Mechanical arm control method, device and equipment in C2F mode and medium

Abstract

The invention discloses a mechanical arm control method, device and equipment in a C2F mode and a medium, and the method comprises the steps that data collection is conducted on a mechanical arm in the mechanical arm control equipment, and multiple pieces of sensor data are obtained; performing data grouping on all the sensor data based on a preset clustering algorithm to obtain a plurality of data groups and clustering tags, and performing clustering evaluation on each data group according to each clustering tag to obtain clustering data; performing feature learning on the clustering data to obtain a high-level feature vector, and generating a task template corresponding to the clustering data according to the high-level feature vector and the clustering label; feedback information fed back when the mechanical arm executes the task at the current moment is obtained; and the task template is adjusted according to the feedback information and the high-level feature vector, a target task at the next moment is generated, and the mechanical arm is controlled according to the target task. In the embodiment of the invention, the task can be dynamically generated according to the feedback, and the flexibility and adaptability of production are improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method, device, equipment and medium for controlling a robot arm in a C2F mode. Background Art

[0002] The Customer to Factory (C2F) model refers to the use of advanced information technology and communication technology to enable end customers to interact directly with manufacturing companies, thereby achieving rapid delivery of personalized customized products and services. In this model, customers can submit specific product requirements, specifications or designs through an online platform, and manufacturers can flexibly adjust production lines based on this information to meet customers' unique requirements. Taking the robotic arm system as an example, many robotic arm control systems rely on predefined control strategies and fixed task processes to further meet the needs of large-scale production.

[0003] However, current robotic arm control systems usually require a large amount of training data and time, and perform unstable in complex environments, making it difficult for the control system to adapt to rapidly changing production needs and lacking flexibility. Summary of the invention

[0004] The embodiments of the present application provide a robot arm control method, device, electronic device and readable storage medium in a C2F mode, which can dynamically generate tasks based on feedback to improve the flexibility and adaptability of production.

[0005] In a first aspect, an embodiment of the present application provides a robot arm control method in a C2F mode, which is applied to a robot arm control device, and the method includes: Collecting data on the robot arm in the robot arm control device to obtain multiple sensor data; Grouping all the sensor data based on a preset clustering algorithm to obtain a plurality of data groups and a clustering label corresponding to each of the data groups, and performing clustering evaluation on each of the data groups according to each of the clustering labels to obtain clustered data; Performing feature learning on the cluster data to obtain a high-level feature vector, and generating a task template corresponding to the cluster data according to the high-level feature vector and the cluster label; Obtaining feedback information from the robot arm when performing the task at the current moment; The task template is adjusted according to the feedback information and the high-level feature vector, a target task at the next moment is generated, and the robotic arm is controlled according to the target task.

[0006] In some embodiments, the data grouping of all the sensor data based on a preset clustering algorithm to obtain a plurality of data groups and a clustering label corresponding to each of the data groups includes: Performing feature encoding on all the sensor data to obtain numerical representations corresponding to the sensor data, and constructing feature vectors corresponding to the sensor data based on the numerical representations; Selecting a preset number of feature vectors from the sensor data as initial clustering centers; For each of the initial cluster centers, determining a distance value between each of the sensor data and the initial cluster center; The sensor data are distributed according to the distance values ​​to obtain a plurality of data groups, and a label is distributed to each of the data groups to obtain a cluster label corresponding to each of the data groups.

[0007] In some embodiments, after allocating the sensor data according to the distance value to obtain a plurality of data groups, and assigning a label to each of the data groups to obtain a cluster label corresponding to each of the data groups, the robot arm control method in the C2F mode further includes: Determining a data mean for all data points in each of the data sets; Updating the initial cluster center of the data group according to the data mean; The step of grouping all the sensor data based on a preset clustering algorithm to obtain a plurality of data groups and a clustering label corresponding to each of the data groups includes: For each updated initial cluster center, determining a target distance value between each of the sensor data and the updated initial cluster center; The sensor data is distributed according to the target distance value to obtain a plurality of target data groups, and a label is distributed to each of the target data groups to obtain a cluster label corresponding to each of the target data groups.

[0008] In some embodiments, performing cluster evaluation on each of the data groups according to each of the cluster labels to obtain cluster data includes: For each of the data groups, selecting a target data point in the data group; Determine a first similarity between the target data point and other data points in the corresponding data group, and determine a second similarity between the target data point and data points in the other data groups; determining a silhouette coefficient according to the first similarity and the second similarity; Comparing the silhouette coefficient with a preset coefficient range to determine a target number of clusters; The sensor data is adjusted according to the target cluster number to obtain cluster data.

[0009] In some embodiments, performing feature learning on the cluster data to obtain a high-level feature vector includes: Reconstructing the cluster data by a decoder to obtain reconstructed data; Training the autoencoder using the reconstructed data and the clustered data to obtain a pre-trained autoencoder; The clustering data is input into a pre-trained autoencoder, so that the pre-trained autoencoder performs feature extraction on the clustering data and outputs a high-level feature vector.

[0010] In some embodiments, the feedback information is obtained by the following steps: Record the success rate, execution time and error rate of the task performed by the robot arm at the current moment; Feedback information is generated according to the success rate, the execution time and the error rate.

[0011] In some embodiments, adjusting the task template according to the feedback information and the high-level feature vector to generate a target task at the next moment includes: Comparing the feedback information with the preset expected target to obtain an evaluation result; determining adjustment parameters based on the evaluation results and the high-level feature vector; The parameters in the task template are modified based on the adjustment parameters to generate a target task at the next moment.

[0012] In some embodiments, before grouping all the sensor data based on a preset clustering algorithm to obtain multiple data groups, the robot arm control method in the C2F mode further includes: Performing a preprocessing operation on all the sensor data, and performing a standardization process on the sensor data after the preprocessing operation, so as to scale all the sensor data to a preset range; The method of grouping all the sensor data based on a preset clustering algorithm to obtain a plurality of data groups includes: The scaled sensor data is grouped based on a preset clustering algorithm to obtain a plurality of data groups.

