A warehouse classification management system and method

Abstract

The application relates to the technical field of warehouse management, in particular to a warehouse classification management system and method, which comprises a dynamic characteristic sensing module, a self-adaptive decision engine, an elastic scheduling execution module and a device execution layer; the dynamic characteristic sensing module acquires and updates warehouse dynamic data in real time; the self-adaptive decision engine dynamically optimizes material classification rules, divides materials into integral carrying types and independent taking and placing types, matches exclusive devices with operation strategies; the elastic scheduling execution module analyzes strategies into device instructions and issues the device instructions; the device execution layer executes operations and feeds back state data, forming a closed-loop scheduling. Through a closed-loop process of data acquisition, rule optimization, material classification, strategy generation, instruction execution and state feedback, the application realizes self-adaptive adjustment of warehouse operations, improves operation efficiency and device utilization, reduces operation and maintenance costs, and is suitable for modern intelligent warehouse scenes.

Description

Technical Field

[0001] This invention relates to the field of warehouse management technology, specifically to a warehouse classification management system and method. Background Technology

[0002] Current warehouse management systems generally suffer from several pressing problems. Firstly, material classification rules are rigid. Existing systems often classify materials based on static, inherent attributes such as volume and weight, without dynamically adjusting to real-time task queue status and equipment operating status. This leads to a disconnect between material classification results and actual operational needs, making it impossible to prioritize material handling efficiency for high-urgency tasks during peak periods. Secondly, equipment scheduling strategies are inflexible. Most systems employ preset, fixed scheduling strategies, with automated guided vehicles (AGVs) and bin-handling robots typically performing separate tasks. The system does not dynamically allocate tasks based on real-time equipment load and power levels, easily resulting in some equipment operating under continuous overload while others remain idle for extended periods, leading to low utilization of warehouse equipment resources. Thirdly, data perception and... The system suffers from several problems: 1) Disconnect between decision-making and execution; 2) Data collection in existing systems is mostly localized or periodic, lacking comprehensive real-time perception of material characteristics, task queues, and equipment status; 3) The decision engine cannot acquire dynamically updated warehouse data, resulting in often lagging operational strategies; 4) Untimely feedback of status data during operation execution prevents dynamic optimization of the decision-making process, hindering the formation of a complete closed loop of data collection, decision generation, instruction execution, and status feedback; 5) High degree of human intervention; Traditional warehouse management still relies on human experience for decision-making in key areas such as material classification, equipment task assignment, and path planning; 6) This significantly increases human resource management costs and is prone to errors due to human error, leading to material misclassification and task delays, making it unsuitable for large-scale automated warehousing operations.

[0003] The aforementioned problems result in overall low efficiency in warehousing operations, insufficient rational allocation of resources, and high operation and maintenance costs, failing to meet the development trends of high efficiency, intelligence, and flexibility in the modern warehousing industry. Therefore, developing a warehouse classification management system and method capable of dynamically optimizing material classification rules, intelligently scheduling handling equipment, and achieving closed-loop management throughout the entire process has become an urgent need for the industry's current development. Summary of the Invention

[0004] The purpose of this invention is to provide a warehouse classification management system and method to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A warehouse classification management system, comprising:

[0007] The dynamic feature perception module is used to acquire and update warehouse dynamic data in real time. The warehouse dynamic data includes at least: material feature data, in-transit task queue status data, and real-time status data of various handling equipment in the equipment execution layer.

[0008] An adaptive decision engine communicates with the dynamic feature perception module; the adaptive decision engine includes:

[0009] The rule optimization unit is used to dynamically calculate and output the current optimal material classification threshold or rule parameters based on real-time task queue status data and real-time equipment status data; the material classification includes at least a first category of materials suitable for overall handling and a second category of materials suitable for independent handling.

[0010] The strategy generation unit is used to classify target materials into corresponding categories based on material characteristic data and the current optimal rule parameters output by the rule optimization unit, and dynamically assemble and generate corresponding integrated operation strategies from a pre-set strategy component library based on real-time equipment status data; wherein, the strategy for assembling the first type of materials specifies at least a hidden automated guided vehicle as the core handling equipment and a mobile shelf as the handling unit; the strategy for assembling the second type of materials specifies at least a bin handling robot as the core handling equipment and a standardized bin as the handling unit.

