Remote intelligent control system of palletizing robot

Abstract

The invention provides a remote intelligent control system of a stacking robot, and relates to the technical field of robot control, a historical stacking log set is obtained through a historical log analysis module, scene clustering analysis is carried out, and a historical clustering stacking log set is obtained; the historical log supplementing module supplements historical data to obtain an updated historical cluster stacking log set; the stacking deviation analysis module is used for carrying out stacking deviation analysis and determining a stacking reliability factor; the retrieval matching module takes the task basic information set and the environment information set obtained by the information acquisition module as indexes to perform retrieval matching on the scene feature set to obtain a target stacking scene and a target stacking reliability factor, and the parameter search optimization module performs stacking control parameter search optimization to obtain target stacking control parameters. The technical problem that in the prior art, the control strategy of the palletizing robot cannot be adjusted in a self-adaptive mode according to changes of environments and tasks is solved, and the accuracy and reliability of palletizing operation are remarkably improved.

Description

Technical Field

[0001] The present application relates to the field of robot control technology, and in particular to a remote intelligent control system for a palletizing robot. Background Art

[0002] As a key equipment in automated production lines, palletizing robots are widely used in warehousing, logistics, production and other fields, undertaking heavy tasks such as cargo stacking and handling. Existing palletizing robot control methods or systems mainly focus on the execution of basic task instructions, usually based on preset palletizing modes and simple sensor feedback to perform specific palletizing tasks.

[0003] With the increasing complexity of industrial production environments, traditional control methods have gradually exposed many shortcomings. First, most of the existing palletizing robot control systems adopt static preset strategies and lack the ability to dynamically adapt to environmental changes and task diversity. For example, it is difficult for robots to adjust operating parameters in real time under different working environments or task conditions, resulting in low operating efficiency and accuracy in complex scenarios. Secondly, existing methods usually rely on real-time sensor data for decision-making, ignoring the accumulation and analysis of historical task data. At the same time, there is a lack of quantitative means to evaluate the reliability of palletizing, and the deviation problem in the robot's working process is not effectively considered, resulting in difficulty in accurately adjusting control parameters in application scenarios with high precision requirements, thereby affecting the palletizing quality and overall efficiency. Summary of the invention

[0004] The present application provides a remote intelligent control system for a palletizing robot, which solves the technical problem that the existing palletizing robot control system cannot adaptively adjust the control strategy according to changes in the environment and tasks, and achieves the technical effect of improving the accuracy and reliability of the palletizing operation of the palletizing robot in different production environments.

[0005] In view of the above problems, the present application provides a remote intelligent control system for a palletizing robot, the system comprising: a historical log analysis module, the historical log analysis module is used to obtain a historical palletizing log set of a target palletizing robot, perform scene clustering analysis on the historical palletizing log set, and obtain M historical clustered palletizing log sets of M palletizing scenes, wherein the M palletizing scenes include M scene feature sets, and M is an integer greater than or equal to 1; a historical log supplement module, the historical log supplement module is used to traverse the M historical clustered palletizing log sets to perform diversity analysis, obtain supplementary instructions according to the analysis results, and obtain the model information of the target palletizing robot, and supplement the M historical clustered palletizing log sets in combination with the supplementary instructions to obtain M updated historical clustered palletizing log sets; a palletizing deviation analysis module, the palletizing deviation analysis module is used to traverse the M updated historical clustered palletizing log sets to perform palletizing deviation analysis an information collection module, the information collection module is used to collect basic information of the target palletizing task based on the visual component of the target palletizing robot to obtain a task basic information set, and use the sensor component to collect environmental data of the execution place of the target palletizing task to obtain an environmental information set; a retrieval and matching module, the retrieval and matching module is used to retrieve and match the M scene feature sets with the task basic information set and the environmental information set as indexes to obtain a target palletizing scene, and match the M palletizing reliability factors according to the target palletizing scene to obtain a target palletizing reliability factor; a parameter search and optimization module, the parameter search and optimization module is used to search and optimize palletizing control parameters according to the target palletizing scene and the target palletizing reliability factor to obtain a target palletizing control parameter, and remotely control the target palletizing robot based on the target palletizing control parameter.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: The historical log analysis module obtains the historical palletizing log set of the target palletizing robot, performs scene clustering analysis on the historical palletizing log set, and obtains M historical clustered palletizing log sets of M palletizing scenes. By analyzing the historical logs, the working mode in different scenes can be identified and understood, and an accurate scene feature model can be formed, which provides a data basis for subsequent task decisions. The historical log supplement module traverses the M historical clustered palletizing log sets for diversity analysis, obtains supplementary instructions according to the analysis results, and obtains the model information of the target palletizing robot, and supplements the M historical clustered palletizing log sets in combination with the supplementary instructions to obtain M updated historical clustered palletizing log sets to enhance the diversity and richness of historical data. The integrity and applicability of the data can be improved by supplementing historical data, thereby enhancing the adaptability to different palletizing scenes and palletizing tasks. The palletizing deviation analysis module traverses the M updated historical clustered palletizing log sets for palletizing deviation analysis and determines M palletizing reliability factors. By evaluating the stability and error range of each scene, potential reliability problems can be identified, so that targeted measures can be taken to reduce errors and losses during the operation process. The information acquisition module collects basic information of the target palletizing task based on the visual component of the target palletizing robot to obtain a task basic information set, and uses the sensor component to collect environmental data of the execution site of the target palletizing task to obtain an environmental information set. The task basic information and environmental information provide key data for subsequent task matching and control optimization, ensuring that the robot can perceive the environment in real time and respond. The retrieval and matching module uses the task basic information set and the environmental information set as indexes to retrieve and match the M scene feature sets to find the target palletizing scene that best suits the current task, and matches the M palletizing reliability factors according to the target palletizing scene to obtain the target palletizing reliability factor, thereby selecting the optimal operation strategy for the current task and ensuring the stable performance of the robot in a complex environment. The parameter search and optimization module performs palletizing control parameter search and optimization according to the target palletizing scene and the target palletizing reliability factor to obtain the target palletizing control parameters, and remotely controls the target palletizing robot based on the target palletizing control parameters, thereby improving the control accuracy and efficiency of the palletizing robot, so that it can perform optimal operations under different working conditions.

