Multi-robot collaborative welding operation method and system

Abstract

This invention provides a method and system for multi-robot collaborative welding operations. The method includes: obtaining a state space pool, the state space pool including a global state space for storing welding task information and attribute information of the workpiece to be processed, and a local state space for storing the actual state information of the robots; decomposing the welding task into several sub-tasks based on the welding task information and the attribute information of the workpiece to be processed, and determining the target state information corresponding to each sub-task; assigning the sub-tasks to several robots based on the matching relationship between the target state information of the sub-tasks and the actual state information of the robots, and controlling the robots to execute the corresponding welding operations according to the assigned sub-tasks. This method utilizes the collaborative structure of the global state space and the local state space to decouple the welding task planning and execution process, and achieves dynamic adaptation between the two.

Description

Technical Field

[0001] This invention relates to the field of intelligent welding technology, and more specifically to a method and system for multi-robot collaborative welding operations. Background Technology

[0002] Welding is a crucial process in modern manufacturing, widely used in automotive, aerospace, shipbuilding, and many other fields. With the continuous expansion of production scale and the increasing complexity of products, the demands for welding efficiency and quality are also rising. Multi-robot collaborative welding, as an advanced production method, hinges on how to efficiently and accurately allocate collaborative tasks to multiple robots to fully leverage their collective advantages and achieve efficient, high-quality welding production. Through a scientifically sound task allocation mechanism, multiple robots can simultaneously weld multiple workpieces or different parts of large, complex workpieces. This not only significantly shortens the production cycle and improves overall efficiency but also enhances welding consistency and stability through standardized task processes, reducing the interference of human factors on welding quality and thus meeting the stringent requirements of high-precision, high-quality welding.

[0003] Currently, multi-robot collaborative welding task allocation mainly adopts two control methods: centralized and distributed. Centralized control relies on a central controller, the "brain," to uniformly command the system. This controller collects the real-time status of all robots, generates control commands based on global objectives, and distributes them via the network. However, if the central controller fails, the entire system will shut down, severely impacting continuous production. Furthermore, as the number of robots or the complexity of tasks increases, the data deluge between the central node and terminals can easily cause delays and congestion, slowing down welding accuracy and response speed.

[0004] Distributed control, on the other hand, adopts a "decentralized" approach. Specifically, each robot makes autonomous decisions based on local information, negotiates task allocation through information exchange with neighboring robots, and dynamically coordinates its actions. However, this model also faces key challenges: the lack of global scheduling may lead to task allocation conflicts. For example, when welding large structural components, different robots may choose conflicting welding paths and parameters due to local optimization, resulting in weld misalignment and quality fluctuations. In addition, autonomous negotiation without a central coordinator can easily lead to resource competition (such as conflicts over the allocation of shared welding torches) or task overlap, resulting in low overall collaborative efficiency. Summary of the Invention

[0005] One of the objectives of this invention is to provide a multi-robot collaborative welding operation method to solve the technical problem in the prior art that it is difficult to balance the local matching of tasks and robots with the global coordination efficiency, resulting in the allocation results being difficult to achieve both effective execution and efficient collaboration.

[0006] To achieve one of the aforementioned objectives, this invention provides a multi-robot collaborative welding operation method, comprising: obtaining a state space pool, the state space pool including a global state space for storing welding task information and attribute information of the workpiece to be processed, and a local state space for storing the actual state information of the robots; decomposing the welding task into several sub-tasks based on the welding task information and the attribute information of the workpiece to be processed, and determining the target state information corresponding to each sub-task; assigning the several sub-tasks to several robots based on the matching relationship between the target state information of the sub-tasks and the actual state information of the robots, and controlling the robots to perform corresponding welding operations according to the assigned sub-tasks.

[0007] As a further improvement of one embodiment of the present invention, the step of decomposing the welding task into several sub-tasks based on the welding task information and the attribute information of the workpiece to be processed includes: parsing the welding task information and determining several constraints, the constraints including at least one of quality constraints, production capacity constraints and process constraints; extracting and determining several structural features of the workpiece to be processed based on the attribute information of the workpiece to be processed; and dividing the welding task into several sub-tasks based on the matching relationship between the several structural features and the several constraints.

[0008] As a further improvement of one embodiment of the present invention, the step of dividing the welding task into several sub-tasks based on the matching relationship between several structural features and several constraints includes: dividing the workpiece to be processed into several areas to be welded based on several structural features; merging several areas to be welded that satisfy at least one of quality constraints, production capacity constraints, and process constraints to obtain several sub-regions; and generating several sub-tasks based on the several sub-regions and the corresponding structural features and constraints.

[0009] As a further improvement of one embodiment of the present invention, the step of merging several regions to be welded that satisfy at least one of quality constraints, production capacity constraints, and process constraints to obtain several sub-regions includes: determining a first sub-region and a second sub-region that satisfy the constraints; determining whether the distance between the first sub-region and the second sub-region is less than or equal to a first preset threshold; if so, merging the first sub-region and the second sub-region and updating the determined sub-regions; the step of generating several sub-tasks based on the several sub-regions and the corresponding structural features and constraints includes: updating the determined sub-tasks based on the updated sub-regions and the corresponding structural features and constraints.

[0010] As a further improvement of one embodiment of the present invention, the target state information includes target parameter configuration and target requirement information, the actual state parameters include actual parameter configuration and actual performance parameters, and the matching relationship between the target state information of the sub-task and the actual state information of the robot, controlling the robot to perform the corresponding welding operation according to the assigned sub-task, includes: generating an initial task allocation result based on the matching relationship between the target parameter configuration corresponding to the sub-task and the actual parameter configuration of the robot; adjusting the initial task allocation result based on the matching relationship between the target requirement information of the sub-task and the actual performance parameters of the robot, and controlling the robot to perform the corresponding welding operation according to the adjusted allocation result.

[0011] As a further improvement of one embodiment of the present invention, the step of generating an initial task allocation result based on the matching relationship between the target parameter configuration corresponding to the sub-task and the actual parameter configuration of the robot includes: determining the corresponding first position information and first process parameter information according to the target parameter configuration; determining the corresponding second position information and second process parameter information according to the actual parameter configuration; determining the first sub-task and the first robot that match the target parameter configuration and the actual parameter configuration according to the order of parameter information and position information, and allocating the first sub-task to the first robot.

[0012] As a further improvement of one embodiment of the present invention, the step of generating an initial task allocation result based on the matching relationship between the target parameter configuration corresponding to the sub-task and the actual parameter configuration of the robot includes: obtaining a first sub-task and a second sub-task, and determining the corresponding third position information and fourth position information; when the distance between the third position information and the fourth position information is less than a second preset threshold, assigning the first sub-task and the second sub-task to the same robot, or assigning the first sub-task to a robot closer to the third position information and / or assigning the second sub-task to a robot closer to the fourth position information.