[0011] The elastic scheduling execution module communicates with the adaptive decision engine and is used to parse the integrated operation strategy into a specific set of equipment drive instructions and send them to the equipment execution layer; the equipment execution layer includes at least one lurking automated guided vehicle and at least one bin handling robot.

[0012] The state changes generated during the execution of tasks by the equipment execution layer are fed back to the dynamic feature perception module in real time, forming a closed-loop scheduling.

[0013] As a preferred solution, the dynamic feature perception module is used to acquire and update warehouse dynamic data in real time, including:

[0014] The raw data acquisition unit is used to periodically capture raw warehouse data through a preset sensor group and warehouse management system interface. The raw warehouse data includes at least the initial characteristic information of the materials, the initial task queue information, and the initial status information of each handling equipment.

[0015] The data preprocessing unit is connected to the raw data acquisition unit and is used to clean and format the raw warehouse data to obtain formatted warehouse data. The formatted warehouse data includes at least a structured material feature dataset, a structured task queue dataset, and a structured equipment status dataset.

[0016] The data verification and fusion unit communicates with the data preprocessing unit and is used to perform logical consistency verification and multi-source data fusion on the formatted warehouse data to generate dynamic warehouse data after fusion and verification.

[0017] The dynamic update unit communicates with the data verification and fusion unit and is used to incrementally update and manage the version of the fused and verified warehouse dynamic data based on the real-time operation feedback from the equipment execution layer and external system instructions, so as to form the final updated warehouse dynamic data.

[0018] The data output unit communicates with the dynamic update unit and is used to push the final updated warehouse dynamic data to the adaptive decision engine in real time according to a preset format, so that the rule optimization unit and strategy generation unit can call it.

[0019] As a preferred embodiment, the rule optimization unit dynamically calculates and outputs the current optimal material classification threshold or rule parameters based on real-time task queue status data and real-time equipment status data, including:

[0020] Receive real-time task queue status data and real-time device status data, and parse and quantify the data to generate a quantitative indicator system that includes task urgency indicators and device load balancing indicators.

[0021] Based on a quantitative indicator system, the overall operational efficiency of the current warehouse and the real-time capacity margin of each handling equipment are assessed, and a warehouse status assessment result including a comprehensive efficiency score and a description of equipment capacity status is generated.

[0022] Based on the warehouse status assessment results, multiple sets of candidate classification rule parameters that match the current warehouse status are extracted from the preset parameter space to form a candidate rule parameter set;

[0023] The preset optimization decision model is used to simulate and predict the benefits of each group of parameters in the candidate rule parameter set, and the candidate rule parameter with the highest comprehensive benefit prediction is selected as the current optimal material classification threshold or rule parameter.

[0024] The current optimal material classification threshold or rule parameters are encapsulated and standardized in format, and output to the strategy generation unit for use. At the same time, the quantitative indicator system and warehouse status assessment results on which this decision is based are archived to the historical rule base for subsequent iterative learning of the optimization decision model.

[0025] As a preferred embodiment, the strategy generation unit is used to classify the target material into the corresponding category based on material characteristic data and the current optimal rule parameters output by the rule optimization unit, and dynamically assemble and generate the corresponding integrated operation strategy from a pre-set strategy component library based on real-time equipment status data, including:

[0026] Receive material feature data and current optimal rule parameters, and match and judge the material feature data according to the classification threshold contained in the current optimal rule parameters, and output the material classification result containing the first type of material identifier set and the second type of material identifier set;

[0027] Receive real-time equipment status data and extract real-time location information, power information, task load information of each hidden automated guided vehicle, as well as real-time availability status information and load status information of each bin handling robot, to generate a structured equipment status description set.

