A Production Scheduling and Collaborative Control Method and System Based on AI Full-Stack Intelligent Agents

Abstract

This invention relates to the field of industrial intelligent technology, specifically to a production scheduling and collaborative control method and system based on an AI full-stack intelligent agent. The method includes: deconstructing production orders through an order parsing intelligent agent to obtain structured order requirements; a collaborative scheduling intelligent agent dividing sub-tasks and distributing them to corresponding workshop intelligent agents, generating departmental task commitments based on capacity and material data, and obtaining a global collaborative scheduling plan through conflict resolution and orchestration; an instruction generation intelligent agent completing step-level decoding and issuing the processing instruction sequence to the manufacturing execution system; a status monitoring intelligent agent detecting production progress timeline deviations and generating deviation alarms; and the status monitoring intelligent agent feeding back the deviation alarms to the collaborative scheduling intelligent agent, which, based on the latest real-time production status data received from the manufacturing execution system interface, adaptively reschedules unfinished production tasks in the production orders to obtain a rescheduled production order plan. This invention can improve the efficiency of production scheduling collaboration.

Description

Technical Field

[0001] This invention relates to the field of industrial intelligent technology, and in particular to a production scheduling and collaborative control method and system based on an AI full-stack intelligent agent. Background Technology

[0002] Current production scheduling relies on manual experience and traditional rule-based management models. Order parsing requires manual structured conversion, which cannot efficiently deconstruct unstructured order information. Task granularity division and workshop allocation lack intelligent collaborative mechanisms, easily leading to sub-task matching deviations. Workshop capacity and material inventory data cannot be linked and shared in real time. Departmental task commitments are generated late and lack accuracy. Conflicts such as process dependencies and resource competition are difficult to resolve automatically. Overall scheduling plan preparation is inefficient and has poor adaptability.

[0003] Production instructions require manual configuration of step parameters, making automatic decoding and rapid issuance at the step level impossible. Production status can only be passively collected, failing to detect progress timing deviations in real time, resulting in slow response to deviation alarms and rescheduling. The traditional model lacks full-stack intelligent agent collaboration support, resulting in a lack of closed-loop control over the entire production scheduling process and weak dynamic adaptive adjustment capabilities. It cannot meet the real-time scheduling needs of complex production scenarios. Therefore, improving the efficiency of production scheduling collaborative control has become an urgent problem to be solved. Summary of the Invention

[0004] This invention provides a production scheduling and collaborative control method and system based on AI full-stack intelligent agents to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a production scheduling and collaborative control method based on an AI full-stack intelligent agent, comprising:

[0006] A1. Through the AI ​​full-stack intelligent agent for order parsing, the received production orders are semantically deconstructed to obtain the structured order requirements of the production orders;

[0007] A2. The structured order requirements are sent to the collaborative scheduling agent in the AI ​​full-stack intelligent agent to divide the tasks into granularities and obtain the sub-task requirements of the production order.

[0008] A3. The collaborative scheduling agent distributes the sub-task requirements to the corresponding workshop agents and, in conjunction with the workshop agents' capacity data and material inventory data, generates departmental task commitments for the sub-task requirements.

[0009] A4. Based on preset process dependency constraints and resource competition constraints, conflict resolution and arrangement of departmental task commitments are carried out to obtain a global collaborative scheduling plan for production orders;

[0010] A5. Send the global collaborative scheduling plan to the AI ​​full-stack intelligent agent to perform step-level decoding, obtain the processing instruction sequence of the production order, and send the processing instruction sequence to the manufacturing execution system interface;

[0011] A6. Through the status monitoring agent in the AI ​​full-stack intelligent agent, the real-time production status data received by the manufacturing execution system interface is compared with the planned progress in the global collaborative scheduling plan to detect the time sequence deviation and generate deviation alarms for production orders.

[0012] A7. The status monitoring agent feeds back deviation alarms to the collaborative scheduling agent. Based on the latest real-time production status data received from the manufacturing execution system interface, it adaptively reschedules the unfinished production tasks in the production order to obtain the rescheduling plan for the production order.

[0013] In a preferred embodiment, the order parsing agent, which utilizes a full-stack AI agent, performs semantic deconstruction on the received production orders to obtain structured order requirements, including:

[0014] The order parsing agent, a full-stack AI agent, performs word segmentation on the text data of production orders to obtain the word segmentation sequence of production orders.

[0015] The word segmentation sequence of orders is grammatically categorized to obtain the part-of-speech tags for production orders;

[0016] Extract product name terms, quantity terms, and date terms from the part-of-speech tagging words. Treat product name terms as product type entities, quantity terms as delivery quantity entities, and date terms as expected delivery time entities.

[0017] The order parsing agent calls the order data structure associated with the product type entity, assigns the delivery quantity entity to the quantity field in the order data structure, and assigns the expected delivery time entity to the delivery date field in the order data structure to generate a structured order demand record for the production order.

[0018] In a preferred embodiment, the step of sending the structured order requirements to the collaborative scheduling agent in the AI ​​full-stack agent for task granularity division to obtain the sub-task requirements of the production order includes:

[0019] In the AI ​​full-stack intelligent agent, the collaborative scheduling intelligent agent receives structured order requirements and parses the product type field value, quantity field value, and delivery date field value from the structured order requirements as parameters for the task division of production orders;

[0020] The process decomposition template associated with the product type field value is retrieved from the pre-stored process template library, and the names of each production stage and the corresponding standard working hours of each production stage are extracted from the process decomposition template to generate the initial task node list of the production order.

[0021] Based on the quantity field value and the delivery date field value, time windows are allocated to the task nodes in the initial task node list to obtain the sub-task requirements of the production order.

[0022] In a preferred embodiment, the collaborative scheduling agent distributes sub-task requirements to corresponding workshop agents, and combines the workshop agents' capacity data and material inventory data to generate departmental task commitments for the sub-task requirements, including:

[0023] Data is extracted from each subtask node in the subtask requirements to obtain the product type identifier and quantity field value of each subtask node.

[0024] Based on the product type identifier, workshop matching is performed on each sub-task node to obtain the distribution mapping relationship table of each sub-task node;

[0025] According to the distribution mapping table, the collaborative scheduling agent sends each subtask node to the corresponding workshop agent and receives the current available capacity value and current available material quantity returned by the workshop agent to obtain the workshop real-time status data of the workshop agent;

[0026] Based on the quantity field value and real-time status data of the workshop, the commitment amount of each sub-task node is calculated to obtain the departmental task commitment required by the sub-task.

[0027] In a preferred embodiment, the formula for calculating the promised delivery time offset in the departmental task commitment is as follows:

[0028] ;

[0029] In the formula, For the first Promised delivery time offset for each subtask node For the first Each sub-task node corresponds to the current available production capacity value of the workshop intelligent agent. For the first The value of the field representing the number of subtask nodes. For the first Each sub-task node corresponds to the current available material quantity of the workshop agent. For the first Each sub-task node corresponds to the unit material consumption of the product.

