A closed-loop method and system for structural optimization of industrial logistics

Abstract

This invention relates to the field of structural optimization technology for logistics systems, and in particular discloses a closed-loop method and system for structural-level optimization of industrial logistics. The method acquires and aligns multi-source operational data in real time to construct a structural state diagram of the logistics system, including nodes and relational edges. Based on this structural state diagram, it identifies structural imbalance units that lead to repetitive inefficiencies. For each imbalance unit, it generates a structural change package consisting of changes to topology, resources, and rules. It combines a baseline scenario with the structural change package to form a simulable scenario variant, and performs simulation evaluation and side effect detection. Finally, based on the evaluation results, it selects effective changes, forms a new structural state baseline, and proceeds to the next round of optimization. This invention achieves a closed-loop process from data to structural expression, imbalance identification, scheme generation, simulation verification, to baseline evolution, solving the technical problem of existing technologies' difficulty in performing structural-level optimization and continuous evolution, and significantly improving the adaptive optimization capability of logistics systems.

Description

Technical Field

[0001] This invention relates to the field of structural optimization technology for logistics systems, and in particular to a closed-loop method and system for structural optimization of industrial logistics. Background Technology

[0002] Currently, the operation of industrial internal logistics systems (such as manufacturing workshops and warehousing centers) generally relies on various scheduling systems or management software for task allocation and execution. However, existing technical solutions mostly focus on "operational-level optimization" or "local scheduling optimization," such as AGV path planning, task prioritization, and improving equipment utilization. While these solutions can achieve certain results in addressing single-point problems, they suffer from the following fundamental flaws:

[0003] First, there is a lack of a unified expression for the "structural level" of the logistics system. Existing methods cannot integrate elements such as regional layout, route network, resource allocation, and operational rules into a computable digital model, making it difficult to fully perceive and quantify the overall operational status of the system.

[0004] Second, it is difficult to identify the "structural imbalances" that lead to repetitive inefficiencies. In actual operation, many inefficiencies (such as periodic congestion, high resource empty running rates, and frequent material shortages) are not caused by single scheduling errors, but by structural defects such as insufficient regional capacity, path network conflicts, and mismatch of resource service ranges. Existing technologies lack the means to pinpoint these structural root causes.

[0005] Third, the optimization solutions remain at the level of local adjustments and cannot form verifiable "structural change solutions". When structural problems are identified, existing methods usually rely on human experience to make local repairs (such as adding an AGV or adjusting a path), and cannot systematically generate structural change solutions that include topology reconstruction, resource reallocation, and rule reset, let alone quantify and verify their effects and side effects before implementation.

[0006] Fourth, there is a lack of a closed-loop mechanism for "simulation verification - side effect detection". Any change to the structure of the logistics system (such as migrating buffer zones or reconstructing the route network) may trigger unpredictable chain reactions, shifting congestion or bottlenecks to other areas. Existing technologies do not provide technical means to "deploy behavior" and "detect side effects" of structural changes in a virtual environment, resulting in high trial-and-error costs and risks in structural optimization.

[0007] Fifth, it cannot achieve continuous evolution of the structural baseline. As production rhythm changes, products are iterated, and layouts are adjusted, the "optimal structure" of the logistics system evolves dynamically. Existing technologies lack a mechanism to solidify validated structural changes into a "new baseline" and continuously record imbalances and new developments, resulting in poor system adaptability. Every business change requires manual remodeling and configuration.

[0008] In summary, existing technologies have not yet achieved a technical solution that can autonomously optimize the entire process of an industrial logistics system, from "structural-level expression → imbalance identification → structural change → simulation verification → baseline evolution".

[0009] Application content

[0010] This invention proposes a closed-loop method and system for structural optimization of industrial logistics to address the aforementioned problems, thereby solving the following technical issues:

[0011] (1) How to establish a unified representation of the current structure of the industrial logistics system and transform multi-source operational data into a structural state diagram;

[0012] (2) How to identify structural imbalance units that lead to repetitive inefficiency based on structural state diagrams;

[0013] (3) How to generate a set of structural change candidates for unbalanced units and map the candidate schemes into a simulateable topology-resource-rule ternary change package;

[0014] (4) How to map structural change packages to scene variants in a simulation environment, perform behavior unfolding and result evaluation on scene variants, and detect structural side effects;

[0015] (5) How to write back the structural changes that have passed the simulation verification to form a new structural state baseline, so as to realize the structural-level optimization closed loop of the logistics system.

[0016] The technical solution of this invention is implemented as follows:

[0017] An industrial logistics structure-level optimization closed-loop system, comprising:

[0018] The structural modeling layer is used to acquire multi-source operational data from the enterprise management system, the field sensing system, and the topology definition source in real time, and to perform time alignment, spatial encoding alignment and object normalization processing on the data to generate a time-series logistics system structural state diagram. The structural state diagram includes at least regional nodes, resource nodes, task nodes and edges representing flow, service and constraint relationships.

[0019] An imbalance identification layer, connected to the structure modeling layer, is used to calculate structural indicators for the region subgraph, path subgraph, and resource subgraph based on the structure state diagram, and to identify structural imbalance units that lead to repetitive inefficiency.

[0020] The structural change layer, connected to the imbalance identification layer, is used to call a preset structural change operator according to the imbalance type, generate multiple candidate structural change schemes, and convert each candidate scheme into a structural change package. The structural change package is a structural change object described by topology dimension, resource dimension, and rule dimension, and contains at least one or more of topology change, resource change, and rule change.