[0007] In summary, this application realizes remote intelligent control of the palletizing robot by the coordinated operation of the above modules, comprehensive analysis of historical data, real-time information, and various environmental characteristics, and optimization of control parameters by quantifying the reliability factor of palletizing. The remote intelligent control system of the palletizing robot provided by this application not only improves the working efficiency and accuracy of the palletizing robot, but also enhances its stability and adaptability in complex production environments, reduces the need for manual intervention, and thus improves the automation and intelligence level of the production line.

[0008] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 A schematic diagram of the structure of a remote intelligent control system for a palletizing robot provided in an embodiment of the present application.

[0010] Figure 2 A schematic diagram of a process for performing scene cluster analysis on a historical palletizing log set in a remote intelligent control system of a palletizing robot provided in an embodiment of the present application.

[0011] Figure 3 A schematic diagram of a flow chart for determining M palletizing reliability factors in a remote intelligent control system of a palletizing robot provided in an embodiment of the present application.

[0012] Description of the accompanying drawings: historical log analysis module 10, historical log supplement module 20, palletizing deviation analysis module 30, information collection module 40, retrieval matching module 50, parameter search optimization module 60. DETAILED DESCRIPTION

[0013] The embodiment of the present application solves the technical problem that the existing control system of the palletizing robot cannot adaptively adjust the control strategy according to changes in the environment and tasks by providing a remote intelligent control system for the palletizing robot, thereby achieving the technical effect of improving the accuracy and reliability of the palletizing operation of the palletizing robot in different production environments.

[0014] like Figure 1 As shown, the embodiment of the present application provides a remote intelligent control system for a palletizing robot, the system comprising: The historical log analysis module 10 is used to obtain a historical palletizing log set of a target palletizing robot, perform scene clustering analysis on the historical palletizing log set, and obtain M historical clustered palletizing log sets of M palletizing scenes, wherein the M palletizing scenes include M scene feature sets, and M is an integer greater than or equal to 1.

[0015] Specifically, the target palletizing robot is the control object of the system described in the embodiment of the present application, and can be any robot that needs to perform a specific palletizing task. The historical log analysis module 10 extracts all historical palletizing log data of the target palletizing robot from a database or a log storage system. These historical palletizing logs record the specific information of the robot when performing palletizing operations in the past, including the environment, execution time, item characteristics, and control parameters of the palletizing robot for each task. The collected historical palletizing log data are summarized into a historical palletizing log set, and then a cluster analysis algorithm (such as K-means or a hierarchical clustering algorithm) performs scene clustering analysis on the historical palletizing log set, and classifies palletizing scenes with similar characteristics together to form different clusters. For example, if the palletizing of the same type of items is often handled within a certain time period, and the conditions such as the location and quantity of the palletizing are similar, then these situations may be clustered into a cluster. After clustering analysis, M clusters are obtained. Here, M is a positive integer, representing the total number of clusters obtained by the scene clustering analysis.

[0016] Each cluster is a historical cluster palletizing log set, corresponding to a palletizing scene. Each palletizing scene contains a scene feature set consisting of multiple scene features. These scene features describe the specific conditions of each palletizing task scene, including environmental conditions, task requirements, robot operations and other factors in the task scene, such as temperature, humidity, stacking method, item type, etc. For example, in the "heavy object stacking" scene, the scene feature set includes item weight (such as 20kg), item shape (such as a cuboid), workspace (such as stacking in a limited space), etc.

[0017] By performing scenario cluster analysis on a collection of historical palletizing logs, we can not only understand the execution modes in different scenarios, but also provide accurate scenario data support for subsequent task optimization.

[0018] The historical log supplement module 20 is used to traverse the M historical clustered palletizing log sets to perform diversity analysis, obtain supplementary instructions according to the analysis results, and obtain the model information of the target palletizing robot, and supplement the M historical clustered palletizing log sets in combination with the supplementary instructions to obtain M updated historical clustered palletizing log sets.

[0019] Specifically, the historical log supplement module 20 traverses M historical clustered palletizing log sets, analyzes the historical log data in each set, and counts the number of types of palletized objects in different logs, the types of palletizing environments (such as the distribution of different environments such as temperature and humidity), and the difference range of operating parameters (such as palletizing speed, grasping force, etc.), etc., to check whether the tasks in each scene cover enough environmental changes and operational differences. If it is found that most of the logs in a historical clustered palletizing log set are palletizing similar objects in similar environments, it indicates that the diversity of this set is low, and the historical log data of the corresponding task scene is not rich enough, and data supplementation is required.

