An intelligent mechanical arm path planning method based on data processing

Abstract

The present application relates to the technical field of data processing, more particularly, the present application relates to a kind of intelligent mechanical arm path planning method based on data processing, method includes: according to the point cloud data distribution in cluster interior obtains the structure feature of each cluster;According to the difference between the structure feature of cluster in one moment and its previous moment, to determine the matching cluster of the cluster;According to the velocity vector of cluster and the drift index between cluster and its matching cluster, obtain the tracking confidence of the cluster;The safety radius corresponding to cluster is obtained, the safety radius corresponding to cluster and the tracking confidence of cluster are negatively correlated, and positively correlated with the size of cluster;Remove the safety area corresponding to each cluster from the movable space of mechanical arm, path planning is carried out in the remaining space, effectively improve the accuracy of mechanical arm path planning.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing. More particularly, the present application relates to a method for intelligent robot arm path planning based on data processing. BACKGROUND

[0002] In tunnel engineering construction, the installation of communication leakage cable is a key task. The traditional manual operation mode has many problems such as low efficiency, poor precision, environmental pollution, high-altitude operation safety risk and the like. Therefore, introducing intelligent construction robots carrying mechanical arms for automatic installation has become an important development direction to ensure tunnel operation safety and improve construction efficiency. In the operation process, the mechanical arm is required to be able to perceive and avoid static and dynamic obstacles in the environment in real time in the long and unstructured tunnel space, so as to ensure the safety and efficiency of the operation.

[0003] To realize environment perception, the prior art usually uses sensors such as laser radars to obtain three-dimensional point cloud data of the environment, and uses clustering algorithms to process the point cloud to identify obstacles. Among them, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm is often used to segment obstacles from original point cloud data because it can discover clusters of arbitrary shape and is not sensitive to noise.

[0004] However, the traditional DBSCAN algorithm is stateless when processing continuous point cloud data, that is, independent clustering analysis is performed on each frame of point cloud. This non-memory mode makes it impossible to establish a connection between the same obstacles in consecutive time frames, so it cannot effectively track dynamic targets such as construction personnel and equipment moving in the tunnel, and cannot predict their movement trends. This leads to the fact that the path planning of the mechanical arm can only make emergency avoidance when the obstacle is very close, which has safety hazards.

[0005] Therefore, how to accurately capture the movement trend of the obstacle in the clustering process to reduce safety hazards and improve the accuracy of path planning is a problem to be solved at present. SUMMARY

[0006] To solve the above technical problem of how to accurately capture the movement trend of the obstacle in the clustering process to reduce safety hazards and improve the accuracy of path planning, the present application proposes a method for intelligent robot arm path planning based on data processing, which comprises the following steps:

[0007] The point cloud data in the tunnel environment is clustered to obtain a plurality of clustering clusters, and the structural features of each clustering cluster are obtained according to the distribution of the point cloud data in the clustering cluster; the drift index between the clustering clusters is obtained according to the difference between the structural features of the clustering clusters at one moment and the previous moment; for any clustering cluster obtained at each moment, the clustering cluster with the minimum drift coefficient between the clustering cluster and the clustering cluster obtained at the previous moment is taken as the matching cluster of the clustering cluster; the tracking confidence of the clustering cluster is obtained according to the velocity vector of the clustering cluster and the drift index between the clustering cluster and the matching cluster thereof; the safety radius corresponding to the clustering cluster is obtained, and the safety radius corresponding to the clustering cluster is negatively correlated with the tracking confidence of the clustering cluster and positively correlated with the size of the clustering cluster; the safety area corresponding to each clustering cluster is removed from the movable space of the mechanical arm, and path planning is performed in the remaining space.

[0008] By matching the obstacle clustering clusters across time frames, the dynamic target in the tunnel can be accurately captured, the possibility of collision between the obstacles and the mechanical arm is reduced, and the accuracy of path planning is improved. In the process of planning the path, the drift index between the clustering clusters at adjacent moments is calculated to establish the correlation across time frames, and the tracking confidence of the clustering cluster is calculated based on this, so that the interference of the environment on the obstacle identification can be reduced. On this basis, the safety radius of the obstacle corresponding to the clustering cluster is adjusted based on the tracking confidence and the physical size of the clustering cluster, so that the mechanical arm can predict the running trend of the obstacle and timely adjust the moving path based on this, thereby effectively improving the accuracy of path planning and increasing the work efficiency of the mechanical arm.

