Intelligent warehouse logistics management and control method and system based on multi-source data fusion

Abstract

The application discloses an intelligent warehouse logistics management and control method and system based on multi-source data fusion, belongs to the technical field of intelligent warehouse logistics management and control, and aims to solve the problems of different data caliber of existing warehouse multi-source data, weak material supply and demand linkage, insufficient wave scheduling precision, and low sorting efficiency. By fusing full-link logistics data to build a unified space-time data benchmark, mining material supply and demand correlation characteristics to optimize inventory management and demand prediction, checking orders in time sequence and locking available inventory, combining multi-constraint optimization and arrival triggering mechanism to complete intelligent wave aggregation, and supporting dynamic path planning to land sorting scheduling, the application can systematically improve warehouse sorting efficiency, inventory guarantee precision and order response speed.

Description

Technical Field

[0001] This invention relates to the field of intelligent warehousing and logistics management technology, and more specifically to an intelligent warehousing and logistics management method and system based on multi-source data fusion. Background Technology

[0002] Currently, the categories of materials in smart warehousing scenarios are continuously expanding, and the proportion of small-batch, multi-batch outbound demand is constantly increasing. Frequent demand plan changes place higher demands on the accuracy of warehouse inventory management, order fulfillment response speed, and sorting operation efficiency. In current warehousing operations, business data from the supply, warehousing, and demand stages are scattered and inconsistent in definition, making it difficult to form a unified data support system. This results in insufficient timeliness and accuracy of inventory status. Inventory management is often carried out independently on a single material basis, without fully considering the supply-demand linkage and matching relationships between materials. This easily leads to unbalanced inventory structures and low matching rates, affecting overall order fulfillment efficiency. The scheduling of outbound orders and the planning of sorting operations lack refinement, resulting in delayed order response. The rationality of sorting operation paths and task arrangement needs improvement, and overall operational efficiency is difficult to adapt to the high-frequency changing business demands. To a certain extent, this restricts the operational quality and efficiency of warehousing logistics and fails to fully meet the warehousing management needs in multi-category, rapidly changing scenarios. Therefore, to overcome these limitations, this invention proposes a smart warehousing logistics management method and system based on multi-source data fusion. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide an intelligent warehousing and logistics management method and system based on multi-source data fusion, which solves the problems of insufficient inventory management accuracy, delayed order fulfillment response, and low overall sorting efficiency in warehousing scenarios with a wide variety of materials and frequent demand changes.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] Intelligent warehousing and logistics management methods based on multi-source data fusion include:

[0006] Collect multi-source logistics data from the supply side, warehousing side, and demand side, perform standardized cleaning, granular decomposition and time-series alignment processing, and build a unified spatiotemporal data benchmark;

[0007] Based on the unified spatiotemporal data benchmark, the supply-side co-transportation correlation characteristics and demand-side co-demand correlation characteristics among material categories are mined to construct a material correlation feature map, which is used to perform inventory level monitoring of single material categories and completeness monitoring of significantly related material groups, generate demand forecast results for each material category, and update the time-series inventory data of the corresponding material categories.

[0008] Calculate the dynamic available inventory for each material category, use the outbound time of the outbound order as the verification node, perform linked time-series verification on the outbound order, and construct a set of executable outbound orders;

[0009] Based on a unified spatiotemporal data benchmark, rigid constraints are extracted and a multi-objective optimization function is constructed. Intelligent wave aggregation is performed on the executable outbound order set to generate a sorting wave plan and reconstruct the execution sequence of picking tasks within each sorting wave.

[0010] Based on the sorting wave operation plan of the execution sequence, dynamic path planning is performed to issue picking tasks to the corresponding operation resources.

[0011] Specifically, constructing a unified spatiotemporal data benchmark includes:

[0012] It receives multi-source logistics data from the supply side, warehousing side, and demand side. After cleaning the multi-source logistics data, it performs field mapping and format unification processing on the multi-source logistics data according to the preset data field specifications to generate a standardized logistics dataset.

[0013] The standardized logistics dataset is decomposed into multi-source logistics data at the material category level, and the inbound operation feature parameters corresponding to each material category are extracted to update the standardized logistics dataset.

[0014] Using a continuous time axis as the baseline dimension and material category as the smallest management unit, the updated standardized logistics dataset is aligned in the time dimension and associated in the spatial dimension to construct a unified spatiotemporal data benchmark that includes inventory level time series and spatial location attributes.

[0015] Specifically, constructing a material association feature map includes:

[0016] Call up historical inbound and outbound data within a preset historical period, and perform cleaning and verification processing based on material category codes to form a sample dataset for correlation analysis;

[0017] Based on the aforementioned correlation analysis sample dataset, correlation statistics were performed from both the supply and demand sides:

[0018] For the supply side, count the number of times any two material categories appear together in the same inbound batch, and calculate the co-transport correlation degree of the corresponding material category pair;

[0019] On the demand side, count the number of times any two material categories appear together in the same outbound order, and calculate the co-demand correlation degree of the corresponding material category pair;

[0020] The co-transportation correlation and co-demand correlation of each material category pair are compared with the preset correlation thresholds to identify strong correlation relationships and mark the strong correlation attributes on the supply side and the strong correlation attributes on the demand side.

[0021] Using strong correlations as graph connection edges and co-operation correlation degree or co-demand correlation degree as edge weight parameters, a material correlation feature graph containing supply correlation sub-graph and demand correlation sub-graph is constructed.

[0022] Specifically, this includes monitoring the inventory level of individual material categories and the availability status of significantly related material groups, including:

[0023] Retrieve real-time inventory data for each material category. When the real-time inventory level of a material category is lower than the corresponding preset safety stock threshold, generate an inventory shortage warning for the corresponding material category.

[0024] Based on the significantly associated material groups in the material association feature map, real-time inventory data of all material categories in each significantly associated material group is retrieved. Combined with the material matching ratio corresponding to each significantly associated material group, the inventory matching rate of each significantly associated material group is calculated. When the inventory matching rate is lower than the preset matching rate threshold, the matching supply risk warning information of the corresponding significantly associated material group is generated.

[0025] Specifically, demand forecasts for each material category are generated, including:

[0026] Historical outbound data within a preset prediction period is retrieved from the unified spatiotemporal data benchmark. The future demand trend of each material category is fitted and calculated, and the expected outbound demand for the corresponding time period is output in a preset time granularity.

[0027] Based on the demand correlation sub-map in the material correlation feature map, the core driving material and supporting follower material of each significantly correlated material group are identified, the demand linkage coefficient of each material in the significantly correlated material group is calculated, and the deviation of the basic demand prediction result is corrected based on the basic demand prediction result of the core driving material to obtain the final demand prediction result of each material category.

[0028] Specifically, a set of executable outbound orders is constructed, including:

[0029] By combining demand forecast results, current real-time inventory data, expected inbound data for each future period, and pre-allocated outbound data, the dynamic available inventory of each material category is calculated periodically according to the preset time granularity, forming a time-segmented dynamic available inventory sequence.

[0030] Using the outbound time of the outbound order as the verification node, and combining the dynamic available inventory sequence of the corresponding time period, perform a full material category linkage time sequence verification to determine the execution status of the outbound order, identify outbound orders that can be executed immediately, and split outbound orders that cannot be executed immediately into immediately executable outbound sub-orders.

[0031] The outbound orders that are determined to be immediately executable and the outbound sub-orders that are immediately executable are aggregated to construct an executable outbound order set.

[0032] Specifically, rigid constraints are extracted and a multi-objective optimization function is constructed, including:

[0033] Based on a unified spatiotemporal data benchmark, rigid constraints are extracted, including supply rhythm constraints, operational resource constraints, and warehouse layout constraints.

[0034] The supply rhythm constraint is that for all outbound orders included in the same sorting wave, the available inventory of all material categories becomes effective earlier than the planned execution start time of the corresponding sorting wave.

[0035] The operational resource constraints are that the total picking volume, total number of material categories, and estimated operation time of a single sorting wave are less than the current operational resource's processing capacity limit.

[0036] The warehouse layout constraint is that the area coverage of the picking location within the same sorting wave meets the preset area limit.

[0037] Configure multi-objective optimization directions and construct multi-objective optimization functions; the multi-objective optimization directions include minimizing the total sorting operation journey, minimizing the average response time of outbound orders, and balancing the load rate of each operation resource.

[0038] Specifically, intelligent wave aggregation is performed on the set of executable outbound orders to generate a sorting wave plan, and the execution sequence of picking tasks within each sorting wave is reconstructed, including:

[0039] Based on inbound arrival event triggering and fixed-duration timed triggering, a dual-trigger scheduling mechanism for intelligent wave aggregation is constructed to start intelligent wave aggregation calculation;

[0040] Based on the current set of executable outbound orders, and combined with the material association feature map, the outbound orders are pre-grouped. Under the boundary constraints of supply rhythm, operation resource, and warehouse layout, the multi-objective optimization function is solved by a multi-objective optimization algorithm to obtain the optimal outbound order combination scheme and generate the initial sorting wave plan.

[0041] Based on rigid constraints, the initial sorting wave plan is subjected to constraint compliance verification, and a sorting wave plan is generated.

[0042] For each sorting wave generated, the execution sequence of picking tasks within each sorting wave is reconstructed by combining the warehouse space layout and material attribute characteristics, thus generating a sorting wave operation plan.

[0043] Specifically, dynamic path planning is performed, and picking tasks are assigned to the corresponding operational resources, including:

[0044] Retrieve the resource allocation information for the corresponding sorting wave, the picking task execution sequence for each resource, and the location coordinates for each picking task;

[0045] Simultaneously retrieve the access cost attribute and functional attribute label data of the rasterized warehouse space data;

[0046] Taking the current location of the operational resource as the starting point of the path, and the first picking task location in the picking task execution sequence as the first waypoint, all picking task locations are sequentially connected; based on the passage cost attribute of the gridded warehouse space data, the lowest cumulative passage cost path between each waypoint is solved, and the paths are spliced ​​together to form the picking travel path of a single operational resource.

[0047] Perform conflict prediction and verification on the picking and driving paths of all generated job resources to generate conflict-free global job path solutions;

[0048] The overall operation path plan and corresponding picking task information are simultaneously distributed to each operation resource.

[0049] An intelligent warehousing and logistics management system based on multi-source data fusion includes:

[0050] The benchmark construction module collects multi-source logistics data from the supply side, warehousing side, and demand side, performs standardized cleaning, granular decomposition, and time-series alignment processing, and constructs a unified spatiotemporal data benchmark.

[0051] The monitoring and prediction module, based on the unified spatiotemporal data benchmark, mines the supply-side co-transportation correlation characteristics and demand-side co-demand correlation characteristics between material categories, constructs a material correlation feature map, and is used to perform single material category inventory level monitoring and significant correlation material group completeness status monitoring, generate demand prediction results for each material category, and update the time series inventory data of the corresponding material category.

[0052] The verification management module calculates the dynamic available inventory for each material category, uses the outbound time point of the outbound order as the verification node, performs linked time-series verification on the outbound order, and constructs a set of executable outbound orders;

[0053] The aggregation and reconstruction module extracts rigid constraints based on a unified spatiotemporal data benchmark and constructs a multi-objective optimization function. It performs intelligent wave aggregation on the executable outbound order set, generates sorting wave plans, and reconstructs the execution sequence of picking tasks within each sorting wave.

[0054] The planning and control module performs dynamic path planning based on the sorting wave operation plan of the execution sequence and issues picking tasks to the corresponding operation resources.

[0055] The beneficial effects of this invention are:

[0056] This invention constructs a unified spatiotemporal data benchmark by integrating multi-source data from the supply, warehousing, and demand sides of the entire logistics chain. This provides consistent data support for end-to-end management, eliminating decision-making biases caused by differences in data caliber and time-series misalignment across multiple systems. Based on the material correlation feature map obtained through mining, it conducts demand forecasting with dual-layer inventory monitoring and linked correction, accurately grasping the supply and demand linkage patterns between materials, optimizing inventory structure and matching availability, and reducing the risk of inventory imbalance. Combined with time-series dynamic inventory calculation and outbound order linkage time-series verification, it can accurately match available inventory with outbound demand, effectively improving order fulfillment response speed and inventory utilization efficiency. Through intelligent wave aggregation and picking task sequence reconstruction under multiple constraints, it adapts to high-frequency changing outbound demand for small batches and multiple shipments, optimizing sorting batch combinations and work arrangements. With supporting full-domain dynamic path planning and traffic conflict avoidance mechanisms, it ensures efficient and orderly sorting operations, ultimately systematically improving the accuracy of warehousing inventory management, order fulfillment response speed, and overall operational efficiency of sorting operations. Attached Figure Description

[0057] Figure 1 This is a flowchart of the intelligent warehousing and logistics management method based on multi-source data fusion according to the present invention;

[0058] Figure 2 The flowchart for constructing a material association feature map for this invention;

[0059] Figure 3 A flowchart for constructing an executable outbound order set for this invention;

[0060] Figure 4 This is a flowchart illustrating the generation of a sorting wave operation scheme for this invention. Detailed Implementation

[0061] Example 1:

[0062] Please see Figure 1 This embodiment introduces an intelligent warehousing and logistics management method based on multi-source data fusion. Addressing industry pain points in intelligent warehousing scenarios, such as a wide variety of materials, small outbound batches for each category, frequent demand plan changes, and overlapping material supply and demand, this method uses multi-source logistics data fusion as a foundation to construct a unified spatiotemporal data benchmark with consistent data across the entire process. Combined with techniques such as material association feature mining, time-series inventory verification, multi-constraint intelligent wave scheduling, and dynamic path closed-loop management, this method systematically improves warehousing and sorting efficiency, inventory accuracy, and order response speed.