[0020] The module generates corresponding supplementary instructions to increase the diversity of data based on the analysis results of diversity analysis of M historical clustered palletizing log sets. The supplementary instruction is an instruction derived from the diversity analysis results, which is used to guide the data supplementation of the historical clustered palletizing log set, and contains information such as the type of data to be supplemented, the source or the direction of supplementation. The supplementary instruction can be generated based on the set supplementary standards, such as the range, humidity, temperature and other conditions of the task environment parameters. For example, if it is found through diversity analysis that a historical clustered palletizing log set lacks palletizing logs in a high temperature environment, a supplementary instruction to supplement the high temperature environment log data is generated.

[0021] At the same time, the module obtains the model information of the target palletizing robot, such as load capacity, range of motion of the robot arm, accuracy, etc., and improves and adjusts the supplementary instructions according to the model information. Finally, the supplementary instructions are executed to obtain data from relevant data sources and supplement them to the corresponding historical clustered palletizing log sets, thereby obtaining M updated historical clustered palletizing log sets. These updated historical clustered palletizing log sets are new log sets obtained after the supplementary operation on the basis of the original historical clustered palletizing log sets. Compared with the original set, it contains more diverse task data.

[0022] By supplementing historical data, a more comprehensive and accurate collection of historical logs can be obtained, which helps to enhance the adaptability of the palletizing robot to diverse environments and tasks, and improves the accuracy of subsequent task optimization and decision-making.

[0023] The palletizing deviation analysis module 30 is used to traverse the M update history clustered palletizing log sets to perform palletizing deviation analysis and determine M palletizing reliability factors.

[0024] Specifically, palletizing deviation refers to the deviation between the actual palletizing position or method completed by the palletizing robot and the expected target when performing the palletizing task. For example, a box may be placed at a specific coordinate position, but due to robot control errors or environmental changes, the actual position is offset by a few centimeters or the angle changes. This deviation will affect the stability and efficiency of the stacking. The palletizing reliability factor is a set of indicators or parameters used to quantify the accuracy and reliability of the execution of the palletizing task.

[0025] The palletizing deviation analysis module 30 traverses the updated M historical clustered palletizing log sets one by one, and analyzes the deviations in the execution of the task in each scenario, such as deviations in the actual position of the palletizing, errors in the stacking order, damage to items, etc. By analyzing the degree of deviation during the execution process, the M palletizing reliability factors corresponding to the M historical clustered palletizing log sets can be obtained. The palletizing reliability factors are a set of indicators or parameters for quantifying the accuracy and reliability of the execution of the palletizing task. The value of the palletizing reliability factor reflects the quality of the task execution. The smaller the degree of deviation, the higher the palletizing reliability factor.

[0026] The palletizing reliability factor determined through palletizing deviation analysis provides feedback on execution accuracy and can be used to optimize the control strategy of the palletizing task and improve the reliability and stability of the robot's task execution.

[0027] The information collection module 40 is used to collect basic information of the target palletizing task based on the visual component of the target palletizing robot to obtain a task basic information set, and to collect environmental data of the execution location of the target palletizing task using a sensor component to obtain an environmental information set.

[0028] Specifically, the information acquisition module 40 observes and collects data on the target task in real time through the visual component of the target palletizing robot. The visual component usually includes a camera or a visual sensor (such as LiDAR, RGB-D camera, etc.), which can capture the real-time situation of the palletizing task, such as the location, shape, size, interval between objects, orientation of objects and other basic information related to the task. The basic information of the task captured by the visual component is summarized to obtain a set of basic information of the task.

[0029] At the same time, the module obtains information related to the task execution environment through the sensor components equipped by the target palletizing robot. The sensor components include various sensors such as temperature sensors, humidity sensors, and lidar sensors, which can obtain specific data on external environmental conditions and working areas, which will affect the execution of the palletizing task. The task execution environment information captured by the sensor components is summarized to obtain an environmental information set.

[0030] The task basic information set and the environmental information set provide important basis for the execution of palletizing tasks. Through these sets, the conditions and environment for task execution can be comprehensively evaluated and the strategy for task execution can be optimized.

[0031] The retrieval and matching module 50 is used to retrieve and match the M scene feature sets with the task basic information set and the environment information set as indexes to obtain a target palletizing scene, and match the M palletizing reliability factors according to the target palletizing scene to obtain a target palletizing reliability factor.

[0032] Specifically, the retrieval and matching module 50 obtains the basic task information and environmental information collected by the information acquisition module 40, and uses this information as an index to perform attention matching with the scene features in the M scene feature sets, and selects the target palletizing scene that best matches the current task. This target palletizing scene represents the most suitable working mode under the current palletizing task and environmental conditions, and the control parameter search optimization is subsequently performed based on this scene. Then, the corresponding palletizing reliability factor is matched from the M palletizing reliability factors according to the retrieved target palletizing scene. The M palletizing reliability factors correspond to different palletizing scenes, and each factor reflects the reliability of palletizing in the corresponding scene. Once the target palletizing scene is determined, the corresponding target palletizing reliability factor can be found from these M palletizing reliability factors.

[0033] Through scene matching, the most suitable working mode for the current task can be selected. Furthermore, the corresponding palletizing reliability factor is matched according to the target palletizing scene, so as to evaluate the execution accuracy of the task in the scene, thereby providing direction and basis for subsequent control parameter optimization and improving palletizing efficiency and accuracy.

[0034] The parameter search optimization module 60 is used to search and optimize the palletizing control parameters according to the target palletizing scenario and the target palletizing reliability factor, obtain the target palletizing control parameters, and remotely control the target palletizing robot based on the target palletizing control parameters.

