Flexible job-shop scheduling method based on multi-agent cooperation

Abstract

The present application relates to the field of flexible workshop production scheduling, and specifically to a flexible workshop production scheduling method based on multi-agent cooperation, which obtains the production scheduling requirements of a production manager in the form of natural language through a main agent, performs intention recognition on the production requirements through an intention-related agent, collects production information through an information agent calling a tool, and then enters a scheduling negotiation stage, in which the main agent and a scheduling agent perform multiple rounds of negotiation; in each round of negotiation, the main agent outputs a production plan according to the results of the intention recognition stage, and the scheduling agent generates a scheduling scheme by calling a scheduling tool according to the production plan of the main agent; if the generated production plan or scheduling scheme does not meet the requirements, the next round of negotiation is entered. The present application can continuously obtain the production scheduling requirements of a production manager in the form of natural language and timely give resource feasible and time feasible production plans and scheduling schemes.

Description

Technical Field

[0001] The present invention relates to the field of flexible workshop production scheduling, and in particular to a flexible workshop production scheduling method based on multi-agent collaboration. Background Art

[0002] Traditional shop floor scheduling algorithms take as input the machine processing sequence and time for each job and generate scheduling solutions using scheduling rules, heuristic algorithms, deep reinforcement learning algorithms, and other methods. These algorithms have the following shortcomings:

[0003] 1) Failure to achieve intelligent scheduling: Traditional job shop scheduling problems use heuristic algorithms or deep reinforcement learning algorithms, which only perform combinatorial optimization on data. When faced with insufficient resources or time constraints, they can only conclude that scheduling is unavailable.

[0004] 2) Inability to achieve natural language interaction: Traditional job scheduling is implemented using programs, which can only accept input in a fixed format. Therefore, it lacks flexibility and adaptability, cannot be intelligently expanded (for example, input multimodal information such as voice and images), and has difficulty in providing personalized solutions based on context. It is even more difficult to participate in real-time workshop management as an artificial intelligence (AI) assistant.

[0005] In today's digital age, the rapid development of large language model (LLM) engines has brought unprecedented opportunities and challenges to various industries. In particular, the application ecosystem built around large language models (such as ChatGLM) can not only significantly enhance the practical application of artificial intelligence, provide intelligent solutions for specific fields, but also promote business model innovation and industrial transformation. As a powerful technical tool, the core capability of large language models lies in understanding and generating natural language.

[0006] The multi-agent system generates multiple agents by building a large language model and using prompt words and the large language model. Each agent in the system has a specific role and background, a set of external tools with different functions, performs specific tasks, and works together to achieve the same goal from different perspectives. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the existing technology and propose a flexible workshop production scheduling method based on multi-agent collaboration, which can continuously obtain the production scheduling needs of production managers in the form of natural language, and promptly provide resource-feasible and time-feasible production plans and scheduling solutions.

[0008] In order to achieve the above object, the technical solutions specifically adopted by the present invention are as follows:

[0009] The flexible workshop production scheduling method based on multi-agent collaboration includes the following steps:

[0010] Step 1: Divide the multi-agent collaborative process for flexible workshop production scheduling into three stages, including:

[0011] The intention recognition stage (InterpretStage) is used to identify the intention of the production manager. Through the collaborative work of multiple intelligent agents, the various recognition results are formatted and stored in the cache area. After completion, it enters the next stage;

[0012] The information retrieval stage (InformationStage) is used to collect current production information, store it in the cache, and then enter the next stage after completion;

[0013] The scheduling negotiation stage (ScheduleStage) is used to control the two agents to continuously negotiate to produce a feasible production plan based on the results of the intention recognition stage, and finally generate a scheduling solution.

[0014] Step 2: Create a main agent, task agent, time agent, quantity agent, preference agent, expedited agent, information agent, and scheduling agent, which are responsible for semantic understanding and intent recognition, process control, information collection, production plan design and verification, and scheduling plan generation, respectively; specifically:

[0015] The MasterAgent is responsible for the control transition between stages. During the intention recognition stage, it receives the needs of the production manager, integrates the results of the intention recognition into formatted data, and stores them in the public cache. During the scheduling negotiation stage, it generates the production plan and verifies the feasibility of the production plan.

