Steel structure multi-robot collaborative assembly and welding method and device, electronic equipment and medium

Abstract

The present application provides a steel structure multi-robot collaborative assembly welding method, device, electronic equipment and medium, the steel structure multi-robot collaborative assembly welding method obtains the steel structure to be assembled and welded, an assembly welding task set, the assembly welding task set includes a plurality of assembly welding sub-task sets, the plurality of assembly welding sub-task sets are trained, the time sequence order of the plurality of assembly welding sub-task sets is obtained, the assembly welding sub-task set includes a grabbing task and a welding task, the grabbing task and the welding task are assigned to different robots, so as to control different robots to assemble and weld the steel structure to be assembled and welded according to the time sequence order of the plurality of assembly welding sub-task sets; the present application realizes the autonomous distribution and collaborative work of multi-robot and multi-task, reduces the overall welding process time, improves the welding efficiency and welding quality, and reduces the labor cost and labor intensity.

Description

Technical Field

[0001] This invention relates to the field of steel structure welding technology, and in particular to a multi-robot collaborative welding method, apparatus, electronic equipment and medium for steel structure components. Background Technology

[0002] In recent years, the proportion of steel structure buildings has gradually increased, the forms of steel structure buildings have become increasingly diverse, and the demand for steel structure components has been increasing. Steel structures include main frames and connecting components, and the assembly process is completed through the connection and welding of the main frames and connecting components. The main frames are welded steel and rolled steel columns, etc., and the connecting components include stiffening plates, supporting brackets, and connecting brackets.

[0003] Currently, steel structural components are mainly welded manually. The general welding sequence is to first weld the brackets and stiffening plates, and then weld them together with the steel columns. This requires repeated operations such as grabbing, spot welding, and flipping, which is complex and difficult. Especially for large-sized and heavy-tonnage steel components, welding corners and seams is difficult, and assisting in hoisting and flipping is strenuous, making the overall welding process too time-consuming, resulting in low welding efficiency and insufficient output.

[0004] Therefore, there is an urgent need for a multi-robot collaborative assembly and welding method to achieve autonomous allocation and collaborative operation of multiple robots and multiple tasks, reduce the overall welding process time, improve welding efficiency, and reduce labor costs and labor intensity. Summary of the Invention

[0005] In view of the shortcomings of the prior art described above, this application provides a multi-robot collaborative welding method, apparatus, electronic product, and storage medium for steel structural components to solve the above-mentioned technical problems.

[0006] According to one aspect of the embodiments of this application, a multi-robot collaborative assembly and welding method for steel structural components is provided, comprising: acquiring a steel structural component to be assembled and welded, and an assembly and welding task set, wherein the assembly and welding task set includes multiple assembly and welding sub-task sets; training the multiple assembly and welding sub-task sets to obtain a temporal sequence of the multiple assembly and welding sub-task sets, wherein the assembly and welding sub-task sets include grasping tasks and welding tasks; and assigning the grasping tasks and the welding tasks to different robots to control different robots to assemble and weld the steel structural component to be assembled and welded according to the temporal sequence of the multiple assembly and welding sub-task sets.

[0007] In one embodiment of this application, the process of training multiple sets of welding sub-tasks to obtain a temporal order of multiple sets of welding sub-tasks includes: obtaining the current welding state of the steel structure to be welded; using the current welding state of the steel structure to be welded as the initial welding state; calculating the probability of selecting all sets of welding sub-tasks based on the initial welding state; determining the next set of welding sub-tasks based on the probability of all sets of welding sub-tasks; and so on until all sets of welding sub-tasks have been selected, thereby obtaining a temporal order of multiple sets of welding sub-tasks.

[0008] In one embodiment of this application, the process of calculating the probability of selecting all sets of assembly / soldering subtasks based on the initial assembly / soldering state includes: inputting the initial assembly / soldering state into a convolutional neural network model to obtain a feature matrix between the initial assembly / soldering state and all sets of assembly / soldering subtasks; inputting the feature matrix between the initial assembly / soldering state and all sets of assembly / soldering subtasks into a first fully connected network model, and converting it through the first fully connected network model to obtain a one-dimensional feature vector; inputting the one-dimensional feature vector into a normalized exponential function, and calculating and outputting the probability of selecting all sets of assembly / soldering subtasks through the normalized exponential function.

[0009] In one embodiment of this application, the process of determining the next assembly / welding sub-task set based on the probabilities of all the assembly / welding sub-task sets until all the assembly / welding sub-task sets have been selected includes: based on the initial assembly / welding state, selecting the largest probability value among the probabilities of all the assembly / welding sub-task sets, and taking the assembly / welding sub-task set corresponding to the largest probability value as the next assembly / welding sub-task set; taking the assembly / welding state corresponding to the next assembly / welding sub-task set as the initial assembly / welding state, and repeatedly selecting the largest probability value among the probabilities of all the assembly / welding sub-task sets until all the assembly / welding sub-task sets have been selected.

[0010] In one embodiment of this application, after obtaining the temporal order of the multiple sets of assembly and welding sub-tasks, the method further includes: sorting multiple assembly and welding states corresponding to the multiple sets of assembly and welding sub-tasks according to the temporal order to form a temporal path, and using the temporal relationship between the multiple assembly and welding states as state transition links in the temporal path; determining the reward value of each state transition link by using the distance value and corresponding weight between the end point of the current assembly and welding state and the starting point of the next assembly and welding state, the robot joint rotation angle and corresponding weight between the current assembly and welding state and the next assembly and welding state, the positioner rotation angle and corresponding weight between the current assembly and welding state and the next assembly and welding state, and the collision and corresponding weight between the current assembly and welding state and the next assembly and welding state; determining the temporal path discount reward based on the reward value of each state transition link and the discount coefficient corresponding to each state transition link; and setting the current assembly and welding state corresponding to each state transition link as the current assembly and welding state. Input a convolutional neural network model and output the feature matrix of the current assembly / soldering state. Input the set of the next assembly / soldering subtasks corresponding to each state transition step into a second fully connected network model and output the feature matrix of the next assembly / soldering subtask set. Perform feature association between the feature matrix of the current assembly / soldering state and the feature matrix of the next assembly / soldering subtask set to obtain the association matrix between the current assembly / soldering state and the next assembly / soldering subtask set. Input the association matrix into a third fully connected network model to obtain the value of each state transition step. Calculate the difference between the value of the start state transition step and the value of the end state transition step in the time-series path, and the difference of the time-series path discount reward. Take the absolute value of the difference to obtain the time-series path value. With the goal of minimizing the time-series path value, update the parameters of the convolutional neural network model, the second fully connected network model, and the third fully connected network model to obtain the first updated convolutional neural network model, the second updated fully connected network model, and the third updated fully connected network model.