[0028] Based on the material classification results and the structured equipment status description set, retrieve and select strategy components that match the current material category and equipment status from the pre-set strategy component library; among them, the strategy components for matching materials in the first type of material identifier set include at least a hidden automated guided vehicle scheduling component, a mobile shelf docking path planning component, and an overall handling task sequence generation component; the strategy components for matching materials in the second type of material identifier set include at least a bin handling robot assignment component, a standardized bin storage and retrieval operation component, and an independent pick-and-place task sequence generation component;

[0029] The selected strategy components are combined and parameter-bound according to the predefined strategy logic architecture to generate an initial integrated operation strategy that includes specific equipment assignment, operation path, task sequence and control parameters.

[0030] Based on the latest structured equipment status description set, conflict detection and efficiency evaluation are performed on the initial integration operation strategy, and the detected equipment resource conflicts or path conflicts are resolved and optimized to output the final executable integration operation strategy.

[0031] The final executable integrated job strategy is encapsulated according to a predetermined protocol format and output to the elastic scheduling execution module.

[0032] As a preferred solution, the elastic scheduling execution module is used to parse the integrated job strategy into a specific set of device driver instructions and distribute it to the device execution layer, including:

[0033] The instruction parsing and decomposition unit is used to receive the integrated operation strategy output by the strategy generation unit, parse the integrated operation strategy, extract the equipment assignment information, operation sequence information and control parameter information, and generate the parsed strategy elements.

[0034] The instruction conversion and adaptation unit is connected in communication with the instruction parsing and decomposition unit. It is used to identify the target equipment type based on the equipment assignment information in the parsed strategy elements, and convert the job sequence information and control parameter information into a set of drive instruction drafts that match the target equipment type based on the preset equipment instruction template library.

[0035] The timing arrangement and verification unit is connected in communication with the instruction conversion and adaptation unit. It is used to receive the driver instruction draft set and, in combination with the real-time device status data obtained from the dynamic feature perception module, arrange the execution timing of each instruction in the driver instruction draft set and perform conflict pre-check, and generate the timing arranged instruction set.

[0036] The instruction distribution and synchronization unit is connected to the timing arrangement and verification unit. It is used to sequentially send each driving instruction to the corresponding target handling device in the device execution layer through the device communication interface according to the instruction execution timing of the instruction set after timing arrangement.

[0037] The instruction confirmation and monitoring unit is connected to the instruction distribution and synchronization unit. It is used to monitor and receive instruction reception confirmation signals and initial execution status signals returned by each handling device in the equipment execution layer in real time after the instruction is issued.

[0038] The execution log generation unit communicates with the instruction confirmation and monitoring unit. It generates a structured instruction execution log based on the instruction receipt confirmation signal, the initial execution status signal, and the content of the issued instruction, and pushes the structured instruction execution log synchronously to the dynamic feature perception module.

[0039] As a preferred solution, the state changes generated during the execution of tasks by the equipment execution layer are fed back to the dynamic feature perception module in real time, forming a closed-loop scheduling; including:

[0040] The status change acquisition unit is used to collect status change data generated by various handling equipment in the equipment execution layer during the execution of operations in real time. The status change data includes at least equipment position update data, task execution progress data, and equipment fault alarm data.

[0041] The state data preprocessing unit is connected in communication with the state change acquisition unit. It is used to clean, timestamp align and format standardize the state change data to generate preprocessed state data.

[0042] The feedback data encapsulation unit is communicatively connected to the state data preprocessing unit and is used to encapsulate the preprocessed state data into a feedback data packet with a unified protocol format.

[0043] The feedback transmission unit is communicatively connected to the feedback data encapsulation unit and is used to transmit the feedback data packet to the dynamic feature sensing module in real time via a wireless communication network.

[0044] The dynamic data update unit is connected to the feedback transmission unit and is used to receive feedback data packets in the dynamic feature perception module, parse them, and then merge and incrementally update them with the current warehouse dynamic data to generate real-time updated warehouse dynamic data.

[0045] The closed-loop triggering unit communicates with the dynamic data update unit. When real-time updated warehouse dynamic data is generated, it automatically triggers the adaptive decision engine to re-execute rule optimization and strategy generation based on the real-time updated warehouse dynamic data, so as to generate a new integrated operation strategy and send it to the equipment execution layer through the elastic scheduling execution module, thereby forming a closed-loop scheduling.

