Spot welding robot path determination method and device, computer equipment, storage medium and computer program product

Abstract

The invention relates to a spot welding robot path determination method and device, computer equipment, a storage medium and a computer program product. The method comprises the steps that welding spot information of a to-be-processed welding spot and the number of spot welding robots are obtained; according to the welding spot information, the to-be-processed welding spots are clustered, and a welding spot set corresponding to the number of the robots is obtained; welding spot sets corresponding to all the spot welding robots are determined from the welding spot sets, and position information of all the spot welding robots is determined; the position information of the to-be-processed welding spots in the welding spot sets corresponding to the spot welding robots and the position information of the spot welding robots are input into a pre-trained welding spot processing path prediction model, and welding spot processing paths of the spot welding robots are obtained; and the welding spot processing path of each spot welding robot is used for representing the spot welding processing path of each spot welding robot on the to-be-processed welding spots in the corresponding welding spot set. By adopting the method, the determination efficiency of the spot welding robot path can be improved.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to the Chinese invention patent application filed with the State Intellectual Property Office on July 12, 2024, with application number 202410932007.0 and application name “Spot welding robot path determination method, device, computer equipment, storage medium and computer program product”, the entire contents of which are incorporated by reference into this application. Technical Field

[0003] The present application relates to the field of computer technology, and in particular to a method, device, computer equipment, computer-readable storage medium, and computer program product for determining a spot welding robot path. Background Art

[0004] In the production process, efficient determination of the spot welding robot path is crucial for the processing of weld spots.

[0005] In traditional technology, the path of the spot welding robot is generally determined through manual planning; however, this method is likely to consume a lot of time and manpower, resulting in low efficiency in determining the path of the spot welding robot. Summary of the Invention

[0006] Based on this, it is necessary to provide a spot welding robot path determination method, device, computer equipment, computer-readable storage medium and computer program product that can improve the efficiency of determining the spot welding robot path in order to address the above technical problems.

[0007] In a first aspect, the present application provides a method for determining a spot welding robot path, comprising:

[0008] Obtaining the welding spot information of the welding spot to be processed and the robot number of the spot welding robot;

[0009] Clustering the weld points to be processed according to the weld point information to obtain a weld point set corresponding to the number of robots;

[0010] Determining, from the welding point sets, welding point sets corresponding to the respective spot welding robots, and determining position information of the respective spot welding robots;

[0011] The position information of the weld points to be processed in the weld point set corresponding to each of the spot welding robots, as well as the position information of each of the spot welding robots, are input into a pre-trained weld point processing path prediction model to obtain the weld point processing path of each of the spot welding robots; the weld point processing path of each of the spot welding robots is used to represent the path along which each of the spot welding robots performs spot welding processing on the weld points to be processed in the corresponding weld point set.

[0012] In one of the embodiments, the clustering the to-be-processed welding points according to the welding point information to obtain the welding point sets corresponding to the number of robots comprises:

[0013] The clustering the to-be-processed welding points according to the welding point information to obtain the candidate welding point sets corresponding to the number of robots and the cluster centers of the candidate welding point sets;

[0014] The loss values corresponding to the candidate welding point sets are determined according to the distances between the to-be-processed welding points in the candidate welding point sets and the cluster centers, respectively.

[0015] In the case that the loss values corresponding to the candidate welding point sets all satisfy the preset loss value condition, the candidate welding point sets are taken as the welding point sets corresponding to the number of robots.

[0016] In one of the embodiments, the determining the welding point sets corresponding to the spot welding robots from the welding point sets comprises:

[0017] The cluster centers of the welding point sets are obtained.

[0018] The welding point sets corresponding to the spot welding robots are determined from the welding point sets according to the cluster centers of the welding point sets and the position information of the spot welding robots.

[0019] In one of the embodiments, the inputting the position information of the to-be-processed welding points in the welding point sets corresponding to the spot welding robots and the position information of the spot welding robots into a pre-trained welding point processing path prediction model to obtain the welding point processing paths of the spot welding robots comprises:

[0020] The position information of the to-be-processed welding points in the welding point sets corresponding to the spot welding robots and the position information of the spot welding robots are input into a pre-trained welding point processing path prediction model to obtain the predicted welding point processing paths of the spot welding robots and the predicted probabilities corresponding to the predicted welding point processing paths of the spot welding robots.

[0021] The predicted welding point processing path corresponding to the maximum predicted probability is selected from the predicted welding point processing paths of the spot welding robots as the welding point processing path of the spot welding robot.

[0022] In one of the embodiments, the pre-trained welding point processing path prediction model is trained by the following way:

[0023] The sample welding point information of sample welding points and the sample number of sample spot welding robots are obtained.

[0024] According to the sample welding point information, the sample welding points are clustered to obtain a sample welding point set corresponding to the sample robot number;

[0025] From the sample welding point set, a sample welding point set corresponding to each sample spot welding robot is determined, and position information of each sample spot welding robot is determined;

[0026] The position information of the sample welding points in the sample welding point set corresponding to each sample spot welding robot and the position information of each sample spot welding robot are input into the welding point processing path prediction model to be trained to obtain a predicted welding point processing path of each sample spot welding robot.

[0027] According to the difference between the predicted welding point processing path of each sample spot welding robot and the actual welding point processing path, the welding point processing path prediction model to be trained is iteratively trained to obtain the pre-trained welding point processing path prediction model.

[0028] In one embodiment, the welding point processing path prediction model to be trained is iteratively trained according to the difference between the predicted welding point processing path of each sample spot welding robot and the actual welding point processing path to obtain the pre-trained welding point processing path prediction model, including:

[0029] The first path length information corresponding to the predicted welding point processing path of each sample spot welding robot is determined, and the second path length information corresponding to the actual welding point processing path of each sample spot welding robot is determined.

[0030] According to the difference between the first path length information and the second path length information, the welding point processing path prediction model to be trained is iteratively trained to obtain the pre-trained welding point processing path prediction model.

[0031] In one embodiment, after inputting the position information of the welding points to be processed in the welding point set corresponding to each spot welding robot and the position information of each spot welding robot into the pre-trained welding point processing path prediction model to obtain the welding point processing path of each spot welding robot, the method further includes:

[0032] According to the welding point processing path of each spot welding robot, a welding point processing instruction of each spot welding robot is generated.

[0033] According to the welding point processing instruction of each spot welding robot, each spot welding robot is controlled to perform spot welding processing on the welding points to be processed in the corresponding welding point set.

[0034] In a second aspect, the application also provides a spot welding robot path determination device, including:

[0035] a data acquisition module, configured to acquire spot information of to-be-processed spots and a number of spot welding robots;

[0036] a spot clustering module, configured to perform clustering processing on the to-be-processed spots according to the spot information, to obtain spot sets corresponding to the number of robots;

[0037] an information determination module, configured to determine, from the spot sets, spot sets corresponding to each of the spot welding robots, and determine position information of each of the spot welding robots;

[0038] a path prediction module, configured to input position information of to-be-processed spots in the spot set corresponding to each of the spot welding robots and the position information of each of the spot welding robots into a pre-trained spot processing path prediction model, to obtain a spot processing path of each of the spot welding robots; the spot processing path of each of the spot welding robots is used to represent a path of each of the spot welding robots for processing the to-be-processed spots in the corresponding spot set.