[0013] As a further improvement to one embodiment of the present invention, the adjustment of the initial task allocation result based on the matching relationship between the target requirement information of the sub-task and the actual performance parameters of the robot includes: Based on the initial task allocation results, determine the number of sub-tasks currently assigned to each robot; based on the priority of each sub-task and the actual parameter configuration of the robot, determine the contribution value of the robot; sum the contribution values ​​of several robots in the initial task allocation results to obtain the corresponding total contribution value; when the total contribution value is less than or equal to a third preset threshold, adjust the initial task allocation results.

[0014] As a further improvement of one embodiment of the present invention, the step of determining the contribution value of the robot based on the priority of each sub-task and the actual state information of the robot includes: determining at least one of the corresponding capability factor and the current load factor based on the actual performance parameters of the robot; and determining the contribution value of the robot according to the priority of each sub-task and the product of the corresponding robot capability factor and the current load factor.

[0015] As a further improvement of one embodiment of the present invention, the adjustment of the initial task allocation result based on the matching relationship between the target requirement information of the sub-task and the actual performance parameters of the robot includes: determining the robot's capability factor and the priority corresponding to the sub-task based on the initial task allocation result; determining whether the sub-task with the highest priority is assigned to the robot with the highest capability factor; if not, when the parameter information of the sub-task with the highest priority matches the parameter configuration of the robot with the highest capability factor, the sub-task with the highest priority is assigned to the robot with the highest capability factor.

[0016] As a further improvement of one embodiment of the present invention, the method further includes: determining whether the timestamp and version information of the robot's local state information are consistent with the corresponding information in the state space pool; if not, updating the corresponding information in the state space pool according to the robot's local state information.

[0017] To achieve one of the aforementioned objectives, this invention provides a multi-robot collaborative welding system, comprising: a state space pool management module for constructing and managing a state space pool, the state space pool including: a global state space for storing welding task information and attribute information of the workpiece to be processed, and a local state space for storing the actual state information of the robots; a task allocation module for decomposing the welding task into several sub-tasks based on the welding task information and the attribute information of the workpiece to be processed, and determining the target state information corresponding to each sub-task; and a working robot configured to communicate with the state space pool management module and the task allocation module, and to perform corresponding welding operations based on the matching relationship between the target state information of the allocated sub-task and its corresponding actual state information.

[0018] Compared with the prior art, the embodiments of the present invention have at least one of the following beneficial effects: This invention discloses a multi-robot collaborative welding operation method. By utilizing the collaborative structure of a global state space (storing total task and workpiece information) and a local state space (storing robot real-time state), the welding task planning and execution process are decoupled, and dynamic adaptation between the two is achieved. At the same time, based on the multi-dimensional matching relationship between the sub-task target state and the actual state of the robot, each robot can accurately undertake sub-tasks that meet its capability boundaries according to its own real-time working conditions, avoiding welding defects caused by resource conflicts or insufficient capabilities. This achieves efficient dynamic allocation and precise execution of multi-robot welding tasks, improving operation quality and efficiency. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the structure of a multi-robot collaborative welding system according to one embodiment of the present invention.

[0020] Figure 2 This is a schematic diagram of the steps of a multi-robot collaborative welding operation method according to an embodiment of the present invention.

[0021] Figure 3 This is a schematic diagram of step S2 in one embodiment of the present invention.

[0022] Figure 4 This is a schematic diagram of step S23 in a specific embodiment of the present invention.

[0023] Figure 5 This is a schematic diagram of step S3 in one embodiment of the present invention.

[0024] Figure 6(a) is a schematic diagram of step S23 in a specific embodiment of the present invention.

[0025] Figure 6(b) is a schematic diagram of step S23 in another specific embodiment of the present invention.

[0026] Figure 7(a) is a schematic diagram of step S32 in a specific embodiment of the present invention.

[0027] Figure 7(b) is a schematic diagram of step S32 in another specific embodiment of the present invention.

[0028] Figure 8 This is a schematic diagram of the steps of a multi-robot collaborative welding operation method in one embodiment of the present invention.

[0029] Figure 9(a) is a flowchart of a multi-robot collaborative welding operation method in a preferred embodiment of the present invention.

[0030] Figure 9(b) is a flowchart of a multi-robot collaborative welding operation method in a preferred embodiment of the present invention.

[0031] Figure 9(c) is a flowchart illustrating a multi-robot collaborative welding operation method in a preferred embodiment of the present invention. Detailed Implementation

[0032] The present invention will now be described in detail with reference to the specific embodiments shown in the accompanying drawings. However, these embodiments do not limit the present invention, and any structural, methodological, or functional modifications made by those skilled in the art based on these embodiments are included within the scope of protection of the present invention.

[0033] It should be noted that the term "comprising" or any other variation thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0034] like Figure 1 As shown, one embodiment of the present invention provides a multi-robot collaborative welding system 100.

[0035] The Multi-Robot Collaborative Welding System 100 is an intelligent production system that utilizes multiple robots to cooperate and improve welding operations. By integrating the functions of multiple robots and combining information management and task allocation mechanisms, it achieves efficient and precise execution of complex welding tasks, improving welding efficiency, ensuring welding quality, and adapting to welding production needs of different scales and complexities.

[0036] The multi-robot collaborative welding system 100 includes a state space pool management module 11, which is used to build and manage the state space pool.

[0037] The state space pool is the core data storage and management area of ​​the multi-robot collaborative welding system 100. It includes a global state space and local state spaces. The global state space stores welding task information and the attribute information of the workpiece to be processed, such as the overall welding objective and time constraints. The local state space stores the actual state information of the robots, such as the position and orientation of each robot.

[0038] The state space pool serves as the system's "data center," ensuring that all modules work collaboratively based on the same information benchmark, thereby achieving information sharing and data consistency.

[0039] The multi-robot collaborative welding system 100 includes a task allocation module 12. The task allocation module 12 is used to decompose the welding task into several sub-tasks based on the welding task information and the attribute information of the workpiece to be processed, and to determine the target state information corresponding to each sub-task.

[0040] The task allocation module 12 can set an ideal execution state for each subtask, such as the position the robot should reach, welding parameters, etc.; and allocate subtasks according to the actual state information of the robot to avoid resource conflicts or idle resources.

[0041] The multi-robot collaborative welding system 100 includes several collaborative robots 13. Each robot 13 is configured to communicate with the state space pool management module 11 and the task allocation module 12, and to perform the corresponding welding operation based on the target state information of the allocated sub-task and the matching relationship between it and the corresponding actual state information.

[0042] Specifically, the robot matches and analyzes the target state information with the actual state information to determine whether it has the capability to perform the sub-task. If the match is successful, the robot performs collaborative welding operations according to the target state information of the sub-task; if the match is unsuccessful, the robot feeds back the relevant information to the task allocation module 12 and the state space pool management module 11 so that it can complete the matching with other robots.

[0043] The following will combine Figure 1 The content shown summarizes the working process of the multi-robot collaborative welding system 100.

[0044] During the welding task performed by the robot, the state space pool management module 11 continuously updates the actual state information of the robot 13 to the local state space. If the actual welding parameters of the robot 13 deviate from the target state information of the sub-task, the local state space will record the deviation in real time and feed the information back to the task allocation module 12.

