EWTMS full-link intelligent warehouse distribution management system

Abstract

The invention relates to the technical field of intelligent logistics management, and discloses an EWTMS full-link intelligent warehouse distribution management system. According to the system, digital mirror image mapping composed of a storage physical space mirror image and a distribution process network mirror image is established, and historical data is used for driving dynamic twinborn evolution of the system. And based on an evolution result, coupling the identified physical efficiency bottleneck region and the flow delay sensitive node, and generating a full-link dynamic vulnerability topological graph. The system further drives the topological graph to perform instant reconstruction according to real-time cargo movement data and equipment state signals, and dynamically generates a warehouse allocation instruction by analyzing reconstruction differences. According to the scheme, deep collaboration and overall vulnerability visualization of storage and distribution links are realized, adaptive strategy adjustment based on real-time data is supported, and the overall efficiency and response agility of a full-link warehouse distribution system are improved.

Description

EWTMS End-to-End Intelligent Warehousing and Distribution Management System Technical Field

[0001] This invention relates to the field of intelligent logistics management technology, specifically to the EWTMS end-to-end intelligent warehousing and distribution management system. Background Technology

[0002] Current smart logistics warehouse and distribution management systems typically manage warehousing and distribution as independent units. Warehousing management technology focuses on the static optimization of physical space and equipment scheduling within the warehouse, with performance analysis relying on periodically collected space utilization rates or equipment operating indicators. Distribution management technology, on the other hand, focuses on the dynamic planning of transportation routes and vehicle status monitoring, with optimization primarily based on route mileage or planned delivery times. Existing technological solutions have established independent data processing and decision-making mechanisms for these two stages.

[0003] This fragmented management model leads to data isolation and decision-making lags within the system. There is a lack of effective correlation analysis mechanisms between physical bottlenecks in warehousing and process delays in delivery, making it impossible to construct a unified model reflecting the causal relationship between the two. Therefore, the system struggles to accurately identify end-to-end vulnerabilities caused by cross-stage linkages. Strategy adjustments based on preset rules or fixed cycles cannot effectively integrate and respond to real-time cargo movement data and equipment status signals, resulting in decision updates lagging behind dynamic changes in the system's state.

[0004] A technological solution is needed that can deeply integrate physical warehouse data with delivery process data. The key challenge this solution must address is how to correlate physical space performance bottlenecks with process network latency nodes in real time, and how to dynamically drive the generation and adjustment of management strategies based on continuously changing real-time data. Summary of the Invention

[0005] The purpose of this invention is to provide an EWTMS end-to-end intelligent warehousing and distribution management system to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, this invention provides an EWTMS end-to-end intelligent warehousing and distribution management system. The system includes: a digital mirror construction module, which establishes a digital mirror mapping of warehousing and distribution operations, the digital mirror mapping consisting of a physical space mirror of the warehousing process and a process network mirror of the distribution process; a dynamic twin evolution module, which performs dynamic twin evolution on the digital mirror mapping, the twin evolution process being driven by the original data stream of historical warehousing and distribution operations; a bottleneck identification and coupling module, which, based on the twin evolution results, identifies performance bottleneck areas in the physical space mirror and identifies latency-sensitive nodes in the process network mirror, logically coupling the performance bottleneck areas and the latency-sensitive nodes to generate a dynamic vulnerability topology map of the entire chain; and a strategy dynamic allocation module, which, based on real-time cargo movement data and equipment status signals, drives the dynamic vulnerability topology map to be reconstructed in real time, and generates dynamic allocation instructions for warehousing and distribution strategies based on the reconstruction differences.

[0007] Preferably, a digital mirror mapping for warehousing and distribution operations is established. This digital mirror mapping consists of a physical space mirror of the warehousing process and a process network mirror of the distribution process. Specifically, it includes: receiving the coordinates of the storage location, storage status, handling equipment identification, and operating speed from the warehousing management unit; constructing a physical space mirror representing the physical layout and resource distribution within the warehouse based on the storage location coordinates, storage status, handling equipment identification, and operating speed; receiving the node locations, path connections, transportation vehicle attributes, and turnover time consumption from the distribution management unit; constructing a process network mirror representing the movement path and sequence dependence of goods based on the node locations, path connections, transportation vehicle attributes, and turnover time consumption; and aligning and binding the physical space mirror and the process network mirror in the time dimension to form a unified digital mirror mapping for warehousing and distribution operations.

[0008] Preferably, the digital mirror mapping undergoes dynamic twin evolution, driven by the raw data stream of historical warehousing and distribution operations. Specifically, this includes: continuously inputting actual order information, inventory change records, equipment operation logs, and vehicle trajectory point sequences occurring within a historical time interval. The actual order information, inventory change records, equipment operation logs, and vehicle trajectory point sequences constitute the raw data stream of historical warehousing and distribution operations. Using the raw data stream as an input sequence, the physical space mirror in the digital mirror mapping is driven to simulate the historical changes in cargo location status and the historical fluctuations in equipment load. Using the raw data stream as an input sequence, the process network mirror in the digital mirror mapping is driven to simulate the historical distribution of path traffic and the historical saturation of node processing capacity.

[0009] Preferably, based on the twin evolution results, performance bottleneck areas are identified in the physical space mirror, and delay-sensitive nodes are identified in the process network mirror. Specifically, this includes: calculating the ratio of the stacking rate to the emptying rate of task requests per unit time for each physical unit during the historical changes in the simulated cargo location status and the historical fluctuations in equipment load in the physical space mirror, and marking the set of physical units whose stacking rate continuously exceeds the emptying rate as performance bottleneck areas; and calculating the downstream waiting time increment caused by each network node when processing a unit of cargo during the historical distribution of simulated path traffic and the historical saturation of node processing capacity in the process network mirror, and marking network nodes whose waiting time increment exceeds a preset threshold as delay-sensitive nodes.

[0010] Preferably, the performance bottleneck region and the latency-sensitive node are logically coupled to generate a dynamic vulnerability topology graph of the entire link. Specifically, this includes: establishing a causal relationship mapping between the performance bottleneck region and the latency-sensitive node, whereby the causal relationship mapping describes how congestion in a specific performance bottleneck region leads to waiting at a specific latency-sensitive node; encoding the performance bottleneck region marked in the physical space mirror, the latency-sensitive node marked in the process network mirror, and the causal relationship mapping together into a weighted directed graph structure, which is the dynamic vulnerability topology graph; the node weights in the dynamic vulnerability topology graph are determined by the accumulation rate or waiting time increment of the corresponding region, and the edge weights are determined by the strength of the causal relationship; the process of encoding the performance bottleneck region marked in the physical space mirror, the latency-sensitive node marked in the process network mirror, and the causal relationship mapping together into a weighted directed graph structure is described. The code is a weighted directed graph structure, specifically including: abstracting each performance bottleneck region as a graph node and assigning a weight value to the graph node, the weight value being obtained by normalizing the ratio of the accumulation rate to the clearing rate of the performance bottleneck region; abstracting each latency-sensitive node as another graph node and assigning a weight value to the graph node, the weight value being obtained by normalizing the average waiting time increment of the latency-sensitive node; determining the existence and direction of edges pointing from performance bottleneck region nodes to latency-sensitive nodes according to the causal association mapping; assigning a weight value to each existing edge, the weight value being jointly determined by the number and severity of the increase in waiting time of associated latency-sensitive nodes caused by the deterioration of the corresponding performance bottleneck region state in historical data; and combining all weighted nodes and weighted edges to construct a complete weighted directed graph structure, namely the dynamic vulnerability topology graph.