[0021] The scenario verification layer, connected to the structure change layer, is used to combine the baseline scenario with the structure change package to form a simulable scenario variant, and to perform behavior expansion and result evaluation on the scenario variant in a discrete event simulation environment, outputting a result evaluation vector, while detecting the side effects of the structure change on other subgraphs or resource networks.

[0022] The structural evolution layer, connected to the scenario verification layer, is used to jointly screen effective structural changes based on the result evaluation vector and the structural side effect detection results. The implemented structural version is recorded as the new structural state baseline, and the elimination of unbalanced units, the occurrence of new unbalanced units, and the long-term effectiveness of the change operator are recorded before entering the next round of optimization loop.

[0023] Furthermore, the structural modeling layer establishes a data synchronization link with the core business database of the enterprise management system through a change data capture middleware or a high-frequency incremental query API; the structural change layer is connected to a structural change operator library, which includes at least spatial operators, path operators, resource operators, and rule operators; the scenario verification layer is connected to a discrete event simulation engine; and the structural evolution layer is connected to a structural baseline version library for storing historical structural baselines and their evolution records.

[0024] A closed-loop method for structural optimization of industrial logistics includes the following steps:

[0025] S1: Multi-source operational data alignment to construct a logistics structure state diagram: Real-time collection of multi-source operational data from enterprise management system, field sensing system and topology definition source, and time alignment, spatial coding alignment and object normalization processing of the data to generate a time-series logistics system structure state diagram;

[0026] S2: Identify structural imbalance units based on the structural state diagram: Calculate structural indicators for the region subgraph, path subgraph, and resource subgraph in the structural state diagram to identify structural imbalance units that lead to repetitive inefficiency;

[0027] S3: Generate a set of candidate structural changes for imbalanced units: Invoke a preset structural change operator according to the imbalance type to generate multiple candidate structural change schemes. Each candidate structural change scheme is represented by a structural change package that includes topology changes, resource changes, and rule changes.

[0028] S4: Map the candidate set to a simulable scene variant: Combine the structural change package with the baseline scene to generate a simulable scene variant, wherein the baseline scene consists of the current topology, current resources and current rules;

[0029] S5: Perform behavioral expansion and result evaluation on scene variants in a simulation environment: Perform behavioral expansion on the scene variants in a discrete event simulation environment, output the result evaluation vector, and detect the side effects of structural changes on other subgraphs or resource networks;

[0030] S6: Screen effective structural changes based on simulation results and form a new structural state baseline: Based on the result evaluation vector and the results of the structural side effect detection, screen effective changes, record the implemented structural version as the new structural state baseline, record the imbalance elimination, the occurrence of new imbalances and the long-term effectiveness of the operator, and return to step S1 to enter the next round of optimization loop.

[0031] Furthermore, the multi-source operational data mentioned in step S1 includes at least: historical task data of the enterprise resource planning system, inventory and storage location data, resource operation data, ultra-wideband trajectory data, arrival / production cycle data, and regional topology definition; the structural state diagram includes at least regional nodes, resource nodes, task nodes, and edges representing flow, service, and constraint relationships.

[0032] Furthermore, the structural indicators mentioned in step S2 include at least one of the following: average waiting time, peak stacking density, task return rate, empty run rate, path duplication rate, time window conflict rate, resource load skewness, and degree of inventory location deviation from the demand center; the structural imbalance unit includes at least one of the following elements: imbalance location, imbalance type, imbalance intensity, propagation range, stability, and dominant constraint; the imbalance type includes at least one of the following: capacity imbalance, temporal imbalance, spatial imbalance, path imbalance, and resource imbalance.

[0033] Furthermore, the structural change operator in step S3 includes at least one of the following categories: spatial operators, path operators, resource operators, and rule operators; the structural change package includes at least one or more of the following fields: target imbalance unit identifier, topology change description, resource change description, rule change description, applicable constraints, and expected effect vector.

[0034] Furthermore, the topology changes include at least one of the following: region merging or splitting, buffer migration, high-turnover material relocation, path channel reconstruction, and line-side supply mapping changes; the resource changes include at least one of the following: reconfiguration of transportation resource quantities, reclassification of resource service areas, adjustment of task division boundaries, and enhancement of resources during specific time periods; the rule changes include at least one of the following: adjustment of replenishment trigger thresholds, change of task priority calculation methods, change of batch release strategies, and change of time window release rules.

[0035] Furthermore, the scenario variant described in step S4 is expanded into at least one of the following parameters before simulation: region layout parameters, resource configuration parameters, path network parameters, task release parameters, and time window constraint parameters.

[0036] Furthermore, the result evaluation vector in step S5 includes at least one of the following indicators: throughput, average latency, maximum queue length, empty run rate, emergency task completion rate, resource utilization balance, lineside material shortage risk, and topology stability score; structural side effect detection includes determining whether candidate solutions will transfer congestion, resource fluctuations, or bottlenecks to other regions or subgraphs.

[0037] Furthermore, the new structural state baseline described in step S6 shall record at least one of the following information: baseline version number, list of eliminated structural imbalance units, list of newly emerging structural imbalance units, list of long-term effective change operators, and list of change operators applicable to a specific cycle interval.

