A collaborative control method and device for a supply chain, an electronic device, a computer readable storage medium and a computer program product

Abstract

This application provides a method, apparatus, electronic device, computer program product, and computer-readable storage medium for collaborative control of a supply chain. The method includes: parsing heterogeneous data packets using a preset semantic communication protocol to obtain scheduling requests, constraint parameters, and real-time status of physical resources from business nodes; determining resource status anomaly events caused by the real-time status in the future based on a dynamic knowledge graph of the supply chain in the time dimension; performing conflict detection on the scheduling requests based on multiple constraint parameters and resource status anomaly events to obtain conflict detection results; determining the optimal action from a set of candidate actions when the conflict detection results indicate a conflict; generating an execution instruction based on the optimal action and the preset semantic communication protocol, and sending the execution instruction to the business node. This application can improve the accuracy of collaborative control of the supply chain.

Description

Technical Field

[0001] This application relates to distributed system technology, and more particularly to a collaborative control method, apparatus, electronic device, computer-readable storage medium, and computer program product for a supply chain. Background Technology

[0002] In modern supply chain scenarios, enterprises typically need to run multiple heterogeneous business systems simultaneously. To cope with dynamic business demands within the supply chain, collaborative control needs to complete cross-system state awareness, constraint verification, action coordination, and execution closure within a short period of time, thereby achieving collaborative control of various business systems within the supply chain. However, related technologies still suffer from low accuracy in supply chain collaborative control. Summary of the Invention

[0003] This application provides a method, apparatus, electronic device, computer-readable storage medium, and computer program product for collaborative control of the supply chain, which can improve the accuracy of collaborative control of the supply chain.

[0004] The technical solution of this application embodiment is implemented as follows: This application provides a collaborative control method for a supply chain, the method comprising: Using a preset semantic communication protocol, multiple heterogeneous data packets sent by multiple business nodes are parsed to obtain the scheduling request of the business node for physical resources in the supply chain, the constraint parameters of the scheduling request, and the real-time status of the physical resources. Based on the dynamic knowledge graph of the supply chain in the time dimension, determine the abnormal resource status events caused by the real-time status in the future. Based on multiple constraint parameters and resource status anomaly events, conflict detection is performed on multiple scheduling requests to obtain conflict detection results. When the conflict detection results indicate that a conflict exists, the optimal action is determined from the candidate action set corresponding to the conflict. Based on the optimal action and the preset semantic communication protocol, an execution instruction is generated and sent to multiple service nodes.

[0005] This application provides a collaborative control device for a supply chain, comprising: The data parsing module is used to parse multiple heterogeneous data packets sent by multiple business nodes using a preset semantic communication protocol, so as to obtain the scheduling request of the business node for physical resources in the supply chain, the constraint parameters of the scheduling request, and the real-time status of the physical resources. The event determination module is used to determine, based on the dynamic knowledge graph of the supply chain in the time dimension, the abnormal resource status events caused by the real-time status in the future. The conflict detection module is used to perform conflict detection on multiple scheduling requests and resource status abnormal events based on multiple constraint parameters and resource status abnormal events, and obtain conflict detection results. The action determination module is used to determine the optimal action from the set of candidate actions corresponding to the conflict when the conflict detection result indicates that a conflict exists; The instruction generation module is used to generate an execution instruction based on the optimal action and the preset semantic communication protocol, and send the execution instruction to multiple service nodes.

[0006] This application provides an electronic device, the electronic device comprising: Memory is used to store executable instructions or computer programs. The processor, when executing computer-executable instructions or computer programs stored in the memory, implements the supply chain collaborative control method provided in the embodiments of this application.

[0007] This application provides a computer-readable storage medium storing a computer program or computer-executable instructions for implementing the supply chain collaborative control method provided in this application when executed by a processor.

[0008] This application provides a computer program product, including a computer program or computer executable instructions. When the computer program or computer executable instructions are executed by a processor, they implement the supply chain collaborative control method provided in this application.

[0009] The embodiments of this application have the following beneficial effects: By using a preset semantic communication protocol, heterogeneous data packets sent by business nodes distributed across different business systems are parsed to obtain information rich in semantics, such as the business nodes' scheduling requests for physical resources, the constraint parameters of the scheduling requests, and the real-time status of the physical resources. Furthermore, based on a dynamic knowledge graph with a time dimension, the actual state is inferred over time to identify potential future resource status anomalies. Then, based on the different constraint parameters of different scheduling requests and the resource status anomalies, it is determined whether each scheduling request will be affected by future resource status anomalies, achieving comprehensive conflict detection. If a conflict exists, the optimal action is determined from the set of candidate actions corresponding to the conflict, and an execution instruction is generated based on the optimal action. In this way, collaborative control of the supply chain can be achieved based on semantically rich information, and the impact of future anomalies is fully considered in the collaborative control, thereby enabling more accurate selection of the optimal action and more accurate collaborative control. In summary, the embodiments of this application can improve the accuracy of collaborative control of the supply chain. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating the collaborative control method for the supply chain provided in this application. Figure 1 ; Figure 2 This is a schematic diagram of the standard message structure generated by the preset semantic communication protocol provided in the embodiments of this application; Figure 3 This is a flowchart illustrating the collaborative control method for the supply chain provided in this application. Figure 2 ; Figure 4 This is a flowchart illustrating the collaborative control method for the supply chain provided in this application. Figure 3 ; Figure 5 This is a schematic diagram illustrating the principle of resource conflict detection provided in the embodiments of this application; Figure 6 This is a schematic diagram of the spatiotemporal conflict detection process provided in an embodiment of this application; Figure 7 This is a structural diagram of a supply chain collaborative control device provided in an embodiment of this application; Figure 8 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0011] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0012] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0013] In the following description, the terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first" and "second" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0014] In the embodiments of this application, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0015] Unless otherwise defined, all technical and scientific terms used in the embodiments of this application have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in the embodiments of this application is for the purpose of describing the embodiments of this application only and is not intended to limit this application.

[0016] In the implementation of this application, the collection and processing of relevant data should strictly comply with the requirements of relevant laws and regulations, obtain the informed consent or separate consent of the personal information subject, and carry out subsequent data use and processing within the scope of laws and regulations and the authorization of the personal information subject.

[0017] Before providing a further detailed description of the embodiments of this application, the nouns and terms involved in the embodiments of this application will be explained, and the nouns and terms involved in the embodiments of this application shall be interpreted as follows.

[0018] 1) Supply chain refers to a complex network structure composed of multiple interactive links such as raw material procurement, finished product processing and manufacturing, warehousing and logistics and last-mile delivery, involving cross-regional flow and collaboration of material flow and information flow between different organizational boundaries.

[0019] 2) A business node is a computing unit deployed within a specific business system of the supply chain. It has the ability to collect data, perform preliminary processing, and send data to the business system it belongs to. Each business node corresponds one-to-one with a business system in the supply chain. By calling the application programming interface of the business system or reading its underlying database, a business node can independently complete the collection of data and the issuance of instructions within that business system.

[0020] 3) Heterogeneous data packets refer to communication packets obtained by business nodes of different business systems after standardizing and encapsulating the collected raw business data according to a preset semantic communication protocol. Different heterogeneous data packets have a unified message structure, but the data sources they carry are heterogeneous, covering the business meanings, dimensions, and contexts of different business domains.

[0021] 4) Physical resources refer to resources that actually exist in each business link of the supply chain, are used to support actual tasks such as production, storage, circulation or delivery, and have a physical form.

[0022] 5) Scheduling request refers to the information sent by the business node to represent its intention to operate or use physical resources. It clarifies the specific business purpose of the business node for applying for, releasing, canceling tasks or reporting status of physical resources, that is, it indicates the specific operation that the business node wants to perform on the physical resources.

[0023] 6) Constraint parameters are the restrictions that scheduling requests must meet during execution, and these conditions determine the feasibility boundary of the scheduling request.

[0024] 7) Resource occupation scope refers to the specific range of physical resources required by the scheduling request in terms of spatial location, time span, or quantity.

[0025] 8) Resource status anomalies refer to specific situations that, based on inference, may occur in the future regarding physical resources, and that do not conform to preset business rules or physical constraints. These represent potential risks in the future, such as anticipated stockouts, insufficient capacity, or arrival delays.

[0026] 9) The scope of resource impact refers to the set of related objects that are triggered by abnormal resource status events and have a chain reaction within the supply chain network, including downstream tasks that may be blocked due to dependencies, related orders, and other related physical resources that are affected.

[0027] 10) The Pareto front solution set is a set of non-dominated candidate actions calculated when dealing with multi-objective resource scheduling optimization problems. This set represents all the best trade-offs. For any candidate action in this set, if the performance of one objective is to be improved, at least one other objective must be sacrificed.

[0028] In modern supply chain scenarios, enterprises typically need to run multiple heterogeneous business systems simultaneously, such as Enterprise Resource Planning (ERP), Warehouse Management System (WMS), Manufacturing Execution System (MES), and Transportation Management System (TMS), etc., and these business systems are often deployed in different physical regions or network environments. To cope with the dynamic business demands in the supply chain, collaborative control of the supply chain needs to complete cross-system state awareness, constraint verification, action coordination, and execution closure within a short period of time to achieve collaborative control of various business systems in the supply chain.

[0029] In related technologies, collaborative control of the supply chain is typically achieved through the following two types of solutions: The first type of solution is to achieve collaborative control of the supply chain based on centralized unified optimization (such as a control tower). This solution first collects data from various business systems and then uses algorithms such as linear programming for centralized solution. However, the solution complexity of this solution increases with the scale of business and the dimensions of constraints, making it difficult to meet real-time requirements, and the central node is susceptible to single point of failure.

[0030] The second approach employs a distributed architecture based on point-to-point API calls and rule engine integration to achieve collaborative control of the supply chain. In this approach, various business systems interact with each other through application programming interfaces (APIs) or message brokers, thereby mitigating the problems associated with the first approach, which is based on centralized unified optimization.