[0009] According to the intelligent mechanical arm path planning method based on data processing provided by the application, the point cloud data in the tunnel environment is clustered to obtain a plurality of clustering clusters, and the method comprises the following steps: continuously collecting three-dimensional point cloud data of the tunnel environment for preprocessing to obtain point cloud data; the point cloud data is clustered by setting a neighborhood radius and a minimum sample number in the DBSCAN algorithm to obtain various clustering clusters, each clustering cluster containing a plurality of point cloud data, and the three-dimensional coordinate mean value of all point cloud data in the clustering cluster being taken as the geometric center of the clustering cluster.

[0010] According to the intelligent mechanical arm path planning method based on data processing provided by the application, the structural features of each clustering cluster are obtained according to the distribution of the point cloud data in the clustering cluster, and the method comprises the following steps: the structural features of each clustering cluster include the number of point cloud data in the clustering cluster, the volume of the minimum circumscribed cube of the point cloud data in the clustering cluster, and the Euclidean distance mean value of each point cloud data in the clustering cluster and the geometric center.

[0011] The application constructs the structural features of the clustering cluster by the number of point clouds, the circumscribed cube volume and the average distance in the point, so that when subsequent obstacles are matched across time frames based on this, the distribution information such as the position, size and shape of the obstacles can be comprehensively matched, thereby effectively improving the dynamic correlation accuracy of the obstacles in the complex tunnel situation.

[0012] According to the intelligent mechanical arm path planning method based on data processing provided by the application, the difference between the structural features of the clustering cluster at a moment and the previous moment is obtained, and the drift index between the clustering clusters is obtained, including: obtaining the Euclidean distance between the geometric centers of any two clustering clusters at the moment and the previous moment as a distance index; subtracting the cosine similarity between the structural features of the two clustering clusters by 1 to obtain the difference index between the clustering clusters; and the product of the difference index and the distance index is normalized, and the value is recorded as the drift index between the clustering clusters.

[0013] The application provides an accurate drift index calculation method, which combines the Euclidean distance representing the physical displacement and the cosine similarity representing the shape change to obtain a comprehensive index that can measure the position and shape stability of the obstacle, so that the real object movement and the apparent change caused by noise or occlusion can be accurately distinguished based on this.

[0014] According to the intelligent mechanical arm path planning method based on data processing provided by the application, the speed vector of the clustering cluster is obtained, including: the difference between the geometric center of the clustering cluster and the corresponding matching cluster geometric center is recorded as the speed vector of the clustering cluster.

[0015] In order to associate the obstacles across time frames and obtain the motion information, the application obtains the possible moving direction of the clustering cluster through the difference between the geometric centers of the clustering cluster and the matching cluster, so that the mechanical arm can actively avoid obstacles by the pre-obtained moving direction of the clustering cluster when planning the path of the mechanical arm.

[0016] According to the intelligent mechanical arm path planning method based on data processing provided by the application, the tracking confidence of the clustering cluster is obtained according to the speed vector of the clustering cluster and the drift index between the clustering cluster and the matching cluster, including:

[0017] ;

[0018] 、 respectively as the tracking confidence, the speed vector, the influence coefficient, the drift index and the matching cluster of the i th clustering cluster collected at the moment ​​​​is the time point, a velocity vector of a matching cluster of the i-th cluster, is an exponential function with base e, is a cosine similarity function.

[0019] The method for path planning of an intelligent mechanical arm based on data processing provided by the application comprises the following steps: obtaining a safety radius corresponding to a cluster, wherein the safety radius of the cluster is the sum of a basic safety radius and an additional safety radius of the cluster; the ratio of the number of point cloud data contained in the cluster to the number of point cloud data contained in the maximum cluster is added by 1 to obtain a reference adjustment coefficient; the product of a preset basic radius reference value of the cluster and the reference adjustment coefficient is taken as the basic safety radius; the difference between 1 and the tracking confidence of the cluster is taken as an additional adjustment coefficient; and the product of a preset additional radius reference value of the cluster and the additional adjustment coefficient is taken as the additional safety radius.

[0020] The application considers that the safety radius of the mechanical arm cannot adapt to the fixed obstacle, and therefore, the safety radius of the obstacle is calculated by using the basic safety radius and the additional safety radius, wherein the basic part is proportional to the physical size of the obstacle, and the basic safety can be guaranteed; the additional part is adjusted according to the tracking confidence, and the uncertainty can be buffered, and the possibility of collision of the mechanical arm is effectively reduced.