[0016] The TaskAgent is used to identify the task intent during the intent recognition phase and distinguish whether the task is production, order insertion, order cancellation, or fault.

[0017] The TimeAgent is used to identify the time nodes of tasks during the intent recognition phase. For production tasks, this time node represents the production deadline; for order insertion tasks, it represents the time point of order insertion; for order cancellation tasks, it represents the time point of order cancellation; and for fault tasks, it represents the time period of the fault.

[0018] AmountAgent is used to identify the products that need to be produced and the quantity of each product during the intention recognition phase;

[0019] The preference agent (PretentAgent) is used to identify task preferences during the intention recognition phase and distinguish whether the task is time-prioritized or quantity-prioritized (the preference only applies to production tasks);

[0020] PriorityAgent is used to identify task priorities during the intent recognition phase and identify tasks that need to be completed urgently.

[0021] The InfoAgent is used to call external tools related to production information to collect production information during the information retrieval phase and store the retrieved information into the cache area one by one;

[0022] The scheduling agent (ScheduleAgent) is used to obtain a scheduling solution according to the production plan using tools during the scheduling negotiation phase.

[0023] Step 3: Build an external tool library, define multiple tools, and interface with external databases or software systems for data interaction;

[0024] Step 4: After the main agent obtains the production manager's production scheduling requirements in the form of natural language, it enters the intent recognition stage. The agent related to the intent recognizes the production requirements, formats the recognition results, and stores them in the cache.

[0025] Step 5: Entering the information retrieval phase, the information agent uses tools to collect production information and stores it in the cache;

[0026] Step 6: Enter the scheduling negotiation phase. The master agent and the scheduling agent conduct multiple rounds of negotiations. In each round, the master agent outputs a production plan based on the results of the intention recognition phase. The scheduling agent uses the scheduling tool to generate a scheduling solution based on the master agent's production plan. If the generated production plan or scheduling solution does not meet the requirements, the next round of negotiation begins.

[0027] Step 7: The main agent continues to wait for new production scheduling requirements from the production manager, and proceeds to step 4 after receiving the new requirements.

[0028] Furthermore, the plurality of tools include:

[0029] Product information query tool, used to query the machine process and material information of the product;

[0030] Inventory query (Storage) tool, used to query current material inventory information;

[0031] Machine query tool, used to query the current list of idle machines;

[0032] Human query tool, used to query the current human resources situation;

[0033] The production plan check tool is used to receive the production plan as input, combine it with the current production resources, and output the judgment result of whether the production plan is feasible;

[0034] Production plan generation (Plan) tool, used to output the maximum production quantity under the current production resource constraints;

[0035] The scheduling plan generation (Schedule) tool is used to take the production plan as input and output a scheduling plan with time span priority based on tasks, preferences, and priorities.

[0036] Furthermore, the step 4 specifically includes the following steps:

[0037] S4.1 uses task agents to identify specific production scheduling tasks corresponding to production managers' needs, including production, order insertion, order cancellation, and failure. Ordinary production tasks are identified as "production," while the production of a specific number of products at a certain point in time is identified as "order insertion," while the reduction of a specific number of products is identified as "order cancellation," and the suspension of a machine for some reason is identified as "failure."

[0038] S4.2 Identify the time and quantity parameters of production through time agents and quantity agents;

[0039] S4.3 identifies the preference of selecting demand between “time priority” and “quantity priority” through preference agent identification. Producing as many target products as possible within a specific time is identified as “time priority”, and requiring the production of a specific number of target products is identified as “quantity priority”.

[0040] S4.4 uses the expedited agent to identify the expedited jobs in the demand and increase the priority of a certain job.

[0041] Furthermore, the step 5 specifically includes the following steps:

[0042] Through the information agent, it calls external tools related to information query to obtain production information from external databases or software interfaces, including product materials, processing process information, material inventory information, idle machine information, and human resource information. The obtained information is formatted and stored in the cache area, which serves as background knowledge and preset conditions for other agents to use when checking the feasibility of production plans, generating production plans, and generating scheduling plans.