[0013] In a second aspect, an embodiment of the present application provides a robot arm control device in a C2F mode, comprising: A data acquisition module, used to collect data from the robot in the robot control device to obtain multiple sensor data; An adaptive clustering module, used for grouping all the sensor data based on a preset clustering algorithm to obtain a plurality of data groups and a clustering label corresponding to each of the data groups, and performing clustering evaluation on each of the data groups according to each of the clustering labels to obtain clustered data; A feature learning module, used for performing feature learning on the cluster data to obtain a high-level feature vector, and generating a task template corresponding to the cluster data according to the high-level feature vector and the cluster label; A data feedback module is used to obtain feedback information fed back by the robot arm when performing the task at the current moment; The task generation module is used to adjust the task template according to the feedback information and the high-level feature vector, generate a target task at the next moment, and control the robotic arm according to the target task.

[0014] In a third aspect, an embodiment of the present application provides an electronic device comprising at least one processor and a memory for communicating with the at least one processor; the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the robot arm control method in the C2F mode as described in the first aspect.

[0015] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the robot arm control method in the C2F mode as described in the first aspect.

[0016] The robot control method, device, electronic device and readable storage medium under the C2F mode provided by the embodiment of the present application have at least the following beneficial effects: first, data collection is performed on the robot in the robot control device to obtain multiple sensor data, and the real-time collection of the robot state is realized, which is convenient for subsequent improvement of operation accuracy and efficiency. Then, all sensor data are grouped based on the preset clustering algorithm, so that the data points in the same group have high similarity, while the data points between different groups have low similarity, and multiple data groups and cluster labels corresponding to each data group are obtained. The operation mode can be identified from the sensor data to generate a cluster label, and each data group is clustered and evaluated according to each cluster label to obtain cluster data, so that the mechanism of adaptively adjusting the number of clusters can be combined to adapt to the dynamic environment. Afterwards, feature learning is performed on the cluster data to obtain a high-level feature vector, so that a high-level, abstract feature representation can be automatically extracted from the above-mentioned clustered data, and the key information and internal structure of the data are retained. A task template corresponding to the cluster data is generated according to the high-level feature vector and the cluster label, which is convenient for the subsequent control of the operation state of the robot according to the task template. Then obtain the feedback information fed back by the robot arm when performing the task at the current moment, realize real-time monitoring of the parameters of the robot arm to perform different tasks, adjust the task template according to the feedback information and high-level feature vectors, and dynamically generate tasks according to the feedback to realize dynamic adjustment of the task template to ensure that it can flexibly respond to different operational requirements, generate the target task at the next moment, and control the robot arm according to the target task, so as to improve the flexibility and adaptability of production and ensure that the system can flexibly respond to personalized customized tasks in complex and dynamic environments.

[0017] Other features and advantages of the present application will be described in the following description, and partly become apparent from the description, or understood by practicing the present application. The purpose and other advantages of the present application can be realized and obtained through the structures particularly pointed out in the description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the architecture of a robotic arm control device provided by an embodiment of the present application; Figure 2 is a specific method flow chart of a robot arm control method in a C2F mode according to an embodiment of the present application; Figure 3 It is a flowchart of a specific method for grouping all sensor data based on a preset clustering algorithm provided in an embodiment of the present application; Figure 4 is a specific method flow chart of a robot arm control method in C2F mode provided by another embodiment of the present application; Figure 5is a flowchart of a specific method for performing cluster evaluation on each data group according to each cluster label provided in an embodiment of the present application; Figure 6 is a flowchart of a specific method for obtaining feedback information provided by an embodiment of the present application; Figure 7 This is a flowchart of a specific method for performing feature learning on cluster data provided by an embodiment of the present application; Figure 8 is a flowchart of a specific method for adjusting a task template according to feedback information and a high-level feature vector provided in an embodiment of the present application; Fig. 9 is a specific method flow chart of a robot arm control method in C2F mode provided by another embodiment of the present application; Fig.10 is a schematic diagram of the structure of a robotic arm control device in a C2F mode provided by an embodiment of the present application; Fig.11 It is a schematic diagram of the structure of an electronic device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0019] This section will describe in detail the specific embodiments of the present invention. The preferred embodiments of the present invention are shown in the accompanying drawings. The purpose of the accompanying drawings is to supplement the description of the text part of the specification with graphics, so that people can intuitively and vividly understand each technical feature and the overall technical solution of the present invention, but it cannot be understood as a limitation on the scope of protection of the present invention.

[0020] In the description of the present invention, if there is a description of first and second, it is only for the purpose of distinguishing the technical features, and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0021] The Customer to Factory (C2F) model refers to the use of advanced information technology and communication technology to enable end customers to interact directly with manufacturing companies, thereby achieving rapid delivery of personalized customized products and services. In this model, customers can submit specific product requirements, specifications or designs through an online platform, and manufacturers can flexibly adjust production lines based on this information to meet customers' unique requirements. Taking the robotic arm system as an example, many robotic arm control systems rely on predefined control strategies and fixed task processes to further meet the needs of large-scale production.

[0022] However, current robotic arm control systems usually require a large amount of training data and time, and perform unstable in complex environments, making it difficult for the control system to adapt to rapidly changing production needs and lacking flexibility.

[0023] In order to solve the above problems, the present embodiment provides a method, device, equipment and medium for controlling a robot arm in a C2F mode, which has at least the following beneficial effects: first, data collection is performed on the robot arm in the robot arm control device in the C2F mode to obtain multiple sensor data, and the real-time collection of the robot arm state is realized, which is convenient for subsequent improvement of operation accuracy and efficiency, and then all sensor data are grouped based on a preset clustering algorithm, so that the data points in the same group have high similarity, while the data points between different groups have low similarity, and multiple data groups and cluster labels corresponding to each data group are obtained, and the operation mode can be identified from the sensor data to generate a cluster label, and each data group is clustered and evaluated according to each cluster label to obtain cluster data, so that the mechanism of adaptively adjusting the number of clusters can be combined to adapt to the dynamic environment. Afterwards, feature learning is performed on the cluster data to obtain a high-level feature vector, so that a high-level, abstract feature representation can be automatically extracted from the above clustered data, and the key information and internal structure of the data are retained, and a task template corresponding to the cluster data is generated according to the high-level feature vector and the cluster label, so that the operation state of the robot arm can be controlled according to the task template in the future. Then obtain the feedback information fed back by the robot arm when performing the task at the current moment, realize real-time monitoring of the parameters of the robot arm to perform different tasks, adjust the task template according to the feedback information and high-level feature vectors, and dynamically generate tasks according to the feedback to realize dynamic adjustment of the task template to ensure that it can flexibly respond to different operational requirements, generate the target task at the next moment, and control the robot arm according to the target task, so as to improve the flexibility and adaptability of production and ensure that the system can flexibly respond to personalized customized tasks in complex and dynamic environments.