[0030] In a preferred embodiment, the step of resolving and orchestrating conflicts in departmental task commitments based on preset process dependency constraints and resource competition constraints to obtain a global collaborative scheduling plan for production orders includes:

[0031] The elements of the departmental task commitments are deconstructed to obtain the set of task attributes to be arranged for the departmental task commitments;

[0032] Based on the preset process dependency constraints, the predecessor and successor relationships of each task in the task attribute set to be arranged are retrieved to obtain the process disorder pairs of the production order.

[0033] The promised delivery time of subsequent tasks in the process disorder pairing is postponed until the process dependency constraint is met, so as to obtain the process alignment task set of the production order.

[0034] Based on the preset resource competition constraints, time window overlap detection is performed on multiple tasks that share the same workshop agent for process alignment tasks to obtain the resource preemption group of production orders;

[0035] The promised delivery time of tasks with lower priority in the resource preemption group is postponed until the resource occupation time windows no longer overlap, so as to obtain a resource-free task set for the production order.

[0036] The promised delivery time of each task in the resource-free task set is sorted and mapped with the corresponding product type identifier and promised delivery quantity on the timeline to generate a global collaborative scheduling plan for production orders.

[0037] In a preferred embodiment, the step of sending the global collaborative scheduling plan to the instruction generation agent in the AI ​​full-stack agent for step-level decoding to obtain the processing instruction sequence of the production order, and then sending the processing instruction sequence to the manufacturing execution system interface, includes:

[0038] The instructions sent to the AI ​​full-stack intelligent agent from the global collaborative scheduling plan generate an intelligent agent, and extract the product type identifier, the promised delivery quantity, and the start and end values ​​of the time window for each task item in the global collaborative scheduling plan.

[0039] The process route template associated with the product type identifier is retrieved from the pre-stored process route library, and the process route template encapsulates the process step sequence corresponding to each task item and the operation code corresponding to each process step into a process step-level decoding mapping table for the production order.

[0040] Based on the step-level decoding mapping table and the start and end values ​​of the time window, the process parameters of the corresponding step sequence of each task item are fused to obtain the processing instruction record of each task item.

[0041] The processing instruction records are serialized and combined to obtain the processing instruction sequence of the production order, and the processing instruction sequence is sent to the manufacturing execution system interface.

[0042] In a preferred embodiment, the step of using the state monitoring agent within the AI ​​full-stack agent to detect the time-series offset between the real-time production status data received from the manufacturing execution system interface and the planned progress in the global collaborative scheduling plan, in order to generate a deviation alarm for the production order, includes:

[0043] Receive real-time production status data returned by the manufacturing execution system interface, and read the start and end timestamps of each task item in the global collaborative scheduling plan to obtain the real-time status set and planned time baseline of the production order;

[0044] By using the state monitoring agent in the AI ​​full-stack agent, the actual start and end timestamps of each task item in the real-time state set are periodically checked against the planned start and end timestamps of the corresponding task items in the planned time baseline to obtain the time-series offset task set of the production order.

[0045] By using a state monitoring agent, violations are identified in the time-series offset task set, and deviation alarms for production orders are obtained.

[0046] In a preferred embodiment, the status monitoring agent feeds back deviation alarms to the collaborative scheduling agent, and adaptively reschedules incomplete production tasks in the production order based on the latest real-time production status data received from the manufacturing execution system interface, thereby obtaining a rescheduling plan for the production order, including:

[0047] The status monitoring agent sends deviation alarms to the collaborative scheduling agent for alarm semantic decoding to obtain the set of rescheduling trigger conditions for production orders;

[0048] The collaborative scheduling agent obtains the latest real-time production status data from the manufacturing execution system interface, and extracts the actual completion progress and remaining workload of unfinished tasks from the latest real-time production status data to obtain a snapshot of the status of unfinished tasks in the production order.

[0049] Based on the rescheduling trigger condition set, and with the snapshot of the incomplete task status as the initial state, the task granularity division and distribution process of the collaborative scheduling agent is invoked again to obtain the rescheduling plan for production orders.

[0050] To address the aforementioned problems, this invention also provides a production scheduling and collaborative control system based on an AI full-stack intelligent agent, the system comprising:

[0051] The order parsing module uses an AI full-stack intelligent agent to perform semantic deconstruction on the received production orders, thereby obtaining the structured order requirements of the production orders.

[0052] The task splitting module sends the structured order requirements to the collaborative scheduling agent in the AI ​​full-stack intelligent agent for task granularity division, thereby obtaining the sub-task requirements of the production order;

[0053] The commitment generation module coordinates the dispatching agent to distribute the sub-task requirements to the corresponding workshop agents, and generates departmental task commitments for the sub-task requirements by combining the workshop agents' capacity data and material inventory data.

[0054] The conflict resolution module, based on preset process dependency constraints and resource competition constraints, performs conflict resolution and orchestration on departmental task commitments to obtain a global collaborative scheduling plan for production orders;

[0055] The instruction decoding module sends the global collaborative scheduling plan to the instruction generation agent in the AI ​​full-stack intelligent agent for step-level decoding to obtain the processing instruction sequence of the production order, and then sends the processing instruction sequence to the manufacturing execution system interface;

[0056] The offset detection module, through the status monitoring agent in the AI ​​full-stack intelligent agent, performs time-series offset detection between the real-time production status data received from the manufacturing execution system interface and the planned progress in the global collaborative scheduling plan, so as to generate deviation alarms for production orders.

[0057] The rescheduling module, where the status monitoring agent feeds back deviation alarms to the collaborative scheduling agent, adaptively reschedules the unfinished production tasks in the production order based on the latest real-time production status data received from the manufacturing execution system interface, thus obtaining the rescheduling plan for the production order.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] 1. This invention utilizes an AI-powered full-stack intelligent agent for order parsing to deconstruct the semantics of production orders, automatically generating structured order requirements. This enables standardized extraction and transformation of order information, improving the accuracy and efficiency of order requirement processing. A collaborative scheduling intelligent agent intelligently divides tasks at the granular level, accurately distributing sub-tasks to corresponding workshop agents. Combined with real-time workshop capacity and material inventory data, departmental task commitments are generated, ensuring precise matching between task allocation and production resource status. Based on process dependency constraints and resource competition constraints, conflict resolution orchestration is completed, constructing a global collaborative scheduling plan. This achieves orderly arrangement of production tasks and rational allocation of resources, ensuring the scheduling plan possesses global adaptability and execution feasibility.