[0038] By adopting the above technical solution, the beneficial effects of the present invention are as follows:

[0039] First, this invention enables structural-level representation and digital modeling of logistics systems. By constructing a "structural state diagram" that includes regional nodes, resource nodes, task nodes, and multi-relationship edges, it transforms multi-source operational data into a unified digital representation of the "structure" of the logistics system. This lays the technical foundation for subsequent identification of structural imbalances and structural changes, and solves the problem of the lack of structural representation capabilities in existing technologies.

[0040] Second, it accurately identifies structural imbalance units and their root causes. Based on the structural state diagram, this invention calculates multi-dimensional structural indicators for regional subgraphs, path subgraphs, and resource subgraphs. It can accurately locate "structural imbalance units" (such as capacity imbalance, path imbalance, and resource imbalance) that lead to repetitive inefficiency, and quantify their imbalance intensity, propagation range, and dominant constraints. This avoids the limitations of traditional methods that only focus on single abnormal events or local manifestations.

[0041] Third, the invention generates verifiable structural change solutions. It transforms candidate optimization solutions into structural change packages consisting of "topology changes, resource changes, and rule changes," elevating the optimization object from "operational parameters" to the "system structure" level. This enables the solutions to systematically address structural defects rather than merely performing localized repairs.

[0042] Fourth, a closed-loop mechanism for simulation verification and side effect detection is established. This invention combines structural change packages with baseline scenarios to generate "scenario variants." These variants are then subjected to behavioral decomposition and quantitative evaluation in a discrete event simulation environment. Specifically, it detects whether structural changes transfer congestion, resource fluctuations, or bottlenecks to other subgraphs. This mechanism allows for effect prediction and risk assessment of structural optimization schemes before implementation, significantly reducing the trial-and-error costs and operational risks of actual deployment.

[0043] Fifth, a closed-loop system for continuous evolution of the structural baseline is formed. This invention writes back the structural changes that have passed simulation verification to the actual system, forming a new "structural state baseline," and automatically records the elimination of unbalanced units, the addition of unbalanced units, and the long-term effectiveness of change operators in each round of evolution. As a result, the logistics system can achieve continuous autonomous optimization and adaptive evolution at the "structural level" in response to changes in production cycle, product structure, layout adjustments, and other factors, maintaining a high-efficiency operating state in the long term.

[0044] Sixth, this invention reduces reliance on human experience and improves system robustness. It achieves full-process automation and closed-loop decision-making from structural representation, imbalance identification, change generation, simulation verification to baseline updating, significantly reducing dependence on manual intervention by domain experts. Simultaneously, by detecting and avoiding structural side effects, it effectively prevents secondary problems during the optimization process, significantly improving the system's robustness in dynamic and complex environments. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a system framework diagram of the present invention;

[0047] Figure 2 This is a flowchart of the method of the present invention;

[0048] Figure 3 These are the node and edge types of the Structural State Diagram (SSG) of this invention;

[0049] Figure 4This is a labeled view of the structural imbalance unit (SIU) of the present invention;

[0050] Figure 5 This is a schematic diagram illustrating the formation of structural change packages and scene variants according to the present invention.

[0051] Figure 6 This is a simulation evaluation and side effect detection view of the present invention. Detailed Implementation

[0052] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0053] like Figure 1 As shown, an industrial logistics structure-level optimization closed-loop system is described. The system constructs a structural state diagram of the industrial logistics system, identifies structural imbalance units that lead to repetitive inefficiency, generates structural change packages for the imbalance units, maps the structural change packages to simulateable scenario variants, and filters effective structural changes through simulation results to form a new structural baseline, thereby realizing a structure-level optimization closed loop for the industrial logistics system.

[0054] The system includes the following functional layers:

[0055] The structural modeling layer is used to acquire multi-source operational data from the enterprise management system, the field sensing system, and the topology definition source in real time, and to perform time alignment, spatial encoding alignment, and object normalization processing on the data to generate a time-series logistics system structural state diagram; the structural state diagram includes at least regional nodes, resource nodes, task nodes, and edges representing flow, service, and constraint relationships;

[0056] An imbalance identification layer, connected to the structure modeling layer, is used to calculate structural indicators for the region subgraph, path subgraph, and resource subgraph based on the structure state diagram, and to identify structural imbalance units that lead to repetitive inefficiency; the structural imbalance unit includes at least the imbalance location, imbalance type, imbalance intensity, propagation range, stability, and dominant constraint.

[0057] The structural change layer, connected to the imbalance identification layer, is used to call a preset structural change operator according to the imbalance type, generate multiple candidate structural change schemes, and convert each candidate scheme into a structural change package. The structural change package is a structural change object described by topology dimension, resource dimension, and rule dimension, and contains at least one or more of topology change, resource change, and rule change.

[0058] The scenario verification layer, connected to the structure change layer, is used to combine the baseline scenario with a certain structure change package to form a simulable scenario variant. The scenario variant is then subjected to behavior expansion and result evaluation in a discrete event simulation environment, and the result evaluation vector is output. At the same time, the side effects of the structure change on other subgraphs or resource networks are detected.

[0059] The structural evolution layer, connected to the scenario verification layer, is used to jointly screen effective structural changes based on the result evaluation vector and the structural side effect detection results. This screening is based at least on the result evaluation vector and the structural side effect detection results. The implemented structural version is recorded as the new structural state baseline, and the elimination of unbalanced units, the occurrence of new unbalanced units, and the long-term effectiveness of the change operator are recorded before entering the next round of optimization loop.

