Industrial production line process flow arrangement optimization method, device and computer equipment

Abstract

The application relates to an industrial production line process arrangement optimization method and device and computer equipment. The method comprises the following steps: acquiring instruction information comprising scene description and task description; performing semantic analysis on the instruction information to obtain an execution process set corresponding to the task description in a target scene, a process allocation set and a same workpiece process sequence set; drawing a corresponding disjunctive graph according to an execution process of the execution process set, an execution robot required by each execution process in the process allocation set and an execution order in the same workpiece process sequence set; performing scheduling solving on the disjunctive graph to determine a process execution order under the condition of meeting the resource allocation constraint, and obtaining a target scheduling graph; and executing each execution process corresponding to the execution order of the target scheduling graph in sequence. In this way, efficient and accurate task planning is realized, the efficiency and accuracy of task planning are effectively improved, and the strict constraints of an industrial scene are met.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, apparatus and computer equipment for optimizing the process flow of an industrial production line. Background Technology

[0002] Against the backdrop of the continuous development of intelligent manufacturing, the demand for multiple robots to collaborate in completing tasks is constantly increasing. Traditional robot writing tasks rely on manual task planning and programming. Faced with the strict process sequence constraints and equipment and robot resource occupation in industrial production lines, it is not only inefficient but also prone to errors, making it difficult to quickly respond to the dynamically changing production needs in industrial production lines.

[0003] With the development of artificial intelligence technologies such as large language models, new ideas have emerged for automatically planning robot tasks and generating execution code. However, unlike home scenarios, industrial scenarios have strict resource mutual exclusion and rigid sequential constraints between processes. Currently, scheduling plans generated in industrial scenarios have problems such as low availability or low efficiency in real production lines.

[0004] Therefore, given a manufacturing task described in natural language and an industrial production line containing multiple robots, how to plan robot tasks and generate corresponding execution code, and how to achieve the rationality, efficiency, and executability of the generated code, has become a key technical bottleneck. Summary of the Invention

[0005] Therefore, it is necessary to provide an industrial production line process scheduling optimization method, apparatus, and computer equipment that can improve the availability and efficiency of scheduling plans in real production lines to address the above-mentioned technical problems.

[0006] Firstly, this application provides a method for optimizing the process flow arrangement of an industrial production line, including:

[0007] Obtain instruction information including scene description and task description;

[0008] Semantic analysis is performed on the instruction information to deduce the set of execution procedures, the set of procedures assigned, and the set of procedures for the same workpiece required to complete the task corresponding to the task description in the scenario corresponding to the scenario description.

[0009] Based on the execution processes of the execution process set, the execution robots required for each execution process in the process allocation set, and the execution order in the process sequence set for the same workpiece, a corresponding disjunction graph is drawn; wherein, the disjunction graph is used to characterize the execution order and resource allocation constraints between different processes in the process set;

[0010] The disjunctive graph is used to solve for scheduling, and the execution order of the processes under the resource allocation constraints is determined to obtain the target scheduling graph; wherein, the target scheduling graph is used to represent the target execution order of the processes.

[0011] According to the execution order of the target scheduling diagram, the execution program corresponding to each execution step is executed sequentially.

[0012] In one embodiment, the step of performing semantic analysis on the instruction information to deduce the set of execution procedures, the set of procedure allocations, and the set of procedure sequences for the same workpiece required to complete the task corresponding to the task description in the scenario corresponding to the scenario description includes:

[0013] The scene description and task description in the instruction information are input into a large language model for semantic analysis to determine the scene configuration corresponding to the scene description.

[0014] Based on the scenario configuration, the task description is decomposed into multiple corresponding execution procedures to obtain the set of execution procedures corresponding to the instruction information;

[0015] Assign a corresponding execution robot to each execution process in the set of execution processes to determine the process assignment set;

[0016] Define the execution order of all processes that process the same workpiece in the set of execution processes to obtain the set of process sequences for the same workpiece.

[0017] In one embodiment, the step of assigning a corresponding execution robot to each execution process in the execution process set to determine the process assignment set includes:

[0018] Based on the robot list in the scenario configuration corresponding to the scenario description, an execution robot with corresponding functions is matched for each execution process in the execution process set;

[0019] Output the matching execution robot for each execution process to obtain the process allocation set.

[0020] In one embodiment, the step of defining the execution order of all processes handling the same workpiece in the set of execution processes to obtain a set of process sequences for the same workpiece includes:

[0021] The dependencies between multiple execution processes in the set of execution processes are derived to determine the dependencies between the execution processes that process the same workpiece.

[0022] Based on the dependencies between the execution processes of the same workpiece, a list of sequences in which each execution robot executes different execution processes is determined, resulting in a set of process sequences for the same workpiece.

[0023] In one embodiment, the step of drawing the corresponding disjunction graph based on the execution processes of the execution process set, the execution robots required for each execution process in the process allocation set, and the execution order in the same workpiece process sequence set includes:

[0024] Construct a directed graph structure with each execution process in the set of execution processes as a node;

[0025] Based on the execution order in the same workpiece process sequence set, a first connection line is added between two corresponding nodes in the directed graph structure; wherein, the first connection line is used to represent the dependency relationship between the two nodes;

[0026] Identify the first conflicting execution process assigned to the same execution robot in the process assignment set; wherein, the first conflicting execution process is the process in which the execution robots conflict;

[0027] In the directed graph structure, a second connection line is added between the nodes corresponding to the first conflict execution process, wherein the second connection line is used to represent the resource conflict constraint relationship of the execution robot between the two nodes;

[0028] Identify a second conflicting execution process that requires the use of the same machine from the set of execution processes; wherein the second conflicting execution process is a process in which machine conflict occurs, and the machine is a specific execution component in the execution robot;

[0029] In the directed graph structure, a third connection line is added between the nodes corresponding to the second conflict execution process; wherein, the third connection line is used to represent the resource conflict constraint relationship between the two nodes.

[0030] The output includes a directed graph structure comprising the node corresponding to each execution step in the execution step set, the first connection line, the second connection line, and the third connection line, thus obtaining the disjunctive graph.

[0031] In one embodiment, before the step of sequentially executing the execution program corresponding to each execution step according to the execution order of the target scheduling diagram, the industrial production line process flow arrangement optimization method further includes:

[0032] Obtain a pre-generated standard process flow tree; wherein, the standard process flow tree includes multiple functional nodes, each functional node represents the corresponding functional code, each functional node is arranged in the execution order, and the functional nodes in different branch structures are used to represent the differences in functional execution under different target scenarios;

[0033] Based on the scenario description, a target execution path is matched for each process in the standard process flow tree; wherein, the target execution path includes the function code from the start function node to the end function node of the execution process;

[0034] The functional code fragments associated with each functional node in the execution path are concatenated to generate the execution program corresponding to the execution procedure.

[0035] In one embodiment, prior to the step of obtaining a pre-generated standard process flow tree, the method further includes:

[0036] Obtain multiple robot execution program examples;

[0037] Analyze the commonalities in the execution functions and the differences in specific scenarios among the execution program examples;

[0038] Executable program examples that share common functionalities will be merged into common functional nodes;

[0039] The example execution program with scenario-specific differences will be generated into different branch structures;

[0040] Write corresponding executable code snippets for each functional node in the branch structure to generate a standard process flow tree that includes branch structures based on scenario differences and common functional nodes with common functionalities.

[0041] Secondly, this application also provides an industrial production line process flow arrangement and optimization device, comprising:

[0042] The acquisition module is used to acquire instruction information, including scene description and task description;

[0043] The semantic analysis module is used to perform semantic analysis on the instruction information to deduce the set of execution procedures, the set of procedures allocation, and the set of procedures sequence for the same workpiece required to complete the task corresponding to the task description in the scenario corresponding to the scenario description.

[0044] The drawing module is used to draw a corresponding disjunction graph based on the execution processes of the execution process set, the execution robots required for each execution process in the process allocation set, and the execution order in the process sequence set of the same workpiece; wherein, the disjunction graph is used to characterize the execution order and resource allocation constraints between different processes in the process set;

[0045] The solution module is used to perform scheduling solution on the disjunctive graph, determine the execution order of the processes under the resource allocation constraints, and obtain the target scheduling graph; wherein, the target scheduling graph is used to represent the target execution order of the processes.

[0046] The execution module is used to execute the execution program corresponding to each execution step in sequence according to the execution order of the target scheduling diagram.

[0047] In one embodiment, the industrial production line process scheduling and optimization device further includes an execution program generation module, which comprises a first acquisition unit, a third matching unit, and a splicing unit, wherein:

[0048] The first acquisition unit is used to acquire a pre-generated standard process flow tree; wherein, the standard process flow tree includes multiple functional nodes, each functional node represents a corresponding functional code, each functional node is arranged in the execution order, and the functional nodes in different branch structures are used to represent the differences in functional execution under different target scenarios;

[0049] The third matching unit is used to match the target execution path in the standard process flow tree for each execution procedure according to the scenario description; wherein, the target execution path includes the function code from the start function node to the end function node of the execution procedure;

[0050] The splicing unit is used to splice together the functional code fragments associated with each functional node in the execution path to generate the execution program corresponding to the execution procedure.

[0051] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the industrial production line process flow arrangement and optimization method of the first aspect.

[0052] The aforementioned industrial production line process flow scheduling optimization method, apparatus, and computer equipment acquire instruction information including scenario descriptions and task descriptions; perform semantic analysis on the instruction information to deduce the set of execution steps, the set of step allocations, and the set of steps for the same workpiece required to complete the task corresponding to the task description in the scenario corresponding to the scenario description; draw a corresponding disjunctive graph based on the execution steps of the execution step set, the execution robots required for each execution step in the step allocation set, and the execution order in the set of steps for the same workpiece; wherein the disjunctive graph is used to represent the execution order and resource allocation constraints between different steps in the step set; perform scheduling solution on the disjunctive graph to determine the execution order of the steps under the resource allocation constraints, and obtain a target scheduling graph; wherein the target scheduling graph is used to represent the target execution order of the execution steps; and execute the corresponding execution steps of each execution step in sequence according to the execution order of the target scheduling graph. Through automated semantic analysis, scheduling solution, and program generation, efficient and accurate task planning is achieved, effectively improving the efficiency, accuracy, and robustness of task planning and meeting the strict constraints of industrial scenarios. Attached Figure Description

[0053] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0054] Figure 1 A schematic diagram of an application environment for an industrial production line process flow arrangement and optimization method;

[0055] Figure 2 A flowchart illustrating the first embodiment of the method for optimizing the process flow of an industrial production line.