[0060] 2. This invention utilizes an instruction generation agent to complete step-by-step decoding, automatically generating processing instruction sequences and directly connecting to the manufacturing execution system interface, enabling rapid issuance and precise implementation of production instructions. A status monitoring agent detects real-time production progress deviations, generating timely deviation alarms to ensure real-time alignment between production status and planned progress. An adaptive rescheduling mechanism triggers the deviation alarm feedback mechanism, dynamically adjusting incomplete tasks based on the latest production status, forming a closed-loop control system for the entire production scheduling process. This entire solution achieves intelligent collaboration and dynamic optimization across the entire production scheduling chain, significantly improving production scheduling execution efficiency, strengthening real-time control capabilities of the production process, ensuring efficient progress of production tasks, and continuously optimizing overall production operation efficiency. Attached Figure Description

[0061] Figure 1 A flowchart illustrating a production scheduling and collaborative control method based on an AI full-stack intelligent agent provided in an embodiment of the present invention;

[0062] Figure 2 A functional block diagram of a production scheduling and collaborative control system based on an AI full-stack intelligent agent provided in an embodiment of the present invention;

[0063] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0064] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0065] This application provides a production scheduling and collaborative control method based on an AI full-stack intelligent agent. The executing entity of this AI full-stack intelligent agent-based production scheduling and collaborative control method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the AI ​​full-stack intelligent agent-based production scheduling and collaborative control method can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0066] Reference Figure 1The diagram shown is a flowchart illustrating a production scheduling and collaborative control method based on an AI full-stack intelligent agent according to an embodiment of the present invention. In this embodiment, the production scheduling and collaborative control method based on an AI full-stack intelligent agent includes:

[0067] A1. Through the AI ​​full-stack intelligent agent for order parsing, the received production orders are semantically deconstructed to obtain the structured order requirements of the production orders;

[0068] In this embodiment of the invention, the order parsing agent, which uses an AI full-stack intelligent agent, performs semantic deconstruction on the received production orders to obtain structured order requirements, including:

[0069] The order parsing agent, a full-stack AI agent, performs word segmentation on the text data of production orders to obtain the word segmentation sequence of production orders.

[0070] The word segmentation sequence of orders is grammatically categorized to obtain the part-of-speech tags for production orders;

[0071] Extract product name terms, quantity terms, and date terms from the part-of-speech tagging words. Treat product name terms as product type entities, quantity terms as delivery quantity entities, and date terms as expected delivery time entities.

[0072] The order parsing agent calls the order data structure associated with the product type entity, assigns the delivery quantity entity to the quantity field in the order data structure, and assigns the expected delivery time entity to the delivery date field in the order data structure to generate a structured order demand record for the production order.

[0073] The AI ​​full-stack intelligent agent for order parsing calls a pre-built vocabulary library specifically for the manufacturing industry. The vocabulary library fully includes standard words such as product names, quantity units, date expressions, and process names in the production scenario. Using complete words in the vocabulary library as the smallest splitting unit, the continuous text data of the production order is split in a targeted manner. During the splitting process, punctuation marks and redundant characters without actual semantics are automatically removed. After the splitting is completed, all valid independent words are arranged strictly in the original order of the production order text to form the order word segmentation sequence of the production order.

[0074] According to the predefined grammatical category annotation rules for production order text, the rules clearly define the specific judgment conditions for grammatical attributes such as nouns, numerals, and time nouns. The order parsing agent reads each independent word unit in the order word segmentation sequence in turn, and matches a unique corresponding grammatical attribute identifier for each word unit according to the judgment conditions. After completing the matching of grammatical attribute identifiers for all word units, the words with identifiers are combined in the original order to form the part-of-speech tag words for the production order.

[0075] Based on the pre-defined core entity extraction rules for production orders, the rules specify that product name terms must match entries in the product name vocabulary library, quantity terms must contain specific numerical values ​​and corresponding units of measurement, and date terms must conform to the standard expression format of year, month, and day. From the part-of-speech tagging words, terms that match the product name category are selected as product name terms, terms that match the quantity category are selected as quantity terms, and terms that match the date category are selected as date terms. Product name terms are directly identified as product type entities, quantity terms are directly identified as delivery quantity entities, and date terms are directly identified as expected delivery time entities.

[0076] The order parsing intelligent agent performs a one-to-one precise match between product type entities and a pre-stored order data structure library. Each product type in the structure library corresponds to a unique order data structure. After a successful match, the dedicated order data structure is retrieved, and the entire contents of the delivery quantity entity are completely filled into the quantity field of the order data structure. The entire contents of the expected delivery time entity are completely filled into the delivery date field of the order data structure. The order data structure after the fields are filled is the structured order requirement record of the production order.

[0077] The beneficial effects are that this process automatically converts unstructured production order text into standardized data through standardized vocabulary splitting, precise grammatical annotation, targeted entity extraction, and structured data encapsulation. The information extraction is complete and without omissions, the matching is error-free, the data format is unified and standardized, and no manual intervention is required throughout the process. This significantly improves the accuracy and efficiency of order parsing, provides stable and standardized data support for subsequent production scheduling task division, and ensures the reliability and consistency of data processing in the early stage of production scheduling.

[0078] A2. The structured order requirements are sent to the collaborative scheduling agent in the AI ​​full-stack intelligent agent to divide the tasks into granularities and obtain the sub-task requirements of the production order.

[0079] In this embodiment of the invention, the step of sending the structured order requirements to the collaborative scheduling agent in the AI ​​full-stack agent for task granularity division to obtain the sub-task requirements of the production order includes:

[0080] In the AI ​​full-stack intelligent agent, the collaborative scheduling intelligent agent receives structured order requirements and parses the product type field value, quantity field value, and delivery date field value from the structured order requirements as parameters for the task division of production orders;

[0081] The process decomposition template associated with the product type field value is retrieved from the pre-stored process template library, and the names of each production stage and the corresponding standard working hours of each production stage are extracted from the process decomposition template to generate the initial task node list of the production order.

[0082] Based on the quantity field value and the delivery date field value, time windows are allocated to the task nodes in the initial task node list to obtain the sub-task requirements of the production order.

[0083] In the AI ​​full-stack intelligent agent, the collaborative scheduling agent receives the structured order demand record generated by the preceding steps. The product type field, quantity field, and delivery date field in the structured order demand record all have fixed storage locations. The collaborative scheduling agent directly reads all the stored content of the product type field to obtain the product type field value, reads all the stored content of the quantity field to obtain the quantity field value, and reads all the stored content of the delivery date field to obtain the delivery date field value, according to the preset field positioning rules. The product type field value, quantity field value, and delivery date field value are jointly determined as the task division parameters for the production order.