[0024] Reference Figure 1 , Figure 1 It is a schematic diagram of the architecture of a robotic arm control device provided in one embodiment of the present application.

[0025] In some embodiments, the robotic arm control device 100 may be an electronic device in an embodiment of the present application, and the robotic arm control device 100 may include but is not limited to a data acquisition component 110, an adaptive clustering learning component 120, a data feedback component 130 and a task generation component 140.

[0026] In some embodiments, the data acquisition component 110 is used to collect various sensor data of the manipulator in the manipulator control device during the execution of the task, such as position, speed, force, temperature, etc. Specifically, the data acquisition component 110 is also used to perform preliminary processing on the collected data, such as denoising, to improve the data quality, and to monitor the state of the manipulator in real time, to ensure that data collection and task execution are carried out in parallel, and to ensure that the system can respond to environmental changes and task requirements in a timely manner. In addition, the data acquisition component 110 can also receive tasks issued by the task execution module, and generate manipulator control instructions to control the various joints and actuators of the manipulator.

[0027] It is understandable that through real-time sensor data collection, data preprocessing, state monitoring, issuing control instructions and other functions, necessary environment and state information is provided to the robot arm control device 100. These data will be used for subsequent adaptive learning and dynamic task generation to ensure that the system can flexibly respond to personalized customized tasks in complex and dynamic environments.

[0028] In some embodiments, the adaptive clustering learning component includes an adaptive clustering component and a feature learning component. The adaptive clustering component combines a mechanism for adaptively adjusting the number of clusters to adapt to dynamic environments, and is intended to identify operating modes from sensor data. The feature learning component is used to extract features and generate operating strategies.

[0029] In some embodiments, the task generation component 140 is used to automatically generate task templates and adjust tasks based on the feedback results provided by the data feedback component 130 and the clustering labels and feature vectors of the adaptive clustering module to ensure that it can flexibly respond to different operational requirements. The module first defines a set of basic task templates (such as grasping, moving, placing, etc.) and collects feedback data (the execution status of the robotic arm, such as success rate, execution time, etc.). The module will dynamically adjust the task parameters (such as target position and grasping strength) according to the feedback execution results to optimize the task execution effect. And send the generated new task to the data acquisition component 110 for the generation of control instructions.

[0030] In some embodiments, the data feedback component 130 is used to obtain the execution status of the robot arm from the data acquisition component 110 in real time, and continuously obtain the status information of the robot arm during the execution of the task, such as the success rate, execution time and error rate, and compare these data with the preset target. And the comparison results are fed back to the task generation component 140. The task generation component 140 automatically adjusts the task parameters (such as the target position and grasping force) or updates the control strategy according to the feedback results to optimize the execution effect of the subsequent tasks. Through this closed-loop feedback mechanism, the system can continuously learn and adapt to new environmental changes and task requirements.

[0031] It can be understood that the robot arm control device 100 and the application scenario described in this embodiment are for the purpose of more clearly illustrating the technical solution of the embodiment of the present invention, and do not constitute a limitation on the technical solution provided by the embodiment of the present invention. Those skilled in the art can know that with the emergence of new application scenarios, the technical solution provided by the embodiment of the present application is also applicable to similar technical problems.

[0032] It can be understood by those skilled in the art that Figure 1 The robot arm control device 100 shown in the figure does not constitute a limitation on the embodiments of the present invention, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0033] Reference Figure 2 , Figure 2 is a specific method flow chart of a robot arm control method in a C2F mode according to an embodiment of the present application. The robot arm control method in the C2F mode is applied but not limited to Figure 1 The robot arm control device 100 shown in the method includes but is not limited to steps S101 to S105.

[0034] It is understandable that the C2F mode in the embodiment of the present application refers to the consumer-to-factory mode, which refers to the use of advanced information technology and communication technology to enable end customers to interact directly with manufacturing companies, thereby achieving rapid delivery of personalized customized products and services. Specifically, the robotic arm in the C2F mode in the embodiment of the present application can flexibly switch working modes according to different production tasks through flexible control algorithms and real-time adjustment capabilities, and can quickly adapt to the customized needs of different customers. The control method of the robotic arm in the C2F mode in the embodiment of the present application is specifically described below.

[0035] Step S101 , collecting data from a robot in the robot control device 100 to obtain a plurality of sensor data.

[0036] In step S101 of some embodiments, data is collected on the robotic arm in the robotic arm control device 100. Specifically, the embodiments of the present application collect data on the robotic arm through sensors. For example, the position information of the robotic arm is collected through a position sensor, the speed information of the robotic arm is collected through a speed sensor, the temperature information of the robotic arm is collected through a temperature sensor, and so on, to obtain multiple sensor data. By acquiring the position and speed data in real time, the robotic arm can perform tasks more accurately, thereby improving operational accuracy and efficiency.

[0037] Step S102, grouping all sensor data based on a preset clustering algorithm to obtain multiple data groups and a clustering label corresponding to each data group, and performing clustering evaluation on each data group according to each clustering label to obtain clustered data.

[0038] In step S102 of some embodiments, all sensor data are grouped based on a preset clustering algorithm to group sensor data with relatively close distances into a group, so that each data point in the sensor data can be assigned to a specific cluster, and a cluster label is assigned to each data group, so as to obtain multiple data groups and cluster labels corresponding to each data group, so as to identify the operation modes of different data groups, thereby simplifying the complexity of the data, and perform clustering evaluation on each data group according to each cluster label, so as to measure the closeness of the data points in their own clusters and the separation in other clusters, and these points can be identified and processed by analyzing the silhouette coefficient to obtain clustered data, thereby improving the accuracy of clustering.

[0039] It should be noted that the clustering algorithm in the embodiment of the present application can be a K-Means algorithm, a DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm, a hierarchical clustering algorithm, etc. The embodiment of the present application takes the clustering algorithm as an adaptive clustering learning (ACL) algorithm as an example for explanation.