[0011] Preferably, based on real-time cargo movement data and equipment status signals, the dynamic vulnerability topology is driven to be reconstructed in real time, and dynamic allocation instructions for warehousing and distribution strategies are generated based on the reconstruction differences. Specifically, this includes: real-time collection of current order sorting information, the location of goods in transit, the real-time load rate of handling equipment, and the real-time speed of transportation vehicles. The current order sorting information, the location of goods in transit, the real-time load rate of handling equipment, and the real-time speed of transportation vehicles constitute real-time cargo movement data and equipment status signals; using the real-time cargo movement data and equipment status signals as stimulus inputs, updating the weight values ​​of corresponding nodes and the connection strength of edges in the dynamic vulnerability topology; comparing the dynamic vulnerability topology before and after the update, identifying newly emerging nodes or edges whose weights exceed the activation threshold, and identifying the physical locations or process links associated with the newly emerging nodes or edges as targets requiring real-time intervention; and calculating resource reallocation schemes or path adjustment schemes based on the targets requiring real-time intervention to form dynamic allocation instructions for warehousing and distribution strategies.

[0012] Preferably, the original data stream is used as the input sequence to drive the physical space mirror in the digital mirror mapping, simulating the historical changes in the status of storage locations and the historical fluctuations in equipment load. Specifically, this includes: extracting the inbound records, outbound records, and storage location adjustment records for each moment within the historical time interval from the original data stream; updating the occupancy status, goods type, and storage duration of each storage location in the physical space mirror step by step based on the inbound records, outbound records, and storage location adjustment records; extracting the equipment task allocation records and task completion records for each moment within the historical time interval from the original data stream; and calculating the queue length of pending tasks, task execution rate, and idle rate of each handling device in the physical space mirror step by step based on the equipment task allocation records and task completion records.

[0013] Preferably, the original data stream is used as the input sequence to drive the process network mirror in the digital mirror mapping, simulating the historical distribution process of path traffic and the historical saturation process of node processing capacity. Specifically, this includes: extracting the quantity of goods, the quantity of vehicles, and the passage timestamp for each transportation path within the historical time interval from the original data stream; calculating the total quantity of goods and the total number of vehicles flowing through each path within each time window in the process network mirror based on the quantity of goods, the quantity of vehicles, and the passage timestamp; extracting the goods arrival record, processing start timestamp, and processing end timestamp for each network node within the historical time interval from the original data stream; and calculating the average processing time and maximum concurrent processing capacity of each network node within each time window in the process network mirror based on the goods arrival record, the processing start timestamp, and the processing end timestamp.

[0014] Preferably, the ratio of the accumulation rate to the clearing rate of task requests for each physical unit within a unit time is calculated, and the set of physical units whose accumulation rate consistently exceeds the clearing rate is marked as a performance bottleneck region. Specifically, this includes: for each physical unit, counting the number of task requests received within multiple consecutive unit time windows, and calculating the growth rate of the number of task requests within each time window as the accumulation rate; counting the number of tasks completed by each physical unit within the same multiple consecutive unit time windows, and calculating the growth rate of the number of completed tasks within each time window as the clearing rate; for each physical unit, calculating the ratio of its accumulation rate to the clearing rate within each time window; when the ratio of the accumulation rate to the clearing rate of a physical unit is greater than one within a consecutive preset number of time windows, the physical unit is marked as a potential bottleneck unit; and clustering spatially adjacent physical units that are simultaneously marked as potential bottleneck units to form a continuous region, which is the performance bottleneck region.

[0015] Preferably, the calculation of the downstream waiting time increment caused by each network node when processing a unit of goods, and the marking of network nodes whose waiting time increment exceeds a preset threshold as delay-sensitive nodes, specifically includes: for each network node, obtaining the timestamp record of each batch of goods that has been processed and delivered to the downstream node; calculating the difference between the actual timestamp and the expected arrival timestamp of each batch of goods received by the downstream node as the single waiting time caused by each batch of goods; for each network node, statistically analyzing all batches of goods processed within a historical period, calculating the average single waiting time caused by all batches as the average waiting time increment of the network node; comparing the average waiting time increment of each network node with a system-preset delay threshold; and marking network nodes whose average waiting time increment is greater than or equal to the delay threshold as delay-sensitive nodes.

[0016] Compared with existing technologies, the beneficial effects of this invention are: logically coupling the performance bottleneck areas of the physical space mirror with the delay-sensitive nodes of the process network mirror, generating a dynamic vulnerability topology map of the entire link. This technology breaks down the information barriers between warehousing and distribution links, and algorithmically correlates bottlenecks such as equipment congestion and low operational efficiency in the physical world with delay nodes such as order backlog and path timeouts in the virtual process. This coupling analysis reveals chain reactions and root causes that cannot be discovered from a traditional isolated perspective. From a systems theory perspective, it presents a complete and interconnected dynamic health map of the warehousing and distribution network, enabling management decisions to shift from patching local symptoms to root-cause governance and collaborative optimization based on the global vulnerability structure.

[0017] Based on real-time cargo movement data and equipment status signals, the system drives the instant reconstruction of a dynamic vulnerability topology graph and generates dynamic allocation instructions for warehousing and distribution strategies based on the reconstruction differences. This achieves millisecond-level synchronization and interaction between the digital model and the physical system. A continuous real-time data stream triggers instantaneous updates to the topology graph node states, edge weights, and connectivity relationships, ensuring the model remains synchronized with reality. By calculating the differences in the graph structure between two reconstructions, the system can automatically identify newly emerging bottlenecks, shifted delays, or alleviated congestion. This decision-making mechanism possesses agility similar to biological neural reflexes; strategy generation no longer relies on preset fixed periods or trigger thresholds but stems directly from instantaneous changes in the system state. This enables precise, proactive, and adaptive adjustments to warehousing and distribution resource scheduling and routing planning, reducing the time lag from detecting anomalies to executing responses. Attached Figure Description

[0018] Figure 1 is a timing diagram of the EWTMS full-link intelligent warehousing and distribution management system described in this invention; Figure 2 is a flowchart of digital mirror mapping construction; Figure 3 is a flowchart of dynamic policy allocation instruction generation; Figure 4 is a heat map of network node throughput time window distribution; Figure 5 is a timing diagram of task requests and completions in the performance bottleneck area of ​​the EWTMS system. Detailed Implementation

[0019] 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.

[0020] Please refer to Figure 1. This invention provides an EWTMS end-to-end intelligent warehousing and distribution management system. The system includes: a digital mirror construction module that establishes a digital mirror mapping covering both warehousing and distribution, which consists of a physical space mirror of the warehousing stage and a process network mirror of the distribution stage; a dynamic twin evolution module that receives and processes the raw data stream of historical warehousing and distribution operations, and uses this data stream to drive the aforementioned digital mirror mapping to perform retrospective dynamic evolution simulation; based on the output of the evolution simulation, a bottleneck identification coupling module locates the efficiency bottleneck area in the physical space mirror and identifies latency-sensitive nodes in the process network mirror, thereby logically associating the two types of bottlenecks and integrating them to generate a dynamic vulnerability topology map of the entire chain; and a strategy dynamic allocation module, as the execution end, continuously monitors real-time cargo movement data and equipment status signals, using this real-time data to stimulate and drive the dynamic vulnerability topology map to be updated and reconstructed in real time. By analyzing the topology differences before and after reconstruction, it automatically generates and issues dynamic allocation instructions for warehousing and distribution resources, thereby completing the end-to-end intelligent management cycle from perception, analysis to decision-making and execution.