[0084] The pre-stored process template library is classified and archived according to product type. Each process decomposition template is bound to a unique corresponding product type identifier. The collaborative scheduling agent performs a character-by-character complete match comparison between the product type field value in the task division basis parameter and the product type identifier of all process decomposition templates in the process template library. When the matching result is completely consistent, the corresponding process decomposition template is immediately retrieved. The collaborative scheduling agent extracts all production stage names one by one according to the preset record order in the process decomposition template. At the same time, it extracts the standard working time value corresponding to each production stage. The extracted production stage names and corresponding standard working time values ​​are combined and organized according to the actual execution order of the process to generate the initial task node list of the production order.

[0085] The collaborative scheduling agent sets the delivery date field value as the final time node that all production tasks must complete, and sets the quantity field value as the basis for the total workload that this production task needs to complete. Combining the standard working hours of each task node in the initial task node list with the fixed execution order, it calculates the execution period of each task node sequentially from the final time node backwards, and determines a unique start execution time and end execution time for each task node. After all task nodes have completed the allocation of execution time ranges, they are integrated to form the sub-task requirements of the production order.

[0086] The beneficial effects are that the implementation process can accurately extract core parameters of orders by locating fixed fields, achieve unique matching and calling based on the classified and archived process template library, complete the task time window allocation according to fixed time sequence and workload, and the execution rules are clear and reproducible throughout the process. The generated initial task node list and sub-task requirements are in a unified format and have accurate content, providing a standardized task basis for subsequent workshop task distribution and collaborative scheduling, and ensuring the standardization and feasibility of production task division.

[0087] A3. The collaborative scheduling agent distributes the sub-task requirements to the corresponding workshop agents and, in conjunction with the workshop agents' capacity data and material inventory data, generates departmental task commitments for the sub-task requirements.

[0088] In this embodiment of the invention, the collaborative scheduling agent distributes sub-task requirements to the corresponding workshop agents, and combines the workshop agents' capacity data and material inventory data to generate departmental task commitments for the sub-task requirements, including:

[0089] Data is extracted from each subtask node in the subtask requirements to obtain the product type identifier and quantity field value of each subtask node.

[0090] Based on the product type identifier, workshop matching is performed on each sub-task node to obtain the distribution mapping relationship table of each sub-task node;

[0091] According to the distribution mapping table, the collaborative scheduling agent sends each subtask node to the corresponding workshop agent and receives the current available capacity value and current available material quantity returned by the workshop agent to obtain the workshop real-time status data of the workshop agent;

[0092] Based on the quantity field value and real-time status data of the workshop, the commitment amount of each sub-task node is calculated to obtain the departmental task commitment required by the sub-task.

[0093] The formula for calculating the delivery time offset that can be committed to the departmental task commitment is as follows:

[0094] ;

[0095] In the formula, For the first Promised delivery time offset for each subtask node For the first Each sub-task node corresponds to the current available production capacity value of the workshop intelligent agent. For the first The value of the field representing the number of subtask nodes. For the first Each sub-task node corresponds to the current available material quantity of the workshop agent. For the first Each sub-task node corresponds to the unit material consumption of the product.

[0096] The collaborative scheduling agent performs data stripping operations on each subtask node in the subtask requirements according to the preset subtask node data filtering rules. The stripping operation retains only the preset product type identifier and quantity field value in the subtask node, and removes unnecessary data such as time window and process name in the subtask node. After completing the data stripping of all subtask nodes, the product type identifier and quantity field value corresponding to each subtask node are obtained.

[0097] The collaborative scheduling agent calls the pre-stored database of product type and production workshop correspondences, and performs a complete match between the product type identifier of each sub-task node and the identifier in the database. After a successful match, the dedicated production workshop corresponding to the sub-task node is determined. The sub-task node, product type identifier, and corresponding production workshop are recorded one by one in a fixed format and organized into a distribution mapping table for each sub-task node.

[0098] Based on the correspondence between sub-task nodes and production workshops recorded in the distribution mapping table, the collaborative scheduling agent sends each sub-task node to the corresponding workshop agent that has been matched. After receiving the sub-task node, each workshop agent collects its own current production capacity data in real time and feeds it back as the current available production capacity value. It also collects its own current available material reserve data in real time and feeds it back as the current available material quantity. The collaborative scheduling agent integrates the current available production capacity value and current available material quantity returned by all workshop agents to form the workshop real-time status data of the workshop agents.

[0099] The collaborative scheduling agent uses the quantity field value of each sub-task node as the benchmark for task requirements. It combines the current available capacity value in the workshop's real-time status data to determine the scale of tasks that the workshop can undertake, and combines the current available material quantity in the workshop's real-time status data to determine the total production volume that the workshop can support. It matches the scale of tasks that can be undertaken with the total production volume that can be supported to determine the final commitment to complete the sub-task node. After integrating the commitment to complete all sub-task nodes, it obtains the departmental task commitment for the sub-task requirements.

[0100] The departmental task commitment includes a committable delivery time offset, which is the [number]th [unit]. The delivery time adjustment value for each sub-task node is obtained through specified calculation logic. It is used to clarify the deviation of the delivery time of the sub-task node from the planned time. The numerical result directly serves as the core basis for the verification of delivery time in the department's task commitment.

[0101] The current available capacity value is from the first Each subtask node corresponds to a workshop agent that collects and generates data in real time. The workshop agent continuously monitors its own production equipment operating status, the number of production personnel, the effective production shift duration, and other production conditions. After integrating all effective resources that can be invested in the production of the subtask, it calculates the actual production capacity value that can be executed for the current subtask.

[0102] The quantity field value comes from the extraction result of the data stripping stage of the preceding subtask node. It is the total production task demand obtained by the collaborative scheduling agent from the subtask nodes of the subtask requirements. This value directly reflects the fixed production demand standard of the subtask node.

[0103] The current available material quantity is from the first Each subtask node corresponds to a real-time inventory count generated by the workshop intelligent agent. The workshop intelligent agent checks the inventory quantities of raw materials, semi-finished products, and supporting materials one by one, and after removing materials that have been occupied or consumed, obtains the material reserve value that can be used for the production of the subtask.

[0104] The unit material consumption is determined based on the historical production data and standard process documents of the corresponding product. By statistically analyzing the average material consumption per unit in multiple actual productions of the product, and combining it with the material consumption quota per unit specified in the standard process documents, a fixed unit product material consumption value is comprehensively determined.

[0105] When calculating the promised delivery time offset, the quantity field value is first doubled, then the current available material quantity is divided by the unit material consumption to obtain the total production that the material can support. The difference between production demand and material support is obtained by subtracting the total production that the material can support from the doubled quantity field value. Finally, the difference is divided by the current available capacity value to obtain the promised delivery time offset.

[0106] This calculation logic is highly correlated with subtask node processing, workshop status acquisition, and departmental task commitment generation. The resulting promised delivery time offset is directly used to correct the delivery time of departmental tasks, ensuring that the promised delivery time matches the actual workshop capacity and material reserve conditions. This provides accurate time adjustment data for conflict resolution and orchestration of subsequent global collaborative scheduling plans.