[0060] Preferably, the structural modeling layer establishes a data synchronization link with the core business database of the enterprise management system through a change data capture middleware or a high-frequency incremental query API. The data acquired in real time includes at least SAP historical task data, inventory and storage location data, resource operation data, UWB trajectory data, arrival / JPH cycle time data, and regional topology definition.

[0061] Preferably, the structural indicators include average waiting time, peak stacking density, task return rate, empty run rate, path duplication rate, time window conflict rate, resource load skewness, and the degree to which inventory location deviates from the demand center; the imbalance types include capacity imbalance, time-series imbalance, spatial imbalance, path imbalance, and resource imbalance; and the structural change operators include spatial operators, path operators, resource operators, and rule operators.

[0062] Preferably, the structural change package includes at least one or more of the following fields: target imbalance unit identifier, topology change description, resource change description, rule change description, applicable constraints, and expected effect vector; the topology change includes region merging / splitting, buffer migration, high-turnover material relocation, path channel reconstruction, and line-side supply mapping change; the resource change includes reconfiguration of transportation resource quantity, reclassification of resource service range, adjustment of task division boundaries, and resource enhancement for specific time periods; the rule change includes adjustment of replenishment trigger threshold, change of task priority calculation method, change of batch release strategy, and change of time window release rule.

[0063] Preferably, the scenario variant is expanded before simulation into regional layout parameters, resource configuration parameters, path network parameters, task release parameters, and time window constraint parameters; the simulation environment is a discrete event simulation, which expands the behavior of the scenario variant under a unified time axis to simulate arrival, shelving, replenishment, outbound, resource occupation, path movement, waiting and conflict, and cycle time consumption; the result evaluation vector includes throughput, average latency, maximum queue length, empty run rate, emergency task completion rate, resource utilization balance, line-side material shortage risk, and topology stability score.

[0064] Preferably, the structural baseline update content recorded by the structural evolution layer includes: baseline version number, list of eliminated structural imbalance units, list of newly emerging structural imbalance units, list of long-term valid change operators, and list of change operators applicable to specific beat intervals.

[0065] like Figure 2 As shown, this invention also provides a closed-loop method for optimizing the structure of industrial logistics, comprising the following steps:

[0066] S1: Multi-source operational data alignment to construct a logistics structure state diagram: Through change data capture middleware or high-frequency incremental query API, real-time collection of logistics orders, inventory status, site information and transportation resource status from enterprise management system, field perception system and topology definition source, and time alignment, spatial coding alignment and object normalization processing of data to generate a time-series structure state diagram sequence;

[0067] S2: Identify structural imbalance units based on structural state diagrams: Calculate structural indicators for the region subgraphs, path subgraphs, and resource subgraphs in the structural state diagrams to identify structural imbalance units that lead to repetitive inefficiency;

[0068] S3: Generate a set of candidate structural changes for unbalanced units: Based on the type of imbalance, call the preset structural change operator to generate multiple candidate structural change schemes. Each candidate structural change scheme is represented by a structural change package consisting of topology change, resource change, and rule change.

[0069] S4: Map the candidate set to a simulateable topology-resource-rule ternary change package and combine it with the baseline scenario to form a scenario variant: Combine the structural change package with the baseline scenario (original topology + original resources + original rules) to generate a simulateable scenario variant, and expand it into regional layout parameters, resource configuration parameters, path network parameters, task release parameters, and time window constraint parameters.

[0070] S5: Behavior unfolding and outcome evaluation of scene variants in a simulation environment: Behavior unfolding of scene variants in a discrete event simulation environment, outputting outcome evaluation vectors, and detecting the side effects of structural changes on other subgraphs or resource networks;

[0071] S6: Filter effective structural changes based on simulation results and write back to form a new structural state baseline: Based on the result evaluation vector and the results of the structural side effect detection, filter effective changes, record the implemented structural version as the new structural state baseline, record the imbalance elimination, the occurrence of new imbalances and the long-term effectiveness of the operator, and return to step S1 to enter the next round of optimization loop.

[0072] Preferably, the identification of the structural imbalance unit in step S2 is based on at least one of the following structural indicators: average waiting time, peak stacking density, task return rate, empty run rate, path duplication rate, time window conflict rate, resource load skewness, and degree of inventory location deviation from demand center.

[0073] Preferably, the structural change operators in step S3 include spatial operators, path operators, resource operators, and rule operators; the topology changes include region merging / splitting, buffer migration, high-turnover material relocation, path channel reconstruction, and line-side supply mapping changes; the resource changes include reconfiguration of transportation resource quantities, reclassification of resource service areas, adjustment of task division boundaries, and resource enhancement for specific time periods; and the rule changes include adjustment of replenishment trigger thresholds, changes in task priority calculation methods, changes in batch release strategies, and changes in time window release rules.

[0074] Preferably, the result evaluation vector in step S5 includes throughput, average latency, maximum queue length, empty run rate, emergency task completion rate, resource utilization balance, lineside material shortage risk, and topology stability score; structural side effect detection includes determining whether the candidate solution will transfer congestion, resource fluctuations, or bottlenecks to other regions or subgraphs.