[0063] Intelligent warehousing and logistics management methods based on multi-source data fusion include:

[0064] Step S1: Collect multi-source logistics data from the supply side, warehousing side, and demand side, and perform standardized cleaning, granular decomposition, and time-series alignment processing to construct a unified spatiotemporal data benchmark with material category as the smallest management granularity and integrating time and spatial dimension attributes. The unified spatiotemporal data benchmark serves as the sole data carrier for full-process control, providing consistent data support for subsequent analysis, decision-making, and execution in each stage, and eliminating problems such as differences in data caliber, time-series misalignment, and granularity mismatch among multiple systems.

[0065] The supply side refers to the upstream business entities and corresponding logistics links that provide material inbound supply to the intelligent warehousing system, including, for example, material suppliers, transportation carriers, upstream collection nodes, and regional distribution centers. The warehousing side refers to the equipment, systems, and human resources within the intelligent warehousing system that perform material storage, sorting, handling, and control operations, including, for example, warehousing management systems, automated storage equipment, sorting and conveying equipment, automated guided vehicles, operators, and control terminals. The demand side refers to the downstream business entities and corresponding business links that submit material outbound requests to the intelligent warehousing system, including, for example, production lines, downstream distribution warehouses, end-customer order nodes, and after-sales spare parts service centers.

[0066] The multi-source logistics data refers to a collection of logistics business data from different links in the supply chain, with different data formats, time granularity, and spatial dimensions. For example, it includes supplier delivery plan data, in-transit transportation status data, mixed loading details data, estimated arrival time data, and material loading and unloading sequence data from the supply side; real-time inventory ledger data, gridded warehouse space data, operational resource status data, historical operation log data, and equipment operating parameter data from the warehousing side; and outbound order data, material demand planning data, order priority data, and outbound timeliness requirements data from the demand side.

[0067] The unified spatiotemporal data benchmark is a standardized data set formed after processing with a unified format, unified granularity, unified time sequence, and unified spatial coordinate system. It is the only data carrier that is jointly called and synchronously updated by all control links in the entire process, and can realize consistent query and call of inventory status, spatial location, and operation parameters of any material category at any point in time.

[0068] Furthermore, the steps for establishing a unified spatiotemporal data benchmark include:

[0069] Step S11: Receive multi-source logistics data from various business systems on the supply side, warehousing side, and demand side. Perform cleaning processing on the multi-source logistics data, including missing value completion, outlier removal, and duplicate data removal. Then, according to the preset data field specifications, perform field mapping and format unification processing on the multi-source logistics data with different formats and naming rules to generate a standardized logistics dataset.

[0070] The cleaning process includes the following execution rules: material category code, inbound / outbound quantity, and business occurrence time are defined as core fields, while other fields are defined as non-core fields; outlier removal uses a 3-standard-deviation judgment rule, where a single data entry's inbound / outbound quantity exceeds 3 times the standard deviation of the corresponding material category's historical inbound / outbound quantity for the same period, and is therefore judged as an outlier and removed; duplicate data removal is based on field completeness, calculating the proportion of non-empty fields in a single record to the total number of fields, and retaining the record with the highest proportion; missing value completion uses time series linear interpolation, numerically completing valid data with missing non-core fields, and directly marking missing core fields as invalid data and removing them. The preset data field specification is a pre-configured set of rules for unifying the naming rules and data storage formats of multi-source data fields. Its setting is based on the data management standards and cross-system data interaction requirements of intelligent warehousing operations. This specification uses the material category code as the core primary key, uniformly configuring the naming rules and data formats of time fields, spatial fields, and attribute fields to ensure the relevance and computability of subsequent cross-system and cross-process data. For example, the preset data field specifications can assign unique identification codes to all material categories using preset coding rules, uniformly convert all time fields to timestamp format accurate to the minute, and uniformly map all spatial fields to two-dimensional coordinate values ​​under the internal coordinate system of the warehouse. The preset coding rules are pre-configured segmented coding rules for unique identification of material categories. The coding structure is as follows: warehouse code segment (2 digits), material category code segment (3 digits), material sub-category code segment (3 digits), and check code segment (1 digit). Each coding segment uses Arabic numerals, and a check code is generated through a preset check algorithm to ensure the uniqueness and verifiability of the code, providing a standardized material identification for cross-system data interaction.

[0071] Step S12: For the multi-source logistics data corresponding to the mixed transportation of multiple categories of materials in the same batch from the supply side in the standardized logistics dataset, based on the preset data field specifications, the multi-source logistics data is decomposed into granularity according to the material category dimension, the warehousing operation feature parameters corresponding to each material category are extracted, and the standardized logistics dataset is updated; the time-series mapping of mixed data from batch granularity to material category granularity is completed, and the problem of inconsistent actual available inventory time for different material categories in the mixed arrival scenario is solved.

[0072] The multi-source logistics data corresponding to the mixed transportation of multiple categories of materials in the same batch is taken from the mixed loading details data, estimated arrival time data and material loading and unloading sequence data in the standardized logistics dataset; the warehousing operation characteristic parameters are a set of parameters characterizing the quantity scale, operation time and available time nodes of a single category of material in the entire warehousing operation process, for example, including warehousing quantity, loading and unloading operation time, shelving operation time, estimated shelving completion time and available inventory effective time.

[0073] For example, granular decomposition of multi-source logistics data includes: obtaining detailed mixed-load cargo data corresponding to a single batch of transportation units, determining all material categories included in the batch and the corresponding inbound quantity of each category; retrieving the estimated arrival time data and material loading and unloading sequence data corresponding to the batch, and, based on the preset material loading and unloading sequence rules, superimposing the corresponding loading and unloading operation time and shelving operation time for each category to calculate the estimated shelving completion time for each material category, and setting the estimated shelving completion time as the effective time point for the available inventory of the corresponding material category; the preset material loading and unloading sequence rules are set according to the material weight and volume from large to small and the operation priority from high to low, prioritizing the loading and unloading of heavy and large-volume materials; the loading and unloading operation time and shelving operation time are obtained based on historical operation log data statistics, calculated by multiplying the average operation time per unit quantity of the corresponding material category by the corresponding inbound quantity, with differentiated operation time parameters corresponding to different categories and different quantity scales. After the granularity decomposition is completed, the inbound operation feature parameters corresponding to each material category are synchronously updated and stored in the standardized logistics dataset, realizing the granularity of the full inbound data, and providing a fine-grained data foundation at the material category level for subsequent time-series inventory status calculation and order executability verification.

[0074] Step S13: Based on the standardized logistics dataset updated by granular decomposition in step S12, using the continuous time axis as the reference dimension and material category as the smallest management unit, perform time dimension alignment and spatial dimension association processing on the updated standardized logistics dataset to construct a unified spatiotemporal data benchmark that includes inventory level time series and spatial location attributes. The unified spatiotemporal data benchmark has real-time read / write and global synchronization capabilities. All control links in the entire process retrieve data from the same source from this benchmark and synchronously write back the processing results, ensuring the consistency of linkage between various control links from the data layer.

[0075] The time dimension alignment refers to dividing the continuous time axis into several continuous time periods according to a preset time granularity, mapping the real-time inventory data, expected inbound data, and allocated outbound data in the standardized logistics dataset to the corresponding time periods, forming a continuous time-series sequence of inventory levels for each time period. Real-time inventory data is included in the corresponding time period based on the end-of-period value of the time period in which the data was collected; expected inbound data is included in the corresponding time period based on the beginning-of-period value of the time period in which available inventory becomes effective; and allocated outbound data is included in the corresponding time period based on the end-of-period value of the time period in which the agreed outbound time period is located. The preset time granularity is a pre-set time segmentation scale, set based on the operational scheduling accuracy requirements and data update frequency of intelligent warehousing. For example, the preset time granularity can be set to 1 hour, that is, dividing time periods by hour and statistically analyzing the inventory and operational data for the corresponding time periods.

[0076] The spatial dimension association refers to mapping material categories in the standardized logistics dataset to their corresponding storage locations, work areas, and sorting nodes based on the warehouse's internal coordinate system. This mapping, combined with gridded warehouse space data, forms a unified spatial location attribute label across the entire area, enabling the spatial dimension association calculation of material categories, storage locations, and work paths. When a single material category corresponds to multiple storage locations, the center point of all storage locations is taken as the representative spatial coordinate of that material category. The gridded warehouse space data refers to a standardized spatial dataset that carries the spatial coordinates, accessibility, and functional attributes of each grid unit, after dividing the entire warehouse operation area into several standardized grid units of uniform size according to a preset grid size, based on the warehouse's internal coordinate system. This serves as the basic spatial data carrier for realizing spatial dimension association calculation and dynamic path planning. The grid division uses the lower left corner of the warehouse's internal coordinate system as the origin, with the horizontal axis pointing to the right and the vertical axis pointing upwards. The grid is uniformly divided row by row and column by column according to the preset grid size. Grids containing fixed obstacles or pillars within the warehouse are marked as impassable grids. The preset grid size is a pre-defined physical scale parameter of a single grid unit. Its setting is based on the overall scale of the storage operation area, the passage accuracy of the operation equipment, and the granularity requirements of the path planning. For example, it can be set as a square grid with a side length of 0.5 meters. In the gridded warehouse space data, each grid cell is configured with a corresponding spatial location attribute label, access cost attribute, and functional attribute label. The spatial location attribute label corresponds to the coordinate value of the grid cell in the internal coordinate system of the warehouse. The access cost attribute characterizes the accessibility of the grid cell, and the value range of the access cost attribute is 1 to 10. A straight standard passage grid is assigned a value of 1, a turning intersection grid is assigned a value of 2, a narrow passage grid is assigned a value of 5, an intersection grid is assigned a value of 3, and an impassable grid is assigned a value of infinity. The functional attribute label corresponds to the business function type of the grid cell. The functional attribute label is divided according to the boundary of the business function area corresponding to the grid. For example, it includes passage grids, storage location grids, and sorting gate grids. Among them, the grid cells located within the storage storage location range are marked as storage location grids, the grid cells located within the operation passage range are marked as passage grids, and the grid cells located within the sorting gate operation range are marked as sorting gate grids. The gridded warehouse space data can be used to map the location of the storage location and the location of the operation node corresponding to the material category to the whole-domain grid network, providing a unified spatial calculation benchmark for subsequent storage location association analysis, picking path planning and operation flow conflict avoidance.

[0077] The inventory level time series includes the real-time inventory of each material category at the current time, the expected inbound volume for each future time period, and the pre-allocated outbound volume for each future time period. The spatial location attributes include the coordinates of the storage location corresponding to each material category, the warehouse area it belongs to, and information on adjacent sorting nodes. Adjacent sorting nodes are determined using a path travel cost rule. When the cumulative travel cost from the storage location of a material category to the sorting node is lower than a preset node proximity threshold, the sorting node is determined to be an adjacent sorting node for the corresponding material category. The preset node proximity threshold is a preset upper limit for path travel costs, exemplarily set to 1.2 times the average travel cost within the warehouse. This threshold can be dynamically adjusted according to the regional division of the warehouse layout, the density of sorting nodes, and the requirements for travel efficiency, ensuring that the determination result matches the distance priority of the actual operating path.

[0078] The unified spatiotemporal data benchmark employs an optimistic locking concurrency control mechanism to ensure data consistency during concurrent read and write operations across multiple stages. All data update operations carry a version number identifier, and writing can only be performed when the version number of the update operation matches the current version number within the benchmark. Data synchronization adopts an event-triggered mechanism supplemented by timed polling. Changes in core business data trigger full-link synchronization immediately, while non-core data undergoes batch synchronization according to a preset synchronization cycle. The preset synchronization cycle can be configured to 15 minutes, 30 minutes, or 60 minutes, with a default setting of 30 minutes. The specific duration is adjusted based on the update frequency of non-core data in warehousing operations and system load, reducing the system resource consumption of frequent synchronization while ensuring data consistency.

[0079] Once this step is completed, the unified spatiotemporal data benchmark will serve as the sole data carrier for the entire process control. Subsequent inventory monitoring, order verification, wave scheduling, and route planning will all be calculated based on this benchmark, and the calculation results will be updated to this benchmark in real time. This will achieve consistency and dynamic linkage of data throughout the entire process, eliminating decision-making biases caused by differences in data calibers, time sequence misalignments, and granularity mismatches among multiple systems.

[0080] Specifically, step S1 uses a three-level progressive processing logic of standardized cleaning and unified format, granular decomposition and unified spatiotemporal alignment to converge multi-source logistics data with scattered sources, heterogeneous formats and uneven granularity into a unified spatiotemporal data benchmark with the same source. Among them, step S11 provides a unified field and format foundation for subsequent granular decomposition and spatiotemporal alignment. Step S12 completes the data granularity sinking for the typical scenario of mixed transportation on the supply side, making up for the deficiency that batch-level data cannot support precise time-series control. Step S13 completes the systematic integration of the full amount of data from the two dimensions of time and space, and finally forms a standardized data base that can be reused in all stages of the process. Throughout the overall operation of this method, the unified spatiotemporal data benchmark is not merely used as an initial input to unidirectionally transmit data to subsequent steps. Instead, it continuously receives dynamic data such as inventory changes, operational status, and execution results from each subsequent control link, enabling real-time iterative updates of the benchmark data. Subsequent steps such as inventory monitoring and demand forecasting in step S2, order feasibility verification in step S3, wave intelligent aggregation in step S4, and path planning and execution feedback in step S5 all use the unified spatiotemporal data benchmark as the sole data interaction carrier. This ensures logical linkage and status synchronization between each step from the data foundation, avoiding data deviations and decision-making disconnects caused by the parallel operation of multiple systems.