[0056] Figure 3 A flowchart illustrating the second embodiment of the method for optimizing the process flow of an industrial production line.

[0057] Figure 4 A flowchart illustrating the third embodiment of the method for optimizing the process flow of an industrial production line;

[0058] Figure 5 A flowchart illustrating the fourth embodiment of the method for optimizing the process flow of an industrial production line;

[0059] Figure 6 A schematic diagram of a scenario for optimizing the process flow of an industrial production line.

[0060] Figure 7 This is a schematic diagram illustrating another scenario for optimizing the process flow of an industrial production line.

[0061] Figure 8 This is a schematic diagram of another scenario of industrial production line process flow in one embodiment;

[0062] Figure 9 This is a schematic diagram of the internal structure of a computer device in one embodiment. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0064] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0065] The industrial production line process flow arrangement optimization method provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 can communicate with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or placed on a cloud or other network server. Server 104 can obtain instruction information from terminal 102, including scene descriptions and task descriptions. By performing semantic analysis on the scene and task descriptions in the instruction information, it determines the scene configuration corresponding to the scene description, as well as the set of execution procedures, the set of procedure allocations, and the set of procedure sequences for the same workpiece in the target scene corresponding to the scene description. After determining the set of execution procedures, the set of procedure allocations, and the set of procedure sequences for the same workpiece, it can also draw a corresponding disjunction graph based on the execution procedures in the set of execution procedures, the execution robots required for each execution procedure in the set of procedure allocations, and the execution order of each execution procedure in the set of procedure sequences for the same workpiece. After determining the disjunction graph, it can perform optimal scheduling solutions on the procedure sequence and workpiece constraints in the disjunction graph to generate a corresponding target scheduling graph. After determining the target scheduling diagram, the execution program of each robot is executed according to the execution order specified in the target scheduling diagram, thereby obtaining the scheduling plan generated in the industrial scenario, and performing multi-robot task scheduling on the real production line.

[0066] The terminal 102 may be, but is not limited to, various robots, personal computers, laptops, smartphones, tablets, drones, low-altitude aircraft, IoT devices, and portable wearable devices that can communicate with the server 104 to execute the planned content corresponding to the task plan.

[0067] Server 104 can be a standalone physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server that provides cloud computing services.

[0068] In traditional multi-robot collaborative task planning for existing industrial production lines, due to the strict sequential dependencies between processes and the mutual exclusion constraints on shared resources such as robots and equipment, existing technologies struggle to accurately achieve end-to-end automatic generation from natural language task descriptions to executable programs.

[0069] In terms of task planning, traditional methods rely on manual offline processing, resulting in slow response times and a high risk of errors. In industrial scenarios, methods based on large language models often generate infeasible scheduling schemes due to strong constraints.

[0070] In terms of program generation, existing methods generate control programs that lack completeness and executability, making them difficult to execute across production lines. Consequently, the rationality of task planning, the efficiency of execution scheduling, and the executability of the programs cannot be guaranteed, affecting the overall reliability and responsiveness of the production system.

[0071] If the above problems are not addressed, the generated task plans will fail to meet the stringent constraints of industrial production lines, leading to frequent scheduling conflicts and execution errors. Furthermore, this will increase the frequency of manual intervention, reduce production efficiency, and may cause equipment damage or safety risks due to program errors, making it difficult to quickly respond to dynamically changing production demands on industrial production lines.

[0072] Based on this, this embodiment proposes a first embodiment of an industrial production line process flow orchestration optimization method. It transforms high-level natural language task descriptions into low-level operation sequences that can be executed by robots, and optimizes the scheduling of multi-robot collaborative processes to achieve efficient and accurate task execution. Figure 2 As shown, this method is applied to Figure 1 Taking server 104 as an example, the explanation includes the following steps 21 to 25, wherein:

[0073] Step 21: Obtain instruction information including scene description and task description.

[0074] The instruction information can be provided in natural language, including a description of the operating environment and the specific tasks to be completed.

[0075] Here, the scene description provides contextual information about the robot's working environment, while the task description clarifies the specific task objectives that the robot needs to perform.

[0076] The instruction information can be manually entered by the operator through a text input interface. As one implementation, the operator inputs a text describing the production task. Alternatively, the instruction information can be obtained by reading a preset configuration file or task script; for example, the system loads a task description from a text file stored in the local file system.

[0077] Step 22: Perform semantic analysis on the instruction information to deduce the set of execution procedures, the set of procedure allocations, and the set of procedure sequences for the same workpiece required to complete the task corresponding to the task description in the scenario corresponding to the scenario description.

[0078] The scenario corresponding to the scenario description can be used to determine the specific configuration and status of the robot's working environment based on the scenario description in the instruction information. It may include the robot's type, number, location, and information such as the location and spatial relationship of other equipment and workpieces.

[0079] It should be noted that semantic analysis of instruction information can be a deep understanding and parsing of instruction information. It can extract structured, machine-understandable information from unstructured natural language, such as identifying target scenes, decomposing tasks into basic procedures, and determining the required resources.

[0080] Here, the set of execution procedures contains all the basic operational steps or subtasks required to complete the task description, and each execution procedure is the smallest schedulable unit of the task.

[0081] The process assignment set can be used to record the robots required for each process, clarifying which robots are assigned to specific processes, and can also be used to characterize the mutual exclusion relationships that may exist between these robots when performing processes.

[0082] A set of process sequences for the same workpiece can be used to define the order and dependencies between the execution processes that process the same workpiece, ensuring that the task proceeds according to the correct logical and physical flow. For example, some processes can only begin after other processes are completed. Here, a workpiece can refer to the object being processed by the robot when performing its processes.

[0083] As an example, instruction information including task description and scenario description can be input into a large language model, which will then analyze the instruction information to determine the configuration of the target scenario and decompose the task description into a series of discrete execution procedures.

[0084] For example, verbs such as "install," "fix," and "inspect" can be identified and associated with specific execution steps. Then, based on the list of robots available in the target scenario, one or more robots with corresponding functions are matched to each execution step, forming a step allocation set. Simultaneously, by analyzing time adverbs or logical connectors in the task description, such as "first," "then," and "next," the sequential dependencies between execution steps can be deduced, thereby constructing a set of steps for the same workpiece.

[0085] Step 23: Draw the corresponding disjunction diagram based on the execution processes of the execution process set, the execution robots required for each execution process in the process allocation set, and the execution order in the process sequence set of the same workpiece.

[0086] Disjunctive graphs can be a mathematical model used to represent task scheduling problems. In a disjunctive graph, nodes represent execution steps, and edges represent dependencies or resource constraints between steps. Disjunctive graphs can be used to clearly show the sequential constraints of execution steps and the mutual exclusion constraints of the robots.

[0087] It should be noted that the disjunction graph can be a data structure in which each execution operation is represented as an independent record or object, and the execution order defined in the same set of workpiece operation sequences is represented by internal references or indexes between each execution operation.

[0088] For example, a process record can contain references to its predecessor and successor processes. Furthermore, for conflicting execution processes identified in the process assignment set that may be executed by the same robot or machine, special mutual exclusion markers can be established between these process records to indicate that they cannot be executed simultaneously by the same robot or machine. This forms a structured representation capable of simultaneously expressing process dependencies and resource mutual exclusion, namely, a disjunction graph.

[0089] Step 24: Solve the disjunctive graph to determine the execution order of processes under the resource allocation constraints, and obtain the target scheduling graph.

[0090] The target scheduling diagram is used to represent the target execution order of the execution process. The target execution order can be the final execution order of the robot or machine involved in the target scheduling diagram when performing the corresponding execution process.

[0091] Here, the target scheduling diagram can be the output of the scheduling solution process, which can be used to characterize the specific execution order of each execution process. The target scheduling diagram can provide a clear execution order for the robot or the machine in the robot to perform the actual operation on different workpieces.

[0092] It should be noted that by analyzing and calculating the disjunction graph, the scheduling and solution of all execution operations in the disjunction graph can be performed to determine the execution operations and resource constraints contained in the disjunction graph, thereby achieving the optimal execution plan.

[0093] Understandably, the target scheduling graph obtained by scheduling based on the process sequence of the execution process in the disjunctive graph can optimize the performance indicators of the entire robot scheduling system during the invocation process while satisfying all process sequences and resource constraints. For example, it can reduce the total completion time of the robot scheduling system and improve the resource utilization of the robot scheduling system.

[0094] When scheduling and solving the process sequence and workpiece constraints in the disjunctive graph, heuristic algorithms can be used, such as priority-based scheduling strategies. In this strategy, each execution process is assigned a priority, and the start and end times of each process are determined progressively based on these priorities, dependencies between processes, and the availability of robots and machines. For example, processes without predecessor processes or whose predecessor processes have been completed can be prioritized for scheduling. When multiple processes compete for the same execution robot, selection is made according to preset rules (such as shortest processing time priority). The final output is a scheduling scheme based on the priority of each process in the execution process, i.e., the target scheduling graph. Here, the target scheduling graph explicitly represents the execution order of all execution processes.

[0095] Step 25: Execute the execution program corresponding to each execution step in sequence according to the execution order of the target scheduling diagram.

[0096] In this context, "robot" can refer to the robot body required to perform a process in an industrial production environment, while "machine" can refer to specific equipment in an industrial setting. A robot can be a physical executor that completes a process, and a machine can be the specific equipment that enables the robot to complete the process.

