A control method and system for a large-torque robot collaborative arm

Abstract

The application provides a control method and system of a large-torque robot collaborative arm, and relates to the technical field of collaborative arm control. The method comprises the following steps: obtaining bedrock ground penetrating radar data of a target sampling area; generating a bedrock primary bedding development degree distribution map of the target sampling area based on the bedrock ground penetrating radar data of the target sampling area; determining a plurality of groups of drilling point information of a peripheral core reference sampling area and a plurality of groups of drilling point information of a primary bedding core sampling area based on the bedrock primary bedding development degree distribution map of the peripheral core reference sampling area and the bedrock primary bedding development degree distribution map of the primary bedding core sampling area; and controlling a robot collaborative arm to perform core drilling operation of the primary bedding core sampling area based on a core drilling collaborative arm control scheme of the primary bedding core sampling area. The method can efficiently and accurately determine a robot collaborative arm control scheme for core drilling of bedrock primary bedding in the target sampling area.

Description

Technical Field

[0001] This invention relates to the field of collaborative arm control technology, and specifically to a control method and system for a high-torque robotic collaborative arm. Background Technology

[0002] In the field of modern geological exploration and engineering drilling, core sampling of complex geological structures places extremely high demands on the control precision, high torque output stability, and operational adaptability of robotic collaborative arms. Traditional core drilling relies heavily on technicians to analyze and divide sampling areas and plan sampling points, which is highly subjective and lacks a systematic analysis of the degree of development of primary bedrock bedding, making it difficult to adapt to the complex conditions of heterogeneous bedrock. Control methods often employ fixed parameters to execute drilling tasks, failing to dynamically optimize and adjust subsequent collaborative arm operating parameters based on prior core drilling quality information. Furthermore, when dealing with the layout of multiple collaborative arm workstations and the matching of drilling torque, existing technologies do not fully explore the spatial and geological correlations of drilling points, making it difficult to effectively model and predict operational risks in complex drilling environments. These technical shortcomings directly lead to positioning deviations and core breakage in high-torque operating scenarios, making it difficult to meet the stringent requirements of operational stability, sampling integrity, and efficiency for high-torque robotic collaborative arms in primary bedding core sampling.

[0003] Therefore, how to efficiently and accurately determine the control scheme of the robotic collaborative arm for drilling bedrock cores in the target sampling area is an urgent problem to be solved. Summary of the Invention

[0004] The main technical problem solved by this invention is how to efficiently and accurately determine the target sampling area for drilling core samples from bedrock primary bedding.

[0005] According to a first aspect, the present invention provides a control method for a high-torque robotic collaborative arm, comprising: acquiring ground penetration radar data of bedrock in a target sampling area; generating a distribution map of the development degree of primary bedrock bedding in the target sampling area based on the ground penetration radar data of bedrock in the target sampling area; determining a peripheral core reference sampling area and a primary bedding core sampling area based on the distribution map of the development degree of primary bedrock bedding in the target sampling area; and determining multiple sets of drilling point information and primary bedding core sampling in the peripheral core reference sampling area based on the distribution map of the development degree of primary bedrock bedding in the peripheral core reference sampling area and the distribution map of the development degree of primary bedrock bedding in the primary bedding core sampling area. The system obtains multiple drilling point information for the outer core reference sampling area; determines a core drilling collaborative arm control scheme for the outer core reference sampling area based on the multiple drilling point information for the outer core reference sampling area; controls the robotic collaborative arm to perform core drilling operations based on the core drilling collaborative arm control scheme for the outer core reference sampling area and obtains core drilling quality information; determines a core drilling collaborative arm control scheme for the primary bedding core sampling area based on the core drilling quality information and the multiple drilling point information for the primary bedding core sampling area; and controls the robotic collaborative arm to perform core drilling operations in the primary bedding core sampling area based on the core drilling collaborative arm control scheme for the primary bedding core sampling area.

[0006] In one possible implementation, determining the core drilling cooperative arm control scheme for the peripheral core reference sampling area based on multiple sets of drilling point information of the peripheral core reference sampling area includes: clustering multiple sets of drilling point information of the peripheral core reference sampling area to obtain multiple clusters; determining multiple cooperative arm work stations and the drilling torque of each cooperative arm work station based on the multiple clusters; and generating the core drilling cooperative arm control scheme for the peripheral core reference sampling area based on the multiple cooperative arm work stations and the drilling torque of each cooperative arm work station.

[0007] In one possible implementation, the step of determining the core drilling cooperative arm control scheme for the primary bedding core sampling area based on the core drilling quality information and multiple sets of drilling point information of the primary bedding core sampling area includes: constructing a drilling point feature map, wherein the drilling point feature map includes multiple drilling point nodes and multiple edges between the drilling point nodes, and the node features of each drilling point node are drilling point information; processing the drilling point feature map based on a graph neural network to determine multiple cooperative arm work stations in the primary bedding core sampling area; and determining the core drilling cooperative arm control scheme based on the core drilling quality information of the peripheral core reference sampling area and the core drilling quality information of the primary bedding core sampling area. Based on multiple drilling point information in the primary bedding core sampling area, simulated core drilling quality information under different drilling torques is generated. Based on this simulated core drilling quality information under different drilling torques and multiple drilling point information corresponding to each cooperative arm work station in the primary bedding core sampling area, the target drilling torque for each cooperative arm work station in the primary bedding core sampling area is determined. Based on these multiple cooperative arm work stations and the target drilling torque for each cooperative arm work station in the primary bedding core sampling area, a core drilling cooperative arm control scheme for the primary bedding core sampling area is generated.

[0008] In one possible implementation, the input of the graph neural network is the feature map of the drilling point, and the output of the graph neural network is multiple cooperative arm work stations in the original bedding core sampling area.

[0009] According to a second aspect, the present invention provides a control system for a high-torque robotic collaborative arm, comprising: a data acquisition module for acquiring bedrock ground penetration radar data of a target sampling area; a bedding distribution map generation module for generating a distribution map of the development degree of primary bedding in the bedrock of the target sampling area based on the bedrock ground penetration radar data of the target sampling area; a sampling area division module for determining an outer core reference sampling area and a primary bedding core sampling area based on the distribution map of the development degree of primary bedding in the bedrock of the target sampling area; and a drilling point determination module for determining multiple sets of drilling point information of the outer core reference sampling area and multiple drilling point information of the primary bedding core sampling area based on the distribution map of the development degree of primary bedding in the bedrock of the outer core reference sampling area and the distribution map of the development degree of primary bedding in the bedrock of the primary bedding core sampling area. The system comprises: a group of drilling point information; a first control scheme generation module, used to determine a core drilling collaborative arm control scheme for the outer core reference sampling area based on multiple groups of drilling point information; a first drilling execution module, used to control a robotic collaborative arm to perform core drilling operations and obtain core drilling quality information based on the core drilling quality information and multiple groups of drilling point information in the primary bedding core sampling area; and a second drilling execution module, used to control a robotic collaborative arm to perform core drilling operations in the primary bedding core sampling area based on the core drilling quality information and multiple groups of drilling point information in the primary bedding core sampling area.