[0081] Step S2: Based on the unified spatiotemporal data benchmark, mine the supply-side co-transportation correlation characteristics and demand-side co-demand correlation characteristics among material categories, and construct a material correlation feature map with material categories as nodes and correlation relationships as edges; based on the material correlation feature map, perform two-layer dynamic inventory monitoring, including single material category inventory level monitoring and significantly related material group completeness status monitoring, and generate demand forecast results for each material category; synchronize the demand forecast results to the unified spatiotemporal data benchmark in real time, and update the time-series inventory data of the corresponding material categories.

[0082] The material association feature map is a data structure that stores the supply and demand linkage patterns among material categories in a graph structure, used to characterize the degree of correlation between different material categories in the inbound supply stage and the outbound demand stage. The co-transport association feature refers to the association attributes that different material categories appear together in the same batch of inbound transportation on the supply side, used to reflect the supply linkage pattern between material categories. The co-demand association feature refers to the association attributes that different material categories appear together in the same outbound order on the demand side, used to reflect the demand matching pattern between material categories. The dual-layer dynamic inventory monitoring is a two-level linkage inventory status monitoring mechanism that carries out monitoring from the individual dimension of a single material category and the group dimension of significantly related material groups, used to simultaneously identify the inventory gap risk of a single category and the supply risk of multiple categories.

[0083] Please see Figure 2 Furthermore, step S2 includes:

[0084] Step S21: Call the historical inbound and outbound data within the preset historical period in the unified spatiotemporal data benchmark, mine the correlation patterns between material categories from the supply and demand sides, construct a material correlation feature map and synchronize it to the unified spatiotemporal data benchmark; the preset historical period is a pre-set historical data time range used for statistical correlation features, and its setting is based on the fluctuation cycle of warehousing business and data sample size requirements. For example, it can be set to the operating data of the past 6 months.

[0085] The historical inbound data refers to the set of inbound business record data from the supply side stored in the unified spatiotemporal data benchmark within a preset historical period. It is used to reflect the historical inbound supply patterns and batch combination characteristics of each material category. For example, it includes historical inbound batch data, material category and inbound quantity data corresponding to each batch, mixed loading details data, arrival time data, and material loading and unloading sequence data.

[0086] The historical outbound data refers to the set of outbound business record data from the demand side stored in the unified spatiotemporal data benchmark within a preset historical period. It is used to reflect the historical outbound demand patterns and order combination characteristics of each material category. For example, it includes historical outbound order data, material category and outbound quantity data corresponding to each order, order priority data, and outbound timeliness requirement data.

[0087] Further, step S21 includes:

[0088] Step S211: Using the material category code as a unique identifier, clean and verify the historical inbound and outbound data, including deduplication, completion, and standardization verification, and remove invalid batches and abnormal order data to form a correlation analysis sample dataset; the field format of the correlation analysis sample dataset is consistent with the preset data field specification to ensure the consistency of the data caliber for subsequent correlation calculation.

[0089] The specific execution logic of the cleaning and verification process is as follows: Deduplication is performed by merging duplicate uploads of the same batch or order data, using the inbound batch code and outbound order code as unique identifiers. Field completeness is quantified by the proportion of non-empty fields in a single record to the total number of fields, retaining the record with the highest proportion. Completion is performed by using time-series interpolation to fill in missing values ​​for valid data with missing non-core fields, based on historical data of the same category and type. Data lacking core fields such as material category code, inbound / outbound quantity, and business time are directly marked as invalid data. Standardization verification is based on the preset data field specifications, verifying the field completeness, format compliance, and numerical rationality of the data, removing abnormal batches and orders with values ​​exceeding the preset reasonable range. The preset reasonable range is set using a 3-standard-deviation rule, i.e., the range of values ​​between the mean of historical inbound / outbound quantities for the corresponding material category and 3 standard deviations; data exceeding this range is considered abnormal data.

[0090] Step S212: Based on the aforementioned correlation analysis sample dataset, perform correlation statistics from both the supply and demand sides: For the supply side, count the number of times any two material categories appear together in the same inbound batch, and calculate the co-transport correlation of the corresponding material category pair; For the demand side, count the number of times any two material categories appear together in the same outbound order, and calculate the co-demand correlation of the corresponding material category pair.

[0091] The co-transport correlation degree is used to characterize the degree of linkage between two material categories in the warehousing and transportation process. Its basic calculation method uses the co-occurrence frequency method, which is the ratio of the number of co-occurring batches of the material category pair to the total number of warehousing batches. Further, a batch warehousing quantity weight can be introduced for weighted calculation to reflect the different contributions of batches of different sizes to the correlation degree, adapting to the need to differentiate the correlation degree between bulk materials and miscellaneous materials. The batch warehousing quantity weight is the ratio of the total warehousing quantity of the corresponding warehousing batch to the total warehousing quantity within the statistical period. The co-demand correlation degree is used to characterize the degree of matching between two material categories in the outbound demand process. Its basic calculation method uses the co-occurrence frequency method, which is the ratio of the number of co-occurring orders of the material category pair to the total number of outbound orders. Further, an order outbound quantity weight can be introduced for weighted calculation to reflect the impact of bulk orders on the material matching correlation, improving the business relevance of the correlation degree statistics. The order outbound quantity weight is the ratio of the total outbound quantity of the corresponding outbound order to the total outbound quantity within the statistical period.

[0092] Step S213: Compare the co-transportation correlation degree and co-demand correlation degree of each material category pair with the preset correlation degree threshold. Material category pairs with correlation degrees higher than the preset correlation degree threshold are determined to have strong correlation relationships and are classified into significantly related material groups. The strong correlation attributes of the supply side and the strong correlation attributes of the demand side are marked respectively. Using each material category as a graph node, the strong correlation relationship as a graph connection edge, and the corresponding co-transportation correlation degree or co-demand correlation degree as the weight parameter of the edge, a material correlation feature graph containing supply correlation sub-graph and demand correlation sub-graph is constructed.

[0093] The preset correlation threshold is a pre-set critical value used to determine whether there is a strong correlation between material categories. It is set based on the distribution characteristics of the co-occurrence frequency of materials obtained by statistical analysis of historical inbound and outbound data in the unified spatiotemporal data benchmark, as well as the intensity of supporting demand from downstream businesses. For example, the preset correlation threshold can be set to 0.7, that is, when the co-occurrence frequency of material category pairs is higher than 70%, they are determined to be significantly related material groups.

[0094] Furthermore, each node in the material association feature map carries a material category classification attribute label corresponding to the material category. After the material association feature map is constructed, it is divided into groups by clustering. Multiple closely connected material categories are grouped into the same significantly associated material group, forming a complete material association feature map containing pairwise relationships and group clustering results. The group clustering is implemented using a community detection algorithm, with material category as the node and association degree as the edge weight. When the average association degree between nodes within a group is higher than a preset group threshold, and the average association degree between nodes between groups is lower than the preset group threshold, they are determined to be the same significantly associated material group. The preset group threshold is pre-set based on business support requirements, and can be set to 0.6 for example.

[0095] Step S214: After converting the constructed material association feature map according to the preset data field specifications, it is synchronously stored in the unified spatiotemporal data benchmark as basic analysis data that can be called throughout the entire process; at the same time, a periodic update mechanism for the map is established, and the correlation degree is recalculated and the map structure is iteratively optimized according to the preset update cycle to ensure the matching of material association features with the actual business operation status.

[0096] The preset update cycle is a pre-set time interval for iterative updates of the map. It is set based on the frequency of category updates and the speed of change in demand structure of the warehousing business. For example, the preset update cycle can be set to update once a month.

[0097] Furthermore, the periodic update mechanism employs a combination of full and incremental updates: a full update is performed when the preset update cycle arrives, recalculating the correlation of all materials within the complete cycle; within the preset update cycle, an incremental update is performed whenever a preset number of inbound batches and outbound orders are added, iteratively calculating the correlation of material category pairs involved in the new data, thus improving the efficiency of the map update. The preset number is set based on the average daily volume of inbound and outbound documents in the warehousing business. For example, it can be set to trigger an incremental update when 50 new inbound batches or outbound orders are added; the incremental update uses a sliding window iteration method, adding the co-occurrence frequency corresponding to the new data to the original statistical results according to weights. The weight setting adopts a time decay coefficient model, assigning higher weights to more recent new data. For example, the weight of new data in the current cycle can be set to 1, and the weight of data in the previous cycle can be reduced by a decay coefficient of 0.8, synchronously updating the correlation values ​​of the corresponding material category pairs. A version management mechanism for the material association feature map is established in parallel, and each updated version of the material association feature map is assigned a version number and archived historically, supporting historical version retrospective and analysis of the trend of changes in association features.

[0098] Step S22: Based on the material association feature map and the real-time inventory data in the unified spatiotemporal data benchmark, perform dual-layer dynamic inventory monitoring, which includes monitoring the inventory level of single material categories and monitoring the completeness status of significantly related material groups. Generate early warning information and synchronize it to the unified spatiotemporal data benchmark to provide decision-making reference for risk dimensions for subsequent order verification and wave scheduling.

[0099] Further, step S22 includes:

[0100] Step S221: Retrieve real-time inventory data for each material category from the unified spatiotemporal data benchmark, and compare the real-time inventory level of each material category with the corresponding preset safety stock threshold one by one; when the real-time inventory level of a certain material category is lower than the corresponding preset safety stock threshold, generate an inventory shortage warning message for the corresponding material category and mark the degree of shortage.

[0101] The preset safety stock threshold is a pre-set minimum inventory level threshold used to ensure the basic supply capacity of materials. It is set based on the turnover cycle data and historical demand fluctuation data of the corresponding material category stored in the unified spatiotemporal data benchmark. For example, it can be set as the average outbound consumption of the corresponding material category over 3 days.

[0102] Furthermore, the single-material category inventory level monitoring adopts a monitoring mechanism combining event triggering and timed polling: whenever a business event occurs that causes inventory changes due to inbound shelving or outbound picking, an inventory level verification is triggered immediately; simultaneously, timed polling verification is performed according to a preset monitoring cycle to ensure the real-time and continuous nature of inventory status monitoring; event-triggered verification only performs level verification on the corresponding material category where inventory changes occur, while timed polling verification performs batch level verification on all material categories. The preset monitoring cycle is set based on the busyness of warehousing operations, and can be set to 15 minutes for example. A preset anti-shake duration is set during the early warning triggering process. When the inventory level is below the preset safety stock threshold for a continuous period exceeding the preset anti-shake duration, an inventory shortage early warning message is officially generated to avoid false early warnings caused by temporary inventory fluctuations; the timing start of the preset anti-shake duration is the point when the inventory level first falls below the preset safety stock threshold, and the early warning triggering condition is only determined to be met when the inventory level remains below the safety stock threshold for a continuous period from that point; if the inventory level rises above the safety stock threshold during the period, the timing start is reset. The preset stabilization duration can be set to 30 minutes, for example.

[0103] Step S222: Based on all significantly related material groups in the material association feature map, retrieve the real-time inventory data of all material categories within each significantly related material group from the unified spatiotemporal data benchmark, and calculate the inventory matching rate of each significantly related material group by combining the material matching ratio corresponding to each significantly related material group; compare the inventory matching rate of each group with the preset matching rate threshold, and when the inventory matching rate is lower than the preset matching rate threshold, generate the matching supply risk warning information of the corresponding significantly related material group, and mark the material category and degree of shortage.

[0104] The inventory completeness rate is a parameter representing the proportion of significantly related material groups that can meet the matching outbound demand. It is calculated as the ratio of the available quantity of bottleneck materials within the group to the theoretical total quantity. Bottleneck materials within the group are the material categories with the smallest available output quantity after conversion according to the standard matching ratio. The available quantity is the real-time available inventory of the corresponding material category, and the theoretical total quantity is the base quantity of a single material set when fully matched according to the standard matching ratio multiplied by the preset total number of matching sets. The preset total number of matching sets is set based on the historical peak matching demand during the same period. The preset completeness rate threshold is a pre-set minimum proportion threshold value used to determine whether a significantly related material group has the capability for matching supply. It is set based on the matching fulfillment requirements of downstream businesses and the historical outbound matching ratio data stored in the unified spatiotemporal data benchmark. For example, it can be set to 95%.

[0105] Furthermore, the material matching ratio is obtained based on historical outbound data statistics in the unified spatiotemporal data benchmark. Specifically, it involves statistically analyzing the quantity ratio of each material category within a significantly related material group in the same current order, and taking the ratio with the highest frequency as the standard matching ratio. The statistical range of the material matching ratio remains consistent with the preset historical period. In addition to monitoring the current real-time inventory matching status, the system also incorporates the inventory level time series in the unified spatiotemporal data benchmark to extrapolate and calculate the expected inventory matching rate within a preset future period. The preset period can be configured as 3 days, 7 days, or 15 days, with a default setting of 7 days. The specific duration can be adjusted according to the supply cycle of the material category, the order fulfillment cycle, and the forecast accuracy requirements of warehousing operations. For categories with longer supply cycles and slower material turnover, the forecast period can be extended to achieve proactive risk warning. The deduction logic for the expected inventory availability rate in the future period is as follows: combining the expected inbound and outbound volumes in the inventory level time series for each future period, calculate the expected available inventory of each material category for each period, and then calculate the expected availability rate for the corresponding period according to the availability rate calculation formula.

[0106] Step S223: Classify and process the inventory shortage warning for single material categories and the supply risk warning for significantly related material groups in a hierarchical manner. Based on the degree of shortage, classify them into different warning levels. After unifying the format of all warning information according to the preset data field specifications, write them synchronously into the unified spatiotemporal data benchmark. The updated warning data is synchronously used for priority adjustment in the subsequent order verification process and resource allocation in the wave scheduling process, so as to realize the linkage support between risk information and business decision-making.