[0011] In one embodiment of this application, after obtaining the updated convolutional neural network model, the method further includes: multiplying the value of each state transition step by the probability of the corresponding state transition step to obtain the expected value of each state transition step; summing the expected values ​​of all state transition steps to obtain a total expected value; updating the parameters of the first updated convolutional neural network model and the first fully connected network model with the goal of maximizing the total expected value to obtain a second updated convolutional neural network model and a first updated fully connected network model; repeatedly training multiple sets of assembly and welding sub-tasks based on the second updated convolutional neural network model, the first updated fully connected network model, the second updated fully connected network model, and the third updated fully connected network model until the number of training times for multiple sets of assembly and welding sub-tasks reaches a preset threshold, and outputting the temporal order of multiple sets of assembly and welding sub-tasks.

[0012] In one embodiment of this application, before obtaining the steel structural components to be welded and the set of welding tasks, the process includes: obtaining the equipment required in the process of welding the steel structural components and the corresponding equipment information; constructing a three-dimensional simulation model based on the equipment and equipment information, so as to control different robots to perform welding operations through the three-dimensional simulation model.

[0013] According to one aspect of the embodiments of this application, a multi-robot collaborative assembly and welding device for steel structural components is provided, comprising: a steel structural component acquisition module, configured to acquire steel structural components to be assembled and welded, and an assembly and welding task set, wherein the assembly and welding task set includes multiple assembly and welding sub-task sets; a timing determination module, configured to train the multiple assembly and welding sub-task sets to obtain a timing sequence of the multiple assembly and welding sub-task sets, wherein the assembly and welding sub-task sets include grasping tasks and welding tasks; and a task allocation module, configured to allocate the grasping tasks and the welding tasks to different robots, so as to control different robots to assemble and weld the steel structural components to be assembled and welded according to the timing sequence of the multiple assembly and welding sub-task sets.

[0014] According to one aspect of the embodiments of this application, an electronic device is provided, the electronic device comprising: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the electronic device enables the multi-robot collaborative welding method for steel structural components as described above.

[0015] According to one aspect of the embodiments of this application, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a computer processor, causes the computer to perform the above-described multi-robot collaborative welding method for steel structural components.

[0016] The beneficial effects of this invention are as follows: By acquiring the steel structural components to be welded and the set of welding tasks, this invention trains multiple sets of welding sub-tasks to obtain the temporal sequence of the multiple welding sub-tasks, and assigns the grasping and welding tasks to different robots. This allows different robots to assemble and weld the steel structural components according to the temporal sequence of the multiple welding sub-tasks, realizing the autonomous allocation and collaborative operation of multiple robots and multiple tasks. This reduces the overall welding process time, improves welding efficiency and welding quality, and reduces labor costs and labor intensity, thereby further promoting the digital, intelligent, and green collaborative development of the steel structure industry.

[0017] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort. In the drawings: Figure 1 This is a schematic diagram illustrating an exemplary system architecture as shown in an exemplary embodiment of this application; Figure 2 This is a flowchart illustrating a multi-robot collaborative welding method for steel structural components, as shown in an exemplary embodiment of this application. Figure 3 This is a schematic diagram illustrating the layout of a robotic assembly and welding system, as shown in an exemplary embodiment of this application. Figure 4 This is a flowchart illustrating a robot welding subtask set training method in an exemplary embodiment of this application; Figure 5 A block diagram is shown that is suitable for implementing a multi-robot collaborative welding apparatus for steel structures in accordance with the embodiments of this application. Figure 6 A schematic diagram of the structure of a computer system suitable for implementing the electronic device of the present application is shown. Detailed Implementation

[0019] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0020] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0021] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0022] In this application, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0023] The technical solutions of this application involve technologies related to welding of steel structural components, which are specifically illustrated through the following embodiments: Figure 1 This is a schematic diagram illustrating an exemplary system architecture as shown in an exemplary embodiment of this application.

[0024] Reference Figure 1 As shown, the system architecture may include a steel structure component storage device 101 and a computer device 102. The computer device 102 may be at least one of a desktop graphics processing unit (GPU) computer, a GPU computing cluster, or a neural network computer. Those skilled in the art can use the computer device 102 to train multiple sets of welding sub-tasks to obtain the temporal order of these sub-tasks, and assign the grasping and welding tasks to different robots to control them to perform assembly and welding according to the temporal order of the multiple welding sub-task sets. The steel structure component storage device 101 stores the steel structure components to be welded and the welding task sets. In this embodiment, the steel structure component storage device 101 uses read-only memory (ROM) or random access memory (RAM) to store the steel structure components to be welded and the welding task sets, and provides them to the computer device 102 for processing.

[0025] Indicatively, after acquiring the steel structural components to be welded and the set of welding tasks from the steel structural component storage device 101, the computer device 102 trains multiple sets of welding sub-tasks to obtain the temporal sequence of the multiple welding sub-task sets, and assigns the grasping task and welding task to different robots to control different robots to perform assembly and welding according to the temporal sequence of the multiple welding sub-task sets. This realizes the autonomous allocation and collaborative operation of multiple robots and multiple tasks, reduces the overall welding process time, improves welding efficiency and welding quality, and reduces labor costs and labor intensity, thereby further promoting the digital, intelligent, and green collaborative development of the steel structure industry.