[0031] However, this approach brings some new problems: First, the interaction between various business systems relies on the transmission of discrete data fields, and there is a lack of standardized protocols that can define the specific scheduling requests, constraint parameters, and required physical resources of each business system. This makes it difficult to effectively coordinate the actions of various business systems, and thus makes it difficult to accurately complete the collaborative control of the supply chain.

[0032] Secondly, this solution is essentially implemented through static rule matching, lacking causal deduction in the time dimension. That is, it cannot take into account the impact of the real-time status of physical resources in the supply chain on the future, making it difficult to effectively coordinate the actions of various business systems, thus making it difficult to accurately complete the collaborative control of the supply chain.

[0033] This shows that the accuracy of collaborative control of the supply chain still remains low in related technologies.

[0034] In addition, this solution lacks a response plan for network jitter or execution failure when implementing collaborative control of the supply chain, thus failing to guarantee the reliability of data status between heterogeneous business systems, which further affects the reliability of collaborative control of the supply chain.

[0035] To address the aforementioned issues, this application provides a collaborative control method for a supply chain, which improves the accuracy of collaborative control. The method includes: parsing multiple heterogeneous data packets sent by multiple business nodes using a preset semantic communication protocol to obtain scheduling requests from business nodes for physical resources in the supply chain, constraint parameters of the scheduling requests, and the real-time status of the physical resources; determining resource status anomalies caused by the real-time status in the future based on a dynamic knowledge graph of the supply chain in the time dimension; performing conflict detection on multiple scheduling requests based on multiple constraint parameters and resource status anomalies to obtain conflict detection results; determining the optimal action from the candidate action set corresponding to the conflict when the conflict detection results indicate a conflict; generating an execution instruction based on the optimal action and the preset semantic communication protocol, and sending the execution instruction to multiple business nodes. This achieves collaborative control of the supply chain based on semantically contained information. Furthermore, the collaborative control not only considers whether different scheduling requests conflict but also fully considers the impact of future anomalies on the scheduling requests, thereby enabling more accurate selection of the optimal action and thus more accurate collaborative control, improving the accuracy of collaborative control of the supply chain.

[0036] The following will describe the collaborative control method for the supply chain provided in the embodiments of this application.

[0037] See Figure 1 , Figure 1 This is a flowchart illustrating the collaborative control method for the supply chain provided in this application. Figure 1 , will combine Figure 1 The steps shown are explained below. Figure 1 The main component of the process is electronic equipment.

[0038] Step 101: Using a preset semantic communication protocol, parse multiple heterogeneous data packets sent by multiple business nodes to obtain the scheduling requests of business nodes for physical resources in the supply chain, the constraint parameters of the scheduling requests, and the real-time status of the physical resources.

[0039] In this embodiment, the electronic device first receives multiple heterogeneous data packets from multiple service nodes via a message bus or network interface. These heterogeneous data packets can be received by the electronic device in parallel or sequentially within a certain time period. Next, the electronic device can identify the type of business system corresponding to each service node based on the source identifier or header information of each heterogeneous data packet. Then, using a preset semantic communication protocol, it performs structured parsing of the heterogeneous data packets of each service node to extract information such as scheduling requests, constraint parameters, and real-time status.

[0040] It should be noted that multiple business nodes are distributed across different business systems within the supply chain. In other words, in this embodiment, to achieve real-time perception and collaborative control of the entire supply chain process, business nodes are deployed in computer management systems responsible for different business functions. Each business node corresponds one-to-one with a specific business function within the supply chain; for example, a business node connected to the WMS is deployed in the warehousing stage, and a business node connected to the TMS is deployed in the transportation stage. Each business node independently completes the collection of internal data and the issuance of instructions within its respective business system by calling the application programming interface (API) or reading its underlying database.

[0041] It should be noted that a supply chain refers to a functional network structure centered around a core enterprise, connecting raw material suppliers, manufacturers, distributors, retailers, and ultimately end users into a unified whole through the control of information flow, logistics, and capital flow. In this embodiment, the supply chain includes not only the physical flow network of materials but also a complex set of business processes, encompassing the entire lifecycle from raw material procurement, semi-finished product processing, finished product storage, trunk transportation, to last-mile delivery. The operational status of the supply chain is jointly determined by the real-time status of the physical resources (such as inventory, production capacity, and transportation capacity) at each link.

[0042] A business system refers to a computer application system used in a specific link of the supply chain to execute specific business logic, manage business data, and control business processes. Each business system has a specific application area. For example, business systems may include Enterprise Resource Planning (ERP), Warehouse Management System (WMS), Manufacturing Execution System (MES), Transportation Management System (TMS), and so on. Specifically, ERP handles order management, financial accounting, and procurement planning; WMS handles material receiving, issuing, inventory counting, and storage location management; MES handles work order scheduling, process control, and equipment inspection in the production workshop; and TMS handles vehicle scheduling and route planning.

[0043] A business node refers to a computing unit deployed within the aforementioned business system, possessing the capability to collect, preliminarily process, and transmit data from the business system it resides in. A business node can be an edge gateway, a local data center, or a cloud-based microservice. For example, a business node could be an industrial gateway connected to the server of an automated warehouse management system (WMS), or an API proxy service integrated into the MES system interface.

[0044] Heterogeneous data packets are communication packets obtained by edge nodes of different business systems after standardizing and encapsulating the collected business data according to a preset semantic communication protocol. Here, heterogeneity refers to the heterogeneity of the data packet content; that is, while heterogeneous data packets sent by different business nodes have a unified message structure, the data sources carried in the data packets are heterogeneous. For example, heterogeneous data packets sent by WMS business nodes carry inventory data, heterogeneous data packets sent by TMS business nodes carry transportation data, and heterogeneous data packets sent by MES business nodes carry production capacity data. The business meaning, dimensions, and context of this data are essentially heterogeneous (i.e., they belong to different business domains).

[0045] Physical resources refer to resources that actually exist in various business links of the supply chain, are used to support actual tasks such as production, storage, circulation, or delivery, and have a physical form. In the embodiments of this application, physical resources may include material resources, such as raw materials, components, semi-finished products, and finished products, for example, chips stored in a warehouse; physical resources may include equipment resources, such as machines and production lines used for processing, assembly, or handling, such as final assembly lines; physical resources may also include logistics resources, such as physical tools or network nodes used for spatial transfer, such as transport fleets or containers; physical resources may also include warehousing space resources, such as physical spaces used for storage, such as cold storage locations or loading docks in a warehouse.

[0046] A scheduling request is a message sent by a business node to represent its intention to operate on or use physical resources. In other words, a scheduling request indicates the operation that a business node wants to perform on the physical resources. For example, a scheduling request from a business node in a production system could be "Request to lock 100 units of raw material A for production work order X".

[0047] Constraint parameters are the restrictions that a scheduling request must meet during execution. These constraints determine the feasibility boundary of the scheduling request. These constraints can involve dimensions such as time, cost, environment, and process. In other words, they determine under what circumstances the scheduling request is feasible and under what circumstances it is not. For example, constraint parameters could be that the request needs to be completed before 10:00 AM on June 1, 2024, or that the temperature during transportation must not exceed 5 degrees Celsius, etc.

[0048] Real-time status represents the operational attributes or availability of physical resources at the current moment. For example, the "running / faulty / idle" status of a production device, or the "in stock / under quality inspection / frozen" status of a batch of materials, etc.

[0049] A predefined semantic communication protocol is a set of rules that are predefined to unify the communication standards between different business systems. In other words, a predefined semantic communication protocol not only defines the format of data transmission, but also defines the business meaning (semantics) of the data, field mapping relationships, etc.

[0050] In this embodiment of the application, a preset semantic communication protocol defines a standard message structure for carrying the heterogeneous data packets. The standard message structure includes at least: a scheduling request field, a constraint parameter field, and a status field.

[0051] It should be noted that the scheduling request field is used to define the type of scheduling request; The constraint parameter field is used to define the constraint parameters that the scheduling request must satisfy when it is executed; The status field is used to carry the real-time status value of physical resources or device identifiers.

[0052] In other words, in this embodiment of the application, heterogeneous data from different business systems can be mapped into a unified semantic expression by using a standard information structure defined by a preset semantic communication protocol, so as to obtain corresponding heterogeneous data messages.

[0053] It should be noted that the standard information structure refers to the data encapsulation method that is mandated by the preset semantic communication protocol and must be followed by all heterogeneous data packets. Therefore, regardless of whether the underlying business system transmits XML, JSON, or a private binary stream, the heterogeneous data packets processed by the corresponding business nodes must logically contain the key-value pairs or data blocks defined by this structure to ensure that electronic devices can read the corresponding data with a unified logic.

[0054] The scheduling request field is a specific data field reserved in the standard message structure to identify the business intent of a business node. This field characterizes the purpose for which the business node sends the message, specifying what the business node intends to do, such as requesting or releasing resources, canceling a task, or simply reporting a status. The value range of the scheduling request field can be a predefined set of intents, such as: Request (representing a request to occupy or lock specific physical resources); Release (representing a request to release the occupation of physical resources upon task completion); Cancel (representing the cancellation of a previously issued scheduling request); Report (representing only proactive status reporting without resource requests); Update (representing changes to previously issued request parameters (such as time or quantity)), etc.

[0055] The constraint parameter field is a specific data field in the standard message structure used to encapsulate restrictive conditions. This field characterizes what limitations exist in the scheduling request, ensuring that the scheduling request is executed within specific boundaries. The value range of the constraint parameter field varies depending on the type of constraint. In some embodiments, constraint parameters may include physical constraint parameters, such as numerical ranges for temperature and humidity limits, or weight / capacity limits, whose value range can be a physically meaningful numerical range, such as a temperature of (-∞, -18] degrees Celsius (cold chain transportation) or a weight of (0, 10000] kilograms. In other embodiments, constraint parameters may also include business constraint parameters, such as order priority, supplier / customer list, order deadline, etc.

[0056] The status field is a specific data field in the standard message structure used to characterize the current physical state. Therefore, the status field indicates the current state of a physical resource. Depending on the attributes of the physical resource, the status field can have both discrete state and continuous numerical values. For example, for the device's operating status, its value range can be [Idle, Running, Error, Maintenance]; for the physical resource's inventory status, its value range can be non-negative integers, and so on.