[0039] In a third aspect, the present application further provides a computer device, comprising a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

[0040] acquire spot information of to-be-processed spots and a number of spot welding robots;

[0041] perform clustering processing on the to-be-processed spots according to the spot information, to obtain spot sets corresponding to the number of robots;

[0042] determine, from the spot sets, spot sets corresponding to each of the spot welding robots, and determine position information of each of the spot welding robots;

[0043] input position information of to-be-processed spots in the spot set corresponding to each of the spot welding robots and the position information of each of the spot welding robots into a pre-trained spot processing path prediction model, to obtain a spot processing path of each of the spot welding robots; the spot processing path of each of the spot welding robots is used to represent a path of each of the spot welding robots for processing the to-be-processed spots in the corresponding spot set.

[0044] In a fourth aspect, the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the following steps:

[0045] acquire spot information of to-be-processed spots and a number of spot welding robots;

[0046] According to the welding point information, the to-be-processed welding points are clustered to obtain welding point sets corresponding to the number of robots;

[0047] From the welding point sets, welding point sets corresponding to each spot welding robot are determined, and position information of each spot welding robot is determined;

[0048] The position information of the to-be-processed welding points in the welding point sets corresponding to each spot welding robot and the position information of each spot welding robot are input into a pre-trained welding point processing path prediction model to obtain welding point processing paths of each spot welding robot. The welding point processing path of each spot welding robot is used to represent the path of each spot welding robot for processing the to-be-processed welding points in the corresponding welding point set.

[0049] In a fifth aspect, the present application also provides a computer program product comprising a computer program which, when executed by a processor, implements the following steps:

[0050] Obtaining welding point information of to-be-processed welding points and the number of spot welding robots;

[0051] According to the welding point information, the to-be-processed welding points are clustered to obtain welding point sets corresponding to the number of robots;

[0052] From the welding point sets, welding point sets corresponding to each spot welding robot are determined, and position information of each spot welding robot is determined;

[0053] The position information of the to-be-processed welding points in the welding point sets corresponding to each spot welding robot and the position information of each spot welding robot are input into a pre-trained welding point processing path prediction model to obtain welding point processing paths of each spot welding robot. The welding point processing path of each spot welding robot is used to represent the path of each spot welding robot for processing the to-be-processed welding points in the corresponding welding point set.

[0054] The above-mentioned spot welding robot path determination method, device, computer equipment, storage medium and computer program product first obtain the welding point information of the welding points to be processed and the number of spot welding robots, and then cluster the welding points to be processed according to the welding point information to obtain a welding point set corresponding to the number of robots. Then, from the welding point set, the welding point set corresponding to each spot welding robot is determined, and the position information of each spot welding robot is determined. Then, the position information of the welding points to be processed in the welding point set corresponding to each spot welding robot and the position information of each spot welding robot are input into a pre-trained welding point processing path prediction model to obtain the path for each spot welding robot to perform spot welding processing on the welding points to be processed in the corresponding welding point set as the welding point processing path of each spot welding robot. In this way, when determining the path of the spot welding robot, through clustering processing and targeted allocation of welding point sets to each spot welding robot, and predicting the processing path, the reasonable allocation and optimization arrangement of the spot welding tasks can be achieved, thereby reducing the idle and waiting time of the robots, and giving full play to the capabilities of each spot welding robot, avoiding waste or excessive use of resources, and achieving optimal allocation of resources, which is conducive to improving the efficiency of the robot spot welding processing; moreover, the entire process does not require human intervention, avoiding the defect of low efficiency in determining the spot welding robot path due to manual planning that easily consumes a lot of time and manpower, thereby improving the efficiency of determining the spot welding robot path. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments of the present application or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying any creative work.

[0056] Figure 1 1 is a flow chart of a method for determining a path for a spot welding robot according to an embodiment;

[0057] Figure 2 A schematic flow chart of the training steps of a pre-trained solder joint processing path prediction model in one embodiment;

[0058] Figure 3 Schematic diagram of a flow chart of a method for determining a path for a spot welding robot in another embodiment;

[0059] Figure 4 Schematic diagram of a Seq2seq model and a pointer network in one embodiment;

[0060] Figure 5 Schematic diagram of an Actor-Critic based reinforcement pointer network in one embodiment;

[0061] Figure 6 Flowchart of the path planning method of the digital twin robot for a multi-welding point station in an embodiment;

[0062] Figure 7 Schematic diagram of the welding point distribution of an automobile rear door assembly in an embodiment;

[0063] Figure 8 Schematic diagram of a multi-welding point robot station in an embodiment;

[0064] Figure 9 Schematic diagram of the reinforcement pointer network training process in an embodiment;

[0065] Figure 10 Schematic diagram of the training network path planning verification in an embodiment;

[0066] Figure 11 Schematic diagram of the reinforcement pointer network training with different learning rates in an embodiment;

[0067] Figure 12 Schematic diagram of the initial welding point allocation based on the K-means algorithm in an embodiment;

[0068] Figure 13 Schematic diagram of the welding point path planning result evolved by the genetic algorithm in an embodiment;

[0069] Figure 14 Schematic diagram of the genetic algorithm with double-chromosome coding in an embodiment;

[0070] Figure 15 Schematic diagram of the evolution result of the minimum value of the individual fitness in the population in an embodiment;

[0071] Figure 16 Schematic diagram of the evolution result of the average value of the individual fitness in the population in an embodiment;

[0072] Figure 17 Schematic diagram of the spot welding task path re-planning result after the robot fails in an embodiment;

[0073] Figure 18 Structural block diagram of the spot welding robot path determination device in an embodiment;

[0074] Figure 19 Internal structure diagram of a computer device in an embodiment. DETAILED DESCRIPTION

[0075] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not intended to limit the present application.

[0076] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0077] In one exemplary embodiment, as shown in Figure 1 A spot welding robot path determination method is provided, and the present embodiment is exemplified by the method applied to a server; it can be understood that the method can also be applied to a terminal, and can also be applied to a system including a terminal and a server, and is realized through the interaction between the terminal and the server. The terminal can be, but is not limited to, various personal computers, notebook computers, smart phones and tablet computers; the server can be realized by an independent server or a server cluster composed of multiple servers. In the present embodiment, the method includes the following steps:

[0078] In step S101, the welding point information of the to-be-processed welding point and the number of robots of the spot welding robot are obtained.

[0079] The to-be-processed welding point refers to a welding point that needs to be processed.

[0080] The welding point information refers to information associated with the to-be-processed welding point, such as the quantity information and position information of the to-be-processed welding point.

[0081] The spot welding robot refers to an industrial robot that processes the to-be-processed welding point.

[0082] The number of robots refers to the quantity information of the spot welding robot, such as 5.