[0045] Based on feedback information and welding task information in the global state space, the task allocation module 12 allocates welding tasks or adjusts the target state information of sub-tasks to ensure the smooth completion of the entire welding operation. For example, when a robot 13 malfunctions or is overloaded, the task allocation module 12 reassigns the sub-tasks undertaken by that robot 13 to other robots 13 with corresponding capabilities.

[0046] like Figure 2 As shown, one embodiment of the present invention provides a method for multi-robot collaborative welding operations.

[0047] The multi-robot collaborative welding operation method is applied to a multi-robot collaborative welding operation system.

[0048] In one embodiment, the multi-robot collaborative welding system can be as follows: Figure 1 The configuration described above is referenced in the generation method provided by this invention. However, the multi-robot collaborative welding system used in the multi-robot collaborative welding operation method provided by this invention is not limited to this configuration structure.

[0049] like Figure 2 As shown, a multi-robot collaborative welding operation method provided in one embodiment of the present invention includes the following steps.

[0050] Step S1: Obtain the state space pool, which includes a global state space for storing welding task information and attribute information of the workpiece to be processed, and a local state space for storing the actual state information of the robot. Step S2: Based on the welding task information and the attribute information of the workpiece to be processed, the welding task is decomposed into several sub-tasks, and the target state information corresponding to each sub-task is determined. Step S3: Based on the matching relationship between the target state information of the sub-task and the actual state information of the robot, several sub-tasks are assigned to several robots, and the robots are controlled to perform corresponding welding operations according to the assigned sub-tasks.

[0051] In this way, by utilizing the collaborative structure of the global state space (which stores the total task and workpiece information) and the local state space (which stores the robot's real-time state), the welding task planning and execution process are decoupled, and dynamic adaptation between the two is achieved. At the same time, based on the multi-dimensional matching relationship between the sub-task target state and the robot's actual state, each robot can accurately undertake sub-tasks that meet its capability boundaries according to its own real-time working conditions, avoiding welding defects caused by resource conflicts or insufficient capabilities. This enables efficient dynamic allocation and precise execution of welding tasks by multiple robots, improving work quality and efficiency.

[0052] In step S1, the state space pool is a data storage and management structure used to centrally store and manage various information related to the welding operation. This information covers the task level, workpiece level, and robot level, providing data support for the planning, execution, and monitoring of the entire welding operation. It also facilitates robots in easily acquiring and sharing the necessary information, enabling collaborative work among multiple robots.

[0053] In one embodiment, the state space pool adopts a distributed database architecture, consisting of a global state space and a local state space, and achieves data synchronization through a communication protocol.

[0054] In this way, the state space pool of this distributed database architecture achieves data synchronization through communication protocols, which not only ensures the consistency of global and local state space data, but also enables the robots managed by the local state space to respond to tasks autonomously and flexibly based on real-time synchronized data. This effectively avoids the computational bottleneck caused by centralized data processing in traditional centralized control methods, and also avoids data inconsistency caused by untimely data synchronization and imperfect coordination mechanisms in traditional distributed control methods.

[0055] In one embodiment, an indexing mechanism is used to perform structured storage of the stored data in the state space pool.

[0056] Specifically, an indexing function is determined based on the stored data type, data source, and timestamp. This indexing function is used to classify and label the data in the state space pool. For example, if the data type is defined as type, the data source as source, the timestamp as timestamp, and the data information as data, the indexing function I(data, type, source, timestamp) = {index1, index2, ...}, where index1 and index2 are the index values ​​generated by the indexing function I after processing the relevant input data parameters.

[0057] In one embodiment, a cache layer is established for frequently accessed data within the state space pool.

[0058] In this way, by reading high-frequency data from the cache, latency can be reduced, access pressure on the underlying state space pool can be reduced, and throughput can be improved (the cache layer can respond to requests in parallel).

[0059] Specifically, a cache hit function HR(data) is defined, which is used to quantify the effect of setting the cache layer. The cache name function HR(data) = access_success(data) / access_total(data), where access_success(data) is the number of successful accesses and access_total(data) is the total number of accesses. This function reflects the probability of directly retrieving data from the cache.

[0060] Furthermore, when the robot's actual state information changes, this information is updated in the state space pool, which can include real-time updates and periodic updates. For example, for critical data (such as the robot's current position information)... If it can be updated in real time, then the updated location information ,in It is the change in position. This is the location information before the update.

[0061] For example, secondary data (such as ambient temperature) can be updated periodically, with an update cycle of [time period]. ,when hour, .

[0062] In step S1, the welding task information is a comprehensive description and requirement of the entire welding project, used to clarify the goals and planning of the welding operation. Specifically, it includes at least the overall requirements and overall planning of the entire welding project, such as the overall welding objectives (e.g., completing the welding assembly of a large workpiece), time constraints (e.g., requiring the welding task to be completed within a specified time), and quality standards (e.g., weld strength, sealing performance), etc.

[0063] The attribute information of the workpiece to be processed is a description of the various properties and characteristics of the workpiece itself, including at least the workpiece's geometry, size, material type, physical properties (such as density, hardness, etc.) and chemical properties (such as corrosion resistance, etc.).

[0064] The global state space is a multi-dimensional data structure used to store the 3D model of the workpiece to be processed (such as a CAD model), key area markers (such as load-bearing part markings), process parameters (such as welding current and voltage), working environment data, and a map of the welding area. It stores common data related to the entire welding operation scenario and task. A welding area map is a technological abstraction of the workpiece to be processed. In other words, the physical boundaries of the welding area map are entirely determined by the geometry of the workpiece. By analyzing the structural features (such as lap gaps and bevel angles) in the 3D model of the workpiece and combining them with the parameters defined in the welding process specification, the physical workpiece is transformed into a digital welding area map with technological attributes.

[0065] The local state space is a storage area that stores the actual state information of each robot. This information reflects the robot's working state and capabilities at the current moment, and has real-time and individual characteristics.

[0066] In one embodiment, the local state space is used to store at least one of the robot's current pose information (including position information and attitude information) in the workspace, the local area information of the workpiece it is responsible for processing, and the surrounding local environment information.

[0067] In this way, the welding position can be accurately located by using the local area information of the workpiece, the welding posture of the robot can be adjusted and the appropriate welding process can be selected. It can also perform personalized task processing for different local area characteristics, improving the welding flexibility and adaptability. At the same time, based on the surrounding local environment information, it can achieve better collaboration by sharing the working status and task information of other robots in the vicinity, thereby improving the efficiency and coordination of the entire welding system.

[0068] In this embodiment, the workpiece local area information includes the geometric features (three-dimensional geometry, size, surface curvature, etc.) and material properties of the local area of ​​the workpiece to be welded by the robot. The surrounding local environment information includes at least one of the following: obstacle information stored around the current workspace (such as workbench, other moving robots, etc.), spatial layout information (such as space size, passage width, etc.), and environmental condition information (such as humidity, temperature, etc.).