[0010] In one possible implementation, the first control scheme generation module is further configured to: cluster multiple clusters based on multiple sets of drilling point information of the peripheral core reference sampling area; determine multiple cooperative arm workstation locations and the drilling torque of each cooperative arm workstation location based on the multiple clusters; and generate a core drilling cooperative arm control scheme for the peripheral core reference sampling area based on the multiple cooperative arm workstation locations and the drilling torque of each cooperative arm workstation location.

[0011] In one possible implementation, the second control scheme generation module is further configured to: construct a drilling point feature map, wherein the drilling point feature map includes multiple drilling point nodes and multiple edges between the drilling point nodes, and the node features of each drilling point node are drilling point information; process the drilling point feature map based on a graph neural network to determine multiple cooperative arm operation stations in the original bedding core sampling area; and generate a source code based on the core drilling cooperative arm control scheme of the outer core reference sampling area, the core drilling quality information, and multiple sets of drilling point information in the original bedding core sampling area. Simulated core drilling quality information under different drilling torques in the primary bedding core sampling area; based on the simulated core drilling quality information under different drilling torques in the primary bedding core sampling area and the multiple drilling point information corresponding to each cooperative arm work station in the primary bedding core sampling area, the target drilling torque for each cooperative arm work station in the primary bedding core sampling area is determined; based on the multiple cooperative arm work stations in the primary bedding core sampling area and the target drilling torque for each cooperative arm work station in the primary bedding core sampling area, a core drilling cooperative arm control scheme for the primary bedding core sampling area is generated.

[0012] In one possible implementation, the input of the graph neural network is the feature map of the drilling point, and the output of the graph neural network is multiple cooperative arm work stations in the original bedding core sampling area.

[0013] According to a third aspect, embodiments of the present invention provide an electronic device, including: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method as described above, the method including: acquiring ground-penetrating radar data of bedrock in a target sampling area; generating a distribution map of the degree of development of primary bedrock bedding in the target sampling area based on the ground-penetrating radar data of bedrock in the target sampling area; determining a peripheral core reference sampling area and a primary bedding core sampling area based on the distribution map of the degree of development of primary bedrock bedding in the target sampling area; and determining the distribution map of the degree of development of primary bedrock bedding in the peripheral core reference sampling area and the distribution map of the degree of development of primary bedrock bedding in the primary bedding core sampling area. The process involves: determining multiple sets of drilling point information for the outer core reference sampling area and multiple sets of drilling point information for the primary bedding core sampling area; determining a core drilling collaborative arm control scheme for the outer core reference sampling area based on the multiple sets of drilling point information for the outer core reference sampling area; controlling the robotic collaborative arm to perform core drilling operations and obtain core drilling quality information based on the core drilling quality information and the multiple sets of drilling point information for the primary bedding core sampling area; and controlling the robotic collaborative arm to perform core drilling operations in the primary bedding core sampling area based on the core drilling collaborative arm control scheme for the primary bedding core sampling area.

[0014] According to the fourth aspect, this embodiment provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the program implements the aforementioned control method for a high-torque robotic collaborative arm. The method includes: acquiring ground-penetrating radar data of a target sampling area; generating a distribution map of the development degree of primary bedding in the bedrock of the target sampling area based on the ground-penetrating radar data of the bedrock of the target sampling area; determining a peripheral core reference sampling area and a primary bedding core sampling area based on the distribution map of the development degree of primary bedding in the bedrock of the target sampling area; and determining a peripheral core reference sampling area and a primary bedding core sampling area based on the distribution map of the development degree of primary bedding in the bedrock of the peripheral core reference sampling area and the distribution map of the development degree of primary bedding in the bedrock of the primary bedding core sampling area. Based on multiple sets of drilling point information in the sampling area and multiple sets of drilling point information in the primary bedding core sampling area, a core drilling collaborative arm control scheme for the outer core reference sampling area is determined. Based on the core drilling collaborative arm control scheme for the outer core reference sampling area, a robotic collaborative arm is controlled to perform core drilling operations and obtain core drilling quality information. Based on the core drilling quality information and multiple sets of drilling point information in the primary bedding core sampling area, a core drilling collaborative arm control scheme for the primary bedding core sampling area is determined. Based on the core drilling collaborative arm control scheme for the primary bedding core sampling area, a robotic collaborative arm is controlled to perform core drilling operations in the primary bedding core sampling area.

[0015] This invention provides a control method and system for a high-torque robotic collaborative arm. The method includes: acquiring bedrock surface penetration radar data of a target sampling area; generating a distribution map of the primary bedding development degree of the bedrock in the target sampling area based on the bedrock surface penetration radar data; determining a peripheral core reference sampling area and a primary bedding core sampling area based on the distribution map of the primary bedding development degree of the bedrock in the target sampling area; determining multiple sets of drilling point information for the peripheral core reference sampling area and the primary bedding core sampling area based on the distribution map of the primary bedding development degree of the bedrock in the peripheral core reference sampling area and the primary bedding core sampling area; and determining multiple sets of drilling point information for the primary bedding core sampling area based on the distribution map of the primary bedding development degree of the bedrock in the peripheral core reference sampling area and the primary bedding core sampling area. The method involves determining a core drilling collaborative arm control scheme for the outer core reference sampling area based on multiple sets of drilling point information in the sampling area; controlling the robotic collaborative arm to perform core drilling operations and obtain core drilling quality information based on the core drilling quality information and multiple sets of drilling point information in the primary bedding core sampling area; and controlling the robotic collaborative arm to perform core drilling operations in the primary bedding core sampling area based on the core drilling collaborative arm control scheme in the primary bedding core sampling area. This method can efficiently and accurately determine the robotic collaborative arm control scheme for drilling primary bedding cores in bedrock of the target sampling area. Attached Figure Description

[0016] Figure 1 A flowchart illustrating a control method for a high-torque robotic collaborative arm provided in an embodiment of the present invention;

[0017] Figure 2 A schematic flowchart illustrating a core drilling cooperative arm control scheme for determining the peripheral core reference sampling area, provided in an embodiment of the present invention;

[0018] Figure 3 A schematic diagram of a robotic collaborative arm provided in an embodiment of the present invention;

[0019] Figure 4 A schematic flowchart illustrating a core drilling cooperation arm control scheme for determining the primary bedding core sampling area, provided in an embodiment of the present invention;

[0020] Figure 5 This is a schematic diagram of a control system for a high-torque robotic collaborative arm provided in an embodiment of the present invention. Detailed Implementation

[0021] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of the invention. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to the present invention are not shown or described in the specification. This is to avoid obscuring the core parts of the invention with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0022] In this embodiment of the invention, the following are provided: Figure 1 The method for controlling a high-torque robotic collaborative arm, as shown, includes steps S1 to S8:

[0023] Step S1: Obtain ground penetration radar data of the bedrock in the target sampling area.