[0026] It should be noted that the multi-robot collaborative welding method for steel structural components provided in this application embodiment is generally executed by computer equipment 102, and correspondingly, the multi-robot collaborative welding device for steel structural components is generally installed in computer equipment 102.

[0027] The implementation details of the technical solutions in the embodiments of this application are described in detail below: Figure 2 This is a flowchart illustrating a multi-robot collaborative welding method for steel structural components, as shown in an exemplary embodiment of this application. This multi-robot collaborative welding method for steel structural components can be executed by a computational processing device, which may be... Figure 1 The computer device 102 shown is illustrated. (Refer to...) Figure 2 As shown, the multi-robot collaborative welding method for steel structural components includes at least steps S210 to S230, which are described in detail below: In step S210, the steel structural components to be welded and the welding task set are obtained.

[0028] In one embodiment of this application, the assembly and welding task set includes multiple assembly and welding sub-task sets, which correspond to multiple assembly and welding steel structural components. The steel structural components to be assembled and welded include stiffening plates and partitions, etc. For example, the assembly and welding sub-task set corresponding to the stiffening plate includes gripping the stiffening plate, spot welding assembly, arc welding, rotation of auxiliary mechanism station, etc.

[0029] In step S220, multiple sets of assembly and welding subtasks are trained to obtain the timing sequence of multiple sets of assembly and welding subtasks.

[0030] In this embodiment, the process of training multiple sets of welding sub-tasks to obtain the temporal order of multiple sets of welding sub-tasks includes: (1) obtaining the current welding state of the steel structure to be welded; (2) taking the current welding state of the steel structure to be welded as the initial welding state, calculating the probability of selecting all sets of welding sub-tasks based on the initial welding state, determining the next set of welding sub-tasks based on the probability of all sets of welding sub-tasks, until all sets of welding sub-tasks are selected, and obtaining the temporal order of multiple sets of welding sub-tasks.

[0031] In this embodiment, the set of welding sub-tasks includes gripping tasks and welding tasks. When the welding sub-tasks are combined to include gripping stiffeners, spot welding assembly, arc welding, and rotation of the auxiliary mechanism station, gripping stiffeners is the gripping task, spot welding assembly and arc welding are the welding tasks, and rotation of the auxiliary mechanism station is the rotation task of the positioner.

[0032] In step S230, the grasping task and the welding task are assigned to different robots so that the different robots can be controlled to assemble and weld the steel structural components to be welded in the order of multiple assembly and welding sub-task sets.

[0033] In this embodiment, by training multiple sets of assembly and welding sub-tasks, the temporal sequence of multiple sets of assembly and welding sub-tasks is obtained, and the grasping task and welding task are assigned to different robots. This allows different robots to perform assembly and welding according to the temporal sequence of multiple sets of assembly and welding sub-tasks, realizing the autonomous allocation and collaborative operation of multiple robots and multiple tasks. This reduces the overall welding process time, improves welding efficiency and welding quality, and reduces labor costs and labor intensity, thereby further promoting the digital, intelligent, and green collaborative development of the steel structure industry.

[0034] In one embodiment of this application, the process of training multiple sets of assembly and welding subtasks to obtain the temporal order of the multiple sets of assembly and welding subtasks includes: Obtain the current welding status of the steel structural components to be welded.

[0035] In this embodiment, the welding state of the steel structural component to be welded is set according to the set of welding sub-tasks. For example, the state corresponding to the steel structural component to be welded when no welding is performed is that the welding task has not been implemented; the state corresponding to the completion of welding of the stiffener plate is that the stiffener plate welding task has been completed; the state corresponding to the completion of welding of the partition plate is that the welding plate welding task has been completed. The states such as the welding task not being implemented, the stiffener plate welding task being completed, or the partition plate welding task being completed can be used as the current welding state of the steel structural component to be welded.

[0036] The current welding state of the steel structural component to be welded is taken as the initial welding state. Based on the initial welding state, the probability of selecting all welding sub-task sets is calculated respectively. Based on the probability of all welding sub-task sets, the next welding sub-task set is determined. This process continues until all welding sub-task sets are selected, resulting in the temporal order of multiple welding sub-task sets.

[0037] In this embodiment, when initially selecting the next set of welding sub-tasks, the set of welding sub-tasks with the highest probability among all sets of welding sub-tasks is selected as the next set of welding sub-tasks. The selected set of welding sub-tasks is marked to distinguish it from the unselected set of welding sub-tasks. When recalculating the probability of selecting all sets of welding sub-tasks, the probability of the selected set of welding sub-tasks is set to 0, and the process continues to select the set of welding sub-tasks with the highest probability among all sets of welding sub-tasks as the next set of welding sub-tasks until all sets of welding sub-tasks have been selected.

[0038] In one embodiment of this application, the process of calculating the probability of selecting all sets of assembly / soldering subtasks based on the initial assembly / soldering state includes: The initial assembly and welding state is input into the convolutional neural network model to obtain the feature matrix between the initial assembly and welding state and the set of all assembly and welding subtasks.

[0039] In this embodiment, the convolutional neural network model includes an input layer, a hidden layer, and an output layer. The hidden layer includes convolutional layers, activation function layers, pooling layers, fully connected layers, etc. The input layer, hidden layer, and output layer extract features between the initial assembly state and the set of all assembly subtasks, and form a feature matrix with the features between the initial assembly state and the set of all assembly subtasks.

[0040] In this embodiment, the convolutional neural network model can be a residual network (ResNet) or a densely connected network (DenseNet), etc.

[0041] The feature matrix between the initial assembly and welding state and the set of all assembly and welding subtasks is input into the first fully connected network model, and then transformed by the first fully connected network model to obtain a one-dimensional feature vector.