[0046] A warehouse classification management method, comprising the following steps:

[0047] S1: Real-time acquisition and updating of warehouse dynamic data; warehouse dynamic data includes at least: material characteristic data, in-transit task queue status data, and real-time status data of various handling equipment in the equipment execution layer; the equipment execution layer includes at least one lurking automated guided vehicle and at least one bin handling robot;

[0048] S2: Based on real-time task queue status data and real-time equipment status data, dynamically calculate and output the current optimal material classification threshold or rule parameters; material classification includes at least a first category of materials suitable for overall handling and a second category of materials suitable for independent handling;

[0049] S3: Based on the material characteristic data and the current optimal rule parameters, classify the target material into either the first category or the second category of materials;

[0050] S4: Based on real-time equipment status data and the category of target materials, dynamically assemble and generate corresponding integrated operation strategies from the pre-set strategy component library; wherein, the strategy assembled for the first type of materials specifies at least a hidden automated guided vehicle as the core handling equipment and a mobile shelf as the handling unit; the strategy assembled for the second type of materials specifies at least a bin handling robot as the core handling equipment and a standardized bin as the handling unit.

[0051] S5: Parse the integrated operation strategy into a specific set of device driver instructions;

[0052] S6: Send the device drive instruction set to the corresponding handling device in the device execution layer to execute the operation;

[0053] S7: Collect the status change data generated when the equipment execution layer performs the operation, and feed back the status change data and update it to the warehouse dynamic data, so as to re-execute the steps starting from S2 based on the updated warehouse dynamic data, forming a closed-loop scheduling.

[0054] As can be seen from the technical solutions provided by the present invention above, the beneficial effects of the warehouse classification management system and method provided by the present invention are:

[0055] Improving overall warehouse operation efficiency: The system acquires multi-dimensional real-time warehouse data through a dynamic feature perception module, and the adaptive decision engine dynamically optimizes material classification rules accordingly; it matches dedicated handling equipment and operation strategies for different categories of materials. The first category of materials is efficiently handled as a whole using a hidden automated guided vehicle and mobile shelves, while the second category of materials is accurately and independently picked up and placed using a bin handling robot and standardized bins; the differentiated operation mode adapts to the handling needs of various materials, effectively shortening the operation cycle and increasing warehouse throughput;

[0056] Improving equipment resource utilization efficiency: The rule optimization unit accurately assesses the warehouse operation status and optimizes classification parameters by calculating quantitative indicators such as task urgency and equipment load balance; the system prioritizes scheduling low-load, high-power equipment to perform operations, avoiding situations where some equipment is overloaded while others are idle; continuous optimization of equipment load balance improves the overall utilization rate of the steerable automated guided vehicles and bin handling robots, reducing equipment resource waste;

[0057] To achieve intelligent and automated warehousing operations: The system constructs a closed-loop scheduling process that includes data collection, rule optimization, strategy execution, feedback, and iteration. The entire process can be completed without human intervention, including material classification, equipment assignment, instruction issuance, and operation monitoring. The automated operation mode reduces errors caused by manual sorting and handling, and the intelligent dynamic decision-making mechanism enables the system to adjust its strategies in real time according to changes in the operational situation, improving operational accuracy and stability.

[0058] Enhanced system flexibility and scalability: The strategy generation unit adopts a component-based architecture, and various components in the strategy component library support adding, deleting, and modifying operations; when new equipment types or material categories are added to the warehousing scenario, only the corresponding equipment scheduling components or classification rules need to be added to the strategy component library, without reconstructing the entire system; the component-based design enables the system to quickly adapt to different warehousing management needs;

[0059] Reduce warehouse operation and management costs: The equipment execution layer collects equipment status data in real time and feeds it back to the dynamic feature perception module. The system can detect equipment failures in a timely manner and trigger alarms; real-time fault warning and accurate fault location reduce equipment fault troubleshooting time and downtime; automated operation mode reduces reliance on manual labor, reduces manual management costs, and improves the economic efficiency of warehouse operation and maintenance. Attached Figure Description

[0060] Fig. 1 This is a schematic diagram of the structure of a warehouse classification management system according to the present invention;

[0061] Fig. 2 This is a schematic diagram of the steps in a warehouse classification management method according to the present invention. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0063] To better understand the above technical solutions, the following will provide a detailed description of the technical solutions in conjunction with the accompanying drawings and specific embodiments.