[0107] The beneficial effects of this implementation process are that it extracts core task information through standardized data stripping, completes the precise matching of tasks and workshops based on fixed correspondences, collects real-time data on actual workshop capacity and materials, verifies the committed quantity in combination with task requirements, and generates a distribution mapping table that matches the department's task commitments with the actual production conditions of the workshop. The data is authentic and traceable, providing accurate and reliable basic data support for subsequent production scheduling conflict resolution and planning.

[0108] A4. Based on preset process dependency constraints and resource competition constraints, conflict resolution and arrangement of departmental task commitments are carried out to obtain a global collaborative scheduling plan for production orders;

[0109] In this embodiment of the invention, the step of resolving and arranging conflict resolution among departmental task commitments based on preset process dependency constraints and resource competition constraints to obtain a global collaborative scheduling plan for production orders includes:

[0110] The elements of the departmental task commitments are deconstructed to obtain the set of task attributes to be arranged for the departmental task commitments;

[0111] Based on the preset process dependency constraints, the predecessor and successor relationships of each task in the task attribute set to be arranged are retrieved to obtain the process disorder pairs of the production order.

[0112] The promised delivery time of subsequent tasks in the process disorder pairing is postponed until the process dependency constraint is met, so as to obtain the process alignment task set of the production order.

[0113] Based on the preset resource competition constraints, time window overlap detection is performed on multiple tasks that share the same workshop agent for process alignment tasks to obtain the resource preemption group of production orders;

[0114] The promised delivery time of tasks with lower priority in the resource preemption group is postponed until the resource occupation time windows no longer overlap, so as to obtain a resource-free task set for the production order.

[0115] The promised delivery time of each task in the resource-free task set is sorted and mapped with the corresponding product type identifier and promised delivery quantity on the timeline to generate a global collaborative scheduling plan for production orders.

[0116] The collaborative scheduling agent extracts the product type identifier, deliverable quantity, deliverable time, corresponding execution workshop agent, and related process information of all tasks in the departmental task commitment according to the preset task element extraction specifications. The core elements of all tasks after decomposition are integrated into a single task unit to form the task attribute set to be arranged in the departmental task commitment.

[0117] The collaborative scheduling agent calls the pre-stored production process dependency constraint file. The constraint file clearly stipulates the execution rules that each process must follow before the successor task can be started after the predecessor task is completed. The execution order of each task in the task attribute set to be orchestrated is checked one by one with the predecessor and successor relationship. Task combinations that start the successor task without meeting the predecessor task completion conditions are screened out to form the process disorder pairs of the production order.

[0118] The collaborative scheduling agent, for the subsequent tasks in the process disorder pair, pushes their promised delivery time forward one by one according to the preset unit production time. After each push, it checks the compliance against the process dependency constraints until the start time of the subsequent task is later than the completion time of the predecessor task and fully complies with the process dependency constraints. After integrating all the qualified tasks, the process alignment task set of the production order is obtained.

[0119] Based on preset resource competition constraints, the collaborative scheduling agent can only execute one production task within the same workshop agent in the same time window. For multiple tasks that are aligned with the same workshop agent, the start and end range of the time window of each task is compared one by one. Multiple tasks with overlapping time windows are classified and combined to form resource preemption groups for production orders.

[0120] The collaborative scheduling agent identifies tasks with lower priority indicators in the resource preemption group according to a preset task priority judgment standard. This standard is based on the order delivery deadline and the production importance level. The promised delivery time of such tasks is successively postponed according to the preset unit production time. Each time it is postponed, the time window overlap status is re-checked until there is no overlap in the resource occupation time windows of all tasks. All adjusted tasks are integrated to obtain the resource-free task set of the production order.

[0121] The collaborative scheduling agent arranges the promised delivery times of all tasks in the resource-free task set in chronological order. Based on a continuous production timeline, it precisely maps the product type identifier and promised delivery quantity of each task to the corresponding time node on the timeline. After completing all mappings, it integrates and generates a global collaborative scheduling plan for production orders.

[0122] The beneficial effects are that the implementation process forms a standardized task attribute set by deconstructing standardized elements, completes disorder adjustment based on process dependency constraints, eliminates task time conflicts by combining resource competition constraints, and generates a global collaborative scheduling plan through time axis mapping, ensuring that the production task process sequence is compliant and the resource allocation is reasonable, improving the scientificity and executability of the scheduling plan, and providing a stable global execution basis for production instruction issuance and status monitoring.

[0123] A5. Send the global collaborative scheduling plan to the AI ​​full-stack intelligent agent to perform step-level decoding, obtain the processing instruction sequence of the production order, and send the processing instruction sequence to the manufacturing execution system interface;

[0124] In this embodiment of the invention, the step of sending the global collaborative scheduling plan to the instruction generation agent in the AI ​​full-stack agent for step-level decoding to obtain the processing instruction sequence of the production order, and then sending the processing instruction sequence to the manufacturing execution system interface, includes:

[0125] The instructions sent to the AI ​​full-stack intelligent agent from the global collaborative scheduling plan generate an intelligent agent, and extract the product type identifier, the promised delivery quantity, and the start and end values ​​of the time window for each task item in the global collaborative scheduling plan.

[0126] The process route template associated with the product type identifier is retrieved from the pre-stored process route library, and the process route template encapsulates the process step sequence corresponding to each task item and the operation code corresponding to each process step into a process step-level decoding mapping table for the production order.

[0127] Based on the step-level decoding mapping table and the start and end values ​​of the time window, the process parameters of the corresponding step sequence of each task item are fused to obtain the processing instruction record of each task item.

[0128] The processing instruction records are serialized and combined to obtain the processing instruction sequence of the production order, and the processing instruction sequence is sent to the manufacturing execution system interface.

[0129] The instruction generating agent receives the global collaborative scheduling plan transmitted by the collaborative scheduling agent. Each task entry in the global collaborative scheduling plan has a preset fixed field storage location. The product type identifier, the promised delivery quantity, and the start and end values ​​of the time window correspond to independent exclusive storage fields. According to the preset field positioning rules, the instruction generating agent accurately locates the exclusive storage field of each task entry, reads the complete content of the product type identifier, the complete content of the promised delivery quantity, and the complete content of the start and end values ​​of the time window for each entry, and completes the extraction of the core data of all task entries.

[0130] The pre-stored process route library is hierarchically classified and archived according to product type identifiers. Each product type identifier is bound to a unique corresponding process route template. The process route template fully records the work step sequence and the unique operation code for each task item. The instruction generation agent performs a character-by-character full match verification between the extracted product type identifier and all product type identifier items in the process route library. If the match result is completely consistent, the process route template corresponding to the product type identifier is immediately retrieved. According to the preset task item order in the template, the work step sequence and the operation code corresponding to each task item are extracted one by one. The extracted work step sequence and operation code are encapsulated and organized according to the one-to-one correspondence of task items to form a work step-level decoding mapping table for the production order.