[0042] In this embodiment, the fully connected network model is also known as a multilayer perceptron, which can map low-dimensional data to high-dimensional data. By increasing the dimensional space in which the data resides, the data becomes linearly separable.

[0043] In this embodiment, the first fully connected network model, the second fully connected network model, and the third fully connected network model are configured with different model parameters according to the structure of the input and output.

[0044] Input the one-dimensional feature vector into the normalized exponential function, and calculate and output the probability of selecting all assembly and welding subtask sets through the normalized exponential function.

[0045] In this embodiment, the normalization exponential function is used to normalize a set of data so that the data is within a certain numerical range, generally in the range of [0, 1].

[0046] In one embodiment of this application, the process of determining the next set of assembly / welding subtasks based on the probabilities of all assembly / welding subtask sets, until all assembly / welding subtask sets have been selected, includes: Based on the initial assembly and welding state, select the highest probability value among all assembly and welding sub-task sets, and use the assembly and welding sub-task set corresponding to the highest probability value as the next assembly and welding sub-task set.

[0047] In this embodiment, after the next set of assembly and welding sub-tasks is determined for the first time, the selected set of assembly and welding sub-tasks needs to be marked. The selected set of assembly and welding sub-tasks is distinguished from the unselected set of assembly and welding sub-tasks by marking, and the probability of the selected set of assembly and welding sub-tasks is set to 0, so as to facilitate the determination of the next set of assembly and welding sub-tasks from the unselected set of assembly and welding sub-tasks.

[0048] Use the welding state corresponding to the next welding sub-task set as the initial welding state, and repeatedly select the highest probability value among all welding sub-task sets until all welding sub-task sets have been selected.

[0049] In this embodiment, after the next set of assembly and welding subtasks is determined for the first time, the assembly and welding state corresponding to the next set of assembly and welding subtasks is taken as the initial assembly and welding state, and the next assembly and welding subtask of the initial assembly and welding state is determined until all sets of assembly and welding subtasks are selected.

[0050] In one embodiment of this application, after obtaining the temporal order of multiple welding sub-task sets, the multi-robot collaborative welding method for steel structural components further includes: According to the timing order of multiple assembly and welding sub-task sets, the assembly and welding states corresponding to the assembly and welding sub-task sets are sorted to form a timing path, and the timing relationship between multiple assembly and welding states is used as the state transition link in the timing path.

[0051] In this embodiment, the timing path includes timing nodes and state transition links. The timing nodes are the assembly and welding states corresponding to the assembly and welding subtask set, and the state transition links are the timing relationships between multiple assembly and welding states.

[0052] The reward value for each state transition stage is determined by the distance between the end point of the current welding state and the start point of the next welding state, the robot joint rotation angle between the current welding state and the next welding state, the positioner rotation angle between the current welding state and the next welding state, and the collision between the current welding state and the next welding state, along with the collision weight.

[0053] In this embodiment, each welding state corresponds to a start point and an end point. Let `dis` represent the distance between the end point of the current welding state and the start point of the next welding state in each state transition. Let `angle1` represent the robot joint rotation angle from the current welding state to the next welding state, `angle2` represent the positioner rotation angle from the current welding state to the next welding state, and `col` represent a collision occurring from the current welding state to the next welding state. Taking an example where `dis` corresponds to a weight of 0.5, `angle1` to a weight of 0.1, `angle2` to a weight of 0.2, and `col` to a weight of 0.1, the calculation formula for the reward value R of each state transition is as follows: R = -(0.5*dis + 0.1*angle1 + 0.2*angle2 + 0.1*col) (Equation 1) Where R is the reward value for each state transition, dis is the distance between the end point of the current welding state and the start point of the next welding state in each state transition, angle1 is the robot joint rotation angle from the current welding state to the next welding state, angle2 is the positioner rotation angle from the current welding state to the next welding state, and col is the value of 100 when a collision occurs from the current welding state to the next welding state, and 0 when no collision occurs.

[0054] In this embodiment, since R is negative, the larger the values ​​of dis, angle1, angle2, and col are, the smaller the reward value R becomes. The reward value of each state transition step comprehensively considers factors such as dis, angle1, angle2, and col, so that the resources consumed when multiple robots complete all tasks are minimized (reflected in no collisions, smooth robot joint movements, and short idle distances), while all tasks (the entire assembly and welding task) can be completed in the shortest possible time.

[0055] The timing path discount reward is determined based on the reward value of each state transition stage and the corresponding discount coefficient.

[0056] In this embodiment, taking a timing path with 5 assembly / soldering states as an example, the timing path has 4 state transition stages R. 12 R 23 R34 and R 45 R 12 The corresponding discount factor is 1, R 23 The corresponding discount factor is r, R 34 The corresponding discount factor is r 2 R 45 The corresponding discount factor is r 3 The formula for calculating the time-series path discount reward U is as follows: U=R 12 +r*R 23 +r 2 *R 34 +r 3 *R 45 Equation (2) Where U represents the time-series path discount reward, and R... 12 R 23 R 34 and R 45 These represent the four state transition stages in the timing path, where r is the discount coefficient, and its value ranges from 0 to 1.

[0057] The current assembly / welding state corresponding to each state transition step is input into a convolutional neural network model, which outputs a feature matrix of the current assembly / welding state. The set of the next assembly / welding subtasks corresponding to each state transition step is input into a second fully connected network model, which outputs a feature matrix of the next assembly / welding subtasks. The feature matrix of the current assembly / welding state and the feature matrix of the next assembly / welding subtasks are correlated to obtain an association matrix between the current assembly / welding state and the next assembly / welding subtasks. The association matrix is ​​then input into a third fully connected network model to obtain the value of each state transition step.

[0058] In this embodiment, the convolutional neural network model includes an input layer, a hidden layer, and an output layer. The hidden layer includes convolutional layers, activation function layers, pooling layers, fully connected layers, etc. The input layer, hidden layer, and output layer extract the feature matrix of the current assembly state corresponding to each state transition step. The fully connected network model extracts the feature matrix of the next assembly subtask set corresponding to each state transition step.