[0083] Exemplarily, the server obtains the welding point identifier of the to-be-processed welding point in response to the processing request information for the to-be-processed welding point; then, the server obtains the welding point information corresponding to the welding point identifier of the to-be-processed welding point from the database as the welding point information of the to-be-processed welding point according to the welding point identifier of the to-be-processed welding point; then, the server screens out the spot welding robots whose corresponding state information meets the preset state information (such as the spot welding robots in the idle state) from the candidate spot welding robots, and determines the quantity information of these spot welding robots as the number of robots of the spot welding robots.

[0084] In step S102, the welding points to be processed are clustered according to the welding point information to obtain welding point sets corresponding to the number of robots.

[0085] The clustering processing refers to a processing procedure of converting the welding points to be processed into the welding point sets.

[0086] The welding point set refers to a set containing multiple welding points to be processed.

[0087] For example, the server clusters the welding points to be processed according to the welding point information to obtain candidate welding point sets corresponding to the number of robots and the clustering centers of the candidate welding point sets; then, the server judges the clustering centers of the candidate welding point sets, and in the case that the clustering centers of the candidate welding point sets all satisfy the preset condition, the candidate welding point sets corresponding to the number of robots are taken as the welding point sets corresponding to the number of robots.

[0088] In step S103, the welding point sets corresponding to the spot welding robots are determined from the welding point sets, and the position information of the spot welding robots is determined.

[0089] The position information of the spot welding robots refers to the three-dimensional coordinate information of the spot welding robots.

[0090] For example, the server obtains the clustering centers of the welding point sets, and determines the welding point sets corresponding to the spot welding robots and the position information of the spot welding robots according to the clustering centers of the welding point sets.

[0091] In step S104, the position information of the welding points to be processed in the welding point sets corresponding to the spot welding robots and the position information of the spot welding robots are input into a pre-trained welding point processing path prediction model to obtain the welding point processing paths of the spot welding robots; the welding point processing paths of the spot welding robots are respectively used to represent the paths of the spot welding robots for spot welding processing on the welding points to be processed in the corresponding welding point sets.

[0092] The position information of the welding points to be processed refers to the three-dimensional coordinate information of the welding points to be processed.

[0093] The welding point processing path prediction model refers to a network model capable of obtaining the welding point processing paths of the spot welding robots by using the position information of the welding points to be processed in the welding point sets corresponding to the spot welding robots and the position information of the spot welding robots.

[0094] The welding point processing paths of the spot welding robots are respectively used to represent the paths of the spot welding robots for spot welding processing on the welding points to be processed in the corresponding welding point sets.

[0095] Exemplarily, the server obtains position information of the to-be-processed welding points in the welding point set corresponding to each spot welding robot; then, the server inputs the position information of the to-be-processed welding points in the welding point set corresponding to each spot welding robot and the position information of each spot welding robot into the pre-trained welding point processing path prediction model, and obtains, through the pre-trained welding point processing path prediction model, a path for each spot welding robot to process the to-be-processed welding points in the welding point set corresponding to the spot welding robot, as the welding point processing path of each spot welding robot.

[0096] In the above spot welding robot path determination method, the welding point information of the to-be-processed welding points and the number of spot welding robots are obtained first, then the to-be-processed welding points are clustered according to the welding point information to obtain welding point sets corresponding to the number of robots, then the welding point set corresponding to each spot welding robot is determined from the welding point sets, and the position information of each spot welding robot is determined, then the position information of the to-be-processed welding points in the welding point set corresponding to each spot welding robot and the position information of each spot welding robot are input into the pre-trained welding point processing path prediction model to obtain the path for each spot welding robot to process the to-be-processed welding points in the welding point set corresponding to the spot welding robot, as the welding point processing path of each spot welding robot. In this way, when determining the spot welding robot path, the welding point sets are allocated to each spot welding robot through clustering and targeted processing, and the processing path is predicted, so that the reasonable allocation and optimization arrangement of the spot welding task can be realized, thereby reducing the idle and waiting time of the robot, fully exerting the capacity of each spot welding robot, avoiding waste or overuse of resources, optimizing the allocation of resources, and improving the efficiency of robot spot welding processing. Moreover, the whole process does not need manual intervention, avoiding the defects of consuming a large amount of time and manpower through manual planning, leading to low efficiency of determining the spot welding robot path, and thus improving the efficiency of determining the spot welding robot path.

[0097] In one exemplary embodiment, the step S102 of clustering the to-be-processed welding points according to the welding point information to obtain welding point sets corresponding to the number of robots specifically includes the following contents: clustering the to-be-processed welding points according to the welding point information to obtain candidate welding point sets corresponding to the number of robots and clustering centers of each candidate welding point set; determining loss values corresponding to each candidate welding point set according to the distance between the to-be-processed welding points in each candidate welding point set and the clustering center; and in the case that the loss values corresponding to each candidate welding point set all satisfy a preset loss value condition, taking each candidate welding point set as a welding point set corresponding to the number of robots.

[0098] The candidate welding point set refers to a welding point set obtained by initially clustering the to-be-processed welding points.

[0099] The cluster center refers to the center point corresponding to the solder joint to be processed in each candidate solder joint set.

[0100] Here, the distance may refer to the Euclidean distance.

[0101] The preset loss value condition refers to a preset loss value threshold. It should be noted that the preset loss value condition may be determined according to the circumstances.

[0102] Exemplarily, the server clusters the welds to be processed based on the weld information to obtain a set of candidate welds corresponding to the number of robots, and a cluster center for each candidate weld set. For example, the server uses a K-means algorithm to perform initial allocation based on the spatial distribution characteristics of the welds, and continuously changes the random number seed to form an initial population of genetic algorithm chromosomes, which serve as a set of candidate welds corresponding to the number of robots, and a cluster center for each candidate weld set. Next, the server determines a loss value corresponding to each candidate weld set based on the distance between the welds to be processed in each candidate weld set and the cluster center. For example, the server determines the sum of the Euclidean distances between the welds to be processed in each candidate weld set and the cluster center as the loss value corresponding to each candidate weld set. Then, if the loss values ​​corresponding to each candidate weld set all meet a preset loss value condition, the server regards each candidate weld set as a weld set corresponding to the number of robots. For example, if the loss values ​​corresponding to each candidate weld set are all less than a preset loss value threshold, the server regards each candidate weld set as a weld set corresponding to the number of robots.

[0103] For example, the server can calculate the loss value using the following formula:

[0104] , formula (1)

[0105] in, Refers to the loss value, N is the number of solder joints, is the solder joint to be processed in each candidate solder joint set, is the cluster center.

[0106] In this embodiment, the welds to be processed are assigned to different spot welding robots through clustering processing, which can more reasonably allocate work tasks according to the number of robots, which is conducive to improving production efficiency; moreover, the loss value is determined according to the distance between the weld point and the cluster center, thereby ensuring that the weld points in each weld point set are relatively uniform and concentrated, thereby improving the welding efficiency of the spot welding robot.

[0107] In an example embodiment, the step S103 determines the welding point set corresponding to each spot welding robot from the welding point set, and specifically includes the following contents: obtaining the clustering center of each welding point set; determining the welding point set corresponding to each spot welding robot from the welding point set according to the clustering center of each welding point set and the position information of each spot welding robot.

