until the quality meets the standard.

[0042] In this embodiment, step 5 is the core link of the present invention to achieve coordinated regulation of process energy and closed-loop control of terminal quality. By using a pre-constructed velocity-jet parameter coupling model, the jet water pressure and flow rate are synchronously matched according to the real-time movement speed of the robotic arm. This ensures that the reference jet energy is stable during the uniform speed section, and the jet parameters are dynamically adjusted according to the speed during acceleration, deceleration and reversal transition sections. This ensures that the jet erosion energy per unit working area is constant, effectively solving the problems of disconnect between motion and jet control and uneven chiseling depth caused by speed fluctuations in traditional solutions, and ensuring the consistency of the operation quality throughout the entire process. At the same time, by accurately registering and calculating the difference between the point clouds before and after the operation, the chiseling quality is evaluated with multi-dimensional quantitative indicators. The under-chiseled areas are accurately identified and the appropriate chiseling plan is automatically generated and executed. This forms an automated quality control closed loop of "operation-inspection-chiseling-verification", eliminating the dependence on manual inspection and chiseling planning. This reduces labor costs and the risk of missed inspections, while ensuring that the final chiseling quality is stable and meets the standards, significantly improving the level of intelligence and the accuracy of quality control throughout the chiseling operation.

[0043] Please see Figure 2 The present invention also provides a control system for a concrete water jetting and scabbing robot arm, the system being used to execute the concrete water jetting and scabbing robot arm control method described in any one of the above claims, comprising: The data acquisition and registration module is used to acquire the original point cloud data of the concrete surface to be worked on, register and convert it to the robot arm base coordinate system, and build a point cloud model of the entire working surface. The parameter mapping module is used to divide the point cloud model of the entire operation surface into several local grid units according to the effective chiseling width of a single high-pressure water jet. The principal component analysis method is used to solve the surface normal vector of the center point of each grid unit. Combined with the preset target chiseling depth and the grade of concrete to be operated, the reference process parameters of each grid unit are matched to generate a grid-level process parameter mapping table. The speed planning module is used to plan the robot arm's working path based on the grid-level process parameter mapping table, grid layout and width parameters, and to perform S-curve acceleration and deceleration planning on the working path, dividing it into multiple nodes and generating the end velocity sequence corresponding to each node. The pose control module is used for the robotic arm to perform chiseling operations with the working path and end velocity sequence as the nominal trajectory. It uses the current grid's reference target distance and surface normal vector as the control targets, collects the actual target distance in real time to calculate the target distance deviation value, and solves the incident angle deviation value between the jet axis and the concrete surface normal vector. Through the position and attitude dual closed-loop control architecture, it corrects the robotic arm's end pose in real time and maintains the target distance deviation value and incident angle deviation value within the preset threshold. The supplementary chiseling module is used to call a preset speed-jet parameter coupling model based on the real-time movement speed of the robotic arm, synchronously adjust the jet parameters, perform a second scan after the full-area chiseling operation is completed, obtain the point cloud data after the operation, compare the point cloud data before and after the operation, calculate the chiseling quality, and if there are under-chiseled areas, control the robotic arm to perform supplementary chiseling operations until the chiseling quality meets the standard.

[0044] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0045] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0046] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0047] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A control method for a robotic arm of a concrete water jetting and roughening robot, characterized in that, The specific steps include: Collect the original point cloud data of the concrete surface to be worked on, register and transform it to the robot arm base coordinate system, and construct a point cloud model of the entire working surface. Based on the effective chiseling width of a single high-pressure water jet, the point cloud model of the entire working surface is divided into several local grid units. The principal component analysis method is used to solve the surface normal vector of the center point of each grid unit. Combined with the preset target chiseling depth and the grade of concrete to be worked, the reference process parameters of each grid unit are matched to generate a grid-level process parameter mapping table. Based on the grid-level process parameter mapping table, grid layout and width parameters, the robot arm operation path is planned, and the operation path is subjected to S-curve acceleration and deceleration planning, divided into multiple nodes, and the end velocity sequence corresponding to each node is generated. The robotic arm performs chiseling operations using the work path and end-effector velocity sequence as the nominal trajectory. It uses the current grid's reference target distance and surface normal vector as control targets, collects the actual target distance in real time to calculate the target distance deviation value, and simultaneously calculates the incident angle deviation value between the jet axis and the concrete surface normal vector. Through a dual closed-loop control architecture of position and attitude, it corrects the robotic arm's end-effector posture in real time to maintain the target distance deviation value and incident angle deviation value within the preset threshold. Based on the real-time movement speed of the robotic arm, a preset speed-jet parameter coupling model is invoked to synchronously adjust the jet parameters. After the full-area chiseling operation is completed, a second scan is performed to obtain the post-operation point cloud data. The point cloud data before and after the operation are compared to calculate the chiseling quality. If there are under-chiseled areas, the robotic arm is controlled to perform supplementary chiseling operations until the chiseling quality meets the standard.