[0043] Furthermore, in step 6, there are different negotiation methods for different production intentions:

[0044] 1) The time-priority production task is divided into the following sub-steps:

[0045] S6.1 The main agent calls the production plan generation (Plan) tool to provide the maximum production quantity of the corresponding product under the current resource conditions;

[0046] S6.2 The scheduling agent calls the scheduling plan generation tool for the production plan to obtain a scheduling plan and obtains scheduling data such as the time span and the production time consumed by each product;

[0047] S6.3 The master agent checks the scheduling data and outputs "Accepted" if the time span is within the specified time; otherwise, the quantity of each product is reduced proportionally based on the production time consumed by each product;

[0048] S6.4 The scheduling agent observes the output of the main agent. If it is "Accepted", the scheduling is completed and the scheduling plan is output; otherwise, repeat step S6.2.

[0049] 2) Quantity-priority production tasks are divided into the following sub-steps:

[0050] S6.1 The main intelligent agent uses the production plan verification (Check) tool to check whether the production plan proposed by the production manager is feasible. If so, it outputs the original production plan. Otherwise, it uses the production plan generation (Plan) tool to generate the maximum production quantity that meets the current resource conditions.

[0051] S6.2 The scheduling agent identifies the output of the main agent and calls the scheduling plan generation tool for the production plan to obtain the scheduling plan and scheduling data including the time span and the production time consumed by each product;

[0052] S6.3 The main agent outputs "Accepted", the scheduling is completed, and the scheduling plan is output.

[0053] 3) The insertion task is divided into the following sub-steps:

[0054] S6.1 The main intelligent agent calls the production plan verification (Check) tool to check whether the insertion order demand meets the current production conditions. If so, the original time point, product and quantity are directly handed over to the scheduling intelligent agent. Otherwise, the production plan generation (Plan) tool is used to change the quantity of the insertion order product to the maximum production quantity that can be met under the current resource conditions.

[0055] The S6.2 scheduling agent uses the scheduling scheme generation (Schedule) tool and passes in the parameters in step S6.1. It will keep the scheduling scheme before the insertion time point unchanged, add new jobs after the time point, and reschedule.

[0056] S6.3 The main agent outputs "Accepted", the scheduling is completed, and the scheduling plan is output.

[0057] 4) The order cancellation task is divided into the following sub-steps:

[0058] S6.1 The main intelligent agent outputs the time point, product and quantity of order cancellation.

[0059] The S6.2 scheduling agent uses the scheduling plan generation (Schedule) tool and passes in the parameters in step S6.1. It will keep the scheduling plan before the order cancellation time point unchanged, reduce the corresponding number of jobs after the time point, and reschedule.

[0060] S6.3 The main agent outputs "Accepted", the scheduling is completed, and the scheduling plan is output.

[0061] 5) The fault task is divided into the following sub-steps:

[0062] S6.1 The main intelligent agent outputs the faulty machine serial number, fault time, and repair time.

[0063] The S6.2 scheduling agent uses the scheduling scheme generation (Schedule) tool and inputs the parameters in step S6.1. It will truncate and postpone the jobs of the faulty machine during the fault time period, and adjust the jobs affected by the association.

[0064] S6.3 The main agent outputs "Accepted", the scheduling is completed, and the scheduling plan is output.

[0065] The present invention has the following characteristics and beneficial effects:

[0066] This invention adds consideration of resources such as people, machines, and materials to the traditional scheduling algorithm, ensuring that a feasible scheduling plan can be generated while optimizing. It can serve as a production AI assistant, continuously obtaining the production scheduling needs of production managers in the form of natural language, and promptly providing resource-feasible and time-feasible production plans and scheduling plans, thereby realizing real-time and flexible production scheduling management at the production site. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 This is a flow chart of a flexible workshop production scheduling method based on multi-agent collaboration according to an embodiment of the present invention;

[0068] Figure 2 This is a collaborative flow chart of various intelligent agents in an embodiment of the present invention;

[0069] Figure 3 This is a scheduling Gantt chart in an embodiment of the present invention. DETAILED DESCRIPTION

[0070] The present invention is described in detail below in conjunction with specific embodiments. The following examples will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other.