[0131] The instruction generation agent uses a step-level decoding mapping table as the core conversion basis, retrieves the start and end values ​​of the time window corresponding to each task item in the global collaborative scheduling plan, and accurately binds the step sequence, operation code and time window start and end values ​​corresponding to each task item. It configures a matching start execution time and end execution time for each step in the step sequence, and fully integrates the process execution requirements and time execution parameters into the step information, completes the comprehensive integration of process parameters and step information, and generates the processing instruction record corresponding to each task item.

[0132] The instruction-generating agent refers to the predetermined execution order of task items in the global collaborative scheduling plan, arranges and integrates all processing instruction records in an orderly manner according to this order, forms a processing instruction sequence for the production order, and transmits the processing instruction sequence to the manufacturing execution system interface through a dedicated data transmission channel preset by the system, thus completing the complete issuance of processing instructions.

[0133] The beneficial effects of this implementation process are that by accurately extracting core data from the scheduling plan, matching exclusive process route templates, and completing the integration of step-level parameters and instruction serialization, it achieves automatic conversion from the global scheduling plan to executable processing instructions. The instructions are standardized and accurate, with clear execution sequences, and can be directly connected to the manufacturing execution system, improving the efficiency and accuracy of production instruction issuance and ensuring a high degree of synchronization between the production execution process and the scheduling plan.

[0134] A6. Through the status monitoring agent in the AI ​​full-stack intelligent agent, the real-time production status data received by the manufacturing execution system interface is compared with the planned progress in the global collaborative scheduling plan to detect the time sequence deviation and generate deviation alarms for production orders.

[0135] In this embodiment of the invention, the step of using the state monitoring agent in the AI ​​full-stack intelligent agent to detect the time-series offset between the real-time production status data received from the manufacturing execution system interface and the planned progress in the global collaborative scheduling plan, in order to generate a deviation alarm for the production order, includes:

[0136] Receive real-time production status data returned by the manufacturing execution system interface, and read the start and end timestamps of each task item in the global collaborative scheduling plan to obtain the real-time status set and planned time baseline of the production order;

[0137] By using the state monitoring agent in the AI ​​full-stack agent, the actual start and end timestamps of each task item in the real-time state set are periodically checked against the planned start and end timestamps of the corresponding task items in the planned time baseline to obtain the time-series offset task set of the production order.

[0138] By using a state monitoring agent, violations are identified in the time-series offset task set, and deviation alarms for production orders are obtained.

[0139] The status monitoring agent continuously receives real-time production status data from the manufacturing execution system interface through a dedicated data transmission channel preset by the system. This data includes complete actual production execution information and actual start and end timestamps for each task item. The status monitoring agent simultaneously and accurately locates each task item from the global collaborative scheduling plan, reads the planned start and end timestamps of the corresponding task item, integrates the actual production execution information and actual start and end timestamps of all task items to form a real-time status set of the production order, and integrates the planned start and end timestamps of all task items to form the planned time baseline of the production order.

[0140] The status monitoring agent performs data comparison according to a preset fixed verification cycle. It compares the actual start and end timestamps of each task item in the real-time status set with the planned start and end timestamps of the same numbered task items in the planned time baseline one by one. Using the preset allowable time deviation value as the judgment criterion, all task items whose difference between the actual start and end timestamps and the planned start and end timestamps exceeds the allowable time deviation value are screened out. All the screened task items are integrated to form the time-series offset task set of the production order.

[0141] The status monitoring agent checks each task item in the time-series deviation task set according to the preset production progress deviation violation judgment criteria. The check includes whether the task deviation duration exceeds the production allowable range and whether the deviation affects the execution of subsequent processes. The task items that meet the violation judgment conditions are identified as deviation tasks. Corresponding warning information is generated for deviation tasks, and all warning information is integrated to form a deviation alarm for the production order.

[0142] The beneficial effects of this implementation process are that by connecting to production execution data in real time, establishing a standardized plan time baseline, and conducting periodic time sequence verification and violation identification, it achieves full-process visual monitoring of production progress, accurately identifies production time sequence deviation problems and outputs deviation alarms in a timely manner, ensuring that production anomalies can be perceived in real time, and providing accurate triggering conditions and data support for adaptive rescheduling of subsequent production tasks.

[0143] A7. The status monitoring agent feeds back deviation alarms to the collaborative scheduling agent. Based on the latest real-time production status data received from the manufacturing execution system interface, it adaptively reschedules the unfinished production tasks in the production order to obtain the rescheduling plan for the production order.

[0144] In this embodiment of the invention, the status monitoring agent feeds back deviation alarms to the collaborative scheduling agent, and adaptively reschedules incomplete production tasks in the production order based on the latest real-time production status data received from the manufacturing execution system interface, thereby obtaining a rescheduling plan for the production order, including:

[0145] The status monitoring agent sends deviation alarms to the collaborative scheduling agent for alarm semantic decoding to obtain the set of rescheduling trigger conditions for production orders;

[0146] The collaborative scheduling agent obtains the latest real-time production status data from the manufacturing execution system interface, and extracts the actual completion progress and remaining workload of unfinished tasks from the latest real-time production status data to obtain a snapshot of the status of unfinished tasks in the production order.

[0147] Based on the rescheduling trigger condition set, and with the snapshot of the incomplete task status as the initial state, the task granularity division and distribution process of the collaborative scheduling agent is invoked again to obtain the rescheduling plan for production orders.

[0148] The status monitoring agent transmits deviation alarms to the collaborative scheduling agent through the system's preset internal data transmission channel. The collaborative scheduling agent, based on the preset alarm semantic parsing specifications, classifies and parses the core information in the deviation alarm, such as the unique identifier of the deviation task, the specific duration of the time offset, and the scope of impact on subsequent production processes. It then extracts all the limiting conditions that must be initiated to reschedule, and organizes and summarizes all the limiting conditions according to type and priority to form a set of rescheduling trigger conditions for the production order.

[0149] The collaborative scheduling agent proactively sends a request to retrieve the latest production data to the manufacturing execution system interface through a pre-set dedicated data interaction port. After receiving the request, the manufacturing execution system interface transmits back the latest execution status data of all current production tasks in real time. The collaborative scheduling agent filters out all production tasks that have not reached the completion status according to the pre-set incomplete task screening criteria. For each incomplete task, it accurately extracts the percentage of actual completed processes and the remaining workload to be executed. It integrates and archives the identifiers, actual completion progress, and remaining workload of all incomplete tasks according to the corresponding relationship, forming a snapshot of the incomplete task status of the production order.