[0021] In one embodiment of the present invention, referring to Figure 2, the digital mirror mapping consists of a physical space mirror of the warehousing process and a process network mirror of the distribution process. The system receives location coordinates, storage status, handling equipment identification, and operating speed from the warehousing management unit. This data defines the physical attributes, occupancy, available equipment, and performance of each storage location within the warehouse. Based on the location coordinates, storage status, handling equipment identification, and operating speed, the system constructs a physical space mirror that accurately represents the physical layout and resource distribution within the warehouse. Simultaneously, the system receives node locations, path connections, transportation vehicle attributes, and turnover time consumption from the distribution management unit. This data characterizes the geographical topology and transportation capacity of the distribution network. Based on the node locations, path connections, transportation vehicle attributes, and turnover time consumption, the system constructs a process network mirror that represents the movement paths and order dependencies of goods. The system aligns and binds the physical space mirror and the process network mirror on a unified timeline to ensure the temporal logical consistency of warehousing events and distribution events, thereby forming a unified, temporally synchronized digital mirror mapping of warehousing and distribution operations.

[0022] In practical implementation, the system establishes a digital mirror mapping of warehousing and distribution operations. This digital mirror mapping consists of a physical space mirror of the warehousing process and a process network mirror of the distribution process. In a specific example scenario, assuming the digital modeling of a large e-commerce warehouse and its corresponding regional distribution network is being performed, the implementation process begins with receiving structured data from the existing business system. This includes receiving location coordinates, storage status, handling equipment identification, and operating rates from the warehouse management unit. Location coordinate data defines the specific spatial location of each standard pallet location or light-duty racking storage location within the warehouse in three-dimensional coordinates. Storage status data reflects in real-time the commodity code, batch number, current quantity, and storage status of the goods at each location. Handling equipment identification data includes the Automated Guided Vehicle (AGV) number, stacker crane number, and forklift number. Operating rate data records in detail the horizontal movement speed, vertical lifting speed, and standard time taken for each piece of equipment to perform forklift and placement actions under both empty and full-load conditions. Based on location coordinates, storage status, handling equipment identification, and operating speed, the system constructs a physical space mirror representing the physical layout and resource distribution within the warehouse. Internally, this physical space mirror is modeled as an overlay of a 3D grid layer with timestamps and an equipment status layer. Each grid cell corresponds to a physical location and is associated with its current storage status. Each piece of equipment is modeled as a mobile intelligent agent, with attributes including real-time location, speed, task queue, and rated performance parameters. In some embodiments, the data provided by the warehouse management unit may further include attributes such as shelf load-bearing capacity, cargo volume, and storage temperature zone. This data is integrated into the physical space mirror, making its representation of the physical world more refined.

[0023] In practical implementation, the system receives node locations, path connectivity, transportation vehicle attributes, and transit time consumption from the delivery management unit. Node location data includes the latitude and longitude coordinates of the regional distribution center, the address coordinates of the last-mile delivery point, and the location information of temporary handover points. Path connectivity data describes the road connectivity between nodes in a graph structure, including attributes such as path length, road grade, and typical travel time. Transportation vehicle attribute data includes the vehicle type, load capacity, volume, average speed, and driver information. Transit time consumption data statistically analyzes the transportation time of historical orders on each path at different times, as well as the average processing time for sorting, loading / unloading, and temporary storage at each node. Based on node locations, path connectivity, transportation vehicle attributes, and transit time consumption, the system constructs a process network mirror representing the path and order dependence of goods movement. The process network mirror is modeled as a dynamic directed weighted graph, where vertices represent logistics nodes, edges represent transportation paths between nodes, and edge weights can be dynamically adjusted based on dimensions such as time, cost, or distance. Each vertex and edge is associated with a series of time-related performance parameters. In some embodiments, the path connection relationship data may include real-time traffic information, and the vehicle attribute data may include real-time fuel consumption and health status of the vehicle. This dynamic data is continuously injected into the process network mirror, enabling the process network mirror to reflect the real-time status of the delivery network.

[0024] In practice, the physical space mirror and the process network mirror are aligned and bound together in the time dimension to form a unified digital mirror mapping of warehousing and distribution operations. The core of this step is to establish a unified time benchmark and event association mechanism for the two mirrors. The system assigns a unified timestamp to every event occurring in the physical space mirror and every event occurring in the process network mirror. When an automated guided vehicle (AGV) completes the loading of a batch of goods in the physical space mirror, the timestamp of this event, the batch number of the goods, and the target vehicle information are synchronized to the process network mirror, triggering the start of movement event of the corresponding transport vehicle in the process network mirror. This binding ensures that the flow of goods from warehousing to distribution is continuous and traceable in the digital world. It can be understood that time dimension alignment includes not only the alignment of event times but also the synchronization of periodic or trend-based time patterns, such as associating the warehouse's shift schedule with the trunk line transportation departure schedule of the distribution network. Optionally, the system maintains a global order lifecycle log. Each record in the log is simultaneously indexed to the specific operation in the physical space mirror and the corresponding trip in the process network mirror, thereby achieving deep coupling between the physical space mirror and the process network mirror at the data level. This unified digital mirror mapping provides a complete, consistent, and spatiotemporally synchronized data foundation for subsequent dynamic simulation and analysis.

[0025] In one embodiment of the present invention, the process of dynamic twin evolution of the digital mirror mapping is driven by the raw data stream of historical warehousing and distribution operations. The system continuously inputs actual order information, inventory change records, equipment operation logs, and vehicle trajectory point sequences occurring within a historical time interval. These data collectively constitute the raw data stream of historical warehousing and distribution operations. Using the raw data stream as the input sequence, the system drives the physical space mirror in the digital mirror mapping to replay historical events in chronological order, simulating the historical changes in the status of storage locations and the historical fluctuations in equipment load. Similarly, using the raw data stream as the input sequence, the system drives the process network mirror in the digital mirror mapping to simulate the historical distribution of path traffic and the historical saturation of node processing capacity. Based on the twin evolution results, performance bottleneck areas are identified in the physical space mirror, while delay-sensitive nodes are identified in the process network mirror. During the historical changes and fluctuations simulated in the physical space mirror, the system calculates the ratio of the stacking rate to the clearing rate of task requests for each physical unit per unit time, and marks the set of physical units whose stacking rate consistently exceeds the clearing rate as performance bottleneck areas. During the historical distribution and saturation process of the process network mirror simulation, the system calculates the downstream waiting time increment caused by each network node when processing a unit of goods, and marks the network node whose waiting time increment exceeds the preset threshold as a delay-sensitive node.