[0071] This paper provides a flexible workshop production scheduling method based on multi-agent collaboration, using a large language and large model. This scheduling method uses Python to develop the multi-agent collaborative system and various external tools. MySQL is used as the database for accessing production information, and ChatGLM3-6B is used as the large model. The present invention is further described below with reference to examples.

[0072] refer to Figure 1 and Figure 2 ,The flexible workshop production scheduling method based on multi-agent collaboration includes the following steps:

[0073] Step 1: Divide into stages

[0074] The multi-agent collaborative process for flexible workshop production scheduling is divided into:

[0075] Intention recognition stage (InterpretStage): Identify the intention of the production manager, and through the collaborative work of multiple intelligent agents, format the various recognition results and store them in the cache area. After completion, enter the next stage.

[0076] Information retrieval stage (InformationStage): Collect current production information, store it in the cache area, and enter the next stage after completion.

[0077] Scheduling negotiation stage (ScheduleStage): Based on the results of the intention recognition stage, the two agents are controlled to continuously negotiate to produce a feasible production plan and finally generate a scheduling solution.

[0078] In this embodiment, the basic attributes of each stage include: stage name, stage identifier, jump condition, included agents, the role and task of each agent in the stage, the workflow of the stage, and the goals to be achieved in the stage; the implementation method of the stage: read the predefined prompt in the configuration file and store it in the short-term memory of the large language model, which defines the background information of each stage.

[0079] Step 2: Create agents, including the main agent, task agent, time agent, quantity agent, preference agent, expedited agent, information agent, and scheduling agent, where:

[0080] MasterAgent: Responsible for the control transition between stages. During the intent recognition phase, it receives requests from production managers, integrates the intent recognition results into formatted data, and stores them in a public cache. During the scheduling negotiation phase, it generates production plans and verifies their feasibility.

[0081] Task Agent: Identifies task intent during the intent recognition phase and distinguishes whether the task is production, order insertion, order cancellation, or fault.

[0082] Time Agent: Identifies the time nodes of tasks during the intent recognition phase. It represents the production deadline for production tasks, the time point for inserting an order, the time point for canceling an order, and the time period of a failure for a failure task.

[0083] AmountAgent: Identifies the products that need to be produced and the quantity of each product during the intention recognition phase.

[0084] PretentAgent: Identifies task preferences during the intent recognition phase and distinguishes whether the task is time-prioritized or quantity-prioritized (preferences only apply to production tasks).

[0085] PriorityAgent: Identifies task priorities during the intent recognition phase and identifies tasks that need to be completed urgently.

[0086] InfoAgent: In the information retrieval stage, it calls an external tool set related to production information to collect production information and stores the retrieved information into the cache area one by one.

[0087] Schedule Agent: In the scheduling negotiation phase, it uses tools to obtain a scheduling solution according to the production plan.

[0088] In this embodiment, the basic attributes of the agent include: the name of the agent, the task of the agent, the rules that the agent should follow, the thinking chain for the agent to refer to when analyzing problems, output format and other restrictions.

[0089] Agent implementation: Reads predefined prompts from the configuration file and stores them in the short-term memory of the large language model. This defines the role each agent plays at the current stage. Program execution: Defines a global control object (SOP). Based on the execution flow, it controls the progress of the entire process and schedules different agents to perform tasks based on the current progress.

[0090]

[0091] Step 3: Create an external toolset

[0092] Build an external tool library and define multiple tools to interact with external databases or software system interfaces. The tools include: product information query (Product) tool, inventory query (Storage) tool, machine query (Machine) tool, manpower query (Human) tool, production plan verification (Check) tool, production plan generation (Plan) tool, and scheduling plan generation (Schedule) tool. Among them:

[0093] The product information query (Product) tool is used to query the machine process and material information of the product; the inventory query (Storage) tool is used to query the current material inventory information; the machine query (Machine) tool is used to query the current list of idle machines; the human query (Human) tool is used to query the current human resource situation; the production plan verification (Check) tool is used to receive the production plan as input, combine the current production resources, and output the judgment result of whether the production plan is feasible; the production plan generation (Plan) tool is used to output the maximum production quantity under the current production resource constraints; the scheduling plan generation (Schedule) tool is used to take the production plan as input and output a scheduling plan with time span priority based on tasks, preferences, and priorities.