[0150] The collaborative scheduling agent uses all the constraints in the rescheduling trigger condition set as the necessary basis for rescheduling execution. It uses the actual execution information of each task in the snapshot of the incomplete task status as the initial state of rescheduling. It re-invokes the complete execution process of task granularity division, sub-task distribution, workshop agent matching, department task commitment generation, and conflict resolution orchestration within the collaborative scheduling agent. It combines the latest production constraints and task status to re-plan and orchestrate all incomplete production tasks, and finally generates a production order rescheduling plan that adapts to the current production status.

[0151] The beneficial effects are that this implementation process determines the rescheduling trigger conditions through accurate alarm semantic decoding, collects the status of incomplete tasks in real time to form a complete snapshot, and restarts the scheduling process based on the trigger conditions and initial status to generate a rescheduling plan. This enables dynamic adaptive scheduling optimization after production anomalies, ensuring that production progress quickly returns to a reasonable track and improving the dynamic adaptability and overall execution efficiency of production scheduling.

[0152] like Figure 2 The diagram shown is a functional block diagram of a production scheduling and collaborative control system based on an AI full-stack intelligent agent provided in an embodiment of the present invention.

[0153] The production scheduling and collaborative control system based on an AI full-stack intelligent agent described in this invention can be installed in an electronic device. Depending on the functions implemented, the production scheduling and collaborative control system based on an AI full-stack intelligent agent may include an order parsing module, a task splitting module, a commitment generation module, a conflict resolution module, an instruction decoding module, an offset detection module, and a rescheduling module. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0154] In this embodiment, the functions of each module / unit are as follows:

[0155] The order parsing module uses an AI full-stack intelligent agent to perform semantic deconstruction on the received production orders, thereby obtaining the structured order requirements of the production orders.

[0156] The task splitting module sends the structured order requirements to the collaborative scheduling agent in the AI ​​full-stack intelligent agent for task granularity division, thereby obtaining the sub-task requirements of the production order;

[0157] The commitment generation module coordinates the dispatching agent to distribute the sub-task requirements to the corresponding workshop agents, and generates departmental task commitments for the sub-task requirements by combining the workshop agents' capacity data and material inventory data.

[0158] The conflict resolution module, based on preset process dependency constraints and resource competition constraints, performs conflict resolution and orchestration on departmental task commitments to obtain a global collaborative scheduling plan for production orders;

[0159] The instruction decoding module sends the global collaborative scheduling plan to the instruction generation agent in the AI ​​full-stack intelligent agent for step-level decoding to obtain the processing instruction sequence of the production order, and then sends the processing instruction sequence to the manufacturing execution system interface;

[0160] The offset detection module, through the status monitoring agent in the AI ​​full-stack intelligent agent, performs time-series offset detection between the real-time production status data received from the manufacturing execution system interface and the planned progress in the global collaborative scheduling plan, so as to generate deviation alarms for production orders.

[0161] The rescheduling module, where the status monitoring agent feeds back deviation alarms to the collaborative scheduling agent, adaptively reschedules the unfinished production tasks in the production order based on the latest real-time production status data received from the manufacturing execution system interface, thus obtaining the rescheduling plan for the production order.

[0162] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0163] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0164] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0165] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0166] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0167] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A production scheduling and collaborative control method based on an AI full-stack intelligent agent, characterized in that, The method includes: A1. Through the AI ​​full-stack intelligent agent for order parsing, the received production orders are semantically deconstructed to obtain the structured order requirements of the production orders; A2. The structured order requirements are sent to the collaborative scheduling agent in the AI ​​full-stack intelligent agent to divide the tasks into granularities and obtain the sub-task requirements of the production order. A3. The collaborative scheduling agent distributes the sub-task requirements to the corresponding workshop agents and, in conjunction with the workshop agents' capacity data and material inventory data, generates departmental task commitments for the sub-task requirements. A4. Based on preset process dependency constraints and resource competition constraints, conflict resolution and arrangement of departmental task commitments are carried out to obtain a global collaborative scheduling plan for production orders; A5. Send the global collaborative scheduling plan to the AI ​​full-stack intelligent agent to perform step-level decoding, obtain the processing instruction sequence of the production order, and send the processing instruction sequence to the manufacturing execution system interface; A6. Through the status monitoring agent in the AI ​​full-stack intelligent agent, the real-time production status data received by the manufacturing execution system interface is compared with the planned progress in the global collaborative scheduling plan to detect the time sequence deviation and generate deviation alarms for production orders. A7. The status monitoring agent feeds back deviation alarms to the collaborative scheduling agent. Based on the latest real-time production status data received from the manufacturing execution system interface, it adaptively reschedules the unfinished production tasks in the production order to obtain the rescheduling plan for the production order.

2. The production scheduling and collaborative control method based on an AI full-stack intelligent agent as described in claim 1, characterized in that, The order parsing agent, which utilizes a full-stack AI agent, performs semantic deconstruction on the received production orders to obtain structured order requirements, including: The order parsing agent, a full-stack AI agent, performs word segmentation on the text data of production orders to obtain the word segmentation sequence of production orders. The word segmentation sequence of orders is grammatically categorized to obtain the part-of-speech tags for production orders; Extract product name terms, quantity terms, and date terms from the part-of-speech tagging words. Treat product name terms as product type entities, quantity terms as delivery quantity entities, and date terms as expected delivery time entities. The order parsing agent calls the order data structure associated with the product type entity, assigns the delivery quantity entity to the quantity field in the order data structure, and assigns the expected delivery time entity to the delivery date field in the order data structure to generate a structured order demand record for the production order.

3. The production scheduling and collaborative control method based on an AI full-stack intelligent agent as described in claim 1, characterized in that, The process of sending structured order requirements to the collaborative scheduling agent within the AI ​​full-stack agent for task granularity division, resulting in sub-task requirements for the production order, includes: In the AI ​​full-stack intelligent agent, the collaborative scheduling intelligent agent receives structured order requirements and parses the product type field value, quantity field value, and delivery date field value from the structured order requirements as parameters for the task division of production orders; The process decomposition template associated with the product type field value is retrieved from the pre-stored process template library, and the names of each production stage and the corresponding standard working hours of each production stage are extracted from the process decomposition template to generate the initial task node list of the production order. Based on the quantity field value and the delivery date field value, time windows are allocated to the task nodes in the initial task node list to obtain the sub-task requirements of the production order.