[0026] In practical implementation, the system dynamically evolves the digital mirror mapping using twins, a process driven by the raw data stream of historical warehousing and distribution operations. In a specific example scenario, assuming historical warehousing and distribution operation data from the past thirty days is selected for backtracking simulation, the system first needs to continuously input actual order information, inventory change records, equipment operation logs, and vehicle trajectory point sequences occurring within the historical time interval. These elements together constitute the raw data stream driving the evolution of historical warehousing and distribution operations. Actual order information includes the creation time of each customer order, the details and quantity of included goods, priority, and final delivery destination. Inventory change records precisely record the operation time, location, operator, and involved product batches for each goods receipt, relocation, picking, and inventory adjustment. Equipment operation logs, sourced from the warehouse management system and the automated guided vehicle (AGV) scheduling system, record the task start time, task content, task end time, and fault alarm information for each stacker crane, AVT, and forklift. The vehicle trajectory point sequence is obtained from the onboard GPS device, which records the latitude and longitude coordinates, instantaneous speed and direction of each delivery vehicle at fixed time intervals.

[0027] In practical implementation, the raw data stream is used as the input sequence to drive the physical space mirror in the digital mirror mapping, simulating the historical changes in the status of storage locations and the historical fluctuations in equipment load. The system reads event records from the raw data stream in chronological order and advances the simulation clock of the physical space mirror with minute or second-level time steps. For the historical changes in storage location status, the system adds virtual goods objects to the corresponding coordinates in the physical space mirror based on the inbound records and updates the storage status; removes virtual goods objects and releases the storage locations based on the outbound records; and updates the coordinates of the virtual goods objects in the physical space mirror based on the shift records. For the historical fluctuations in equipment load, the system adds task items to the pending task queue of the corresponding handling equipment virtual model based on the equipment task allocation records; removes task items from the queue based on the task completion records; and calculates the instantaneous value of the queue length. Simultaneously, the system records the actual execution time of the handling equipment virtual model in each task and compares it with the standard operating rate to calculate the real-time load rate and idle rate, thus fully reproducing the fluctuations in the operational intensity and equipment busyness of various areas of the warehouse over historical periods.

[0028] In practical implementation, the raw data stream is used as the input sequence to drive the process network mirror in the digital mirror mapping, simulating the historical distribution of path traffic and the historical saturation of node processing capacity. The system also processes delivery-related events in the raw data stream in chronological order. For the historical distribution of path traffic, the system replays the movement trajectory of each vehicle on the road network map of the process network mirror based on the vehicle trajectory point sequence, and counts the total number of vehicles and the total weight of goods passing through different delivery directions of each road segment virtual edge within each time window, such as every fifteen minutes. For the historical saturation of node processing capacity, the system determines the timestamp of each batch of goods arriving at each network node virtual vertex based on order information and vehicle trajectories. Combined with node operation records, it calculates the waiting time and processing time of goods at the network node virtual vertex, and then counts the average backlog of goods, processing time distribution, and peak number of concurrent processing tasks for each network node virtual vertex in different time periods, simulating the historical throughput pressure changes of key nodes such as the distribution center sorting line and transfer station.

[0029] In practical implementation, based on twin evolution results, performance bottleneck areas are identified in the physical space mirror, while latency-sensitive nodes are marked in the process network mirror. For the physical space mirror, the system analyzes historical changes in storage location status and historical fluctuations in equipment load obtained from simulation. For each physical unit, such as a specific aisle, a set of shelves, or a sorting station, the ratio of its task request stacking rate to its clearing rate per unit time is calculated. The stacking rate is obtained by statistically analyzing the trend of the number of new tasks received by the physical unit within a continuous time window, and the clearing rate is obtained by statistically analyzing the trend of the number of tasks completed by the physical unit within the same time window. The system applies a judgment rule: when the stacking rate of a physical unit continuously exceeds its clearing rate for multiple consecutive preset time windows, the physical unit is considered to be in a state of continuous congestion. The system marks all physical units that meet this condition in the physical space mirror and aggregates adjacent marked physical units in spatial location. The resulting continuous area is defined as the performance bottleneck area, which may be a set of high-frequency picking aisles in warehouse area A.

[0030] In practical implementation, for the process network mirror, the system analyzes the simulated path traffic distribution data and node processing capacity data to calculate the downstream waiting time increment caused by each network node when processing a unit of goods. The system retrieves the data recorded in the process network mirror to obtain the actual timestamp of each batch of goods received by the downstream node and processed by the upstream network node. Simultaneously, it determines the expected arrival timestamp of the batch of goods based on the standard path duration or planned timetable; the difference between the two is the single-time waiting time. For each network node, the system statistically analyzes all batches of goods processed within a historical period and calculates the average single-time waiting time caused by all batches, which is taken as the average waiting time increment of the network node. The system compares this average waiting time increment with a preset global or hierarchical delay threshold, and identifies network nodes with an average waiting time increment greater than or equal to the delay threshold as delay-sensitive nodes in the process network mirror. It can be understood that the identification of efficiency bottleneck areas focuses on the spatial local decline in work efficiency, while the identification of delay-sensitive nodes focuses on the time delay transmission in process links. In some embodiments, the unit time window length used to calculate the backlog rate and clearing rate can be adjusted according to the business rhythm, for example, a shorter window can be used during peak promotional periods to capture faster dynamic changes. Optionally, a weighted average can be used when calculating the average wait time increment, giving higher weight to recent data so that the identification results better reflect the recent network status.

[0031] In one embodiment of the present invention, referring to Figure 3, performance bottleneck regions and latency-sensitive nodes are logically coupled to generate a dynamic vulnerability topology graph of the entire link. The system first establishes a causal relationship mapping between performance bottleneck regions and latency-sensitive nodes. This mapping describes how congestion in a specific performance bottleneck region leads to waiting at a specific latency-sensitive node. The performance bottleneck regions marked in the physical space mirror and the latency-sensitive nodes marked in the process network mirror, along with their causal relationship mapping, are jointly encoded into a weighted directed graph structure, which is the dynamic vulnerability topology graph. The node weights in the dynamic vulnerability topology graph are determined by the accumulation rate or waiting time increment of the corresponding region, and the edge weights are determined by the strength of the causal relationship. The specific encoding process is as follows: Each performance bottleneck region is abstracted into a graph node, and a weight value is assigned to this graph node. The weight value is obtained by normalizing the ratio of the accumulation rate to the clearing rate of the performance bottleneck region. Each latency-sensitive node is abstracted into another graph node, and a weight value is assigned to this graph node. The weight value is obtained by normalizing the average waiting time increment of the latency-sensitive node. Based on the causal association mapping, the existence and direction of the edges pointing from the performance bottleneck region nodes to the latency-sensitive nodes are determined. Each existing edge is assigned a weight value, which is determined by the number and severity of the increase in the waiting time of the associated latency-sensitive nodes caused by the deterioration of the corresponding performance bottleneck region in historical data. All weighted nodes and weighted edges are combined to construct a complete weighted directed graph structure, namely the dynamic vulnerability topology graph.

[0032] Based on real-time cargo movement data and equipment status signals, the system drives the dynamic vulnerability topology graph to be reconstructed in real time. Dynamic allocation instructions for warehousing and distribution strategies are generated based on the reconstruction differences. The system collects real-time order sorting information, the location of goods in transit, the real-time load rate of handling equipment, and the real-time speed of transport vehicles. This information constitutes real-time cargo movement data and equipment status signals. These real-time cargo movement data and equipment status signals are used as stimulus inputs to update the weight values ​​of corresponding nodes and the connection strength of edges in the dynamic vulnerability topology graph. The system compares the dynamic vulnerability topology graph before and after the update, identifying newly emerging nodes or edges whose weights exceed the activation threshold. The physical locations or process links associated with these newly emerging nodes or edges are identified as targets requiring immediate intervention. Based on the targets requiring immediate intervention, the system calculates resource reallocation schemes or path adjustment schemes, forming specific dynamic allocation instructions for warehousing and distribution strategies.