[0069] For ease of understanding, the global state space is defined as GSS, and the information stored in the global state space GSS is defined as follows: ,in, Map of the welding area. For the 3D model of the workpiece, This is information about the welding task.

[0070] Define the local state space as LSS, and in the local state space LSS, the first... The actual state information of a robot is defined as follows: .in, It is the first The pose state of the robot (including position information) Attitude information wait), It is the first Information on the local area of ​​the workpiece handled by each robot. It is the first Information about the local environment surrounding the robot.

[0071] In one specific embodiment, the attribute information of the workpiece to be processed and the actual state information of the robot are detected and determined based on a sensor array.

[0072] Specifically, each robot is equipped with a sensor array, which includes at least one of a vision sensor, a force sensor, a displacement sensor, a temperature sensor, and an environmental sensor. The vision sensor captures images of the workpiece, and image processing algorithms (such as edge detection and contour extraction) are used to obtain the workpiece's outer contour information.

[0073] The outer contour information includes shape, location, and weld information; the data from its detection is recorded as follows: ,in, Indicates different times, This represents the pixels or feature points in the image captured by the vision sensor at different times. The observation data can reflect the dynamic changes in the position and shape of the workpiece.

[0074] Force sensors monitor welding forces, and the data they detect is recorded as force data. , The sampling sequence number represents the sampling time of the force, and the force value corresponding to each sampling time. This reflects the magnitude and direction of the force applied to the workpiece by the robot during the welding process. Appropriate welding force ensures good contact between weld beads and improves weld strength.

[0075] Displacement sensors record the robot's joint displacement information, and the data they detect is recorded as displacement data. These are displacement sampling points, representing different sampling times, and the displacement value corresponding to each sampling time. It reflects the motion state of each joint of the robot. By recording the joint displacements, the robot's motion trajectory and posture can be precisely controlled, thereby achieving accurate welding operations.

[0076] Temperature sensors detect the temperature of the welding area and record the detected data as follows: m represents the temperature sampling time, and the temperature value corresponding to each sampling time. This reflects the temperature changes in the welding area during the welding process. Different welding methods and materials require appropriate welding temperature ranges. For example, in laser welding, excessively high temperatures may cause the material to burn through, while excessively low temperatures will not allow the material to melt sufficiently, affecting the weld strength.

[0077] Environmental sensors include light sensors and humidity sensors, which record the detected data as follows: and n and p represent the temperature sampling time and humidity sampling time, respectively, and the light intensity value corresponding to each sampling time. This reflects the lighting conditions and corresponding humidity levels within the welding workshop. It reflects the air humidity in the welding workshop.

[0078] In one embodiment, the data detected by the sensor is preprocessed, including at least one of data cleaning, data format conversion, and data compression. Data cleaning refers to removing obvious outliers, such as force data. When the force exceeds the normal range, set the force data after cleaning. .

[0079] Data format conversion refers to the process of reorganizing and representing raw data collected by different types of sensors according to a unified standard or specification. For example, converting image data captured by a vision sensor into feature vector form, or converting displacement data recorded by a displacement sensor into a standard displacement matrix.

[0080] Data compression refers to the process of processing data using specific algorithms to reduce the storage space or transmission bandwidth occupied by the data. In robotic welding, a large amount of sensor data needs to be transmitted to the state space pool for processing and analysis in real time. Excessive data volume will lead to increased transmission latency, affecting the real-time performance and overall system performance. By employing data compression, the amount of data can be reduced, data transmission efficiency can be improved, and communication costs can be lowered.

[0081] In one embodiment, a Kalman filter method is used to fuse first data detected by a vision sensor and second data detected by a force sensor to determine the current state of the robot or weld at the current moment.

[0082] Specifically, an iterative loop operation is performed using a state prediction equation and a measurement update equation (i.e., "prediction-update") to obtain the optimal state information. The state prediction equation is defined as follows: The measurement update equation is defined as follows: ,in, , The coefficient matrix, For control inputs (such as robot joint torque, motion commands, etc.). For measured values ​​(such as vision or force sensor data). For Kalman gain. The posterior state estimate of k-1 at the previous time step is the optimal state estimate obtained after the previous round of "prediction-update". It combines the predicted and measured values ​​at the previous time step and is the output of this Kalman filter and the input of the next time step. The prior state estimate for the current time k is a preliminary prediction obtained solely through the state prediction equation (without considering new measurements).

[0083] Of course, other types of data collected by various sensors can also be fused using the Kalman filter method, which will not be elaborated here.

[0084] In one embodiment, when multiple robots work collaboratively, a collaborative decision-making model is constructed based on reinforcement learning, and the optimal decision information of the robots is determined according to the actual state information of the robots and the attribute information of the workpiece to be processed.

[0085] Specifically, based on several structural features of the workpiece to be processed, the state space S is determined; based on the robot's actual state information, the corresponding motion space A (such as welding speed adjustment, path planning adjustment, etc.) is determined; and a reward function is constructed based on the task completion quality Q, resource consumption E, task completion time T, and robot collaboration efficiency C. ,in, These are weighting coefficients. The robot determines the weighting coefficients based on the current state. Select Action By interacting with the environment, the optimal policy is learned, thus enabling the cumulative reward function to... maximum, This is the discount factor. (Constructing a collaborative decision-making model) The robot acquires its own and its environment's state information, performs information fusion and feature extraction to obtain a state s that conforms to the state space S, and inputs it into the constructed collaborative decision-making model. The model then bases its decisions on the strategy... Choose action 'a', specifically based on the action probability. ( (Using temperature parameters) Select action a from action space A; after the robot executes the selected action a, the environment provides feedback on the new state s′ and the reward r calculated based on the reward function, thus consolidating this interaction experience. Store the experience in the experience replay buffer D; and randomly sample a batch of interactive experiences from buffer D. Based on the estimated value of Q(s,a) and the target value (Target network parameters) The error is used to update the online network parameters θ, and the loss function is... And use gradient descent method (α is the learning rate) Parameters are updated, and the online network parameters θ are periodically updated to the target network. This allows for continuous optimization of the decision-making model to achieve better collaborative decision-making. (Collaborative Decision-Making Model Training) like Figure 3 As shown, in one embodiment, step S2, which involves decomposing the welding task into several sub-tasks based on the welding task information and the attribute information of the workpiece to be processed, includes the following steps.

[0086] Step S21: Analyze the welding task information and determine several constraints, including at least one of quality constraints, production capacity constraints, and process constraints; Step S22: Extract and determine several structural features of the workpiece to be processed based on its attribute information. Step S23: Based on the matching relationship between several structural features and several constraints, the welding task is divided into several sub-tasks.

[0087] In this way, by analyzing constraints such as quality, capacity, and process, and combining them with the structural characteristics of the workpiece, the welding task is divided into sub-tasks that meet the actual production requirements, thus avoiding rework or defects caused by neglecting constraints in traditional methods.