[0059] In this embodiment, the method of performing feature association between the feature matrix of the current assembly and welding state and the feature matrix of the next assembly and welding sub-task set to obtain the association matrix between the current assembly and welding state and the next assembly and welding sub-task set is not specifically limited.

[0060] In this embodiment, the third fully connected network model transforms the correlation matrix between the current assembly / soldering state and the next assembly / soldering subtask set into a one-dimensional feature vector, and calculates the value of each state transition step through the one-dimensional feature vector.

[0061] Calculate the difference between the value of the start state transition stage and the value of the end state transition stage in the time sequence path, and the time sequence path discount reward, and take the absolute value of the difference to obtain the time sequence path value.

[0062] In this embodiment, the timing path has 5 assembly / soldering states, and the value of the initial state transition stage is Q. 12 The value of the final state transition stage is Q. 45 For example, the formula for calculating the time-series path value Q is as follows: Q=|Q 12 -r 4 *Q 45 -U| Formula (3) Where Q is the time-series path value, Q 45 To determine the value of the final state transition phase, Q 12 Let U be the value of the initial state transition stage, U be the time-series path discount reward, and r be the discount coefficient, with the value of r ranging from (0,1).

[0063] With the goal of minimizing the temporal path value, the parameters of the convolutional neural network model, the second fully connected network model, and the third fully connected network model are updated to obtain the first updated convolutional neural network model, the second updated fully connected network model, and the third updated fully connected network model.

[0064] In this embodiment, the process of minimizing the temporal path value is to calculate the partial derivative of formula (3), update the parameters of the convolutional neural network model, the second fully connected network model, and the third fully connected network model according to the parameter values ​​obtained by calculating the partial derivative, and then continuously calculate the partial derivative of formula (3) according to the preset learning rate, so as to minimize the temporal path value.

[0065] In this embodiment, the policy function network model consists of a convolutional neural network model and a first fully connected network model. The parameters of the convolutional neural network model and the first fully connected network model in the policy function network model are represented by... express.

[0066] In this embodiment, the value function network model consists of a convolutional neural network model, a second fully connected network model, and a third fully connected network model. The parameters of the convolutional neural network model, the second fully connected network model, and the third fully connected network model in the value function network model are represented by... This indicates that the convolutional neural network model in the value function network model and the convolutional neural network model in the policy function network model are shared.

[0067] In one embodiment of this application, after obtaining the updated convolutional neural network model, the multi-robot collaborative welding method for steel structural components further includes: The expected value of each state transition is obtained by multiplying the value of each state transition by the probability of the corresponding state transition.

[0068] In this embodiment, the probability corresponding to each state transition stage is determined by the initial assembly / soldering state of that stage and the set of the next assembly / soldering subtasks. The formula for calculating the expected value of each state transition stage is as follows: Equation (4) in, This indicates the initial assembly and soldering state for each state transition stage. This represents the expected value of each state transition step. Indicates the set of the next welding task. This indicates that the initial assembly / soldering state is 's' and the next assembly / soldering subtask set is 's'. At that time, the value of each state transition step, This indicates that the initial assembly / soldering state is 's' and the next assembly / soldering subtask set is 's'. At that time, the probability of each state transition, where, Value function network model The parameters, For policy function network models The parameters.

[0069] The total expected value is obtained by summing the expected values ​​of all state transition stages.

[0070] In this embodiment, the formula for calculating the total expected value is as follows: Equation (5) in, The total expected value, This indicates the initial assembly and soldering state for each state transition stage. This represents the expected value of each state transition step. This indicates the set of the next welding task. This indicates that the initial assembly / soldering state is 's' and the next assembly / soldering subtask set is 's'. At that time, the value of each state transition step, This indicates that the initial assembly / soldering state is 's' and the next assembly / soldering subtask set is 's'. At that time, the probability of each state transition, where, Value function network model The parameters, For policy network model The parameters.

[0071] With the goal of maximizing the total expected value, the parameters of the first updated convolutional neural network model and the first fully connected network model are updated to obtain the second updated convolutional neural network model and the first updated fully connected network model.

[0072] In this embodiment, after obtaining the first updated convolutional neural network model and the first updated fully connected network model, the parameters of the value function network model are... As known values, the total expected value is maximized. The process of maximizing the total expected value involves taking the partial derivative of formula (5) and applying the parameter values ​​obtained from the partial derivatives to the parameters in the policy network model. The model is updated (i.e., the first updated convolutional neural network model and the first fully connected network model are updated), and then, based on the pre-set learning rate, the partial derivatives of formula (5) are continuously calculated to maximize the total expected value.

[0073] Based on the second updated convolutional neural network model, the first updated fully connected network model, the second updated fully connected network model, and the third updated fully connected network model, the training is repeated on multiple sets of assembly and welding sub-tasks until the number of training times for multiple sets of assembly and welding sub-tasks reaches a preset threshold, and the temporal order of multiple sets of assembly and welding sub-tasks is output.

[0074] In this embodiment, the preset number of times threshold can be set according to the actual situation.

[0075] In one embodiment of this application, before obtaining the steel structural components to be welded and the set of welding tasks, the process includes: Obtain the equipment and corresponding equipment information required during the assembly and welding of steel structural components.

[0076] In this embodiment, the equipment required for assembling and welding steel structural components includes multiple robots, robot end effectors, positioners, and sensors. The robot end effectors include welding torches, suction cups, and grippers. When a welding torch is installed at the end of a robot arm, it enables spot welding and other welding functions. When suction cups or grippers are installed at the end of a robot arm, it enables grasping functions. The positioner is used to flip the steel structural components to be welded. Sensors include cameras, scanners, and contact detection devices to acquire the position of the workpiece and determine if a collision has occurred. Equipment information includes the model and dimensions of the equipment, its location information, and the coordination relationships between the various devices.

[0077] A 3D simulation model is built based on the equipment and equipment information, so as to control different robots to perform welding operations through the 3D simulation model.