[0033] In its implementation, the system logically couples performance bottleneck areas and latency-sensitive nodes to generate a dynamic vulnerability topology map of the entire link. The system first establishes a causal relationship mapping between performance bottleneck areas and latency-sensitive nodes. This mapping describes how congestion in a specific performance bottleneck area leads to waiting times at a specific latency-sensitive node. In a specific example scenario, the warehouse area designated WH-Zone-A is identified as the performance bottleneck area in the physical space mirror, and the distribution center node designated DC-Node-X is identified as the latency-sensitive node in the process network mirror. The system retrieves the raw data stream of historical warehousing and distribution operations and analyzes that when the equipment load rate in the WH-Zone-A area exceeds 85% and the task queue continues to grow, subsequent shipments from the WH-Zone-A area to the DC-Node-X node generally experience delays in actual arrival time compared to the planned time, with a significant increase in the proportion of batches delayed by more than 30 minutes. Based on this data pattern, the system establishes a causal relationship mapping between the WH-Zone-A performance bottleneck region and the DC-Node-X latency-sensitive node. This mapping records the number of historical causal events, the average latency increment, and the correlation strength.

[0034] In practical implementation, the performance bottleneck regions marked in the physical space mirror and the latency-sensitive nodes marked in the process network mirror, along with the causal relationship mapping, are jointly encoded into a weighted directed graph structure, which is the dynamic vulnerability topology graph. The node weights in the dynamic vulnerability topology graph are determined by the accumulation rate or waiting time increment of the corresponding region, while the edge weights are determined by the strength of the causal relationship. The encoding process specifically includes abstracting each performance bottleneck region into a graph node and assigning a weight value to each node. The weight value is obtained by normalizing the ratio of the accumulation rate to the clearing rate of the performance bottleneck region. For example, the historical average ratio of the accumulation rate to the clearing rate of region WH-Zone-A is 1.5; after maximum and minimum value normalization, its node weight value is set to 0.75. Simultaneously, each latency-sensitive node is abstracted into another graph node and assigned a weight value. The weight value is obtained by normalizing the average waiting time increment of the latency-sensitive node. For example, the average waiting time increment of the DC-Node-X node is 45 minutes, which is normalized across all nodes in the system to obtain a weight value of 0.80.

[0035] In practical implementation, the existence and direction of edges pointing from performance bottleneck nodes to latency-sensitive nodes are determined based on causal relationship mapping. In the example, due to the existence of a causal relationship mapping from WH-Zone-A to DC-Node-X, the system creates a directed edge from node WH-Zone-A to node DC-Node-X. Each existing edge is assigned a weight value, which is determined by the number and severity of instances in historical data where the deterioration of the corresponding performance bottleneck state leads to an increase in the waiting time of the associated latency-sensitive node. In some embodiments, the edge weight can be quantitatively calculated using the following formula:

[0036] Where: symbol Represents the node Pointing to node The weight of the edge, sign This represents the performance bottleneck area within the historical observation period. The state deteriorates and causes latency-sensitive nodes to become unstable. Total number of events that cause additional wait times, symbol Represents the node in the i-th event. The difference between the actual latency increment and the baseline latency, sign... Represents the node in the i-th event. Quantitative indicators of the degree of condition deterioration, such as the extent to which the load rate exceeds the limit, and the sign. Represents the difference among all N events The maximum value is used for normalization. All weighted nodes and weighted edges are combined to construct a complete weighted directed graph structure, namely the dynamic vulnerability topology graph. The dynamic vulnerability topology graph is stored and maintained in the system using an adjacency list or adjacency matrix data structure, and allows for dynamic updates.

[0037] In practical implementation, the system drives the dynamic vulnerability topology map to be reconstructed in real time based on real-time cargo movement data and equipment status signals. Dynamic allocation instructions for warehousing and distribution strategies are generated based on the reconstruction differences. The system collects current order sorting information, the location of goods in transit, the real-time load rate of handling equipment, and the real-time speed of transport vehicles in real time. These data constitute real-time cargo movement data and equipment status signals.

[0038] In one embodiment of the present invention, the original data stream is used as the input sequence to drive the physical space mirror in the digital mirror mapping, simulating the historical changes in the status of storage locations and the historical fluctuations in equipment load. Specific operations include: extracting inbound records, outbound records, and storage location adjustment records for each moment within the historical time interval from the original data stream; updating the occupancy status, goods type, and storage duration of each storage location in the physical space mirror step-by-step based on the inbound, outbound, and storage location adjustment records. Simultaneously, extracting equipment task allocation records and task completion records for each moment within the historical time interval from the original data stream; and calculating the queue length of pending tasks, task execution rate, and idle rate for each handling device in the physical space mirror step-by-step based on the equipment task allocation records and task completion records.

[0039] Using the raw data stream as input, this method drives a process network mirror in a digital mirror mapping, simulating the historical distribution of path traffic and the historical saturation of node processing capacity. Specific operations include: extracting the quantity of goods, the number of vehicles, and the passage timestamps for each transportation path within a historical time interval from the raw data stream; and calculating the total quantity of goods and vehicles flowing through each path within each time window in the process network mirror based on these data. Simultaneously, it extracts the goods arrival record, processing start timestamp, and processing end timestamp for each network node within the historical time interval from the raw data stream; and calculating the average processing time and maximum concurrent processing capacity of each network node within each time window in the process network mirror based on these data.

[0040] In practical implementation, the system uses the raw data stream as the input sequence to drive the physical space mirror in the digital mirror mapping, simulating the historical changes in the status of the storage location and the historical fluctuations in the equipment load. The specific operations include extracting the inbound records, outbound records, and storage location adjustment records for each moment in the historical time interval from the raw data stream. The inbound records contain information about the goods being placed in the designated storage location after they arrive at the warehouse and are inspected. The outbound records contain information about the goods required for the order being picked from the storage location and moved to the shipping area. The storage location adjustment records contain information about the relocation of goods within the warehouse to optimize storage. Based on inbound, outbound, and location adjustment records, the system updates the occupancy status, goods type, and storage duration of each location in the physical space mirror at time steps. The physical space mirror maintains a virtual location grid that perfectly corresponds to the actual warehouse layout. Whenever an inbound record is processed, the system marks the corresponding coordinate in the virtual location grid as "occupied" and associates it with the goods' commodity code and inbound timestamp. Whenever an outbound record is processed, the system updates the status of the corresponding virtual location to "idle" and clears the associated commodity information. Whenever a location adjustment record is processed, the system migrates the commodity's associated information from one virtual location coordinate to another and updates the commodity's storage duration. The system also extracts equipment task allocation records and task completion records for each moment within the historical time interval from the raw data stream. The equipment task allocation record indicates which specific handling equipment was assigned a handling task at a specific time point, while the task completion record indicates which handling equipment completed which task at what time. Based on the equipment task allocation records and task completion records, the system calculates the length of the pending task queue, task execution rate, and idle rate for each handling device at each time step in the physical space mirror. The physical space mirror maintains a task queue data structure for each virtual model of the handling device. When processing a device task allocation record, the corresponding task item is added to the end of the task queue of the corresponding device's virtual model. When processing a task completion record, the completed task item is removed from the head of the queue. The system calculates the length of the pending task queue at any given time by monitoring changes in queue length, calculates the task execution rate by counting the number of tasks completed per unit time, and calculates the idle rate by accumulating the time the device is in a task-free state. See Table 1.