[0040] It can be understood that the clustering label in the embodiment of the present application represents the operation mode to which the data point belongs, that is, the crawling strategy.

[0041] Step S103, performing feature learning on the cluster data to obtain a high-level feature vector, and generating a task template corresponding to the cluster data according to the high-level feature vector and the cluster label.

[0042] In step S103 of some embodiments, feature learning is performed on the cluster data to obtain a high-level feature vector, so that key features can be retained while removing redundant information in the data, thereby achieving the extraction of high-level, abstract feature representations, and generating a task template corresponding to the cluster data based on the high-level feature vector and the cluster label, so as to facilitate the subsequent control of the write-only ratio to execute corresponding actions based on relevant parameters of the task template.

[0043] Step S104, obtaining feedback information fed back by the robot arm when executing the task at the current moment.

[0044] In step S104 of some embodiments, feedback information fed back by the robot arm when performing the task at the current moment is obtained, that is, feedback information generated by the task currently being performed by the robot arm is obtained, so that it is possible to determine through the feedback information whether the robot arm's performance of the current task meets expectations, thereby facilitating subsequent optimization of task execution efficiency, and further dynamically adjusting the task according to the feedback of the robot arm, thereby improving production flexibility and adaptability.

[0045] It can be understood that the initial value of the feedback information in the embodiment of the present application is 0 or empty, and after the robotic arm performs each task, the embodiment of the present application will obtain the information fed back by the robotic arm when performing the task, so as to facilitate the subsequent dynamic adjustment of the task parameters.

[0046] Step S105, adjusting the task template according to the feedback information and the high-level feature vector, generating the target task at the next moment, and controlling the robotic arm according to the target task.

[0047] In step S105 of some embodiments, the task template is adjusted according to the feedback information and the high-level feature vector, so that the task parameters can be adjusted in real time in combination with the feedback of the robot arm, and the target task at the next moment is generated to achieve optimization of the target task, thereby ensuring that the system can flexibly respond to personalized customized tasks in a complex and dynamic environment, and control the robot arm according to the target task, thereby improving the flexibility and adaptability of production.

[0048] It is worth noting that the newly generated target task in the embodiment of the present application includes but is not limited to task ID, task type, target object characteristics and grasping parameters, wherein the task ID is a number that uniquely identifies the task, the task type is used to describe the type of task (grasping, moving, placing, etc.), the target object characteristics include but are not limited to the shape, size and position of the object, and the grasping parameters include but are not limited to grasping force, grasping method, etc.

[0049] In some embodiments, after controlling the robotic arm according to the target task, the embodiments of the present application will also obtain the target feedback information fed back by the robotic arm when performing the target task, and then fine-tune the task parameters of the target task based on the target feedback information, thereby realizing dynamic adjustment of the task parameters. Specifically, the task parameters of the target task are adjusted according to the target feedback information and high-level feature vectors to generate a new task, and then the robotic arm is controlled to execute the task parameters corresponding to the new task, so as to adapt to a variety of application scenarios, and can dynamically adjust the task according to the execution situation, thereby improving the response speed and adaptability of the system.

[0050] Reference Figure 3 , Figure 3 This is a flowchart of a specific method for grouping all sensor data based on a preset clustering algorithm provided in an embodiment of the present application. The method includes but is not limited to steps S201 to S204.

[0051] Step S201 , feature encoding is performed on all sensor data to obtain numerical representations corresponding to the sensor data, and feature vectors corresponding to the sensor data are constructed based on the numerical representations.

[0052] Step S202: Select a preset number of feature vectors from the sensor data as initial cluster centers.

[0053] Step S203: for each initial cluster center, determine the distance value between each sensor data and the initial cluster center.

[0054] Step S204 , allocating the sensor data according to the distance value to obtain a plurality of data groups, and allocating a label to each data group to obtain a cluster label corresponding to each data group.

[0055] In steps S201 to S204 of some embodiments, in the process of grouping all sensor data based on a preset clustering algorithm, the embodiments of the present application first perform feature encoding on all sensor data. Taking the sensor data including shape data and size data as an example, the embodiments of the present application will convert the shape and size into numerical representations, for example, converting a circle to 0, a square to 1, a triangle to 2, and so on; if the sensor data includes three size types: large, medium, and small, the small size is converted to 1, the medium size is converted to 2, and the large size is converted to 3 to obtain a numerical representation corresponding to the sensor data, so that the characteristics of the object can be accurately reflected through the numerical representation, and a feature vector corresponding to the sensor data is constructed based on the numerical representation, that is, a feature vector is constructed for each object, so that the feature vector includes all the information of the above numerical representations. Afterwards, the data is grouped through a clustering algorithm so that the data points in the same group have a high similarity, while the data points between different groups have a low similarity. Specifically, a preset number of feature vectors are selected from the sensor data as initial cluster centers. For each initial cluster center, the distance value between each sensor data and the initial cluster center is determined, and then the sensor data is allocated according to the distance value, and sensor data with relatively close distances are grouped together to obtain multiple data groups, so that each data point in the sensor data can be assigned to a specific cluster, and a label is assigned to each data group to obtain a cluster label corresponding to each data group, so as to identify the operation modes of different data groups, thereby simplifying the complexity of the data, wherein the cluster label is used to characterize the operation mode to which the cluster belongs, that is, the cluster label indicates that each data point is assigned to a specific cluster and the operation mode to which it belongs, so that the operation mode can be identified from the sensor data.

[0056] It should be noted that the preset number in the embodiment of the present application can be set according to the needs of the user. For example, the preset number is 3, 4, 5, etc., and the embodiment of the present application does not make any specific restrictions.

[0057] Specifically, the embodiment of the present application will select K data points as the initial cluster centers. In cluster analysis, K refers to the number of clusters, that is, the number of clusters into which the data set is divided. In the K-Means clustering algorithm, the user needs to pre-specify this parameter K so that the algorithm can divide the data into K different clusters. K determines the number of clusters that are ultimately generated. For example, if K=3, the algorithm will divide the data set into 3 clusters.