[0078] In this embodiment, after establishing a three-dimensional simulation model according to the model and size of the equipment, the location information of the equipment, and the coordination relationship of each piece of equipment, a system environment simulation platform is built according to the arrangement of the equipment. The simulation platform can control the movement of moving parts (such as robots and guide rails) in the assembly and welding system in the form of instructions.

[0079] This application acquires the steel structural components to be assembled and welded, and sets of assembly and welding tasks. It trains multiple sets of assembly and welding sub-tasks to obtain the temporal sequence of these sub-tasks. The grasping and welding tasks are then assigned to different robots to control them to assemble and weld the steel structural components according to the temporal sequence of the multiple assembly and welding sub-tasks. This achieves autonomous allocation and collaborative operation of multiple robots and multiple tasks, reduces the overall welding process time, improves welding efficiency and quality, and reduces labor costs and labor intensity. This further promotes the digital, intelligent, and green collaborative development of the steel structure industry.

[0080] Please see Figure 3 , Figure 3 This is a schematic diagram illustrating the layout of a robotic assembly and welding system, as shown in an exemplary embodiment of this application. It includes a robot module, an end effector module, a positioner, and a sensor module. The robot module comprises 2-3 cooperating robots (e.g., robot 1 and robot 2) and external robot axes (moving guides, controlled by the robot's internal system) for linear motion, used to perform the assembly and welding of steel structural components. The end effector module includes welding torches, suction cups, grippers, and other robot end tools for grasping and welding steel structural components. The positioner rotates the steel structural components, enabling the robot to assemble and weld various parts of the workpiece. The sensor module mainly includes cameras, scanners, and contact detectors, used to acquire the workpiece's position and determine if a collision has occurred. The control module allocates the assembly and welding sequence tasks, issues robot movement paths, and controls the robot end effector to ultimately achieve the assembly and welding process.

[0081] In this embodiment, for the steel structure welding task that will be completed in the subsequent production, the name and type of the steel structure to be welded are input. Then, according to the multi-robot collaborative welding method for steel structure, the global assembly and welding task set is divided into a set of sub-tasks with a temporal order. These sub-tasks are then allocated and distributed to the task storage area of ​​each designated robot in the robot assembly and welding system, and the robots are controlled to cooperate with each other to realize the assembly and welding of the steel structure.

[0082] Please see Figure 4 , Figure 4This is a flowchart illustrating a robot welding sub-task set training method in an exemplary embodiment of this application. The method includes: (1) initializing the simulation experimental environment, i.e., obtaining the equipment and corresponding equipment information required in the welding process of steel structural parts, constructing a three-dimensional simulation model based on the equipment and equipment information, so as to control different robots to perform welding operations through the three-dimensional simulation model; (2) calculating the probability of selecting all welding sub-task sets based on the initial welding state, determining the next welding sub-task set based on the probability of all welding sub-task sets, and forming a time-series path; (3) calculating the time-series path value, minimizing the time-series path value, and updating the parameters of the value function network model; (4) calculating the total expected value, maximizing the total expected value, and updating the parameters in the policy function network model when the updated parameters of the value function network model are known; (5) executing the loop until the training ends.

[0083] The following describes an embodiment of the apparatus described in this application, which can be used to execute the multi-robot collaborative welding method for steel structural components described in the above embodiments of this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the multi-robot collaborative welding method for steel structural components described above in this application.

[0084] Figure 5 This is a block diagram illustrating a multi-robot collaborative welding device for steel structural members, as shown in an exemplary embodiment of this application. The device can be applied to... Figure 1 The implementation environment shown is specifically configured in computer device 102. This device can also be applied to other exemplary implementation environments and specifically configured in other devices. This embodiment does not limit the implementation environment to which the device is applicable.

[0085] like Figure 5 As shown, this exemplary multi-robot collaborative welding device for steel structural components includes: The steel structure component acquisition module 501 is used to acquire steel structure components to be welded and a set of welding tasks. The set of welding tasks includes multiple sets of welding sub-tasks.

[0086] The timing determination module 502 is used to train multiple sets of assembly and welding sub-tasks to obtain the timing order of multiple sets of assembly and welding sub-tasks, which include grasping tasks and welding tasks.

[0087] The task allocation module 503 is used to assign gripping tasks and welding tasks to different robots, so as to control different robots to assemble and weld the steel structural parts to be welded in the order of multiple assembly and welding sub-task sets.

[0088] In one embodiment of this application, the assembly and welding task set includes multiple assembly and welding sub-task sets, which correspond to multiple assembly and welding steel structural components. The steel structural components to be assembled and welded include stiffening plates and partitions, etc. For example, the assembly and welding sub-task set corresponding to the stiffening plate includes gripping the stiffening plate, spot welding assembly, arc welding, rotation of auxiliary mechanism station, etc.

[0089] In this embodiment, the process of training multiple sets of welding sub-tasks to obtain the temporal order of multiple sets of welding sub-tasks includes: (1) obtaining the current welding state of the steel structure to be welded; (2) taking the current welding state of the steel structure to be welded as the initial welding state, calculating the probability of selecting all sets of welding sub-tasks based on the initial welding state, determining the next set of welding sub-tasks based on the probability of all sets of welding sub-tasks, until all sets of welding sub-tasks are selected, and obtaining the temporal order of multiple sets of welding sub-tasks.

[0090] In this embodiment, the set of welding sub-tasks includes gripping tasks and welding tasks. When the welding sub-tasks are combined to include gripping stiffeners, spot welding assembly, arc welding, and rotation of the auxiliary mechanism station, gripping stiffeners is the gripping task, spot welding assembly and arc welding are the welding tasks, and rotation of the auxiliary mechanism station is the rotation task of the positioner.