[0021] The method for path planning of an intelligent mechanical arm based on data processing provided by the application comprises the following steps: removing the safety area corresponding to each cluster from the movable space of the mechanical arm, and performing path planning in the remaining space, wherein the total space range of the tunnel perceived by the mechanical arm is taken as the movable space, the safety area of each cluster is removed from the movable space to obtain the remaining space; the remaining space is taken as the path planning space of the mechanical arm, and the RRT algorithm is used to plan the moving path of the mechanical arm in the remaining space.

[0022] The method for path planning of an intelligent mechanical arm based on data processing provided by the application comprises the following steps: removing the safety area corresponding to each cluster from the movable space of the mechanical arm, and performing path planning in the remaining space, wherein the total space range of the tunnel perceived by the mechanical arm is taken as the movable space, the safety area of each cluster is removed from the movable space to obtain the remaining space; the remaining space is taken as the path planning space of the mechanical arm, and the RRT algorithm is used to plan the moving path of the mechanical arm in the remaining space.

[0023] The method for path planning of an intelligent mechanical arm based on data processing provided by the application comprises the following steps: removing the safety area corresponding to each cluster from the movable space of the mechanical arm, and performing path planning in the remaining space, wherein the total space range of the tunnel perceived by the mechanical arm is taken as the movable space, the safety area of each cluster is removed from the movable space to obtain the remaining space; the remaining space is taken as the path planning space of the mechanical arm, and the RRT algorithm is used to plan the moving path of the mechanical arm in the remaining space.

[0024] The application has the following beneficial effects:

[0025] Based on the technical scheme, the intelligent mechanical arm path planning method based on data processing provided by the application can accurately capture dynamic targets in the tunnel, reduce the possibility of collision between the obstacles and the mechanical arm, and improve the accuracy of path planning. In the process of planning the path, the application establishes a cross-time frame association by calculating the drift index between the clustering clusters at adjacent time points, and calculates the tracking confidence of the clustering cluster based on this, which can reduce the interference of the environment on obstacle identification. On this basis, the application adjusts the safety radius of the obstacle corresponding to the clustering cluster by the tracking confidence and the physical size of the clustering cluster, so that the mechanical arm can predict the running trend of the obstacle and adjust the moving path in time based on this, effectively improving the accuracy of path planning and increasing the work efficiency of the mechanical arm. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 A step flow chart of an intelligent mechanical arm path planning method based on data processing provided by the embodiment of the application is provided.

[0027] Figure 2 A mechanical arm path planning result schematic diagram from a perspective provided by the embodiment of the application is provided.

[0028] Figure 3 A mechanical arm path planning result schematic diagram from another perspective provided by the embodiment of the application is provided. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are part of the embodiments of the application, rather than all the embodiments of the application.

[0030] Please refer to Figure 1 , Figure 1 A step flow chart of an intelligent mechanical arm path planning method based on data processing provided by the embodiment of the application is provided. The method comprises the following steps:

[0031] S1: Obtain point cloud data of a tunnel environment.

[0032] For example, in the embodiment of the application, the point cloud data of the tunnel environment is obtained, including: continuously collecting three-dimensional point cloud data of the tunnel environment for preprocessing to obtain the point cloud data.

[0033] Specifically, the laser radar carried on the mechanical arm continuously collects three-dimensional point cloud data in the tunnel environment at each time according to a preset collection frequency, and then removes discrete noise and false points on the wall surface through Gaussian filtering to reduce the influence of noise on obstacle identification.

[0034] The key features such as the extraction of the tunnel axis, the wall normal, the pipeline center, and the simplified environment model are used for preprocessing, and finally, the point cloud data of the tunnel environment at each time is obtained, so as to prepare for subsequent data processing.

[0035] The acquisition frequency can be set to 30Hz, and the acquisition frequency and the preprocessing mode can be set according to actual needs.

[0036] According to the above steps, the point cloud data of the tunnel environment can be acquired, and the clustering clusters corresponding to the obstacles in the tunnel can be obtained by clustering the point cloud data.

[0037] However, the robot arm and the obstacle can be in a dynamic running state, so that the robot arm cannot dynamically plan the path according to the distance between the robot arm and the obstacle and the position of the obstacle, and finally affect the safety and accuracy of the robot arm operation.

[0038] Therefore, in the embodiment of the application, the similarity change between the point cloud data clustering clusters collected by the mobile robot arm at adjacent time points is used to obtain the clustering cluster with the highest similarity to the current time clustering cluster at the previous time as the matching cluster of the current clustering cluster for analysis, so as to accurately obtain the possible moving direction of the obstacle; when determining the safety radius of the clustering cluster, the basic safety radius and the additional safety radius of the reference are dynamically adjusted according to the change and size of the clustering cluster, so as to accurately plan the running path of the robot arm and reduce the possibility of collision, that is, the following steps are executed.