[0088] In step S21, quality constraints are used to determine the quality standards that the product must meet in the welding task, including material properties, structural strength, surface quality, etc. For example, the weld joint must reach a certain strength to avoid cracking, or the surface roughness of the workpiece must meet assembly requirements.

[0089] Capacity constraints refer to the limitations imposed on production capacity by resources (equipment, manpower, time, etc.) during the production process. Examples include the maximum processing capacity of a robot per unit of time, such as welding 20 workpieces per hour, and the allocation of time.

[0090] Process constraints are used to determine limitations of manufacturing processes, including material handling methods (such as the sequence and parameters of welding and cutting), robot capabilities (such as machining accuracy and production efficiency), etc. For example, aluminum alloys require specific welding processes to avoid hot cracking.

[0091] In step S22, the structural features include at least one of the following: the geometry of the workpiece to be processed, its components, and the connection methods and relative positional relationships between the components. For example, whether the workpiece has grooves, protrusions, etc.; whether it is welded from multiple simple parts or formed as a single piece, etc.

[0092] For example, the weld length can be extracted from the attribute information of the workpiece to be processed. ,in, These are the coordinates of the points on the weld.

[0093] In step S23, "matching relationship" refers to the inherent correlation between the structural features of the workpiece to be processed and the pre-determined quality, capacity, and process constraints in the welding task. By comprehensively analyzing the degree of influence and interaction of structural features on each constraint, it is identified which structural features require specific quality standards, which structural features need adjustment due to capacity limitations, and which structural features depend on specific process methods. Based on this multi-dimensional correlation, the welding task is then decomposed into several sub-tasks, ensuring that each sub-task has independent execution feasibility and overall task synergistic optimization within the matching framework of structural features and constraints.

[0094] like Figure 4 As shown, in one specific embodiment, step S23 specifically includes the following steps.

[0095] Step S231: Based on several structural features, the workpiece to be processed is divided into several areas to be welded; Step S232: Merge several regions to be welded that satisfy at least one of the quality constraints, production capacity constraints, and process constraints to obtain several sub-regions; Step S233: Generate several sub-tasks based on several sub-regions and their corresponding structural features and constraints.

[0096] In this way, by merging the areas to be welded that meet the constraints, the frequent switching of tasks or paths by the robot due to area fragmentation is avoided, thereby improving the efficiency of continuous operation.

[0097] In this embodiment, the physical structure of the workpiece to be processed is transformed into weldable logical units, and the welding task is "spatially decoupled" through the analysis of structural features. Specifically, the division can be based on at least one of geometric features, connection features, and material features.

[0098] It should be noted that in this embodiment, "sub-region" is a logical unit formed by abstracting the structural features and constraints of the physical workpiece, rather than a physical unit. Its relationship with the physical region may be "one-to-many" (i.e., one sub-region corresponds to multiple physically discontinuous areas to be welded).

[0099] For ease of understanding, for example, the workpiece to be processed is divided into region A (butt joint, steel material), region B (T-joint, steel material), and region C (butt joint, aluminum alloy material) based on structural characteristics. Regions A and B are merged into sub-region 1 because they are made of the same material but have different joint types, but are compatible in terms of process constraints. Region C is divided into sub-region 2 because it is made of different materials.

[0100] If several sub-regions that meet the constraints are physically adjacent, they can be merged into one sub-region and the same welding parameters (such as current, voltage, etc.) can be used uniformly. In this case, the merged sub-regions are physically connected.

[0101] In one specific embodiment, step S232 specifically includes the following steps.

[0102] Step S2321: Determine the first sub-region and the second sub-region that satisfy the constraints; Step S2322: Determine whether the distance between the first sub-region and the second sub-region is less than or equal to a first preset threshold. If so, proceed to step S2323, merge the first sub-region and the second sub-region, and update the determined sub-region; Based on this, step S233 includes the following steps.

[0103] Step S233': Based on the updated sub-region, and the corresponding structural features and constraints, update and determine several sub-tasks.

[0104] In this way, by merging adjacent sub-regions by setting a distance threshold, the number of sub-tasks is reduced, thereby reducing communication overhead and allocation complexity.

[0105] In one embodiment, the target state information includes target parameter configuration and target requirement information, and the actual state parameters include actual parameter configuration and actual performance parameters.

[0106] In this embodiment, the target state information and the actual state parameters are a decoupled design of task requirements and equipment capabilities. The former focuses on the technical requirements of the sub-task, while the latter reflects the robot's actual execution capability. The two achieve local adaptation of tasks and resources through dynamic matching.

[0107] like Figure 5 As shown, in one embodiment, step S3 may specifically include the following steps.

[0108] Step S31: Based on the matching relationship between the target parameter configuration corresponding to the sub-task and the actual parameter configuration of the robot, generate the initial task allocation result; Step S32: Based on the matching relationship between the target requirement information of the sub-task and the actual performance parameters of the robot, the initial task allocation result is adjusted, and the robot is controlled to perform the corresponding welding operation according to the adjusted allocation result.

[0109] In this way, by breaking down the target state information into target parameter configuration and target requirement information, and matching them with the robot's actual parameter configuration and actual performance parameters respectively, it is ensured that the task allocation meets both technical requirements (parameter matching) and operating conditions (performance matching), thereby achieving precise matching between task allocation and robot state, avoiding welding failures caused by inconsistent states, and improving the stability and reliability of welding operations.

[0110] In step S31, the target parameter configuration defines the set of process parameters that the sub-task should follow. In other words, it defines the technical baseline of "what process parameters must be used to complete the sub-task". For example, for the sub-task of butt welding of thick deck plates of ships, the target parameter configuration may include: welding method (submerged arc welding), current (600-650A), voltage (32-34V), welding speed (45-55cm / min), welding diameter (4.0mm), etc.

[0111] Target requirements information is used to define the process quality requirements of sub-tasks, guiding task allocation and resource scheduling. It defines "what effect this sub-task needs to achieve," including dimensions such as quality level, capacity requirements, and cost constraints.

[0112] Actual parameter configuration reflects the actual process parameters used by the robot when performing a sub-task, revealing the degree of matching between the robot's current process capabilities and the target parameter configuration. It reflects the robot's equipment capability of "with what parameters can it complete the task". For example, for the sub-task of butt welding of thick deck plates on ships, the actual parameter configuration may include: welding method (submerged arc welding), current (620A), voltage (33V), welding speed (458cm / min), welding diameter (4.0mm), etc.

[0113] Actual performance parameters are the maximum capabilities of a robot. They define the robot's limit operation capabilities under specific working conditions, including but not limited to load capacity, capability factor, and motion accuracy. They provide a quantitative basis for the matching degree of "equipment capability - task requirements" for task allocation, avoiding equipment overload or inefficiency due to over-allocation, and achieving efficient operation of welding tasks.