[0058] It can be understood that the specific formula for allocating sensor data according to distance value is as follows: ; in, is the set of the i-th cluster (data group), containing all the data points assigned to this cluster. : A data point in the dataset. is the center of the ith cluster. is the center of other clusters. To represent the data point x to the cluster center The square of the Euclidean distance.

[0059] Reference Figure 4 , Figure 4 This is a specific method flow chart of a robot arm control method in C2F mode provided by another embodiment of the present application, and the method includes but is not limited to steps S301 to S302.

[0060] It should be noted that step S301 to step S302 occur after the sensor data is distributed according to the distance value to obtain multiple data groups, and a label is allocated to each data group to obtain a cluster label corresponding to each data group.

[0061] Step S301, determining the data mean of all data points in each data group.

[0062] Step S302: Update the initial cluster center of the data group according to the data mean.

[0063] In steps S301 to S302 of some embodiments, after obtaining multiple data groups, the embodiments of the present application will also determine the data mean of all data points in each data group, and then update the initial cluster center of the data group according to the data mean to update the cluster center. By updating the cluster center to the mean of all data points in the cluster, the distance from each data point to its corresponding cluster center can be minimized, thereby reducing the intra-cluster distance and improving the compactness of the clustering.

[0064] It can be understood that updating the initial distance class center of the data group according to the data mean can be expressed as follows: ; in, is the set of the i-th cluster (data group), containing all the data points assigned to this cluster. represents the sum of all data points in the i-th cluster.

[0065] In some embodiments, after the initial cluster center of the data group is updated according to the data mean, for each updated initial cluster center, the embodiment of the present application will again determine the target distance value between each sensor data and the updated initial cluster center, and then distribute the sensor data according to the target distance value to obtain multiple target data groups, and assign a label to each target data group to obtain a cluster label corresponding to each target data group, so as to minimize the distance from each data point to its corresponding cluster center, realize the update of the cluster center, and further realize the accurate grouping of the data group.

[0066] Reference Figure 5 , Figure 5 It is a flowchart of a specific method for performing cluster evaluation on each data group according to each cluster label provided in an embodiment of the present application, and the method includes but is not limited to steps S401 to S405.

[0067] Step S401: for each data group, select a target data point in the data group.

[0068] Step S402, determining a first similarity between the target data point and other data points in the corresponding data group, and determining a second similarity between the target data point and data points in other data groups.

[0069] Step S403: determining a silhouette coefficient according to the first similarity and the second similarity.

[0070] Step S404: Compare the silhouette coefficient with a preset coefficient range to determine the target cluster number.

[0071] Step S405: adjusting the sensor data according to the target cluster number to obtain cluster data.

[0072] In steps S401 to S405 of some embodiments, in the process of performing clustering evaluation on each data group according to each clustering label, the embodiments of the present application use the silhouette coefficient to evaluate the clustering effect and measure the similarity between a data point and the cluster to which it belongs. Specifically, for each data group, a target data point is selected in the data group, and then a first similarity between the target data point and other data points in the corresponding data group is determined, that is, the similarity between the target data point and other data points in the data group to which it belongs, and a second similarity between the target data point and data points in other data groups is determined, that is, the similarity between the target data point and data points in data groups other than the data group to which it belongs. Then, a silhouette coefficient is determined based on the first similarity and the second similarity, so that the similarity between a data point and the cluster to which it belongs can be measured by the silhouette coefficient, and then the silhouette coefficient is compared with a preset coefficient range to determine the target number of clusters, so that the target number of clusters for each cluster can be determined. Then, the sensor data is adjusted according to the target number of clusters to obtain cluster data, so that the closeness of the data point in its own cluster and the separation in other clusters can be measured. By analyzing the silhouette coefficient, these points can be identified and processed to improve the accuracy of clustering.

[0073] It is worth noting that in the process of determining the silhouette coefficient according to the first similarity and the second similarity, the embodiment of the present application first determines the number of data points in the data group to which the target data point belongs, and then calculates the distance between the target data point and other data points in the data group to which it belongs, which is specifically expressed as follows: ; in, is the set of all points assigned to the same cluster. For clustering The number of points in , that is, the total number of data points belonging to this cluster. is the feature vector of data point i. is the feature vector of other points in the same cluster. is the distance between data point i and data point j, usually calculated using Euclidean distance.

[0074] In addition, the embodiment of the present application also determines the nearest neighbor cluster from the target data point, and determines the number of points in the nearest neighbor cluster and the feature vectors of all data points in the nearest neighbor cluster, which are specifically expressed as follows: ; in, is the nearest neighbor cluster The number of points in . is the eigenvector of all points in the nearest neighbor cluster.

[0075] After obtaining the above two parameters, the specific process of determining the silhouette coefficient in the embodiment of the present application is as follows: ; in, is the silhouette coefficient of the ith data point; It is the average distance from data point i to all other points in the same cluster. It represents the similarity between data point i and other points in its cluster, that is, the first similarity. It is the average distance from data point i to all points in the nearest neighbor cluster, which represents the similarity between data point i and other clusters, that is, the second similarity.

[0076] In some embodiments, the preset coefficient range can be set according to the needs of the user. Specifically, the embodiment of the present application takes the initial cluster number range of 2-10 as an example. In the process of comparing the silhouette coefficient with the preset coefficient range to determine the target cluster number, the silhouette coefficient is calculated for each cluster number, and the preset silhouette coefficient range is set to -1 to 1. The specific evaluation criteria are: when the silhouette coefficient is between 0.7-1.0, the clustering of the data group is determined to be a high-quality clustering; when the silhouette coefficient is between 0.5-0.7, the clustering of the data group is determined to be a reasonable clustering; when the silhouette coefficient is between 0.25-0.5, the clustering of the data group is determined to be a poor clustering; when the silhouette coefficient is less than 0.25, the clustering of the data group is determined to be an unacceptable clustering.

[0077] It is understandable that when the silhouette coefficient is less than 0.5, the number of clusters is increased and clustering is performed again. When the silhouette coefficient continues to stabilize at a high level, the number of clusters is reduced and the model is simplified. The embodiment of the present application records the results of each cluster evaluation. When the cluster results are stable for 3 to 5 consecutive times and the silhouette coefficient changes less than 0.1, it is considered that the current number of clusters is optimal.

[0078] Reference Figure 6 , Figure 6 It is a flowchart of a specific method for obtaining feedback information provided in an embodiment of the present application, and the method includes but is not limited to steps S501 to S502.

