A control method of a cantilever vibrating robot for concrete vibrating operation

Abstract

The application discloses a control method of a cantilever vibrating robot for concrete vibrating operation, and relates to the technical field of concrete vibrating robot control. The application constructs a three-dimensional digital grid model, divides vibrating into three functional stages of initial vibrating, re-vibrating and vibrating arrangement according to a concrete pouring operation process, and finds out candidate vibrating points in the three-dimensional digital grid model for each stage. The application collects vibration resistance feedback signals of a vibrating rod and operation cumulative duration signals of the robot, extracts core features of the two signals to determine the current stage of vibrating, dynamically extracts effective vibrating points from the three-dimensional digital grid model, generates a continuous operation path, and simultaneously, the robot cantilever adaptively adjusts vibrating depth and posture according to the continuous operation path and stage requirements. The application improves the working efficiency of the vibrating robot.

Description

Technical Field

[0001] This invention relates to the field of concrete vibration robot control technology, specifically to a control method for a cantilever vibration robot for concrete vibration operations. Background Technology

[0002] Concrete vibration is a crucial process for ensuring the density and mechanical properties of concrete structures. With the development of automation technology, cantilever vibration robots are gradually replacing manual labor. However, existing robot control methods still have the following shortcomings: Firstly, existing control methods lack the ability to deeply integrate with the design drawings of the pouring silo, relying mostly on single sensor scanning and failing to accurately associate with boundary information such as embedded parts and silo walls, which can easily lead to misjudgment of vibration points. Secondly, the division of the vibration stage is not integrated with the entire concrete pouring process and relies heavily on manual experience. Third, the grid density was not adjusted according to the difference in vibration requirements between the area around the embedded parts and the leveled area, and the path planning did not take into account the connectivity of the area, resulting in the robot frequently moving across areas and low work efficiency. Therefore, there is an urgent need for a control method for a cantilever vibrating robot for concrete vibration operations. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a control method for a cantilever vibrating robot for concrete vibration operations, which solves the problems of inaccurate three-dimensional digital modeling, stage misjudgment, and poor posture adaptation in existing concrete vibration robots.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a control method for a cantilever vibrating robot for concrete vibration operations, comprising: Step 1: Based on the design drawings of the concrete pouring silo, the robot is equipped with a lidar and vision camera to scan the working surface in real time to generate a three-dimensional digital mesh model. According to the concrete pouring operation process, the vibration is divided into three functional stages: initial vibration, secondary vibration, and vibration discharge. For each stage, the candidate vibration points in the corresponding three-dimensional digital mesh model are identified. Step 2: Collect the vibration resistance feedback signal of the vibrating rod and the cumulative operation time signal of the robot to determine the current stage of vibration, and dynamically extract the effective vibration points from the three-dimensional digital mesh model to generate a continuous operation path. The robot arm adaptively adjusts the vibration depth and posture according to the continuous operation path and stage requirements.

[0005] As a further aspect of the present invention, the specific operation for generating a three-dimensional digital mesh model is as follows: Analyze the design drawings of the concrete pouring silo, extract the length, width, and height of the silo, as well as the three-dimensional coordinates of the embedded parts, and place n cubic reference blocks with unique reflective codes at the four corners and the midpoints of the two long sides of the silo. The bottom of the reference blocks is completely fixed to the bottom surface of the pouring silo, and the surface is unobstructed. Where n≥6. The robot is equipped with a lidar and vision camera to scan all reference blocks one by one. The unique identity of each reference block is accurately identified through reflective coding. The pixel coordinates of the reference blocks collected by the two sensors are established to form a one-to-one correspondence, and the calibration of the same coordinate system is completed. The robot's movement trajectory is planned in a bow shape, with the trajectory spacing set to 0.8 times the scanning width of the LiDAR. The robot moves at a constant speed along the planned trajectory. The LiDAR on board continuously collects three-dimensional point cloud data of the working surface and the warehouse wall at a frame rate f. The vision camera simultaneously takes an image at a high resolution every fixed distance, and the image must contain at least one reflective coded reference block. At the same time, the robot's odometer records the three-dimensional position coordinates of the robot body in real time at the moment of each data acquisition. By using the reflective coding of the reference block, the point cloud data of the corresponding area is matched with the image data. Two types of geometric features, namely the concrete surface corners and the edges of the embedded parts, are extracted from the point cloud. Two types of visual features, namely the surface depressions and texture differences, are extracted from the image, forming a complete point cloud dataset containing four types of features. A cubic mesh is generated according to the differentiation rule: a closed space is formed by uniformly extending a distance c in all directions based on the outer contour of the embedded part, and a mesh with a side length of a is used in the flat area, while a mesh with a side length of b is used, and a < b. Traverse all grid vertices, remove vertices that fall on embedded parts, silo walls, and reference blocks, mark the remaining vertices as candidate vibration points, and output the final three-dimensional digital grid model.