[0057] The aforementioned scheduling request field, constraint parameter field, and status field can be set in the message body. That is, the business node can fill the local business data into the standard slots of the above four dimensions according to the preset semantic communication protocol to form the message body. Then, a message header containing basic information such as the protocol version and message ID of the preset semantic communication protocol is generated for the message body. Then, the message header and message body form a heterogeneous data message.

[0058] It should be noted that, in some embodiments of this application, the standard message structure defined by the preset semantic communication protocol may include, in addition to the scheduling request field, constraint parameter field and status field mentioned above, a resource requirement field, wherein the resource requirement stage is used to carry the scope of physical resource occupation for the scheduling request.

[0059] The resource requirement field is a specific data field in the standard message structure used to quantify the needs of a scheduling request. It characterizes how much physical resources the scheduling request requires and when it is needed. The resource requirement field must contain values ​​in at least two dimensions: time and space. The time dimension of the resource requirement field can be a time period consisting of a start time and an end time, while the space dimension can be a positive integer or a positive real number.

[0060] The standard message structure can also include fields to support more complex engineering control logic, such as a source identifier field, a timestamp field, an authorization information field, and an idempotency control field. The source identifier field uniquely identifies the ID of the service node that sent the heterogeneous data packet, or the ID of its associated service system; the timestamp field records the precise time the heterogeneous data packet was generated or sent; the authorization information field records the authorization information for the actions required by the electronic device for the service node; and the idempotency control field carries a unique serial number to prevent duplicate executions caused by network retries.

[0061] For example, Figure 2 This is a schematic diagram of a standard message structure generated by the preset semantic communication protocol provided in the embodiments of this application. See also... Figure 2The standard message structure 2-1 generated by the pre-defined semantic communication protocol is logically divided into two parts: a message header 2-2 and a message body 2-3. The message header 2-2 contains basic fields for ensuring communication security and transmission control, such as the sender source identifier field 2-21, the timestamp field 2-22, the authorization information field 2-23, and the idempotency control field 2-24. The message body 2-3 encapsulates specific fields used for collaborative control processing, such as the scheduling request field 2-31, the constraint parameter field 2-32, the status field 2-33, and the resource requirement field 2-34. Through this standardized message structure, electronic devices can parse heterogeneous data packets from different business systems using a unified logic.

[0062] It is understood that in this embodiment of the application, a standard message structure is defined by a preset semantic communication protocol, so that the electronic device can directly extract relevant information from heterogeneous data packets according to the standard message structure. In this way, the semantic understanding deviation caused by different field definitions of different services can be eliminated, thereby ensuring that the electronic device can clearly understand the different negotiation actions required by each service node, so as to improve the accuracy of collaborative control.

[0063] Next, the specific parsing process of heterogeneous data packets is explained in conjunction with the specific content of the preset semantic communication protocol. The electronic device parses the heterogeneous data packets using the preset semantic communication protocol, which can be achieved by directly locating and reading the values ​​of corresponding fields according to the standard message structure defined by the preset semantic communication protocol. More specifically, in this embodiment, the electronic device can first verify the integrity of the heterogeneous data packets according to the preset semantic communication protocol. After successful verification, the electronic device scans the values ​​of the scheduling request field in the heterogeneous data packets to obtain the specific scheduling request of the business node; scans the values ​​of the constraint parameter field in the heterogeneous data packets and determines the values ​​as the constraint parameters of the scheduling request; scans the values ​​of the status field in the heterogeneous data packets and uses the scanned values ​​as the real-time status of the physical resources; it can also scan the values ​​of the fields used to define resource requirements in the heterogeneous data packets and determine the scanned values ​​as the resource occupancy range of the scheduling request.

[0064] Step 102: Based on the dynamic knowledge graph of the supply chain in the time dimension, determine the abnormal resource status events caused by the real-time status in the future.

[0065] Electronic devices perform reasoning based on a dynamic knowledge graph in the time dimension to proactively identify potential resource status anomalies caused by the real-time state of physical resources. The dynamic knowledge graph of the supply chain in the time dimension is a graph data model that includes entities within the supply chain, entity relationships, and the evolution of entity states. Graph nodes in the dynamic knowledge graph can represent physical entities in the supply chain, directed edges represent dependencies, and node attributes can record the states of physical entities.

[0066] Resource anomaly events refer to predicted abnormal events that may occur in physical resources in the future. These anomaly events are sudden changes in state that do not conform to preset business rules or physical constraints, representing potential risks to physical resources in the future. For example, resource anomaly events may include events such as a predicted material shortage in 2 hours, a predicted insufficient production capacity tomorrow, or a predicted delay in transportation arrival.

[0067] See Figure 3 , Figure 3 This is a flowchart illustrating the collaborative control method for the supply chain provided in this application. Figure 2 In some embodiments of this application, Figure 1 Step 102, which involves determining resource status anomalies caused by the real-time status in the future based on the dynamic knowledge graph of the supply chain in the time dimension, can be achieved through the following processing: Step 1021: From the dynamic knowledge graph, determine the first graph node corresponding to the physical resource, and update the node attributes of the first graph node based on the real-time status.

[0068] The electronic device first indexes the graph node corresponding to the physical resource in the dynamic knowledge graph, takes the graph node as the first graph node, then extracts the original node attributes of the first graph node, and updates the original node attributes using the real-time status of the physical resource to obtain the updated node attributes of the first graph node.

[0069] It should be noted that node attributes are data combinations that describe the characteristics of graph nodes. Node attributes can change over time and can include status values ​​(such as quantity and location), timestamps, and validity markers. Based on the real-time status, updating the node attributes of the first graph node can be achieved by directly overwriting the original values ​​of the node attributes in the original node attributes of the first graph node. For example, if the real-time status is "equipment failure," the original value of the node attribute of the first graph node can be directly changed from "running" to "failure." Alternatively, this process can involve cumulatively updating the original values ​​of the node attributes of the first graph node. For example, if the real-time status is "50 items received," and the current value of the first graph node is 100, adding the newly added value will update the value of the node attribute to 150.

[0070] Step 1022: Based on the directed edges representing physical dependencies in the dynamic knowledge graph, search for downstream nodes affected by the nodes in the first graph in the dynamic knowledge graph.

[0071] After updating the node attributes of the first graph node, the electronic device can use the first graph node as a starting point and search for other graph nodes based on the topological structure of the dynamic knowledge graph, i.e., the connection of directed edges, to identify all graph nodes that are logically dependent on the first graph node and determine these graph nodes as downstream nodes of the first graph node.

[0072] In some embodiments of this application, the electronic device can start from the first graph node, first search all first-level downstream nodes directly connected to it, then search the second-level downstream nodes of the first-level downstream nodes, and so on, until a preset search depth (such as 3 layers) is reached, so as to fully determine all downstream nodes affected by the first graph node.

[0073] In some other embodiments of this application, the electronic device can search only along directed edges with specific labels based on a specific business type (such as focusing only on the logistics link), while filtering out directed edges with irrelevant labels (such as edges representing financial relationships), thereby filtering out graph nodes that depend on the first graph node but are irrelevant to the specific business type, thereby accurately locating downstream nodes related to the business.

[0074] Step 1023: Based on the updated node attributes of the first graph nodes and the attribute passing logic of the directed edges, update the node attributes of the downstream nodes.

[0075] The electronic device utilizes the attribute propagation logic of directed edges to calculate how changes in the node attributes of the first graph node, after being propagated through physical dependencies, manifest as changes in the node attributes of downstream nodes. Based on this, the device modifies the future values ​​of the downstream node attributes. In other words, in this step, the electronic device uses the attribute propagation logic of directed edges to calculate how changes in the node attributes of the upstream first graph node affect the node attributes of the downstream node, thereby deducing the future node attributes of the downstream node.

[0076] It should be noted that the attribute propagation logic for directed edges is a predefined set of calculation rules or functions used to quantify how changes in the attributes of upstream graph nodes affect the attributes of downstream graph nodes. For example, the attribute propagation logic could be that a 1-hour delay upstream leads to a 1-hour delay downstream, or a material shortage of 1 upstream node leads to a production reduction of 1 downstream node, and so on.

[0077] Step 1024: Based on the updated node attributes of downstream nodes, determine the abnormal resource status events.

[0078] The electronic equipment system checks whether the updated node attribute values ​​of downstream nodes exceed the preset business baseline or security threshold. If they do, the system identifies this state as a resource status anomaly event.

[0079] Electronic devices can first obtain the normal attribute range (such as inventory safety level, delivery deadline) for downstream nodes, and then determine whether the updated node attribute value of the downstream node exceeds the normal attribute range. If the updated downstream node attribute value exceeds the range, a serious delay event is identified.

[0080] Electronic devices can also check whether the updated node attributes of downstream nodes logically contradict other attributes (or physical constraints) of the node itself. If a logical contradiction exists, this contradictory state is identified as a resource status anomaly event. For example, if the updated estimated start time is later than the node's latest completion time, the task cannot be completed logically, thus identifying a performance failure event.

[0081] For example, in an automotive manufacturing supply chain, there are two business links: tire transportation and vehicle assembly. Edge nodes report the real-time status of a physical resource, namely a tire transport vehicle, stating "Tire transport vehicle has malfunctioned and stopped moving." Electronic devices locate the first graph node representing this transport vehicle (e.g., node ID: Truck_01) in a dynamic knowledge graph and update its status attribute to malfunction, locking its location attribute at point A. Next, the electronic devices perform a node search along directed edges with specific labels (assuming the specific label here is: transported to), discovering that the downstream node pointed to by Truck_01 is the "final assembly workshop raw material warehouse" (node ​​ID: Warehouse_Ref_01). The electronic devices then follow the attribute propagation logic: estimated arrival time = current time + (remaining distance / average speed), and when a malfunction is detected in the transportation resource, the average speed is set to unavailable, and the estimated arrival time is updated accordingly. The time interval is marked as unreachable or re-estimated based on the repair time (e.g., estimated arrival time = current time + repair time + (remaining distance / average speed after recovery)). The estimated arrival time is determined to change from "today 10:00" to "unknown" or postponed to "today 18:00" (assuming repair takes 8 hours). Then, the estimated warehouse entry time attribute of the downstream node (Warehouse_Ref_01) is updated according to this information. Finally, the electronic device checks the specific value of the updated node attribute of the downstream node (Warehouse_Ref_01) and finds that the estimated warehouse entry time (18:00) is later than the latest material feeding time of the workshop's production plan (14:00). Therefore, a resource status abnormal event is determined, namely "the final assembly workshop is expected to experience a material shortage and production stoppage at 14:00 today".