[0097] Here, the execution program corresponding to each execution process can be the low-level control code required for the robot to perform a specific execution process. The execution program can include the action instructions that the robot needs to execute, sensor interaction logic, and communication protocols with other devices.

[0098] It should be noted that after determining the target scheduling diagram, the execution process that needs to be executed and the robot corresponding to the execution process can be identified according to the instructions of the target scheduling diagram. The execution program required by the robot corresponding to the execution process can be generated through the large language model.

[0099] Here, each robot can sequentially call the required programs according to the execution order in the target scheduling diagram, and then start the robot to execute the program. After one process is completed, the robot can continue to execute the program code corresponding to the next process according to the target scheduling diagram, until all processes are completed.

[0100] The following example illustrates this embodiment. Suppose that on an industrial production line in scenario S, an instruction I described in natural language can also be obtained: "Polish all workpieces (processing objects) and place them on the conveyor belt."

[0101] After determining scenario S, the robots, machines, and workpieces included in scenario S can be identified. Here, a robot can be an independent robot involved in scenario S, such as robot 1, robot 2, and robot 3; a machine can be a specific execution device among different robots in scenario S, such as a conveyor belt, a grinding table, and a pallet; and a workpiece can be a processing object to be processed in scenario S, such as a workpiece.

[0102] After obtaining instruction I, task planning can be performed based on the large language model. Specifically, semantic analysis can be performed on instruction I to derive three sets corresponding to instruction I: the process set O, the workpiece manipulation sequence Q, and the process allocation A. Here, the process set O can be used to represent the multiple processes required to complete the function corresponding to instruction I, for example, it can include process 1 and process 2; the workpiece manipulation sequence Q can be used to represent the execution order of the same robot (either deleted or R1 / robot) when performing multiple processes, for example, the execution order can be O1->O2->O3; the process allocation A can be used to represent the robots assigned to multiple processes in the process set O, for example, {(process 1, robot 1); (process 2, robot 2)}.

[0103] After determining the process set O, the workpiece manipulation sequence Q, and the process assignment A, a corresponding disjunctive graph G can be drawn based on the constraints between the processes contained in the process set O, the executing machine information, the sequence information contained in the workpiece manipulation sequence Q, and the robot information contained in the process assignment A. In this disjunctive graph G, with S as the starting point and T as the ending point, each process can be represented as an independent node in S and T. For example, it can include 1 to 9 different processes. Taking process 1 to process 3 as an example, there are sequential dependencies between process 1 and process 2, and between process 2 and process 3. Simultaneously, since the executing robot needs to execute processes sequentially at different times, there are mutual exclusion constraints between these processes that need to be executed by the same robot.

[0104] Furthermore, after determining the disjunctive graph, a scheduling solution can be performed on the disjunctive graph to generate a target scheduling graph F. Here, the target scheduling graph F can be used to represent the execution order of multiple processes corresponding to those in the disjunctive graph G. In addition, the disjunctive graph G can also consider the dependencies between each process and the availability of robots. This target scheduling graph F clearly defines the execution order of each process and which robot will perform it.

[0105] After determining the target scheduling diagram F, the execution program of each robot can be executed sequentially according to the execution order of the target scheduling diagram F. Specifically, for example, the scheduling diagram F can include different operations from 1 to 9. The scheduling result can be as follows: For robot 1, it can first execute operation 1 and operation 4. After completion, robot 1 waits for operation 3 to complete. After confirming that operation 3 is completed, robot 1 continues to execute operation 7. In this way, robot 1 completes its required execution steps. For robot 2, it can wait for operation 1 to complete. After confirming that operation 1 is completed, robot 2 continues to execute operation 2 and operation 5 sequentially. After completion, robot 2 waits for operation 7 to complete. After confirming that operation 7 is completed, robot 2 continues to execute operation 8. In this way, robot 2 completes its required execution steps. For robot 3, it can wait for operation 2 to complete. After confirming that operation 2 is completed, robot 3 continues to execute operation 3 and operation 6 sequentially. After completion, robot 3 waits for operation 8 to complete. After confirming that operation 8 is completed, robot 3 continues to execute operation 9. In this way, robot 3 completes its required execution steps. In this way, the strict sequential constraints between processes and the resource mutual exclusion constraints of the execution robot can be effectively handled, thereby achieving the rationality of task planning, the efficiency of execution scheduling, and the executability of the execution program.

[0106] Based on the above examples, the industrial production line process orchestration optimization method proposed in this embodiment solves the problem of multi-robot collaborative task planning and execution. Compared with traditional methods that rely on manual offline planning and programming, it achieves end-to-end automation from natural language instructions to robot executable programs. Traditional methods are inefficient, error-prone, and difficult to respond quickly when facing complex and ever-changing production demands. This embodiment significantly improves the efficiency and accuracy of task planning through automated semantic analysis and scheduling solutions.

[0107] Furthermore, compared to existing planning methods based on large language models but primarily focused on indoor home scenarios, this method can effectively adapt to the requirements of strong constraints and high reliability in industrial fields. Existing methods often struggle to generate feasible and efficient scheduling schemes when dealing with strict process sequence constraints and mutual exclusion access to shared resources such as robots and equipment in industrial scenarios. This embodiment introduces a disjunctive graph—an explicit mathematical model—to represent process sequence and workpiece constraints, and performs specialized scheduling solutions, ensuring the feasibility and efficiency of the generated scheduling plan in a real production line.

[0108] In the above-mentioned industrial production line process flow arrangement optimization method, not only is the problem of how to achieve high-level task planning and low-level execution program automatic generation for multi-robot collaboration under the condition of a manufacturing task described in natural language and an industrial production line containing multiple robots solved, but also the rationality of task planning, the efficiency of execution scheduling and the executability of execution program are achieved under the premise of satisfying strict process sequence constraints and mutual exclusion constraints of equipment and robot resources.

[0109] In some embodiments described above in this application, semantic analysis of instruction information is proposed to derive the set of execution steps, the set of step assignments, and the set of steps sequence for the same workpiece. However, in practical applications, instruction information is usually given in natural language, and its semantic understanding and structured information extraction face challenges. Especially when it is necessary to accurately identify the target scene from the scene description and task description, parse the task into specific steps, match the required workpiece, and determine the step sequence, traditional methods may be inefficient or inaccurate, and are difficult to effectively deal with complex and ever-changing instructions.

[0110] Based on this, this embodiment further proposes a second embodiment of an industrial production line process flow arrangement optimization method. In the second embodiment of the industrial production line process flow arrangement optimization method, as follows: Figure 3 As shown, step 22 includes steps 221 to 224, wherein:

[0111] Step 221: Input the scene description and task description in the instruction information into the large language model for semantic analysis to determine the scene configuration corresponding to the scene description.

[0112] The instruction information can be input data containing scene descriptions and task descriptions, and is usually presented in the form of natural language text.

[0113] Here, the scenario description provides the environmental background information for task execution, while the task description specifies the operations or objectives that need to be completed.

[0114] It should be noted that a large language model is an artificial intelligence model trained on massive amounts of text data. Its core capability lies in understanding, generating, and processing human language. Large language models can be implemented based on the Transformer architecture.

[0115] For example, GPT or DeepSeek models can be language models fine-tuned for specific domains. Semantic analysis refers to a deep understanding of natural language text to extract the inherent meaning of instructions, identify entities, and extract relationships. This can be achieved through techniques such as named entity recognition, relation extraction, intent recognition, and referential resolution, or by directly utilizing the contextual understanding and reasoning capabilities of large language models for end-to-end semantic parsing.

[0116] The scene configuration corresponding to the scene description can be the specific environmental state or robot configuration information determined after semantic analysis, used to represent the execution process when performing the task. It can be a digital model of the physical environment, an abstract logical environment configuration, or a configuration image that exists as a topological structure diagram.

[0117] Step 222: Based on the scenario configuration, decompose the task description into multiple corresponding execution procedures to obtain the set of execution procedures corresponding to the instruction information.

[0118] After determining the scene configuration corresponding to the scene description, the specific layout and status of physical or logical resources such as robots, machines (equipment), and workpieces available in the target scene can be determined. Furthermore, the task description can be parsed into multiple corresponding execution procedures, decomposing the high-level, abstract task description into a series of more granular operation steps that can be performed by the robot.

[0119] It should be noted that the task decomposition can be performed entirely by the large language model based on its internal knowledge, or by combining a rule-based expert system with the large language model to obtain multiple execution procedures corresponding to the task description, i.e., a set of execution procedures. Here, the set of execution procedures can be the set of all subtasks obtained after the task description is decomposed.

[0120] Step 223: Assign a corresponding execution robot to each execution process in the execution process set to determine the process assignment set.

[0121] In this context, an execution robot can refer to a robot body, end effector, tool, or other auxiliary device that performs specific operations in a robot scheduling system. The process assignment set is the set of specific execution robots to which each execution process is assigned.

[0122] It should be noted that, based on the functional requirements of each execution step, suitable robots can be found and allocated from the available resources in the scenario configuration corresponding to the scenario description. Specifically, matching can be based on functional tags, such as matching a robot arm with gripping function for a grasping step; matching capability parameters, such as matching a robot with strong load-bearing capacity for a heavy object handling step; or using knowledge graphs for reasoning matching. Furthermore, factors such as robot availability, load balancing, and energy consumption can also be considered when selecting the execution robot corresponding to the execution step.

[0123] In real-world industrial scenarios, there are numerous types of execution robots with varying functions, and they may have specific operational limitations or availability constraints. Simply matching robots without fully considering the available resources, functions, and constraints of the robots in the target scenario can lead to inaccurate matching results. The selected robot may be unable to effectively execute the corresponding process, thus affecting subsequent scheduling and task execution efficiency. Therefore, this embodiment further proposes an implementation method whereby matching the execution robot required for each execution process in the execution process set to determine the process allocation set can specifically include:

[0124] (1) Based on the robot list in the scenario configuration corresponding to the scenario description, match an execution robot with the corresponding function for each execution process in the execution process set.