[0075] Preferably, the new structural state baseline described in step S6 records at least the following information: baseline version number, list of eliminated structural imbalance units, list of newly emerging structural imbalance units, list of long-term effective change operators, and list of change operators applicable to specific time intervals.

[0076] Example 1: Migration scenario of the buffer zone at the edge of the automobile assembly line

[0077] like Figures 3-6 As shown, this embodiment takes the frequent congestion of the line-side buffer zone in the final assembly workshop of an automobile manufacturing enterprise due to the introduction of a new model as an example to fully illustrate the closed-loop operation process of "structural-level optimization" in this invention.

[0078] Scenario: At workstation WS01 on a certain final assembly line, the increased size of parts for a new model has led to insufficient capacity in the original buffer zone B01, resulting in excessively long unloading times for AGVs and causing congestion on path P01. The system needs to identify this structural imbalance and propose an optimization solution that includes topology changes.

[0079] Step 1: Business Event Triggering and Structural State Diagram Update. As the production of the new model ramps up, the structural modeling layer continuously acquires multi-source data from the enterprise management system and the on-site sensing system. Specifically, the system uses a change data capture middleware to collect material requirement orders and inventory change records from the SAP system in real time; it receives real-time AGV location and status data from the UWB positioning system via the MQTT protocol; and it obtains workstation production cycle time (JPH) and material consumption data from the MES database. All raw data first enters a stream processing engine (such as Apache Flink) for the following preprocessing:

[0080] ① Time alignment: Convert different source data (SAP transaction time, UWB timestamp, MES timestamp) into a system clock in milliseconds and group them into 100ms time windows to ensure that all types of data participating in the calculation within the same time slice have a synchronized time base.

[0081] ② Spatial encoding alignment: Establish a spatial mapping table to associate the logical library location encoding (such as "WS01-B01") in MES with the physical coordinates (x, y) in the UWB coordinate system, so as to unify the expression of all spatial data.

[0082] ③ Object normalization: Using an object mapping library, different identifiers for the same entity in different systems (such as material number "M1001" in SAP and material code "MAT-001" in MES) are mapped to globally unique internal object IDs.

[0083] The preprocessed data is fed into a graph database (such as Neo4j) to construct a temporal structured state graph SSG(t). SSG(t) is stored as an attribute graph, and its node types include:

[0084] (1) Area node: such as warehouse WH01, line edge buffer B01, workstation WS01, attributes include current inventory, capacity limit, coordinates, etc.

[0085] (2) Resource nodes: such as AGV-05 and AGV-08, whose attributes include location, status (idle / busy / waiting), current task, etc.

[0086] (3) Task node: such as replenishment task T123 and outbound task T456. Attributes include task type, target location, issuance time, time limit, etc.

[0087] The edges between nodes represent the dynamic relationships in the logistics system:

[0088] Flow: Represents the movement path of materials or resources, such as AGV-05-[Execute]->T123. Attributes include path distance and historical average travel time.

[0089] Service edge: Represents the coverage relationship of resources to regions, such as AGV-05-[Serves]->WS01. Attributes include service priority and response time.

[0090] Constraint edge: Represents capacity limit or timing constraint between regions. For example, B01-[capacity constraint]->WS01 means that B01 provides a cache for WS01, and its maximum inventory cannot exceed the physical capacity.

[0091] At the current time t0, SSG(t0) shows that the stacking density (current inventory / capacity limit) of the line-edge buffer B01 of regional node WS01 continues to exceed 0.9, and the service edge between it and resource nodes AGV-05 and AGV-08 is in a frequent waiting state (the average waiting time for AGV-05 to unload at B01 increases from 5 minutes to 15 minutes). On the flow edge connecting B01 and the central warehouse regional node WH01, the task return rate (the proportion of AGVs returning empty due to unloading port blockage) increases to 25%.

[0092] Step 2: Identify structurally imbalanced units. The imbalance identification layer analyzes SSG(t) and triggers a round of indicator calculations periodically (e.g., every 5 minutes) for regional subgraphs, path subgraphs, and resource subgraphs. The specific calculation method is as follows:

[0093] Peak packing density ρ max = max{t∈[t0-1h, t0]} (current inventory / capacity limit), ρ is calculated for B01. max = 0.95.

[0094] Task return rate r return = (Number of AGV tasks returning empty from B01 per unit time) / (Total number of delivered tasks), calculated for path P01 to obtain r return = 25%.

[0095] Average waiting time w avg = Total waiting time for all AGVs unloading at B01 / Number of unloading operations, calculated for the service edge (AGV-05, B01). avg = 15 minutes.

[0096] The system compares each indicator with a preset dynamic threshold (generated based on historical data statistics, such as ρ). max Threshold 0.85, r returnThe threshold of 20% is compared. The ρ value in the WS01 region... max and r return If all conditions exceed the threshold and persist for more than 30 minutes (stability condition), the system identifies this state as a composite structural imbalance unit SIU-01, characterized by both "capacity imbalance" and "path imbalance". The imbalance location of SIU-01 is "buffer zone B01 and connected path P01", the imbalance type is "capacity-path composite imbalance", the imbalance intensity is (0.95-0.85)+(0.25-0.20)=0.15 (after normalization), the propagation range affects the outbound efficiency of upstream warehouse WH01 (the AGV waiting queue of WH01 increases), and the dominant constraint is "upper limit of buffer zone physical space".