[0024] Ground-penetrating radar (GPR) data of the bedrock in the target sampling area is obtained by using ground-penetrating radar (GPR) equipment to detect electromagnetic wave reflection signals in the bedrock of the target sampling area.

[0025] The bedrock ground penetration radar data of the target sampling area includes the dielectric constant difference data of the underground medium in the target sampling area, the reflection amplitude data of radar waves at the bedrock interface at different depths, and the two-way travel time data of radar waves.

[0026] Ground-penetrating radar data of the bedrock in the target sampling area can reflect the internal structural characteristics, stratification distribution, and potential fracture development of the bedrock in the target sampling area.

[0027] Step S2: Generate a distribution map of the degree of development of primary bedrock bedding in the target sampling area based on the ground penetration radar data of the bedrock in the target sampling area.

[0028] In some embodiments, a primary bedding analysis model can be used to generate a distribution map of the primary bedding development degree of the bedrock in the target sampling area. The primary bedding analysis model is a deep neural network model. The input to the primary bedding analysis model is ground-penetrating radar data of the bedrock in the target sampling area, and the output of the primary bedding analysis model is a distribution map of the primary bedding development degree of the bedrock in the target sampling area.

[0029] Deep neural network models include deep neural networks (DNNs). A deep neural network is a machine learning model with multiple layers of nonlinear transformation units. By simulating the connection patterns of neurons in the human brain, deep neural network models utilize multiple hidden layers to extract and abstract features from input data layer by layer. Deep neural networks can continuously adjust network weights through the backpropagation algorithm, thereby establishing a complex mapping relationship between input data and output results. Deep neural networks possess powerful nonlinear fitting capabilities, enabling them to uncover deep-seated underlying patterns from high-dimensional data.

[0030] The distribution map of the primary bedding development of the bedrock in the target sampling area is a three-dimensional mapping distribution map of the density and development level of bedrock bedding structures at different spatial locations within the target sampling area, output by the primary bedding analysis model.

[0031] The distribution map of the development level of primary bedrock bedding in the target sampling area uses the numerical value of voxel blocks to quantify the development level of primary bedding, and can intuitively show the continuous changes in dense and sparse bedding areas and bedding orientation.

[0032] The ground penetration radar data of the bedrock in the target sampling area contains rich electromagnetic wave echo characteristics. Due to the presence of the original bedrock bedding, the dielectric constant will undergo a slight abrupt change, which will produce specific radar wave reflection modes, such as the continuity of the reflection phase axis, the intensity of the amplitude, and the degree of waveform distortion. These characteristics have a close physical correspondence with the degree of bedding development.

[0033] Deep neural networks (DNNs) can perform in-depth analysis of ground-penetrating radar (GPR) data on bedrock in a target sampling area through nonlinear computation of multiple layers of neurons. The model first extracts fundamental time-frequency features of the radar data in the shallow network, such as instantaneous frequency, instantaneous amplitude, and phase continuity indices. Then, in the deep network, the model fuses and abstracts these fundamental features to construct a nonlinear mapping relationship between radar signal characteristics and geological bedding structures. The DNN can identify the characteristic textures representing bedding interfaces in radar profiles and quantify the bedding development index by analyzing the density and discontinuity of phase axes. By learning from a large amount of known GPR data, the DNN can grasp the electromagnetic response patterns under different bedding development conditions, thus enabling point-by-point calculation of the bedding development degree at each location within the target sampling area. Finally, the model maps the calculated development degree values ​​to corresponding spatial coordinates, and then generates a continuous distribution map of the primary bedrock bedding development degree of the target sampling area through interpolation and smoothing.

[0034] Step S3: Determine the outer core reference sampling area and the primary bedding core sampling area based on the distribution map of the bedrock primary bedding development degree of the target sampling area.

[0035] In some embodiments, a sampling partitioning model can be used to determine the peripheral core reference sampling area and the primary bedding core sampling area. The sampling partitioning model is a convolutional neural network model. The input to the sampling partitioning model is a distribution map of the development degree of primary bedding in the bedrock of the target sampling area, and the output of the sampling partitioning model is the peripheral core reference sampling area and the primary bedding core sampling area.

[0036] Convolutional Neural Network (CNN) models are a type of artificial neural network capable of processing data with a grid-like structure. A CNN consists of convolutional layers, pooling layers, and fully connected layers. Convolutional layers use kernels to perform sliding operations on the input data to extract local features. Pooling layers reduce data dimensionality while preserving key information, and fully connected layers map the extracted features to the final output category. CNNs possess translation invariance and local awareness capabilities, and can efficiently extract spatial hierarchical features from image data.

[0037] The peripheral core reference sampling area is a region delineated within the target sampling area by a sampling division model. It is used for preliminary drilling and reference sampling to obtain basic physical and mechanical parameters of the bedrock and drilling adaptability data.

[0038] The outer core reference sampling area is a geological block located at the edge of the target sampling area and with a relatively stable and regular bedding structure.

[0039] The primary bedding core sampling area is the core operational area for obtaining primary bedding cores, which is defined in the target sampling area by a sampling division model.

[0040] The primary bedding core sampling area is the central working block within the target sampling area, characterized by the most significant bedding features, core research value, and relatively complex structure. The primary bedding core sampling area is the final target range for sampling performed by the robotic collaborative arm.

[0041] The distribution map of the primary bedding development of bedrock in the target sampling area can visually present different texture patterns and gray-scale gradients. Among them, densely bedding areas can be characterized by complex textures and dramatic changes in numerical gradients, while sparsely bedding areas are characterized by smooth textures and uniform numerical distribution.

[0042] Convolutional neural networks (CNNs) use convolutional layers to scan the spatial features of the distribution map of the primary bedding development of bedrock in the target sampling area, extracting texture information and gradient features from different regions. Through pooling and classification mechanisms within the model, CNNs can identify stable regions that meet reference conditions and feature regions that meet sampling conditions, thereby accurately defining and outputting the outer core reference sampling area and the primary bedding core sampling area in the spatial coordinate system.

[0043] In some embodiments, determining the peripheral core reference sampling area and the primary bedding core sampling area based on the distribution map of the bedrock primary bedding development degree of the target sampling area includes steps S31 to S33:

[0044] Step S31: Based on the distribution map of the primary bedding development degree of the bedrock in the target sampling area, determine the distribution data of bedding density, bedding dip angle sequence, and rock hardness distribution data of the target sampling area.