[0107] The gap degree is a parameter used to quantify the deviation between the inventory gap size and demand fulfillment capacity of a material category or significantly related material group. It is calculated as the ratio of the difference between the current available inventory and the corresponding benchmark inventory to the average outbound volume of that material category per unit period. The benchmark inventory for a single material category is a preset safety stock threshold, and the benchmark inventory for a significantly related material group is the minimum matching inventory required to meet the standard matching ratio. For example, the gap degree can be divided into three levels: mild gap, moderate gap, and severe gap. A mild gap corresponds to a gap less than the average outbound volume of that material category in one day; a moderate gap corresponds to a gap between the average outbound volume of that material category in one to three days; and a severe gap corresponds to a gap greater than the average outbound volume of that material category in three days. The average outbound volume per unit period is derived from the historical outbound statistics of the corresponding material category stored in the unified spatiotemporal data benchmark. The warning levels correspond one-to-one with the degree of the shortage. For example, a slight shortage corresponds to a Level 1 warning, a moderate shortage corresponds to a Level 2 warning, and a severe shortage corresponds to a Level 3 warning. The warning information is uniformly included in the preset data field specifications, including the warning type, associated material category, degree of shortage, warning level, and trigger time field.

[0108] Step S23: Based on the historical outbound data in the unified spatiotemporal data benchmark, perform basic demand forecasting for a single material category, and correct the deviation of the forecasting results by combining the demand linkage pattern in the material association feature map. Derive the multi-period demand forecasting results for each material category, and synchronize the demand forecasting results to the unified spatiotemporal data benchmark in real time to update the time-series inventory data of the corresponding material category.

[0109] Furthermore, step S23 includes:

[0110] Step S231: Retrieve historical outbound data within the preset forecast period from the unified spatiotemporal data benchmark, and use a time-series forecasting algorithm to fit and calculate the future demand trend of each material category to generate basic demand forecast results for each material category; the basic demand forecast results are output in the preset time granularity, showing the expected outbound demand for the corresponding time period.

[0111] The preset forecast period is a future time range covered by the pre-set demand forecast. It is set based on the replenishment cycle and scheduling planning requirements of the warehousing business. For example, it can be set to the next 7 days. The time series forecasting algorithm can be, but is not limited to, time convolutional networks, differential integrated moving average autoregressive models, and other time series forecasting algorithms.

[0112] Specifically, the steps for fitting and calculating the future demand trends of each material category to generate basic demand forecasts for each material category include:

[0113] Step S2311: Extract historical outbound data of the corresponding material category within a preset historical statistical period from the unified spatiotemporal data benchmark, and simultaneously extract the time feature identifiers of the corresponding time periods to construct a basic prediction input feature set. The time feature identifiers include weekday identifiers, holiday identifiers, monthly / seasonal identifiers, and weekly / cycle identifiers, all of which are stored in the unified spatiotemporal data benchmark. The preset historical statistical period is set in accordance with the preset historical period, and can be exemplarily set to the past 6 months. The basic prediction input feature set uses a preset time granularity to match the historical outbound volume at each moment with the time feature identifier of the corresponding time period, forming a time-series feature sample sequence.

[0114] Step S2312: Perform stationarization preprocessing on historical outbound data, eliminate long-term trend terms and seasonal fluctuation terms in time series data through difference operation to obtain stationary time series data; at the same time, perform one-hot encoding on categorical features in the basic prediction input feature set, and normalize numerical features to generate a standardized feature dataset.

[0115] Step S2313: Input the stationary time-series data and standardized feature dataset into the time-series prediction algorithm. Through model training, a demand trend fitting model for the corresponding material category is obtained. Based on the demand trend fitting model, the initial demand prediction result within a preset prediction period is output. The hyperparameters of the time-series prediction algorithm are determined using a grid optimization method based on historical outbound data in the unified spatiotemporal data benchmark, ensuring that the model adapts to the demand fluctuation characteristics of different material categories. The search range of the grid optimization method is preset based on the historical data volume of the corresponding material category. The optimization hyperparameters include the model input window length, convolution kernel size, and number of model layers, with the minimum mean absolute error of the prediction result as the optimization objective.

[0116] Step S2314: Retrieve historical forecast residual data for the corresponding material category from the unified spatiotemporal data benchmark, statistically analyze the distribution pattern of historical forecast deviations, calculate the residual correction coefficient, and correct the initial demand forecast result based on the residual correction coefficient to obtain the basic demand forecast result for each material category; the basic demand forecast result is output in units of a preset time granularity, corresponding to the expected outbound demand for each time period. The residual correction coefficient is calculated based on the mean of historical forecast residuals, and the correction logic is: corrected forecast quantity = initial forecast quantity - mean of historical residuals; where the historical residual is the difference between the forecast value and the actual outbound value for the same historical period, and the statistical range is consistent with the preset historical statistical period.

[0117] Step S232: Based on the demand association sub-map in the material association feature map, identify the core driving material and supporting follow-up material of each significantly associated material group, calculate the demand linkage coefficient of each material in the significantly associated material group, and correct the deviation of the basic demand forecast result based on the basic demand forecast result of the core driving material to obtain the final demand forecast result of each material category.

[0118] The demand linkage coefficient is a parameter that characterizes the degree of linkage between demand changes of various material categories within a significantly related material group. It is calculated based on historical outbound data in the unified spatiotemporal data benchmark and is used to correct the deviation caused by independent forecasting of a single product category and improve the forecasting accuracy of demand for multiple product categories.

[0119] Further, step S232 includes:

[0120] Step S2321: Based on the demand association sub-graph in the material association feature graph, traverse all significantly associated material groups; for all material categories within each group, retrieve the historical outbound data of the corresponding material category from the unified spatiotemporal data benchmark, and identify the core driving material and supporting follower material within the group through Granger causality test; the core driving material is the material category whose demand changes precede other materials in the group and plays a dominant role in the overall demand of the group, and the supporting follower material is the material category whose demand changes synchronously with the core driving material. The significance level of the Granger causality test is set to 0.05, and the lag order is automatically determined based on the information criterion; when the test result rejects the null hypothesis, it is determined that the demand change of the corresponding material category is a Granger cause of the demand change of another material category, that is, the former is the core driving material, and the latter is the supporting follower material.

[0121] Step S2322: For each significantly related material group, calculate the Pearson correlation coefficient of the demand time series data between the core driving material and each supporting follower material in the group, and set the Pearson correlation coefficient as the demand linkage coefficient of the corresponding material pair; the value range of the demand linkage coefficient is 0 to 1, and the larger the value, the higher the degree of demand linkage between the two types of materials.

[0122] Step S2323: Based on the basic demand forecast results of the core driving material, calculate the demand deviation rate of the core driving material. The demand deviation rate is the ratio of the deviation between the basic demand forecast of the core driving material and the historical average outbound quantity for the same period. Combined with the demand linkage coefficient of the corresponding supporting materials, correct the deviation of the basic demand forecast results of the supporting materials. The correction logic is: corrected expected outbound quantity = basic demand forecast quantity × (1 + demand linkage coefficient × core driving material demand deviation rate). After completing the correction calculation for all significantly related material groups, obtain the final demand forecast results for each material category. The historical average outbound quantity for the same period is the average outbound quantity for the corresponding period within the preset historical statistical period. When the demand deviation rate of the core driving material is positive, the forecast value of the supporting materials is increased accordingly; when the demand deviation rate is negative, the forecast value of the supporting materials is decreased accordingly. The correction result must be greater than or equal to 0.

[0123] Step S233: After standardizing the demand forecast results according to the preset data field specifications, write them into the unified spatiotemporal data benchmark in real time to update the inventory level time series of the corresponding material category; the updated time series inventory data is synchronously used for subsequent order executability verification and wave intelligent aggregation, so as to realize the linkage support between forecast data and business decision-making.

[0124] Furthermore, before writing the demand forecast results back to the unified spatiotemporal data benchmark, data format verification and temporal continuity verification are performed to ensure that the written-back data conforms to the preset data field specifications and that the time series has no breaks or repetitions. The temporal continuity verification rules are as follows: verify whether the number of time periods in the forecast results is consistent with the total number of time periods within the preset forecast period, and whether the time nodes of adjacent time periods are continuous without overlap or breaks; the write-back operation can only be performed after the verification passes. After the write-back is completed, the recalculation of the future time period matching rate in step S22 and the batch update of the order executability verification in step S3 are automatically triggered, realizing automatic linkage of downstream decision-making links after the forecast data is updated, and ensuring real-time consistency of data and decisions throughout the entire process.

[0125] Specifically, step S2, through a progressive analytical logic of correlation pattern mining, dual-layer status monitoring, and trend prediction correction, deeply mines the supply and demand linkage characteristics between material categories based on the unified spatiotemporal data benchmark constructed in step S1, breaking through the limitations of traditional independent management of single materials. Step S21 extracts material correlation patterns from historical data and consolidates them into the unified spatiotemporal data benchmark, providing analytical basis for the entire process based on correlation dimensions. Step S22 achieves refined inventory risk identification by combining individual and group data. Step S23 corrects prediction accuracy through correlation features, outputting inventory trend data that better aligns with business linkage patterns. All analytical processes in this step are based on the same source data of the unified spatiotemporal data benchmark, and the analysis results are written back to update the benchmark data in real time. This provides more accurate time-series inventory support for order executability verification in step S3 and provides optimization basis for wave aggregation scheduling in step S4, achieving two-way linkage between data and decision-making.

[0126] Step S3: Based on the unified spatiotemporal data benchmark, calculate the dynamic available inventory of each material category in each time period; for outbound orders containing combinations of multiple material categories, use the outbound time point agreed in the outbound order as the verification node, perform linked time-series verification on all material categories in the outbound order, and construct a set of executable outbound orders that are updated in real time; synchronize the inventory pre-occupancy data generated during the outbound order verification process to the unified spatiotemporal data benchmark in real time, and update the dynamic available inventory data for each time period accordingly.

[0127] The dynamic available inventory refers to the total effective inventory of the corresponding material category that can be allocated for outbound delivery within a specified period, representing the actual capacity of the material category to support outbound demand during that period. The executable outbound order set is the set of all outbound orders that meet the outbound inventory conditions and can enter the sorting operation process, serving as the input basis for subsequent intelligent aggregation. The outbound order is a material outbound demand certificate submitted by the demand side to the intelligent warehousing system, which, for example, includes production material requisition orders, finished product distribution orders, after-sales spare parts transfer orders, and cross-warehouse transfer orders.

[0128] Inventory pre-holding is a pre-locking operation of available inventory for the corresponding time period after the order verification is passed. It is used to ensure the uniqueness of the inventory resources of the allocated orders and to avoid multiple orders repeatedly occupying the same inventory resources.

[0129] Please see Figure 3 Furthermore, step S3 includes:

[0130] Step S31: Combining the final demand forecast results with the current real-time inventory data, the expected inbound data for each future period, and the pre-allocated outbound data in the unified spatiotemporal data benchmark, calculate the dynamic available inventory of each material category for each period according to the preset time granularity to form a time-segmented dynamic available inventory sequence.

[0131] Further, step S31 includes:

[0132] Step S311: Retrieve the current real-time inventory data, the original expected inbound data for each future time period, and the pre-allocated outbound data for each material category from the unified spatiotemporal data benchmark. Simultaneously, retrieve the expected outbound volume data for each time period from the final demand forecast result generated in Step S2. Using the material category code as a unique index, complete the time-series alignment of all data according to a preset time granularity to form a unified time-series calculation basis dataset. The time-series alignment process follows the mapping rules consistent with the time dimension alignment in Step S1, matching data from different sources to the corresponding time period according to material category, ensuring that the same time period contains complete data dimensions of beginning inventory, expected inbound, expected outbound, and pre-allocated outbound.

[0133] Step S312: For the original expected inbound data for each time period, match and confirm each time period according to the material category dimension and the preset time granularity to obtain the current expected inbound quantity of the corresponding material category for each time period; at the same time, deduct the safety stock buffer of the corresponding material category from the calculation base dataset, and lock this part of the inventory as unallocable inventory to ensure the basic supply capacity of each material category. The safety stock buffer is a reserved inventory of material categories that is locked in advance and does not participate in the allocation of outbound order inventory in response to uncertainties such as demand fluctuations and supply time sequence deviations of material categories. It is used to ensure the basic supply capacity of each material category and avoid the risk of supply gaps caused by full allocation of inventory; its setting is based on the turnover cycle data and historical demand fluctuation data of the corresponding material category stored in the unified spatiotemporal data benchmark, and its value is consistent with the preset safety stock threshold of the corresponding material category in step S22. For example, it can be set as the average outbound consumption of the corresponding material category over 3 days.

[0134] Step S313: Calculate the dynamic available inventory using a rolling recursive method for each time period. Use the current real-time inventory as the initial inventory for the initial time period, and calculate the ending inventory and dynamic available inventory for each time period in sequence. The formula for calculating the dynamic available inventory for a single time period is: Dynamic available inventory = Initial inventory for the current time period + Expected inbound quantity for the current period - Expected outbound quantity for the current period - Pre-allocated outbound quantity for the current period - Safety stock buffer. After the calculation is completed, a time-segmented dynamic available inventory sequence covering the entire forecast period is formed. The calculation results are then processed in a unified format according to the preset data field specifications and synchronously written into the unified spatiotemporal data benchmark in real time to provide a consistent inventory basis for subsequent outbound order verification.

[0135] Step S32: For each outbound order containing a combination of multiple material categories, take the outbound time point agreed in the order as the verification node, combine it with the dynamic available inventory sequence of the corresponding time period, perform a full material category linkage time sequence verification, determine the executable status of the order and output the corresponding processing result.

[0136] Furthermore, step S32 includes:

[0137] Step S321: Retrieve the detailed outbound order data of all outbound orders to be verified from the unified spatiotemporal data benchmark, extract all material categories and corresponding demand quantities, agreed outbound time points, and outbound order priority level information within the outbound orders; sort all outbound orders to be verified in descending order of outbound order priority level to determine the verification execution order and ensure that high-priority outbound orders occupy inventory resources first; the outbound order priority level is pre-divided based on the urgency of downstream business, for example, including three levels: urgent, normal, and deferred.