[0097] Step 3: Generate a candidate set of structural changes. The structural change layer calls the space-based operator "Buffer Migration" for "Capacity Imbalance" in SIU-01, and the path-based operator "Path Channel Reconstruction" for "Path Imbalance". The operator library stores the definition of each operator in JSON format.

[0098] The system automatically fills in specific parameters for the operators based on the attributes of the current SIU (target region B01, associated path P01). By combining different operator parameters, two candidate structural change packages are generated.

[0099] Each SCP also contains metadata: the target imbalance unit identifier "SIU-01", the applicable constraints "only applicable during daytime peak hours (8:00-11:00, 13:00-16:00)", and the expected effect vector (e.g., "expected waiting time reduction of 30%").

[0100] Step 4: Mapping to Simulated Scene Variants. The scene verification layer combines the baseline scene (current layout) with SCP-A and SCP-B respectively to generate scene variants V1 and V2. The baseline scene is defined by the topology, resources, and rules in the current structural state diagram SSG(t0). The combination process is completed through the simulation configuration generator, which converts the change descriptions in the SCP into a parameter table that the simulation engine can recognize.

[0101] Region layout parameters: such as the coordinates of B01' (120, 85), and the buffer capacity of 200.

[0102] Resource configuration parameters: such as adding 2 AGVs to the pool and dividing the service area.

[0103] Path network parameters: such as the node sequence of path P01' [(115,80), (120,85)], and the travel time is automatically calculated based on the distance and AGV speed.

[0104] Task release parameters: Generate a material demand sequence for the next hour based on historical data, including task type, target workstation, demand quantity, and release time.

[0105] Time window constraint parameters: such as keeping the AGV arrival time window width unchanged at 30 minutes.

[0106] The generated scene variants V1 and V2 are saved as simulation project files (such as XML format), containing the definitions of all entities and rules.

[0107] Step 5: Simulation Deployment and Result Evaluation. Run V1 and V2 in a discrete event simulation environment (such as AnyLogic or Simio). The simulation duration is 2 hours of virtual time, and each simulation is repeated 5 times to eliminate the influence of randomness. The simulation engine advances the time according to the event scheduling mechanism, recording the occurrence time and state changes of each event (task issuance, AGV start-up, arrival, unloading, waiting, conflict, etc.). After the simulation is completed, the result evaluation vector is output, which is calculated as follows:

[0108] Throughput: The total number of tasks completed during the simulation.

[0109] Average delay: The average time interval between all AGVs from receiving a task to completing unloading.

[0110] Maximum queue length: The maximum number of AGVs waiting to be unloaded in front of B01'.

[0111] Empty driving rate: The proportion of AGV's empty driving time to its total driving time.

[0112] Urgent task completion rate: The percentage of tasks with a priority of "urgent" that are completed within the specified time limit.

[0113] Resource utilization balance: the standard deviation of the busy time of each AGV; the smaller the value, the more balanced the resource utilization.

[0114] Material shortage risk at the line: The total duration during which material shortage occurs at workstation WS01 (with a buffer of 0).

[0115] Topology stability score: Calculated based on the proportion of path congestion time to total simulation time; the less congestion time, the higher the score.

[0116] After the simulation, the side effect detection module analyzes the changes in key indicators of other subgraphs. For example, for V1, the system checks whether the average waiting time of the adjacent workstation WS02 exceeds a threshold (e.g., 10 minutes). If it does not exceed the threshold, and the stacking density of WS02 does not exceed 0.7, then it is determined that there is no side effect. For V2, the system finds that a new time window conflict occurs at the path fork point (the entrances of B01-a and B01-b) (the number of AGVs arriving at the same time exceeds the entrance capacity), causing the material shortage risk of WS02 to increase from 0.1 to 0.25, exceeding the side effect threshold of 0.2. Therefore, it is determined that V2 has a side effect.

[0117] The simulation results vector for V1 are: throughput +12%, average latency -40%, maximum queue length -50%, empty run rate -5%, material shortage risk reduced to 0, topology stability score of 92, and no side effects.

[0118] Step 6: Establish a new structural state baseline. The structural evolution layer selects SCP-A as a valid change based on the result evaluation vector (high benefit) and side effect detection results (no side effects). The system issues rule changes (adjusting the replenishment threshold to 50 pieces) through the rule engine, issues AGV new path instructions through the MES interface, and updates the position and capacity of buffer B01' in the topology database. The updated system state is persisted to the structural baseline version library, denoted as Baseline v2.0. The baseline record includes: baseline version number v2.0, timestamp, list of eliminated SIUs [SIU-01], list of newly added SIUs [], list of long-term valid change operators ["Buffer Migration"], and list of operators applicable to specific beat intervals ["Path Reconstruction" applicable to peak periods]. Then, return to Step 1 and wait for the next round of optimization.

[0119] The specific implementation details of integration with enterprise management systems in this embodiment, such as SAP HANA, Debezium, and Kafka, can be implemented using conventional middleware technologies, and will not be elaborated here.

[0120] Example 2: Path Restructuring and Resource Allocation Linkage Scenario in High-Beat Production Areas

[0121] This embodiment uses the example of frequent congestion in the AGV path network of a certain workshop during high-frequency periods to illustrate the ability of the present invention to handle more complex structural changes.

[0122] Scenario: The workshop has three high-cycle production areas, Z1, Z2, and Z3, sharing a main road, MAIN. During peak hours, Z1 and Z2 simultaneously require material replenishment, causing a deadlock on the main road, MAIN, and a sharp increase in the AGV's empty-run rate.