[0082] It is understood that, in the embodiments of this application, the electronic device can automatically search for downstream nodes of the first graph node corresponding to the physical resource through directed edge connections in the dynamic knowledge graph, and predict the node attributes of the downstream node and determine the abnormal resource status event based on the attribute transfer logic and the updated node attributes of the first graph node. In this way, the specific impact of the change of the upstream first graph node on the downstream node can be determined in advance, thereby accurately determining the abnormal resource status event and improving the accuracy of the abnormal resource status event.

[0083] In other embodiments of this application, Figure 1 Step 102, which is to determine the abnormal resource status events caused by the real-time status in the future based on the dynamic knowledge graph of the supply chain in the time dimension, can also be achieved through the following processing: from the dynamic knowledge graph, determine the first graph node corresponding to the physical resource, and update the node attributes of the first graph node based on the real-time status; through the graph neural network model, predict the probability distribution of the status of each physical resource in the future based on the dynamic knowledge graph with updated node attributes; determine the predicted status value of each physical resource based on the probability distribution; when the predicted status value exceeds the normal range, determine the abnormal resource status event.

[0084] The graph neural network model here learns the propagation pattern of physical resource state changes within the graph topology over historical time periods. The electronic device receives the dynamic knowledge graph after node attribute updates as input to this graph neural network model. Through multiple layers of nonlinear transformations, the model determines the potential impact of real-time state changes on other nodes in the network over the time dimension, thus outputting a probability distribution of the state of each physical resource in the future time period. The electronic device then samples or performs maximum likelihood estimation on the obtained probability distribution to obtain the predicted state value.

[0085] Step 103: Based on multiple constraint parameters and abnormal resource status events, perform conflict detection on multiple scheduling requests to obtain conflict detection results.

[0086] Since the carrying capacity of physical resources is limited within a specific time span, multiple scheduling requests initiated by different service nodes are highly likely to compete for the same physical resource. Furthermore, because the dynamic knowledge graph has already deduced future resource status anomalies, these anomalies will inevitably further alter the actual availability of physical resources. Therefore, in this embodiment, it is necessary to jointly evaluate the constraint parameters reflecting static service limitations and the resource status anomalies reflecting dynamic temporal risks, and perform conflict verification on all scheduling requests at a global level.

[0087] See Figure 4 , Figure 4This is a flowchart illustrating the collaborative control method for the supply chain provided in this application. Figure 3 In some embodiments of this application, when parsing multiple heterogeneous data packets sent by multiple service nodes using a preset semantic communication protocol, the resource usage range of the call request can also be obtained.

[0088] The resource occupancy range refers to the specific range of physical resources required by the scheduling request in terms of spatial location, time span, or quantity. In this embodiment, the resource occupancy range can be the spatial range occupied by the scheduling request, such as occupying the third row of shelves in warehouse A; the resource occupancy range can also be the time range occupied by the scheduling request, such as occupying the period from June 1, 2024 to June 2, 2024, etc.

[0089] When the resource usage range is obtained by parsing heterogeneous data packets, Figure 1 Step 103, which involves performing conflict detection on multiple scheduling requests based on multiple constraint parameters and abnormal resource status events to obtain the conflict detection results, can be achieved through the following processing: Step 1031: Based on multiple constraint parameters, perform resource conflict detection on multiple scheduling requests to obtain resource conflict detection results.

[0090] It should be noted that resource conflict refers to a situation where multiple scheduling requests operate on the same physical resource, but these scheduling requests are logically contradictory due to the limitations of the actual state of the physical resource. Resource conflict detection is the process of identifying potential resource conflicts.

[0091] In this embodiment of the application, resource conflicts may include resource occupation conflicts, such as two different edge nodes simultaneously requesting to lock the last item A in the same warehouse; resource conflicts may also include time overlap conflicts, such as two work orders simultaneously reserving the same equipment for production during the time period of 10:00-11:00 AM; resource conflicts may also be conflicting conditions, such as scheduling request A requiring express delivery, i.e., time is limited, while its constraint parameters limit the minimum cost, thereby restricting the use of air freight, resulting in no feasible solution under the existing conditions.

[0092] Constraint parameters essentially define the specific boundaries of a scheduling request during execution. They define the exclusivity of the scheduling request on physical resources, quantify the supply and demand relationship of physical resources, and reflect the compatibility of the scheduling request's business logic. Therefore, by using the constraint parameters of different scheduling requests, it is possible to determine whether there are resource conflicts between different scheduling requests.

[0093] In some embodiments of this application, the constraint parameters include physical constraint parameters and business constraint parameters. In this case, the above-mentioned method of performing resource conflict detection on multiple scheduling requests based on multiple constraint parameters to obtain a first detection result can be achieved through the following processing: performing overlap detection on multiple physical constraint parameters of multiple scheduling requests to obtain a first sub-detection result; performing compatibility detection on multiple business constraint parameters of multiple scheduling requests to obtain a second sub-detection result; and determining the resource conflict of multiple scheduling requests based on the first sub-detection result and the second sub-detection result to obtain a resource conflict detection result.

[0094] It should be noted that physical constraint parameters refer to hard, unchangeable limitations determined by the objective properties of physical resources. These parameters reflect the inherent capability boundaries of physical resources in the physical world; requests exceeding physical constraint parameters are physically unfeasible. For example, physical constraint parameters could be the maximum load capacity of a transport vehicle (5 tons), the minimum refrigeration temperature of cold chain equipment (-20°C), etc.

[0095] Business constraint parameters refer to soft restrictions set by supply chain companies based on their management strategies, business rules, compliance requirements, or contractual terms. These restrictions are not directly related to the physical attributes of resources but must be followed. In other words, business constraint parameters reflect human-defined business logic; violating these parameters is non-compliant, but may still be physically permissible. For example, business constraint parameters can be order priority levels (VIP / normal), lists of material suppliers, or rules prohibiting the mixing of goods (such as food and chemicals not being allowed to be mixed).

[0096] In this embodiment, the electronic device can extract the physical constraint parameters of all scheduling requests, project these physical constraint parameters onto the coordinate system of the same physical resource, and calculate their overlap in time, space, or quantity. If the overlap is unacceptable, such as overlapping mutually exclusive intervals or exceeding the total limit, a first sub-detection result representing that the overlap does not meet the requirements is generated.

[0097] Here, for physical constraint parameters in the time dimension, such as start time and end time, the electronic device can construct a timeline graph. The time period of each scheduling request is represented as a line segment on the axis. If two or more mutually exclusive (exclusive resource) line segments are found to have overlapping areas on the timeline, or if the overlapping area is greater than a threshold, a first sub-detection result representing the corresponding situation is generated. For physical constraint parameters related to quantity, such as demand, the electronic device calculates the sum of the demand of all scheduling requests at a specific time. If this sum is greater than the current available amount of physical resources, a first sub-detection result representing that the overlap does not meet the requirements is generated.

[0098] The electronic device performs a coexistence check on the business constraint parameters of each scheduling request within the logic of preset business rules. If they cannot coexist, the device determines that these business constraint parameters are incompatible and generates a second sub-detection result indicating that the compatibility requirement is not met. Here, the electronic device can construct a rule matching matrix. Rows represent the business constraint parameters of one scheduling request (e.g., cold chain transportation), and columns represent the business constraint parameters of another scheduling request (e.g., ordinary truck transportation). The electronic device checks whether the two match. If they do not match, for example, if ordinary trucks cannot achieve cold chain transportation, a second sub-detection result representing that the compatibility requirement is not met is generated.

[0099] After receiving the first and second sub-detection results, the electronic device can generate a resource conflict detection result indicating resource conflicts among multiple scheduling requests if either the first or second sub-detection result fails to meet the requirements. The electronic device can also determine the first detection result as a medium-level resource conflict if only the first sub-detection result fails to meet the requirements, triggering subsequent time adjustments or resource replacements to resolve the conflict; determine the first detection result as a low-level resource conflict if only the second sub-detection result fails to meet the requirements, triggering subsequent resolution through negotiation or authorization; and determine the first detection result as a high-level resource conflict if both the first and second sub-detection results fail to meet the requirements, triggering more advanced conflict resolution actions.

[0100] Here is an example illustrating the process of determining the first detection result. For example, Figure 5 This is a schematic diagram illustrating the principle of resource conflict detection provided in this application embodiment. If two scheduling requests, A and B, simultaneously request the use of the same transport truck (e.g., requesting truck Truck_01), and the physical constraint parameter of scheduling request A is the occupation time of 10:00-12:00, while the physical constraint parameter of scheduling request B is the occupation time of 11:00-13:00, the electronic device performs an overlap detection of the physical parameter constraints on these two requests 5-1. It finds that the two time periods overlap between 11:00 and 12:00, thus obtaining a first sub-detection result 5-3 indicating that the overlap does not meet the requirements. If the business constraint parameter of scheduling request A is "general merchandise" and the business constraint parameter of scheduling request B is "flammable and dangerous goods," the electronic device performs a compatibility detection of the business constraint parameters on these two requests 5-2. It finds that the goods types are incompatible, thus obtaining a second sub-detection result 5-4 indicating that the compatibility does not meet the requirements. Finally, the electronic device makes a comprehensive judgment 5-5 based on the first sub-detection result 5-3 and the second sub-detection result 5-4 to determine that the two scheduling requests cannot be satisfied simultaneously, generating a resource conflict detection result 5-6 indicating that a resource conflict exists.