[0039] S2: clustering the point cloud data in the tunnel environment to obtain a plurality of clustering clusters, obtaining the structural features of each clustering cluster according to the distribution of the point cloud data inside the clustering cluster; obtaining the drift index between the clustering clusters according to the difference between the structural features of the clustering clusters at one time and the previous time.

[0040] For example, in the DBSCAN algorithm, the neighborhood radius and the minimum sample number are set to cluster the point cloud data, and each clustering cluster contains a plurality of point cloud data, and the three-dimensional coordinate mean of all point cloud data in the clustering cluster is taken as the geometric center of the clustering cluster.

[0041] The neighborhood radius and the minimum sample number can be set according to actual needs, for example, in the embodiment of the application, the neighborhood radius can be set to 0.5, and the minimum sample number can be set to 20. The specific steps of obtaining the clustering cluster of the point cloud data according to the DBSCAN algorithm can be realized by the prior art, and the embodiment of the application will not be repeated here.

[0042] It should be noted that, since the structure of the obstacle is in a relatively stable state in a short time, that is, the distribution of the size, spatial volume and position state of the point cloud data in the cluster corresponding to the obstacle in a short time is generally close.

[0043] Therefore, in order to stably identify the same obstacle in continuous time frames and better plan the path of the robot arm, the embodiment of the present application can obtain the structural features of the cluster as a whole according to the distribution state of the cluster, and analyze the state change of the obstacle corresponding to the cluster based on the change of the structural features.

[0044] For example, in the embodiment of the present application, the structural features of each cluster are obtained according to the distribution of the point cloud data in the cluster, including the number of point cloud data in each cluster, the volume of the minimum circumscribed cube of the point cloud data in the cluster, and the average Euclidean distance between each point cloud data and the geometric center in the cluster.

[0045] Specifically, when obtaining the volume of the minimum circumscribed cube of the point cloud data in the cluster, the maximum and minimum values of the point cloud data in the cluster on the X, Y and Z axes can be obtained first; the edge length of the cube in the X axis direction is determined according to the difference between the maximum and minimum values on the X axis, the edge length of the cube in the Y axis direction is determined according to the difference between the maximum and minimum values on the Y axis, and the edge length of the cube in the Z axis direction is determined according to the difference between the maximum and minimum values on the Z axis; and the volume of the minimum circumscribed cube of the point cloud data in the cluster is obtained according to the product of the edge lengths of the cube in the X, Y and Z axes.

[0046] It can be understood that the path planning of the robot arm is planned in advance, and therefore, when analyzing the cluster composed of the point cloud data at the current time, the time point one time before the current time can be taken as the initial state of the cluster, and the difference between the structural features of the cluster in the two adjacent time points from the initial state to the current state is analyzed.

[0047] Different clusters correspond to different obstacles, and therefore their sizes, shapes and densities are different. With the movement of the robot arm, the positions of the clusters corresponding to different obstacles relative to the robot arm will change, and therefore the cluster corresponding to the same obstacle at different time points may change. The higher the possibility that the two clusters in the two time points correspond to the same obstacle, the smaller the difference between their structural features due to the structural stability, and the closer the spatial distance, and therefore the smaller the corresponding drift index. Based on this, the possibility that the clusters in the two adjacent time points correspond to the same obstacle can be evaluated.

[0048] For example, in the embodiment of the present application, the drift index between the clustering clusters is obtained according to the difference between the structural features of the clustering clusters at one time and the previous time, including: obtaining the Euclidean distance between the geometric centers of any two clustering clusters at the time and the previous time as a distance index; subtracting the cosine similarity between the structural features of the two clustering clusters by 1 to obtain a difference index between the clustering clusters; and normalizing the product of the difference index and the distance index, denoted as the drift index between the clustering clusters.

[0049] As an example, the embodiment of the present application obtains the time The drift index between the i th clustering cluster collected at the time The drift index between the i th clustering cluster collected at the time The drift index between the i th clustering cluster collected at the time

[0050]

[0051] The drift index between the i th clustering cluster collected at the time The drift index between the i th clustering cluster collected at the time The drift index between the i th clustering cluster collected at the time The drift index between the i th clustering cluster collected at the time The drift index between the i th clustering cluster collected at the time The drift index between the i th clustering cluster collected at the time The drift index between the i th clustering cluster collected at the time The drift index between the i th clustering cluster collected at the time The drift index between the i th clustering cluster collected at the time The drift index between the i th clustering cluster collected at the time The drift index between the i th clustering cluster collected at the time The drift index between the i th clustering cluster collected at the time The drift index between the i th clustering cluster collected at the time

[0052] In the calculation method of the drift index, is the difference index, is the distance index.