[0041] Table 1: Update Table of Cargo Location Status During Physical Space Mirror Simulation

[0042] In practical implementation, the raw data stream is used as the input sequence to drive the process network mirror in the digital mirror mapping, simulating the historical distribution of path traffic and the historical saturation of node processing capacity. Specific operations include extracting the quantity of goods, the number of vehicles, and the passage timestamp for each transportation path within a historical time interval from the raw data stream. The quantity of goods on the transportation path comes from waybill data, identifying the total number or weight of goods passing through a specific path in a given transport. The number of vehicles comes from scheduling logs or GPS trajectory aggregation, and the passage timestamp precisely records the start and end times of vehicles entering and leaving a path. Based on the quantity of goods, the number of vehicles, and the passage timestamp, the total quantity of goods and the total number of vehicles flowing through each path within each time window are counted in the process network mirror. Each virtual edge of a path in the process network mirror is associated with a time series database. The system accumulates the quantity of goods and the number of vehicles for each transport into the corresponding time window interval based on the passage timestamp, for example, every 15 minutes or every hour, thereby reconstructing the historical traffic fluctuation curve. The system extracts cargo arrival records, processing start timestamps, and processing end timestamps for each network node within a historical time interval from the raw data stream. The cargo arrival record indicates the physical time when a batch of goods arrives at a distribution center or transit node. The processing start timestamp records the time when the batch of goods begins sorting or loading / unloading operations within the node. The processing end timestamp records the time when the batch of goods completes all processing steps within the node and is ready to leave. Based on the cargo arrival records, processing start timestamps, and processing end timestamps, the system calculates the average processing time and maximum concurrent processing capacity of each network node within each time window in the process network mirror. The system first calculates the processing time for each batch of goods within the node, i.e., the difference between the processing end timestamp and the processing start timestamp. Then, it averages the processing times of all batches within each time window to obtain the average processing time for that window. The maximum concurrent processing capacity is determined by analyzing the peak number of cargo batches in the "processing" state within the node at the same time point.

[0043] In practical implementation, process network mirroring can quantify the historical processing pressure of nodes through simulation. For example, the system calculates the throughput efficiency index of network nodes during a certain peak period, which is defined by the following formula: Where: symbol Represents the throughput efficiency of a network node within a selected time window, symbol Represents the total number of all shipments processed by the network node within that time window, denoted by [symbol]. Represents the number of standard containers or standard weight units contained in the b-th shipment, symbol Represents the length of the time window, symbol This represents the historical average time it takes for a network node to process a single standard box or standard weight unit.

[0044] Referring to Figure 4, in the process network mirror simulation of the EWTMS end-to-end intelligent warehousing and distribution management system, the graph uses the time window (08:00:00 to 18:00:00) as the horizontal axis and network nodes (distribution point E, distribution center A, etc.) as the vertical axis. A color gradient (corresponding to throughput values ​​of 400-1400) visualizes the throughput distribution of each node at different times. Specifically, the values ​​in the graph represent the amount of goods processed by the corresponding node within that time window; the darker the color (e.g., dark green), the higher the throughput, and the lighter the color (e.g., dark red), the lower the throughput. For example, distribution point E has a throughput of 1560 (dark green) at 08:00:00, while distribution center A has a throughput of only 380 (dark red) at 12:00:00. This heatmap is a visual output of the "historical saturation process simulation of node processing capacity" in the process network mirror, which can support the subsequent identification of delay-sensitive nodes. By analyzing the throughput fluctuations of each node in the time dimension, combined with the downstream waiting time increment per unit of goods processed, delay-sensitive nodes can be located.

[0045] In one embodiment of the present invention, the ratio of the accumulation rate to the clearing rate of task requests for each physical unit within a unit time is calculated, and the set of physical units whose accumulation rate consistently exceeds the clearing rate is marked as a performance bottleneck region. The specific steps include: for each physical unit, counting the number of task requests received within a consecutive multiple unit time windows, and calculating the growth rate of the number of task requests within each time window as the accumulation rate; counting the number of tasks completed by each physical unit within the same consecutive multiple unit time windows, and calculating the growth rate of the number of completed tasks within each time window as the clearing rate; for each physical unit, calculating the ratio of its accumulation rate to the clearing rate within each time window; when the ratio of the accumulation rate to the clearing rate of a physical unit is greater than one within a consecutive preset number of time windows, the physical unit is marked as a potential bottleneck unit; and clustering the spatially adjacent physical units that are simultaneously marked as potential bottleneck units to form a continuous region, which is the performance bottleneck region.

[0046] The calculation of the downstream waiting time increment caused by each network node when processing a unit of goods, and the marking of network nodes whose waiting time increment exceeds a preset threshold as delay-sensitive nodes, includes the following steps: For each network node, obtain the timestamp record of each batch of goods that has been processed and delivered to the downstream node; calculate the difference between the actual timestamp and the expected arrival timestamp of each batch of goods received by the downstream node as the single waiting time caused by each batch of goods; for each network node, count all batches of goods processed within a historical period, calculate the average single waiting time caused by all batches as the average waiting time increment of the network node; compare the average waiting time increment of each network node with the system's preset delay threshold; and mark network nodes whose average waiting time increment is greater than or equal to the delay threshold as delay-sensitive nodes.

[0047] In practical implementation, the system calculates the ratio of the accumulation rate to the clearing rate of task requests in each physical unit per unit time. Physical units whose accumulation rate consistently exceeds the clearing rate are marked as performance bottleneck areas. Specific steps include, for each physical unit, counting the number of task requests received within multiple consecutive unit time windows, and calculating the growth rate of the number of task requests within each time window as the accumulation rate. In a specific example scenario, "High-Frequency Picking Lane A1" in the warehouse is selected as a physical unit for analysis. The unit time window is set to 15 minutes. The system continuously counts the number of new picking tasks received in this lane within 12 time windows between 8:00 and 11:00, recorded as the sequence [10, 15, 22, 18, 25, 30, 28, 35, 40, 38, 45, 50]. For each time window, the accumulation rate is obtained by dividing the difference between the number of tasks in the current window and the number of tasks in the previous window by the number of tasks in the previous window. For example, the accumulation rate of the second window is (15-10) / 10=0.5, representing the growth rate of task requests. The system counts the number of tasks completed in each physical unit within the same consecutive time windows, and calculates the growth rate of the number of completed tasks within each time window as the clearing rate. The system synchronously counts the actual number of tasks completed in "High-Frequency Picking Lane A1" within the same 12 time windows, recording it as a sequence [8, 12, 15, 14, 20, 22, 25, 28, 30, 32, 35, 36]. Similarly, the system calculates the growth rate of the number of completed tasks in each window relative to the previous window as the clearing rate. For example, the clearing rate for the second window is... .