[0101] It is understood that in this embodiment, the electronic device can determine whether different scheduling requests meet the requirements in both the physical and business dimensions to detect resource conflicts. This not only identifies explicit physical resource overlaps but also implicit incompatibilities in business rules. Thus, the dual-dimensional detection mechanism minimizes missed detections and improves the accuracy of resource conflict detection results.

[0102] In other embodiments of this application, Figure 1 Step 103, which involves performing resource conflict detection on multiple scheduling requests based on multiple constraint parameters to obtain conflict detection results, can also be achieved through the following processing: converting all constraint parameters of the scheduling requests into logical expressions; processing the obtained logical expressions using a constraint satisfaction problem (CSP) solver; if the solver cannot solve the problem, determining that multiple scheduling requests have resource conflicts, and obtaining a resource conflict detection result indicating the existence of resource conflicts; if the solver has a solution, determining that multiple scheduling requests do not have resource conflicts, and obtaining a resource conflict detection result indicating the absence of resource conflicts.

[0103] Step 1032: Based on the resource occupancy range and the resource impact range of abnormal resource status events, perform spatiotemporal conflict detection on each scheduling request and abnormal resource status event to obtain the spatiotemporal conflict detection results.

[0104] The electronic device first extracts the resource occupancy range from each scheduling request and simultaneously determines the resource impact range from resource status anomaly events. Then, it maps these two sets of data to a unified spatiotemporal reference system. Within this system, it analyzes the geometric relationship between the resource occupancy range and the resource impact range. This analysis may include calculating the intersection, analyzing relative positional relationships, and so on. Finally, based on the geometric relationship analysis results, the electronic device determines whether there is a spatiotemporal conflict between the scheduling request and the resource status anomaly event, thus obtaining a second detection result.

[0105] The resource impact range of an abnormal resource status event can be extracted from the node attributes of the corresponding graph node in the dynamic knowledge graph. That is, when an electronic device identifies an abnormal resource status event, it can determine the resource impact range based on its node attributes. For example, by extracting the time interval from the node attributes, the time-related resource impact range can be determined. It should be noted that spatiotemporal conflict refers to the process by which a scheduling request's specific business operation may be interfered with or blocked by potential future resource status anomalies within a given spatiotemporal coordinate system. Spatiotemporal conflict can include time conflicts caused by spatial overlap. For example, a scheduling request plans to pass through road segment A at 14:00, while a resource status anomaly event indicates that road segment A will experience severe congestion from 13:30 to 15:00. The two coincide in spatiotemporal coordinates, causing the scheduling request to fail. Spatiotemporal conflict can also include spatial conflicts caused by time overlap. For example, a scheduling request plans to use equipment B for production, while a resource status event indicates that equipment B will shut down due to overheating during the same period, causing the call request to fail.

[0106] In some embodiments of this application, Figure 4 Step 1032, which is to perform spatiotemporal conflict detection on each scheduling request and resource status anomaly event based on the resource occupancy range and the resource influence range of the resource status anomaly event, and obtain the spatiotemporal conflict detection result, can be achieved by the following processing: calculating the spatiotemporal intersection between the resource occupancy range and the resource influence range; when the spatiotemporal intersection is not empty, determining that there is a spatiotemporal conflict between the scheduling request and the resource status anomaly event, so as to obtain the spatiotemporal conflict detection result.

[0107] Electronic devices can map the resource occupancy range of scheduling requests and the resource impact range of abnormal resource status events to the same spatiotemporal coordinate system. Then, through geometric or algebraic operations, they can calculate whether there is an overlap between the two sets, thus obtaining the spatiotemporal intersection.

[0108] Here, the resource range can be defined as an object in a four-dimensional coordinate system (x, y, z, t). Thus, the electronic device can represent the resource occupancy range as a four-dimensional polyhedron and the resource influence range as another four-dimensional polyhedron. Then, by calculating the volume intersection of these two four-dimensional polyhedra, it can be determined whether there is an overlapping part.

[0109] Resources can be divided into discrete grid cells. Thus, the resource occupancy range can be mapped to a set of grid IDs marked as occupied, while the resource influence range can be mapped to a set of grid IDs marked as abnormal. Electronic devices can determine whether there is overlap by calculating the intersection of these two sets of grid IDs.

[0110] If the spatiotemporal intersection is not an empty set, it means that the two overlap in spatiotemporal space. This means that the scheduling request attempts to use the affected resources at the time and place where the abnormal resource status event occurs. This means that the scheduling request is physically or logically infeasible. Therefore, it is determined that there is a spatiotemporal conflict and the corresponding spatiotemporal conflict detection result is obtained.

[0111] Here is an example to illustrate the process of solving the second detection result. Figure 6 This is a schematic diagram of the spatiotemporal conflict detection process provided in an embodiment of this application. See also... Figure 6 If the resource usage range of a certain call request is 6-1: the transport vehicle will pass through intersection X at 14:00, cross bridge Y at 14:30, and arrive at warehouse Z at 15:00, while the resource status abnormal event, such as the resource impact range of traffic events, is 6-2: bridge Y is closed from 14:15 to 16:00, electronic devices can perform spatiotemporal intersection calculation 6-3 by comparison, and know that the planned arrival time of the transport vehicle at 14:30 falls into the abnormal period of 14:15-16:00, so the spatiotemporal intersection is not empty. Therefore, a spatiotemporal conflict detection result 6-4 will be generated, which indicates that there is a spatiotemporal conflict between the scheduling request and the resource status abnormal event.

[0112] It is understood that, in the embodiments of this application, electronic devices can transform abstract risk assessments into precise geometric operations by using the mathematical means of calculating spatiotemporal intersection, thereby enabling them to determine with high accuracy whether scheduling requests will actually be affected by abnormal resource status events, thus improving the accuracy of spatiotemporal conflict detection results.

[0113] In other embodiments of this application, Figure 4 Step 1032, which involves performing spatiotemporal conflict detection on each scheduling request and resource status anomaly event based on the resource occupancy range and the resource impact range of the resource status anomaly event, to obtain the spatiotemporal conflict detection result, can also be achieved through the following processing: mapping the resource impact range of the resource status anomaly event to a risk probability field that decays with distance and time, and treating the resource occupancy range of the scheduling request as a path traversing this field; calculating the integral of the risk probability along this path to obtain the cumulative risk value; if the cumulative risk value exceeds the risk threshold, it is determined that a spatiotemporal conflict exists, and a spatiotemporal conflict detection result characterizing the existence of a spatiotemporal conflict is generated.

[0114] This approach can be used in scenarios where risk boundaries are blurred or gradually spread, such as the impact of severe weather.

[0115] Step 1033: When at least one of the representations in the resource conflict detection result and the spatiotemporal conflict detection result is in conflict, generate a conflict detection result in which the representations are in conflict.

[0116] When either the resource conflict detection result or the spatiotemporal conflict detection result is in conflict, or when both the resource conflict detection result and the spatiotemporal conflict detection result are in conflict, the electronic device will determine that multiple scheduling requests are in conflict and obtain the corresponding conflict detection result.

[0117] Step 104: When the conflict detection results indicate that a conflict exists, determine the optimal action from the set of candidate actions corresponding to the conflict.

[0118] After receiving the collision detection result, the electronic device will confirm whether the result indicates the existence of a collision. Once a collision is detected, the device will initiate a collision resolution decision process. During collision resolution, the electronic device will determine a set of candidate actions for the collision, and then evaluate each candidate action to determine the optimal action.

[0119] It should be noted that the candidate action set is a set of alternative actions that the electronic device can physically and logically generate for the detected conflict, and each element in the set is a candidate action.

[0120] The candidate action set can be automatically generated by the electronic device based on conflict type mapping and rule template matching. In this process, the electronic device first identifies the conflict type of the conflict detection result, such as time overlap or insufficient inventory. Then, it searches for feasible solution templates for this type of conflict in a pre-built conflict-strategy knowledge base. Based on the current conflict context, such as the resource ID involved and the conflict magnitude, it instantiates the abstract strategy template into specific candidate actions. For example, if the conflict is a shortage of raw materials, and the conflict context is a shortage of 50 units of raw materials, then the electronic device can retrieve the strategies: transfer and replenishment, and instantiate them to generate a candidate action set: {Action A: Transfer 50 units from warehouse B, Action B: Emergency replenishment of 50 units from the supplier}.

[0121] In some embodiments of this application, Figure 1 Step 104, determining the optimal action from the candidate action set corresponding to the conflict, can be achieved through the following process: from the dynamic knowledge graph, determine the second graph node related to the conflict, and construct the conflict state vector based on the node attributes of the second graph node; through the reward model, calculate the reward value for each candidate action in the candidate action set based on the state vector; and determine the candidate action corresponding to the largest reward value as the optimal action.

[0122] In other words, the electronic device first locates the corresponding graph node of the conflict in the dynamic knowledge graph, i.e., the second graph node. Then, based on the node attributes of the second graph node, it generates a vector form that the model can read. This second graph node can be the graph node corresponding to the physical resource involved in the conflict, or it can be an upstream or downstream node of that graph node. The electronic device pairs the state vector with each candidate action in the candidate action set, thus obtaining an input pair with the same number of candidate actions. It then reads this input pair through the reward model to begin processing, and uses the output of the reward model as the reward value. Finally, the electronic device determines the largest reward value from multiple reward values ​​and directly uses the candidate action corresponding to the largest reward value as the optimal action.

[0123] It should be noted that the degree to which a candidate action improves the physical scenario of the conflict after execution refers to the extent to which various indicators of the physical scenario of the supply chain improve after the candidate action is executed. Therefore, the reward value output by the reward model is directly proportional to the degree to which each candidate action improves the physical scenario of the conflict after execution. The larger the reward value, the greater the contribution of the corresponding candidate action to the conflict in the supply chain after execution. Conversely, the smaller the reward value, the smaller the contribution of the corresponding candidate action to the conflict in the supply chain.