[0035] Specifically, the parameter search optimization module 60 obtains the target palletizing scene and the target palletizing reliability factor, searches in the historical palletizing log set according to the target palletizing scene, and determines the historical palletizing control parameters under the corresponding scene, including moving speed, gripping force, stacking angle, path planning, etc. Then, the historical palletizing control parameters are optimized according to the target palletizing reliability factor to determine the target palletizing control parameters that are more suitable for the current target palletizing scene and have higher control accuracy. The target palletizing control parameters are transmitted to the target palletizing robot through remote control means (such as wireless network, cloud platform, control terminal, etc.), and the target palletizing robot is controlled to perform the actual palletizing task according to these parameters. Ensure that the target palletizing robot can accurately perform tasks and improve work efficiency and palletizing quality.

[0036] Further, such as Figure 2 As shown, the historical log analysis module 10 described in the embodiment of the present application is also used to perform the following steps: Step P11: Utilize a scene feature extractor to perform scene feature analysis on the historical palletizing log set to obtain a historical palletizing scene feature cluster, wherein each historical palletizing log corresponds to a historical palletizing scene feature set.

[0037] Step P12: Taking the preset scene feature similarity as the cluster analysis scale, a historical palletizing scene feature set is randomly selected from the historical palletizing scene feature cluster without replacement for cluster analysis, and a first historical clustered palletizing log set is obtained according to the clustering result.

[0038] Step P13: performing mean calculation on a plurality of historical palletizing scene feature sets corresponding to the first historical clustered palletizing log set to determine a first scene feature set.

[0039] Step P14: remove multiple historical palletizing scene feature sets corresponding to the first historical clustered palletizing log set from the historical palletizing scene feature cluster, and perform scene clustering analysis on the removed historical palletizing scene feature cluster until all historical palletizing logs in the historical palletizing scene feature cluster are divided into corresponding historical clustered palletizing log sets, thereby obtaining M historical clustered palletizing log sets and M scene feature sets.

[0040] Step P15: Perform semantic recognition on the M scene feature sets to determine M palletizing scenes.

[0041] Specifically, the scene feature extractor is a tool or algorithm used to extract scene features such as the environment features, item features, and operation features of palletizing from the historical palletizing log set. The scene feature extractor is used to analyze the historical palletizing log set, extract the historical palletizing scene feature set corresponding to each historical palletizing log, and summarize these historical palletizing scene feature sets into historical palletizing scene feature clusters. These feature sets usually include key data points in the log, such as execution time, equipment status, working environment, item size, palletizing method, etc.

[0042] When performing cluster analysis, a pre-set measurement standard, namely the preset scene feature similarity, is used to judge the similarity between different historical palletizing scene feature sets, which is used as the scale of cluster analysis. A historical palletizing scene feature set (referred to as set A) is randomly extracted from the historical palletizing scene feature cluster without replacement for cluster analysis. Set A will not be drawn again in the subsequent extraction process. Use the clustering algorithm to traverse all historical palletizing scene feature sets in the historical palletizing scene feature cluster except set A, and group the historical palletizing scene feature sets whose similarity with the scene features of set A meets the preset scene feature similarity into the same cluster, and obtain a clustering result, which is recorded as the first historical clustered palletizing log set.

[0043] Each scene feature dimension in all historical palletizing scene feature sets contained in the first historical clustered palletizing log set is averaged to obtain the feature mean of each feature dimension, and then all feature means are summarized to construct the first scene feature set. The first scene feature set represents the typical features of the first scene corresponding to the first historical clustered palletizing log set.

[0044] The historical palletizing scene feature set in the divided first historical clustered palletizing log set is removed from the historical scene feature cluster, and then steps P12 to P13 are repeated to continue clustering analysis on the remaining scene feature sets. After each clustering, the processed clusters need to be removed until all historical palletizing scene feature sets in the historical scene feature cluster are clustered and divided, and finally M clustering results and M mean processing results are obtained, namely, M historical clustered palletizing log sets and M scene feature sets.

[0045] The semantic analysis algorithm in natural language processing technology is used to perform semantic recognition on the M scene feature sets obtained, so as to determine the palletizing scene corresponding to each scene feature set. For example, the scene feature set contains descriptive information such as "palletizing in a high temperature environment" and "palletizing of fragile items". The semantic recognition algorithm is used to understand and analyze this information and convert it into a semantic representation that can be understood by the computer, so as to determine M palletizing scenes such as high temperature palletizing scene and fragile item palletizing scene.

[0046] Furthermore, step P12 further includes: Step P121: randomly extracting a historical palletizing scene feature set from the historical palletizing scene feature cluster without replacement as the first historical palletizing scene feature set.

[0047] Step P122: using the cosine similarity calculation formula to respectively calculate the similarity between the historical palletizing scene feature set in the historical palletizing scene feature cluster and the first historical palletizing scene feature set, to obtain a plurality of similarity recognition results.

[0048] Step P123: Determine whether the multiple similarity recognition results are greater than or equal to the cluster analysis scale; if so, add the corresponding historical palletizing scene feature set into the first cluster historical palletizing scene feature set.

[0049] Step P124: according to the one-to-one mapping relationship between the historical palletizing logs and the historical palletizing scene feature sets, the first historical clustered palletizing log set is acquired according to the first clustered historical palletizing scene feature set mapping.