[0064] like Figs. 1-2 As shown, an embodiment of the present invention provides a warehouse classification management system, including:

[0065] The dynamic feature perception module is used to acquire and update warehouse dynamic data in real time. The warehouse dynamic data includes at least: material feature data, in-transit task queue status data, and real-time status data of various handling equipment in the equipment execution layer.

[0066] An adaptive decision engine communicates with the dynamic feature perception module; the adaptive decision engine includes:

[0067] The rule optimization unit is used to dynamically calculate and output the current optimal material classification threshold or rule parameters based on real-time task queue status data and real-time equipment status data; the material classification includes at least a first category of materials suitable for overall handling and a second category of materials suitable for independent handling.

[0068] The strategy generation unit is used to classify target materials into corresponding categories based on material characteristic data and the current optimal rule parameters output by the rule optimization unit, and dynamically assemble and generate corresponding integrated operation strategies from a pre-set strategy component library based on real-time equipment status data; wherein, the strategy for assembling the first type of materials specifies at least a hidden automated guided vehicle as the core handling equipment and a mobile shelf as the handling unit; the strategy for assembling the second type of materials specifies at least a bin handling robot as the core handling equipment and a standardized bin as the handling unit.

[0069] The elastic scheduling execution module communicates with the adaptive decision engine and is used to parse the integrated operation strategy into a specific set of equipment drive instructions and send them to the equipment execution layer; the equipment execution layer includes at least one lurking automated guided vehicle and at least one bin handling robot.

[0070] The state changes generated during the execution of tasks by the equipment execution layer are fed back to the dynamic feature perception module in real time, forming a closed-loop scheduling.

[0071] In this embodiment, the dynamic feature perception module is the data core of the warehouse classification management system. It provides high-precision and non-redundant dynamic warehouse data for the adaptive decision engine through real-time data collection, cleaning, verification, fusion and incremental updates in multiple dimensions and with multiple means. It is the basic support for realizing intelligent warehouse classification and elastic scheduling.

[0072] The dynamic feature perception module is primarily responsible for capturing three core types of data throughout the warehousing operation process: material characteristic data, in-transit task queue status data, and real-time status data of various handling equipment at the equipment execution layer. This module, through the collaborative operation of five sub-units, completes the entire process from raw data capture to standardized data output. The processing includes raw data acquisition, data preprocessing, data verification and fusion, dynamic incremental updates, and standardized data push. The final warehousing dynamic data output by the module possesses three key characteristics: real-time performance, consistency, and accuracy. This data directly supports the rule optimization and strategy generation units of the adaptive decision engine. Simultaneously, the module can receive real-time operation feedback data from the equipment execution layer, enabling continuous iterative updates of warehousing dynamic data and providing a stable data link for the system's closed-loop scheduling. The dynamic feature perception module includes:

[0073] Raw data acquisition unit:

[0074] Multi-type sensor data capture: Deploy multiple types of sensor groups to cover the entire warehousing scenario; RFID sensors are used to collect material identification, specification, and storage location information; position sensors are used to collect the real-time position coordinates of steerable guided vehicles and bin handling robots; status sensors are used to collect equipment power load and operating status information; sensors collect data in real time at preset intervals to ensure the timeliness of material and equipment status data;

[0075] Data retrieval from warehouse management system interface: Connects to the warehouse management system through a standardized interface; periodically retrieves information on the task queue in progress, including task number, task type, task priority, and task execution progress; the retrieval frequency can be dynamically adjusted according to the busyness of warehouse operations, shortening the retrieval interval during peak operations and extending the retrieval interval during off-peak operations to balance data timeliness and system resource consumption.