[0045] In some embodiments, deep neural networks can be used to determine the distribution data of bedding density, bedding dip angle sequence, and rock hardness distribution data of the target sampling area.

[0046] The distribution data of bedding density in the target sampling area is a dataset of the number and spacing of bedding interfaces per unit volume within the target sampling area, output by a deep neural network.

[0047] The bedding tilt angle sequence is a sequence of data consisting of the angle values ​​between the bedding planes and the horizontal plane at various locations within the target sampling area, arranged in spatial order, output by a deep neural network.

[0048] Rock hardness distribution data is the distribution data of physical indicators of the ability of rocks at different spatial locations within a target sampling area to resist external mechanical intrusion, output by a deep neural network.

[0049] Deep neural networks analyze the distribution map of primary bedding development in the target sampling area block by block. The deep neural network tracks the extension direction of the bedding texture in the distribution map and generates a sequence of bedding dip angles by calculating the angle between the texture tangent and the horizontal direction. By statistically analyzing the numerical density in the distribution map, the model calculates the bedding index within a local window, thus obtaining the distribution data of bedding density in the target sampling area. Based on the learned mapping relationship between bedding development and rock mechanical properties, the model can convert the development values ​​in the distribution map of primary bedding development in the target sampling area into estimated rock hardness values. By comprehensively analyzing the roughness, contrast, and spatial continuity of the texture, the deep neural network can infer the degree of rock fragmentation and then inversely deduce the distribution of rock hardness throughout the region.

[0050] Step S32: Based on the distribution data of bedding density in the target sampling area, the bedding dip angle sequence, and the rock hardness distribution data, determine multiple geological feature blocks to be divided, the structural stability index of each block, the geological morphology complexity, and the estimated drilling resistance parameters.

[0051] In some embodiments, deep neural networks can be used to determine multiple geological features to be divided into blocks, the structural stability index of each block, the geological morphology complexity, and the estimated drilling resistance parameters.

[0052] Multiple geological feature blocks to be divided are obtained by spatially discretizing the target sampling area based on the similarity of geological attributes using a deep neural network, resulting in multiple sub-regions with independent boundaries.

[0053] The structural stability index of each block is a numerical indicator of the continuity and homogeneity of the geological structure within each geological feature block to be divided, determined by a deep neural network.

[0054] Geological morphological complexity is a quantitative indicator determined by deep neural networks, representing the degree of drastic change in bedding trend and the degree of tectonic disorder within each geological feature to be divided block.

[0055] The drilling resistance prediction parameter is a value determined by a deep neural network that predicts the magnitude of the mechanical resistance experienced by the drill bit when drilling in a geologically defined block.

[0056] Deep neural networks can spatially register and fuse multi-source geological data to construct a multi-dimensional feature space. They can identify continuous spatial ranges with high consistency in bedding density, dip angle, and hardness properties, thus defining them as multiple geological feature blocks. For each block, the deep neural network calculates the variance of the bedding dip angle and the standard deviation of bedding density; smaller variances and standard deviations indicate a more stable structure, thus deriving a structural stability index for each block. The model analyzes the rate of curvature change and irregularity of bedding strike within the block, combined with the frequency of bedding intersections, to calculate the geological morphological complexity. Furthermore, by combining rock hardness distribution data and bedding density distribution data, and comprehensively considering the cutting resistance caused by hardness and the frictional resistance caused by bedding, the deep neural network can ultimately calculate the estimated drilling resistance parameters for the block.

[0057] Step S33: Based on the multiple geological features to be divided into blocks, the structural stability index of each block, the geological morphology complexity, and the drilling resistance prediction parameters, determine the peripheral core reference sampling area and the primary bedding core sampling area.

[0058] In some embodiments, deep neural networks can be used to determine the peripheral core reference sampling area and the primary bedding core sampling area.

[0059] Deep neural networks can use the physical indicators of each block as input vectors. Through pre-set learning logic, the model can weigh the scientific research value and operational risks of different blocks, automatically selecting blocks with extremely complex geological morphology and original characteristics as primary bedding core sampling areas, and using blocks with highly stable geological structures as peripheral core reference sampling areas.

[0060] Step S4: Based on the distribution map of the development degree of primary bedding in the bedrock of the outer core reference sampling area and the distribution map of the development degree of primary bedding in the bedrock of the primary bedding core sampling area, determine the drilling point information of multiple sets of the outer core reference sampling area and the drilling point information of multiple sets of the primary bedding core sampling area.

[0061] The distribution map of the primary bedding development of the bedrock in the outer core reference sampling area is a local distribution map of the primary bedding development of the bedrock in the target sampling area, corresponding to the range of the outer core reference sampling area.

[0062] The distribution map of the development degree of primary bedding in the bedrock of the primary bedding core sampling area is a local distribution map of the development degree of primary bedding in the bedrock of the target sampling area, corresponding to the range of the primary bedding core sampling area.

[0063] In some embodiments, a drilling point determination model can be used to determine multiple sets of drilling point information for the peripheral core reference sampling area and multiple sets of drilling point information for the primary bedding core sampling area. The drilling point determination model is a deep neural network model. The inputs to the drilling point determination model are the distribution maps of the primary bedding development degree of the bedrock in the peripheral core reference sampling area and the primary bedding core sampling area. The outputs of the drilling point determination model are multiple sets of drilling point information for the peripheral core reference sampling area and multiple sets of drilling point information for the primary bedding core sampling area.

[0064] The multiple drilling point information for the peripheral core reference sampling area is output by the drilling point determination model for multiple drilling point locations within the peripheral core reference sampling area. This information includes the specific spatial coordinates, preset drilling depth, and azimuth angle for exploratory drilling operations.

[0065] The multiple drilling point information for the primary bedding core sampling area is output by the drilling point determination model. This information includes the specific spatial coordinates, preset drilling depth, and azimuth of the drilling points for the formal drilling operation.

[0066] The distribution maps of the development degree of primary bedrock bedding in the peripheral core reference sampling area and the primary bedding core sampling area can represent the variation in the strength of geological structures. The values ​​in the distribution maps of the development degree of primary bedrock bedding can intuitively reflect the density and development state of the bedding. Among them, the regional characteristics of stable and uniform values ​​can support the model in locating peripheral reference points with low interference, while the regional characteristics of significant values ​​and clear textures provide a basis for the model to identify primary bedding points with typical research value. The spatial continuity characteristics in the distribution maps can help the model assess the environmental integrity around the points, thereby selecting the best drilling location that balances representativeness and safety.