[0125] Here, the robot list can refer to the collection of all available, registered or identified execution robots in the target scenario. The robot list may include the function, model, current status, performance parameters and any specific operational limitations or compatibility information of each execution robot.

[0126] It should be noted that the robot list provides the actual resource basis for the matching process, ensuring that the matched execution robots are real and available. The robot list can be updated in real time by scanning execution robots in the target scene through sensors, or by loading known execution robot information in the target scene through a preset database or configuration file.

[0127] Understandably, matching one or more execution robots with corresponding functions and satisfying constraints for each execution process in the set of execution processes can mean finding one or more execution robots for each process to be executed that can complete the required functions of the process and meet all relevant constraints.

[0128] Here, the "corresponding function" of the execution process can refer to the robot's ability to perform specific operations, such as gripping, welding, assembling, and grinding; the execution robot that "meets the constraints" may include the robot's availability, load capacity, working range, accuracy requirements, time limits, environmental adaptability, and compatibility parameters with other workpieces.

[0129] Furthermore, matching execution robots for each execution process in the execution process set can be done entirely using large language models, filtering based on process requirements and execution robot attributes. Alternatively, machine learning-based methods can be employed, using pre-trained classification models to predict the most suitable execution robot for the corresponding function of the execution process.

[0130] (2) Output the matching execution robot for each execution process to obtain the process allocation set.

[0131] Here, the process assignment set clearly indicates which robot(s) will perform each process in the process set. This provides explicit input for subsequent disjunctive graph drawing and scheduling solutions, ensuring that each process has a corresponding executor. The output of the process assignment set can be a data structure, such as a list or dictionary, where each entry associates a process and its corresponding robot identifier.

[0132] In this embodiment, after semantic analysis of the instruction information and parsing the task description into a set of execution procedures, a detailed list of all available execution robots in the current target scenario can be obtained to accurately determine which robot will perform each execution procedure. This list includes not only the robot's identification information but also its functional characteristics, performance parameters, and potential operational constraints. For each execution procedure in the set, intelligent matching can be performed based on the specific function required for that procedure and the robot list in the target scenario. During the matching process, robots that meet the procedure requirements functionally and in terms of constraints such as load, accuracy, working range, and environmental adaptability can be rigorously selected. For example, if an execution procedure requires high-precision welding, a robot with welding capabilities and a qualified accuracy level can be selected from the list. If multiple robots meet the criteria, one or more can be selected based on the conditions of highest efficiency, lowest energy consumption, or lowest cost. Finally, each execution procedure is associated with its matched robot to form a clear procedure allocation set. This ensures that the robot allocation used for subsequent disjunction graph drawing and scheduling solutions is based on actual resources and constraints, avoiding scheduling failures or low execution efficiency caused by resource mismatch or constraint violations.

[0133] Therefore, based on the above-described implementation method for determining the process allocation set, when determining the allocation relationship between the execution process and the execution robot, the actual available execution robot resources, their functions, and constraints in the target scenario can be fully considered. This avoids problems such as infeasible scheduling schemes or low execution efficiency caused by blind matching or resource mismatch. Specifically, it ensures that each execution process can be allocated to one or more execution robots that truly possess the corresponding functions and meet all operational constraints, thereby improving the accuracy and reliability of task planning.

[0134] Step 224: Define the execution order of all processes that process the same workpiece in the execution process set to obtain the process sequence set for the same workpiece.

[0135] After determining the set of execution steps, the sequential relationships between the various execution steps that process the same workpiece can be determined. Specifically, this can be achieved through analysis using a large language model to identify the dependencies between execution steps.

[0136] Here, the set of process sequences for the same workpiece can be used to describe the logical execution order of all processes that handle the same workpiece.

[0137] However, in real-world industrial scenarios, when determining the order in which each robot executes different processes within a process allocation set, failure to fully consider the inherent logical connections and execution constraints between processes may result in an unreasonable process sequence, thereby affecting the accuracy of subsequent scheduling solutions and the efficiency of task execution. Therefore, this embodiment proposes an implementation method that defines the execution order of all processes handling the same workpiece within the process allocation set to obtain a set of process sequences for the same workpiece. The specific steps include:

[0138] (1) Derive multiple execution processes in the execution process set to determine the dependencies between execution processes that process the same workpiece.

[0139] Here, identifying the logical relationships between different execution processes in the set of execution processes, such as their sequence, resource usage, or data transfer, ensures that the task is executed correctly.

[0140] (2) Based on the dependency relationship between each two execution steps, determine the order list of execution steps for processing the same workpiece, and obtain the set of execution steps sequence for the same workpiece.

[0141] After clarifying the dependencies between each pair of execution steps, the execution order of the steps that process the same workpiece can be determined, resulting in a set of steps for the same workpiece.

[0142] For example, the dependencies between multiple processes required to be executed for the same workpiece can be directly output through a large language model, and further, the execution order of processing each workpiece can be determined based on the dependencies.

[0143] It should be noted that dependencies can include sequence dependencies, resource dependencies, or data dependencies, and can be used to represent the inherent constraints for the correct execution of a process. Once these key dependencies are identified, a list of one or more process execution sequences that satisfy the inherent constraints for a given task can be generated based on these dependencies. This list of sequences constitutes a set of process sequences for the same task, ensuring that when each robot executes its assigned process, the process sequence within the same task follows the correct logical order, avoiding errors or inefficiencies caused by incorrect process execution. Therefore, this embodiment, when generating a set of process sequences for the same task, fully considers the inherent logic between the multiple processes required to be executed for the same task, providing a solid and logically rigorous foundation for subsequent disjunctive graph drawing and scheduling solutions, thereby significantly improving the accuracy and reliability of task planning.

[0144] Therefore, based on the above implementation method of obtaining a set of process sequences for the same workpiece, it is possible to accurately identify the inherent dependencies between the processes executed within the same workpiece, and generate a logically correct process execution order list for each workpiece based on these dependencies. This effectively solves the problem of execution errors or low efficiency caused by improper process order in task planning, ensures the accuracy of subsequent scheduling solutions and the reliability of task execution, and enables the execution robot to complete complex tasks according to preset, logical steps, significantly improving the intelligence level and execution efficiency of automated task planning.

[0145] This application's embodiments effectively process complex natural language instruction information by introducing a large language model. The large language model performs deep semantic analysis on the scene and task descriptions in the instruction information, leveraging its powerful contextual understanding and reasoning capabilities to overcome the limitations of traditional methods in handling unstructured and ambiguous instructions. Based on the determined target scene and its resource allocation, the large language model intelligently parses the high-level task description into a series of specific, executable atomic processes, forming an execution process set. According to the functional requirements of each execution process, intelligent matching and allocation are performed among the available execution robots in the target scene, ensuring that each process can obtain a robot with the corresponding capabilities, thus obtaining a process allocation set. Furthermore, the sequential constraints of multiple processes for the same workpiece are considered, defining the precise order of the execution processes corresponding to each workpiece, generating a set of process sequences for the same workpiece.

[0146] This makes the process of converting natural language instructions into structured task planning more efficient, accurate, and automated.

[0147] In this embodiment, by leveraging the powerful semantic understanding and reasoning capabilities of a large language model, complex natural language instructions are efficiently and accurately transformed into structured task planning elements. This significantly improves the automation level of task planning and its adaptability to diverse instructions, reduces the need for manual intervention, and provides clear and explicit input for subsequent scheduling and execution, thereby enhancing the robustness and efficiency of the entire task planning process.

[0148] In some embodiments described above in this application, a technical solution is proposed to draw a disjunction graph based on the execution processes of the execution process set, the execution robots required for each execution process in the process allocation set, and the execution order in the process sequence set of the same workpiece. However, in the actual process of drawing the disjunction graph, how to effectively express the dependencies between execution processes and the resource conflicts of the same execution robot in different processes is a key step in generating the target scheduling graph. If the disjunction graph cannot fully represent the complex relationship between the process sequence and resource constraints, it may lead to inaccurate scheduling results and failure to effectively avoid resource conflicts.

[0149] Based on this, this embodiment further proposes a third embodiment of an industrial production line process flow arrangement optimization method. In this exemplary embodiment, as follows: Figure 4 As shown, step 23 includes steps 231 to 237, wherein:

[0150] Step 231: Construct a directed graph structure with each execution process in the set of execution processes as a node.

[0151] Here, each execution step is abstracted as an independent node, and these nodes constitute the basic elements of the disjunctive graph. This directed graph structure can be implemented using graph data structures in a programming language, such as adjacency matrices or adjacency lists, to store nodes and their connections. Alternatively, an object-oriented design approach can be used to encapsulate each execution step as a node object, which contains a unique identifier for the step, attribute information, and maintains a list of references to other nodes.

[0152] Step 232: Based on the execution order in the same workpiece process sequence set, add a first connection line between the corresponding two nodes in the directed graph structure.

[0153] The first connecting line is used to represent the dependency relationship between two nodes.

[0154] Here, the first connecting line can be used to clearly indicate the sequential execution order between processes, ensuring that the task is carried out according to the preset logical flow.

[0155] For example, if process A must be completed before process B, then a directed edge from the node representing process A to the node representing process B is added as the first connecting line.

[0156] Step 233: Identify the first conflicting execution process assigned to the same execution robot in the process assignment set.

[0157] The first conflict execution process can be a process in which execution robots conflict among multiple nodes in a directed structure graph, that is, there may be resource competition points for execution robots during task execution.

[0158] Specifically, the process assignment set can be traversed, and for each execution robot, all execution processes assigned to it can be collected. If an execution robot is assigned to two or more execution processes, then these processes are identified as the first conflicting execution processes.

[0159] Another approach is to construct a hash table or dictionary with the unique identifier of the executing robot as the key and the list of execution steps assigned to that robot as the value. The first conflicting execution step is determined by checking if the list of values ​​corresponding to the execution step key contains more than one execution step.