[0048] In practice, for each physical unit, the ratio of its stacking rate to its emptying rate within each time window is calculated. The system performs element-wise division on the calculated stacking rate and emptying rate values ​​for each window to obtain a sequence of ratios. When the ratio of a physical unit's stacking rate to its emptying rate is greater than one for a consecutive preset number of time windows, the physical unit is marked as a potential bottleneck unit. The preset number can be set to three consecutive windows. In the example, assuming that the calculated ratios from the 4th to the 6th time windows are 1.29, 1.25, and 1.36 respectively, and these three ratios are all greater than one, the system marks "High-Frequency Picking Lane A1" as a potential bottleneck unit. The system clusters adjacent physical units that are simultaneously marked as potential bottleneck units, forming a continuous region that is the performance bottleneck region. For example, if adjacent "high-frequency picking lane A2" and "high-frequency picking lane A3" are also marked as potential bottleneck units based on their respective data, the system will merge and cluster them according to the physical layout coordinates of these lanes, identifying them as a continuous region called "East Side Picking Lane Group of Area A". This region is defined as the performance bottleneck region.

[0049] In practical implementation, the system calculates the downstream waiting time increment caused by each network node when processing a unit of goods. Network nodes with waiting time increments exceeding a preset threshold are marked as delay-sensitive nodes. Specific steps include obtaining the timestamp record of each batch of goods processed and delivered to the downstream node for each network node. In a specific example scenario, "Regional Distribution Center RDC1" is selected as a network node for analysis, and "City Distribution Station C1" is the downstream node. The system retrieves all batch records of goods processed by "Regional Distribution Center RDC1" and delivered to "City Distribution Station C1" from historical data. Each record includes a batch ID, the departure timestamp from "Regional Distribution Center RDC1," the planned arrival timestamp at "City Distribution Station C1," and the actual arrival timestamp at "City Distribution Station C1." The difference between the actual timestamp and the expected arrival timestamp of each batch of goods received by the downstream node is calculated as the single waiting time caused by each batch of goods. The system calculates this for each batch of goods using the formula: Single waiting time = Actual arrival timestamp - Expected arrival timestamp. For example, for goods with batch ID "T20231026001", the expected arrival timestamp is 2023-10-26 14:00:00, and the actual arrival timestamp is 2023-10-26 14:40:00, then the single waiting time is 40 minutes.

[0050] In practice, for each network node, all batches of goods processed within a historical period are statistically analyzed, and the average waiting time per batch is calculated as the average waiting time increment for the network node. The system sets the statistical period to the past 7 days. During this period, a total of 560 batches of goods were transported from "Regional Distribution Center RDC1" to "City Distribution Station C1". The system calculates the arithmetic mean of the single waiting time for these 560 batches to obtain the average waiting time increment, for example, a result of 38 minutes. The average waiting time increment of each network node is compared with a system-preset delay threshold. The delay threshold may be set according to the service level agreement. For example, for trunk transportation from the distribution center to the distribution station, the system's preset delay threshold is 30 minutes. Network nodes with an average waiting time increment greater than or equal to the delay threshold are marked as delay-sensitive nodes. In the example, the average waiting time increment of "Regional Distribution Center RDC1" of 38 minutes is greater than the preset delay threshold of 30 minutes. Therefore, the system marks "Regional Distribution Center RDC1" as a delay-sensitive node in the network topology. It is understandable that the calculation of the average waiting time increment depends on accurate timestamp records and a reliable expected time benchmark.

[0051] Referring to Figure 5, the efficiency bottleneck identification process of the EWTMS end-to-end intelligent warehousing and distribution management system presents the dynamic relationship between the number of task requests and the number of completed tasks for a specific physical unit (such as a high-frequency picking aisle) within a continuous time window. Specifically, the figure shows the temporal fluctuations of the number of task requests (blue line) and the number of completed tasks (orange line) from 8:00 to 10:45, with 15-minute time windows. The pink area, marked "continuous bottleneck windows (windows 4-6)," corresponds to the period when the system determines that the accumulation rate of the physical unit continuously exceeds the emptying rate. Within this interval, the ratio of the growth rate of the number of task requests (accumulation rate) to the growth rate of the number of completed tasks (emptying rate) is continuously greater than 1, meeting the identification criteria for potential bottleneck units. The fluctuation characteristics of the data sequence in the figure can serve as a direct observation basis for the accumulation rate and emptying rate of physical units, providing temporal quantitative support for the subsequent clustering of efficiency bottleneck regions by the system.

[0052] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0053] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. The EWTMS end-to-end intelligent warehousing and distribution management system is characterized by, The process includes the following steps: a digital mirror construction module, which establishes a digital mirror mapping of warehousing and distribution operations, wherein the digital mirror mapping consists of a physical space mirror of the warehousing process and a process network mirror of the distribution process; and a dynamic twin evolution module, which performs dynamic twin evolution on the digital mirror mapping, wherein the twin evolution process is driven by the original data stream of historical warehousing and distribution operations. The bottleneck identification coupling module, based on the twin evolution results, identifies the performance bottleneck region in the physical space mirror and identifies the latency-sensitive node in the process network mirror. It logically couples the performance bottleneck region and the latency-sensitive node to generate a dynamic vulnerability topology map of the entire link. The strategy dynamic allocation module drives the dynamic vulnerability topology map to be reconstructed in real time based on real-time cargo movement data and equipment status signals, and generates dynamic allocation instructions for warehousing and distribution strategies based on the reconstruction differences.

2. The EWTMS end-to-end intelligent warehousing and distribution management system according to claim 1, characterized in that, A digital mirror mapping for warehousing and distribution operations is established. This digital mirror mapping consists of a physical space mirror of the warehousing process and a process network mirror of the distribution process. Specifically, it includes: receiving the coordinates of the storage location, storage status, handling equipment identification, and operating rate from the warehousing management unit; constructing a physical space mirror representing the physical layout and resource distribution within the warehouse based on the storage location coordinates, storage status, handling equipment identification, and operating rate; receiving the node location, path connection relationship, transportation vehicle attributes, and circulation time consumption from the distribution management unit; constructing a process network mirror representing the movement path and sequence dependency of goods based on the node location, path connection relationship, transportation vehicle attributes, and circulation time consumption; and aligning and binding the physical space mirror and the process network mirror in the time dimension to form a unified digital mirror mapping for warehousing and distribution operations.

3. The EWTMS end-to-end intelligent warehousing and distribution management system according to claim 2, characterized in that, The digital mirror mapping undergoes dynamic twin evolution, driven by the raw data stream of historical warehousing and distribution operations. Specifically, this includes: continuously inputting actual order information, inventory change records, equipment operation logs, and vehicle trajectory point sequences occurring within a historical time interval. These elements constitute the raw data stream of historical warehousing and distribution operations. Using this raw data stream as the input sequence, the physical space mirroring in the digital mirror mapping is driven to simulate the historical changes in cargo location status and the historical fluctuations in equipment load. Furthermore, using the raw data stream as the input sequence, the process network mirroring in the digital mirror mapping is driven to simulate the historical distribution of path traffic and the historical saturation of node processing capacity.