[0067] Deep neural networks analyze the distribution maps of the primary bedding development in the bedrock of the peripheral core reference sampling area and the primary bedding core sampling area to find the optimal sampling location within a local range. For the peripheral core reference sampling area, the deep neural network can identify the center of the region with the smallest numerical fluctuation and the gentlest bedding in the distribution map as the drilling point to ensure stable acquisition of reference samples. By calculating the gradient field of the distribution map, the deep neural network can locate the zero gradient point or low-gradient flat area to generate multiple sets of drilling point information for the peripheral core reference sampling area. The deep neural network can also capture the information of the richest bedding development in the primary bedding core sampling area. The model can identify locations in the distribution map with clear bedding texture, a medium-to-high level of development, and a certain degree of continuity. Through grid search, the deep neural network can maximize the sampling information entropy of each point while ensuring the uniformity of the spatial distribution of points, thereby determining multiple sets of drilling point information for the primary bedding core sampling area.

[0068] Step S5: Determine the core drilling cooperative arm control scheme for the outer core reference sampling area based on multiple sets of drilling point information of the outer core reference sampling area.

[0069] In some embodiments, Figure 2 This is a flowchart illustrating a core drilling support arm control scheme for determining the peripheral core reference sampling area, provided by an embodiment of the present invention. The core drilling support arm control scheme for determining the peripheral core reference sampling area includes steps S51 to S53:

[0070] Step S51: Based on the multiple sets of drilling point information of the peripheral core reference sampling area, clustering is performed to obtain multiple clusters.

[0071] The clustering method used is K-means clustering. K-means clustering is a classic unsupervised learning algorithm. Its core principle is to automatically divide a given dataset into K clusters, thereby ensuring that data points within the same cluster have high similarity and data points between different clusters have significant differences, thus achieving rapid grouping and classification of data.

[0072] In some embodiments, the value of K can be determined using a preset relationship table between the value of K and the total number of drilling points in the peripheral core reference sampling area. The larger the total number of drilling points in the peripheral core reference sampling area, the larger the value of K. The preset relationship table between the value of K and the total number of drilling points in the peripheral core reference sampling area is artificially constructed in advance.

[0073] Multiple clusters are obtained by clustering multiple sets of drilling point information from the peripheral core reference sampling area using the K-means clustering algorithm. The number of clusters is K. Each cluster contains a group of drilling point information with similar characteristics. Drilling points within the same cluster have high similarity in terms of spatial distribution and geological feature correlation.

[0074] As an example, specifically: K drilling point information is randomly selected as the initial cluster centers. For each set of drilling point information in the peripheral core reference sampling area, its spatial distance to all initial cluster centers is calculated, and the set of drilling point information is assigned to the nearest cluster center, thus forming K clusters. For each cluster, the average spatial characteristic of all drilling point information within the cluster is calculated, and this average value is used as the new cluster center. The above process is repeated until the positional change of the cluster center is less than a preset threshold or the predetermined number of iterations is reached, at which point the clustering ends.

[0075] Clustering multiple drilling point information from the peripheral core sampling area into several clusters allows for grouping drilling points with similar spatial distribution characteristics and geological conditions into one category, thus achieving structured grouping of dispersed drilling points. This grouping method clearly distinguishes point units under different operating conditions and avoids indiscriminate analysis and planning for all drilling points. Subsequently, the working position of the cooperative arm and the drilling torque can be accurately determined based on the overall characteristics of the cluster, improving the efficiency and specificity of core drilling scheme design.

[0076] Step S52: Determine multiple cooperative arm workstation locations and the drilling torque for each cooperative arm workstation location based on the multiple clusters.

[0077] In some embodiments, a borehole analysis model can be used to determine multiple collaborative arm workstation locations and the borehole torque at each collaborative arm workstation location. The borehole analysis model is a deep neural network model. The input to the borehole analysis model is the plurality of clusters, and the output of the borehole analysis model is the multiple collaborative arm workstation locations and the borehole torque at each collaborative arm workstation location.

[0078] The multiple collaborative arm work stations are the set of optimal spatial coordinates of the end effector of the robot collaborative arm that needs to stop when performing drilling operations, determined by the drilling analysis model.

[0079] The drilling torque at each working station of the collaborative arm is a torque parameter for drilling operations determined by the drilling analysis model for each working station.

[0080] Multiple clusters represent the spatial distribution characteristics of the points. The center and extent of each cluster can provide geometric constraints for the selection of station points, thereby ensuring that the workspace of the collaborative arm can cover all points within the cluster.

[0081] By analyzing the geometric center and dispersion of points within each cluster, and combining this with the kinematic reachability of the collaborative arm, deep neural networks can calculate station coordinates that balance efficiency and accessibility. Simultaneously, deep neural networks can use estimated geological hardness values ​​corresponding to each point within the cluster for feature mapping to deduce the reasonable power output required for operation in the current area, thereby outputting multiple collaborative arm operating station locations and the drilling torque for each collaborative arm operating station location.

[0082] Step S53: Generate a core drilling cooperative arm control scheme for the outer core reference sampling area based on the multiple cooperative arm work stations and the drilling torque of each cooperative arm work station.

[0083] In some embodiments, a scheme generation model can be used to generate a control scheme for the core drilling cooperative arm in the peripheral core reference sampling area. The scheme generation model is a Transformer model. The inputs to the scheme generation model are the plurality of cooperative arm workstations and the drilling torque at each cooperative arm workstation, and the output of the scheme generation model is the control scheme for the core drilling cooperative arm in the peripheral core reference sampling area.

[0084] The Transformer model is a deep learning model based on the self-attention mechanism. It abandons traditional recurrent and convolutional structures, relying entirely on attention to capture global dependencies in the input sequence. The Transformer model mainly consists of an encoder and a decoder. The encoder maps the input sequence to hidden layer representations, and the decoder uses these representations to generate the target sequence. The Transformer model features strong parallel computing capabilities and the ability to handle long-range dependencies, making it suitable for sequence generation and path planning tasks.

[0085] The core drilling collaborative arm control scheme for the outer core reference sampling area is a detailed sequence of instructions output by the scheme generation model to control the robotic collaborative arm to perform drilling tasks in the outer area.

[0086] The control scheme for the core drilling collaborative arm in the outer core reference sampling area includes the robot collaborative arm's station movement sequence, the robotic arm's posture adjustment trajectory, the timing of drilling action triggering, and torque control parameters.

[0087] The Transformer model utilizes position encoding technology to map the spatial coordinates of multiple collaborative arm workstations and their corresponding drilling torques into a high-dimensional feature vector sequence. The encoder, through a multi-head self-attention mechanism, analyzes the spatial distance relationships and torque parameter correlations between different workstations, and calculates the mutual weights between all workstations, thereby capturing the global operational logic and constraints. When generating a control scheme, the decoder can predict the optimal work sequence step-by-step based on the context information output by the encoder. The Transformer model can balance the principle of shortest travel distance with the principle of operational safety, thus generating an optimal path connecting all workstations. Simultaneously, the Transformer model can embed drilling torque parameters into the corresponding motion commands, ultimately generating a complete sequence containing movement commands, attitude adjustment commands, and drilling parameter setting commands.