[0160] Step 234: In the directed graph structure, add a second connection line between the nodes corresponding to the conflicting execution steps.

[0161] The second connecting line is used to represent the resource conflict constraint relationship between the two nodes for the execution robot.

[0162] Here, the second connecting line can explicitly represent the mutual exclusion or sequence constraints caused by resource sharing among execution robots, thereby preventing the same execution robot from being occupied simultaneously. For the first conflicting execution process that is identified as having resource conflict constraints among execution robots, an undirected edge can be added between the nodes corresponding to the first conflicting execution process as the second connecting line, indicating that there is a mutual exclusion relationship between these processes and they cannot be executed simultaneously.

[0163] Step 235: Identify a second conflicting execution process in the set of execution processes that requires the use of the same machine.

[0164] The second conflict execution process is the process in which machine conflict occurs, and the machine is the specific execution component in the robot.

[0165] Here, the second conflict execution process can be a process in which robot machine conflicts occur in multiple nodes of a directed structural graph. That is, during the task execution process, there may be resource competition points between specific execution components in the same execution robot.

[0166] Specifically, the execution machine required for each execution process can be queried by iterating through the process allocation set. If an execution machine is assigned to two or more execution processes simultaneously, then these processes are identified as second conflicting execution processes.

[0167] Step 236: In the directed graph structure, add a third connection line between the nodes corresponding to the second conflict execution steps.

[0168] The third connecting line is used to represent the resource conflict constraint relationship between the machines of the two nodes.

[0169] Here, the third connecting line can be used to represent mutual exclusion or sequence constraints caused by resource sharing of specific execution components of the robot, thereby preventing the same machine from being occupied simultaneously. For a second conflicting execution step that is identified as having resource conflict constraints between machines, an undirected edge can be added as a third connecting line between the nodes corresponding to the second conflicting execution step to indicate that there is a mutual exclusion relationship between these steps and they cannot be executed simultaneously.

[0170] Step 237: Output a directed graph structure including the node, first connection line, second connection line and third connection line corresponding to each execution process in the execution process set, to obtain the disjunctive graph.

[0171] To facilitate storage, transmission, or visualization, the extract graph can be serialized into a specific data format, such as JSON, XML, or DOT language.

[0172] This application embodiment constructs a structured disjunctive graph to clearly represent the execution steps of a task, their dependencies, and execution resource constraints. Each step to be executed is abstracted and constructed as a node in a directed graph. Based on a pre-determined set of steps for the same workpiece, a first connection line is added between steps with a sequential order. This first connection line indicates the logical dependencies between steps, ensuring the correct execution flow of the task. Simultaneously, to resolve resource contention issues, the system identifies first conflicting execution steps assigned to the same execution robot and second conflicting execution steps assigned to the same machine. A second connection line is added between nodes corresponding to the first conflicting steps, and a third connection line is added between nodes corresponding to the second conflicting steps. These second and third connection lines represent the mutual exclusion or sequence constraints resulting from resource sharing; that is, these steps cannot be executed simultaneously or must be executed in a predetermined order. Thus, the disjunctive graph not only includes the logical dependencies of steps but also incorporates the resource constraints of the execution robot and the machine, forming a comprehensive and accurate task model. This scheduling model can transform complex task planning problems into scheduling problems in graph theory, generating a target scheduling graph that avoids resource conflicts and satisfies all dependencies.

[0173] The above technical solution allows for a clear and structured representation of process dependencies and resource constraints in complex task planning problems within a disjunctive graph. This not only ensures the logical correctness of the execution order but also clarifies the mutual exclusion constraints arising from resource sharing, thus avoiding resource conflicts. This enables subsequent scheduling solutions to be based on a comprehensive and accurate task model, significantly improving the rationality and efficiency of scheduling results. It ensures that tasks proceed according to the expected order and resource allocation, effectively solving the scheduling problems caused by incomplete graph representation in traditional methods.

[0174] In some of the embodiments described above in this application, a scheme is proposed to run the execution program of each execution robot according to the execution order of the target scheduling diagram. However, in practical applications, facing diverse task requirements and complex scene changes, how to efficiently and flexibly generate or obtain the execution program required by each execution robot to ensure that the execution robot can accurately complete specific execution procedures is a challenge. If the running program needs to be customized for each task and scenario, it will greatly increase development costs and time, and reduce the efficiency of task planning and the adaptability of the system.

[0175] Based on this, this embodiment further proposes a fourth embodiment of an industrial production line process flow arrangement optimization method. In the fourth embodiment, as follows: Figure 5 As shown, prior to step 25, the industrial production line process flow arrangement optimization method further includes steps B251 to B253, wherein:

[0176] Step B251: Obtain the pre-generated standard process flow tree.

[0177] The standard process flow tree includes multiple functional nodes, each representing a corresponding functional code. Each functional node is arranged in the execution order, and the functional nodes in different branch structures are used to represent the differences in functional execution under different target scenarios.

[0178] Here, the pre-generated standard process flow tree can refer to a structured data model that the system pre-builds and stores. This model is organized in a tree-like manner and is used to describe the general process of the robot performing tasks and its variations in different scenarios.

[0179] This standard process flow tree can be an XML file, a JSON file, or a database structure, defining the hierarchical relationships, sequential relationships, and associations with functional nodes and their corresponding code. Its purpose is to provide a reusable and configurable foundation for program generation, avoiding the need to write robot execution programs from scratch for each specific task.

[0180] It's important to note that functional nodes are the basic building blocks in the process tree, representing a specific function or operation performed by the robot, such as "taking a picture," "grabbing," or "placing." Each functional node is associated with a piece of program code that can be directly executed by the robot. These code snippets can be API calls specific to the robot brand, scripting language snippets, or function entry points of pre-compiled binary modules. By breaking down complex tasks into modular functional nodes, the modularity and maintainability of the program can be improved.

[0181] Functional nodes in the process tree have a clear sequential relationship, which can be defined by directed edges or by the sequence identifier in the node data to ensure the logical correctness of task execution. Furthermore, the standard process tree can contain conditional branches, allowing the system to select different execution paths based on different scenario parameters. For example, one branch might correspond to "processing type A workpieces," while another branch might correspond to "processing type B workpieces," thereby improving the program's versatility and adaptability to diverse scenarios.

[0182] Step B252: Based on the scenario description, match the target execution path for each execution process in the standard process flow tree.

[0183] The target execution path includes the function codes from the starting function node to the ending function node of the execution process.

[0184] Here, the target execution path utilizes the scenario information associated with the execution process. Through a large language model, the scenario description is compared with the branch conditions in the process tree, thereby finding the most suitable node sequence for the execution process in the standard process process tree.

[0185] It should be noted that the target execution path includes the functional code from the starting functional node to the ending functional node of the execution procedure, which clarifies the scope of code required to generate the program for the current execution procedure. This path can consist of a series of functional node identifiers or a subtree structure containing all relevant nodes and connections.

[0186] Step B253 involves concatenating the functional code fragments associated with each functional node in the execution path to generate the execution program corresponding to the execution procedure.

[0187] Specifically, the code snippets corresponding to all functional nodes on the matched target execution path can be combined according to their order in the path. The concatenation process can be a simple code string concatenation, or it can be achieved by using a code generator to assemble the code snippets into a script file or executable file that can be recognized by a specific robot platform.

[0188] This embodiment introduces a standard process flow tree, transforming the robot program generation process from manual programming to an automated, configurable coding process. After the industrial production line process flow arrangement optimization method derives the set of execution processes, the set of process assignments, and the set of process sequences for the same workpiece based on instruction information, and draws the corresponding disjunction diagram and generates the target scheduling diagram, before the robot executes the corresponding program, it can intelligently match the target execution path that meets the current task and scenario requirements within the pre-constructed standard process flow tree for each execution process, using its corresponding scenario description. The target execution path can consist of a series of functional nodes, each associated with an executable functional code fragment. Subsequently, these functional code fragments can be concatenated according to the order of the target execution path to quickly and accurately generate the robot program required for that execution process. This allows the robot to execute highly customized programs adapted to specific scenarios according to the instructions of the target scheduling diagram, improving the flexibility and efficiency of task planning.

[0189] Through the above technical solution, this application can significantly improve the automation and flexibility of robot task planning. By pre-constructing a standard process flow tree and dynamically matching and assembling functional code according to the task scenario description, the tedious work of manually writing robot execution programs for each new task or scenario is avoided, thereby greatly shortening the task development cycle and reducing development costs. At the same time, this modular and parameterized program generation method enables the system to quickly adapt to diverse task requirements and complex environmental changes, improving the efficiency of task planning and the overall adaptability of the system.

[0190] In practical applications, constructing a standard process flow tree often requires human experience and a significant amount of time, and it is difficult to flexibly adapt to constantly changing robot task requirements and scenario differences, resulting in limited efficiency and adaptability in task planning. In one specific implementation, prior to step B251, which involves obtaining the pre-generated standard process flow tree, the industrial production line process orchestration optimization method further includes a step of creating a standard process flow tree, specifically comprising:

[0191] (1) Obtain multiple robot execution program examples.

[0192] The execution program example can be existing program code or operation records of multiple robots completing various tasks.

[0193] Specifically, existing, validated robot program files can be exported from the robot controller or programming environment, or obtained by recording teaching data or programming instructions from the robot operator during actual task execution.

[0194] (2) Analyze the commonalities and scenario-specific differences in the execution functions of the execution program examples.

[0195] Among them, common functionalities and scenario-specific differences can be general patterns in robot operation and variations in specific situations, respectively.

[0196] Specifically, program code can be parsed using large language modeling techniques or code analysis tools to identify recurring instruction sequences or functional modules and extract scenario-related parameters or conditions; or, through expert systems or machine learning models, pattern recognition and cluster analysis can be performed on a large number of executable program examples to distinguish core functions from variations in specific scenarios.

[0197] (3) Merge executable program examples with common functions into common function nodes.