2. The control method for the robotic arm of a concrete water jetting and roughening robot according to claim 1, characterized in that, The method used to construct the global operation surface point cloud model is as follows: Using a 3D laser scanning unit mounted at the end of a robotic arm, the robotic arm drives the scanning unit to traverse the area to be worked along a preset scanning path to collect the original point cloud data of the concrete surface to be worked. based on The statistical filtering algorithm of the criterion is used to denoise the original point cloud data, and the hand-eye calibration method is used to solve the homogeneous transformation matrix between the coordinate system of the three-dimensional laser scanning unit and the coordinate system of the end effector of the robot arm. Combined with the end effector pose matrix obtained by solving the forward kinematics of the robot arm, the denoised point cloud data is mapped from the coordinate system of the scanning unit to the base coordinate system of the robot arm through matrix operations, and finally a global working surface point cloud model reflecting the shape of the working surface is constructed.

3. The control method for the robotic arm of a concrete water jetting and roughening robot according to claim 1, characterized in that, Based on the effective chiseling width of a single high-pressure water jet, the point cloud model of the entire working surface is divided into several local grid cells. Principal component analysis is used to solve for the surface normal vector of the center point of each grid cell. The method used is as follows: The point cloud model of the entire working surface is divided into grids along the tangent plane of the working surface. The width of a single grid is set to be consistent with the effective chiseling width of a single high-pressure water jet. Several local grid units with uniform specifications, no overlap and full coverage are obtained. Each grid unit corresponds to a section of the area to be chiseled. For each local grid cell, extract all point cloud data within it, and determine the average coordinates of all point cloud data within the grid cell as the coordinates of the grid cell's center point. A covariance matrix is ​​constructed for the point cloud data within each local grid cell. The covariance matrix is ​​then decomposed into eigenvalues ​​using principal component analysis to obtain the eigenvector corresponding to the smallest eigenvalue of the covariance matrix. This eigenvector is then used as the surface normal vector of the center point of the corresponding grid cell.

4. The control method for the robotic arm of a concrete water jetting and roughening robot according to claim 1, characterized in that, By combining the target roughening depth with the grade of concrete to be processed, and matching the baseline process parameters of each grid cell, a grid-level process parameter mapping table is generated. The method used is as follows: The baseline process parameters include baseline target distance, baseline moving speed, baseline jet water pressure, and baseline flow rate; Based on the jet impact dynamics model, a quantitative mapping relationship is established between the chiseling depth and the target distance, the robotic arm moving speed, the jet water pressure, and the jet flow rate. Combining the preset target chiseling depth and the grade of concrete to be processed, the corresponding reference target distance, reference moving speed, reference jet water pressure and reference flow rate are matched for each local grid cell. The reference process parameters of each local grid cell and its corresponding reference process parameters are sorted out, and a grid-level process parameter mapping table with one-to-one correspondence between grid cells and process parameters is generated.

5. The control method for the robotic arm of a concrete water jetting and roughening robot according to claim 4, characterized in that, Based on the grid-level process parameter mapping table, grid layout, and width parameters, the robotic arm's working path is planned using the following method: The operation path type is determined to be a parallel strip reciprocating path, the strip extension direction is consistent with the row arrangement direction of the grid cells, and the center line spacing between adjacent strips is set to be equal to the difference between the effective chiseling width of a single high-pressure water jet and the preset overlap allowance. The preset overlap allowance is the overlap width of adjacent strips along the width direction. Within each strip, discrete path points are arranged sequentially along the strip extension direction. Each discrete path point corresponds one-to-one with the center point of the corresponding grid cell, and each discrete path point carries the reference moving speed of the corresponding grid cell. The reversal transition section between adjacent strips is set outside the boundary of the work area, and the reversal transition section adopts an arc transition path.

6. The control method for the robotic arm of a concrete water jetting and roughening robot according to claim 5, characterized in that, The method used to perform S-curve acceleration / deceleration planning on the work path, divide it into multiple nodes, and generate the end-velocity sequence for each node is as follows: Constraints are set for the rate of change of acceleration, maximum tangential acceleration, and limit speed. A seven-segment S-curve algorithm with continuously changing acceleration is used to perform seven-segment nodal velocity interpolation on the straight work strip. The uniform speed segment matches the grid reference speed. The circular arc reversal segment is speed-limited according to the normal acceleration constraint and smoothly transitions with the speed of the straight segment. Discrete motion sequence containing position, velocity, and acceleration is obtained by discrete interpolation of the entire path in Cartesian space according to a fixed control period.

7. The method for controlling the robotic arm of a concrete water jetting and roughening robot according to claim 1, characterized in that, The robot arm's end-effector pose is corrected in real time using a dual closed-loop control architecture for both position and attitude. The method used is as follows: The robotic arm performs the chiseling operation with the working path and end velocity sequence as the nominal trajectory, and uses the reference target distance and surface normal vector of the grid cell corresponding to the current working position as the control target; Three laser rangefinders, evenly distributed around the end of the robotic arm, collect multi-point range data from the nozzle to the concrete working surface in real time. Based on this range data, the actual target distance is calculated. The difference between the actual target distance and the reference target distance of the corresponding grid cell is calculated to obtain the target distance deviation value. At the same time, the incident angle deviation value between the jet axis and the normal vector of the concrete surface is calculated. A dual closed-loop control architecture for position and attitude is constructed. The target distance deviation value is input into the position control loop, and a fuzzy adaptive PID algorithm is used for calculation to output the displacement correction value along the normal vector of the working surface. The incident angle deviation value is input into the attitude control loop, and a quaternion spherical linear interpolation algorithm is used to output the attitude correction value. The displacement and attitude correction values ​​output by the position control loop and attitude control loop are superimposed on the nominal trajectory of the robotic arm to correct the position and pose of the robotic arm end in real time, and maintain the target distance deviation and incident angle deviation values ​​within the corresponding preset thresholds.