[0094] In this embodiment, the basic attributes of the tool include: name, usage scenario, input parameters, and output parameters.

[0095] Tool implementation: Use a configuration file to import the tool's definition, including how to use it, when to use it, and what parameters are required. When the agent generates a "tool_call" request using the large model, the system calls the tool according to the requested tool name.

[0096] Step 4: Intent Recognition

[0097] Step 4.1: The main agent obtains the production manager's requirements in the form of natural language: "Tomorrow morning at 6 o'clock, 50 colorful sticks, 50 engine cylinder blocks, and 50 engine cylinder heads need to be produced" and forwards it to the task agent, time agent, quantity agent, preference agent, and expedited agent.

[0098] Step 4.2: The Task Agent identifies the task intent as "production." The Time Agent identifies the time parameter as "none." The Quantity Agent identifies the quantity parameters as "colorful sticks (50), engine block (50), and engine cylinder head (50). The Preference Agent identifies the task preference as "quantity." The Expedited Agent identifies the expedited task as "none." All data is formatted and stored in the cache.

[0099] Step 4.3: The main agent formats the recognition results and stores them in the buffer, outputs "Accepted", and enters the next stage.

[0100] Step 5: Information retrieval.

[0101] Step 5.1: Enter the state information retrieval phase and the information agent starts working.

[0102] Step 5.2: The system parses the product names from the cache: colorful stick, engine block, and engine cylinder head.

[0103] Step 5.3: The information agent executes a tool call (function_call) to query the product's material information and processing sequence, material inventory, machine information, and manpower information.

[0104]

[0105]

[0106] Step 5.4: The information agent stores this production information in the cache area as parameters for production scheduling and context information for generating natural language text.

[0107] Step 6: Scheduling Negotiation

[0108] Step 6.1: Entering the scheduling negotiation phase, the master agent retrieves the current demand intent information (production tasks, quantity priority) from the cache and invokes the production plan verification tool. It carefully checks whether material resources are sufficient, whether idle machines are sufficient, and whether human resources are sufficient. The current production plan indicates insufficient material resources.

[0109] Step 6.2: The main agent calls the production plan generation tool to query the production information such as products, processes, machines, etc. in the cache area, and obtains the maximum production quantity supported by the current resources: [{"colorful sticks": 34}, {"engine block": 8}, {"engine cylinder head": 22}], modifies the quantity of the cache area, and initiates a scheduling instruction to the scheduling agent.

[0110] The method to generate a production plan is as follows:

[0111]

[0112]

[0113]

[0114] Step 6.3: The scheduling agent receives and calls the scheduling plan generation tool (Schedule).

[0115]

[0116]

[0117] Step 6.4: The scheduling scheme generation tool calls the scheduling algorithm to obtain the processing sequence after the longest processing time (makespan) is optimized, and the shortest completion time makespan=53, and prints the scheduling Gantt chart (such as Figure 3 As shown in the figure): the horizontal axis is time (in minutes), the vertical axis is the machine number, the label contains the workpiece produced this time, and the color block in the figure represents the processing progress of the corresponding process of the current workpiece during this period: From the figure, you can see the work details of this scheduling and the time span of this processing, namely makspan.

[0118] Step 6.5: The main agent outputs "Accepted", and the scheduling ends. Returning to step 4, the main agent continues to monitor the next demand.

[0119] In summary, the flexible workshop production scheduling method based on multi-agent collaboration of the present invention adds consideration of resources such as people, machines, and materials to the traditional scheduling algorithm, ensures that a feasible scheduling plan can be generated while optimizing, and can continuously obtain the production scheduling needs of production managers in the form of natural language, and then provide timely and accurate information based on actual conditions. In the future, the system functions can be further expanded. For example, in a production workshop, workers can use voice commands to inquire about the current production progress, equipment status, or inventory information, and can handle emergencies such as production, order insertion, order cancellation, and fault repair in real time, and flexibly adjust the production scheduling plan. This real-time response not only saves workers time in looking for information, but also reduces the loss of production efficiency when production managers are busy with emergencies.