[0108] For example, the server obtains the clustering center of each welding point set; then, the server determines the distance between each welding point set and the spot welding robot according to the clustering center of each welding point set and the position information of each spot welding robot; then, the server determines the welding point set corresponding to each spot welding robot from the welding point set according to the distance between each welding point set and the spot welding robot; for example, the server obtains the combination between the plurality of spot welding robots and the welding point set, and sums the distance between each welding point set and the spot welding robot for each combination to obtain the total distance, and takes the combination corresponding to the minimum total distance as the welding point set corresponding to each spot welding robot.

[0109] In this embodiment, the welding point set is reasonably distributed according to the clustering center of the welding point set and the position information of the spot welding robot, which is beneficial to reduce the time and distance of robot movement, thereby speeding up the overall progress of the welding work; moreover, by accurately distributing the welding point set to the corresponding spot welding robot, the work load of each robot is more balanced, and resource idling or overuse is avoided.

[0110] In an example embodiment, the step S104 inputs the position information of the to-be-processed welding points in the welding point set corresponding to each spot welding robot and the position information of each spot welding robot into the pre-trained welding point processing path prediction model to obtain the welding point processing path of each spot welding robot, and specifically includes the following contents: inputting the position information of the to-be-processed welding points in the welding point set corresponding to each spot welding robot and the position information of each spot welding robot into the pre-trained welding point processing path prediction model to obtain the predicted welding point processing path of each spot welding robot and the predicted probability corresponding to the predicted welding point processing path of each spot welding robot; selecting the predicted welding point processing path with the maximum predicted probability from the predicted welding point processing path of each spot welding robot as the welding point processing path of each spot welding robot.

[0111] The predicted welding point processing path refers to the predicted value of the welding point processing path.

[0112] The predicted probability is used to represent the possibility corresponding to the predicted welding point processing path.

[0113] Exemplarily, the server inputs the position information of the welds to be processed in the weld point set corresponding to each spot welding robot and the position information of each spot welding robot into a pre-trained weld point processing path prediction model. Through the pre-trained weld point processing path prediction model, the position information of the welds to be processed in the weld point set corresponding to each spot welding robot and the position information of each spot welding robot are all used as the original input sequence; then, the server performs feature extraction processing on the original input sequence to obtain a feature vector corresponding to the original input sequence; then, the server obtains an encoder output feature vector corresponding to the feature vector and a decoder output feature vector corresponding to the feature vector based on the feature vector corresponding to the original input sequence; then, the server obtains a target feature vector corresponding to the feature vector based on the encoder output feature vector and the decoder feature vector corresponding to the feature vector; then, the server obtains the predicted weld point processing path of each spot welding robot and the prediction probability corresponding to the predicted weld point processing path of each spot welding robot based on the target feature vector; finally, the server selects the predicted weld point processing path with the largest corresponding prediction probability from the predicted weld point processing paths of each spot welding robot as the weld point processing path of each spot welding robot.

[0114] For example, the server can calculate the predicted probability corresponding to the predicted welding spot processing path of each spot welding robot using the following formula:

[0115] , formula (2)

[0116] , formula (3)

[0117] in, The encoder The hidden layer output of the step, The decoder The hidden layer output of the step, 、 、 are the parameters of the pointer network to be trained, is the probability vector output for each solder joint location.

[0118] In this embodiment, by screening out the path with the highest prediction probability as the final processing path, a more efficient and accurate processing path can be selected; moreover, a predicted solder point processing path is quickly generated through a pre-trained solder point processing path prediction model, thereby saving a large amount of computing time required by traditional path planning methods, thereby improving overall production efficiency.

[0119] In an exemplary embodiment, Figure 2 As shown, the spot welding robot path determination method provided by the present application also includes a training step of a pre-trained weld spot processing path prediction model, specifically including the following steps:

[0120] In step S201, sample spot welding information of sample spot welds and a sample number of sample spot welding robots are obtained.

[0121] In step S202, the sample spot welds are clustered according to the sample spot welding information, to obtain sample spot welding sets corresponding to the sample number of sample spot welding robots.

[0122] In step S203, sample spot welding sets corresponding to each sample spot welding robot are determined from the sample spot welding sets, and position information of each sample spot welding robot is determined.

[0123] In step S204, position information of sample spot welds in the sample spot welding sets corresponding to each sample spot welding robot and position information of each sample spot welding robot are input into the welding spot processing path prediction model to be trained, to obtain predicted welding spot processing paths of each sample spot welding robot.

[0124] In step S205, the welding spot processing path prediction model to be trained is iteratively trained according to differences between the predicted welding spot processing paths of each sample spot welding robot and actual welding spot processing paths, to obtain a pre-trained welding spot processing path prediction model.

[0125] The sample spot welding information is information associated with the sample spot welds.

[0126] The sample spot welding information is information associated with the sample spot welds.

[0127] The sample spot welding information is information associated with the sample spot welds.

[0128] The sample number of sample spot welding robots refers to quantity information of the sample spot welding robots.

[0129] The sample spot welding set refers to a set containing multiple sample spot welds.

[0130] The actual welding spot processing path refers to an actual value of the welding spot processing path.

[0131] Exemplarily, in response to a model training instruction for a weld processing path prediction model to be trained, the server obtains sample weld information of the sample welds and the number of sample spot welding robots from a database; then, the server clusters the sample welds based on the sample weld information to obtain a set of candidate welds corresponding to the number of sample robots and a cluster center of each candidate weld set; then, the server determines the cluster center of each candidate weld set, and when the cluster center of each candidate weld set meets a preset condition, the set of candidate welds corresponding to the number of sample robots is used as the sample weld set corresponding to the number of sample robots; then, the server obtains the cluster center of each sample weld set, and determines the sample weld set corresponding to each sample spot welding robot based on the cluster center of each sample weld set, and determines the position information of each sample spot welding robot; Then, the server inputs the position information of the sample welds in the sample weld set corresponding to each sample weld robot, as well as the position information of each sample weld robot, into the weld processing path prediction model to be trained to obtain the predicted weld processing path of each sample weld robot; then, the server obtains the actual weld processing path of each sample weld robot, and obtains the loss value based on the difference between the predicted weld processing path and the actual weld processing path of each sample weld robot; then, the server adjusts the model parameters of the weld processing path prediction model to be trained based on the loss value; then, the server retrains the weld processing path prediction model after the model parameters are adjusted until the loss value obtained by the trained weld processing path prediction model is less than the loss value threshold, then stops training, and uses the trained weld processing path prediction model as the pre-trained weld processing path prediction model.

[0132] In this embodiment, by pre-training the weld processing path prediction model, it is convenient to predict the weld processing path of each spot welding robot after obtaining the position information of the welds to be processed in the weld point set corresponding to each spot welding robot and the position information of each spot welding robot in actual application; moreover, the weld processing path prediction model receives new data in each round of iteration, and performs internal improvements and optimizations on the model, so that predictions can be made more effectively, which is conducive to improving the prediction accuracy of the weld processing path prediction model.