[0050] Specifically, a historical palletizing scene feature set is randomly extracted from the historical palletizing scene feature cluster without replacement as the first historical palletizing scene feature set. This set will serve as the starting point for subsequent cluster analysis, i.e., the initial cluster center. The cosine similarity calculation formula is used to calculate the similarity between each historical palletizing scene feature set in the historical palletizing scene feature cluster and the first historical palletizing scene feature set, and multiple similarity recognition results are obtained to evaluate the similarity between different scenes.

[0051] Compare the calculated multiple similarity recognition results with the preset scene feature similarity (i.e., cluster analysis scale). Add the historical palletizing scene feature set whose similarity recognition result is greater than or equal to the cluster analysis scale to the first cluster historical palletizing scene feature set to form a new cluster. Exemplarily, the cluster analysis scale is set to 0.6, and 5 similarity recognition results are obtained through similarity analysis, which are 0.7, 0.5, 0.8, 0.4, and 0.6, respectively. Then, the historical palletizing scene feature sets corresponding to the 1st, 3rd, and 5th similarity recognition results are added to the first cluster historical palletizing scene feature set.

[0052] Since there is a one-to-one mapping relationship between historical palletizing logs and historical palletizing scene feature sets, the corresponding historical palletizing logs are found through this mapping relationship according to the historical palletizing scene feature sets in the first clustered historical palletizing scene feature set, and these historical palletizing logs are combined to form the first historical clustered palletizing log set.

[0053] Through the above process, similar palletizing scenarios can be extracted from a large number of historical logs, thereby providing reliable historical data support for subsequent palletizing task optimization and decision-making.

[0054] Furthermore, the historical log supplement module 20 of the embodiment of the present application is also used to perform the following steps: Step P21: extracting a first historical clustered stacking log set from the M historical clustered stacking log sets.

[0055] Step P22: Enumerate the first historical clustered palletizing log sets in pairs to obtain multiple enumeration combinations.

[0056] Step P23: Perform similarity identification based on the multiple enumerated combinations to determine the similarities of multiple combinations.

[0057] Step P24: Count the number of the plurality of combinations whose similarities are greater than or equal to the preset combination similarity, and compare the statistical result with the total number of the plurality of enumerated combinations to obtain a similarity ratio.

[0058] Step P25: Determine whether the similarity ratio is greater than or equal to a preset similarity ratio, and if so, obtain a supplementary instruction.

[0059] Specifically, the first historical clustered palletizing log set is extracted from the M historical clustered palletizing log sets, and the historical palletizing scene feature sets in the selected first historical clustered palletizing log set are enumerated in pairs to form multiple enumeration combinations. For example, the selected log set contains three historical palletizing scene feature sets: set A, set B, and set C. After two-by-two enumeration, the following combinations will be formed: AB, AC, and BC. For each enumeration combination, the similarity between the two historical palletizing scene feature sets contained in each combination is calculated by a similarity calculation method (such as cosine similarity, Pearson correlation coefficient, etc.), and multiple combination similarities are obtained. A preset combination similarity value is pre-set, and the preset combination similarity value is the upper limit of the similarity between the two historical palletizing scene feature sets. For each combination similarity, the size of its similarity with the preset combination is compared, and the number of all enumeration combinations whose combination similarity is greater than or equal to the preset combination similarity is counted, and the statistical result is compared with the total number of multiple enumeration combinations to obtain a similarity ratio. This ratio can reflect whether the data in the historical clustered palletizing log set has a high consistency. Determine whether the calculated similarity ratio is greater than or equal to the preset similarity ratio. If so, it indicates that the historical palletizing scene feature set within the cluster has a high consistency, but the data diversity is insufficient. At this time, the module generates supplementary instructions to supplement the data for the historical cluster palletizing log set. Among them, the preset similarity ratio is a pre-set similarity ratio limit, which is used to determine whether the similarity ratio meets the requirements, thereby determining whether to obtain supplementary instructions. If the similarity is too high, it means that the samples in this scenario are not rich enough, and it is easy to fall into the local optimal solution during subsequent control analysis, and cannot face diversified application scenarios.

[0060] By enumerating and similarity analyzing the historical clustered palletizing log sets multiple times and calculating the similarity ratio, it is decided whether it is necessary to supplement the historical clustered palletizing log sets, thereby ensuring the diversity and richness of historical data and improving the adaptability of the palletizing robot to different environments.

[0061] Further, such as Figure 3 As shown, the stacking deviation analysis module 30 described in the embodiment of the present application is also used to perform the following steps: Step P31: extracting palletizing deviation data based on the M update history clustered palletizing log sets to obtain M palletizing deviation data sets.

[0062] Step P32: Calculate the mean of the M palletizing deviation data sets to obtain the M palletizing deviation mean values.

[0063] Step P33: Taking the mean values ​​of the M palletizing deviations as a starting point, the M palletizing deviation data sets are iteratively analyzed according to a preset deviation analysis step length to determine M concentrated palletizing deviation data.

[0064] Step P34: respectively taking the ratio of the M concentrated palletizing deviation data to the sum of the M concentrated palletizing deviation data as the M palletizing reliability factors.

[0065] Specifically, M update history clustered palletizing log sets are traversed to extract the palletizing deviation data in each update history clustered palletizing log set, including the offset of the palletizing position relative to the standard position, the deviation of the palletizing angle from the standard angle, etc. The palletizing deviation data extracted from each update history clustered palletizing log set are summarized to form M palletizing deviation data sets.