[0088] Step S6: Based on the core drilling collaborative arm control scheme of the outer core reference sampling area, control the robot collaborative arm to carry out core drilling operations and obtain core drilling quality information.

[0089] Core drilling quality information is information obtained by evaluating the quality of core samples obtained by a robotic collaborative arm during core drilling operations in the outer core reference sampling area. Figure 3 This is a schematic diagram of a robotic collaborative arm provided in an embodiment of the present invention.

[0090] Core drilling quality information includes core recovery rate (RQD), core column integrity, fracture surface characteristics, and real-time feedback data recorded during the drilling process.

[0091] Real-time feedback data includes drilling speed fluctuations, abnormal torque peaks, etc.

[0092] Core drilling quality information can reflect the actual effect of the core drilling cooperative arm control scheme and geological conditions in the current peripheral core reference sampling area.

[0093] Step S7: Based on the core drilling quality information and the multiple drilling point information of the primary bedding core sampling area, determine the core drilling cooperative arm control scheme for the primary bedding core sampling area.

[0094] In some embodiments, Figure 4 This is a flowchart illustrating a core drilling cooperation arm control scheme for determining a primary bedding core sampling area, provided by an embodiment of the present invention. The core drilling cooperation arm control scheme for determining a primary bedding core sampling area includes steps S71 to S75:

[0095] Step S71: Construct a drilling point feature map, which includes multiple drilling point nodes and multiple edges between drilling point nodes. The node features of each drilling point node are drilling point information.

[0096] The drilling point feature map consists of a set of nodes and a set of edges. The drilling point feature map can structurally represent the spatial topological relationships and attribute characteristics of each drilling point in the primary bedding core sampling area.

[0097] Multiple drilling point nodes are the basic units in the map, with each node representing a specific drilling point within the primary bedding core sampling area. The node characteristics of each drilling point node contain drilling point information.

[0098] The multiple edges between drilling point nodes are components of the drilling point feature map, which characterizes the association features of different drilling point nodes. The features of the edges are the positional relationships between the drilling point nodes.

[0099] The location relationship specifically includes the azimuth information and straight-line distance data between the drilling points corresponding to each node.

[0100] Step S72: Based on the graph neural network, the feature map of the drilling point is processed to determine the locations of multiple cooperative arm operation stations in the original bedding core sampling area.

[0101] Graph Neural Networks (GNNs) are deep learning models that can process graph data. Through message passing mechanisms, GNNs allow each node to aggregate the features and relationships of its neighbors, thereby capturing the topological structure and global dependencies of the data. The input to the GNN is the feature map of the drilling point, and the output is the locations of multiple collaborative arm workstations in the original bedding core sampling area.

[0102] The multiple collaborative arm work stations in the original bedding core sampling area are a set of optimal stopping and working positions for the robot collaborative arms planned by graph neural networks for the complex point distribution in the original bedding core sampling area.

[0103] In the drill point feature map, each node represents a drill point within the primary bedding core sampling area. The node features are drill point information, directly reflecting the geological conditions and drilling requirements of the corresponding point. Edges between nodes represent the relationships between different drill points. Processing the drill point feature map using a graph neural network allows for precise determination of the collaborative arm's workstation locations. This helps optimize the drill point layout, achieving efficient coverage of all drill points while avoiding operational interference.

[0104] Graph neural networks (Graph Neural Networks) fuse the features of each node in a drilling site feature map with the features of its neighboring nodes through multi-layer graph convolution operations. Graph Neural Networks can identify densely populated sub-maps (communities) within the drilling site feature map; these sub-maps correspond to spatially adjacent groups of points suitable for concentrated operations. Through pooling operations, Graph Neural Networks aggregate each sub-map into a supernode, whose features represent the central location and comprehensive geological attributes of the area. Based on the features of the aggregated supernodes, the model can calculate the central location that best covers all original nodes in each sub-map and determine it as the collaborative arm work station location. In this way, Graph Neural Networks utilize the topological structure of the graph to achieve optimized grouping and site selection of points, thereby ensuring that multiple collaborative arm work stations in the primary bedding core sampling area can efficiently cover all target points in the entire complex network.

[0105] Step S73: Based on the core drilling cooperative arm control scheme of the outer core reference sampling area, the core drilling quality information, and the multiple drilling point information of the original bedding core sampling area, generate simulated core drilling quality information under different drilling torques in the original bedding core sampling area.

[0106] In some embodiments, a quality information generation model can be used to generate simulated core drilling quality information under different drilling torques in the native bedding core sampling area. The quality information generation model is a generative adversarial network (GAN). The inputs to the quality information generation model are the core drilling cooperative arm control scheme of the peripheral core reference sampling area, the core drilling quality information, and multiple sets of drilling point information in the native bedding core sampling area. The output of the quality information generation model is the simulated core drilling quality information under different drilling torques in the native bedding core sampling area.

[0107] Generative Adversarial Networks (GANs) are game-like deep learning models that consist of a generator and a discriminator. The generator is responsible for producing simulated data that is as realistic as possible, while the discriminator is responsible for distinguishing real data from the data generated by the generator. During training, the two compete against each other and continuously optimize, ultimately enabling the generator to produce samples with distribution characteristics highly consistent with real data. GANs excel in data augmentation, image generation, and complex system simulation.

[0108] The simulated core drilling quality information for primary bedding core sampling areas under different drilling torques is predicted by a quality information generation model and describes the core quality assessment data that can be obtained from the primary bedding region under different drilling torque parameters. The simulated quality information includes simulated core recovery rate, simulated core column integrity, simulated fracture rate, and simulated drilling stability coefficient.

[0109] The core drilling coordination arm control scheme and core drilling quality information from the peripheral core reference sampling area constitute a real sample pair of operational parameters and results, encompassing the response patterns of drilling operations under this geological environment. Multiple sets of drilling point information from the native bedding core sampling area provide the target environmental parameters to be predicted. The generator of the generative adversarial network receives multiple sets of drilling point information from the native bedding core sampling area as input vectors. The generator can extract the mapping features between drilling parameters and operational results between the core drilling coordination arm control scheme and core drilling quality information from the peripheral core reference sampling area, and generate corresponding simulated core drilling quality information under different borehole torques in the native bedding core sampling area based on these mapping features. The discriminator can receive the core drilling quality information and the simulated core drilling quality information under different borehole torques output by the generator in the native bedding core sampling area, and attempt to distinguish which set of data originates from real drilling operation feedback. During adversarial training, the generator adjusts weights using a backpropagation algorithm to ensure that the generated simulated core drilling quality information under different drilling torques in the native bedding core sampling area approximates the actual core drilling quality information as closely as possible in terms of statistical distribution and physical properties. The model uses multiple sets of drilling point information from the native bedding core sampling area as constraints, enabling the generator to deduce the specific impact of different torque parameters on core integrity within the native bedding geological environment. Ultimately, the generator outputs simulated core drilling quality information under different drilling torques in the native bedding core sampling area.