[0006] As a further aspect of the present invention, the vibration is divided into three functional stages: initial vibration, secondary vibration, and vibration discharge, specifically including: The operation position signal and concrete placement volume data of the concrete placing machine are acquired in real time. When the placing machine completes the placement operation in a certain area and leaves the area, and the accumulation height of the concrete surface in that area is ≤h1, it is determined as the initial vibration stage start condition. The end condition is that the leveling equipment in that area begins to enter the operation range. Here, h1 is the accumulation height threshold. Based on the work progress feedback signal of the leveling equipment, when the leveling equipment completes the leveling work in a certain area and withdraws, it is determined as the start condition of the re-vibration stage, and the end condition is that the finishing equipment in that area begins to move to the edge of the work area. The positioning signal of the finishing equipment and the hardness data of the concrete surface are monitored in real time. When the finishing equipment moves to the edge of a certain area and the hardness of the concrete surface in that area reaches the specified threshold, it is determined as the start condition of the vibration dissipation stage. The end condition is that the finishing equipment in that area begins to formally contact the concrete surface to perform finishing work.

[0007] As a further aspect of the present invention, for each stage, candidate vibration points in the corresponding three-dimensional digital mesh model are identified, specifically as follows: Extract the original node spacing of the three-dimensional digital mesh model and set spacing parameters that match the mesh side length for each stage: the spacing parameters for the initial vibration stage are set to a and b, the spacing parameters for the re-vibration stage are set to 2a and 2b, and the spacing parameters for the vibration discharge stage are set to 3a and 3b. In the robot control system, a linkage logic for selecting stage points is established: if it is the initial vibration stage, the initial vibration spacing parameters (a, b) are automatically called to select all grid nodes from the three-dimensional digital grid model that satisfy the adjacent node spacing of a or b, and mark them as candidate vibration points for initial vibration; if it is the re-vibration stage, the re-vibration spacing parameters (2a, 2b) are automatically called to select grid nodes according to the rule of selecting one every other node, and mark them as candidate vibration points for re-vibration; if it is the de-vibration stage, the de-vibration spacing parameters (3a, 3b) are automatically called to select one every other two nodes, and mark them as candidate vibration points for de-vibration.

[0008] As a further aspect of the present invention, the specific steps for determining the current vibration stage are as follows: Collect N sets of complete historical operation data for each of the three stages of initial vibration, re-vibration, and vibration exhaustion. Each set of data synchronously includes the vibration resistance feedback signal of the vibrating rod and the cumulative operation time signal of the robot. Extract the core features of resistance and core features of time respectively, and stitch the two together to obtain the feature fusion vector of each set of data. For each stage, N feature fusion vectors are obtained, and density peak clustering is performed on them to obtain several clusters. The feature center vector of each cluster is recorded. When the robot is working, it generates a real-time feature fusion vector of the current working area and calculates the cosine similarity between the real-time feature fusion vector and the center of each cluster in the three stages. It finds the minimum cosine similarity d1, d2, and d3 with all clusters in the three stages of initial vibration, re-vibration, and exhaust vibration, selects the stage corresponding to min{d1,d2,d3}, and locks it as the current vibration stage.

[0009] As a further aspect of the present invention, the core resistance features include the average resistance value and the resistance linear fitting slope, and the core duration features include the ratio of the current stage's cumulative duration to the corresponding stage's standard total duration and the frequency of resistance changes within a unit of cumulative duration.

[0010] As a further aspect of the present invention, before the vibration operation of a new batch of concrete, three sets of initial stage data of the first operation area of ​​the batch are collected, covering the initial vibration and secondary vibration stages. After generating the feature fusion vector of the new batch, it is substituted into the cluster set of the corresponding stage and the cosine similarity with each cluster center is calculated. If the cosine similarity between the feature fusion vector of a new batch and all clusters of the corresponding stage is lower than a preset threshold, the feature fusion vector of the new batch is added to the cluster set of the corresponding stage as a new cluster center. At the same time, the cluster with the lowest cosine similarity with the feature fusion vector of the new batch is deleted from the original cluster.

[0011] As a further aspect of the present invention, the specific steps for generating a continuous job path are as follows: Obtain candidate vibration points in the three-dimensional digital mesh model for the current vibration stage, and mark each mesh node as either unvibrated or vibrated using the operation status label library of the robot control system. Initially, all nodes are unvibrated. The candidate vibration point set is sorted according to regional connectivity, with priority given to continuous nodes within the same connected region, and then other connected regions are covered in turn. Using the sorted candidate vibration points as vertices, an initial path is generated according to the principle of connecting adjacent points with straight lines. For turning nodes in the path, the angle between two adjacent path segments is calculated. If the angle is greater than the threshold θ, a transition point is inserted between the two nodes, and finally a complete continuous operation path is output.