4. The EWTMS end-to-end intelligent warehousing and distribution management system according to claim 3, characterized in that, Based on twin evolution results, performance bottleneck regions are identified in the physical space mirror, and delay-sensitive nodes are marked in the process network mirror. Specifically, this includes: calculating the ratio of the stacking rate to the emptying rate of task requests per unit time for each physical unit during the historical changes in the simulated cargo location status and the historical fluctuations in equipment load in the physical space mirror, and marking the set of physical units whose stacking rate consistently exceeds the emptying rate as performance bottleneck regions; and calculating the downstream waiting time increment caused by each network node when processing a unit of cargo during the historical distribution of simulated path traffic and the historical saturation of node processing capacity in the process network mirror, and marking network nodes whose waiting time increment exceeds a preset threshold as delay-sensitive nodes.

5. The EWTMS end-to-end intelligent warehousing and distribution management system according to claim 4, characterized in that, Logically coupling the performance bottleneck region and the latency-sensitive node generates a dynamic vulnerability topology graph for the entire link. Specifically, this includes: establishing a causal relationship mapping between the performance bottleneck region and the latency-sensitive node, whereby the causal relationship mapping describes how congestion in a specific performance bottleneck region leads to waiting times in a specific latency-sensitive node; encoding the performance bottleneck region marked in the physical space mirror, the latency-sensitive node marked in the process network mirror, and the causal relationship mapping together into a weighted directed graph structure, which is the dynamic vulnerability topology graph; the node weights in the dynamic vulnerability topology graph are determined by the accumulation rate or waiting time increment of the corresponding region, and the edge weights are determined by the strength of the causal relationship; the encoding of the performance bottleneck region marked in the physical space mirror, the latency-sensitive node marked in the process network mirror, and the causal relationship mapping together into a weighted directed graph structure, which is the dynamic vulnerability topology graph; the node weights in the dynamic vulnerability topology graph are determined by the accumulation rate or waiting time increment of the corresponding region, and the edge weights are determined by the strength of the causal relationship; the encoding of the performance bottleneck region marked in the physical space mirror, the latency-sensitive node marked in the process network mirror, and the causal relationship mapping together into a weighted directed graph structure... A weighted directed graph structure specifically includes: abstracting each performance bottleneck region into a graph node and assigning a weight value to the graph node, the weight value being obtained by normalizing the ratio of the accumulation rate to the clearing rate of the performance bottleneck region; abstracting each latency-sensitive node into another graph node and assigning a weight value to the graph node, the weight value being obtained by normalizing the average waiting time increment of the latency-sensitive node; determining the existence and direction of edges pointing from performance bottleneck region nodes to latency-sensitive nodes according to the causal association mapping; assigning a weight value to each existing edge, the weight value being jointly determined by the number and severity of the increase in waiting time of associated latency-sensitive nodes caused by the deterioration of the corresponding performance bottleneck region state in historical data; and combining all weighted nodes and weighted edges to construct a complete weighted directed graph structure, namely the dynamic vulnerability topology graph.

6. The EWTMS end-to-end intelligent warehousing and distribution management system according to claim 5, characterized in that, Based on real-time cargo movement data and equipment status signals, the dynamic vulnerability topology is driven to be reconstructed in real time. Dynamic allocation instructions for warehousing and distribution strategies are generated based on the reconstruction differences. Specifically, this includes: real-time collection of current order sorting information, the location of goods in transit, the real-time load rate of handling equipment, and the real-time speed of transport vehicles. The current order sorting information, the location of goods in transit, the real-time load rate of handling equipment, and the real-time speed of transport vehicles constitute real-time cargo movement data and equipment status signals. These real-time cargo movement data and equipment status signals are used as stimulus inputs to update the weight values ​​of corresponding nodes and the connection strength of edges in the dynamic vulnerability topology. The dynamic vulnerability topology before and after the update is compared, and newly emerging nodes or edges with weights exceeding the activation threshold are identified. The physical locations or process links associated with these newly emerging nodes or edges are identified as targets requiring immediate intervention. Based on the targets requiring immediate intervention, resource reallocation schemes or path adjustment schemes are calculated to form dynamic allocation instructions for warehousing and distribution strategies.

7. The EWTMS end-to-end intelligent warehousing and distribution management system according to claim 3, characterized in that, Using the original data stream as the input sequence, the physical space mirror in the digital mirror mapping is driven to simulate the historical changes in the status of storage locations and the historical fluctuations in equipment load. Specifically, this includes: extracting inbound records, outbound records, and storage location adjustment records for each moment within the historical time interval from the original data stream; updating the occupancy status, goods type, and storage duration of each storage location in the physical space mirror at each time step based on the inbound records, outbound records, and storage location adjustment records; extracting equipment task allocation records and task completion records for each moment within the historical time interval from the original data stream; and calculating the queue length of pending tasks, task execution rate, and idle rate of each handling device in the physical space mirror at each time step based on the equipment task allocation records and task completion records.

8. The EWTMS end-to-end intelligent warehousing and distribution management system according to claim 3, characterized in that, Using the original data stream as the input sequence, the process network mirror in the digital mirror mapping is driven to simulate the historical distribution process of path traffic and the historical saturation process of node processing capacity. Specifically, this includes: extracting the quantity of goods, the number of vehicles, and the passage timestamp for each transportation path within the historical time interval from the original data stream; calculating the total quantity of goods and the total number of vehicles flowing through each path within each time window in the process network mirror based on the quantity of goods, the number of vehicles, and the passage timestamp; extracting the goods arrival record, processing start timestamp, and processing end timestamp for each network node within the historical time interval from the original data stream; and calculating the average processing time and maximum concurrent processing capacity of each network node within each time window in the process network mirror based on the goods arrival record, the processing start timestamp, and the processing end timestamp.

9. The EWTMS end-to-end intelligent warehousing and distribution management system according to claim 4, characterized in that, The process involves calculating the ratio of the accumulation rate to the clearing rate of task requests for each physical unit within a unit of time. Physical units whose accumulation rate consistently exceeds their clearing rate are designated as performance bottleneck regions. Specifically, this includes: for each physical unit, counting the number of task requests received within multiple consecutive unit time windows and calculating the growth rate of the number of task requests within each time window as the accumulation rate; counting the number of tasks completed by each physical unit within the same multiple consecutive unit time windows and calculating the growth rate of the number of completed tasks within each time window as the clearing rate; for each physical unit, calculating the ratio of its accumulation rate to its clearing rate within each time window; when a physical unit's ratio of its accumulation rate to its clearing rate is greater than one within a predetermined number of consecutive time windows, the physical unit is designated as a potential bottleneck unit; and clustering spatially adjacent physical units that are simultaneously designated as potential bottleneck units to form a continuous region, which is the performance bottleneck region.

10. The EWTMS end-to-end intelligent warehousing and distribution management system according to claim 4, characterized in that, The calculation of the downstream waiting time increment caused by each network node when processing a unit of goods, and the marking of network nodes whose waiting time increment exceeds a preset threshold as delay-sensitive nodes, specifically includes: for each network node, obtaining the timestamp record of each batch of goods that has been processed and delivered to the downstream node; calculating the difference between the actual timestamp and the expected arrival timestamp of each batch of goods received by the downstream node as the single waiting time caused by each batch of goods; for each network node, statistically analyzing all batches of goods processed within a historical period, calculating the average single waiting time caused by all batches as the average waiting time increment of the network node; comparing the average waiting time increment of each network node with a system-preset delay threshold; and marking network nodes whose average waiting time increment is greater than or equal to the delay threshold as delay-sensitive nodes.

Images