[0066] For each of the M palletizing deviation data sets, the average value of all the data is calculated (the sum of all deviation data divided by the number of data points), and M palletizing deviation means are obtained. These palletizing deviation means reflect the average deviation size of the palletizing task execution.

[0067] Taking each palletizing deviation mean as the starting point, the corresponding palletizing deviation data set is iteratively analyzed according to the preset deviation analysis step. In each analysis, the current palletizing deviation mean is centered and the preset deviation analysis step is the radius. The data within the radius is searched in the palletizing deviation data set, the current data concentration density is calculated, and the current palletizing deviation mean is iterated according to the data concentration density. After step-by-step iterative analysis, M concentrated palletizing deviation data corresponding to M palletizing deviation data sets are finally obtained. These concentrated palletizing deviation data are relatively concentrated deviation data that can be gathered and reflected in multiple iterative analyses, and can usually more accurately describe the most common or important deviations in palletizing tasks.

[0068] Each concentrated palletizing deviation data is compared with the sum of the previous M concentrated palletizing deviation data to obtain the palletizing reliability factor corresponding to each updated history clustered palletizing log set.

[0069] Through the above-mentioned palletizing deviation analysis process, the execution effect of the palletizing task can be accurately analyzed and evaluated, providing direction and basis for subsequent control parameter optimization.

[0070] Furthermore, step P33 also includes: Step P331: Taking the M palletizing deviation means as M starting points, iterate the M palletizing deviation data sets in any direction according to the preset deviation analysis step length to obtain M iterative palletizing deviation data and M first iterative directions, wherein the first iterative direction is the direction in which the palletizing deviation mean points to the iterative palletizing deviation data.

[0071] Step P332: Determine whether the concentration density of the M palletizing deviation means is greater than or equal to the concentration density of the M iterative palletizing deviation data. If not, update the M iterative palletizing deviation data to M starting points, and continue to iterate in the M palletizing deviation data sets according to the M first iteration directions until the preset number of iterations is met, and use the M iterative palletizing deviation data obtained in the last iteration as the M concentrated palletizing deviation data.

[0072] Furthermore, step P332 also includes: If so, the M first iteration directions are added into the M iteration taboo spaces, with the M palletizing deviation means as M starting points, and the M palletizing deviation data sets are iterated in directions other than the iteration directions stored in the M iteration taboo spaces according to a preset deviation analysis step.

[0073] Specifically, the specific process of iterative analysis of M palletizing deviation data sets is as follows: for each palletizing deviation mean, take it as the starting point, and iterate in any direction in the corresponding palletizing deviation data set according to the preset deviation analysis step length. Exemplarily, each palletizing deviation data set is regarded as a set of data points in a multidimensional space (such as coordinate points on a two-dimensional plane, each coordinate point represents the deviation data of a palletizing operation), starting from the palletizing deviation mean, with the preset deviation analysis step length as the step length, move and explore in different directions (such as up, down, left, right, diagonal, etc.). The palletizing deviation data obtained after the movement is recorded as the iterative palletizing deviation data, and the direction from the palletizing deviation mean to the iterative palletizing deviation data is recorded as the first iteration direction.

[0074] Use the k-nearest neighbor algorithm or Gaussian distribution model to calculate the concentration density of each pallet deviation mean and the corresponding iterative pallet deviation data, and compare the two calculated density values ​​to determine which density is higher. The higher density data can better represent the degree of pallet deviation in the scene corresponding to the entire set. If the concentration density of the pallet deviation mean is less than the concentration density of the corresponding iterative pallet deviation data, then update the corresponding iterative pallet deviation data to a new starting point, and continue to iterate in the corresponding pallet deviation data set according to the first iteration direction recorded previously.

[0075] If the concentration density of the palletizing deviation from the mean is greater than or equal to the concentration density of the iterative palletizing deviation data, this indicates that the first iterative direction has failed to achieve the expected effect, and the first iterative direction is added to the iterative taboo space. In the next iteration, this direction will no longer be used to avoid redundant calculations and falling into a local optimal solution. In subsequent iterations, the palletizing deviation from the mean is still used as the starting point, and the palletizing deviation data set is iterated in directions other than the iterative directions stored in the iterative taboo space according to the preset deviation analysis step. For example, if a direction does not meet the requirements and is added to the iterative taboo space in the first iteration, this direction will no longer be used for iteration in the second iteration, but other allowed directions will be selected for iteration. Among them, the iterative taboo space refers to a space composed of certain directions or paths that are avoided from being used in iterative analysis. Whenever an iterative direction fails to achieve the expected effect, the direction will be recorded as a taboo direction to avoid continued use in subsequent iterations.

[0076] The above iterative process is repeated continuously. After each iteration, the concentration density of the current palletizing deviation from the mean is compared with the concentration density of the iterative palletizing deviation data. Then, the iterative direction is selected until the preset number of iterations is met. The iterative palletizing deviation data obtained in the last iteration is used as the concentrated palletizing deviation data.

[0077] After multiple iterations and adjustments, M representative concentrated palletizing deviation data are finally obtained to provide a more reliable basis for subsequent control optimization.

[0078] Furthermore, the parameter search optimization module 60 described in the embodiment of the present application is also used to perform the following steps: Step P61: Obtain matching palletizing control parameters based on the target palletizing scenario.

[0079] Step P62: Utilizing a parameter optimization controller to optimize the matching palletizing control parameters and the target palletizing reliability factor to obtain the target palletizing control parameters.