[0012] As a further aspect of the present invention, the step of sorting the candidate vibration point set according to regional connectivity is as follows: Traverse all candidate vibration points, select one unmarked candidate vibration point as the starting point, and search its six spatially adjacent grid nodes. If the adjacent nodes are not marked, add the adjacent nodes to the same temporary set as the starting point. Then, take all points in the temporary set as the new starting point and repeat the above adjacency search and set merging operation until no new candidate vibration points are added to the temporary set. At this time, the temporary set is a complete connected region. Mark the region identifier of all candidate vibration points in the region and repeat the above process until all candidate vibration points are assigned to the corresponding connected regions. Calculate the center coordinates (Xc, Yc) of each connected region, that is, the average X coordinate of all points in the region is Xc and the average Y coordinate is Yc. Sort all connected regions in the order of "Xc from small to large → Yc from small to large" to form the region operation priority. If there are regions with the same Xc and Yc, then supplement them by sorting in the order of the number of candidate vibration points in the region from large to small. Starting from the center point of each connected region, sort the points according to the rule of clockwise spiral expansion: first mark the center point, then mark the 6 adjacent points of the center point in turn, then mark the adjacent points of the adjacent points, until all points in the region are marked.

[0013] As a further aspect of the present invention, the adaptive adjustment of vibration depth and posture specifically includes: Set the vibration depth of the cantilever foundation according to the vibration stage: h1 for the initial vibration stage, h2 for the secondary vibration stage, and h3 for the exhaust vibration stage. Read the three-dimensional coordinates (X,Y,Z) of the current target point, subtract the foundation vibration depth from the Z-axis coordinate to obtain the initial vibration depth, where h3 < h2 < h1. Calculate the difference ΔZ between the Z-axis coordinate of the current target point and the average Z-axis coordinate of the grid nodes in the area. If ΔZ is positive, decrease the initial vibration depth by ΔZ. If ΔZ is negative, increase the initial vibration depth by |ΔZ|. Bind the current stage to the cantilever attitude angle: set the attitude angle to 0° for the initial vibration stage, α° for the re-vibration stage, and β° for the vibration discharge stage. Read the Z-axis coordinates of the three adjacent grid nodes around the target point, calculate the slope of the area, and if the slope is greater than the threshold, shift the attitude angle by the same angle in the slope direction.

[0014] This invention provides a control method for a cantilever vibrating robot in concrete vibration operations, which has the following advantages compared with the prior art: (1) This invention constructs a three-dimensional digital grid model by combining the design drawings of the concrete pouring silo with the data from the dual sensors, accurately extracts the silo boundary and embedded part information, sets the grid density according to the differentiated rules, eliminates invalid vertices to determine candidate vibration points, and solves the problems of existing methods that are easy to deviate from reality in modeling and easy to misjudge the points. (2) This invention establishes feature models for each stage by clustering historical operation data, and generates feature vectors by combining real-time vibration resistance and operation duration signals. It locks the current stage by similarity calculation and supports dynamic calibration of clusters for new batches of data to ensure that initial vibration, re-vibration and vibration discharge are all started in the concrete-suitable state. (3) The present invention sorts candidate points according to regional connectivity to generate continuous paths, reduces the frequency of robot movement across regions, optimizes path smoothness through transition points, and adaptively adjusts vibration depth and attitude according to stages, solving the problems of path redundancy and parameter rigidity in existing methods and improving vibration efficiency. Attached Figure Description