[0053] The greater the distance index, the farther the distance between the clustering clusters at the two times, the smaller the possibility that the clustering clusters correspond to the same obstacle, and the greater the corresponding drift index; on the contrary, the smaller the distance index, the closer the distance between the clustering clusters at the two times, the smaller the possibility that the clustering clusters correspond to the same obstacle, and the smaller the corresponding drift index.

[0054] ​Since the distances between the cluster groups corresponding to different obstacles in the two time instants before and after the movement of the mechanical arm can be close, in order to reduce the influence of such obstacles on the drift degree, the embodiment of the present application introduces a difference index to correct the distance index. The greater the difference index, the smaller the similarity between the cluster group structure features in the two time instants before and after, and the more likely the two correspond to different obstacles, so the offset index needs to be increased to reduce the possibility of matching different obstacles in different time instants; on the contrary, the smaller the difference index, the greater the similarity between the cluster group structure features in the two time instants before and after, and the more likely the two correspond to the same obstacle, so the offset index needs to be reduced to increase the possibility of matching the same obstacle in different time instants.

[0055] According to the above calculation method, the drift index between all cluster groups in the time instant and the time instant can be obtained.

[0056] After obtaining the drift index between the cluster groups in the current time instant and the cluster groups in the previous time instant based on the above steps, the cluster group with the closest similarity to each cluster group in the current time instant can be obtained from the cluster groups in the previous time instant.

[0057] In order to facilitate understanding, the embodiment of the present application provides a set of cluster group data according to the two time instants before and after, the drift index between all cluster group data in the two time instants calculated and the matching cluster of each current time instant cluster group, which can be specifically referred to in Table 1 as follows:

[0058] Table 1

[0059]

[0060] The values in the above Table 1 are the drift index corresponding between each current time instant cluster group and the cluster group in the previous time instant, for example, the drift index 0 of the current time instant cluster group 1 and the previous time instant cluster group 1 is the minimum value among the drift indexes 0, 0.0125, 0.3648, 0.3149 between the current time instant cluster group 1 and each cluster group in the previous time instant, so the cluster group 1 in the previous time instant is the matching cluster of the current time instant cluster group 1.

[0061] The dynamic association of the current time instant and the previous time instant makes it possible to realize the cross-frame dynamic tracking of obstacles based on this, and improve the obstacle avoidance accuracy.

[0062] S3: For any cluster group obtained in each time instant, the cluster group with the smallest drift coefficient between the cluster group obtained in the previous time instant is taken as the matching cluster of the cluster group; according to the velocity vector of the cluster group and the drift index between the cluster group and its matching cluster, the tracking confidence of the cluster group is obtained.

[0063] It should be noted that during the movement of the mechanical arm, the obstacles in the static state and the moving state will all have relative movement with the mechanical arm, in order to reduce the possibility of collision between the mechanical arm and the obstacles, the running trend of each obstacle geometric center can be obtained according to the change of the geometric center of each obstacle, and the trajectory of the mechanical arm can be adjusted in time to avoid obstacles before the mechanical arm approaches the obstacles, so that the possibility of collision is reduced.

[0064] For example, in the embodiment of the application, the speed vector of the cluster is obtained in the following manner: the difference between the geometric center of the cluster and the corresponding matching cluster geometric center is denoted as the speed vector of the cluster.

[0065] Since the geometric center is a coordinate, the speed vector obtained according to the difference between the geometric center of the cluster and the corresponding matching cluster geometric center is a vector from the matching cluster of the cluster to the cluster, that is, the speed vector represents the possible moving direction of the obstacle corresponding to the cluster.

[0066] It should be further explained that due to the complex environment in the tunnel scene, during the relative movement between the mechanical arm and the obstacles, some obstacles can be temporarily blocked, and the distance between the obstacles with similar shapes can be relatively close, so that when the matching cluster of each cluster is obtained based on the above steps, the position tracking of the continuous time frames of the obstacles can be deviated, which affects the accuracy of the subsequent mechanical arm path planning.

[0067] However, the motion of the same obstacle in adjacent time is continuous, so the similarity of the speed vectors of the cluster and its matching cluster in adjacent time is high, and if the clusters of different obstacles are associated, the speed vectors of the two will have a large difference.

[0068] Therefore, in order to reduce the influence of the blocking of some obstacles and the close distance between the obstacles with similar shapes on the path planning, the drift index between the cluster and its matching cluster and the similarity between the speed vectors can be combined to measure the credibility of the obstacle tracking in the embodiment of the application.