[0138] Step S322: For a single outbound order, using the agreed outbound time as the verification node, retrieve the dynamic available inventory values ​​corresponding to all material categories within the outbound order at that time, and compare the demand for each material category with the corresponding dynamic available inventory. If the dynamic available inventory of all material categories within the outbound order is greater than or equal to the corresponding demand, the outbound order is determined to be an immediately executable outbound order. If the dynamic available inventory of some material categories is insufficient, then, by combining the dynamic available inventory sequence by time period, we can deduce the earliest time point when all material categories meet the demand, and output the earliest executable time point of the outbound order. Deduction by time period refers to starting from the agreed outbound time point and verifying the dynamic available inventory of each material category along the time axis until the first time period in which the dynamic available inventory of all material categories is greater than or equal to the corresponding demand is found. The starting time point of this time period is the earliest executable time point of the outbound order.

[0139] The verification process synchronously retrieves early warning information from the unified spatiotemporal data benchmark. For urgent outbound orders corresponding to material categories with severe inventory shortage warnings, the verification priority is automatically increased and available inventory is matched first to ensure the fulfillment of core production needs. Automatically increasing the verification priority means prioritizing eligible urgent outbound orders in the same priority verification queue, and performing inventory matching and pre-occupancy operations first.

[0140] Step S323: For outbound orders that cannot be executed immediately, perform splitting processing according to the preset outbound order splitting rules: split the material categories in the outbound order that meet the current inventory conditions into immediately executable outbound sub-orders, and split the material categories with insufficient inventory into pending outbound sub-orders; the pending outbound sub-orders are marked with the earliest executable time point and added to the pending outbound order queue, waiting for subsequent inventory updates to automatically trigger re-verification, thereby achieving refined allocation of order resources and maximizing inventory utilization efficiency. The preset outbound order splitting rules are pre-set based on the fulfillment requirements of downstream businesses, and the splitting function can be enabled or disabled according to business scenarios to adapt to different outbound order management needs. The preset rules for splitting outbound orders clearly define the splitting boundaries: a splitting operation can only be performed when some material categories within an outbound order meet the current inventory conditions and others do not; both outbound sub-orders that can be executed immediately after splitting and outbound sub-orders that are yet to be executed inherit the priority level and timeliness attributes of the original outbound order; the queue of outbound orders to be executed is sorted in order from earliest to latest execution time and from highest to lowest priority level.

[0141] Step S33: Summarize all outbound orders and immediately executable outbound sub-orders to construct a dynamic set of executable outbound orders; simultaneously complete the inventory pre-occupancy lock for the corresponding outbound orders, write the inventory pre-occupancy data back to the unified spatiotemporal data benchmark in real time, and establish a rolling update mechanism to ensure the real-time effectiveness of the set of executable outbound orders.

[0142] Furthermore, step S33 includes:

[0143] Step S331: Sort all verified immediately executable outbound orders and immediately executable outbound sub-orders in ascending order according to outbound order priority level and agreed outbound time, and construct an executable outbound order set; synchronously label the material details, required quantity, agreed outbound time, priority and corresponding storage location information of each outbound order in the set, so as to provide complete outbound order dimension input data for subsequent intelligent aggregation.

[0144] Step S332: Perform inventory pre-occupancy operation on all outbound orders within the executable outbound order set. Based on the outbound time corresponding to each outbound order, deduct the dynamic available inventory of each material category for the corresponding time period to generate inventory pre-occupancy data. After standardizing the format of the inventory pre-occupancy data according to the preset data field specifications, write it synchronously to the unified spatiotemporal data benchmark in real time, and synchronously update the dynamic available inventory sequence by time period. This ensures the accuracy and uniqueness of inventory data for subsequent outbound order verification and wave scheduling, avoiding multiple outbound orders repeatedly occupying the same inventory resource. Inventory pre-occupancy is strictly executed according to the verification order of outbound orders, with high-priority outbound orders completing inventory pre-occupancy locking first. The release trigger conditions for pre-occupied inventory include three categories: outbound order execution completion, outbound order cancellation, and outbound order re-verification. After inventory release, the dynamic available inventory for the corresponding time period is synchronously replenished.

[0145] Step S333: Establish a rolling update trigger mechanism for the executable outbound order set. When four types of trigger events occur—new inbound materials completed and put away, new outbound orders submitted, existing outbound orders completed, and inventory forecast results updated—the entire set of pending outbound orders is automatically re-verified, and the executable outbound order set is updated. During the update process, the pre-occupied inventory corresponding to completed outbound orders is released and reallocated to the pending outbound order queue, realizing the dynamic flow and efficient utilization of inventory resources. The execution flow of the rolling update is as follows: After the trigger event occurs, the pre-occupied inventory corresponding to all outbound orders that have not yet entered the sorting execution stage is released first. Then, all pending outbound orders and pending outbound sub-orders are reordered according to priority rules, and the linked timing verification and inventory pre-occupancy are executed sequentially to generate the updated executable outbound order set. Among them, the events of new inbound materials completed and put away and inventory forecast results updated trigger the re-verification of all orders. The events of new outbound order submission and existing outbound order completion only trigger incremental verification, matching only the newly added orders with the pending orders corresponding to the released inventory.

[0146] Specifically, step S3 uses multi-dimensional inventory data from a unified spatiotemporal data benchmark as its core basis. Through a closed-loop processing logic of time-series inventory capacity quantification, full-material linkage verification, and dynamic set rolling updates, it upgrades traditional static single-point inventory verification to time-series, linkage-based dynamic verification. Step S31 utilizes the benchmark data from step S1 and the prediction results from step S2 to achieve refined time-series quantification of available inventory. Step S32 combines material combination characteristics and outbound order priorities to achieve accurate fulfillment determination. Step S33 ensures the uniqueness and real-time nature of inventory allocation through pre-occupancy locking and rolling update mechanisms. All verification and calculation processes in this step are based on the same source data from the unified spatiotemporal data benchmark. Verification results and pre-occupancy data are written back to update the benchmark data in real time. This not only leverages the predictive capabilities of upstream inventory control to achieve proactive decision-making but also provides accurate and reliable outbound order inputs for the subsequent wave intelligent aggregation in step S4, achieving seamless linkage between inventory control and sorting scheduling.

[0147] Step S4: Extract three types of rigid constraints—supply rhythm constraints, operational resource constraints, and warehouse layout constraints—from the unified spatiotemporal data benchmark. With the multi-objective optimization direction of improving overall sorting efficiency, shortening the average response time of outbound orders, and balancing operational resource load, an arrival-triggered rolling optimization mechanism is used to perform intelligent wave aggregation on the executable outbound order set, generating a sorting wave plan. The execution sequence of picking tasks within each sorting wave is reconstructed to generate a sorting wave operation plan. The operational resource pre-occupancy data corresponding to the sorting wave plan is synchronized to the unified spatiotemporal data benchmark in real time, and the core parameter rules of wave aggregation are dynamically iteratively optimized based on historical operational log data in the unified spatiotemporal data benchmark.

[0148] The sorting wave refers to a work unit that aggregates multiple outbound orders into the same work batch and organizes picking operations in a unified manner. This is used to improve the overall sorting efficiency and reduce repetitive work trips through batch operation mode. The arrival-triggered rolling optimization mechanism is a wave aggregation scheduling mechanism with the completion of material warehousing and shelving as the core trigger condition. It is different from the traditional fixed time window aggregation mode and is used to shorten the response time of outbound orders after the materials arrive in the warehouse, thereby improving the demand response speed. The operation resources are a collective term for the equipment and human resources that perform sorting and handling operations within the intelligent warehousing system. For example, they include automated guided vehicles, manual picking personnel, sorting robots, sorting stations, and conveyor line equipment.

[0149] Please see Figure 4 Furthermore, step S4 includes:

[0150] Step S41: Extract multi-dimensional rigid constraints and construct a multi-objective optimization function to clarify the constraint boundaries and optimization direction of wavelet aggregation, and provide calculation rules and judgment criteria for intelligent wavelet aggregation.

[0151] Further, step S41 includes:

[0152] Step S411: Extract three types of rigid constraints from the unified spatiotemporal data benchmark: The first type is supply rhythm constraint, which means that the available inventory of all required material categories for all outbound orders included in the same sorting wave must be effective earlier than the planned start time of the sorting wave to avoid waiting for materials after the wave starts; the second type is operational resource constraint, which means that the total picking volume, total number of material categories, and estimated operation time of a single sorting wave are less than the upper limit of the processing capacity of the currently available operational resources; the third type is warehouse layout constraint, which means that the regional coverage of picking locations within the same sorting wave must meet the preset regional restrictions, and outbound orders of adjacent locations should be aggregated first to reduce long-distance movement across areas. The preset regional restrictions are determined by a regional cost threshold. When the cumulative passage cost between the two farthest picking locations within a sorting wave is lower than the preset regional cost threshold, it is determined that the warehouse layout constraint is met; the preset regional cost threshold is pre-set based on the overall warehouse scale and operational efficiency requirements. The estimated operation time for resource constraints is estimated based on processing capacity parameters. The estimated operation time for a single sorting wave = total picking volume in a single wave ÷ processing volume per unit time of the corresponding resource + total estimated travel time in a single wave, where the total estimated travel time is calculated based on the cumulative passage cost between storage locations and the average travel speed of the resource.

[0153] Step S412: Configure multi-objective optimization directions and construct multi-objective optimization functions; the multi-objective optimization directions include three categories: minimizing the total sorting operation journey, minimizing the average response time of outbound orders, and balancing the load rate of each operation resource. The multi-objective optimization function is constructed by weighted summation; the weight coefficients corresponding to each optimization direction are pre-set based on the core operational objectives of warehousing business. For example, when improving outbound response speed is the core objective, the weight coefficient corresponding to the average response time of outbound orders can be increased.

[0154] Before constructing the multi-objective optimization function, the values ​​of the three types of optimization objectives are normalized, mapping each objective value to the interval between 0 and 1. Normalization uses the extreme value method, i.e., the normalized value of a single objective = (current solution objective value - worst-case objective value) ÷ (optimal objective value - worst-case objective value). The comprehensive score of the multi-objective optimization function = normalized value of total sorting operation travel × travel weight + normalized value of average response time for outbound orders × timeliness weight + normalized value of standard deviation of operational resource load rate × load weight. A lower comprehensive score indicates a better solution; the standard deviation of operational resource load rate is used to quantify the degree of load balance, with a smaller standard deviation indicating a more balanced load.

[0155] Minimizing the total sorting operation journey is the core efficiency optimization objective of wave aggregation. It refers to aggregating outbound orders with adjacent storage locations and high overlap in operation paths into the same sorting wave, thereby reducing the total movement distance of operation resources throughout the warehouse, reducing ineffective handling time and equipment energy consumption, and improving the material picking and processing efficiency per unit time. The quantitative calculation of this objective is based on the rasterized warehouse space data and storage location information in the unified spatiotemporal data benchmark, and the total operation journey of a single wave is obtained by accumulating the path length.

[0156] Minimizing the average response time of outbound orders is the timeliness optimization goal of wave aggregation. It refers to reducing the average time from when an outbound order is included in the set of executable outbound orders to when the sorting operation is completed. By promptly initiating aggregation through an arrival trigger mechanism and prioritizing the inclusion of high-priority outbound orders in the wave, the order waiting time and overall fulfillment cycle are shortened, thereby improving the outbound response speed on the demand side.

[0157] The load rate of equalizing each task resource is the resource optimization target of wave aggregation. It means to reasonably allocate the total sorting task according to the rated processing capacity of each task resource, so as to avoid the polarization of some task resources being overloaded and some task resources being idle, improve the overall utilization efficiency of task resources, and ensure the smooth and continuous operation of sorting operations. The load rate is quantitatively calculated by the ratio of the expected operation time of a single task resource to its rated available time.

[0158] Step S42: Under the constraints of three types of rigid constraints—supply rhythm constraints, operational resource constraints, and warehouse layout constraints—intelligent wave aggregation calculation is performed on the executable outbound order set to generate a sorting wave plan that meets the requirements of multi-objective optimization.

[0159] Further, step S42 includes:

[0160] Step S421: Construct a dual-trigger scheduling mechanism for intelligent wave aggregation. This mechanism uses inbound arrival events as the core triggering method and fixed-duration timed triggering as a supplementary fallback triggering method to jointly ensure the responsiveness and operational continuity of wave aggregation. When a new batch of materials is detected to have completed shelving operations and the dynamic available inventory data of the corresponding material category within the unified spatiotemporal data benchmark is synchronized and updated, an intelligent wave aggregation calculation is automatically initiated. The scope of arrival-triggered aggregation includes all unassigned wave executable outbound orders, with priority given to outbound orders corresponding to newly arrived material categories. A preset maximum aggregation interval is configured. When the continuous duration reaches the preset maximum aggregation interval and no new inbound arrival triggering event is generated, a timed fallback aggregation calculation is automatically initiated to prevent executable outbound orders from accumulating for extended periods. The preset maximum aggregation interval is the maximum allowable time interval between two adjacent wave aggregation operations, set based on the outbound timeliness requirements and operational busyness of the warehousing business. For example, the preset maximum aggregation interval can be set to 2 hours.