[0079] Step S501, recording the success rate, execution time and error rate fed back by the robot arm when executing the task at the current moment.

[0080] Step S502: Generate feedback information according to the success rate, execution time and error rate.

[0081] In steps S501 to S502 of some embodiments, in the process of obtaining feedback information, the embodiments of the present application will record the success rate, execution time and error rate fed back by the robot arm when performing the task at the current moment, wherein the success rate is the proportion of successful completion of the robot arm's tasks, the execution time is the time required for the robot arm to complete the current task, and the error rate is the proportion of errors that occur in the robot arm during the task execution process, for example, grasping failure, position deviation, etc., and then generate feedback information based on the success rate, execution time and error rate, so that it can be determined through the feedback information whether the robot arm's execution of the current task meets expectations, which is convenient for subsequent optimization of the task execution efficiency, and further can dynamically adjust the task according to the feedback of the robot arm to improve the flexibility and adaptability of production.

[0082] Reference Figure 7 , Figure 7 It is a flowchart of a specific method for performing feature learning on cluster data provided in an embodiment of the present application, and the method includes but is not limited to steps S601 to S603.

[0083] Step S601: reconstruct cluster data through a decoder to obtain reconstructed data.

[0084] Step S602, training the autoencoder by reconstructing the data and clustering the data to obtain a pre-trained autoencoder.

[0085] Step S603: input the clustering data into a pre-trained autoencoder, so that the pre-trained autoencoder extracts features from the clustering data and outputs a high-level feature vector.

[0086] In steps S601 to S603 of some embodiments, in the process of feature learning of cluster data, the embodiments of the present application first reconstruct the cluster data through a decoder to achieve data compression and decompression, and can retain key features while removing redundant information in the data to obtain reconstructed data. The autoencoder is then trained with the reconstructed data and the cluster data to obtain a pre-trained autoencoder, thereby improving the reconstruction capability of the autoencoder. Thereafter, the cluster data is input into the pre-trained autoencoder so that the pre-trained autoencoder performs feature extraction on the cluster data, so as to extract high-level features from complex sensor data, output high-level feature vectors, and achieve the extraction of high-level, abstract feature representations, which can retain the key information and intrinsic structure of the data.

[0087] It should be noted that the specific process of training the autoencoder by reconstructing data and clustering data is as follows: ; Where L is the reconstruction error, which represents the difference between the input data and the reconstructed data. X is the original input data. is the data reconstructed by the autoencoder, that is, the reconstructed data.

[0088] It is understandable that when training an autoencoder, the reconstruction error is calculated and recorded at each training cycle. During the training process, the reconstruction error is used as a loss function, and the optimization algorithm attempts to minimize the error to improve the reconstruction ability of the model. By monitoring the changes in the reconstruction error, the training effect of the model can be judged. If the error continues to decrease, it means that the model is learning the characteristics of the data; if the error no longer decreases, the learning rate or other hyperparameters may need to be adjusted. After the training is completed, the reconstruction error can be used to evaluate the performance of the model. A lower reconstruction error means that the model reconstructs the input data better, and vice versa, it means that the model performance is poor.

[0089] Reference Figure 8 , Figure 8 This is a flowchart of a specific method for adjusting a task template according to feedback information and a high-level feature vector provided in an embodiment of the present application. The method includes but is not limited to steps S701 to S703.

[0090] Step S701, compare the feedback information with the preset expected target to obtain an evaluation result.

[0091] Step S702: determining adjustment parameters according to the evaluation result and the high-level feature vector.

[0092] Step S703, modifying the parameters in the task template based on the adjustment parameters to generate the target task at the next moment.

[0093] In steps S701 to S703 of some embodiments, in the process of adjusting the task template according to the feedback information and the high-level feature vector, the embodiment of the present application compares the feedback information with the preset expected target to determine whether the various parameters in the feedback information meet the requirements, and obtains an evaluation result, and then determines the adjustment parameters based on the evaluation result and the high-level feature vector, so that the task parameters can be adjusted in real time in combination with the feedback of the robotic arm, and then the parameters in the task template are modified based on the adjustment parameters to achieve dynamic adjustment of the task parameters, generate the target task at the next moment, and optimize the target task, thereby ensuring that the system can flexibly respond to personalized customized tasks in a complex and dynamic environment.

[0094] It should be noted that the expected goal can be set according to the type of feedback information. Taking the feedback information including success rate, execution time and error rate as an example, the expected goal at this time includes the success rate target, execution time target and error rate target. Specifically, the success rate target is a success rate greater than 90%, the execution time target is an execution time of less than 3 seconds for each task, and the error rate target is an error rate less than 5%. The embodiments of this application do not impose specific restrictions.

[0095] It can be understood that in the process of comparing the feedback information with the preset expected goals, taking the success rate goal as a success rate greater than 90%, the execution time goal as the execution time of each task less than 3 seconds, and the error rate target as an error rate less than 5% as an example, when the success rate in the feedback information is detected to be 85%, which is lower than the success rate target, it is recorded as an area that needs improvement; when the current execution time is detected to be 4 seconds, which exceeds the execution time target, it is recorded as an area that needs improvement; when the current error rate is detected to be 6%, which exceeds the error rate target, it is recorded as an area that needs improvement. At this time, an evaluation result will be generated to facilitate the subsequent dynamic adjustment of task parameters.

[0096] Reference Fig. 9 , Fig. 9 It is a specific method flow chart of a robot arm control method in C2F mode provided by another embodiment of the present application, and the method includes but is not limited to step S801.

[0097] It should be noted that step S801 occurs before all sensor data are grouped based on a preset clustering algorithm to obtain multiple data groups.

[0098] Step S801 : performing a preprocessing operation on all sensor data, and performing a standardization process on the sensor data after the preprocessing operation, so as to scale all sensor data to a preset range.

[0099] In step S801 of some embodiments, before grouping all sensor data based on a preset clustering algorithm to obtain multiple data groups, the embodiments of the present application will also perform preprocessing operations on all sensor data, specifically, perform integrity detection on the sensor data to confirm whether the data is complete and whether there are missing values ​​or duplicate values, so that outliers can be removed or corrected to reduce noise, make the data more consistent with the actual distribution, and avoid a significant reduction in data volume due to deletion of missing values. The sensor data after the preprocessing operation is standardized to scale all sensor data to a preset range. Specifically, the embodiments of the present application will standardize the sensor data to a range of 0-1 and scale the data to the same range to eliminate the dimensional impact of different sensor data.