[0015] Figure 1 This is a flowchart of the steps of the present invention; Figure 2 This is a flowchart illustrating the steps involved in determining the current vibration stage in this invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1 like Figure 1 This invention provides a control method for a cantilever vibrating robot for concrete vibration operations, comprising: Step 1: Based on the design drawings of the concrete pouring silo, and using the lidar and vision camera mounted on the robot, a three-dimensional digital mesh model is generated by scanning the work surface in real time. The specific operation is as follows: Analyze the design drawings of the concrete pouring silo, extract the length, width, and height of the silo, as well as the three-dimensional coordinates of the embedded parts (such as steel reinforcement frames and reserved holes), and place n cubic reference blocks with unique reflective codes at the four corners and the midpoints of the two long sides of the silo. The bottom of the reference blocks is completely fixed to the bottom surface of the pouring silo, and the surface is unobstructed. Where n≥6. By utilizing the length, width, and height of the silo, the maximum range of the three-dimensional mesh model can be directly determined. Extracting the three-dimensional coordinates of the embedded parts can lock the "vibration restricted area" in advance. Furthermore, n≥6 can provide sufficient redundant points. Even if a small number of reference blocks fail, a stable on-site absolute coordinate system can be constructed using the remaining reference blocks to avoid calibration failure. The robot is equipped with a lidar and vision camera to scan all reference blocks one by one. The unique identity of each reference block is accurately identified through reflective coding. The pixel coordinates of the reference blocks collected by the two sensors are established to form a one-to-one correspondence, and the calibration of the same coordinate system is completed. The design with a unique reflective code allows LiDAR and vision cameras to quickly identify the identity of each reference block, ensuring that the three-dimensional coordinates of each reference block correspond to its identity; The robot's movement trajectory is planned in a bow shape, with the trajectory spacing set to 0.8 times the width of the LiDAR scan, ensuring that adjacent scan areas have a 20% overlap; The robot moves at a constant speed along the planned trajectory. The onboard lidar continuously collects three-dimensional point cloud data of the working surface and the warehouse wall at a frame rate f. The vision camera simultaneously takes an image at a high resolution every fixed distance, and the image must contain at least one reflective coded reference block. At the same time, the robot's three-dimensional position coordinates are recorded in real time at the moment of each data collection by the onboard odometer. The collected point cloud data is filtered by a radius filtering algorithm to remove isolated noise points in the point cloud, such as reflection points of dust and debris in the air, so as to avoid noise causing false bumps or holes in the mesh model. By using the reflective coding of the reference block, the point cloud data of the corresponding area is matched with the image data. Two types of geometric features, namely the concrete surface corners and the edges of the embedded parts, are extracted from the point cloud. Two types of visual features, namely the surface depressions and texture differences, are extracted from the image, forming a complete point cloud dataset containing four types of features. A cubic mesh is generated according to the differentiation rule: a closed space is formed by uniformly extending a distance c in all directions based on the outer contour of the embedded part, and a mesh with a side length of a is used in the flat area, while a mesh with a side length of b is used, and a < b. The embedded parts have a complex structure and small spacing. If a large grid is used, the points may fall on the edge or gap of the embedded parts. However, a small grid can make the nodes denser, making it easier to accurately avoid restricted areas during screening and ensuring that the vibrator only works in the concrete area. The leveling area is generally larger. Using a small grid will generate a large number of redundant points, which will cause the robot to move frequently and vibrate repeatedly. A large grid can reduce the number of points, shorten the path distance, and greatly improve the operation speed. Traverse all grid vertices, remove vertices that fall on embedded parts, silo walls, and reference blocks, mark the remaining vertices as candidate vibration points, and output the final three-dimensional digital grid model. According to the concrete pouring process, vibration is divided into three functional stages: initial vibration, secondary vibration, and discharge vibration, which specifically include: The operation position signal and concrete placement volume data of the concrete placing machine are acquired in real time. When the placing machine completes the placement operation in a certain area and leaves the area, and the accumulation height of the concrete surface in that area is ≤h1, it is determined as the initial vibration stage start condition. The end condition of this stage is that the leveling equipment (such as the leveling machine) in that area begins to enter the operation range, where h1 is the accumulation height threshold. After the concrete is placed, it is in a loose state with a high porosity. At this time, it needs to be compacted by vibration. This stage is the foundation for the formation of the concrete structure and must be started closely after the concrete placement operation is completed. Based on the work progress feedback signal of the leveling equipment, when the leveling equipment completes the leveling work in a certain area and withdraws, it is determined as the start condition of the re-vibration stage. The end condition of this stage is that the finishing equipment (such as the finishing machine) in that area begins to move to the edge of the work area. Leveling work will change the surface morphology of concrete and initially stabilize the internal structure, but there may be local insufficient density. Therefore, re-vibration should be carried out when the concrete is in a semi-solid state (not fully set). The positioning signal of the finishing equipment and the hardness data of the concrete surface are monitored in real time. When the finishing equipment moves to the edge of a certain area and the hardness of the concrete surface in that area reaches the specified threshold (measured by a hardness tester), it is determined to be the start condition of the vibration dissipation stage. The end condition of this stage is that the finishing equipment in that area begins to formally contact the concrete surface to perform finishing work. Before applying the plaster, the concrete surface has entered the initial setting stage and the internal structure is basically formed. Only surface air bubbles need to be treated. If the process is started too early, the internal structure will be damaged, and if it is started too late, the air bubbles cannot be effectively removed. For each stage, candidate vibration points are identified in the corresponding 3D digital mesh model. The specific steps are as follows: Extract the original node spacing of the 3D digital mesh model (the side length of the mesh around the embedded part is a, and the side length of the mesh in the flat area is b), and set the spacing parameters matching the mesh side length for each stage according to the stage design principle: the spacing parameters for the initial vibration stage are set to a and b, the spacing parameters for the re-vibration stage are set to 2a and 2b, and the spacing parameters for the vibration discharge stage are set to 3a and 3b, to ensure that the spacing of adjacent effective points in the same stage is uniform, and the effective range overlap ratio is [1 / 3, 1 / 2], with no vibration blind zone; For example, if the original node spacing of the flat area in the 3D digital mesh model is b, the flat area spacing parameter is set to b in the initial vibration stage, that is, all mesh nodes in the flat area are selected as valid points, while in the re-vibration stage it is set to 2b, that is, one is selected every other mesh node. In the robot control system, a linkage logic for selecting stage points is established: If it is the initial vibration stage, the initial vibration spacing parameters (a, b) are automatically called to select all grid nodes from the three-dimensional digital grid model that meet the adjacent node spacing of a (around the embedded part) or b (flat area), and they are marked as candidate vibration points for initial vibration; if it is the re-vibration stage, the re-vibration spacing parameters (2a, 2b) are automatically called to select grid nodes according to the rule of selecting one every other node, and they are marked as candidate vibration points for re-vibration; if it is the de-vibration stage, the de-vibration spacing parameters (3a, 3b) are automatically called to select one every other two nodes, and they are marked as candidate vibration points for de-vibration.