[0124] In this embodiment of the application, the reward model can be a weighted function containing multiple improvement indicators, for example, equation (1): (1) in, The benefits that candidate actions can bring. The execution cost of the candidate action. Improve latency for conflicting candidate actions. Risks associated with candidate actions , , and All are weighted weights. Here, each indicator is normalized / dimensionless before calculating the reward value.

[0125] The reward model can also be a deep Q-network, whose inputs are a state vector and candidate actions, and whose output is the expected long-term cumulative reward for each candidate action.

[0126] It should be noted that, based on the node attributes of the second graph nodes, the conflict state vector can be constructed by extracting the key attribute values ​​of the second graph nodes (such as inventory level and temperature) and the attributes of the associated edges (such as transportation time). The electronic device directly concatenates these values ​​into a one-dimensional array in sequence to serve as the state vector. Alternatively, the conflict state vector can also be constructed based on the node attributes of the second graph nodes through a graph representation learning algorithm. That is, the electronic device can map the second graph nodes and their neighborhood structure into a fixed-length low-dimensional dense vector based on the topology and attributes of the dynamic knowledge graph to serve as the state vector.

[0127] For example, if the conflict is due to insufficient inventory, and the context of the conflict is a shortage of 100 items, the electronic device can first determine that the relevant second graph nodes are warehouse W and its upstream suppliers S1 and S2. Then, based on the attributes of these second graph nodes, namely, warehouse W's current inventory = 0, supplier S1 distance = 10km (near), supplier S2 distance = 500km (far), and current time = 17:00 (evening peak), a state vector S = [0, 10, 500, 17:00] is constructed. Then, based on this state vector, the electronic device calculates the reward value for each candidate action in the candidate action set {Action A: Emergency replenishment to S1, Action B: Emergency replenishment to S2}. For Action A, although its unit price is slightly higher, it can be delivered within 1 hour, so the calculated reward value is 0.9. For Action B, although S2's unit price is lower, it requires cross-city transportation and encounters the evening peak, with delivery expected tomorrow, so the calculated reward value is 0.4. Therefore, the electronic device will select Action A as the optimal action.

[0128] It is understood that in this embodiment of the application, the electronic device extracts second graph nodes from the dynamic knowledge graph and constructs state vectors to ensure that the decision-making process can take into account the deep context behind the conflict. At the same time, it limits the reward value to be proportional to the degree of physical improvement, ensuring that the selected optimal action can maximize efficiency in actual business, thereby improving the accuracy of the selection of the optimal action.

[0129] In other embodiments of this application, Figure 1 Step 104, determining the optimal action from the candidate action set corresponding to the conflict, can also be achieved through the following processing: calculating multiple evaluation values ​​for each candidate action under multiple objectives; determining the Pareto front solution set from the candidate action set based on the multiple evaluation values; and determining the optimal action from the Pareto front solution set according to the preset business strategy.

[0130] It's important to note that in multi-objective optimization, if solution A is no worse than solution B on all objectives, and at least one objective is better than the other, then A is said to dominate B. Conversely, if no other solution can dominate A, then A is said to be a non-dominated solution. The Pareto front solution set includes the non-dominated candidate actions in the candidate action set. Therefore, the Pareto front solution set is the set of all non-dominated candidate actions. This set can be viewed as the set of all optimal trade-offs, where any action in this set that improves the performance of one objective must sacrifice the performance of another.

[0131] It should also be noted that the multiple objectives include at least: scheduling timeliness objectives, physical resource cost objectives, and business compliance objectives. Among them, scheduling timeliness objectives are optimization indicators that measure the efficiency of actions in the time dimension, aiming to minimize task completion time or delay time; physical resource cost objectives are optimization indicators that measure the cost of actions in the resource consumption dimension, aiming to minimize resource consumption; and business compliance objectives are optimization indicators that measure the degree to which actions comply with contractual agreements, laws and regulations, or corporate policies, aiming to maximize compliance scores or minimize the risk of violations.

[0132] In this embodiment of the application, the electronic device can use the calculation formula of each target, such as timeliness = action distance / average speed, cost value = action unit price * quantity, and compliance value = Lookup (supplier credit table), to calculate the evaluation value under different targets. Alternatively, the candidate action can be substituted into the digital twin model for simulation and the scheduling implementation, physical resource cost and compliance record number at the end of the simulation can be recorded as the evaluation value.

[0133] After obtaining multiple evaluation values, the electronic device can compare the evaluation values ​​of different candidate actions to determine the performance of all candidate actions on the multi-dimensional objective. Then, it eliminates dominated solutions, retaining only those solutions that have an irreplaceable advantage in at least one dimension to form the Pareto front solution set. Here, the electronic device can perform pairwise comparisons of candidate actions in the candidate action set, recording how many opponents each candidate action dominates (obtaining the dominance count) and how many opponents dominate each candidate action (obtaining the dominated count). It then sets the Pareto front solution set for all candidate actions with a dominated count of 0. Alternatively, the electronic device can iterate through all candidate actions. For a candidate action A, it checks whether there exists another candidate action B that makes candidate action B superior to A on all objectives. If not, candidate action A is added to the Pareto front solution set.

[0134] There may be more than one candidate action in the Pareto front solution set. Therefore, electronic devices can also determine a candidate action from the Pareto front solution set based on the current preset business strategy, as the final optimal action. Here, the preset business strategy can be understood as a preference. For example, if the preference is to ensure delivery, then the candidate action with the highest timeliness is scheduled as the optimal action; or if the preference is to ensure revenue, then the candidate action with the lowest physical resource cost is selected as the optimal action.

[0135] For example, regarding the conflict of insufficient transport capacity, the corresponding candidate action set includes three candidate actions: air transport, high-speed train, and regular train. Action A is air transport, with a transit time of 3 hours and a cost of 5000 yuan, and high compliance. Action B is high-speed train, with a transit time of 6 hours and a cost of 1000 yuan, and high compliance. Action C is regular train, with a transit time of 24 hours and a cost of 2000 yuan, and medium compliance. The electronic device first compares actions A and B: action A is fast but costly, while action B is slow but costly, and they do not dominate each other. Then, it compares actions B and C: action B is not only faster than action C (6 < 24), but also cheaper (1000 < 2000), and has better compliance. Therefore, action B dominates action C, and the resulting Pareto front solution set is {action A, action B}. Afterwards, the electronic device, according to a preset business strategy (prioritizing timeliness), selects action A as the final optimal action.

[0136] It is understood that in the embodiments of this application, the electronic device first determines multiple evaluation values ​​of candidate actions under multiple objectives, and then eliminates poorly performing solutions through the Pareto front screening mechanism, thereby greatly narrowing the range of selectable candidate actions to obtain the optimal solution or the best balance point in different dimensions. Then, it screens from these candidate actions to ensure that the obtained optimal action is the optimal solution or the best balance point in a certain dimension, thereby ensuring the accuracy of the optimal action.

[0137] Step 105: Based on the optimal action and the preset semantic communication protocol, generate execution instructions and send the execution instructions to multiple business nodes.

[0138] After determining the optimal action, the electronic device generates execution instructions based on a preset semantic communication protocol and then sends these instructions to each business node to achieve collaborative control. Here, the electronic device can directly encapsulate the optimal action according to the preset semantic communication protocol to obtain the execution instructions, or it can encapsulate the optimal action together with idempotent control keys, exception rollback strategies, etc., to obtain execution quality.

[0139] In other words, in some embodiments of this application, Figure 1 Step 105, namely generating an execution instruction based on the optimal action and a preset semantic communication protocol, can be achieved through the following processing: for the optimal action, determine the idempotent control key that uniquely identifies the execution instruction, as well as the exception rollback strategy when the execution of the optimal action encounters an exception; encapsulate the optimal action, the exception rollback strategy, and the idempotent control key into an execution instruction according to the preset semantic communication protocol.

[0140] It should be noted that the idempotent control key refers to a unique anti-duplicate token string encapsulated in the execution instruction, used to uniquely identify the operation request in a distributed network environment. This string can be a unique string generated by the electronic device based on its UUID. The exception rollback strategy refers to the handling logic pre-encapsulated in the execution instruction, explicitly specifying how edge nodes should automatically restore the system state when the optimal action fails, times out, or partially succeeds. This logic can be obtained by consulting a relevant knowledge base.

[0141] After determining the optimal action, the electronic device first generates an instruction header and an instruction body. Then, based on the standard message structure defined by a pre-defined semantic communication protocol, the electronic device fills the corresponding fields of the instruction body with the business parameters of the optimal action, such as resource ID, quantity, and time. The electronic device generates a globally unique idempotent control key and embeds it into the instruction header. Based on the type of the optimal action, the electronic device matches the corresponding exception rollback policy from a pre-defined policy library and encapsulates the relevant parameters of the exception rollback policy into the instruction body. Finally, the electronic device concatenates the instruction header and instruction body into an execution instruction and pushes the resulting execution instruction to the message bus or API gateway. Based on the target address of the instruction (i.e., the address of the edge node to which the physical resource belongs), the electronic device routes and distributes the execution instruction to multiple edge nodes distributed across various business systems, enabling the edge nodes to drive their business systems to execute the optimal action, thereby achieving collaborative control of the supply chain.

[0142] Understandably, compared to related technologies, there are issues with the accuracy and reliability of supply chain collaborative control. In this embodiment, a pre-defined semantic communication protocol is first used to parse heterogeneous data packets from business nodes distributed across different business systems. This yields information rich in semantics, including business nodes' scheduling requests for physical resources, the constraint parameters of these requests, and the real-time status of the physical resources. Furthermore, based on a dynamic knowledge graph with a time dimension, temporal reasoning is performed on the current state to identify potential future resource status anomalies. Then, based on the different constraint parameters of different scheduling requests and the resource status anomalies, it is determined whether each scheduling request will be affected by future resource status anomalies, achieving comprehensive conflict detection. If any conflict exists, the optimal action is determined from the set of candidate actions corresponding to the conflict, and an execution instruction is generated based on this optimal action. This enables collaborative supply chain control based on semantically rich information, fully considering the impact of future anomalies on scheduling requests, thus allowing for more accurate selection of the optimal action and more precise collaborative control. In summary, this embodiment improves the accuracy of supply chain collaborative control.