[0198] Abstract program fragments that perform the same or similar basic operations into reusable modules.

[0199] For example, identify code segments or operation sequences that appear repeatedly in multiple executable program examples and have the same semantics, and abstract them into a unified functional identifier or module as a common functional node; or through manual review and summarization, merge program fragments that implement the same or similar basic operations into a common functional node.

[0200] (4) Generate different branch structures from the scenario-specific execution program examples.

[0201] To handle operational variations under different task scenarios, an independent branch path can be created for each different scenario based on the analyzed scenario differences; or conditional logic or decision tree structure can be used to represent the execution path under different scenarios, with each path corresponding to a branch structure.

[0202] (5) Write the corresponding executable code fragments for each functional node in the branch structure to generate a standard process flow tree that includes branch structures based on scenario differences and common functional nodes with common functions.

[0203] Branching structures can associate abstract functional nodes with specific execution instructions.

[0204] Specifically, code segments corresponding to specific functional nodes and branch structures in the original execution program example can be directly copied or referenced into the functional node; or an interface or template can be defined for each functional node, and then the corresponding executable code can be filled or generated according to the specific scenario and functional requirements to generate a standard process flow tree that includes branch structures based on scenario differences and common functional nodes with common functionalities. The standard process flow tree can be a structured knowledge base that organizes the above processing results.

[0205] By organizing the aforementioned merged functional nodes and branch structures into a tree-like data structure, where the root node can represent the start of a task, and intermediate nodes and branches can represent different functional and scenario paths; or by using a graphical programming interface or a domain-specific language to construct and visualize this standard process flow tree, its logic is clear and easy to maintain.

[0206] In the above implementation, a large number of robot execution program examples are first obtained and analyzed in depth to identify common execution functions and differences arising from different scenarios. Then, these common functions are abstracted and merged into reusable functional nodes, while scenario-specific differences are transformed into different branch structures. Next, specific, executable code snippets are associated with the functional nodes in each branch structure. Finally, these functional nodes and branch structures are organized into a hierarchical standard process flow tree. This method transforms the generation process of the standard process flow tree from being driven by manual experience to being data-driven and automated, greatly improving construction efficiency and accuracy. The standard process flow tree generated in this way not only includes the general logic of robot task execution but also flexibly considers the specific operational details under different scenarios, thus providing a more comprehensive, accurate, and easily maintainable knowledge base for subsequent task planning.

[0207] The above technical solutions enable the automated or semi-automated generation of standard process flow trees, significantly reducing the cost and complexity of manual construction and maintenance. Simultaneously, by effectively identifying and organizing functional commonalities and scenario-specific differences, the reusability of robot programs and the flexibility of task planning are improved, allowing the robot system to adapt to new task requirements and working environments more quickly and accurately.

[0208] The following example illustrates this embodiment; please refer to it. Figure 6 , Figure 6 This is a scenario illustration. Specifically, to address the technical problems of existing technologies relying on manual offline planning and programming, and the difficulty of adapting existing solutions based on large language models to the strong constraints and high reliability requirements of industrial fields, leading to infeasible planning and unexecutable programs, an industrial production line process flow arrangement optimization method is proposed in the following example.

[0209] The core inputs are a machine-readable industrial production scenario description S and task instructions I issued by the user in natural language. The scenario description S defines the physical and logical configuration of the target production line in a structured format (e.g., JSON). The scenario description S includes at least: the identifiers, types, poses, and end effector types of all robots; the identifiers, locations, and functions of all key workstations, equipment, and sensors; the initial state and location of all materials to be processed; and the spatial and mutual exclusion constraints between resources.

[0210] After determining the scenario description S and task instruction I, the industrial production line process flow arrangement optimization method proposed in this embodiment can automatically generate a high-level task plan H and a low-level execution program P that can drive the production line to perform tasks.

[0211] Among them, the task planning H specifies all the procedures required to complete the instructions, the executor of each procedure, and the optimized execution sequence that satisfies all constraints; the execution program P generates specific control code for each procedure that can be directly run on the corresponding robot controller.

[0212] In the task planning H phase, fuzzy instructions from natural language can be transformed into precise and conflict-free scheduling instructions. This includes semantic parsing, model building, and optimization solutions, thereby obtaining the corresponding target scheduling graph.

[0213] For example, the scenario description S and task instruction I can be input into the large language model. By adopting a chain-like thinking prompt strategy and a general rule-based prompt template, the LLM is guided to perform coherent reasoning and output the following in sequence: (1) Based on the available resource capabilities in the scenario, the high-level task instruction is decomposed into a series of discrete, executable atomic processes to form a set of execution processes; (2) According to the robot capability list in the scenario description, a suitable execution robot is assigned to each process in the set of execution processes. It is necessary to ensure that the robot has the required functions and meets the constraints such as the working range, so as to form a mapping relationship between processes and robots, i.e., a set of process assignments; (3) For each independent processing object, the inherent and irreversible logical sequence of processes is inferred to form a set of process sequences for the same workpiece.

[0214] After determining the set of execution operations, the set of operation assignments, and the set of operation sequences for the same job, a disjunctive graph for the job shop scheduling problem can be automatically generated. Specifically, an initial graph structure is created with each execution operation as a node. Next, the set of operation sequences for the same job is traversed. For each defined sequence constraint, a directed edge, called a conjunctive arc, is added between the corresponding two operation nodes to represent this mandatory sequential dependency. Finally, the set of operation assignments is analyzed to identify all pairs of operations assigned to the same robot or machine. Due to resource exclusivity, these operations cannot be executed simultaneously. An undirected edge, called a disjunctive arc, is added between each such pair of nodes to represent the resource conflict constraint where the order is undetermined and a choice must be made. After constructing the disjunctive graph, the nodes represent execution operations, the conjunctive arcs represent logical constraints, and the disjunctive arcs represent resource contention constraints.

[0215] After generating the disjunction graph, optimal scheduling can be solved based on it to generate the corresponding target scheduling graph. It should be noted that the disjunction graph contains logical constraints between execution steps and resource competition constraints between robots. By calling a heuristic scheduling algorithm to solve the disjunction graph, a direction can be determined for each disjunction arc, resulting in a completely deterministic and feasible scheduling sequence. In a preferred embodiment, a scheduler based on the first-in, first-out (FIFO) rule can be used. This scheduler maintains a global simulation clock and event queue to simulate the execution process. Whenever a robot becomes idle, the scheduler checks all currently completed preceding steps that require the robot to be "ready," and selects the one with the earliest entry time for scheduling. It then determines that other conflicting steps must be executed after it. By iteratively advancing the simulation clock and processing events, the direction of all disjunction arcs is determined sequentially. After the solution is completed, the system generates a visual target scheduling graph based on the determined start and end times of each step. This target scheduling graph precisely specifies the execution order of each step.

[0216] After determining the target scheduling diagram, in order to transform the high-level task planning into specific actions that the robot can execute, this embodiment proposes a program generation method based on the process flow tree to replace the traditional method of directly prompting code generation, which significantly improves the completeness, correctness and reusability of the generated code.

[0217] Construct a unified standard process flow tree T that covers all expected process types, serving as the core knowledge base. Once completed, it can be reused in all subsequent tasks.

[0218] Specifically, a set E of robot execution program examples covering various processes, task instructions, and production line scenarios can be collected. These examples are input into a large language model, guiding it to analyze and summarize the common operational logic and implementation differences caused by scenario configurations in different examples. Further, a structured tree description is output based on the implementation differences. In this tree, operation steps with completely identical functions and execution logic are merged into the same common node; operations with the same function but different code implementations due to different scenario parameters form branch structures in the tree. Each node records its supported process types, functional descriptions, and corresponding executable code snippets. After manual verification, the completed process tree is persistently stored in JSON format.

[0219] For example, a process tree may include common nodes with multiple common functions and branching structures with differences, including transport nodes, polish nodes, and transport nodes.

[0220] When it is necessary to generate an execution program for a certain execution procedure in the task planning, the following code generation steps can be performed. Specifically, (1) Path matching. Based on the type of the procedure (e.g., "polishing") and the specific configuration parameters in the current scene description S (e.g., whether the camera is fixed or handheld), a unique execution path from the start node to the end node is automatically searched and matched in the pre-generated standard procedure flow tree T. This path has adaptively selected a function branch that matches the current scene, and the function is called based on the determined function branch. (2) Code synthesis. The code fragments associated with all nodes on the matched execution path are packaged into functions according to their order on the path, so as to splice and combine the code of the called function to form a complete general execution function for the procedure. (3) Output a packaged program that calls this general execution function as the final execution function of the procedure.

[0221] For example, after determining the conveyor belt, W1, transport pallet, and polishing disc 01 for process 1, a function call can be made to determine the use of the following function: operation1("Transport","W1","Pallet", "Polishing table 1", "Robot 1"). That is, the workpiece is transported to the pallet according to the workpiece operation sequence W1, and polished on table 1 by Robot 1. After determining the workpiece operation sequence, code is concatenated and combined based on the following packaging function:

[0222] def operation1(operation_type, workpiece, machine_1, machine_2,robot):

[0223] transport_1(workpiece, machine_1, machine_2, robot).

[0224] The final output includes the executed functions for each node, as follows:

[0225] def transport_1(workpiece_name, machine_1, machine_2, robot):

[0226] # Path in process tree:

[0227] # Start -> Node 2 -> Node 3 -> Node 4 -> Node 5 -> Node 7 -> End.

[0228] After obtaining both the target scheduling diagram H and the execution programs P for each process, the multi-robot collaborative execution phase begins. The central scheduler or distributed control system on the production line strictly follows the timeline and dependencies specified in the target scheduling diagram H, triggering the execution of each process sequentially. When the triggering conditions for a certain process are met, the system automatically loads and runs its corresponding execution program P, precisely controlling the designated robot to complete operations such as grasping, moving, and processing. In this way, the process sequence constraints and resource mutual exclusion constraints are strictly guaranteed during operation, thereby enabling safe, orderly, and efficient collaboration among multiple robots in a shared resource environment, ultimately completing the entire complex manufacturing task.