[0018] Step 2: Collect the vibration resistance feedback signal of the vibrating rod and the cumulative operation time signal of the robot to determine the current vibration stage; Based on the currently identified vibration stage, effective vibration points are dynamically extracted from the 3D digital mesh model, and a continuous operation path is generated. The specific operation is as follows: Obtain candidate vibration points in the three-dimensional digital mesh model for the current vibration stage, and mark each mesh node as either unvibrated or vibrated using the operation status label library of the robot control system. Initially, all nodes are unvibrated. The candidate vibration point set is sorted according to regional connectivity, prioritizing consecutive nodes within the same connected region, and then covering other connected regions in turn. Specifically, this includes: Traverse all candidate vibration points, select one unmarked candidate vibration point as the starting point, and search its six spatially adjacent grid nodes. If the adjacent nodes are not marked, add the adjacent nodes to the same temporary set as the starting point. Then, take all points in the temporary set as the new starting point and repeat the above adjacency search and set merging operation until no new candidate vibration points are added to the temporary set. At this time, the temporary set is a complete connected region. Mark the region identifier of all candidate vibration points in the region and repeat the above process until all candidate vibration points are assigned to the corresponding connected regions. Calculate the center coordinates (Xc, Yc) of each connected region, that is, the average X coordinate of all points in the region is Xc and the average Y coordinate is Yc. Sort all connected regions in the order of "Xc from small to large → Yc from small to large" to form the region operation priority. If there are regions with the same Xc and Yc, then supplement them by sorting in the order of the number of candidate vibration points in the region from large to small. Starting from the center point of each connected region, sort the points according to the rule of clockwise spiral expansion: first mark the center point, then mark the 6 adjacent points of the center point in turn, then mark the adjacent points of the adjacent points, until all points in the region are marked. Using the sorted candidate vibration points as vertices, an initial path is generated according to the principle of connecting adjacent points with straight lines. For turning nodes in the path, the angle between two adjacent path segments is calculated. If the angle is greater than the threshold θ, a transition point is inserted between the two nodes to make the path turning smoother. Finally, the complete operation trajectory is output. The robotic cantilever adaptively adjusts the vibration depth and posture based on the point coordinates and stage requirements, including: Set the vibration depth of the cantilever foundation according to the vibration stage: h1 for the initial vibration stage, h2 for the secondary vibration stage, and h3 for the exhaust vibration stage. Read the three-dimensional coordinates (X,Y,Z) of the current target point, subtract the foundation vibration depth from the Z-axis coordinate to obtain the initial vibration depth, where h3 < h2 < h1. Calculate the difference ΔZ between the Z-axis coordinate of the current target point and the average Z-axis coordinate of the grid nodes in the area. If ΔZ is positive, decrease the initial vibration depth by ΔZ. If ΔZ is negative, increase the initial vibration depth by |ΔZ| to ensure that the vibrator always fits the actual concrete surface and the insertion depth is consistent. The current stage is bound to the cantilever attitude angle: the attitude angle is set to 0° (perpendicular to the working surface) in the initial vibration stage, α° (slightly inclined along the tangent of the working surface) in the re-vibration stage, and β° (parallel to the surface of the working surface) in the exhaust stage. The Z-axis coordinates of the three adjacent grid nodes around the target point are read, and the slope of the area is calculated. If the slope is greater than the threshold, the attitude angle is shifted by the same angle in the slope direction to ensure that the vibrator and the concrete surface always maintain the set attitude relationship.