[0123] Step 1: Construction of the structural state diagram. The system collects the AGV positions, task status, and path occupancy in each area in real time. At time t0, the structural state diagram SSG(t0) shows that on the main road MAIN path sub-graph, the number of AGVs running simultaneously reaches 15 (exceeding its capacity of 12), causing the average travel time to increase from 2 minutes to 8 minutes; the line-side buffers in areas Z1 and Z2 are about to be exhausted, but the AGVs are blocked on MAIN and cannot be delivered in time, forming a composite imbalance unit SIU-02 of "path imbalance" and "timing imbalance".

[0124] Step 2: Imbalance Identification. The imbalance identification layer calculates the "path congestion" (current AGV quantity / capacity) of the path subgraph MAIN to be 1.25, exceeding the threshold of 1.0, and the duration exceeds 10 minutes; the "material shortage risk index" (expected sustaining time / replenishment cycle) of regions Z1 and Z2 are both below 0.5. The system marks SIU-02 as a high-priority imbalance unit requiring immediate intervention.

[0125] Step 3: Generate a candidate set of structural changes. To resolve the main road sharing conflict, the structural change layer invokes the path-type operator "Path Channel Reconstruction" and the resource-type operator "Resource Service Scope Reassignment" to generate a single candidate solution SCP-C containing two linked changes:

[0126] Topology Change: The shared main road MAIN is reconstructed into two unidirectional loops. Specifically, the system generates two new regional nodes, Loop-A and Loop-B, using topology operators. Loop-A connects Z1 and Z2, and Loop-B connects Z2 and Z3, with an exchange zone set up at Z2. The coordinates of the new paths are automatically generated by the path planning algorithm based on the workshop layout, ensuring no conflict with existing facilities.

[0127] Resource Changes: The AGV fleet has been reorganized. The AGV pool is divided into three groups using resource operators: the Red Team (5 AGVs) serves Loop-A, the Blue Team (5 AGVs) serves Loop-B, and the Yellow Team (2 AGVs) acts as a mobile team, capable of temporarily entering any loop to perform cross-regional tasks or support additional needs in Z2. Simultaneously, the task allocation rules have been adjusted: tasks from Z1 are assigned only to the Red Team, tasks from Z3 are assigned only to the Blue Team, and tasks from Z2 are prioritized for the Yellow Team. If the Yellow Team is busy, tasks are then distributed to the Red and Blue Teams based on load balancing.

[0128] Rule Change: Set independent time window rules for each loop to avoid mutual interference between loops.

[0129] SCP-C is packaged as a structural change package containing detailed descriptions of the aforementioned topology, resources, and rules.

[0130] Step 4: Scene Variant Generation and Simulation. The scene verification layer combines the baseline scene with SCP-C to generate scene variant V3. In the simulation parameters, the path network is updated to two loops, and the AGV grouping and task allocation rules are adjusted accordingly. The simulation runs the peak task sequence for the next 2 hours.

[0131] Step 5: Result Evaluation and Side Effect Detection. Simulation results show that V3 completely eliminated congestion on the main road MAIN, reduced the average AGV travel time from 8 minutes to 2.5 minutes, decreased the overall empty-run rate by 25%, and improved the emergency task completion rate by 30%. The side effect detection module checked key indicators in each area: the material shortage risk in Z1 and Z2 decreased to below 0.1; the buffer supply in Z3 was normal; and no new congestion points or resource competition were found. The system also evaluated the utilization rate of the yellow mobile team, finding that it remained within a healthy range of 60%~80% and did not become a new bottleneck.

[0132] Step 6: Baseline Update. The structural evolution layer confirms that SCP-C is a valid change, writes the new path network, AGV grouping, and task allocation rules into the enterprise management system, forming a new structural state baseline, Baseline v3.0. In the baseline record, the eliminated SIU is SIU-02, no new SIUs are added, and the operator combination "path channel reconstruction + resource service scope rezoning" is recorded as effective in this type of high-frequency, multi-region scenario.

[0133] Components not described in detail in this article are existing technologies.

[0134] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An industrial logistics structure-level optimization closed-loop system, characterized in that: include: The structural modeling layer is used to acquire multi-source operational data from the enterprise management system, the field sensing system, and the topology definition source in real time, and to perform time alignment, spatial encoding alignment and object normalization processing on the data to generate a time-series logistics system structural state diagram. The structural state diagram includes at least regional nodes, resource nodes, task nodes and edges representing flow, service and constraint relationships. An imbalance identification layer, connected to the structure modeling layer, is used to calculate structural indicators for the region subgraph, path subgraph, and resource subgraph based on the structure state diagram, and to identify structural imbalance units that lead to repetitive inefficiency. The structural change layer, connected to the imbalance identification layer, is used to call a preset structural change operator according to the imbalance type, generate multiple candidate structural change schemes, and convert each candidate scheme into a structural change package. The structural change package is a structural change object described by topology dimension, resource dimension, and rule dimension, and contains at least one or more of topology change, resource change, and rule change. The scenario verification layer, connected to the structure change layer, is used to combine the baseline scenario with the structure change package to form a simulable scenario variant, and to perform behavior expansion and result evaluation on the scenario variant in a discrete event simulation environment, outputting a result evaluation vector, while detecting the side effects of the structure change on other subgraphs or resource networks. The structural evolution layer, connected to the scenario verification layer, is used to jointly screen effective structural changes based on the result evaluation vector and the structural side effect detection results. The implemented structural version is recorded as the new structural state baseline, and the elimination of unbalanced units, the occurrence of new unbalanced units, and the long-term effectiveness of the change operator are recorded before entering the next round of optimization loop.