[0133] In an exemplary embodiment, the above-mentioned step S205, based on the difference between the predicted weld spot processing path and the actual weld spot processing path of each sample spot welding robot, iteratively trains the weld spot processing path prediction model to be trained to obtain a pre-trained weld spot processing path prediction model, specifically including the following contents: determining the first path length information corresponding to the predicted weld spot processing path of each sample spot welding robot, and determining the second path length information corresponding to the actual weld spot processing path of each sample spot welding robot; based on the difference between the first path length information and the second path length information, iteratively trains the weld spot processing path prediction model to be trained to obtain a pre-trained weld spot processing path prediction model.

[0134] The first path length information refers to the path length information corresponding to the predicted welding spot processing path of each sample spot welding robot.

[0135] The second path length information refers to the path length information corresponding to the actual welding point processing path of each sample spot welding robot.

[0136] Exemplarily, the server determines the path length information corresponding to the predicted weld spot processing path of each sample spot welding robot as the first path length information; then, the server determines the path length information corresponding to the actual weld spot processing path of each sample spot welding robot as the second path length information; then, the server obtains a loss value based on the difference between the first path length information and the second path length information; then, the server adjusts the model parameters of the weld spot processing path prediction model to be trained based on the loss value; then, the server re-trains the weld spot processing path prediction model after the model parameters are adjusted until the loss value obtained by the trained weld spot processing path prediction model is less than the loss value threshold, then stops training, and uses the trained weld spot processing path prediction model as the pre-trained weld spot processing path prediction model.

[0137] In this embodiment, iterative training is performed by comparing the difference between the predicted path length and the actual path length, which can continuously optimize the model and make its prediction results closer to the actual situation, thereby improving the prediction accuracy of the solder joint processing path prediction model and providing more effective guidance for the actual production process.

[0138] In an exemplary embodiment, the above-mentioned step S104, after inputting the position information of the welds to be processed in the weld point set corresponding to each spot welding robot and the position information of each spot welding robot into a pre-trained weld point processing path prediction model to obtain the weld point processing path of each spot welding robot, specifically includes the following contents: generating weld point processing instructions for each spot welding robot according to the weld point processing path of each spot welding robot; and controlling each spot welding robot to perform spot welding on the welds to be processed in the corresponding weld point set according to the weld point processing instructions of each spot welding robot.

[0139] The spot welding point processing instruction is an instruction information for controlling the spot welding robot to perform spot welding processing on the to-be-processed spot welding points in the corresponding spot welding point set.

[0140] For example, the server generates the spot welding point processing instruction corresponding to the spot welding point processing path of each spot welding robot as the spot welding point processing instruction of each spot welding robot according to the spot welding point processing path of each spot welding robot, and then controls each spot welding robot to perform spot welding processing on the to-be-processed spot welding points in the corresponding spot welding point set according to the spot welding point processing instruction of each spot welding robot.

[0141] In this embodiment, the spot welding robot can be automatically controlled to operate according to the generated instruction, manual intervention is reduced, and the automation degree and efficiency of production are improved. Moreover, errors and safety hazards that may occur in manual operation are avoided, and the safety of the production process is ensured.

[0142] In one example embodiment, as shown in Figure 3 Another method for determining the path of a spot welding robot is provided, which is applied to a server for example and includes the following steps:

[0143] In step S301, the spot welding point information of the to-be-processed spot welding points and the number of spot welding robots are obtained.

[0144] In step S302, the to-be-processed spot welding points are clustered according to the spot welding point information to obtain candidate spot welding point sets corresponding to the number of spot welding robots and the cluster centers of the candidate spot welding point sets.

[0145] In step S303, the loss values corresponding to the candidate spot welding point sets are determined according to the distances between the to-be-processed spot welding points in each candidate spot welding point set and the cluster center.

[0146] In step S304, in the case where the loss values corresponding to each candidate spot welding point set all satisfy a preset loss value condition, each candidate spot welding point set is taken as a spot welding point set corresponding to the number of spot welding robots.

[0147] In step S305, the cluster centers of the spot welding point sets are obtained, and the spot welding point set corresponding to each spot welding robot is determined from the spot welding point sets according to the cluster centers of the spot welding point sets and the position information of each spot welding robot.

[0148] In step S306, the position information of the to-be-processed spot welding points in the spot welding point set corresponding to each spot welding robot and the position information of each spot welding robot are input into a pre-trained spot welding point processing path prediction model to obtain the predicted spot welding point processing path of each spot welding robot and the predicted probability corresponding to the predicted spot welding point processing path of each spot welding robot.

[0149] Step S307: Filter out the predicted weld spot processing paths of each spot welding robot with the highest corresponding predicted weld spot processing path as the weld spot processing path of each spot welding robot; the weld spot processing paths of each spot welding robot are respectively used to represent the paths by which each spot welding robot performs spot welding processing on the weld spots to be processed in the corresponding weld spot set.

[0150] In the above-mentioned spot welding robot path determination method, when determining the spot welding robot path, clustering processing and targeted allocation of welding point sets to each spot welding robot, and prediction of the processing path can achieve reasonable allocation and optimal arrangement of spot welding tasks, thereby reducing the idle and waiting time of the robot, and can give full play to the capabilities of each spot welding robot, avoid waste or excessive use of resources, and achieve optimal allocation of resources, which is conducive to improving the efficiency of robot spot welding processing; moreover, the entire process does not require human intervention, avoiding the defect of low efficiency in determining the spot welding robot path due to manual planning that easily consumes a lot of time and manpower, thereby improving the efficiency of determining the spot welding robot path.

[0151] In an exemplary embodiment, in order to more clearly illustrate the spot welding robot path determination method provided in the embodiment of the present application, the spot welding robot path determination method is specifically described below with a specific embodiment. In one embodiment, the present application also provides a multi-welding station digital twin robot path planning method based on an enhanced pointer network. When determining the spot welding robot path, the weld point information of the weld points to be processed and the number of spot welding robots are first obtained. Then, according to the weld point information, the weld points to be processed are clustered to obtain a weld point set corresponding to the number of robots. Then, from the weld point set, the weld point set corresponding to each spot welding robot is determined, and the position information of each spot welding robot is determined. Then, the position information of the weld points to be processed in the weld point set corresponding to each spot welding robot and the position information of each spot welding robot are input into a pre-trained weld point processing path prediction model to obtain the path for each spot welding robot to perform spot welding processing on the weld points to be processed in the corresponding weld point set, as the weld point processing path of each spot welding robot. Specifically including the following contents:

[0152] On the welding production line, robots gather There are A robot, that is , body welding point collection There are solder joints, i.e. Among them, the solder joint The three-dimensional spatial position of ,robot The initial position is Assume that the moving speed of the robot end is the movement time between the welding points and is denoted by , i.e.

[0153] , formula (4)

[0154] For the spot welding task assignment, decision variables are set if the welding point is assigned to the robot . Since one welding point can only be assigned to one robot, the following constraint is satisfied:

[0155] , formula (5)

[0156] Afterwards, the robot is assigned a set of spot welding tasks , where denotes the size of the set , satisfying the following constraint:

[0157] , formula (6)

[0158] , formula (7)

[0159] , formula (8)

[0160] The set of spot welding tasks for the robot is path planned, introducing decision variables . If in the optimized welding sequence for the robot , the robot moves from welding point to welding point , then , satisfying the following constraint:

[0161] , formula (9)

[0162] A possible optimized welding path is:

[0163] , formula (10)

[0164] Therefore, for the robot , the time to complete the optimized welding sequence , where is the welding time of the welding point .