[0120] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A flexible workshop production scheduling method based on multi-agent collaboration is characterized by: The steps include: Step 1: Divide the multi-agent collaborative process for flexible workshop production scheduling into the intention recognition stage, information retrieval stage, and scheduling negotiation stage; Step 2: Create the main agent, task agent, time agent, quantity agent, preference agent, expedited agent, information agent, and scheduling agent; Step 3: Build an external tool library, define multiple tools, and interface with external databases or software systems for data interaction; Step 4: After the main agent obtains the production manager's production scheduling requirements in the form of natural language, it enters the intent recognition stage. The agent related to the intent recognizes the production requirements, formats the recognition results, and stores them in the cache. The step 4 specifically includes the following steps: S4.1 uses task agents to identify specific production scheduling tasks corresponding to production managers' needs, including production, order insertion, order cancellation, and failure. A typical production task is identified as "production," while the production of a specific number of products at a given time point is identified as "order insertion," while the reduction of a specific number of products is identified as "order cancellation," and the suspension of a machine for some reason is identified as "failure." S4.2 Identify the time and quantity parameters of production through time agents and quantity agents; S4.3 uses a preference agent to identify the preference for selecting demand between "time priority" and "quantity priority". Producing as many target products as possible within a specific time is identified as "time priority", and requiring the production of a specific number of target products is identified as "quantity priority". S4.4 uses the expedited agent to identify the expedited jobs in the demand and increase the priority of a job; Step 5: Entering the information retrieval phase, the information agent uses tools to collect production information and stores it in the cache; Step 6: Enter the scheduling negotiation phase. The master agent and the scheduling agent conduct multiple rounds of negotiations. In each round, the master agent outputs a production plan based on the results of the intention recognition phase. The scheduling agent uses the scheduling tool to generate a scheduling solution based on the master agent's production plan. If the generated production plan or scheduling solution does not meet the requirements, the next round of negotiation begins. Step 7: The main agent continues to wait for new production scheduling requirements from the production manager, and proceeds to step 4 after receiving the new requirements.

2. The flexible workshop production scheduling method based on multi-agent collaboration according to claim 1 is characterized in that: The intention recognition stage is used to identify the intention of the production manager. Through the collaborative work of multiple intelligent agents, the various recognition results are formatted and stored in the cache area. After completion, it enters the next stage; The information retrieval phase is used to collect current production information, store it in the cache, and then proceed to the next phase after completion; The scheduling negotiation phase is used to control the two agents to continuously negotiate to produce a feasible production plan based on the results of the intention recognition phase, and finally generate a scheduling solution.

3. The flexible workshop production scheduling method based on multi-agent collaboration according to claim 1 is characterized in that: In step 2: The master agent is responsible for the control transition of the phases; it receives the needs of the production manager during the intention recognition phase, integrates the results of the intention recognition into formatted data, and stores them in the public buffer area; Generate production plans and verify the feasibility of production plans during the scheduling negotiation stage; The task agent is used to identify the task intent in the intent recognition phase and distinguish whether the task is production, order insertion, order cancellation, or fault; The time agent is used to identify the time nodes of the task during the intention recognition phase. In production tasks, this time node represents the production deadline; in order insertion tasks, it represents the time point of order insertion; in order cancellation tasks, it represents the time point of order cancellation; in fault tasks, it represents the time period of the fault; The quantity agent is used to identify the products that need to be produced and the quantity of each product during the intention recognition phase; The preference agent is used to identify task preferences during the intention recognition phase and distinguish whether the task is time-prioritized or quantity-prioritized; The expedited agent is used to identify task priorities during the intent recognition phase and identify tasks that need to be completed urgently. The information agent is used to call the external tool set related to production information to collect production information during the information retrieval phase, and store the retrieved information into the cache area one by one; The scheduling agent is used to obtain a scheduling solution using tools according to the production plan during the scheduling negotiation phase.