[0114] In the above embodiments, the initial task allocation result that meets the basic execution conditions is quickly generated by first matching the target configuration parameters of the sub-task with the actual configuration parameters of the robot, eliminating physically inaccessible allocation combinations, and avoiding execution failure due to parameter incompatibility. Then, the initial task allocation result is globally adjusted by combining the target requirement information of the sub-task with the real-time performance parameters of the robot to achieve deep adaptation. This method adopts a hierarchical and progressive matching mechanism to achieve accurate matching of "global optimal - local adaptation", which can not only ensure the basic feasibility of the allocation scheme, but also give full play to the advantages of each robot and improve the overall work efficiency.

[0115] As shown in Figure 6(a), in one specific embodiment, step S31 includes the following steps.

[0116] Step S3111: Determine the corresponding first position information and first process parameter information according to the target parameter configuration; Step S3112: Determine the corresponding second position information and second process parameter information based on the actual parameter configuration; Step S3113: According to the order of parameter information and position information, determine the first subtask and the first robot that match the target parameter configuration and the actual parameter configuration, and assign the first subtask to the first robot.

[0117] In this way, by matching position information and process parameter information (such as current and speed), sub-tasks and robots with suitable parameters and positions are prioritized for allocation, thereby improving the accuracy and efficiency of initial task allocation and reducing subsequent adjustment costs.

[0118] In step S3111, the position information refers to the spatial coordinate attributes of the welding operation. This can be the absolute position coordinates in the workpiece coordinate system or the relative position relative to a reference point; there are no specific restrictions. The process parameters define the technical specifications of the welding process.

[0119] The first position information refers to the target welding position required by the subtask, such as the start position and the end position. It defines "where the subtask needs to complete the welding" and belongs to the target attribute. The second position information refers to the actual position range that the robot can reach. It defines "where the robot can complete the welding" and belongs to the actual attribute.

[0120] The first process parameter refers to the welding process conditions of the sub-task, which defines "how the sub-task needs to be welded" and belongs to the target attribute; the second process parameter refers to the actual welding process capability supported by the robot, which defines "how the robot can weld" and belongs to the actual attribute.

[0121] As shown in Figure 6(b), in another specific embodiment, step S31 includes the following steps.

[0122] Step S3121: Obtain the first subtask and the second subtask, and determine the corresponding third and fourth position information; Step S3122: When the distance between the third location information and the fourth location information is less than the second preset threshold, the first subtask and the second subtask are assigned to the same robot, or the first subtask is assigned to the robot closer to the third location information and / or the second subtask is assigned to the robot closer to the fourth location information.

[0123] In this way, by prioritizing the assignment of two subtasks to the same robot or the nearest robot when their locations are close together, the risk of collision or path redundancy caused by frequent cross-movement of multiple robots can be avoided.

[0124] In step S3121, the third position information and the fourth position information are the target welding positions defined in the first subtask and the second subtask, respectively. When the distance between the two is less than the second preset threshold, it is determined that the first subtask and the second subtask have spatial clusterability, and the task allocation is adjusted.

[0125] In one specific embodiment, the priority of the corresponding subtask is determined based on the target parameter configuration.

[0126] In this embodiment, the priority of a subtask is dynamically generated by mapping the workpiece structural features (such as load-bearing / non-load-bearing parts) and process reliability requirements (such as welding strength) in real time in the global state space, resulting in a quantifiable priority score.

[0127] To make it easier to understand, let's take an example and assume that the welding task is divided into several sub-tasks. Subtask There is a starting position End position Process parameter information (such as welding current) ,Voltage (etc.) and task priority , can be defined as , Each subtask Matches with global process standards recorded in GSS (such as welding parameter thresholds for different materials).

[0128] As shown in Figure 7(a), in one specific embodiment, step S32 may specifically include the following steps.

[0129] Step S3211: Based on the initial task allocation results, determine the number of sub-tasks currently assigned to each robot; Step S3212: Based on the priority of each subtask and the actual parameter configuration of the robot, determine the contribution value of the robot, and sum the contribution values ​​of several robots in the initial task allocation result to obtain the corresponding total contribution value. Step S3213: When the total contribution value is less than or equal to the third preset threshold, the initial task allocation result is adjusted.

[0130] In this way, by calculating the total contribution value and setting a threshold, task redistribution can be triggered, which helps to dynamically balance the robot's load, avoid efficiency bottlenecks caused by uneven task distribution, and ensure overall production capacity balance.

[0131] In step S3212, the contribution value is a comprehensive evaluation index that quantifies the ability of a single robot to complete its assigned sub-tasks under the current task allocation result. The total contribution value is the weighted sum of the contribution values ​​of all robots, reflecting the balance and effectiveness of the overall task allocation.

[0132] When the total contribution value fails to reach the third preset threshold, it indicates that the current task allocation result has problems such as a mismatch between robot capabilities and task requirements, an unreasonable priority execution order, or an unbalanced load distribution. In this case, it is necessary to optimize the combination of capability factors and current load factors by adjusting task allocation (such as reallocating sub-tasks or switching execution order) to increase the total contribution value and approach the threshold, thereby achieving optimal or adaptive optimization of the overall goal.

[0133] In one specific embodiment, the step S3212, which determines the robot's contribution value based on the priority of each subtask and the robot's actual state information, includes the following steps.

[0134] Step S32121: Based on the robot's actual performance parameters, determine at least one of the corresponding capability factor and the current load factor; Step S32122: Determine the contribution value of the robot based on the priority of each subtask and the product of the corresponding robot's capability factor and current load factor.

[0135] In this way, by quantifying the robot's comprehensive ability to perform sub-tasks, the rationality of contribution value calculation can be improved.

[0136] In step S32121, the current load factor is used as a dynamic indicator to quantify the degree of robot occupancy at a certain moment. It is calculated by comparing the amount of tasks currently undertaken by the robot (such as the number of tasks being executed, task duration, computing resource consumption, energy consumption, etc.) with its maximum carrying capacity (such as rated working time, theoretical maximum load, etc.). These two factors together constitute a binary representation of the robot's actual performance parameters, providing a dynamic adjustment basis for the calculation of the contribution value.

[0137] For ease of understanding, the subtask priority is defined as follows: The robot's capability factor is Robot's current load factor The contribution value of the robot can be calculated using formula (1).

[0138]

[0139] In one specific embodiment, a genetic algorithm is used to adjust the initial task allocation results based on a state space pool.

[0140] Specifically, based on the initial task allocation results, the allocation relationship between each robot and its subtasks is configured as a chromosome. In other words, each robot corresponds to a gene locus in the chromosome, and the gene value represents the subtask assigned to that robot. All genes arranged in the order of the robots constitute a complete chromosome. Let's define the robots as follows: ,chromosome ,Gene Indicates task Assigned to robots The fitness function is defined in formula (1). .

[0141] Each chromosome is evaluated using a fitness function. Based on the evaluation results, high-scoring chromosomes (i.e., high-fitness allocation schemes) are selected and retained, while low-scoring chromosomes are filtered out. The selected results are then used to generate a new generation of chromosomes through recombination and mutation operations. This process is repeated iteratively until the fitness function converges or the number of iterations meets the preset value, thus obtaining the final task allocation result.