[0019] Example 2 This embodiment further discloses a method for determining the current vibration stage based on Embodiment 1, such as... Figure 2 As shown, the specific content includes: Collect N sets of complete historical operation data for each of the three stages of initial vibration, re-vibration, and vibration exhaustion. Each set of data synchronously includes the vibration resistance feedback signal of the vibrating rod and the cumulative operation time signal of the robot. Extract the core features of resistance and core features of time respectively, and stitch the two together to obtain the feature fusion vector of each set of data. The core characteristics of the resistance include the average resistance value and the slope of the linear fitting of the resistance: The average resistance value is calculated by taking the arithmetic mean of all the instantaneous vibration resistance values ​​collected over several consecutive collection cycles. The average resistance value is the smallest when the concrete is loose in the initial vibration stage, the average resistance value is in the middle when the concrete is semi-solid in the re-vibration stage, and the average resistance value is the largest when the surface of the concrete is initially set in the dissipation stage. The resistance linear fitting slope is obtained by using the acquisition period as the horizontal axis and the resistance value at the corresponding moment as the vertical axis, and performing linear fitting using the least squares method. In the initial vibration stage, the concrete changes from loose to preliminary compaction, and the resistance rises rapidly, resulting in the largest resistance linear fitting slope. In the re-vibration stage, the concrete compaction increases slowly, the resistance rises gradually, and the resistance linear fitting slope is in the middle. In the dissipation stage, the surface hardness of the concrete is stable, the resistance rises very slowly, and the resistance linear fitting slope is the smallest. The core characteristics of the duration include the ratio of the current stage's cumulative duration to the corresponding stage's standard total duration, and the frequency of resistance changes within a unit of cumulative duration. The ratio of the current stage's cumulative time to the corresponding stage's standard total time is obtained by first calculating the standard total time of the three stages using historical data, then recording the current task's cumulative time, and comparing the two to obtain this value. The frequency of resistance change within a unit cumulative time period is determined by setting a minimum resistance change threshold, counting the number of times within the cumulative time period when the absolute value of the difference between the instantaneous value of vibration resistance and the value of the previous cycle is greater than or equal to the minimum resistance change threshold, and comparing the number of times with the cumulative time period to obtain the value. For each stage, N feature fusion vectors are obtained, and density peak clustering is performed on them to obtain several clusters. Each cluster represents a typical operation mode of that stage, and the feature center vector of each cluster is recorded. Before the vibration operation of a new batch of concrete, three sets of initial stage data are collected from the first work area of ​​the batch, covering the initial vibration and secondary vibration stages. The feature fusion vector of the new batch is generated according to the feature extraction method described above. The feature fusion vector of the new batch is substituted into the cluster set of the corresponding stage and the cosine similarity with each cluster center is calculated. If the cosine similarity of the feature fusion vector of a new batch with all clusters of the corresponding stage is lower than the preset threshold, the feature fusion vector of the new batch is added to the cluster set of the corresponding stage as a new cluster center. At the same time, the cluster with the lowest cosine similarity with the feature fusion vector of the new batch is deleted. This completes the batch dynamic calibration of the clusters and ensures that the cluster center always fits the physical properties of the current concrete and the status of the operating equipment. When the robot is working, it generates a real-time feature fusion vector of the current working area and calculates the cosine similarity between the real-time feature fusion vector and the center of each cluster in the three stages. It finds the minimum cosine similarity d1, d2, and d3 with all clusters in the three stages of initial vibration, re-vibration, and exhaust vibration, selects the stage corresponding to min{d1,d2,d3}, and locks it as the current vibration stage.