[0091] In this embodiment, by training multiple sets of assembly and welding sub-tasks, the temporal sequence of multiple sets of assembly and welding sub-tasks is obtained, and the grasping task and welding task are assigned to different robots. This allows different robots to perform assembly and welding according to the temporal sequence of multiple sets of assembly and welding sub-tasks, realizing the autonomous allocation and collaborative operation of multiple robots and multiple tasks. This reduces the overall welding process time, improves welding efficiency and welding quality, and reduces labor costs and labor intensity, thereby further promoting the digital, intelligent, and green collaborative development of the steel structure industry.

[0092] It should be noted that the multi-robot collaborative welding device for steel structures provided in the above embodiments and the multi-robot collaborative welding method for steel structures provided in the above embodiments belong to the same concept. The specific methods of operation of each module and unit have been described in detail in the method embodiments and will not be repeated here. In practical applications, the multi-robot collaborative welding device for steel structures provided in the above embodiments can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. This is not a limitation here.

[0093] Embodiments of this application also provide an electronic device, including: one or more processors; and a storage device for storing one or more programs, which, when executed by one or more processors, enable the electronic device to implement the multi-robot collaborative welding method for steel structural components provided in the above embodiments.

[0094] Figure 6 A schematic diagram of a computer system suitable for implementing the embodiments of this application is shown. It should be noted that... Figure 6 The computer system 600 of the electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0095] like Figure 6 As shown, the computer system 600 includes a Central Processing Unit (CPU) 601, which can perform various appropriate actions and processes, such as executing the methods described in the above embodiments, based on programs stored in Read-Only Memory (ROM) 602 or programs loaded from storage portion 608 into Random Access Memory (RAM) 603. The RAM 603 also stores various programs and data required for system operation. The CPU 601, ROM 602, and RAM 603 are interconnected via a bus 604. An Input / Output (I / O) interface 605 is also connected to the bus 604.

[0096] The following components are connected to I / O interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to I / O interface 605 as needed. A removable medium 611, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 610 as needed so that computer programs read from it can be installed into storage section 608 as needed.

[0097] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 609, and / or installed from removable medium 611. When the computer program is executed by central processing unit (CPU) 601, it performs various functions defined in the system of this application.

[0098] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable computer program. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0099] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0100] The units described in the embodiments of this application can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0101] Another aspect of this application provides a computer-readable storage medium storing computer-readable instructions that, when executed by a computer's processor, cause the computer to perform the multi-robot collaborative welding method for steel structures provided in the various embodiments above. This computer-readable storage medium may be included in the electronic devices described in the above embodiments, or it may exist independently and not assembled into the electronic devices.

[0102] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0103] Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, touch terminal, or network device, etc.) to execute the method according to the embodiments of this application.

[0104] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the embodiments disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0105] It should be understood that the above content is only a preferred exemplary embodiment of this application and is not intended to limit the implementation of this application. Those skilled in the art can easily make corresponding modifications or alterations based on the main concept and spirit of this application. Therefore, the scope of protection of this application should be determined by the scope of protection claimed in the claims.

Claims

1. A multi-robot collaborative welding method for steel structural components, characterized in that, include: Obtain the steel structural components to be assembled and welded, and the set of assembly and welding tasks, wherein the set of assembly and welding tasks includes multiple sets of assembly and welding sub-tasks; Training is performed on multiple sets of assembly and welding subtasks to obtain a temporal order of the multiple sets of assembly and welding subtasks. The assembly and welding subtask sets include grasping tasks and welding tasks. The process includes obtaining the current assembly and welding state of the steel structure to be assembled and welded; using the current assembly and welding state of the steel structure to be assembled and welded as the initial assembly and welding state, calculating the probability of selecting each of the assembly and welding subtask sets based on the initial assembly and welding state, determining the next assembly and welding subtask set based on the probability of each of the assembly and welding subtask sets, until all the assembly and welding subtask sets have been selected, thus obtaining a temporal order of the multiple sets of assembly and welding subtasks. The process of calculating the probability of selecting all sets of assembly and soldering subtasks based on the initial assembly and soldering state includes: inputting the initial assembly and soldering state into a convolutional neural network model to obtain a feature matrix between the initial assembly and soldering state and all sets of assembly and soldering subtasks; inputting the feature matrix between the initial assembly and soldering state and all sets of assembly and soldering subtasks into a first fully connected network model, and transforming it through the first fully connected network model to obtain a one-dimensional feature vector; inputting the one-dimensional feature vector into a normalized exponential function, and calculating and outputting the probability of selecting all sets of assembly and soldering subtasks through the normalized exponential function. According to the temporal order of the multiple sets of assembly and welding sub-tasks, the multiple assembly and welding states corresponding to the sets of assembly and welding sub-tasks are sorted to form a temporal path, and the temporal relationship between the multiple assembly and welding states is used as the state transition link in the temporal path. The reward value for each state transition step is determined by the distance between the end point of the current welding state and the start point of the next welding state and the corresponding weight, the robot joint rotation angle between the current welding state and the next welding state and the corresponding weight, the positioner rotation angle between the current welding state and the next welding state and the corresponding weight, and the collision between the current welding state and the next welding state and the corresponding weight. The timing path discount reward is determined based on the reward value of each state transition step and the discount coefficient corresponding to each state transition step. The current assembly / soldering state corresponding to each state transition step is input into a convolutional neural network model, which outputs the feature matrix of the current assembly / soldering state. The set of next assembly / soldering subtasks corresponding to each state transition step is input into a second fully connected network model, which outputs the feature matrix of the next assembly / soldering subtasks. The feature matrix of the current assembly / soldering state and the feature matrix of the next assembly / soldering subtasks are correlated to obtain the correlation matrix between the current assembly / soldering state and the next assembly / soldering subtasks. The correlation matrix is ​​input into a third fully connected network model to obtain the value of each state transition step. Calculate the difference between the value of the start state transition stage and the value of the end state transition stage in the time series path, and the discount reward of the time series path, and take the absolute value of the difference to obtain the value of the time series path; With the goal of minimizing the temporal path value, the parameters of the convolutional neural network model, the second fully connected network model, and the third fully connected network model are updated to obtain the first updated convolutional neural network model, the second updated fully connected network model, and the third updated fully connected network model. The grasping task and the welding task are assigned to different robots so that the different robots can be controlled to assemble and weld the steel structural components to be welded in a sequential order of multiple sets of assembly and welding sub-tasks.