[0143] The following will describe an exemplary application of the embodiments of this application in a real-world application scenario.

[0144] The embodiments of this application are implemented in the scenario of collaborative decision-making and conflict resolution in the supply chain.

[0145] The structure of the supply chain collaborative decision-making and conflict resolution system based on a multi-layered, multi-agent architecture provided in this application embodiment is described below. This system includes business system-side edge agent nodes, a message bus and protocol layer, an agent cluster, and a dynamic knowledge graph storage and inference module. The hardware devices of the business system-side edge agent nodes may include industrial control computers, edge gateways, and virtual machines, with network interfaces connected to the enterprise intranet. The software modules of the business system-side edge agent nodes include connectors, field mappers, semantic converters, signature and authentication modules, etc. The business system-side edge agent nodes (referred to as business nodes) are used to read status from their respective business systems, such as ERP / WMS / MES / TMS, and encapsulate them into standard messages (referred to as heterogeneous data messages), as well as receive instructions from the upper layer (referred to as execution instructions) and convert them into executable transaction operations for each business system. The hardware devices of the message bus and protocol layer can be servers or cloud message services, and the software modules include message queues. The message bus and protocol layer are used to carry asynchronous communication between various agents. The hardware for the intelligent agent cluster includes multiple servers, and the software system may include a microservice deployment platform, which is used to deploy L1 / L2 / L3 intelligent agent microservices and conflict resolution engines (the L2 / L3 intelligent agent microservices and conflict resolution engines can run on servers, which are referred to as electronic devices). The hardware for the dynamic knowledge graph storage and inference module may include a graph database server, and the software module may include temporal attribute extensions, rule / graph query engines, etc., which are used to store the knowledge graph and perform incremental updates and subscription triggers.

[0146] In this embodiment, semantic exchange is achieved through a message structure (referred to as a standard message structure) defined by a standardized semantic communication protocol (referred to as a preset semantic communication protocol). This message structure may include the following fields: Performative (also known as the scheduling request field): Values ​​can be Request / Report / Empower / Reject / CounterOffer; OntologyVersion: Semantic ontology version number; SenderAgentId / ReceiverAgentId: Identifier of the sending / receiving agent; CorrelationId: Session association ID; IdempotencyKey: Idempotency key (used for deduplication and replay protection); Content (also known as the status field): Structured business workload (e.g., material, quantity, location, time window, cost limit, etc.). This field can also be called the business workload field. Constraints (also known as constraint parameter fields): a set of constraints (capacity calendar, inventory holding rules, shipping windows, compliance restrictions); Auth: Signature / Certificate / Permissions Declaration; Deadline: The deadline for negotiation; AuditTrail: Audit fields (traceability, version, rollback point).

[0147] Through the above message structure, different business systems can not only exchange data, but also exchange semantic information such as "intent, constraints, and authorization".

[0148] The following describes the specific process of collaborative decision-making and conflict resolution. This process includes: Step 1: Agent Initialization and Semantic Adaptation At the L1 execution layer, there are intelligent agents such as procurement intelligent agents, inventory intelligent agents, production intelligent agents, and logistics intelligent agents. These intelligent agents are used to process the data of their business systems. For example, they read the values ​​of fields in their own business systems (such as the on_hand and allocated fields of WMS), assemble these field values ​​according to the message structure defined by the above protocol to obtain message data, and then write the message data into the corresponding nodes, edges, and attributes of the dynamic knowledge graph.

[0149] Step 2: Dynamic knowledge graph construction and incremental updates.

[0150] The control tower layer or L2 coordination layer updates the dynamic knowledge graph based on data packets and triggers inference rules to generate risk reports (referred to as resource status anomaly events) when node attributes change.

[0151] Step 3: Conflict detection.

[0152] The control tower layer or L2 coordination layer is configured with a conflict detector. The conflict detector shall detect at least the following types of conflicts: resource conflicts, such as the same inventory batch being locked by multiple orders at the same time, or the same production capacity period being competed for by multiple work orders; target conflicts, such as the contradictory solutions proposed by multiple agents; and constraint conflicts, such as solutions violating hard constraints, such as cold chain temperature control and dangerous goods transportation restrictions.

[0153] The conflict detector then organizes the conflicts into a ConflictCase data structure for subsequent processing. The ConflictCase data structure may include fields such as: the set of participating agents, conflict resources, candidate actions, set of hard constraints, evaluation metric vector, and deadline.

[0154] Step 4: Conflict Resolution 1. Conflict resolution based on reinforcement learning.

[0155] The conflict resolution engine learns a state vector S by extracting features (inventory gap, order priority, penalties, switching costs, transportation capacity, supply risks, etc.) from a knowledge graph. Then, it maps each action in the action set A (called the candidate action set) to an executable instruction of the business system. For example, the action set A may include actions such as air freight, cross-warehouse transfer, and order insertion. Next, the conflict resolution engine calculates the reward value based on the reward function of equation (1), and determines the optimal action or optimal action set based on the reward value to generate the instruction message Empower (containing the authorization scope and rollback point).

[0156] 2. Conflict resolution based on Pareto multi-objective optimization.

[0157] The conflict resolution engine calculates the evaluation value of each action in the action set A, and the evaluation value under multiple objectives. These objectives may include: minimizing cost, minimizing latency, minimizing risk, and maximizing service level. Then, it generates a Pareto frontier candidate set based on the evaluation values ​​under multiple objectives. Finally, according to the strategy (called the preset business strategy), such as order level and risk preference, it selects the final action and outputs the instruction message Empower.

[0158] Step 5: Implement closed-loop and consistency guarantees.

[0159] After receiving the Empower instruction message, the intelligent agent of the edge agent node translates it into an operation executable by the corresponding business system. For example, it might execute an inventory locking transaction in the WMS or issue a work order adjustment in the MES. The instruction message includes an idempotent key (called an idempotent control key) to prevent duplicate execution, as well as a rollback strategy for critical operations (called an exception rollback strategy) to enable rollback in case of exceptions.

[0160] The above scheme enables semantic-based negotiation and prediction of future risk events. Based on the prediction of risk events, it can identify potential conflicts and then resolve them, thereby comprehensively improving the accuracy of collaborative control.

[0161] The supply chain collaborative control method based on the above embodiments, Figure 7This is a structural diagram of a supply chain collaborative control device provided in an embodiment of this application. The supply chain collaborative control device 100 can be a device in an electronic device (e.g., a server). The supply chain collaborative control device can be implemented in software, which can be software in the form of programs and plug-ins, including the following software modules: data parsing module 1001, event determination module 1002, conflict detection module 1003, action determination module 1004, and instruction generation module 1005. These modules are logical, so they can be arbitrarily combined or further split according to the functions implemented.

[0162] The data parsing module 1001 is used to parse multiple heterogeneous data packets sent by multiple business nodes using a preset semantic communication protocol, so as to obtain the scheduling request of the business node for physical resources in the supply chain, the constraint parameters of the scheduling request, and the real-time status of the physical resources. The event determination module 1002 is used to determine, based on the dynamic knowledge graph of the supply chain in the time dimension, the abnormal resource status event caused by the real-time status in the future. The conflict detection module 1003 is used to perform conflict detection on multiple scheduling requests based on multiple constraint parameters and resource status abnormal events, and obtain conflict detection results. The action determination module 1004 is used to determine the optimal action from the candidate action set corresponding to the conflict when the conflict detection result indicates that a conflict exists; The instruction generation module 1005 is used to generate an execution instruction based on the optimal action and the preset semantic communication protocol, and send the execution instruction to multiple service nodes.

[0163] In the above scheme, the event determination module 1002 is further configured to: determine the first graph node corresponding to the physical resource from the dynamic knowledge graph; update the node attributes of the first graph node based on the real-time state; search for downstream nodes affected by the first graph node from the dynamic knowledge graph according to the directed edges representing physical dependencies in the dynamic knowledge graph; update the node attributes of the downstream nodes based on the updated node attributes of the first graph node and the attribute passing logic of the directed edges; and determine the resource state abnormal event based on the updated node attributes of the downstream nodes.

[0164] In the above scheme, when parsing multiple heterogeneous data packets sent by multiple business nodes using a preset semantic communication protocol, the resource usage range of the call request is also obtained. The conflict detection module 1003 is further configured to perform resource conflict detection on multiple scheduling requests based on multiple constraint parameters to obtain resource conflict detection results; perform spatiotemporal conflict detection on each scheduling request and the resource status abnormal event based on the resource occupancy range and the resource influence range of the resource status abnormal event to obtain spatiotemporal conflict detection results; and generate the conflict detection result with conflicting representations when at least one of the resource conflict detection results and the spatiotemporal conflict detection results is in conflict.

[0165] In the above scheme, the constraint parameters include physical constraint parameters and business constraint parameters; the conflict detection module 1003 is further configured to perform overlap detection on multiple physical constraint parameters of multiple scheduling requests to obtain a first sub-detection result; perform compatibility detection on multiple business constraint parameters of multiple scheduling requests to obtain a second sub-detection result; and determine the resource conflict of multiple scheduling requests based on the first sub-detection result and the second sub-detection result to obtain the resource conflict detection result.

[0166] In the above scheme, the conflict detection module 1003 is further used to calculate the spatiotemporal intersection between the resource occupancy range and the resource influence range; when the spatiotemporal intersection is not empty, it is determined that there is a spatiotemporal conflict between the scheduling request and the resource status abnormal event, so as to obtain the spatiotemporal conflict detection result.

[0167] In the above scheme, the action determination module 1004 is further configured to determine the second graph node related to the conflict from the dynamic knowledge graph, and construct the state vector of the conflict based on the node attributes of the second graph node; calculate a reward value for each candidate action in the candidate action set based on the state vector through a reward model, wherein the reward value is proportional to the degree of improvement of the physical scene of the conflict after each candidate action is executed; and determine the candidate action corresponding to the largest reward value as the optimal action.