[0161] Step S422: Based on the latest set of executable outbound orders at the current moment, and combined with the demand relationships in the material association feature map, the outbound orders are pre-grouped. Under the boundary constraints of supply rhythm, operational resource, and warehouse layout, a multi-objective optimization algorithm is used to solve the multi-objective optimization function to obtain the optimal outbound order combination scheme and generate the initial sorting wave plan. The aggregation calculation process follows the outbound order priority from high to low, prioritizing high-priority outbound orders to be included in the sorting wave range to ensure the timeliness of core outbound orders. The multi-objective optimization algorithm can be, but is not limited to, genetic algorithms and particle swarm optimization algorithms. The execution rule of the pre-grouping process is: traverse all outbound orders to be aggregated; if two outbound orders contain material categories within the same significantly related material group, they are assigned to the same candidate group; the same outbound order can belong to multiple candidate groups, and subsequent optimization algorithms combine and solve based on candidate groups to improve the aggregation degree of materials in the same group.

[0162] Furthermore, the steps for generating the initial sorting wave plan include:

[0163] Step S4221: Retrieve the list of available job resources and corresponding processing capacity parameters for the current moment from the unified spatiotemporal data benchmark. Simultaneously, retrieve all outbound orders from the currently executable outbound order set to determine the outbound order pool to be aggregated and the boundaries of allocable job resources, serving as the input basis for wave aggregation calculation. The list of available job resources is a detailed set of job resources that are idle and available for sorting tasks in the current time period, statistically obtained based on the job resource status data in the unified spatiotemporal data benchmark. The list of available job resources includes the unique identifier, resource type, current location, current status, and continuous operation duration of the corresponding job resource, used to clarify the range of schedulable job resources during wave aggregation and ensure the executability of task allocation.

[0164] The processing capacity parameters are a set of quantitative parameters characterizing the sorting scale that each operational resource can handle per unit time. Different types of operational resources correspond to different dimensions of processing capacity parameters. For example, the processing capacity parameters corresponding to automated guided vehicles include the number of materials handled per unit time, the maximum weight carried in a single trip, and the average driving speed. The processing capacity parameters corresponding to manual pickers include the number of material categories picked per unit time and the number of outbound orders that can be processed per hour. The processing capacity parameters are statistically obtained based on historical operation log data in the unified spatiotemporal data benchmark and are used to define the upper limit of the processing capacity of operational resources, providing a quantitative basis for determining operational resource constraints.

[0165] Step S4222: Sort the outbound orders in the aggregated outbound order pool according to their priority from high to low. Combine the demand association relationship in the material association feature map, pre-group the outbound orders that contain the same significantly related material group to improve the aggregation degree of the same group of materials and provide an initial grouping scheme for subsequent optimization solutions.

[0166] Step S4223: Input the pre-grouped outbound orders, rigid constraints, and multi-objective optimization functions into the multi-objective optimization algorithm, and obtain multiple candidate wave combination schemes through iterative optimization calculation; during the iteration process, the optimal comprehensive score of the multi-objective optimization function is used as the convergence condition, and the candidate wave combination scheme with the best comprehensive performance is selected as the initial sorting wave plan.

[0167] Step S4224: Match corresponding operational resources for each sorting wave in the initial sorting wave plan, marking the planned execution start time, outbound order details, estimated total operation time, and corresponding operational resource information for each sorting wave, forming a structured initial sorting wave plan, which is then output to the subsequent constraint compliance verification stage. Operational resource matching follows the principles of type matching and load balancing. First, it matches operational resources of the corresponding type according to the operational type of the sorting wave. Then, based on the ratio of the allocated task volume to the rated processing capacity of each operational resource, the wave tasks are allocated to the operational resource with the lowest current load rate. The estimated total operation time is calculated based on the total picking volume, total location access cost, and the processing capacity parameters of the corresponding operational resource for that sorting wave.

[0168] Step S423: Perform constraint compliance verification on the initial sorting wave plan, verifying whether each sorting wave simultaneously meets all three types of rigid constraint conditions; for sorting waves that do not meet the constraint conditions, adjust the outbound order combination, remove outbound orders that exceed the constraint boundary and add them to the next batch aggregation range, and finally generate a compliant and valid sorting wave plan. The operation scale of a single sorting wave is limited by a preset maximum operation volume threshold for a single wave. The preset maximum operation volume threshold for a single wave is a pre-set critical value for the maximum operation scale that a single sorting wave can carry. Its setting is based on the rated processing capacity of the operation resources and the historical operation efficiency data in the unified spatiotemporal data benchmark. For example, the preset maximum operation volume threshold for a single wave can be set to the total picking volume corresponding to 20 outbound orders. Order rejection follows a priority-to-high rule, first rejecting the lowest priority outbound orders. If the constraints are still not met, the next lowest priority orders are rejected, until the sorting wave meets all three types of rigid constraints. Rejected outbound orders are returned to the outbound order pool to be aggregated and included in the next batch of aggregation.

[0169] Step S43: For each generated sorting wave, combine the warehouse space layout and material attribute characteristics to reconstruct the execution sequence of picking tasks within each sorting wave, optimize the operation flow within the wave, and form a complete and executable sorting wave operation plan; synchronize the pre-occupancy data of operation resources corresponding to the sorting wave operation plan to the unified spatiotemporal data benchmark in real time, and update the timing status of operation resources.

[0170] Further, step S43 includes:

[0171] Step S431: Retrieve the storage location data and gridded storage space data of all material categories within the corresponding sorting wave from the unified spatiotemporal data benchmark, and simultaneously retrieve the weight and volume attribute data of each material category as the basic input data for reconstructing the picking task execution sequence.

[0172] Step S432: With the goal of minimizing the total picking distance per wave, and combining the principles of continuous regional picking and heavy-goods-first picking, calculate the optimal travel path between each picking location based on the travel cost attribute in the gridded warehouse space data. Sort and reconstruct all picking tasks within the wave to generate the picking task execution sequence within the wave. The principle of continuous regional picking prioritizes continuous picking tasks within the same warehouse area to reduce long-distance movement across areas. The principle of heavy-goods-first picking prioritizes materials with larger weight and volume to reduce the risk of accumulated handling load during operations. The same warehouse area refers to pre-divided physical warehouse zones, each corresponding to a set of continuous grid cell ranges. The principle of heavy-goods-first picking adopts a comprehensive judgment rule based on weight priority, first sorting materials by weight from largest to smallest, and then sorting them by volume from largest to smallest if the weights are the same.

[0173] Furthermore, the generation of the intra-wave picking task execution sequence includes:

[0174] Step S4321: Extract the coordinates of the storage locations of all picking tasks within the corresponding sorting wave, and at the same time match the passage cost attribute of the grid cell where each storage location is located, as well as the weight and volume attribute data corresponding to each material category from the gridded storage space data to construct the basic dataset of picking tasks.

[0175] Step S4322: Based on the global access cost attribute of the gridded warehouse space data, calculate the cumulative access cost between any two picking locations and generate an access cost matrix between locations; the cumulative access cost corresponds to the sum of the access costs of all grid cells on the optimal path between the two locations, which is used to quantitatively characterize the operational travel cost between the two locations.

[0176] Step S4323: According to the principle of continuous picking by region, the picking tasks belonging to the same storage area are divided into the same picking task group; within each picking task group, according to the principle of picking heavy goods first, the picking tasks in the group are initially sorted from largest to smallest according to the weight and volume attributes of the material category, forming the initial sorting result of the group.

[0177] Step S4324: Based on the inter-location cost matrix, optimize the connection order between each picking task group. Using the lowest cumulative cost across all waves as the convergence criterion, determine the optimal execution order for each group. Combine the initial intra-group order with the optimal inter-group order to generate the final intra-wave picking task execution sequence. The inter-group connection order uses a traversal optimization method, calculating the cumulative cost of connecting the first and last locations for all permutations of picking task groups, and selecting the permutation with the lowest total cost as the optimal inter-group order.

[0178] Step S433: Synchronously incorporate the reconstructed picking task execution sequence into the corresponding sorting wave plan, assign a corresponding picking task set and execution order to each operational resource, forming a complete and executable sorting wave operation plan; after unifying the format of the operational resource pre-occupancy data corresponding to the sorting wave operation plan according to the preset data field specifications, write it into the unified spatiotemporal data benchmark in real time to update the temporal status data of the operational resources, providing a basis for subsequent dynamic path planning and operation execution scheduling. The operational resource pre-occupancy data includes the unique identifier of the corresponding operational resource, the pre-occupancy period, the allocated sorting wave number, and the estimated operation duration information; after the pre-occupancy takes effect, the status of the corresponding operational resource during the pre-occupancy period is marked as allocated, and it will no longer be included in the range of available operational resources for subsequent waves.

[0179] Step S44: Based on the historical operation log data in the unified spatiotemporal data benchmark, the core parameter rules of wave aggregation are periodically iterated and optimized. The optimized core parameter rules are synchronously updated to the unified spatiotemporal data benchmark to realize the autonomous adaptation and continuous optimization of the wave scheduling strategy.

[0180] Further, step S44 includes:

[0181] Step S441: According to the preset parameter optimization cycle, retrieve historical operation log data within the corresponding parameter optimization cycle from the unified spatiotemporal data benchmark, and statistically analyze the sorting wave execution efficiency indicators under different time periods, different outbound order characteristics, and different operation resource configurations, as the quantitative basis for core parameter rule iterative optimization. The sorting wave execution efficiency indicators include three categories: total operation time per wave, average response time of outbound orders, and average load rate of operation resources; the preset parameter optimization cycle is a pre-set time interval for parameter rule iterative adjustment, and its setting is based on the fluctuation frequency of warehousing operations and strategy adjustment needs. For example, the preset parameter optimization cycle can be set to optimize once a week.

[0182] Step S442: Based on the statistically obtained sorting wave execution efficiency index, iteratively optimize the core parameter rules for wave aggregation. These core parameter rules include a preset maximum single-wave workload threshold, a preset maximum aggregation interval, weight coefficients corresponding to each optimization direction in the multi-objective optimization function, and corresponding threshold parameters for the three types of rigid constraints. The optimization process aims to improve overall sorting efficiency and outbound fulfillment timeliness. By comparing the effects of multiple parameter combinations, the optimal parameter combination suitable for the current business scenario is obtained. Parameter iterative optimization uses historical backtesting, substituting different parameter combinations into historical business data for simulation aggregation. The comprehensive efficiency scores corresponding to each parameter combination are compared, and the parameter combination with the highest score is selected as the optimal parameter. The corresponding threshold parameters for the three types of rigid constraints include a preset regional cost threshold, an upper limit coefficient for operational resource processing capacity, and a supply rhythm time margin threshold.

[0183] Step S443: Synchronously update the optimized core parameter rules to the unified spatiotemporal data benchmark, replacing the original parameter rules for subsequent intelligent wavelet aggregation calculations; at the same time, establish a parameter version archiving mechanism, assign version numbers and retain historical records for each version of the optimized core parameter rules, and support historical parameter backtracking and comparative analysis of the execution effects of different versions.

[0184] Specifically, step S4, through a closed-loop scheduling logic involving constraint system construction, triggered wave aggregation, sequence optimization reconstruction, and parameter iterative optimization, takes over the set of executable outbound orders output from step S3 and achieves refined intelligent wave scheduling supported by source data with a unified spatiotemporal data benchmark. Step S41 clarifies the boundary rules and optimization objectives of wave aggregation; step S42 improves the response speed of outbound orders through an arrival-triggered dual-trigger mechanism; step S43 further explores the potential for improving single-wave efficiency from the operational flow perspective; and step S44 achieves continuous adaptive optimization of the scheduling strategy through iterative historical operational log data. All calculations and scheduling processes in this step are based on source data with a unified spatiotemporal data benchmark. The generated sorting wave plan and operational resource pre-occupancy data are written back to update the benchmark data in real time, inheriting the results of upstream inventory control and outbound order verification, and providing input for the dynamic path planning and operational execution in the downstream step S5, achieving seamless linkage between outbound order scheduling and sorting operation execution.

[0185] Step S5: Based on the sorting wave operation plan, and combining the gridded warehouse space data and real-time operation status data in the unified spatiotemporal data benchmark, perform full-domain dynamic path planning and issue picking tasks to the corresponding operation resources; collect operation status data in real time during the operation execution process, and carry out operation progress tracking and dynamic handling of anomalies; synchronously write the operation execution result data back to the unified spatiotemporal data benchmark, iteratively optimize the path planning strategy based on historical execution data, and form a dynamic path closed-loop management and control mechanism of path planning, execution monitoring, and feedback optimization.

[0186] The operational status data comprises a multi-dimensional dataset generated in real-time by each operational resource during the execution of picking tasks. This dataset reflects the operational status of the resources and the progress of tasks, supporting progress tracking, identification of abnormal events, and dynamic scheduling decisions. The operational status data includes the real-time location coordinates of the operational resources, the number of currently completed picking tasks, a list of pending picking tasks, equipment operating conditions, and remaining continuous operating time. All operational status data is formatted according to the preset data field specifications and then synchronized in real-time to the unified spatiotemporal data benchmark for storage and updating. Dynamic path planning, based on the travel cost attributes of the gridded warehouse space data, plans the optimal travel path for operational resources to start from their current location, sequentially complete all picking tasks, and arrive at the sorting point. This reduces ineffective travel, avoids traffic conflicts, and improves the efficiency and safety of picking operations.

[0187] Further, step S5 includes:

[0188] Step S51: Based on the sorting wave operation plan and the gridded warehouse space data, perform full-domain dynamic path planning for the picking task set allocated to each operation resource, generate the picking travel path for each operation resource, and issue operation tasks and path instructions to the corresponding operation resources.

[0189] Further, step S51 includes:

[0190] Step S511: Retrieve the operation resource allocation information for the corresponding sorting wave, the picking task execution sequence corresponding to each operation resource, and the location coordinate data of each picking task from the unified spatiotemporal data benchmark. At the same time, retrieve the global access cost attribute and functional attribute label data of the rasterized warehouse space data as the basic input data for dynamic path planning.