[0110] Step S74: Based on the simulated core drilling quality information under different drilling torques in the original bedding core sampling area and the information of multiple drilling points corresponding to each cooperative arm work station in the original bedding core sampling area, determine the target drilling torque for each cooperative arm work station in the original bedding core sampling area.

[0111] In some embodiments, a target torque determination model can be used to determine the target borehole torque for each cooperative arm workstation in the primary bedding core sampling area. The target torque determination model is a deep neural network model. The inputs to the target torque determination model are simulated core drilling quality information under different borehole torques in the primary bedding core sampling area, and multiple drilling point information corresponding to each cooperative arm workstation in the primary bedding core sampling area. The output of the target torque determination model is the target borehole torque for each cooperative arm workstation in the primary bedding core sampling area.

[0112] The target drilling torque for each working arm station in the primary bedding core sampling area is the optimal drilling rotation torque value determined by the target torque determination model for all drilling points within the coverage area of ​​each working arm station.

[0113] Simulated core drilling quality data under different drilling torques in the primary bedding core sampling area reveals a nonlinear relationship between torque variation and sampling quality. For each cooperative arm workstation, which covers multiple drilling points with specific geological characteristics, the model can analyze the optimal torque to balance the quality requirements of all drilling points based on the simulated core drilling quality data.

[0114] Deep neural networks, using a multilayer perceptron structure, can analyze simulated core drilling quality information under different drilling torques in input native bedding core sampling areas. For each cooperative arm workstation, the deep neural network can calculate the simulated quality score of all drilling points within its coverage area under different torques. The model aims to maximize the average core integrity of all points within that station while constraining the torque values ​​within the equipment's safe range. It then uses a grid search strategy to select the optimal solution from the simulated quality responses corresponding to discrete torque parameters. The deep neural network can analyze the trade-off between the increased cutting efficiency brought about by increased torque and the potential risk of core breakage, thereby identifying the torque range with the highest quality score. Finally, the model selects the optimal value within this range as the best parameter for that station and determines this optimal value as the target drilling torque for each cooperative arm workstation in the native bedding core sampling area, ensuring the acquisition of high-quality undisturbed cores in actual operations.

[0115] Step S75: Based on the multiple cooperative arm work stations in the original bedding core sampling area and the target drilling torque of each cooperative arm work station in the original bedding core sampling area, generate a core drilling cooperative arm control scheme for the original bedding core sampling area.

[0116] In some embodiments, a second scheme generation model can be used to generate a core drilling cooperative arm control scheme for the native bedding core sampling area. The second scheme generation model is a Transformer model. The inputs to the second scheme generation model are multiple cooperative arm workstation locations in the native bedding core sampling area and the target drilling torque for each cooperative arm workstation location in the native bedding core sampling area. The output of the second scheme generation model is the core drilling cooperative arm control scheme for the native bedding core sampling area.

[0117] The control scheme for the core drilling collaborative arm in the primary bedding core sampling area is a comprehensive command sequence output by the second scheme generation model, which controls the robotic collaborative arm to perform sampling tasks in the primary bedding region. The control scheme for the core drilling collaborative arm in the primary bedding core sampling area includes the robotic collaborative arm's stationary movement sequence, the robotic arm's attitude adjustment trajectory, the drilling action triggering timing, and torque control parameters.

[0118] The Transformer model leverages its sequence modeling capabilities to process multiple cooperative arm workstations and target borehole torque data for each workstation in an unconfined bedding core sampling area. The Transformer model's encoder spatially encodes the workstation locations within the unconfined bedding region and combines this with the numerical characteristics of the target torque to construct a contextual representation of the task. The model's attention mechanism assigns higher weights to workstations with complex geological features and prioritizes the timing of these key nodes. The decoder autoregressively generates a sequence of control commands based on the encoded information. The Transformer model optimizes the transfer paths between workstations to avoid collisions with complex terrain and ensures that the drilling rig smoothly adjusts to the target torque during each workstation switch. Ultimately, the model generates a cooperative arm control scheme for core drilling in an unconfined bedding core sampling area, containing precise motion commands.

[0119] Step S8: Based on the core drilling collaborative arm control scheme of the original bedding core sampling area, control the robot collaborative arm to perform core drilling operations in the original bedding core sampling area.

[0120] Once the control scheme for the core drilling collaborative arm in the native bedding core sampling area is determined, the robotic collaborative arm is controlled to perform operations according to the preset station movement sequence, attitude adjustment trajectory and drilling sequence based on the control scheme for the core drilling collaborative arm in the native bedding core sampling area. The target drilling torque is accurately matched at each station to smoothly complete the drilling action and finally obtain high-quality native bedding core samples.

[0121] Based on the same inventive concept Figure 5 This is a schematic diagram of a control system for a high-torque robotic collaborative arm provided in an embodiment of the present invention. The control system for the high-torque robotic collaborative arm includes:

[0122] Data acquisition module 81 is used to acquire bedrock ground penetration radar data of the target sampling area;

[0123] The bedding distribution map generation module 82 is used to generate a distribution map of the degree of development of primary bedding in the bedrock of the target sampling area based on the bedrock ground penetration radar data of the target sampling area.

[0124] The sampling area division module 83 is used to determine the outer core reference sampling area and the primary bedding core sampling area based on the distribution map of the development degree of the bedrock primary bedding of the target sampling area;

[0125] The drilling point determination module 84 is used to determine multiple sets of drilling point information for the outer core reference sampling area and the primary bedding core sampling area based on the distribution map of the development degree of the primary bedding of the bedrock in the outer core reference sampling area and the distribution map of the development degree of the primary bedding of the bedrock in the primary bedding core sampling area.

[0126] The first control scheme generation module 85 is used to determine the core drilling cooperative arm control scheme of the outer core reference sampling area based on multiple sets of drilling point information of the outer core reference sampling area.

[0127] The first drilling execution module 86 is used to control the robot collaborative arm to carry out core drilling operations based on the core drilling collaborative arm control scheme of the outer core reference sampling area, and to obtain core drilling quality information.

[0128] The second control scheme generation module 87 is used to determine the core drilling cooperative arm control scheme of the original bedding core sampling area based on the core drilling quality information and the multiple sets of drilling point information of the original bedding core sampling area.