[0080] Specifically, according to the target palletizing scenario, the palletizing control parameters matching the scenario are searched from the historical palletizing logs, i.e., matching palletizing control parameters. These parameters include the movement speed of the palletizing robot arm, the gripping force, the arrangement of the palletizing, and other related parameters.

[0081] The parameter optimization controller is a tool or algorithm for optimizing palletizing control parameters. It can adjust and optimize the matching palletizing control parameters according to the target palletizing reliability factor using optimization algorithms such as genetic algorithm, particle swarm optimization algorithm, gradient descent method, etc., to find the optimal parameter combination. The matching palletizing control parameters and the target palletizing reliability factor are input into the parameter optimization controller. The parameter controller continuously adjusts the matching palletizing control parameters through the optimization algorithm to improve the target palletizing reliability factor, and finally outputs the target palletizing control parameters to improve the execution accuracy and quality of the palletizing task.

[0082] In summary, the remote intelligent control system of a palletizing robot provided in the embodiment of the present application has the following technical effects: The historical log analysis module 10 performs scene cluster analysis on the historical palletizing log set of the target palletizing robot, identifies and extracts different palletizing scenes and their feature sets. Through cluster analysis, the historical data is divided into several scenes, each of which represents a typical palletizing task mode. By analyzing the historical logs, the working mode under different scenes can be identified and understood, and an accurate scene feature model can be formed to provide a data basis for subsequent task decisions. The historical log supplement module 20 traverses the M historical clustered palletizing log sets for diversity analysis, obtains supplementary instructions according to the analysis results, and obtains the model information of the target palletizing robot, and supplements the M historical clustered palletizing log sets in combination with the supplementary instructions to obtain M updated historical clustered palletizing log sets. The diversity of existing historical data can be enhanced by supplementing historical data, so that the palletizing robot can cope with more changing situations, avoid data deviations or limitations, and ensure wider scene applicability. The palletizing deviation analysis module 30 calculates the palletizing reliability factor of each scene by analyzing the degree of deviation of the palletizing operation under different scenes. Deviation analysis helps to identify potential problems and errors in various scenarios, thereby providing an accurate reference for subsequent optimization work and ensuring that the robot can stably perform tasks in complex environments. The information acquisition module 40 collects basic information on the target palletizing task based on the visual component of the target palletizing robot to obtain a task basic information set, and uses the sensor component to collect environmental data on the execution location of the target palletizing task to obtain an environmental information set. The task basic information and environmental information provide key data for subsequent task matching and control optimization, ensuring that the robot can perceive the environment and respond in real time. The retrieval and matching module 50 retrieves and matches the appropriate palletizing scene based on the collected task basic information and environmental data. After finding the best target palletizing scene, the corresponding target palletizing reliability factor is further matched to provide a clear decision-making basis for the subsequent control strategy optimization and ensure the stable performance of the robot in a complex environment. The parameter search optimization module 60 optimizes the palletizing control parameters according to the matched target palletizing scene and reliability factor, and remotely controls the target palletizing robot based on the target palletizing control parameters, thereby improving the control accuracy and efficiency of the palletizing robot and enabling it to perform optimal operations under different working conditions.

[0083] In general, the embodiment of the present application realizes remote intelligent control of the palletizing robot by quantifying the reliability factor of palletizing and optimizing the control parameters through the coordinated operation of the above modules, comprehensively analyzing the historical data, real-time information, and various environmental characteristics. The embodiment of the present application not only improves the working efficiency, palletizing accuracy and reliability of the palletizing robot, but also enhances its stability and adaptability in complex production environments, reduces manual intervention, thereby improving the automation and intelligence level of the production line and better meeting the palletizing needs in industrial production.

[0084] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A remote intelligent control system for a palletizing robot, characterized in that: The system comprises: A historical log analysis module, the historical log analysis module is used to obtain a historical palletizing log set of a target palletizing robot, perform scene clustering analysis on the historical palletizing log set, and obtain M historical clustered palletizing log sets of M palletizing scenes, wherein the M palletizing scenes include M scene feature sets, and M is an integer greater than or equal to 1; A historical log supplement module, the historical log supplement module is used to traverse the M historical clustered palletizing log sets to perform diversity analysis, obtain supplementary instructions according to the analysis results, and obtain the model information of the target palletizing robot, and supplement the M historical clustered palletizing log sets in combination with the supplementary instructions to obtain M updated historical clustered palletizing log sets; A palletizing deviation analysis module, the palletizing deviation analysis module is used to traverse the M update history clustered palletizing log sets to perform palletizing deviation analysis and determine M palletizing reliability factors; An information collection module, the information collection module is used to collect basic information of the target palletizing task based on the visual component of the target palletizing robot to obtain a task basic information set, and to collect environmental data of the execution location of the target palletizing task using a sensor component to obtain an environmental information set; A retrieval and matching module is used to retrieve and match the M scene feature sets with the task basic information set and the environment information set as indexes to obtain a target palletizing scene, and to match the M palletizing reliability factors according to the target palletizing scene to obtain a target palletizing reliability factor: A parameter search optimization module is used to perform a palletizing control parameter search optimization according to the target palletizing scenario and the target palletizing reliability factor, obtain the target palletizing control parameters, and remotely control the target palletizing robot based on the target palletizing control parameters.