8. The control method for the robotic arm of a concrete water jetting and roughening robot according to claim 6, characterized in that, Based on the real-time movement speed of the robotic arm, a preset speed-jet parameter coupling model is invoked to synchronously adjust the jet parameters. The method used is as follows: The jet parameters include jet water pressure and jet flow rate; A velocity-pressure-flow coupling model based on the principle of constant jet energy per unit area is preset. During the roughening operation of the robotic arm, the actual movement speed of the robotic arm end is acquired in real time and input into the coupling model. The high-pressure water jet pressure and flow rate are synchronously matched through the coupling model. Specifically, for the uniform speed section of the operation path, the reference jet water pressure and reference jet flow rate are kept constant. For the acceleration / deceleration section and the reversal transition section, the jet water pressure and jet flow rate are linearly and synchronously adjusted according to the rate of change of the robotic arm speed, so that the jet impact energy received per unit operation area remains consistent throughout the entire operation.

9. The control method for the robotic arm of a concrete water jetting and roughening robot according to claim 8, characterized in that, The method used to calculate the roughening quality by comparing point cloud data before and after the operation is as follows: After the full-area roughening operation is completed, the robotic arm is controlled to drive the three-dimensional laser scanning unit mounted on its end to perform a full scan of the concrete working surface along a preset secondary scanning path to obtain the post-operation point cloud model. The post-operation point cloud model is registered and aligned with the constructed global operation surface point cloud model. The global chiseling depth distribution data is obtained through point cloud difference calculation. The average chiseling depth, chiseling depth uniformity coefficient, and effective chiseling coverage rate are used as chiseling quality evaluation indicators. The global average chiseling depth, the standard deviation of chiseling depth in each region, and the proportion of effective chiseling area are statistically calculated to obtain the chiseling quality. Based on the chiseling depth distribution data, under-chiseled areas with chiseling depths below the preset target chiseling depth threshold are identified. The location, range, and area of ​​the under-chiseled areas are determined. According to the location, range, and grid division of the under-chiseled areas, an appropriate supplementary chiseling operation path is automatically generated. The robotic arm is controlled to drive the jet nozzle and perform a second supplementary chiseling operation on the under-chiseled areas according to the supplementary chiseling operation path and the optimized baseline process parameters. After the supplementary chiseling is completed, point cloud scanning and quality calculation are performed again until the average chiseling depth, chiseling depth uniformity coefficient, and effective chiseling coverage all meet the preset chiseling quality threshold requirements, and the chiseling quality is judged to be up to standard.

10. A control system for a concrete water jetting and roughening robot arm, characterized in that, The system is used to execute the concrete water jetting and roughening robot robotic arm control method according to any one of claims 1-9, including: The data acquisition and registration module is used to acquire the original point cloud data of the concrete surface to be worked on, register and convert it to the robot arm base coordinate system, and build a point cloud model of the entire working surface. The parameter mapping module is used to divide the point cloud model of the entire operation surface into several local grid units according to the effective chiseling width of a single high-pressure water jet. The principal component analysis method is used to solve the surface normal vector of the center point of each grid unit. Combined with the preset target chiseling depth and the grade of concrete to be operated, the reference process parameters of each grid unit are matched to generate a grid-level process parameter mapping table. The speed planning module is used to plan the robot arm's working path based on the grid-level process parameter mapping table, grid layout and width parameters, and to perform S-curve acceleration and deceleration planning on the working path, dividing it into multiple nodes and generating the end velocity sequence corresponding to each node. The pose control module is used for the robotic arm to perform chiseling operations with the working path and end velocity sequence as the nominal trajectory. It uses the current grid's reference target distance and surface normal vector as the control targets, collects the actual target distance in real time to calculate the target distance deviation value, and solves the incident angle deviation value between the jet axis and the concrete surface normal vector. Through the position and attitude dual closed-loop control architecture, it corrects the robotic arm's end pose in real time and maintains the target distance deviation value and incident angle deviation value within the preset threshold. The collaborative quality closed-loop module is used to call the preset speed-jet parameter coupling model according to the real-time movement speed of the robotic arm, synchronously adjust the jet parameters, perform a second scan after the whole-domain chiseling operation is completed, obtain the point cloud data after the operation, compare the point cloud data before and after the operation, calculate the chiseling quality, and if there are under-chiseled areas, control the robotic arm to perform supplementary chiseling operations until the chiseling quality meets the standard.

Images