[0129] The second drilling execution module 88 is used to control the robot collaborative arm to perform core drilling operations in the original bedding core sampling area based on the core drilling collaborative arm control scheme of the original bedding core sampling area.

[0130] It should be noted that, in order to simplify the descriptions disclosed herein and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of embodiments of this specification may sometimes combine multiple features into a single embodiment, drawing, or description thereof. However, this method of disclosure does not imply that the subject matter of this specification requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of a single embodiment disclosed above.

[0131] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A control method for a high-torque robotic collaborative arm, characterized in that, include: Acquire bedrock ground penetration radar data of the target sampling area; A distribution map of the degree of development of primary bedrock bedding in the target sampling area is generated based on the ground penetration radar data of the bedrock in the target sampling area; Based on the distribution map of the development degree of primary bedrock bedding in the target sampling area, determine the outer core reference sampling area and the primary bedding core sampling area; Based on the distribution map of the development degree of primary bedrock bedding in the outer core reference sampling area and the distribution map of the development degree of primary bedrock bedding in the primary bedding core sampling area, multiple sets of drilling point information in the outer core reference sampling area and multiple sets of drilling point information in the primary bedding core sampling area are determined. Based on multiple sets of drilling point information of the peripheral core reference sampling area, a core drilling cooperative arm control scheme for the peripheral core reference sampling area is determined. This scheme includes: Multiple clusters were obtained by clustering based on the drilling point information of the surrounding core reference sampling area; Based on the multiple clusters, multiple cooperative arm workstation locations and the drilling torque of each cooperative arm workstation location are determined; A core drilling control scheme for the outer core reference sampling area is generated based on the multiple cooperative arm work stations and the drilling torque of each cooperative arm work station. Based on the core drilling collaborative arm control scheme of the outer core reference sampling area, the robot collaborative arm is controlled to carry out core drilling operations and obtain core drilling quality information. Based on the core drilling quality information and multiple sets of drilling point information in the primary bedding core sampling area, a core drilling cooperative arm control scheme for the primary bedding core sampling area is determined. This scheme includes: Construct a feature map of drilling points, which includes multiple drilling point nodes and multiple edges between drilling point nodes. The node features of each drilling point node are drilling point information. The drilling point feature map is processed using a graph neural network to determine multiple cooperative arm workstations in the original bedding core sampling area. Based on the core drilling cooperative arm control scheme of the outer core reference sampling area, the core drilling quality information, and the multiple drilling point information of the original bedding core sampling area, simulated core drilling quality information under different drilling torques in the original bedding core sampling area is generated. Based on the simulated core drilling quality information under different drilling torques in the original bedding core sampling area and the information of multiple drilling points corresponding to each cooperative arm work station in the original bedding core sampling area, the target drilling torque for each cooperative arm work station in the original bedding core sampling area is determined. Based on the multiple cooperative arm work stations in the original bedding core sampling area and the target drilling torque of each cooperative arm work station in the original bedding core sampling area, a core drilling cooperative arm control scheme for the original bedding core sampling area is generated. Based on the core drilling collaborative arm control scheme of the original bedding core sampling area, the robot collaborative arm is controlled to perform core drilling operations in the original bedding core sampling area.

2. The control method for a high-torque robotic collaborative arm as described in claim 1, characterized in that, The input of the graph neural network is the feature map of the drilling point, and the output of the graph neural network is the multiple cooperative arm operation stations in the original bedding core sampling area.

3. A control system for a high-torque robotic collaborative arm, characterized in that, include: The data acquisition module is used to acquire bedrock ground penetration radar data of the target sampling area; The bedding distribution map generation module is used to generate a distribution map of the degree of development of primary bedding in the bedrock of the target sampling area based on the bedrock ground penetration radar data of the target sampling area; The sampling area division module is used to determine the outer core reference sampling area and the primary bedding core sampling area based on the distribution map of the development degree of the bedrock primary bedding of the target sampling area; The drilling point determination module is used to determine multiple sets of drilling point information for the outer core reference sampling area and the primary bedding core sampling area based on the distribution map of the development degree of the primary bedding of the bedrock in the outer core reference sampling area and the distribution map of the development degree of the primary bedding of the bedrock in the primary bedding core sampling area. The first control scheme generation module is used to determine the core drilling cooperation arm control scheme of the outer core reference sampling area based on multiple sets of drilling point information of the outer core reference sampling area. The first control scheme generation module is also used for: Multiple clusters were obtained by clustering based on the drilling point information of the surrounding core reference sampling area; Based on the multiple clusters, multiple cooperative arm workstation locations and the drilling torque of each cooperative arm workstation location are determined; A core drilling control scheme for the outer core reference sampling area is generated based on the multiple cooperative arm work stations and the drilling torque of each cooperative arm work station. The first drilling execution module is used to control the robot collaborative arm to carry out core drilling operations based on the core drilling collaborative arm control scheme of the outer core reference sampling area, and to obtain core drilling quality information. The second control scheme generation module is used to determine the core drilling cooperation arm control scheme for the primary bedding core sampling area based on the core drilling quality information and multiple sets of drilling point information of the primary bedding core sampling area. The second control scheme generation module is also used for: Construct a feature map of drilling points, which includes multiple drilling point nodes and multiple edges between drilling point nodes. The node features of each drilling point node are drilling point information. The drilling point feature map is processed using a graph neural network to determine multiple cooperative arm workstations in the original bedding core sampling area. Based on the core drilling cooperative arm control scheme of the outer core reference sampling area, the core drilling quality information, and the multiple drilling point information of the original bedding core sampling area, simulated core drilling quality information under different drilling torques in the original bedding core sampling area is generated. Based on the simulated core drilling quality information under different drilling torques in the original bedding core sampling area and the information of multiple drilling points corresponding to each cooperative arm work station in the original bedding core sampling area, the target drilling torque for each cooperative arm work station in the original bedding core sampling area is determined. Based on the multiple cooperative arm work stations in the original bedding core sampling area and the target drilling torque of each cooperative arm work station in the original bedding core sampling area, a core drilling cooperative arm control scheme for the original bedding core sampling area is generated. The second drilling execution module is used to control the robot collaborative arm to perform core drilling operations in the original bedding core sampling area based on the core drilling collaborative arm control scheme of the original bedding core sampling area.

4. The control system for the high-torque robotic collaborative arm as described in claim 3, characterized in that, The input of the graph neural network is the feature map of the drilling point, and the output of the graph neural network is the multiple cooperative arm operation stations in the original bedding core sampling area.

5. An electronic device, characterized in that, include: processor; Memory; And a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the control method for a high-torque robotic collaborative arm as described in any one of claims 1 to 2.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the control method for a high-torque robotic collaborative arm as described in any one of claims 1 to 2.

Images