[0229] In other embodiments, in order to verify the effectiveness, advancement and industrial applicability of the present invention, systematic testing and verification were carried out in both simulated environments and real physical production lines.

[0230] On the one hand, please refer to Figure 7 , Figure 7 This paper presents a scenario illustration in a simulation environment. Fifty typical manufacturing tasks, encompassing three collaborative modes—single robot, simple multi-robot, and complex multi-robot—were constructed in the simulator. Five key indicators were used for quantitative evaluation: process correctness, scheduling efficiency, program executability, target state recall rate, and overall success rate. Comparative experiments show that the method proposed in this embodiment significantly outperforms existing methods that rely entirely on LLM direct planning or use a few example hints to generate code in all indicators. In particular, by entrusting the modeling and solving of complex resource conflicts to a deterministic scheduling algorithm and introducing a reusable process flow tree in program generation, which can be determined by keyframe images 1 to 6 collected in the actual application scenario, a breakthrough improvement has been achieved in both the feasibility of the scheduling scheme and the executability of the generated program, thus achieving a high overall task success rate.

[0231] On the other hand, please refer to Figure 8 , Figure 8This diagram illustrates a scenario on a real physical production line. The system of this invention was deployed and tested in a real experimental production line consisting of three six-axis collaborative robotic arms, a conveyor belt, and a fixed vision system. The given task instruction was "Remove the workpiece from the material frame, locate it visually, and collaboratively place it at the designated position on the conveyor belt." Upon receiving this natural language instruction, the system automatically completed the following processes: parsing the instruction and decomposing it into multiple handling steps; allocating tasks according to robot capabilities; constructing and solving a disjunctive graph to generate a collision-avoiding and efficient three-robot collaborative scheduling scheme; and generating robot control code for each step that could be directly executed based on a pre-built process flow tree. Keyframe images 1 to 6, collected in the actual application scenario and related to robots 1, 2, and 3, confirmed that the three robots worked collaboratively strictly according to the generated target scheduling graph, with smooth sequential connections and no resource conflicts, successfully completing the workpiece transfer task. Real-world testing powerfully demonstrates the invention's closed-loop capability from natural language to physical execution, and its applicability and reliability in real industrial scenarios.

[0232] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0233] Based on the same inventive concept, this application also provides an industrial production line process flow arrangement optimization device for implementing the above-mentioned industrial production line process flow arrangement optimization method. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the industrial production line process flow arrangement optimization device provided below can be found in the limitations of the industrial production line process flow arrangement optimization method above, and will not be repeated here.

[0234] In one exemplary embodiment, an industrial production line process flow orchestration and optimization device is provided, comprising: an acquisition module, a semantic analysis module, a drawing module, a solution module, and an execution module, wherein:

[0235] The acquisition module is used to acquire instruction information, including scene description and task description;

[0236] The semantic analysis module is used to perform semantic analysis on the instruction information to deduce the set of execution procedures, the set of procedures allocation, and the set of procedures sequence for the same workpiece required to complete the task corresponding to the task description in the scenario corresponding to the scenario description.

[0237] The drawing module is used to draw a corresponding disjunction graph based on the execution processes of the execution process set, the execution robots required for each execution process in the process allocation set, and the execution order in the process sequence set of the same workpiece; wherein, the disjunction graph is used to characterize the execution order and resource allocation constraints between different processes in the process set;

[0238] The solution module is used to perform scheduling solution on the disjunctive graph, determine the execution order of the processes under the resource allocation constraints, and obtain the target scheduling graph; wherein, the target scheduling graph is used to represent the target execution order of the processes.

[0239] The execution module is used to execute the execution program corresponding to each execution step in sequence according to the execution order of the target scheduling diagram.

[0240] In one embodiment, the semantic analysis module includes an input unit, a parsing unit, an allocation unit, and a definition unit, wherein:

[0241] The input unit is used to input the scene description and task description in the instruction information into the large language model for semantic analysis in order to determine the scene configuration corresponding to the scene description.

[0242] The parsing unit is used to decompose the task description into multiple corresponding execution procedures according to the scenario configuration, so as to obtain the set of execution procedures corresponding to the instruction information;

[0243] An allocation unit is used to allocate a corresponding execution robot to each execution process in the execution process set, so as to determine the process allocation set;

[0244] A definition unit is used to define the execution order of all processes that process the same workpiece in the execution process set, resulting in a process sequence set for the same workpiece.

[0245] In one embodiment, the allocation unit includes a first matching subunit and a first output subunit, wherein:

[0246] The first matching subunit is used to match an execution robot with corresponding functions for each execution process in the execution process set based on the robot list in the scenario configuration corresponding to the scenario description.

[0247] The first output subunit is used to output the matching execution robot for each execution process, thereby obtaining the process allocation set.

[0248] In one embodiment, the defining unit includes a derivation subunit and a determining subunit, wherein:

[0249] A derivation subunit is used to derive multiple execution processes in the set of execution processes to determine the dependencies between the execution processes that process the same workpiece.

[0250] A subunit is defined to determine a list of sequences in which each execution robot performs different execution processes based on the dependencies between the execution processes of the same workpiece, thereby obtaining a set of process sequences for the same workpiece.

[0251] In one embodiment, the drawing module includes a construction unit, a first connection unit, a first recognition unit, a second connection unit, a second recognition unit, a third connection unit, and a second output unit, wherein:

[0252] Construction unit, used to construct a directed graph structure with each execution operation in the set of execution operations as a node;

[0253] The first connection unit is used to add a first connection line between two corresponding nodes in the directed graph structure according to the execution order in the same workpiece process sequence set; wherein, the first connection line is used to represent the dependency relationship between the two nodes;

[0254] The first identification unit is used to identify conflicting execution processes assigned to the same robot or machine in the process assignment set;

[0255] The first identification unit is used to identify, in the process allocation set, a first conflicting execution process assigned to the same execution robot; wherein, the first conflicting execution process is the process in which a conflict occurs between execution robots.

[0256] The second connection unit is used to add a second connection line between the nodes corresponding to the first conflict execution process in the directed graph structure, wherein the second connection line is used to represent the resource conflict constraint relationship of the execution robot between the two nodes;

[0257] The second identification unit is used to identify a second conflicting execution process that needs to use the same machine in the set of execution processes; wherein the second conflicting execution process is a process in which machine conflict occurs, and the machine is a specific execution component in the execution robot;

[0258] The third connection unit is used to add a third connection line between the nodes corresponding to the second conflict execution process in the directed graph structure; wherein the third connection line is used to represent the resource conflict constraint relationship between the two nodes.

[0259] The second output unit is used to output a directed graph structure including the node corresponding to each execution process in the execution process set, the first connection line, the second connection line and the third connection line, to obtain the disjunctive graph.

[0260] In one embodiment, the industrial production line process scheduling and optimization device further includes an execution program generation module, which comprises a first acquisition unit, a second matching unit, and a splicing unit, wherein:

[0261] The first acquisition unit is used to acquire a pre-generated standard process flow tree; wherein, the standard process flow tree includes multiple functional nodes, each functional node represents a corresponding functional code, each functional node is arranged in the execution order, and the functional nodes in different branch structures are used to represent the differences in functional execution under different target scenarios;

[0262] The second matching unit is used to match the target execution path in the standard process flow tree for each execution process according to the scenario description; wherein, the target execution path includes the function code from the start function node to the end function node of the execution process;

[0263] The splicing unit is used to splice together the functional code fragments associated with each functional node in the execution path to generate the execution program corresponding to the execution procedure.

[0264] In one embodiment, the industrial production line process orchestration optimization device further includes a process tree creation module, which comprises a second acquisition unit, an analysis unit, a first generation unit, a second generation unit, and a third generation unit, wherein:

[0265] The second acquisition unit is used to acquire multiple robot execution program examples;

[0266] The analysis unit is used to analyze the commonalities in the execution functions and the differences in specific scenarios among the execution program examples.

[0267] The first generation unit is used to merge executable program examples with common functionalities into common functional nodes;

[0268] The second generation unit is used to generate different branch structures from the execution program examples with scenario-specific differences.

[0269] The third generation unit is used to write corresponding executable code fragments for each functional node in the branch structure, and generate a standard process flow tree that includes branch structures based on scenario differences and common functional nodes with common functionalities.

[0270] Each module in the aforementioned industrial production line process flow arrangement and optimization device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0271] Specifically, the acquisition module receives instruction information entered by the operator through a text input interface, or scene and task descriptions read from a preset configuration file. The semantic analysis module performs deep parsing of this instruction information based on a large language model, identifies the resource configuration of the target scene, decomposes the task description into a discrete set of execution procedures, matches the required execution robots for each execution procedure to form a procedure allocation set, and derives the dependencies between procedures to construct a set of procedure sequences for the same workpiece. The drawing module constructs a disjunctive graph based on the execution procedure set, the procedure allocation set, and the set of procedure sequences for the same workpiece. Execution procedures are used as nodes, with first connecting lines representing the sequential dependencies defined in the set of procedure sequences for the same workpiece, and second connecting lines identifying the mutual exclusion constraints between conflicting execution procedures assigned to the same execution robot. The solution module applies a heuristic scheduling algorithm to the disjunctive graph, comprehensively considering procedure sequence constraints and mutual exclusion constraints of execution robot resources, to generate an optimized target scheduling graph. The execution module coordinates each execution robot to call its execution program sequentially to complete the task according to the execution order and allocation constraints of the target scheduling graph.