[0165] , formula (11)

[0166] On this basis, in order to minimize the maximum completion time of the robot spot welding task, an optimization objective function is constructed to reduce the tact time of the welding production line as much as possible:

[0167] , formula (12)

[0168] At the same time, a series of space constraints need to be met, including reachability, obstacle avoidance constraint, collision avoidance constraint, etc.

[0169] Reachability: Indicates the welding robot The range of the reachable spherical cylindrical work area, Indicates the maximum radius of the spherical cylindrical space, Indicates the minimum radius of the spherical cylindrical space, and the reachability constraint indicates that the distance between the assigned welding spot and the welding robot is within the range of the robot work space.

[0170] , formula (13)

[0171] Obstacle avoidance constraint: at all times, the path of the welding robot Cannot collide with the welding workpiece .

[0172] , formula (14)

[0173] Collision avoidance constraint: at all times, the paths of any two robots And Cannot interfere with each other.

[0174] , formula (15)

[0175] The application provides a single robot path planning method based on a reinforcement pointer network, which trains and obtains an optimal single robot path planning strategy through the reinforcement pointer network.

[0176] The path planning of multiple robots is an NP (non-deterministic polynomial, non-deterministic polynomial) -Hard problem, which can be regarded as a combination of the knapsack problem and the TSP (Traveling Salesman Problem, Traveling Salesman Problem) problem. Among them, the path planning sub-problem of each robot is a TSP problem, which needs to complete a certain number of spot welding tasks. However, due to the uncertainty of task allocation, in order to cope with collision avoidance, reachability constraints and other dynamic adjustments, the path planning of each robot for a variable number of welding spots is very challenging. Therefore, in order to solve the welding spot path planning decision variable of the first robot The pointer network is introduced and combined with the policy gradient based reinforcement learning method to train and obtain the optimal path planning strategy of the single robot.

[0177] The structure of the pointer network is a Seq2Seq (sequence to sequence) model, which includes an encoder and a decoder, both of which are composed of neurons of recurrent neural networks and long short-term memory networks. As shown in Figure 4 , the left green part is the encoder, and the right purple part is the decoder. However, the prediction output target size of the traditional Seq2Seq model is fixed, which is not suitable for the combinatorial optimization problem with variable input quantity. The pointer network introduces a new attention mechanism. The encoder converts the original input sequence into an intermediate vector representation, and the decoder selects the element with the maximum weight in the input sequence at each step using the attention mechanism. Therefore, the pointer network can receive input sequences of any length and output pointers to input elements.

[0178] , formula (2)

[0179] , formula (3)

[0180] wherein, is the hidden layer output of the encoder at the i-th step, is the hidden layer output of the decoder at the i-th step, , , is the parameter to be trained of the pointer network, is the probability vector output for each weld point position. The pointer network for robot weld point path planning has the following two advantages: (1) its network responds to new data faster than traditional local search algorithms. Compared with traditional heuristic algorithms, it does not need to recalculate without prior knowledge every time a case is input. (2) Compared with other Seq2Seq networks, the output of the pointer network is independent of the dictionary size and is related to the length of the input, which has high universality and can adapt to more complex spot welding task allocation and path planning problems. In order to solve the problem of label required in the training process of the pointer network, reinforcement learning based on policy gradient and actor-critic structure is used for autonomous training. As shown in , it is a schematic diagram of the reinforced pointer network.

[0181] Figure 5 The actor-critic model uses a defined loss function to update the parameters of the pointer network, is the parameterized policy of the pointer network, and its weight is

[0182] .

[0183] ​​, formula (16)

[0184] , formula (17)

[0185] wherein, is a set of coordinates required for a robot to complete a spot welding task, is the path planning result. is the length of the route for a single robot to complete a series of spot welding tasks and return to the origin position. is a baseline included to improve the performance of the algorithm, by dividing the set of spot welding tasks into a sequence of input to predict the expected maximum completion time. The Critic network predicts the value based on the final state of the Actor network given the known input , and is optimized based on the policy gradient. The Critic network trains the predicted value with the actual value mean square error as the optimization goal, using stochastic gradient descent.

[0186] , formula (18)

[0187] wherein, is a baseline function parameterized by the Critic network, and its weight is , is the training batch size.

[0188] The trained reinforcement pointer network only considers the spatial characteristics of the welding points and does not consider complex spatial constraints such as obstacle avoidance and collision avoidance. The digital twin of the spot welding robot includes a geometric model and a motion model, which can be used to detect conflicts and accessibility during the welding process of the spot welding robot through simulation. The related methods are relatively mature and have commercial software support.

[0189] The path planning method of the digital twin robot of the multi-spot welding station is shown in Figure 6 . In order to realize the effective task allocation of the multi-spot welding station robot, the decision variable of the spot welding task allocation is solved. First, the K-means algorithm is used to make the initial allocation according to the spatial distribution characteristics of the welding points, and the random seed is changed constantly to form the initial population of the genetic algorithm chromosome. The K-means algorithm is a data partition method based on Euclidean distance. As a typical unsupervised learning method, it has the advantages of linear computational complexity, fast convergence speed, and strong interpretability. In this algorithm, the selection of the hyperparameter is the number of robots on the welding station, that is, all the welding points are divided into welding point groups, so that each welding point to the center of the weld group to which it belongs and the minimum.

[0190] Equation (1)

[0191] The chromosome population using integer coding is evolved by crossover, mutation, and selection operators. Each chromosome contains the assignment of the welding points to the welding robots. After decoding, the assignment is input to the trained pointer network to get the optimal path planning for each robot. The digital twin of the multi-point welding robot runs the simulation according to the task assignment and path planning method, and returns the fitness value for reselecting the chromosome individuals.

[0192] To verify the effectiveness of the proposed method, the welding points of an automobile rear door assembly are taken as an example. On the automobile production line, this station contains 5 industrial robots, which are distributed on both sides of the production line to complete the spot welding task of this station, as shown in Figure 7 and Figure 8 It can be seen that the X coordinate of all welding points ranges from 0 to 250, the Y coordinate ranges from 50 to 200, and the Z coordinate ranges from 0 to 50. Then, the reinforcement pointer network for single robot path planning is constructed and trained. In order to adapt to the three-dimensional space data of the welding points of this station, 20 points are randomly selected in the three-dimensional cubic space surrounded by the point [0, 50, 0] and the point [250, 200, 50] for network training. During the 20000 training process, the reward value continuously decreases and finally converges in the range of 910-950, as shown in Figure 9 and Figure 10 The trained pointer network can effectively plan for a set of 20 welding points.

[0193] By changing the learning rate, the pointer network of the same structure is trained, as shown in Figure 11 When the learning rate is 0.05, the training process of the pointer network cannot converge, and other learning rates can converge, and the effect is almost the same. The reward value of the pointer network will decrease relatively quickly in the first 1000 training process, and then slowly decrease gradually and stabilize in an interval.