[0142] Furthermore, in multi-robot collaborative scenarios, the optimal robot can be selected based on capability factors. If the actual state of the optimal robot changes, its capability factors are updated in real time. When the optimal robot malfunctions, a robot with the next best capability factor is selected as a replacement to maintain the continuity of the operation.

[0143] As shown in Figure 7(b), in another specific embodiment, step S32 includes the following steps.

[0144] Step S3221: Based on the initial task allocation results, determine the robot's capability factor and the priority of the sub-tasks; Step S3222: Determine whether the subtask with the highest priority is assigned to the robot with the highest capability factor; If not, proceed to step S3223. When the parameter information of the subtask with the highest priority matches the parameter configuration of the robot with the highest capability factor, assign the subtask with the highest priority to the robot with the highest capability factor.

[0145] In this way, by using performance scoring and priority matching, the highest priority subtasks are assigned to the best-performing robot, ensuring that high-priority subtasks are executed by the best robot and improving the reliability of critical welding processes.

[0146] In one specific embodiment, a capability factor is determined based on the robot's historical state information (such as historical welding accuracy) and current actual state information. For example, the capability factor can be determined based on a weighted combination of historical state scores and current state scores. like Figure 8 As shown, in one embodiment, the method further includes the following steps.

[0147] Step M11: Determine whether the timestamp and version information of the robot's local state information are consistent with the corresponding information in the state space pool; If not, proceed to step M12 and update the corresponding information in the state space pool based on the robot's local state information.

[0148] In this way, by verifying timestamps and version information, the robot updates its local state, ensuring consistency with the information in the state space pool and avoiding collaborative errors caused by data asynchrony.

[0149] In step M11, the robot can read state information from the state space pool in two ways: active reading and passive reading.

[0150] Specifically, on the one hand, the robot reads the state space pool at fixed time intervals to ensure that the local state is synchronized with the global state, such as updating the welding area map and / or task list, and monitoring resource status.

[0151] On the other hand, when a local state anomaly is detected and / or a state update notification is received, the robot reads the state space pool to obtain the latest information.

[0152] In step M12, the robot's local state information refers to the actual state information maintained by the robot itself (such as position, completion progress, fault alarms, etc.). It can use sensors to monitor its state or environmental changes in real time and update the detected data to the corresponding position in the state space pool in real time.

[0153] After the task allocation is completed, the multi-robot collaborative operation enters the resource coordination stage. This stage is based on the state space pool to realize real-time interaction between the global state space and the local state space, so as to avoid resource conflicts and achieve the goal of efficient resource utilization.

[0154] Taking welding resources as an example, through the equipment ,energy And manage and coordinate workspaces (WS), etc. Equipment allocation is handled through an equipment allocation matrix. To reflect, when Time indicates device Assigned to the robot ,on the contrary .

[0155] Resource allocation strategies are closely linked to task progress and resource usage. Taking energy allocation as an example, the energy allocated to machine i at time t... ,in, For total energy, Let be the task load of robot i at time t.

[0156] When a robot completes a task or malfunctions, it immediately releases the resources it occupied and reallocates these released resources. During resource coordination, if the robot's workload changes in the Local State Space (LSS) or the resource usage state is updated, this information is uploaded to the Global State Space (GSS), triggering the resource reallocation process. If the resource state in the GSS changes, such as a device... In the event of a malfunction, notifications will be immediately sent to the affected robots (i.e., the device resource list in the LSS). Include The robot sends a resource release command.

[0157] In addition, when two robots have conflicting LSS display resource requirements, such as when they apply for equipment at the same time. At this point, the task priority and robot capability factor in GSS will be called for comprehensive evaluation. Priority will be given to high-priority tasks or robots with better capabilities. At the same time, the other robot will be notified through the state space pool to adjust resource requirements, such as switching to backup equipment, so as to ensure the smooth coordination of resources and the efficient operation of the entire collaborative operation system.

[0158] The various embodiments, examples, or specific examples provided by this invention can be combined with each other to ultimately form multiple better embodiments.

[0159] Figures 9(a), 9(b), and 9(c) illustrate a flowchart of a multi-robot collaborative welding operation method in a preferred embodiment. The process will be summarized below with reference to Figures 9(a), 9(b), and 9(c).

[0160] As shown in Figure 9(a), multiple sensors are used to detect the attribute information of the workpiece to be processed and the actual state information of the robot. The collected data is preprocessed and the preprocessed data is transmitted to the state space pool for data storage and indexing. When the attribute information and actual state information change, the data is updated and synchronized.

[0161] As shown in Figure 9(b), after the state space pool is constructed, managed, and runs stably, various types of data are acquired from the state space pool. Information fusion techniques (such as Kalman filtering) are used to organically fuse information from different sensors and data sources, eliminating redundancy and contradictions between data and improving the accuracy and completeness of information. Then, feature extraction algorithms are used to mine information features that have a key impact on collaborative decision-making from the fused data.

[0162] Based on the data and extracted features in the state space pool, artificial intelligence algorithms such as reinforcement learning and deep learning are used to construct a collaborative decision-making model that can make reasonable decisions based on state and task requirements. In the construction process, the state space, action space and reward function of the model are defined. The model is trained and optimized through a large amount of training data to obtain a well-trained collaborative decision-making model.

[0163] When welding operations begin, the model retrieves current state information from the state space pool in real time and inputs it into the trained collaborative decision-making model. Based on the strategies learned internally, the model quickly generates action instructions that each robot should execute. These instructions are sent to the corresponding robots via the communication network. The robots execute the operations according to the instructions and feed the results back to the state space pool. Simultaneously, the model is further adjusted and optimized based on the feedback results, forming a closed-loop decision-making cycle that continuously improves the accuracy and adaptability of collaborative decision-making.

[0164] As shown in Figure 9(c), while the collaborative decision-making algorithm provides action guidance for the robot, the multi-robot task allocation and resource coordination strategy begins to play its role. First, the complex overall task is decomposed according to factors such as function, work area, and time sequence, forming multiple relatively independent and easily manageable sub-tasks. Each sub-task is described in detail, clarifying its objectives, requirements, constraints, and relationships with other tasks. Based on the task description and the robot's state information, optimization algorithms (such as genetic algorithms) are used to comprehensively consider factors such as the robot's operational capabilities, current position, and task load, to rationally allocate the sub-tasks to each robot.

[0165] After task allocation, the system manages all resources required for the operation, such as equipment, energy, and workspace. A state space pool monitors resource usage and robot resource needs in real time. When resource conflicts occur, such as multiple robots requesting the same equipment simultaneously, or when resources are insufficient, the system makes a comprehensive judgment based on preset resource coordination rules, such as task priority and robot capabilities, and adjusts the resource allocation scheme accordingly. By releasing idle resources, allocating allocated resources, or adjusting the task execution order, resource conflicts are resolved, ensuring that each robot can execute tasks smoothly and achieving efficient collaborative operation of the multi-robot system.