4. The flexible workshop production scheduling method based on multi-agent collaboration according to claim 1 is characterized in that: The multiple tools include: Product information query tool, used to query the product's machine process and material information; Inventory query tool, used to query current material inventory information; Machine query tool, used to query the current list of idle machines; Human resource query tool, used to query the current human resource situation; The production plan verification tool is used to receive the production plan as input, combine it with the current production resources, and output the judgment result on whether the production plan is feasible; Production plan generation tool, used to output the maximum production quantity under the current production resource constraints; The scheduling plan generation tool is used to take the production plan as input and output a time-span-priority scheduling plan based on tasks, preferences, and priorities.

5. The flexible workshop production scheduling method based on multi-agent collaboration according to claim 1 is characterized in that: The step 5 specifically includes the following steps: Through the information agent, it calls external tools related to information query to obtain production information from external databases or software interfaces, including product materials, processing process information, material inventory information, idle machine information, and human resource information. The obtained information is formatted and stored in the cache area, which serves as background knowledge and preset conditions for other agents to use when checking the feasibility of production plans, generating production plans, and generating scheduling plans.

6. The flexible workshop production scheduling method based on multi-agent collaboration according to claim 1 is characterized in that: In step 6, there are different negotiation methods for different production intentions: 1) The time-priority production task is divided into the following sub-steps: S6.1 The main agent calls the production plan generation tool to provide the maximum production quantity of the corresponding product under the current resource conditions; S6.2 The scheduling agent calls the scheduling plan generation tool for the production plan to obtain a scheduling plan, and obtains the time span and the production time consumed by each product; S6.3 The main agent checks the scheduling data and outputs if the time span is within the specified time. "Accepted"; otherwise, reduce the quantity of each product proportionally based on the production time of each product; S6.4 The scheduling agent observes the output of the master agent. If it is "Accepted", the scheduling ends and the scheduling plan is output; otherwise, repeat step S6.2; 2) Quantity-priority production tasks are divided into the following sub-steps: S6.1 The master agent uses the production plan verification tool to check whether the production plan proposed by the production manager is feasible. If so, it outputs the original production plan. Otherwise, it uses the production plan generation tool to generate the maximum production quantity that meets the current resource conditions. S6.2 The scheduling agent identifies the output of the main agent and calls a scheduling plan generation tool for the production plan to obtain a scheduling plan and scheduling data including the time span and the production time consumed by each product; S6.3 The master agent outputs "Accepted", the scheduling is completed, and the scheduling plan is output; 3) The insertion task is divided into the following sub-steps: S6.1 The master agent calls the production plan verification tool to check whether the insertion order requirement meets the current production conditions. If so, it directly passes the original time point, product, and quantity to the scheduling agent. Otherwise, it uses the production plan generation tool to change the quantity of the inserted product to the maximum production quantity that can be met under the current resource conditions. S6.2 The scheduling agent uses the scheduling plan generation tool, inputs the parameters in step S6.1, keeps the scheduling plan before the insertion time point unchanged, adds the new job after the time point, and reschedules; S6.3 The master agent outputs "Accepted", the scheduling is completed, and the scheduling plan is output; 4) The order cancellation task is divided into the following sub-steps: S6.1 The main agent outputs the time point, product, and quantity of the order cancellation; S6.2 The scheduling agent uses the scheduling plan generation tool, inputs the parameters in step S6.1, keeps the scheduling plan before the order cancellation time point unchanged, reduces the corresponding number of jobs after the time point, and reschedules; S6.3 The master agent outputs "Accepted", the scheduling is completed, and the scheduling plan is output; 5) The fault task is divided into the following sub-steps: S6.1 The main agent outputs the faulty machine serial number, fault time, and repair time; S6.2 The scheduling agent uses the scheduling solution generation tool, inputs the parameters in step S6.1, truncates and postpones the jobs of the faulty machine during the fault time period, and adjusts the jobs affected by the associated tasks. S6.3 The main agent outputs "Accepted", the scheduling is completed, and the scheduling plan is output.

Images