[0194] For the spot welding task of the automobile rear door, the initial assignment result based on the K-means algorithm and the evolution result of the genetic algorithm are compared, as shown in Figure 12 and Figure 13The home point positions of 5 industrial robot spot welding tasks are: [50, 50, 50], [50, 200, 50], [250, 200, 50], [250, 50, 50], [150, 50, 50]. The robot moving speed is 20 cm / s. The genetic algorithm can fine-tune the initial assignment of welding spots by the K-means algorithm, so as to balance the task load of each robot and reduce the completion time of all spot welding tasks.

[0195] Using the trained pointer network, the path planning of the spot welding task assigned to each robot can be performed, so that only one chromosome is used to encode the spot welding task assignment. If the pointer network is not used, the traditional genetic algorithm needs two chromosomes to encode an individual, including the task assignment chromosome and the path planning chromosome, as shown in Figure 14 . Figure 15 and Figure 16 The evolution process of the proposed method and the traditional double-chromosome encoding method is compared. The design of the evolution operator in the traditional method is more complex, and the search space is greatly increased, so it is not easy to quickly find the optimal solution.

[0196] At the same time, since the trained pointer network can receive different numbers of spot welding task inputs, the proposed method can also quickly adapt to some abnormal situations. For example, a robot on the workstation fails, and the spot welding task assignment and path planning need to be recalculated. Both the single-chromosome encoding genetic algorithm and the pointer network can adapt to this adjustment. For example, when robot No. 5 fails, the spot welding task assignment and path planning results are as shown in Figure 17 .

[0197] The above embodiment, when determining the spot welding robot path, the welding spot set is allocated to each spot welding robot through clustering processing and targeted prediction processing path, which can realize reasonable allocation and optimization arrangement of spot welding tasks, thereby reducing the idle and waiting time of the robot, can fully play the ability of each spot welding robot, avoid waste or overuse of resources, realize optimization of resources, and is conducive to improving the efficiency of robot spot welding processing; Moreover, the whole process does not need manual intervention, which avoids the defect that the manual planning method easily consumes a lot of time and manpower, leading to low efficiency of determining the spot welding robot path, thereby improving the efficiency of determining the spot welding robot path.

[0198] It should be understood that although the steps in the flowcharts involved in the embodiments described above are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, at least some of the steps in the flowcharts involved in the embodiments described above can include multiple steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but can be alternately or alternately executed with at least part of other steps or steps or stages in other steps.

[0199] Based on the same inventive concept, the embodiments of the present application also provide a spot welding robot path determination device for implementing the above-mentioned spot welding robot path determination method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more spot welding robot path determination device embodiments provided below can refer to the limitations of the spot welding robot path determination method in the above text, which will not be repeated here.

[0200] In one exemplary embodiment, as shown in Figure 18 a spot welding robot path determination device is provided, comprising: a data acquisition module 1801, a welding point clustering module 1802, an information determination module 1803, and a path prediction module 1804, wherein:

[0201] The data acquisition module 1801 is configured to acquire welding point information of to-be-processed welding points and a number of spot welding robots.

[0202] The welding point clustering module 1802 is configured to cluster the to-be-processed welding points according to the welding point information to obtain welding point sets corresponding to the number of robots.

[0203] The information determination module 1803 is configured to determine welding point sets corresponding to each spot welding robot from the welding point sets and determine position information of each spot welding robot.

[0204] The path prediction module 1804 is configured to input position information of to-be-processed welding points in the welding point set corresponding to each spot welding robot and the position information of each spot welding robot into a pre-trained welding point processing path prediction model to obtain a welding point processing path of each spot welding robot. The welding point processing path of each spot welding robot is used to represent a path for each spot welding robot to process to-be-processed welding points in the corresponding welding point set.

[0205] In an example embodiment, the welding spot clustering module 1802 is further configured to: cluster the to-be-processed welding spots according to the welding spot information, to obtain candidate welding spot sets corresponding to the number of robots and clustering centers of the candidate welding spot sets; determine loss values corresponding to the candidate welding spot sets according to distances between the to-be-processed welding spots in the candidate welding spot sets and the clustering centers; and in a case where the loss values corresponding to the candidate welding spot sets all satisfy a preset loss value condition, take the candidate welding spot sets as the welding spot sets corresponding to the number of robots.

[0206] In an example embodiment, the information determining module 1803 is further configured to: obtain the clustering centers of the welding spot sets; and determine, from the welding spot sets, welding spot sets corresponding to the spot welding robots according to the clustering centers of the welding spot sets and the position information of the spot welding robots.

[0207] In an example embodiment, the path prediction module 1804 is further configured to: input the position information of the to-be-processed welding spots in the welding spot sets corresponding to the spot welding robots and the position information of the spot welding robots into a pre-trained welding spot processing path prediction model, to obtain predicted welding spot processing paths of the spot welding robots and predicted probabilities corresponding to the predicted welding spot processing paths of the spot welding robots; and select, from the predicted welding spot processing paths of the spot welding robots, a predicted welding spot processing path corresponding to a maximum predicted probability as the welding spot processing path of the spot welding robot.

[0208] In an example embodiment, the spot welding robot path determination apparatus further includes a model training module configured to: obtain sample welding spot information of sample welding spots and sample robot numbers of sample spot welding robots; cluster the sample welding spots according to the sample welding spot information, to obtain sample welding spot sets corresponding to the sample robot numbers; determine sample welding spot sets corresponding to the sample spot welding robots from the sample welding spot sets and determine position information of the sample spot welding robots; input the position information of the sample welding spots in the sample welding spot sets corresponding to the sample spot welding robots and the position information of the sample spot welding robots into a welding spot processing path prediction model to be trained, to obtain predicted welding spot processing paths of the sample spot welding robots; and perform iterative training on the welding spot processing path prediction model to be trained according to differences between the predicted welding spot processing paths of the sample spot welding robots and actual welding spot processing paths, to obtain the pre-trained welding spot processing path prediction model.

[0209] In an example embodiment, the model training module is further configured to determine first path length information corresponding to the predicted weld point processing path of each sample spot welding robot, and determine second path length information corresponding to the actual weld point processing path of each sample spot welding robot; and perform iterative training on the weld point processing path prediction model to be trained according to the difference between the first path length information and the second path length information, to obtain the pre-trained weld point processing path prediction model.

[0210] In an example embodiment, the spot welding robot path determination apparatus further comprises an instruction control module configured to generate weld point processing instructions for each spot welding robot according to the weld point processing path of each spot welding robot, and control each spot welding robot to perform spot welding processing on the weld points to be processed in the corresponding weld point set according to the weld point processing instructions of each spot welding robot.

[0211] The modules in the spot welding robot path determination apparatus described above can be implemented in whole or in part by software, hardware, and combinations thereof. The modules described above can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory in the computer device in software form, so as to be called and executed by the processor to perform the operations corresponding to the modules.