[0191] Step S512: Taking the current location of the work resource as the starting point of the path, and the first picking task location in the picking task execution sequence as the first waypoint, all picking task locations are sequentially connected, and the corresponding sorting port location is the final destination of the path. Based on the passage cost attribute of the gridded warehouse space data, a path search algorithm is used to solve the path with the lowest cumulative passage cost between each waypoint, and the paths are spliced ​​together to form a complete picking travel path for a single work resource. The path search algorithm may be, but is not limited to, the A* algorithm and the Dijkstra algorithm; the cumulative passage cost is the sum of the passage cost attribute values ​​of all grid cells passed through by the path, which is used to quantify the travel cost of the path. The path planning strictly follows the predetermined order of the picking task execution sequence, only optimizing the travel path between adjacent task points, without adjusting the execution priority and order of the picking tasks; during the path search process, only grid cells with the functional attribute label of "channel grid" are selected as passable nodes, and the location grid and sorting port grid are only used as starting and ending waypoints.

[0192] Step S513: Perform conflict prediction and verification on the picking routes of all generated work resources to identify overlapping sections and simultaneous occupation risks in different work resource routes; for routes with traffic conflict risks, perform time-series staggered adjustments or route detour optimization to generate conflict-free full-domain work route schemes. The traffic conflict determination rule is: when the expected arrival time difference of two or more work resources to the same grid cell is less than a preset conflict time threshold, it is determined that there is a risk of simultaneous occupation; the preset conflict time threshold is set based on the size of the work resources and the safe driving speed, and can be set to 10 seconds for example. Overlapping sections refer to sections where the number of consecutively overlapping grid cells of two or more paths exceeds a preset overlap length threshold, which is set based on the channel width and the safe working distance, and can be set to 5 consecutive grid cells for example. Timing-based off-peak adjustment refers to delaying the departure time or the time of passing through conflict points of low-priority work resources so that the time difference meets the conflict time threshold requirement; route detour optimization refers to replanning alternative routes for conflicting road sections. When the cumulative toll cost of detour is lower than the preset detour cost upper limit, route detour optimization is given priority; otherwise, timing-based off-peak adjustment is used.

[0193] Step S514: The optimized global operation path plan and corresponding picking task information are synchronously distributed to each operation resource. Simultaneously, the expected travel trajectory and estimated arrival time at each storage location for the operation resource corresponding to the path plan are written into the unified spatiotemporal data benchmark, updating the temporal status data of the operation resources and providing a planning benchmark for subsequent operation monitoring. The picking task information is a set of detailed instruction data distributed to the corresponding operation resource to guide it in completing the specified picking operation. For example, it includes the sorting wave identifier, corresponding outbound order number, category and quantity of materials to be picked, target storage location coordinates, task execution sequence, estimated operation time, and target sorting port location information. This clarifies the operation content and execution standards of the operation resource, ensuring accurate and orderly picking operations. The estimated arrival time at each storage location is calculated based on the cumulative toll cost of the corresponding road segment, the average travel speed of the operation resource, and the time consumed by the preceding picking operation. The total travel time is obtained by converting the cumulative toll cost into the actual path length and then dividing it by the average travel speed of the operation resource.

[0194] Step S52: Collect real-time operational status data of each operational resource, track the progress of sorting operations, and dynamically adjust and handle any abnormal progress events during the execution process to ensure the smooth and orderly execution of sorting operations.

[0195] Further, step S52 includes:

[0196] Step S521: Collect location data, task execution progress, and equipment operating status data of each work resource in real time according to a preset status collection cycle. Synchronously upload the collected real-time work status data to the unified spatiotemporal data benchmark to update the real-time status information of the work resources. The preset status collection cycle is a pre-set time interval for collecting work status data, set based on the positioning accuracy and scheduling response requirements of the work equipment. For example, the preset status collection cycle can be set to 5 seconds. Task execution progress is quantified by the ratio of the number of completed picking tasks to the total number of picking tasks. Simultaneously, it is combined with the ratio of the cumulative travel cost of the traveled path to the total cumulative travel cost of the path for auxiliary verification to ensure the accuracy of progress statistics.

[0197] Step S522: Based on real-time operation status data and the planned progress in the sorting wave operation plan, compare the deviation between the actual execution progress and the planned progress of each operation resource. When the progress deviation rate exceeds a preset progress deviation threshold, it is marked as a progress anomaly event. The preset progress deviation threshold is a pre-set critical value for judging progress anomalies, and its setting is based on the timeliness control requirements of warehousing operations. For example, the preset progress deviation threshold can be set to 10%. The progress deviation rate is calculated as follows: Progress deviation rate = (Planned progress - Actual progress) ÷ Planned progress × 100%; a progress anomaly event is marked only when the actual progress lags behind the planned progress and the progress deviation rate exceeds the preset progress deviation threshold; the planned progress is calculated based on the expected time nodes in the sorting wave operation plan, according to the corresponding proportion at the current time.

[0198] Step S523: For different types of progress anomalies identified, execute corresponding dynamic handling strategies: If it is a work resource equipment malfunction, the unfinished picking tasks of that work resource are reassigned to nearby idle work resources, and the corresponding picking path is replanned based on the updated task sequence; if it is a temporary road obstruction anomaly, the path is recalculated based on the real-time updated grid traffic status, and the obstructed road section is bypassed. After the anomaly handling is completed, the sorting wave operation plan and work resource status data are updated synchronously to the unified spatiotemporal data benchmark. Work resource equipment malfunctions are identified through equipment operating status data. When a work resource reports a fault status code, or when no work status data is returned for a consecutive preset number of collection cycles, it is determined to be an equipment malfunction. The preset number can be set to 3 to 5 collection cycles, exemplarily set to 3 collection cycles, to match the data collection frequency and avoid misjudgment due to single signal fluctuations. Temporary road obstruction anomalies are determined by the real-time dwell time of operational resources. When the dwell time of an operational resource in a non-task node grid exceeds a preset dwell threshold, the corresponding road segment is determined to be obstructed, and the passage cost of the corresponding grid cell is simultaneously updated to an impassable state. The preset dwell threshold is exemplarily set to 60 seconds, and can be adjusted according to the complexity of the warehousing environment and the average movement speed of operational resources. Adjacent idle operational resources are determined using a cumulative passage cost threshold. When the cumulative passage cost from the current location of an idle operational resource to the next task location of a faulty operational resource is lower than a preset adjacent threshold, it is determined to be an adjacent idle operational resource. The preset adjacent threshold is a preset maximum allowable detour cost, exemplarily set to 1.5 times the normal passage cost to ensure the timeliness of task reallocation. Task reallocation follows the rule of outbound order priority from high to low, prioritizing the transfer of unfinished picking tasks corresponding to high-priority outbound orders.

[0199] Step S524: Once all picking tasks within a picking wave are completed and materials are delivered to the corresponding sorting point, mark the picking wave as completed. Simultaneously release the pre-occupancy status of the corresponding operational resources and the pre-occupancy status of the corresponding outbound orders, deduct the real-time inventory data of the corresponding material category, and update the inventory data and operational resource status data in the unified spatiotemporal data benchmark. The criterion for determining the completion of a picking wave is: all materials corresponding to all picking tasks are delivered to the target sorting point and have completed barcode verification. The data update execution order is as follows: first, release the pre-occupancy status of the corresponding outbound orders; then, deduct the real-time inventory data of the corresponding material category according to the actual picking and outbound quantity; finally, release the pre-occupancy status of the corresponding operational resources. All update operations are synchronously written to the unified spatiotemporal data benchmark to ensure the consistency between inventory data and operational status.

[0200] Step S53: Synchronize and archive the execution data of the entire wave of operations to the unified spatiotemporal data benchmark, and iteratively optimize the core parameters and strategies of path planning based on historical execution data to achieve continuous improvement in path control capabilities.

[0201] Furthermore, step S53 includes:

[0202] Step S531: After the sorting wave is completed, summarize the actual total operation time of the wave, the actual mileage of each operation resource, the actual passage time of each road segment, and the record data of progress abnormal events to form a complete operation execution log, which is then archived and stored in the historical operation log data in the unified spatiotemporal data benchmark.

[0203] Step S532: According to the preset path optimization cycle, retrieve the historical operation execution logs for the corresponding cycle from the unified spatiotemporal data benchmark, statistically analyze the actual traffic efficiency data for different time periods and regions, update the actual traffic cost attributes of each grid cell in the gridded warehouse space data, and correct the cost calculation benchmark for path planning. The preset path optimization cycle is a pre-set time interval for iterative adjustments to the path strategy, and its setting is based on the frequency of changes in the warehouse operation environment. For example, the preset path optimization cycle can be set to optimize once a week. The statistical dimension of the actual traffic efficiency data is the actual traffic time per unit distance for each grid cell.

[0204] Step S533: Based on the actual execution results obtained from statistics, iteratively optimize the core parameters of the path planning. These core parameters include conflict prediction threshold, path detour weight, and anomaly handling response priority. After optimization, the updated path planning parameters are synchronously stored in the unified spatiotemporal data benchmark, replacing the original parameters for subsequent dynamic path planning calculations. A parameter version archiving mechanism is established synchronously, assigning version numbers and retaining historical records for each optimized version of the path planning parameters, supporting parameter backtracking and effect comparison. The iterative optimization of the core parameters adopts a historical backtesting method, substituting different parameter combinations into historical operation data for simulated path planning. The comprehensive traffic efficiency of the simulated scheme and the actual execution is compared, and the parameter combination with the highest comprehensive efficiency is selected as the optimal parameter. The conflict prediction threshold includes a conflict time threshold and an overlap length threshold. The path detour weight refers to the penalty coefficient of the increased detour cost in path optimization. The anomaly handling response priority is divided into three response levels according to the priority of outbound orders and the scope of operation impact, from high to low.

[0205] Specifically, step S5, through a three-layer closed-loop logic of global path planning, real-time execution monitoring, and iterative strategy optimization, follows the sorting wave operation plan output from step S4. Supported by source data from a unified spatiotemporal data benchmark, it achieves refined dynamic path control. Specifically, step S51 completes globally optimal path planning and conflict avoidance based on rasterized spatial data; step S52 ensures the stability of operation execution through real-time status acquisition and dynamic anomaly handling; and step S53 achieves continuous adaptive optimization of the path strategy through iterative iteration of historical execution data. All planning, monitoring, and optimization processes in this step are conducted based on source data from a unified spatiotemporal data benchmark. Execution results and optimization parameters are written back to update the benchmark data in real time. This not only supports the implementation of upstream wave scheduling outputs but also provides actual operational data support for upstream inventory management and wave aggregation, ultimately forming an intelligent warehousing and logistics management system with unified data across the entire process, collaborative decision-making, and closed-loop optimization.

[0206] Example 2:

[0207] This embodiment introduces an intelligent warehousing and logistics management system based on multi-source data fusion, including:

[0208] The benchmark construction module collects multi-source logistics data from the supply, warehousing, and demand sides, performing standardized cleaning processes such as missing value completion, outlier removal, and duplicate data deduplication. It then maps and unifies the fields of the multi-source data according to preset data field specifications, generating a standardized logistics dataset. For logistics data related to the mixed transportation of multiple material categories in the same batch on the supply side, it performs granular decomposition by material category dimension, extracting the inbound operation characteristic parameters of each material category, and completing the time-series mapping of mixed data from batch granularity to material category granularity. Next, using a continuous time axis as the benchmark dimension and material category as the smallest management unit, it performs time dimension alignment and spatial dimension association processing to construct a unified spatiotemporal data benchmark including inventory level time-series sequences and spatial location attributes. Simultaneously, an optimistic locking concurrency control mechanism ensures data consistency during concurrent reads and writes across multiple stages, and a synchronization mechanism primarily based on event triggering and supplemented by timed polling achieves global synchronization of data throughout the entire process, providing a unique data carrier consistent with the same source for all subsequent management and control stages.

[0209] The monitoring and forecasting module is used to call historical inbound and outbound data within a preset historical period from a unified spatiotemporal data benchmark. It calculates the co-transport correlation of material category pairs from the supply side and the co-demand correlation of material category pairs from the demand side. Based on preset correlation thresholds, it determines strong correlations and constructs a material correlation feature map containing supply correlation sub-maps and demand correlation sub-maps. The map structure is iteratively optimized through a periodic update mechanism combining full and incremental updates. Based on the material correlation feature map and real-time inventory data, it simultaneously conducts two-layer dynamic inventory monitoring, including monitoring the inventory level of individual material categories and the availability status of significantly related material groups. It identifies inventory gaps and associated supply risks and generates tiered early warning information. At the same time, it performs basic demand forecasting for individual material categories based on historical outbound data. It corrects the forecast results by combining the demand linkage patterns in the material correlation feature map, generates multi-period final demand forecast results for each material category, and synchronizes the early warning information and demand forecast results to the unified spatiotemporal data benchmark in real time.

[0210] The verification management module combines final demand forecasts, real-time inventory data, expected inbound data for future periods, and pre-allocated outbound data to calculate the dynamic available inventory sequence for each material category using a rolling, time-by-time approach. For outbound orders containing combinations of multiple material categories, it uses the agreed outbound time point as the verification node to perform a full-material-category linked time-series verification to determine the immediate executable status of the order. For orders that cannot be executed immediately, it deduces and locates the earliest executable time point. Based on preset outbound order splitting rules, it splits orders where some materials meet inventory conditions, generating immediately executable outbound sub-orders and pending outbound sub-orders. It aggregates all immediately executable outbound orders and immediately executable outbound sub-orders to construct an executable outbound order set, simultaneously performing inventory pre-occupancy locking and release management, and establishing a multi-event triggered rolling update mechanism to ensure the real-time validity and uniqueness of the executable outbound order set and inventory data.