[0020] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0021] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A control method for a cantilever vibrating robot for concrete vibration operations, characterized in that, include: Step 1: Based on the design drawings of the concrete pouring silo, and combined with the lidar and vision camera mounted on the robot, a 3D digital mesh model is generated by scanning the working surface in real time. Following the concrete pouring operation process, vibration is divided into three functional stages: initial vibration, secondary vibration, and final vibration. For each stage, candidate vibration points in the corresponding 3D digital mesh model are identified. Specifically, the original node spacing of the 3D digital mesh model is extracted, and spacing parameters matching the mesh edge length are set for each stage: initial vibration stage spacing parameters are set to a and b, secondary vibration stage spacing parameters are set to 2a and 2b, and final vibration stage spacing parameters are set to 3a and 3b. (The last sentence appears to be incomplete and requires further context.) The robot control system establishes a linkage logic for selecting stage points: In the initial vibration stage, the initial vibration spacing parameters (a, b) are automatically invoked to select all grid nodes from the three-dimensional digital grid model that satisfy the adjacent node spacing of a or b, and these nodes are marked as candidate vibration points for the initial vibration; In the secondary vibration stage, the secondary vibration spacing parameters (2a, 2b) are automatically invoked to select grid nodes according to the rule of selecting one node every other node, and these nodes are marked as candidate vibration points for the secondary vibration; In the de-vibration stage, the de-vibration spacing parameters (3a, 3b) are automatically invoked to select nodes according to the rule of selecting one node every two nodes, and these nodes are marked as candidate vibration points for the de-vibration. Step 2: Collect the vibration resistance feedback signal of the vibratory rod and the cumulative operation time signal of the robot to determine the current stage of vibration. Dynamically extract effective vibration points from the 3D digital mesh model to generate a continuous operation path. The robot arm adaptively adjusts the vibration depth and posture according to the continuous operation path and stage requirements. The specific steps for determining the current stage of vibration are as follows: Collect N sets of complete historical operation data for each of the three stages: initial vibration, secondary vibration, and final vibration. Each set of data synchronously includes the vibration resistance feedback signal of the vibratory rod and the cumulative operation time signal of the robot, and extract the core characteristics of resistance and the core characteristics of time respectively. The two are concatenated to obtain the feature fusion vector of each data set; for each stage, N feature fusion vectors are obtained, and they are clustered by density peak clustering to obtain several clusters. The feature center vector of each cluster is recorded; when the robot is working, the real-time feature fusion vector of the current working area is generated, and the cosine similarity between the real-time feature fusion vector and the center of each cluster in the three stages is calculated. The minimum cosine similarity d1, d2, d3 with all clusters in the three stages of initial vibration, re-vibration, and exhaust vibration are found respectively. The stage corresponding to min{d1,d2,d3} is selected and locked as the current vibration stage.

2. The control method for a cantilever vibrating robot for concrete vibration operation according to claim 1, characterized in that, The specific steps for generating a 3D digital mesh model are as follows: Analyze the design drawings of the concrete pouring silo, extract the length, width, and height of the silo, as well as the three-dimensional coordinates of the embedded parts, and place n cubic reference blocks with unique reflective codes at the four corners and the midpoints of the two long sides of the silo. The bottom of the reference blocks is completely fixed to the bottom surface of the pouring silo, and the surface is unobstructed. Where n≥6. The robot is equipped with a lidar and vision camera to scan all reference blocks one by one. The unique identity of each reference block is accurately identified through reflective coding. The pixel coordinates of the reference blocks collected by the two sensors are established to form a one-to-one correspondence, and the calibration of the same coordinate system is completed. The robot's movement trajectory is planned in a bow shape, with the trajectory spacing set to 0.8 times the scanning width of the LiDAR. The robot moves at a constant speed along the planned trajectory. The LiDAR on board continuously collects three-dimensional point cloud data of the working surface and the warehouse wall at a frame rate f. The vision camera simultaneously takes an image at a high resolution every fixed distance, and the image must contain at least one reflective coded reference block. At the same time, the robot's odometer records the three-dimensional position coordinates of the robot body in real time at the moment of each data acquisition. By using the reflective coding of the reference block, the point cloud data of the corresponding area is matched with the image data. Two types of geometric features, namely the concrete surface corners and the edges of the embedded parts, are extracted from the point cloud. Two types of visual features, namely the surface depressions and texture differences, are extracted from the image, forming a complete point cloud dataset containing four types of features. A cubic mesh is generated according to the differentiation rule: a closed space is formed by uniformly extending a distance c in all directions based on the outer contour of the embedded part, and a mesh with a side length of a is used in the flat area, while a mesh with a side length of b is used, and a < b. Traverse all grid vertices, remove vertices that fall on embedded parts, silo walls, and reference blocks, mark the remaining vertices as candidate vibration points, and output the final three-dimensional digital grid model.

3. The control method for a cantilever vibrating robot for concrete vibration operation according to claim 1, characterized in that, The vibration process is divided into three functional stages: initial vibration, secondary vibration, and vibration discharge. Specifically, these include: The operation position signal and concrete placement volume data of the concrete placing machine are acquired in real time. When the placing machine completes the placement operation in a certain area and leaves the area, and the accumulation height of the concrete surface in that area is ≤h1, it is determined as the initial vibration stage start condition. The end condition is that the leveling equipment in that area begins to enter the operation range. Here, h1 is the accumulation height threshold. Based on the work progress feedback signal of the leveling equipment, when the leveling equipment completes the leveling work in a certain area and withdraws, it is determined as the start condition of the re-vibration stage, and the end condition is that the finishing equipment in that area begins to move to the edge of the work area. The positioning signal of the finishing equipment and the hardness data of the concrete surface are monitored in real time. When the finishing equipment moves to the edge of a certain area and the hardness of the concrete surface in that area reaches the specified threshold, it is determined as the start condition of the vibration dissipation stage. The end condition is that the finishing equipment in that area begins to formally contact the concrete surface to perform finishing work.