[0212] In an example embodiment, a computer device is provided, which can be a server, and an internal structure diagram of the computer device can be as shown in Figure 19 The computer device comprises a processor, a memory, an input / output interface (I / O), and a communication interface. The processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The database of the computer device is configured to store weld point information, the number of robots, and other data. The input / output interface of the computer device is configured to exchange information between the processor and external devices. The communication interface of the computer device is configured to communicate with external terminals through network connection. The computer program is executed by the processor to implement a spot welding robot path determination method.

[0213] Those skilled in the art can understand that Figure 19 The structure shown in the above-mentioned embodiments is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can comprise more or fewer components than those shown in the diagram, or combine certain components, or have a different arrangement of components.

[0214] In an example embodiment, a computer device is also provided, including a memory and a processor, the memory storing a computer program, and the processor implementing the steps in the above method embodiments when executing the computer program.

[0215] In an example embodiment, a computer readable storage medium is provided, storing a computer program, and the computer program implementing the steps in the above method embodiments when executed by a processor.

[0216] In an example embodiment, a computer program product is provided, including a computer program, and the computer program implementing the steps in the above method embodiments when executed by a processor.

[0217] A person of ordinary skill in the art can understand that all or part of the processes in the above method embodiments can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above method embodiments. Any reference to a memory, database or other medium used in the embodiments provided in the present application can include at least one of a non-volatile and volatile memory. The non-volatile memory can include a read-only memory (ROM), a magnetic tape, a floppy disk, a flash memory, an optical storage, a high-density embedded non-volatile memory, a resistive memory (ReRAM), a magnetoresistive random access memory (MRAM), a ferroelectric memory (FRAM), a phase change memory (PCM), a graphene memory, etc. The volatile memory can include a random access memory (RAM) or an external cache memory, etc. As an illustration but not limitation, the RAM can be in various forms, such as a static random access memory (SRAM) or a dynamic random access memory (DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., without being limited thereto.

[0218] The technical features of the above embodiments can be combined in any manner. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described, however, as long as the combinations of the technical features do not contradict each other, they should be considered to be within the scope of the present disclosure.

[0219] The above embodiments only express several implementation manners of the present application, and the description is relatively specific and detailed, but it should not be understood as a limitation on the patent scope of the present application. It should be pointed out that, for ordinary skilled persons in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A method for determining a spot welding robot path, characterized in that: The method comprises: Obtaining the welding spot information of the welding spot to be processed and the robot number of the spot welding robot; Clustering the weld points to be processed according to the weld point information to obtain a weld point set corresponding to the number of robots; Determining, from the welding point sets, welding point sets corresponding to the respective spot welding robots, and determining position information of the respective spot welding robots; The position information of the weld points to be processed in the weld point set corresponding to each of the spot welding robots, as well as the position information of each of the spot welding robots, are input into a pre-trained weld point processing path prediction model to obtain the weld point processing path of each of the spot welding robots; the weld point processing path of each of the spot welding robots is used to represent the path along which each of the spot welding robots performs spot welding processing on the weld points to be processed in the corresponding weld point set.

2. The method according to claim 1, characterized in that The clustering process of the to-be-processed weld points according to the weld point information to obtain a weld point set corresponding to the number of robots includes: Clustering the weld points to be processed according to the weld point information to obtain a set of candidate weld points corresponding to the number of robots and a cluster center of each candidate weld point set; Determining the loss value corresponding to each candidate weld point set according to the distance between the weld point to be processed in each candidate weld point set and the cluster center; In the case that the loss values ​​corresponding to each of the candidate welding point sets meet a preset loss value condition, each of the candidate welding point sets is used as a welding point set corresponding to the number of robots.

3. The method according to claim 1, characterized in that Determining, from the welding point sets, the welding point sets corresponding to the spot welding robots, includes: Obtaining the cluster center of each of the solder joint sets; According to the cluster centers of the respective welding point sets and the position information of the respective spot welding robots, the welding point sets corresponding to the respective spot welding robots are determined from the welding point sets.

4. The method according to claim 1, wherein The step of inputting the position information of the weld points to be processed in the weld point set corresponding to each of the spot welding robots and the position information of each of the spot welding robots into a pre-trained weld point processing path prediction model to obtain the weld point processing path of each of the spot welding robots comprises: Inputting the position information of the weld points to be processed in the weld point set corresponding to each of the spot welding robots and the position information of each of the spot welding robots into a pre-trained weld point processing path prediction model to obtain a predicted weld point processing path of each of the spot welding robots and a prediction probability corresponding to the predicted weld point processing path of each of the spot welding robots; From the predicted weld spot processing paths of the spot welding robots, the predicted weld spot processing path with the largest corresponding prediction probability is screened out as the weld spot processing path of the spot welding robots.

5. The method according to claim 1, wherein The pre-trained solder joint processing path prediction model is trained in the following manner: obtaining sample welding point information of the sample welding point and the sample robot number of the sample spot welding robot; performing clustering processing on the sample welding points according to the sample welding point information to obtain a set of sample welding points corresponding to the number of the sample robots; Determining, from the sample welding point sets, a sample welding point set corresponding to each of the sample spot welding robots, and determining position information of each of the sample spot welding robots; Inputting the position information of the sample weld points in the sample weld point set corresponding to each of the sample weld point robots and the position information of each of the sample weld point robots into the weld point processing path prediction model to be trained to obtain the predicted weld point processing path of each of the sample weld point robots; Determining first path length information corresponding to a predicted weld spot processing path of each of the sample spot welding robots, and determining second path length information corresponding to an actual weld spot processing path of each of the sample spot welding robots; The solder joint processing path prediction model to be trained is iteratively trained according to the difference between the first path length information and the second path length information to obtain the pre-trained solder joint processing path prediction model.

6. The method according to any one of claims 1 to 5, characterized in that After inputting the position information of the weld points to be processed in the weld point set corresponding to each of the spot welding robots and the position information of each of the spot welding robots into a pre-trained weld point processing path prediction model to obtain the weld point processing path of each of the spot welding robots, the method further includes: generating a welding spot processing instruction for each of the spot welding robots according to the welding spot processing path of each of the spot welding robots; According to the welding spot processing instructions of each of the spot welding robots, each of the spot welding robots is controlled to perform spot welding processing on the welding spots to be processed in the corresponding welding spot set.

7. A spot welding robot path determination device, characterized in that: The device comprises: A data acquisition module is used to obtain the welding spot information of the welding spot to be processed and the robot number of the spot welding robot; A welding point clustering module, configured to perform clustering processing on the welding points to be processed according to the welding point information, and obtain a welding point set corresponding to the number of the robots; An information determination module is used to determine the welding point sets corresponding to each of the spot welding robots from the welding point sets, and to determine the position information of each of the spot welding robots; The path prediction module is used to input the position information of the unprocessed welds in the weld point set corresponding to each of the spot welding robots and the position information of each of the spot welding robots into a pre-trained weld point processing path prediction model to obtain the weld point processing path of each of the spot welding robots; the weld point processing path of each of the spot welding robots is used to represent the path along which each of the spot welding robots performs spot welding processing on the unprocessed welds in the corresponding weld point set.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images