[0166] In summary, this invention provides a method and system for multi-robot collaborative welding operations. By utilizing a collaborative structure of a global state space (storing overall task and workpiece information) and a local state space (storing real-time robot states), the welding task planning and execution processes are decoupled, and dynamic adaptation between the two is achieved. Furthermore, based on a multi-dimensional matching relationship between the sub-task target state and the robot's actual state, each robot can accurately undertake sub-tasks that conform to its capability boundaries according to its own real-time operating conditions, avoiding welding defects caused by resource conflicts or insufficient capabilities. This achieves efficient dynamic allocation and precise execution of multi-robot welding tasks, improving work quality and efficiency.

[0167] It should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This way of describing the specification is only for clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

[0168] The detailed descriptions listed above are merely specific descriptions of feasible embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. All equivalent embodiments or modifications made without departing from the spirit of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for multi-robot collaborative welding operations, characterized in that, include: A state space pool is obtained, which includes a global state space for storing welding task information and attribute information of the workpiece to be processed, and a local state space for storing the actual state information of the robot. Based on the welding task information and the attribute information of the workpiece to be processed, the welding task is decomposed into several sub-tasks, and the target state information corresponding to each sub-task is determined. Based on the matching relationship between the target state information of the sub-tasks and the actual state information of the robots, several sub-tasks are assigned to several robots, and the robots are controlled to perform corresponding welding operations according to the assigned sub-tasks.

2. The multi-robot collaborative welding operation method according to claim 1, characterized in that, Based on the welding task information and the attribute information of the workpiece to be processed, the welding task is decomposed into several sub-tasks, including: Analyze welding task information and determine several constraints, including at least one of quality constraints, production capacity constraints, and process constraints. Based on the attribute information of the workpiece to be processed, extract and determine several structural features of the workpiece to be processed; Based on the matching relationship between several structural features and several constraints, the welding task is divided into several sub-tasks.

3. The multi-robot collaborative welding operation method according to claim 2, characterized in that, Based on the matching relationship between several structural features and several constraints, the welding task is divided into several sub-tasks, including: Based on several structural features, the workpiece to be processed is divided into several areas to be welded; Merge several regions to be welded that satisfy at least one of the quality constraints, capacity constraints, and process constraints to obtain several sub-regions; Based on several sub-regions, and their corresponding structural features and constraints, several sub-tasks are generated.

4. The multi-robot collaborative welding operation method according to claim 3, characterized in that, The merging of several regions to be welded that satisfy at least one of the quality constraints, production capacity constraints, and process constraints yields several sub-regions, including: Determine the first and second sub-regions that satisfy the constraints. Determine whether the distance between the first sub-region and the second sub-region is less than or equal to a first preset threshold. If so, merge the first and second sub-regions and update the determined sub-regions; The process of generating several sub-tasks based on several sub-regions, their corresponding structural features, and constraints includes: Based on the updated sub-regions, as well as the corresponding structural features and constraints, several sub-tasks are updated and determined.

5. The multi-robot collaborative welding operation method according to claim 1, characterized in that, The target state information includes target parameter configuration and target requirement information; the actual state information includes actual parameter configuration and actual performance parameters; the matching relationship between the target state information based on sub-tasks and the robot's actual state information controls the robot to perform corresponding welding operations according to the assigned sub-tasks, including: Based on the matching relationship between the target parameter configuration corresponding to the sub-task and the actual parameter configuration of the robot, the initial task allocation result is generated; Based on the matching relationship between the target requirements information of the sub-tasks and the actual performance parameters of the robot, the initial task allocation result is adjusted, and the robot is controlled to perform the corresponding welding operation according to the adjusted allocation result.

6. The multi-robot collaborative welding operation method according to claim 5, characterized in that, The initial task allocation result is generated based on the matching relationship between the target parameter configuration corresponding to the sub-task and the actual parameter configuration of the robot, including: Based on the target parameter configuration, determine the corresponding first position information and first process parameter information; Based on the actual parameter configuration, determine the corresponding second position information and second process parameter information; Based on the order of parameter information and location information, determine the first subtask and the first robot that match the target parameter configuration with the actual parameter configuration, and assign the first subtask to the first robot.

7. The multi-robot collaborative welding operation method according to claim 5, characterized in that, The initial task allocation result is generated based on the matching relationship between the target parameter configuration corresponding to the sub-task and the actual parameter configuration of the robot, including: Obtain the first and second subtasks, and determine the corresponding third and fourth position information; When the distance between the third location information and the fourth location information is less than the second preset threshold, the first subtask and the second subtask are assigned to the same robot, or the first subtask is assigned to the robot closer to the third location information and / or the second subtask is assigned to the robot closer to the fourth location information.

8. The multi-robot collaborative welding operation method according to claim 5, characterized in that, The adjustment of the initial task allocation result based on the matching relationship between the target requirement information of the sub-task and the actual performance parameters of the robot includes: Based on the initial task allocation results, determine the number of sub-tasks currently assigned to each robot; Based on the priority of each subtask and the actual parameter configuration of the robot, the contribution value of the robot is determined, and the contribution values ​​of several robots in the initial task allocation result are summed to obtain the corresponding total contribution value. When the total contribution value is less than or equal to a third preset threshold, the initial task allocation result is adjusted.

9. The multi-robot collaborative welding operation method according to claim 8, characterized in that, The process of determining the robot's contribution value based on the priority of each subtask and the robot's actual state information includes: Based on the robot's actual performance parameters, determine at least one of the corresponding capability factor and the current load factor; The robot's contribution value is determined based on the priority of each subtask and the product of the corresponding robot's capability factor and current load factor.

10. The multi-robot collaborative welding operation method according to claim 5, characterized in that, The adjustment of the initial task allocation result based on the matching relationship between the target requirement information of the sub-task and the actual performance parameters of the robot includes: Based on the initial task allocation results, determine the robot's capability factor and the priority of the sub-tasks; Determine whether the subtask with the highest priority is assigned to the robot with the highest capability factor; If not, when the parameter information of the subtask with the highest priority matches the parameter configuration of the robot with the highest capability factor, the subtask with the highest priority will be assigned to the robot with the highest capability factor.

11. The multi-robot collaborative welding operation method according to claim 1, characterized in that, The method further includes: Determine whether the timestamp and version information of the robot's local state information are consistent with the corresponding information in the state space pool; If not, update the corresponding information in the state space pool based on the robot's local state information.

12. A multi-robot collaborative welding system, characterized in that, include: The state space pool management module is used to construct and manage the state space pool, which includes: a global state space for storing welding task information and attribute information of the workpiece to be processed, and a local state space for storing the actual state information of the robot. The task allocation module is used to decompose the welding task into several sub-tasks based on the welding task information and the attribute information of the workpiece to be processed, and to determine the target state information corresponding to each sub-task. The work robot is configured to communicate with the state space pool management module and the task allocation module, and to perform corresponding welding operations based on the target state information of the allocated sub-task and the matching relationship between it and the corresponding actual state information.

Images