4. The production scheduling and collaborative control method based on an AI full-stack intelligent agent as described in claim 1, characterized in that, The collaborative scheduling agent distributes sub-task requirements to the corresponding workshop agents, and combines the workshop agents' capacity data and material inventory data to generate departmental task commitments for the sub-task requirements, including: Data is extracted from each subtask node in the subtask requirements to obtain the product type identifier and quantity field value of each subtask node. Based on the product type identifier, workshop matching is performed on each sub-task node to obtain the distribution mapping relationship table of each sub-task node; According to the distribution mapping table, the collaborative scheduling agent sends each subtask node to the corresponding workshop agent and receives the current available capacity value and current available material quantity returned by the workshop agent to obtain the workshop real-time status data of the workshop agent; Based on the quantity field value and real-time status data of the workshop, the commitment amount of each sub-task node is calculated to obtain the departmental task commitment required by the sub-task.

5. The production scheduling and collaborative control method based on an AI full-stack intelligent agent as described in claim 4, characterized in that, The formula for calculating the delivery time offset that can be committed to the departmental task commitment is as follows: ； In the formula, For the first Promised delivery time offset for each subtask node For the first Each sub-task node corresponds to the current available production capacity value of the workshop intelligent agent. For the first The value of the field representing the number of subtask nodes. For the first Each sub-task node corresponds to the current available material quantity of the workshop agent. For the first Each sub-task node corresponds to the unit material consumption of the product.

6. The production scheduling and collaborative control method based on an AI full-stack intelligent agent as described in claim 1, characterized in that, The process of resolving conflicts and arranging departmental task commitments based on preset process dependency constraints and resource competition constraints to obtain a global collaborative scheduling plan for production orders includes: The elements of the departmental task commitments are deconstructed to obtain the set of task attributes to be arranged for the departmental task commitments; Based on the preset process dependency constraints, the predecessor and successor relationships of each task in the task attribute set to be arranged are retrieved to obtain the process disorder pairs of the production order. The promised delivery time of subsequent tasks in the process disorder pairing is postponed until the process dependency constraint is met, so as to obtain the process alignment task set of the production order. Based on the preset resource competition constraints, time window overlap detection is performed on multiple tasks that share the same workshop agent for process alignment tasks to obtain the resource preemption group of production orders; The promised delivery time of tasks with lower priority in the resource preemption group is postponed until the resource occupation time windows no longer overlap, so as to obtain a resource-free task set for the production order. The promised delivery time of each task in the resource-free task set is sorted and mapped with the corresponding product type identifier and promised delivery quantity on the timeline to generate a global collaborative scheduling plan for production orders.

7. The production scheduling and collaborative control method based on an AI full-stack intelligent agent as described in claim 1, characterized in that, The step of sending the global collaborative scheduling plan to the instruction generation agent in the AI ​​full-stack agent for step-level decoding to obtain the processing instruction sequence of the production order, and then sending the processing instruction sequence to the manufacturing execution system interface, includes: The instructions sent to the AI ​​full-stack intelligent agent from the global collaborative scheduling plan generate an intelligent agent, and extract the product type identifier, the promised delivery quantity, and the start and end values ​​of the time window for each task item in the global collaborative scheduling plan. The process route template associated with the product type identifier is retrieved from the pre-stored process route library, and the process route template encapsulates the process step sequence corresponding to each task item and the operation code corresponding to each process step into a process step-level decoding mapping table for the production order. Based on the step-level decoding mapping table and the start and end values ​​of the time window, the process parameters of the corresponding step sequence of each task item are fused to obtain the processing instruction record of each task item. The processing instruction records are serialized and combined to obtain the processing instruction sequence of the production order, and the processing instruction sequence is sent to the manufacturing execution system interface.

8. The production scheduling and collaborative control method based on an AI full-stack intelligent agent as described in claim 1, characterized in that, The process involves using a state monitoring agent within the AI ​​full-stack intelligent agent to detect time-series offsets between the real-time production status data received from the manufacturing execution system interface and the planned progress in the global collaborative scheduling plan, thereby generating deviation alarms for production orders. This includes: Receive real-time production status data returned by the manufacturing execution system interface, and read the start and end timestamps of each task item in the global collaborative scheduling plan to obtain the real-time status set and planned time baseline of the production order; By using the state monitoring agent in the AI ​​full-stack agent, the actual start and end timestamps of each task item in the real-time state set are periodically checked against the planned start and end timestamps of the corresponding task items in the planned time baseline to obtain the time-series offset task set of the production order. By using a state monitoring agent, violations are identified in the time-series offset task set, and deviation alarms for production orders are obtained.

9. The production scheduling and collaborative control method based on an AI full-stack intelligent agent as described in claim 1, characterized in that, The status monitoring agent feeds back deviation alarms to the collaborative scheduling agent. Based on the latest real-time production status data received from the manufacturing execution system interface, it adaptively reschedules incomplete production tasks in the production order to obtain a rescheduling plan for the production order, including: The status monitoring agent sends deviation alarms to the collaborative scheduling agent for alarm semantic decoding to obtain the set of rescheduling trigger conditions for production orders; The collaborative scheduling agent obtains the latest real-time production status data from the manufacturing execution system interface, and extracts the actual completion progress and remaining workload of unfinished tasks from the latest real-time production status data to obtain a snapshot of the status of unfinished tasks in the production order. Based on the rescheduling trigger condition set, and with the snapshot of the incomplete task status as the initial state, the task granularity division and distribution process of the collaborative scheduling agent is invoked again to obtain the rescheduling plan for production orders.

10. A production scheduling and collaborative control system based on an AI full-stack intelligent agent, characterized in that: The system for implementing the production scheduling and collaborative control method based on an AI full-stack intelligent agent as described in claim 1 includes: The order parsing module uses an AI full-stack intelligent agent to perform semantic deconstruction on the received production orders, thereby obtaining the structured order requirements of the production orders. The task splitting module sends the structured order requirements to the collaborative scheduling agent in the AI ​​full-stack intelligent agent for task granularity division, thereby obtaining the sub-task requirements of the production order; The commitment generation module coordinates the dispatching agent to distribute the sub-task requirements to the corresponding workshop agents, and generates departmental task commitments for the sub-task requirements by combining the workshop agents' capacity data and material inventory data. The conflict resolution module, based on preset process dependency constraints and resource competition constraints, performs conflict resolution and orchestration on departmental task commitments to obtain a global collaborative scheduling plan for production orders; The instruction decoding module sends the global collaborative scheduling plan to the instruction generation agent in the AI ​​full-stack intelligent agent for step-level decoding to obtain the processing instruction sequence of the production order, and then sends the processing instruction sequence to the manufacturing execution system interface; The offset detection module, through the status monitoring agent in the AI ​​full-stack intelligent agent, performs time-series offset detection between the real-time production status data received from the manufacturing execution system interface and the planned progress in the global collaborative scheduling plan, so as to generate deviation alarms for production orders. The rescheduling module, where the status monitoring agent feeds back deviation alarms to the collaborative scheduling agent, adaptively reschedules the unfinished production tasks in the production order based on the latest real-time production status data received from the manufacturing execution system interface, thus obtaining the rescheduling plan for the production order.

Images