[0168] In the above scheme, the action determination module 1004 is further configured to calculate multiple evaluation values ​​under multiple objectives for each candidate action in the candidate action set; wherein the multiple objectives include at least: scheduling timeliness objective, physical resource cost objective, and business compliance objective; based on the multiple evaluation values, determine a Pareto front solution set from the candidate action set, wherein the Pareto front solution set includes non-dominated candidate actions in the candidate action set; and determine the optimal action from the Pareto front solution set according to a preset business strategy.

[0169] In the above scheme, the instruction generation module 1005 is further configured to determine, for the optimal action, an idempotent control key that uniquely identifies the execution instruction, and an exception rollback strategy when the execution of the optimal action encounters an exception; and to encapsulate the optimal action, the exception rollback strategy, and the idempotent control key into the execution instruction according to a preset semantic communication protocol.

[0170] It should be noted that the description of the apparatus in this application embodiment is similar to the description of the method embodiment described above, and has similar beneficial effects as the method embodiment; therefore, it will not be repeated. For technical details not disclosed in this apparatus embodiment, please refer to the description of the method embodiment of this application for understanding.

[0171] This application provides an electronic device. Figure 8 This is a schematic diagram of the structure of the electronic device provided in an embodiment of this application. For example... Figure 8 As shown, the electronic device 130 includes: at least one processor 131 ( Figure 8 (Only one is shown in the image), memory 132, and computer-executable instructions 133 stored in memory 132 and executable on at least one processor 131, wherein the processor 131 executes the computer-executable instructions 133 to implement the steps in the embodiments of the collaborative control methods for any of the above-described supply chains.

[0172] The electronic device may include, but is not limited to, a processor 131 and a memory 132. Those skilled in the art will understand that... Figure 8 This is merely an example of electronic device 130 and does not constitute a limitation on electronic device 130. It may include more or fewer components than shown, or combine certain components, or different components, such as input / output devices, network access devices, etc.

[0173] Processor 131 can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0174] In some embodiments, memory 132 may be an internal storage unit of electronic device 130, such as a hard disk or memory of electronic device 130. In other embodiments, memory 132 may be an external storage device of electronic device 130, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on electronic device 130. Furthermore, memory 132 may include both internal and external storage units of electronic device 130. Memory 132 is used to store operating system, application programs, boot loader, data, and other programs, such as program code of computer programs. Memory 132 may also be used to temporarily store data that has been output or will be output.

[0175] This application provides a computer program product comprising a computer program or computer-executable instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer-executable instructions from the computer-readable storage medium and executes the computer-executable instructions, causing the electronic device to perform the supply chain collaborative control method described above in this application.

[0176] This application provides a computer-readable storage medium storing computer-executable instructions or a computer program. When the computer-executable instructions or the computer program are executed by a processor, the processor will execute the supply chain collaborative control method provided in this application, for example... Figure 1 The method of collaborative control of the supply chain is shown.

[0177] In some embodiments, the computer-readable storage medium may be a memory such as RAM, ROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or it may be a variety of devices including one or any combination of the above-mentioned memories.

[0178] In some embodiments, computer-executable instructions may take the form of programs, software, software modules, scripts, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as stand-alone programs or as modules, components, subroutines, or other units suitable for use in a computing environment.

[0179] As an example, computer-executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple co-located files (e.g., files that store one or more modules, subroutines, or code sections).

[0180] As an example, computer-executable instructions can be deployed to execute on a single electronic device, or on multiple electronic devices located at one location, or on multiple electronic devices distributed across multiple locations and interconnected via a communication network.

[0181] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of this application are included within the scope of protection of this application.

Claims

1. A method of collaborative control of a supply chain, characterized by, The method includes: Using a preset semantic communication protocol, multiple heterogeneous data packets sent by multiple business nodes are parsed to obtain the scheduling request of the business node for physical resources in the supply chain, the constraint parameters of the scheduling request, and the real-time status of the physical resources. Based on the dynamic knowledge graph of the supply chain in the time dimension, determine the abnormal resource status events caused by the real-time status in the future. Based on multiple constraint parameters and resource status anomaly events, conflict detection is performed on multiple scheduling requests to obtain conflict detection results. When the conflict detection results indicate that a conflict exists, the optimal action is determined from the candidate action set corresponding to the conflict. Based on the optimal action and the preset semantic communication protocol, an execution instruction is generated and sent to multiple service nodes.

2. The method of claim 1, wherein, The determination of resource status anomaly events caused by the real-time state in the future, based on the dynamic knowledge graph of the supply chain in the time dimension, includes: From the dynamic knowledge graph, a first graph node corresponding to the physical resource is determined, and the node attributes of the first graph node are updated based on the real-time status. Based on the directed edges representing physical dependencies in the dynamic knowledge graph, search for downstream nodes affected by the nodes in the first graph from the dynamic knowledge graph; Based on the updated node attributes of the first graph node and the attribute passing logic of the directed edge, update the node attributes of the downstream node; Based on the updated node attributes of the downstream node, the abnormal resource status event is determined.

3. The method of claim 1, wherein, When parsing multiple heterogeneous data packets sent by multiple service nodes using a preset semantic communication protocol, the resource occupancy range of the scheduling request is also obtained. The method involves performing conflict detection on multiple scheduling requests based on multiple constraint parameters and resource status anomaly events to obtain conflict detection results, including: Based on multiple constraint parameters, resource conflict detection is performed on multiple scheduling requests to obtain resource conflict detection results; Based on the resource occupancy range and the resource impact range of the abnormal resource status event, spatiotemporal conflict detection is performed on each scheduling request and the abnormal resource status event to obtain spatiotemporal conflict detection results; When at least one of the resource conflict detection results and the spatiotemporal conflict detection results conflicts, the conflict detection result with conflicting representations is generated.

4. The method of claim 3, wherein, The constraint parameters include: physical constraint parameters and business constraint parameters; the step of performing resource conflict detection on multiple scheduling requests based on multiple constraint parameters to obtain resource conflict detection results includes: Overlap detection is performed on multiple physical constraint parameters of multiple scheduling requests to obtain a first sub-detection result; A second sub-detection result is obtained by performing compatibility checks on multiple service constraint parameters of multiple scheduling requests. Based on the first sub-detection result and the second sub-detection result, resource conflicts of multiple scheduling requests are determined to obtain the resource conflict detection result.

5. The method of claim 3, wherein, Based on the resource occupancy range and the resource impact range of the abnormal resource status event, spatiotemporal conflict detection is performed on each scheduling request and the abnormal resource status event to obtain spatiotemporal conflict detection results, including: Calculate the spatiotemporal intersection between the resource occupancy range and the resource influence range; When the spatiotemporal intersection is not empty, it is determined that there is a spatiotemporal conflict between the scheduling request and the resource status abnormal event, so as to obtain the spatiotemporal conflict detection result.

6. The method according to any one of claims 1 to 5, characterized in that, The preset semantic communication protocol defines a standard message structure for carrying the heterogeneous data packets. The standard message structure includes at least: a scheduling request field, a constraint parameter field, and a status field. The scheduling request field is used to define the type of the scheduling request; The constraint parameter field is used to define the constraint parameters that the scheduling request must satisfy when it is executed; The status field is used to carry the value of the real-time status of the physical resource or the device identifier.

7. The method according to any one of claims 1 to 5, characterized in that, Determining the optimal action from the set of candidate actions corresponding to the conflict includes: From the dynamic knowledge graph, determine the second graph node related to the conflict, and construct the state vector of the conflict based on the node attributes of the second graph node; Using a reward model, a reward value is calculated for each candidate action in the candidate action set based on the state vector. The reward value is proportional to the degree of improvement of the conflicting physical scenario after each candidate action is executed. The candidate action corresponding to the highest reward value is determined as the optimal action.

8. The method according to any one of claims 1 to 5, characterized in that, Determining the optimal action from the set of candidate actions corresponding to the conflict includes: For each candidate action in the candidate action set, multiple evaluation values ​​are calculated under multiple objectives; wherein, the multiple objectives include at least: scheduling timeliness objective, physical resource cost objective, and business compliance objective; Based on multiple evaluation values, a Pareto front solution set is determined from the candidate action set, wherein the Pareto front solution set includes non-dominated candidate actions in the candidate action set; The optimal action is determined from the Pareto front solution set according to the preset business strategy.

9. The method according to any one of claims 1 to 5, characterized in that, The step of generating execution instructions based on the optimal action and the preset semantic communication protocol includes: For the optimal action, determine the idempotent control key that uniquely identifies the execution instruction, and the exception rollback strategy when the execution of the optimal action fails. The optimal action, the exception rollback strategy, and the idempotent control key are encapsulated into the execution instructions according to a preset semantic communication protocol.

10. A cooperative control device of a supply chain, characterized by, The device includes: The data parsing module is used to parse multiple heterogeneous data packets sent by multiple business nodes using a preset semantic communication protocol, so as to obtain the scheduling request of the business node for physical resources in the supply chain, the constraint parameters of the scheduling request, and the real-time status of the physical resources. The event determination module is used to determine, based on the dynamic knowledge graph of the supply chain in the time dimension, the abnormal resource status events caused by the real-time status in the future. The conflict detection module is used to perform conflict detection on multiple scheduling requests based on multiple constraint parameters and resource status abnormal events, and obtain conflict detection results; The action determination module is used to determine the optimal action from the set of candidate actions corresponding to the conflict when the conflict detection result indicates that a conflict exists; The instruction generation module is used to generate an execution instruction based on the optimal action and the preset semantic communication protocol, and send the execution instruction to multiple service nodes.

11. An electronic device, characterized in that, The electronic device includes: Memory is used to store executable instructions or computer programs. A processor, when executing computer-executable instructions or computer programs stored in the memory, implements the method according to any one of claims 1 to 9.

12. A computer-readable storage medium storing computer-executable instructions or a computer program, characterized in that, When the computer-executable instructions or computer program are executed by a processor, they implement the method described in any one of claims 1 to 9.

13. A computer program product comprising computer-executable instructions or a computer program, characterized in that, When the computer-executable instructions or computer program are executed by a processor, they implement the method according to any one of claims 1 to 9.

Images