[0100] It is understandable that the embodiment of the present application takes the standardization of position data and force data to the range of 0-1 as an example for explanation, and the data standardization operation can be performed with reference to the following formula: ; in, is a certain feature value in the original data. is the mean of the feature, indicating the center location of the data. is the standard deviation of the feature, which indicates the degree of dispersion of the data. is the eigenvalue after standardization. Through standardization, the mean of the data becomes 0 and the standard deviation becomes 1, so that data with different characteristics can be compared on the same scale and the dimension effect can be eliminated.

[0101] In some embodiments, after scaling all sensor data to a preset range, the embodiments of the present application will group the scaled sensor data based on a preset clustering algorithm to obtain multiple data groups, thereby eliminating the dimensional effects of different features.

[0102] In some embodiments, during the process of preprocessing the sensor data, the embodiment of the present application will perform data cleaning operations and processing missing value operations on the sensor data to check abnormal or erroneous data. Specifically, during the data cleaning process, the embodiment of the present application uses statistical methods (such as mean, standard deviation, quartiles) to identify data points beyond the normal range, and then draws a box plot or scatter plot to visually check the abnormal values. For classified data, check whether the category distribution is reasonable and obtain the inspection result. When the inspection result is that the abnormal value is caused by data entry error or measurement error, it can be corrected to a reasonable value. If it cannot be corrected, you can choose to delete the abnormal value. In the process of processing missing value operations, the embodiment of the present application checks whether there are missing values ​​in the book. If there are missing values ​​in the sensor data, for numerical data, the missing values ​​can be filled with the mean, median or mode of the column; for time series data, the missing values ​​can be filled by linear interpolation, polynomial interpolation and other methods. Filling missing values ​​can retain more data points and avoid a significant reduction in data volume due to deletion of missing values.

[0103] See also Fig.10 , Fig.10 It is a structural schematic diagram of a robot arm control device in C2F mode provided by an embodiment of the present application.

[0104] In some embodiments, the present application also provides a robot arm control device in C2F mode, which is applied to a robot arm control device. The robot arm control device in C2F mode includes: The data acquisition module 901 is used to collect data from the robot in the robot control device to obtain multiple sensor data; The adaptive clustering module 902 is used to group all sensor data based on a preset clustering algorithm to obtain multiple data groups and a clustering label corresponding to each data group, and perform clustering evaluation on each data group according to each clustering label to obtain clustered data; The feature learning module 903 is used to perform feature learning on the cluster data to obtain a high-level feature vector, and generate a task template corresponding to the cluster data according to the high-level feature vector and the cluster label; The data feedback module 904 is used to obtain feedback information of the robot arm when performing the task at the current moment; The task generation module 905 is used to adjust the task template according to the feedback information and the high-level feature vector, generate the target task at the next moment, and control the robot arm according to the target task.

[0105] The specific implementation of the robot arm control is basically the same as the specific implementation of the robot arm control method in the above-mentioned C2F mode, and will not be repeated here.

[0106] In order to more clearly illustrate the above-mentioned robot arm control method in the C2F mode, a specific example is given below.

[0107] Example 1: Example 1 Figure 1 The robot arm control device 100 shown is taken as an example for explanation.

[0108] First, when the robot arm control device 100 is started, the sensors and actuators of the robot arm are configured, and initial parameters (such as the number of clusters, task templates, etc.) are set to ensure that the system can operate normally and is ready for data collection and task execution.

[0109] During the process of the robot arm performing the task, this example collects sensor data in real time through the data acquisition component 110, including position, speed, force, temperature, etc. Sufficient environment and state information is obtained for subsequent analysis and learning, and the execution status of the task is collected, including the success rate of grasping, execution time, etc.

[0110] In some embodiments, the robot arm control device in the C2F mode further includes a data preprocessing module and an adaptive adjustment module, and the different modules are described in detail below.

[0111] The data preprocessing module is also used to perform data preprocessing operations on sensor data. Specifically, the collected sensor data is standardized to eliminate the dimensional impact of different sensor data. This ensures that the data is compared on the same scale and improves the accuracy of subsequent learning and analysis.

[0112] After the data preprocessing module processes, the adaptive clustering module 902 uses an adaptive clustering learning algorithm to perform cluster analysis on the sensor data, identify the operation mode, initialize the cluster center and iteratively update it.

[0113] The adaptive adjustment module is used to evaluate the clustering effect according to the silhouette coefficient and dynamically adjust the number of clusters. After that, the feature learning module 903 uses the autoencoder to extract high-level features.

[0114] Next, the basic task templates are defined through the task generation component 140, such as grabbing, moving, placing, etc., and the feedback results are collected in real time through the data feedback component 130 to check the task execution status, compare the execution data with the expected goals, and generate evaluation results.

[0115] After obtaining the evaluation results, the embodiment of the present application adjusts the task parameters according to the evaluation results, and then generates the task based on the feature vector, cluster label and task parameters. Thereafter, the control instructions of the robotic arm are generated according to the task instructions. Finally, the control instructions are sent to the robotic arm so that the robotic arm works according to the expected goals.

[0116] It is worth noting that after each task is executed, the robot arm control device 100 continuously updates the adaptive clustering learning model and task generation using new data and feedback information to form a closed-loop learning system. Through continuous learning and optimization, it ensures that the system can adapt to new task requirements and environmental changes.

[0117] The embodiment of the present application forms a complete robotic arm control device 100 through steps such as data collection, adaptive clustering learning, dynamic task generation, real-time feedback and adjustment. The system can flexibly respond to personalized customized tasks in complex and dynamic environments, improve production efficiency and flexibility. Through continuous learning and optimization, the system can continuously adapt to new task requirements and environmental changes.