2. The system according to claim 1, characterized in that The historical log analysis module is used to obtain a historical stacking log set of a target stacking robot, and perform scene cluster analysis on the historical stacking log set. The execution steps include: Using a scene feature extractor to perform scene feature analysis on the historical palletizing log set to obtain a historical palletizing scene feature cluster, wherein each historical palletizing log corresponds to a historical palletizing scene feature set; Taking the preset scene feature similarity as the cluster analysis scale, randomly extracting a historical palletizing scene feature set from the historical palletizing scene feature cluster without replacement for cluster analysis, and obtaining a first historical clustered palletizing log set according to the clustering result; Performing mean calculation on a plurality of historical palletizing scene feature sets corresponding to the first historical clustered palletizing log set to determine a first scene feature set; Eliminate multiple historical palletizing scene feature sets corresponding to the first historical clustered palletizing log set from the historical palletizing scene feature cluster, and perform scene clustering analysis on the eliminated historical palletizing scene feature cluster until all historical palletizing logs in the historical palletizing scene feature cluster are divided into corresponding historical clustered palletizing log sets, thereby obtaining M historical clustered palletizing log sets and M scene feature sets; Semantic recognition is performed on the M scene feature sets to determine M palletizing scenes.

3. The system according to claim 2, characterized in that Taking the preset scene feature similarity as the cluster analysis scale, a historical palletizing scene feature set is randomly selected from the historical palletizing scene feature cluster without replacement for cluster analysis, and a first historical clustered palletizing log set is obtained according to the clustering result. The execution steps include: Randomly extracting a historical palletizing scene feature set from the historical palletizing scene feature cluster without replacement as a first historical palletizing scene feature set; The similarity between the historical palletizing scene feature set in the historical palletizing scene feature cluster and the first historical palletizing scene feature set is calculated using a cosine similarity calculation formula to obtain a plurality of similarity recognition results; Determine whether the multiple similarity recognition results are greater than or equal to the cluster analysis scale, and if so, add the corresponding historical palletizing scene feature set into the first cluster historical palletizing scene feature set; And according to the one-to-one mapping relationship between the historical palletizing log and the historical palletizing scene feature set, the first historical clustered palletizing log set is acquired according to the first clustered historical palletizing scene feature set mapping.

4. The system according to claim 1, characterized in that The historical log supplement module is used to traverse the M historical clustered palletizing log sets to perform diversity analysis, obtain supplementary instructions according to the analysis results, and obtain the model information of the target palletizing robot, and supplement the M historical clustered palletizing log sets in combination with the supplementary instructions to obtain M updated historical clustered palletizing log sets. The execution steps include: Extracting a first historical clustered stacking log set from the M historical clustered stacking log sets; Enumerate the first historical clustered stacking log sets in pairs to obtain multiple enumeration combinations; Perform similarity identification based on the multiple enumerated combinations to determine multiple combination similarities; Counting the number of the plurality of combinations whose similarities are greater than or equal to the preset combination similarity, and comparing the statistical result with the total number of the plurality of enumerated combinations to obtain a similarity ratio; It is determined whether the similarity ratio is greater than or equal to a preset similarity ratio, and if so, a supplementary instruction is obtained.

5. The system according to claim 1, wherein: The palletizing deviation analysis module is used to traverse the M update history clustered palletizing log sets to perform palletizing deviation analysis and determine M palletizing reliability factors. The execution steps include: Extracting palletizing deviation data based on the M update history clustered palletizing log sets to obtain M palletizing deviation data sets; Calculate the mean of the M stacking deviation data sets to obtain the M stacking deviation mean values; Taking the mean value of the M palletizing deviations as a starting point, performing iterative analysis on the M palletizing deviation data sets according to a preset deviation analysis step length to determine M concentrated palletizing deviation data; The ratio of the M concentrated palletizing deviation data to the sum of the M concentrated palletizing deviation data is respectively used as the M palletizing reliability factors.

6. The system according to claim 5, characterized in that Taking the M palletizing deviation mean as the starting point, performing iterative analysis on the M palletizing deviation data sets according to a preset deviation analysis step length to determine M concentrated palletizing deviation data, the execution steps include: Taking the M palletizing deviation mean values ​​as M starting points, iterating the M palletizing deviation data sets in any direction according to a preset deviation analysis step length, to obtain M iterative palletizing deviation data and M first iterative directions, wherein the first iterative direction is the direction in which the palletizing deviation mean value points to the iterative palletizing deviation data; It is determined whether the concentration density of the M palletizing deviation means is greater than or equal to the concentration density of the M iterative palletizing deviation data. If not, the M iterative palletizing deviation data are updated to M starting points, and the iteration is continued in the M palletizing deviation data sets according to the M first iteration directions until the preset number of iterations is met, and the M iterative palletizing deviation data obtained in the last iteration are used as the M concentrated palletizing deviation data.

7. The system according to claim 6, characterized in that Taking the M palletizing deviation mean as the starting point, performing iterative analysis on the M palletizing deviation data sets according to a preset deviation analysis step length to determine M concentrated palletizing deviation data, the execution step further includes: If so, the M first iteration directions are added into the M iteration taboo spaces, with the M palletizing deviation means as M starting points, and the M palletizing deviation data sets are iterated in directions other than the iteration directions stored in the M iteration taboo spaces according to a preset deviation analysis step.

8. The system of claim 1, wherein: The parameter search and optimization module is used to perform a palletizing control parameter search and optimization according to the target palletizing scenario and the target palletizing reliability factor to obtain the target palletizing control parameter. The execution steps include: Obtaining matching palletizing control parameters based on the target palletizing scenario; The matching palletizing control parameters and the target palletizing reliability factor are optimized by using a parameter optimization controller to obtain the target palletizing control parameters.

Images