[0211] The aggregation and reconstruction module extracts three types of rigid constraints—supply rhythm constraints, operational resource constraints, and warehouse layout constraints—from a unified spatiotemporal data benchmark. It configures multi-objective optimization directions: minimizing the total sorting operation journey, minimizing the average response time of outbound orders, and balancing the load rate of various operational resources. After extreme value normalization, a multi-objective optimization function is constructed through weighted summation. A dual-trigger scheduling mechanism is adopted, with inbound arrival event triggering as the core and fixed-duration timed triggering as a supplement. Combined with material association feature maps, pre-grouping of outbound orders to be aggregated is performed. The optimal outbound order combination scheme is solved iteratively through a multi-objective optimization algorithm, generating an initial sorting wave plan and performing compliance checks on the three types of rigid constraints. Based on the principles of continuous regional picking and heavy goods picking first, and combined with the passage cost attributes of gridded warehouse space data, the picking task execution sequence within each sorting wave is reconstructed to form a complete and executable sorting wave operation plan. Simultaneously, operational resource pre-occupancy data is written into the unified spatiotemporal data benchmark, and the core parameter rules of wave aggregation are periodically iteratively optimized based on historical operation log data to achieve continuous adaptive optimization of the scheduling strategy.

[0212] The planning and control module, based on sorting wave operation plans and gridded warehouse space data, sequentially connects all picking task locations, starting from the current location of the task resource and ending at the sorting gate. It uses a path search algorithm to solve for the lowest cumulative travel cost path between each waypoint, piecing together a complete picking route for a single task resource. The module performs conflict prediction and verification on the picking routes of all task resources, identifying overlapping sections and concurrent occupancy risks. Through time-series peak-shifting adjustments or route detour optimization, it generates conflict-free global operation path solutions and issues picking task information and routes to the corresponding task resources. The instructions include: collecting operational status data of each operational resource in real time according to a preset status collection cycle; comparing the data with the planned progress to identify abnormal events; and implementing dynamic handling strategies such as task reassignment and route replanning for abnormal scenarios such as equipment failure and temporary road obstruction. After the sorting wave is completed, the pre-occupied inventory and operational resources are released in a predetermined order, real-time inventory is deducted, the full operational execution log is archived, and the grid unit passage cost attributes and route planning core parameters are updated regularly based on historical traffic efficiency data, forming a dynamic closed-loop management mechanism for route planning, execution monitoring, and feedback optimization.

[0213] Working principle and its effects:

[0214] This invention takes the integration and unification of multi-source logistics data as its core underlying logic. By connecting the business data of the supply side, warehousing side and demand side, it constructs a unified spatiotemporal data benchmark with the same source. Based on this, it connects the entire process of inventory control, order fulfillment, wave scheduling and sorting execution. Relying on the data source, it realizes the coordinated linkage of decision-making in each link, solves the decision bias problem caused by the fragmentation of data from multiple systems, and systematically drives the dual improvement of warehousing and logistics control accuracy and operational efficiency.

[0215] At the foundational data layer, through standardized cleaning, category-level granular decomposition, and alignment across time and space, scattered and heterogeneous multi-source business data are transformed into standardized datasets with the same caliber and sequence. This clarifies the effective inventory timing for different materials in mixed-delivery scenarios and provides a unique and reliable data support for all subsequent control processes, ensuring consistency in the entire decision-making logic from the source. At the inventory demand layer, based on unified data mining, the correlation patterns between material supply and demand are analyzed and feature maps are constructed. A two-layer control model combining single-material level monitoring and strongly correlated group-based monitoring is adopted. Simultaneously, the demand linkage relationships between materials are used to correct individual... The category forecasting results not only optimized the inventory structure and improved the completeness of fulfillment, but also made demand predictions more aligned with the actual business linkage patterns, effectively reducing the risk of inventory structure imbalance. At the scheduling and execution layer, relying on time-series dynamic inventory accounting and order linkage verification, inventory resources and outbound demand are accurately matched. Combined with three types of rigid constraints and arrival triggering mechanisms, intelligent wave aggregation is completed. With the addition of picking sequence reconstruction and conflict-free path planning, sorting operations are implemented, which not only speeds up the response speed of order fulfillment, but also optimizes the task arrangement and travel path of sorting operations, improving the overall sorting operation efficiency while balancing the workload of operational resources.

[0216] In summary, this invention uses a unified spatiotemporal data benchmark as its core link to form a closed-loop management system that integrates data fusion support, pre-emptive inventory prediction, precise scheduling adaptation, and efficient execution. Each link relies on the same source data to achieve decision-making linkage and effect transmission, ultimately achieving synergistic improvement in inventory control accuracy, order response speed, and sorting operation efficiency. It can fully adapt to warehousing business scenarios with multiple material categories and rapid demand changes.

[0217] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A smart warehousing and logistics management method based on multi-source data fusion, characterized in that, include: Collect multi-source logistics data from the supply side, warehousing side, and demand side, perform standardized cleaning, granular decomposition and time-series alignment processing, and build a unified spatiotemporal data benchmark; Based on the unified spatiotemporal data benchmark, the supply-side co-transportation correlation characteristics and demand-side co-demand correlation characteristics among material categories are mined to construct a material correlation feature map, which is used to perform inventory level monitoring of single material categories and completeness monitoring of significantly related material groups, generate demand forecast results for each material category, and update the time-series inventory data of the corresponding material categories. Calculate the dynamic available inventory for each material category, use the outbound time of the outbound order as the verification node, perform linked time-series verification on the outbound order, and construct a set of executable outbound orders; Based on a unified spatiotemporal data benchmark, rigid constraints are extracted and a multi-objective optimization function is constructed. Intelligent wave aggregation is performed on the executable outbound order set to generate a sorting wave plan and reconstruct the execution sequence of picking tasks within each sorting wave. Based on the sorting wave operation plan of the execution sequence, dynamic path planning is performed to issue picking tasks to the corresponding operation resources.

2. The intelligent warehousing and logistics management method based on multi-source data fusion as described in claim 1, characterized in that, The construction of the material association feature map includes: Call up historical inbound and outbound data within a preset historical period, and perform cleaning and verification processing based on material category codes to form a sample dataset for correlation analysis; Based on the aforementioned correlation analysis sample dataset, correlation statistics were performed from both the supply and demand sides: For the supply side, count the number of times any two material categories appear together in the same inbound batch, and calculate the co-transport correlation degree of the corresponding material category pair; On the demand side, count the number of times any two material categories appear together in the same outbound order, and calculate the co-demand correlation degree of the corresponding material category pair; The co-transportation correlation and co-demand correlation of each material category pair are compared with the preset correlation thresholds to identify strong correlation relationships and mark the strong correlation attributes on the supply side and the strong correlation attributes on the demand side. Using strong correlations as graph connection edges and co-operation correlation degree or co-demand correlation degree as edge weight parameters, a material correlation feature graph containing supply correlation sub-graph and demand correlation sub-graph is constructed.

3. The intelligent warehousing and logistics management method based on multi-source data fusion as described in claim 1, characterized in that, The execution of single-material category inventory level monitoring and significantly related material group completeness status monitoring includes: Retrieve real-time inventory data for each material category. When the real-time inventory level of a material category is lower than the corresponding preset safety stock threshold, generate an inventory shortage warning for the corresponding material category. Based on the significantly associated material groups in the material association feature map, real-time inventory data of all material categories in each significantly associated material group is retrieved. Combined with the material matching ratio corresponding to each significantly associated material group, the inventory matching rate of each significantly associated material group is calculated. When the inventory matching rate is lower than the preset matching rate threshold, the matching supply risk warning information of the corresponding significantly associated material group is generated.

4. The intelligent warehousing and logistics management method based on multi-source data fusion as described in claim 1, characterized in that, The generation of demand forecast results for each material category includes: Historical outbound data within a preset prediction period is retrieved from the unified spatiotemporal data benchmark. The future demand trend of each material category is fitted and calculated, and the expected outbound demand for the corresponding time period is output in a preset time granularity. Based on the demand correlation sub-map in the material correlation feature map, the core driving material and supporting follower material of each significantly correlated material group are identified, the demand linkage coefficient of each material in the significantly correlated material group is calculated, and the deviation of the basic demand prediction result is corrected based on the basic demand prediction result of the core driving material to obtain the final demand prediction result of each material category.

5. The intelligent warehousing and logistics management method based on multi-source data fusion as described in claim 1, characterized in that, The construction of the executable outbound order set includes: By combining demand forecast results, current real-time inventory data, expected inbound data for each future period, and pre-allocated outbound data, the dynamic available inventory of each material category is calculated periodically according to the preset time granularity, forming a time-segmented dynamic available inventory sequence. Using the outbound time of the outbound order as the verification node, and combining the dynamic available inventory sequence of the corresponding time period, perform a full material category linkage time sequence verification to determine the execution status of the outbound order, identify outbound orders that can be executed immediately, and split outbound orders that cannot be executed immediately into immediately executable outbound sub-orders. The outbound orders that are determined to be immediately executable and the outbound sub-orders that are immediately executable are aggregated to construct an executable outbound order set.

6. The intelligent warehousing and logistics management method based on multi-source data fusion as described in claim 1, characterized in that, The extraction of rigid constraints and the construction of a multi-objective optimization function include: Based on a unified spatiotemporal data benchmark, rigid constraints are extracted, including supply rhythm constraints, operational resource constraints, and warehouse layout constraints. The supply rhythm constraint is that for all outbound orders included in the same sorting wave, the available inventory of all material categories becomes effective earlier than the planned execution start time of the corresponding sorting wave. The operational resource constraints are that the total picking volume, total number of material categories, and estimated operation time of a single sorting wave are less than the current operational resource's processing capacity limit. The warehouse layout constraint is that the area coverage of the picking location within the same sorting wave meets the preset area limit. Configure multi-objective optimization directions and construct multi-objective optimization functions; the multi-objective optimization directions include minimizing the total sorting operation journey, minimizing the average response time of outbound orders, and balancing the load rate of each operation resource.

7. The intelligent warehousing and logistics management method based on multi-source data fusion as described in claim 1, characterized in that, The process of intelligently aggregating the executable outbound order set to generate a sorting wave plan and reconstructing the execution sequence of picking tasks within each sorting wave includes: Based on inbound arrival event triggering and fixed-duration timed triggering, a dual-trigger scheduling mechanism for intelligent wave aggregation is constructed to start intelligent wave aggregation calculation; Based on the current set of executable outbound orders, and combined with the material association feature map, the outbound orders are pre-grouped. Under the boundary constraints of supply rhythm, operation resource, and warehouse layout, the multi-objective optimization function is solved by a multi-objective optimization algorithm to obtain the optimal outbound order combination scheme and generate the initial sorting wave plan. Based on rigid constraints, the initial sorting wave plan is subjected to constraint compliance verification, and a sorting wave plan is generated. For each sorting wave generated, the execution sequence of picking tasks within each sorting wave is reconstructed by combining the warehouse space layout and material attribute characteristics, thus generating a sorting wave operation plan.

8. The intelligent warehousing and logistics management method based on multi-source data fusion as described in claim 1, characterized in that, The execution of dynamic path planning, which involves issuing picking tasks to corresponding operational resources, includes: Retrieve the resource allocation information for the corresponding sorting wave, the picking task execution sequence for each resource, and the location coordinates for each picking task; Simultaneously retrieve the access cost attribute and functional attribute label data of the rasterized warehouse space data; Taking the current location of the operational resource as the starting point of the path, and the first picking task location in the picking task execution sequence as the first waypoint, all picking task locations are sequentially connected; based on the passage cost attribute of the gridded warehouse space data, the lowest cumulative passage cost path between each waypoint is solved, and the paths are spliced ​​together to form the picking travel path of a single operational resource. Perform conflict prediction and verification on the picking and driving paths of all generated job resources to generate conflict-free global job path solutions; The overall operation path plan and corresponding picking task information are simultaneously distributed to each operation resource.

9. The intelligent warehousing and logistics management method based on multi-source data fusion as described in claim 1, characterized in that, The construction of a unified spatiotemporal data benchmark includes: It receives multi-source logistics data from the supply side, warehousing side, and demand side. After cleaning the multi-source logistics data, it performs field mapping and format unification processing on the multi-source logistics data according to the preset data field specifications to generate a standardized logistics dataset. The standardized logistics dataset is decomposed into multi-source logistics data at the material category level, and the inbound operation feature parameters corresponding to each material category are extracted to update the standardized logistics dataset. Using a continuous time axis as the baseline dimension and material category as the smallest management unit, the updated standardized logistics dataset is aligned in the time dimension and associated in the spatial dimension to construct a unified spatiotemporal data benchmark that includes inventory level time series and spatial location attributes.

10. An intelligent warehousing and logistics management system based on multi-source data fusion, used to implement the intelligent warehousing and logistics management method based on multi-source data fusion as described in any one of claims 1-9, characterized in that, include: The benchmark construction module collects multi-source logistics data from the supply side, warehousing side, and demand side, performs standardized cleaning, granular decomposition, and time-series alignment processing, and constructs a unified spatiotemporal data benchmark. The monitoring and prediction module, based on the unified spatiotemporal data benchmark, mines the supply-side co-transportation correlation characteristics and demand-side co-demand correlation characteristics between material categories, constructs a material correlation feature map, and is used to perform single material category inventory level monitoring and significant correlation material group completeness status monitoring, generate demand prediction results for each material category, and update the time series inventory data of the corresponding material category. The verification management module calculates the dynamic available inventory for each material category, uses the outbound time point of the outbound order as the verification node, performs linked time-series verification on the outbound order, and constructs a set of executable outbound orders; The aggregation and reconstruction module extracts rigid constraints based on a unified spatiotemporal data benchmark and constructs a multi-objective optimization function. It performs intelligent wave aggregation on the executable outbound order set, generates sorting wave plans, and reconstructs the execution sequence of picking tasks within each sorting wave. The planning and control module performs dynamic path planning based on the sorting wave operation plan of the execution sequence and issues picking tasks to the corresponding operation resources.

Images