2. The industrial logistics structure-level optimization closed-loop system according to claim 1, characterized in that: The structural modeling layer establishes a data synchronization link with the core business database of the enterprise management system through a change data capture middleware or a high-frequency incremental query API; the structural change layer is connected to a structural change operator library, which includes at least spatial operators, path operators, resource operators, and rule operators. The scenario verification layer is connected to a discrete event simulation engine; the structure evolution layer is connected to a structure baseline version library, which is used to store historical structure baselines and their evolution records.

3. A closed-loop method for structural optimization of industrial logistics, characterized in that: Includes the following steps: S1: Multi-source operational data alignment to construct a logistics structure state diagram: Real-time collection of multi-source operational data from enterprise management system, field sensing system and topology definition source, and time alignment, spatial coding alignment and object normalization processing of the data to generate a time-series logistics system structure state diagram; S2: Identify structural imbalance units based on the structural state diagram: Calculate structural indicators for the region subgraph, path subgraph, and resource subgraph in the structural state diagram to identify structural imbalance units that lead to repetitive inefficiency; S3: Generate a set of candidate structural changes for imbalanced units: Invoke a preset structural change operator according to the imbalance type to generate multiple candidate structural change schemes. Each candidate structural change scheme is represented by a structural change package that includes topology changes, resource changes, and rule changes. S4: Map the candidate set to a simulable scene variant: Combine the structural change package with the baseline scene to generate a simulable scene variant, wherein the baseline scene consists of the current topology, current resources and current rules; S5: Perform behavioral expansion and result evaluation on scene variants in a simulation environment: Perform behavioral expansion on the scene variants in a discrete event simulation environment, output the result evaluation vector, and detect the side effects of structural changes on other subgraphs or resource networks; S6: Screen effective structural changes based on simulation results and form a new structural state baseline: Based on the result evaluation vector and the results of the structural side effect detection, screen effective changes, record the implemented structural version as the new structural state baseline, record the imbalance elimination, the occurrence of new imbalances and the long-term effectiveness of the operator, and return to step S1 to enter the next round of optimization loop.

4. The closed-loop method for optimizing the industrial logistics structure according to claim 3, characterized in that: The multi-source operational data mentioned in step S1 includes at least: historical task data of the enterprise resource planning system, inventory and storage location data, resource operation data, ultra-wideband trajectory data, arrival / production cycle data, and regional topology definition; the structural state diagram includes at least regional nodes, resource nodes, task nodes, and edges representing flow, service, and constraint relationships.

5. The closed-loop method for optimizing the industrial logistics structure according to claim 3, characterized in that: The structural indicators mentioned in step S2 include at least one of the following: average waiting time, peak stacking density, task return rate, empty run rate, path duplication rate, time window conflict rate, resource load skewness, and degree of inventory location deviation from the demand center; the structural imbalance unit includes at least one of the following elements: imbalance location, imbalance type, imbalance intensity, propagation range, stability, and dominant constraint; the imbalance type includes at least one of the following: capacity imbalance, temporal imbalance, spatial imbalance, path imbalance, and resource imbalance.

6. The closed-loop method for optimizing the industrial logistics structure according to claim 3, characterized in that: The structural change operator mentioned in step S3 includes at least one of the following categories: spatial operators, path operators, resource operators, and rule operators; the structural change package includes at least one or more of the following fields: target imbalance unit identifier, topology change description, resource change description, rule change description, applicable constraints, and expected effect vector.

7. The industrial logistics structure-level optimization closed-loop method according to claim 6, characterized in that: The topology changes include at least one of the following: region merging or splitting, buffer migration, high-turnover material relocation, path channel reconstruction, and line-side supply mapping changes; the resource changes include at least one of the following: reconfiguration of transportation resource quantities, redrawing of resource service ranges, adjustment of task division boundaries, and enhancement of resources during specific time periods; the rule changes include at least one of the following: adjustment of replenishment trigger thresholds, change of task priority calculation methods, change of batch release strategies, and change of time window release rules.

8. The closed-loop method for optimizing the industrial logistics structure according to claim 3, characterized in that: The scenario variant described in step S4 is expanded to at least one of the following parameters before simulation: region layout parameters, resource configuration parameters, path network parameters, task release parameters, and time window constraint parameters.

9. The closed-loop method for optimizing the industrial logistics structure according to claim 3, characterized in that: The result evaluation vector in step S5 includes at least one of the following indicators: throughput, average latency, maximum queue length, empty run rate, emergency task completion rate, resource utilization balance, lineside material shortage risk, and topology stability score; structural side effect detection includes determining whether candidate solutions will transfer congestion, resource fluctuations, or bottlenecks to other regions or subgraphs.

10. The industrial logistics structure-level optimization closed-loop method according to claim 3, characterized in that: The new structural state baseline described in step S6 shall record at least one of the following information: baseline version number, list of eliminated structural imbalance units, list of newly emerging structural imbalance units, list of long-term effective change operators, and list of change operators applicable to a specific cycle interval.

Images