[0069] It can be understood that the matching cluster of the current cluster in the previous time is also a cluster, so the matching cluster can be recorded as a target cluster, and the speed vector of the target cluster can be obtained by the change of the geometric centers of the target cluster in the previous time and the matching cluster of the target cluster in the previous time in the above steps.

[0070] For example, in the embodiment of the application, the tracking confidence of the cluster is obtained according to the speed vector of the cluster and the drift index between the cluster and its matching cluster, comprising:

[0071] ;

[0072] is a time point, is a tracking confidence of the i-th cluster collected, is an influence coefficient, is a time point, is a drift index between the i-th cluster collected and its matched cluster, is a time point, is a velocity vector of the i-th cluster collected, is a time point, is a velocity vector of the matched cluster of the i-th cluster collected, is an exponential function with base e, is a cosine similarity function.

[0073] The influence coefficient is used to adjust the degree of influence of the tracking confidence by the drift index. As an example, the influence coefficient can be set to 3 in the embodiment of the present application, and can be set according to actual needs.

[0074] In the calculation method of the tracking confidence, is a cosine similarity value of the velocity vector.

[0075] The closer the tracking confidence is to 1, the smaller the drift index and the higher the similarity of the velocity vector, the higher the possibility that the current cluster corresponds to the same obstacle as its matched cluster, and the higher the credibility of the tracking result.

[0076] The closer the tracking confidence is to 0, the larger the drift index and the lower the similarity of the velocity vector, the lower the possibility that the current cluster corresponds to the same obstacle as its matched cluster, and the lower the credibility of the tracking result.

[0077] The closer the tracking confidence is to 0, the smaller the drift index but the velocity vector has a difference, or the velocity vector is relatively similar but the drift index is relatively large, which may correspond to an obstacle being temporarily blocked to cause point cloud data deviation, so that the credibility of the tracking result is uncertain.

[0078] According to the above calculation method, the tracking confidence of all clusters at the current time point can be obtained. In order to facilitate understanding, the present embodiment provides a set of cluster data of two time points before and after, and calculates the tracking confidence of the cluster according to the drift index and the similarity of the velocity vector between the cluster and its matched cluster in the two time points before and after, which can be seen from Table 2 below: Table 2

[0079]

[0080]

[0081] ​The tracking confidence of the cluster can be calculated according to the drift index and the speed vector similarity between the cluster and the matched cluster. As can be seen from Table 2, the tracking confidence is higher only when the drift index is smaller and the speed vector cosine similarity value is higher.

[0082] S4: Obtain the safety radius corresponding to the cluster; remove the safety area corresponding to each cluster from the movable space of the robot arm, and perform path planning in the remaining space.

[0083] The safety radius corresponding to the cluster is negatively correlated with the tracking confidence of the cluster and positively correlated with the size of the cluster.

[0084] It should be noted that the tracking confidence of each cluster can be obtained based on the above steps. Due to the complex environment in the tunnel scene, if the position sensing of the object is not accurate during the movement of the robot arm, a collision between the robot arm and the obstacle may occur. The safety radius of the obstacle is the minimum distance between the robot arm and the obstacle, which consists of a basic safety radius and an additional safety radius.

[0085] Therefore, in order to avoid collision between the robot arm and the obstacle during movement, the basic radius reference value and the additional radius reference value can be adjusted according to the tracking confidence and the size of the cluster, so as to accurately obtain the safety radius of the obstacle corresponding to each cluster.

[0086] The basic radius reference value and the additional radius reference value are set according to the physical parameters of the obstacle itself.

[0087] As an example, the basic radius reference value of the obstacle can be set to 0.5, and the additional radius reference value can be set to 0.4, which can be set according to actual needs.

[0088] For example, in the embodiment of the application, obtaining the safety radius corresponding to the cluster comprises: taking the sum of the basic safety radius and the additional safety radius of the cluster as the safety radius of the cluster, wherein the ratio of the number of point cloud data contained in the cluster to the number of point cloud data contained in the maximum cluster is added by 1 to obtain a reference adjustment coefficient; the product of the preset basic radius reference value of the cluster and the reference adjustment coefficient is taken as the basic safety radius; the difference between 1 and the tracking confidence of the cluster is taken as an additional adjustment coefficient, and the product of the preset additional radius reference value of the cluster and the additional adjustment coefficient is taken as the additional safety radius.