2. The multi-robot collaborative welding method for steel structural components according to claim 1, characterized in that, The process of determining the next set of assembly / welding subtasks based on the probabilities of all the aforementioned assembly / welding subtask sets, until all the aforementioned assembly / welding subtask sets have been selected, includes: Based on the initial assembly and welding state, select the highest probability value among all the probabilities of the assembly and welding sub-task sets, and take the assembly and welding sub-task set corresponding to the highest probability value as the next assembly and welding sub-task set; The welding state corresponding to the next welding sub-task set is used as the initial welding state. The highest probability value among all the welding sub-task sets is repeatedly selected until all welding sub-task sets have been selected.

3. The multi-robot collaborative welding method for steel structural components according to claim 1, characterized in that, After obtaining the updated convolutional neural network model, the method further includes: The expected value of each state transition step is obtained by multiplying the value of the corresponding state transition step by the probability of the state transition step. The total expected value is obtained by summing the expected values ​​of all the state transition steps. With the goal of maximizing the total expected value, the parameters of the first updated convolutional neural network model and the first fully connected network model are updated to obtain the second updated convolutional neural network model and the first updated fully connected network model. Based on the second updated convolutional neural network model, the first updated fully connected network model, the second updated fully connected network model, and the third updated fully connected network model, the multiple sets of assembly and welding sub-tasks are repeatedly trained until the number of training times for the multiple sets of assembly and welding sub-tasks reaches a preset threshold, and the temporal order of the multiple sets of assembly and welding sub-tasks is output.

4. The multi-robot collaborative welding method for steel structural components according to any one of claims 1-3, characterized in that, Before obtaining the steel structural components to be assembled and welded, and the set of assembly and welding tasks, the following steps are included: Obtain the equipment and corresponding equipment information required during the assembly and welding of steel structural components; A three-dimensional simulation model is constructed based on the equipment and equipment information, so as to control different robots to perform welding operations through the three-dimensional simulation model.

5. A multi-robot collaborative welding device for steel structural components, characterized in that, include: The steel structure component acquisition module is used to acquire steel structure components to be assembled and welded and a set of assembly and welding tasks, wherein the set of assembly and welding tasks includes multiple sets of assembly and welding sub-tasks. A timing determination module is used to train multiple sets of assembly and welding sub-tasks to obtain a timing order of the multiple sets of assembly and welding sub-tasks. Each set of assembly and welding sub-tasks includes a grasping task and a welding task. The module includes: acquiring the current assembly and welding state of the steel structure component to be assembled and welded; using the current assembly and welding state of the steel structure component to be assembled and welded as an initial assembly and welding state; calculating the probability of selecting each of the assembly and welding sub-task sets based on the initial assembly and welding state; determining the next assembly and welding sub-task set based on the probabilities of all the assembly and welding sub-task sets; and continuing until all the assembly and welding sub-task sets have been selected, thus obtaining a timing order of the multiple sets of assembly and welding sub-tasks. Specifically, the module calculates the probability of selecting each assembly and welding sub-task set based on the initial assembly and welding state. The probability calculation process includes: inputting the initial assembly / soldering state into a convolutional neural network model to obtain a feature matrix between the initial assembly / soldering state and all the assembly / soldering sub-task sets; inputting the feature matrix between the initial assembly / soldering state and all the assembly / soldering sub-task sets into a first fully connected network model, and transforming it into a one-dimensional feature vector; inputting the one-dimensional feature vector into a normalized exponential function, calculating and outputting the probability of selecting all the assembly / soldering sub-task sets through the normalized exponential function; sorting the assembly / soldering states corresponding to the assembly / soldering sub-task sets according to their temporal order to form a temporal path, and establishing the temporal relationship between the multiple assembly / soldering states. As state transition links in the aforementioned time-series path; the reward value for each state transition link is determined by the distance between the end point of the current welding state and the start point of the next welding state and its corresponding weight, the robot joint angle between the current welding state and the next welding state and its corresponding weight, the positioner rotation angle between the current welding state and the next welding state and its corresponding weight, and the collision between the current welding state and the next welding state and its corresponding weight; the time-series path discount reward is determined based on the reward value of each state transition link and the discount coefficient corresponding to each state transition link; the current welding state corresponding to each state transition link is input into a convolutional neural network. The network model outputs the feature matrix of the current assembly and soldering state. The set of next assembly and soldering sub-tasks corresponding to each state transition stage is input into the second fully connected network model, which outputs the feature matrix of the next assembly and soldering sub-task set. The feature matrix of the current assembly and soldering state is correlated with the feature matrix of the next assembly and soldering sub-task set to obtain the correlation matrix between the current assembly and soldering state and the next assembly and soldering sub-task set. This correlation matrix is ​​then input into the third fully connected network model to obtain the value of each state transition stage. The difference between the value of the start state transition stage and the value of the end state transition stage in the time-series path, and the difference in the time-series path discount reward, is calculated. The absolute value of this difference is then taken to obtain the time-series path value.With the goal of minimizing the temporal path value, the parameters of the convolutional neural network model, the second fully connected network model, and the third fully connected network model are updated to obtain the first updated convolutional neural network model, the second updated fully connected network model, and the third updated fully connected network model. The task allocation module is used to assign the grasping task and the welding task to different robots, so as to control different robots to assemble and weld the steel structure to be welded in the order of the multiple sets of assembly and welding sub-tasks.

6. An electronic device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the electronic device to implement the multi-robot collaborative welding method for steel structures as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, It stores computer-readable instructions, which, when executed by the computer's processor, cause the computer to perform the multi-robot collaborative welding method for steel structures as described in any one of claims 1 to 4.

Images