[0118] Reference Fig.11 , taking the control processor 1001 and memory 1002 in the electronic device 1000 as an example that can be connected via a bus. The memory 1002, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer executable programs. In addition, the memory 1002 may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one disk memory, a flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory 1002 may optionally include a memory remotely arranged relative to the control processor 1001, and these remote memories may be connected to the electronic device 1000 via a network. Examples of the above-mentioned networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0119] Those skilled in the art will understand that Fig.11 The device structure shown in the figure does not constitute a limitation on the electronic device 1000, and may include more or less components than shown in the figure, or combine certain components, or arrange the components differently.

[0120] The embodiment of the present application also provides a computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are executed by one or more control processors, for example, Fig.11 One of the control processors 1001 executes, which can enable the above-mentioned one or more control processors to execute the robot arm control method under the C2F mode in the above-mentioned method embodiment.

[0121] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0122] It will be appreciated by those skilled in the art that all or some of the steps and systems in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or transient medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, it is well known to those skilled in the art that communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

Claims

1. A robot arm control method in C2F mode, characterized in that: Applied to a robotic arm control device, the method comprises: Collecting data on the robot arm in the robot arm control device to obtain multiple sensor data; Grouping all the sensor data based on a preset clustering algorithm to obtain a plurality of data groups and a clustering label corresponding to each of the data groups, and performing clustering evaluation on each of the data groups according to each of the clustering labels to obtain clustered data; Performing feature learning on the cluster data to obtain a high-level feature vector, and generating a task template corresponding to the cluster data according to the high-level feature vector and the cluster label; Obtaining feedback information from the robot arm when performing the task at the current moment; The task template is adjusted according to the feedback information and the high-level feature vector, a target task at the next moment is generated, and the robotic arm is controlled according to the target task.

2. The robot arm control method in C2F mode according to claim 1, characterized in that: The step of grouping all the sensor data based on a preset clustering algorithm to obtain a plurality of data groups and a clustering label corresponding to each of the data groups includes: Performing feature encoding on all the sensor data to obtain numerical representations corresponding to the sensor data, and constructing feature vectors corresponding to the sensor data based on the numerical representations; Selecting a preset number of feature vectors from the sensor data as initial clustering centers; For each of the initial cluster centers, determining a distance value between each of the sensor data and the initial cluster center; The sensor data are distributed according to the distance values ​​to obtain a plurality of data groups, and a label is distributed to each of the data groups to obtain a cluster label corresponding to each of the data groups.

3. The robot arm control method in C2F mode according to claim 2, characterized in that: After allocating the sensor data according to the distance values ​​to obtain a plurality of data groups, and assigning a label to each of the data groups to obtain a cluster label corresponding to each of the data groups, the method further includes: Determining a data mean for all data points in each of the data sets; Updating the initial cluster center of the data group according to the data mean; The step of grouping all the sensor data based on a preset clustering algorithm to obtain a plurality of data groups and a clustering label corresponding to each of the data groups includes: For each updated initial cluster center, determining a target distance value between each of the sensor data and the updated initial cluster center; The sensor data is distributed according to the target distance value to obtain a plurality of target data groups, and a label is distributed to each of the target data groups to obtain a cluster label corresponding to each of the target data groups.

4. The robot arm control method in C2F mode according to claim 1, characterized in that: The step of performing cluster evaluation on each of the data groups according to each of the cluster labels to obtain cluster data includes: For each of the data groups, selecting a target data point in the data group; Determine a first similarity between the target data point and other data points in the corresponding data group, and determine a second similarity between the target data point and data points in the other data groups; determining a silhouette coefficient according to the first similarity and the second similarity; Comparing the silhouette coefficient with a preset coefficient range to determine a target number of clusters; The sensor data is adjusted according to the target cluster number to obtain cluster data.

5. The robot arm control method in C2F mode according to claim 1, characterized in that: The performing feature learning on the clustering data to obtain a high-level feature vector includes: Reconstructing the cluster data by a decoder to obtain reconstructed data; Training the autoencoder using the reconstructed data and the clustered data to obtain a pre-trained autoencoder; The clustering data is input into a pre-trained autoencoder, so that the pre-trained autoencoder performs feature extraction on the clustering data and outputs a high-level feature vector.

6. The robot arm control method in C2F mode according to claim 1, characterized in that: The feedback information is obtained by the following steps: Record the success rate, execution time and error rate of the task performed by the robot arm at the current moment; Feedback information is generated according to the success rate, the execution time and the error rate.

7. The robot arm control method in C2F mode according to claim 1, characterized in that: The step of adjusting the task template according to the feedback information and the high-level feature vector to generate a target task at the next moment includes: Comparing the feedback information with the preset expected target to obtain an evaluation result; determining adjustment parameters based on the evaluation results and the high-level feature vector; The parameters in the task template are modified based on the adjustment parameters to generate a target task at the next moment.

8. The robot arm control method in C2F mode according to claim 1 or 2, characterized in that: Before grouping all the sensor data based on a preset clustering algorithm to obtain a plurality of data groups, the method further includes: Performing a preprocessing operation on all the sensor data, and performing a standardization process on the sensor data after the preprocessing operation, so as to scale all the sensor data to a preset range; The method of grouping all the sensor data based on a preset clustering algorithm to obtain a plurality of data groups includes: The scaled sensor data is grouped based on a preset clustering algorithm to obtain a plurality of data groups.

9. A robot arm control device in C2F mode, characterized in that: Applied to a robot arm control device, the robot arm control device in the C2F mode includes: A data acquisition module, used to collect data from the robot in the robot control device to obtain multiple sensor data; An adaptive clustering module, used for grouping all the sensor data based on a preset clustering algorithm to obtain a plurality of data groups and a clustering label corresponding to each of the data groups, and performing clustering evaluation on each of the data groups according to each of the clustering labels to obtain clustered data; A feature learning module, used for performing feature learning on the cluster data to obtain a high-level feature vector, and generating a task template corresponding to the cluster data according to the high-level feature vector and the cluster label; A data feedback module is used to obtain feedback information fed back by the robot arm when performing the task at the current moment; The task generation module is used to adjust the task template according to the feedback information and the high-level feature vector, generate a target task at the next moment, and control the robotic arm according to the target task.

10. An electronic device, characterized in that: It includes at least one processor and a memory for communicating with the at least one processor; the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the robot arm control method under the C2F mode as described in any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the robot arm control method in the C2F mode as described in any one of claims 1 to 8.

Images