[0089] In the safety radius calculation method of the cluster, the ratio of the number of point cloud data contained in the cluster to the number of point cloud data contained in the maximum cluster represents the relative size of the cluster. The more the number of point cloud data in the cluster, the larger the range occupied by the obstacle corresponding to the cluster in the space.

[0090] Therefore, the embodiment of the present application can adjust the basic safety radius of the cluster by the ratio to obtain a basic safety radius as a bottom guarantee safety distance, and increase the safety radius of the obstacle with a high physical size ratio, so that the safety radius of the obstacle can match the physical size, and avoid the space of the obstacle with a large physical size from being included in the moving path of the robot arm.

[0091] The additional safety radius is a buffer radius in the safety radius, and when there is uncertainty in obstacle tracking, the additional safety distance can compensate for the perception deviation. When the tracking confidence is close to 1, the additional adjustment coefficient plays a smaller role, and the safety radius is determined by the basic radius reference value and the volume adjustment, so that the robot arm can plan a path in a larger space and improve the work efficiency. On the contrary, when the tracking confidence is low, the additional safety radius takes effect, and the possibility of collision caused by tracking error can be reduced by conservative obstacle avoidance.

[0092] In this way, the embodiment of the present application can keep enough safety space for the robot arm and the obstacle in the complex tunnel environment by the bottom guarantee safety distance of the basic safety radius and the uncertainty buffer of the additional safety radius, and reduce the risk of collision.

[0093] For example, in the embodiment of the present application, the safety area corresponding to the cluster is obtained in the following manner: a spherical area with the geometric center of the cluster as the center and the safety radius corresponding to the cluster as the radius is taken as the safety area corresponding to the cluster.

[0094] For example, in the embodiment of the present application, the safety area corresponding to each cluster is removed from the movable space of the robot arm, and the path is planned in the remaining space, including: taking the whole space range of the tunnel perceived by the robot arm as the movable space, removing the safety area of each cluster from the movable space to obtain the remaining space; taking the remaining space as the path planning space of the robot arm, and using the RRT algorithm to plan the moving path of the robot arm in the remaining space.

[0095] Wherein, the step of planning the moving path of the robot arm according to the RRT algorithm can be realized by the prior art, and the embodiment of the present application will not be repeated here.

[0096] Figure 2 A schematic diagram of the path planning result of the robot arm from the perspective provided by the embodiment of the present application is shown in the following figure, Figure 3 A schematic diagram of the path planning result of the robot arm from another perspective provided by the embodiment of the present application is shown in the following figure, Figure 2 And Figure 3In the figure, X represents the X-axis, Y represents the Y-axis, Z represents the Z-axis, the red line segment represents the mechanical arm movement path, the green dot represents the starting point of the movement path, that is, the current position of the mechanical arm, the red dot represents the end point of the movement path, that is, the position to which the mechanical arm finally reaches, and the mesh aggregation area is the safety area of the obstacle corresponding to the clustering cluster.

[0097] In combination Figure 2 And Figure 3 It can be seen that the number of clustering clusters collected at the current moment is 4, that is, the number of obstacles is 4, and the movement path planned for the mechanical arm avoids the safety areas of all obstacles, effectively reducing the possibility of collision between the mechanical arm and the obstacles.

[0098] According to the above steps, the path planning of the mechanical arm can be realized, and the path planning is updated by collecting the point cloud data of the tunnel in real time, so that the mechanical arm can timely adapt to the complex situation in the tunnel. However, during the operation of the mechanical arm, the mechanical arm may not operate according to the planned path due to installation problems. In order to ensure the accuracy of the operation of the mechanical arm, the actual operation route of the mechanical arm and the planned path can also be matched and monitored in the embodiment of the application.

[0099] For example, in the embodiment of the application, after the path planning in the remaining space, the operation position of the mechanical arm is obtained according to the operation data of the mechanical arm, and the matching result of the operation position of the mechanical arm and the planned movement path is obtained. Abnormality detection is performed.

[0100] The specific steps of monitoring the abnormality of the mechanical arm can be realized by the prior art, which will not be repeated here.

[0101] It can be seen that in the embodiment of the application, when the data of the mechanical arm is processed for path planning, the point cloud data in the tunnel environment can be clustered to obtain a plurality of clustering clusters, the structural features of each clustering cluster are obtained according to the distribution of the point cloud data inside the clustering cluster; the drift index between the clustering clusters is obtained according to the difference between the structural features of the clustering clusters at one moment and the previous moment; for any clustering cluster obtained at each moment, the clustering cluster with the smallest drift coefficient between the clustering cluster and the matching cluster of the clustering cluster obtained at the previous moment is taken as the matching cluster of the clustering cluster; the tracking confidence of the clustering cluster is obtained according to the velocity vector of the clustering cluster and the drift index between the clustering cluster and its matching cluster; the safety radius corresponding to the clustering cluster is obtained, and the safety radius corresponding to the clustering cluster is negatively correlated with the tracking confidence of the clustering cluster and positively correlated with the size of the clustering cluster; the safety areas corresponding to the clustering clusters are removed from the movable space of the mechanical arm, and the path planning is performed in the remaining space, effectively improving the accuracy of the path planning of the mechanical arm.