[0272] Through the above technical solution, this application effectively solves the task planning problem in industrial multi-robot collaboration. Due to the strict constraints on process sequence and resource exclusivity in industrial scenarios, traditional methods relying on manual offline planning are prone to errors and are inefficient. Existing methods based on large language models lack specialized processing for industrial constraints. This device accurately extracts industrial constraint elements through a semantic analysis module, constructs a disjunctive graph containing process dependencies and resource exclusivity through a drawing module, and performs targeted scheduling and solving through a solution module. This generates a feasible and efficient scheduling scheme while satisfying strong industrial constraints. Compared to basic solutions, this device achieves end-to-end automation from natural language instructions to executable programs, significantly improving the rationality of task planning, the efficiency of execution scheduling, and the executability of the executable programs, providing key technical support for the large-scale application of embodied intelligence in industry.

[0273] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 9 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores data such as execution procedures corresponding to scene and task descriptions, and the execution robots allocated to these procedures. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements an industrial production line process flow scheduling optimization method.

[0274] It should be noted that by combining industrial production line process flow orchestration optimization methods with computer equipment hardware architecture, the processor can execute computer programs stored in memory, thereby achieving end-to-end automatic generation from natural language descriptions to multi-robot collaborative executable programs. Specifically, the computer equipment can acquire instruction information including scene descriptions and task descriptions, perform semantic analysis on the instruction information to deduce the target scene and the corresponding set of execution processes, process allocation sets, and sets of process sequences for the same workpiece. Based on this information, a disjunctive graph is drawn and a scheduling solution is performed, ultimately generating a target scheduling graph to guide the execution program of the robots. After running the execution program, the corresponding task planning objectives can be achieved.

[0275] Through the above technical solution, this computer device effectively solves the problems of strict process sequence constraints and resource mutual exclusion constraints in industrial scenarios. In the context of intelligent manufacturing, traditional methods rely on manual offline task planning and programming, which is inefficient, error-prone, and unable to quickly respond to complex and diverse production demands. Existing planning methods based on large language models are mostly concentrated in indoor home scenarios and fail to effectively adapt to the strong constraints and high reliability requirements of the industrial field, resulting in scheduling plans that are often infeasible or inefficient in real production lines. The solution in this application integrates high-level task planning and low-level execution program generation into a unified computer device, ensuring the rationality of task planning, the efficiency of execution scheduling, and the executability of the execution program.

[0276] Those skilled in the art will understand that Figure 9The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0277] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described industrial production line process flow arrangement optimization method embodiment.

[0278] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0279] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0280] The above embodiments merely illustrate several implementation methods of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for optimizing the process flow arrangement of an industrial production line, characterized in that, The method includes: Obtain instruction information including scene description and task description; Semantic analysis is performed on the instruction information to deduce the set of execution procedures, the set of procedures assigned, and the set of procedures for the same workpiece required to complete the task corresponding to the task description in the scenario corresponding to the scenario description. Based on the execution processes of the execution process set, the execution robots required for each execution process in the process allocation set, and the execution order in the process sequence set for the same workpiece, a corresponding disjunction graph is drawn; wherein, the disjunction graph is used to characterize the execution order and resource allocation constraints between different processes in the process set; The disjunctive graph is used to solve for scheduling, and the execution order of the processes under the resource allocation constraints is determined to obtain the target scheduling graph; wherein, the target scheduling graph is used to represent the target execution order of the processes. According to the execution order of the target scheduling diagram, the execution program corresponding to each execution step is executed sequentially; The step of drawing the corresponding disjunction diagram based on the execution processes of the execution process set, the execution robots required for each execution process in the process allocation set, and the execution order in the same workpiece process sequence set includes: Construct a directed graph structure with each execution process in the set of execution processes as a node; Based on the execution order in the same workpiece process sequence set, a first connection line is added between two corresponding nodes in the directed graph structure; wherein, the first connection line is used to represent the dependency relationship between the two nodes; Identify the first conflicting execution process assigned to the same execution robot in the process assignment set; wherein, the first conflicting execution process is the process in which the execution robots conflict; In the directed graph structure, a second connection line is added between the nodes corresponding to the first conflict execution process, wherein the second connection line is used to represent the resource conflict constraint relationship of the execution robot between the two nodes; Identify a second conflicting execution process that requires the use of the same machine from the set of execution processes; wherein the second conflicting execution process is a process in which machine conflict occurs, and the machine is a specific execution component in the execution robot; In the directed graph structure, a third connection line is added between the nodes corresponding to the second conflict execution process; wherein, the third connection line is used to represent the resource conflict constraint relationship between the two nodes. The output includes a directed graph structure comprising the node corresponding to each execution step in the execution step set, the first connection line, the second connection line, and the third connection line, thus obtaining the disjunctive graph.

2. The method according to claim 1, characterized in that, The step of performing semantic analysis on the instruction information to deduce the set of execution procedures, the set of procedures assigned, and the set of procedures for the same workpiece required to complete the task corresponding to the task description in the scenario corresponding to the scenario description includes: The scene description and task description in the instruction information are input into a large language model for semantic analysis to determine the scene configuration corresponding to the scene description. Based on the scenario configuration, the task description is decomposed into multiple corresponding execution procedures to obtain the set of execution procedures corresponding to the instruction information; Assign a corresponding execution robot to each execution process in the set of execution processes to determine the process assignment set; Define the execution order of all processes that process the same workpiece in the set of execution processes to obtain the set of process sequences for the same workpiece.

3. The method according to claim 2, characterized in that, The step of assigning a corresponding execution robot to each execution process in the execution process set to determine the process assignment set includes: Based on the robot list in the scenario configuration corresponding to the scenario description, an execution robot with corresponding functions is matched for each execution process in the execution process set; Output the matching execution robot for each execution process to obtain the process allocation set.

4. The method according to claim 2, characterized in that, The step of defining the execution order of all processes handling the same workpiece in the set of execution processes to obtain the set of process sequences for the same workpiece includes: The dependencies between multiple execution processes in the set of execution processes are derived to determine the dependencies between the execution processes that process the same workpiece. Based on the dependencies between the execution processes of the same workpiece, a list of sequences in which each execution robot executes different execution processes is determined, resulting in a set of process sequences for the same workpiece.

5. The method according to claim 1, characterized in that, Before the step of sequentially executing the execution program corresponding to each execution step according to the execution order of the target scheduling diagram, the industrial production line process flow arrangement optimization method further includes: Obtain a pre-generated standard process flow tree; wherein, the standard process flow tree includes multiple functional nodes, each functional node represents a corresponding functional code, each functional node is arranged in execution order, and the functional nodes in different branch structures are used to represent the differences in functional execution under different target scenarios; According to the scenario description, a target execution path is matched in the standard process flow tree for each execution procedure, wherein the target execution path includes the function code from the start function node to the end function node of the execution procedure; The functional code fragments associated with each functional node in the execution path are concatenated to generate the execution program corresponding to the execution procedure.

6. The method according to claim 5, characterized in that, Prior to the step of obtaining the pre-generated standard process flow tree, the method further includes: Obtain multiple robot execution program examples; Analyze the commonalities and scenario-specific differences in the execution functions of the execution program examples; The executable program examples that share common functionalities are merged into common functional nodes; The executable program examples with scenario-specific differences will be generated into different branch structures; Write corresponding executable code fragments for each functional node in the branch structure to generate a standard process flow tree that includes the branch structure based on scenario differences and the common functional nodes with common functionalities.

7. An industrial production line process flow arrangement and optimization device, characterized in that, The device includes: The acquisition module is used to acquire instruction information, including scene description and task description; The semantic analysis module is used to perform semantic analysis on the instruction information to deduce the set of execution procedures, the set of procedures allocation, and the set of procedures sequence for the same workpiece required to complete the task corresponding to the task description in the scenario corresponding to the scenario description. The drawing module is used to draw a corresponding disjunction graph based on the execution processes of the execution process set, the execution robots required for each execution process in the process allocation set, and the execution order in the process sequence set of the same workpiece; wherein, the disjunction graph is used to characterize the execution order and resource allocation constraints between different processes in the process set; The solution module is used to perform scheduling solution on the disjunctive graph, determine the execution order of the processes under the resource allocation constraints, and obtain the target scheduling graph; wherein, the target scheduling graph is used to represent the target execution order of the processes. The execution module is used to execute the execution program corresponding to each execution step in sequence according to the execution order of the target scheduling diagram; The drawing module is also used to construct a directed graph structure with each execution process in the set of execution processes as a node; Based on the execution order in the same workpiece process sequence set, a first connection line is added between two corresponding nodes in the directed graph structure; wherein, the first connection line is used to represent the dependency relationship between the two nodes; Identify the first conflicting execution process assigned to the same execution robot in the process assignment set; wherein, the first conflicting execution process is the process in which the execution robots conflict; In the directed graph structure, a second connection line is added between the nodes corresponding to the first conflict execution process, wherein the second connection line is used to represent the resource conflict constraint relationship of the execution robot between the two nodes; Identify a second conflicting execution process that requires the use of the same machine from the set of execution processes; wherein the second conflicting execution process is a process in which machine conflict occurs, and the machine is a specific execution component in the execution robot; In the directed graph structure, a third connection line is added between the nodes corresponding to the second conflict execution process; wherein, the third connection line is used to represent the resource conflict constraint relationship between the two nodes. The output includes a directed graph structure comprising the node corresponding to each execution step in the execution step set, the first connection line, the second connection line, and the third connection line, thus obtaining the disjunctive graph.

8. The industrial production line process flow arrangement and optimization device according to claim 7, characterized in that, The industrial production line process flow arrangement and optimization device further includes an execution program generation module, which comprises a first acquisition unit, a third matching unit, and a splicing unit, wherein: The first acquisition unit is used to acquire a pre-generated standard process flow tree; wherein, the standard process flow tree includes multiple functional nodes, each functional node represents a corresponding functional code, each functional node is arranged in the execution order, and the functional nodes in different branch structures are used to represent the differences in functional execution under different target scenarios; The third matching unit is used to match the target execution path in the standard process flow tree for each execution procedure according to the scenario description; wherein, the target execution path includes the function code from the start function node to the end function node of the execution procedure; The splicing unit is used to splice together the functional code fragments associated with each functional node in the target execution path to generate the execution program corresponding to the execution procedure.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

Images