4. The control method for a cantilever vibrating robot for concrete vibration operation according to claim 1, characterized in that, The core features of resistance include the average resistance value and the slope of the linear fitting of resistance. The core features of duration include the ratio of the cumulative duration of the current stage to the standard total duration of the corresponding stage and the frequency of resistance changes within a unit of cumulative duration.

5. The control method for a cantilever vibrating robot for concrete vibration operation according to claim 1, characterized in that, Before the vibration operation of a new batch of concrete, three sets of initial stage data are collected from the first operation area of ​​the batch, covering the initial vibration and secondary vibration stages. After generating the feature fusion vector of the new batch, it is substituted into the cluster set of the corresponding stage and the cosine similarity with each cluster center is calculated. If the cosine similarity between the feature fusion vector of a new batch and all clusters of the corresponding stage is lower than the preset threshold, the feature fusion vector of the new batch is added to the cluster set of the corresponding stage as a new cluster center. At the same time, the cluster with the lowest cosine similarity with the feature fusion vector of the new batch is deleted from the original cluster.

6. The control method for a cantilever vibrating robot for concrete vibration operation according to claim 1, characterized in that, The specific steps for generating a continuous job path are as follows: Obtain candidate vibration points in the three-dimensional digital mesh model for the current vibration stage, and mark each mesh node as either unvibrated or vibrated using the operation status label library of the robot control system. Initially, all nodes are unvibrated. The candidate vibration point set is sorted according to regional connectivity, with priority given to continuous nodes within the same connected region, and then other connected regions are covered in turn. Using the sorted candidate vibration points as vertices, an initial path is generated according to the principle of connecting adjacent points with straight lines. For turning nodes in the path, the angle between two adjacent path segments is calculated. If the angle is greater than the threshold θ, a transition point is inserted between the two nodes, and finally a complete continuous operation path is output.

7. The control method for a cantilever vibrating robot for concrete vibration operation according to claim 6, characterized in that, The steps for sorting the candidate vibration point set according to regional connectivity are as follows: Traverse all candidate vibration points, select one unmarked candidate vibration point as the starting point, and search its six spatially adjacent grid nodes. If the adjacent nodes are not marked, add the adjacent nodes to the same temporary set as the starting point. Then, take all points in the temporary set as the new starting point and repeat the above adjacency search and set merging operation until no new candidate vibration points are added to the temporary set. At this time, the temporary set is a complete connected region. Mark the region identifier of all candidate vibration points in the region and repeat the above process until all candidate vibration points are assigned to the corresponding connected regions. Calculate the center coordinates (Xc, Yc) of each connected region, that is, the average X coordinate of all points in the region is Xc and the average Y coordinate is Yc. Sort all connected regions in the order of "Xc from small to large → Yc from small to large" to form the region operation priority. If there are regions with the same Xc and Yc, then supplement them by sorting in the order of the number of candidate vibration points in the region from large to small. Starting from the center point of each connected region, sort the points according to the rule of clockwise spiral expansion: first mark the center point, then mark the 6 adjacent points of the center point in turn, then mark the adjacent points of the adjacent points, until all points in the region are marked.

8. The control method for a cantilever vibrating robot for concrete vibration operation according to claim 1, characterized in that, Adaptive adjustment of vibration depth and posture specifically includes: Set the vibration depth of the cantilever foundation according to the vibration stage: h1 for the initial vibration stage, h2 for the secondary vibration stage, and h3 for the exhaust vibration stage. Read the three-dimensional coordinates (X,Y,Z) of the current target point, subtract the foundation vibration depth from the Z-axis coordinate to obtain the initial vibration depth, where h3 < h2 < h1. Calculate the difference ΔZ between the Z-axis coordinate of the current target point and the average Z-axis coordinate of the grid nodes in the area. If ΔZ is positive, decrease the initial vibration depth by ΔZ. If ΔZ is negative, increase the initial vibration depth by |ΔZ|. Bind the current stage to the cantilever attitude angle: set the attitude angle to 0° for the initial vibration stage, α° for the re-vibration stage, and β° for the vibration discharge stage. Read the Z-axis coordinates of the three adjacent grid nodes around the target point, calculate the slope of the area, and if the slope is greater than the threshold, shift the attitude angle by the same angle in the slope direction.

Images