[0102] The above merely provides the preferred embodiment of the present application, and is not used to limit the present application, and any modification, equivalent replacement, improvement, etc. within the principle of the present application should be included in the protection scope of the present application.

Claims

1. A data processing based intelligent robotic arm path planning method, characterized in that, include: Multiple clusters are obtained by clustering point cloud data in the tunnel environment. This includes: continuously collecting and preprocessing 3D point cloud data from the tunnel environment to obtain point cloud data; using the DBSCAN algorithm to set the neighborhood radius and minimum number of samples to cluster the point cloud data, resulting in clusters containing multiple point cloud data. The mean of the 3D coordinates of all point cloud data in a cluster is used as the geometric center of that cluster; and the structural features of each cluster are obtained based on the distribution of point cloud data within the cluster. The structural features of each cluster include the number of point cloud data in the cluster, the volume of the minimum bounding cube of the point cloud data in the cluster, and the mean Euclidean distance between each point cloud data in the cluster and the geometric center. Based on the difference between the structural features of clusters at a given time and those at the previous time, the drift index between clusters is obtained, including: obtaining the Euclidean distance between the geometric centers of any two clusters at the given time and those at the previous time as a distance index; subtracting the cosine similarity between the structural features of the two clusters from 1 to obtain the difference index between clusters; normalizing the product of the difference index and the distance index, and recording it as the drift index between the clusters; for any cluster obtained at each time, the cluster with the smallest drift coefficient among the clusters obtained at the previous time is taken as the matching cluster of the cluster. Based on the velocity vector of the cluster and the drift index between the cluster and its matching cluster, the tracking confidence of the cluster is obtained, including: ; , They are time points The tracking confidence and velocity vector of the i-th cluster are collected. The influence coefficient, For a moment The drift index between the i-th cluster and its matching cluster. For a moment The velocity vector of the matching cluster of the i-th cluster is collected. It is an exponential function with base e. Let be the cosine similarity function; the difference between the geometric center of a cluster and the geometric center of its corresponding matching cluster is denoted as the velocity vector of that cluster. Obtain the safety radius corresponding to the cluster. The safety radius corresponding to the cluster is negatively correlated with the tracking confidence of the cluster and positively correlated with the size of the cluster. Remove the safe zones corresponding to each cluster from the movable space of the robotic arm, and perform path planning in the remaining space.

2. The intelligent robotic arm path planning method based on data processing according to claim 1, wherein, The step of obtaining the safe radius corresponding to the cluster includes: The sum of the basic safety radius and the additional safety radius of the cluster is used as the safety radius of the cluster. The ratio of the number of point cloud data contained in the cluster to the number of point cloud data contained in the largest cluster plus 1 is used as the baseline adjustment coefficient. The product of the preset basic radius baseline value of the cluster and the baseline adjustment coefficient is used as the basic safety radius. The difference between 1 and the tracking confidence of the cluster is used as the additional adjustment coefficient, and the product of the preset additional radius benchmark value of the cluster and the additional adjustment coefficient is used as the additional safety radius.

3. The intelligent robotic arm path planning method based on data processing according to claim 1, wherein, The step of removing the safe zones corresponding to each cluster from the movable space of the robotic arm and performing path planning in the remaining space includes: The entire tunnel space perceived by the robotic arm is taken as the movable space. The safe areas of each cluster are removed from the movable space to obtain the remaining space. The remaining space is used as the path planning space of the robotic arm, and the RRT algorithm is used to plan the movement path of the robotic arm in the remaining space.

4. The intelligent robotic arm path planning method based on data processing according to claim 2, wherein, Methods for obtaining the safe zone corresponding to a cluster include: The spherical region with the geometric center of the cluster as its center and the corresponding safety radius as its radius is defined as the safe region for that cluster.

5. The intelligent robotic arm path planning method based on data processing according to claim 1, wherein, The process of path planning within the remaining space also includes: The robot arm's operating position is obtained based on its operational data. Anomaly detection is then performed based on the matching result between the robot arm's operating position and the planned movement path.

Images