[0076] Multi-source data aggregation and transmission: Real-time data collected by sensors and business data pulled from system interfaces are aggregated; the aggregated data is stored in a local cache in a unified temporary format; when the data in the cache reaches a preset threshold or a preset time node, the data transmission process is automatically triggered, and the aggregated data is pushed to the data preprocessing unit.

[0077] Data preprocessing unit:

[0078] Data cleaning:

[0079] Outlier handling: The 3σ principle is used to identify and handle outliers in numerical data; when a data value exceeds the mean plus or minus three standard deviations, it is judged as an outlier and removed; for non-numerical data, such as equipment status information, it is verified by a preset list of valid statuses and invalid data not in the list is removed.

[0080] Missing value handling: For missing fields in material characteristic data, fill them with the average value of materials of the same category; for missing fields in equipment status data, fill them with valid data from the previous collection period; for missing fields in task queue data, mark them as invalid tasks and push them to the warehouse management system for confirmation.

[0081] Duplicate value handling: Duplicate value determination is based on the unique identifier field of the data; material data uses the material identifier as the unique identifier, equipment data uses the equipment number as the unique identifier, and task data uses the task number as the unique identifier; the most recently collected data is retained and the rest of the duplicate data is deleted;

[0082] Data formatting:

[0083] Data structure standardization: The cleaned heterogeneous data is converted into a unified structured data format; after structuring, the material characteristic data includes five major fields: material identifier, specification parameters, storage location, inventory quantity, and material type; after structuring, the task queue data includes five major fields: task identifier, task type, task priority, execution device, and task status; after structuring, the equipment status data includes five major fields: equipment number, equipment type, real-time location, remaining power, and load status.

[0084] Data format standardization: Standardize the field types of all structured data; use floating-point or integer for numeric fields, fixed-length strings for character fields, and timestamp format for date and time fields; the standardized dataset is the formatted warehouse data.

[0085] Preprocessed data verification and transmission: Perform field integrity verification on the formatted warehouse data; ensure that the core fields of each data are not missing, and push the data that passes the verification to the data verification and fusion unit; data that fails the verification is marked as invalid data and stored in the abnormal data log database for subsequent manual investigation and analysis;

[0086] Data verification and fusion unit:

[0087] Logical consistency check:

[0088] Internal data logic validation: Validate the logical relationships between fields within a single data entry; material inventory quantity cannot be negative, equipment load status and task execution status must be consistent, and task priority must conform to the preset priority level range; data with logical contradictions are marked as abnormal data and removed from the formatted dataset.

[0089] Data cross-logic verification: Verify the logical associations between different types of datasets; the executing devices in the task queue must exist in the device status dataset, and the storage location in the material characteristic data must match the reachable location in the device location data; if data with contradictory cross-logic is found, it is synchronously pushed to the abnormal data log library.

[0090] Multi-source data fusion:

[0091] Same-dimensional data weighted fusion: Weighted fusion of multi-source data for the same monitoring dimension; taking equipment location data as an example, the weight of real-time location data collected by location sensors is set to 0.7, and the weight of historical equipment location data recorded by the warehouse management system is set to 0.3; the final equipment location data is obtained through weighted calculation, and the calculation formula is: ,in, For the final location data of the device, For sensor location data weights, Location data collected by the sensor, Weights for historical location data in the system. Historical location data recorded by the system;

[0092] Heterogeneous data association and fusion: Establish association indexes between different types of datasets; associate material feature data and task queue data with material identifier as index, and associate equipment status data and task queue data with equipment number as index; the data after association and fusion can fully present the correspondence between materials, tasks and equipment;

[0093] Data fusion output: Data that has passed logical consistency verification and completed multi-source fusion is defined as fused and verified warehouse dynamic data; this data is pushed to the dynamic update unit for further processing according to the preset batch size;

[0094] Dynamically updated unit:

[0095] Incremental data reception and identification: Two types of incremental data are received in real time; the first type is operation feedback data from the equipment execution layer, including equipment location update data, task execution progress data, and equipment fault alarm data; the second type is external system instruction data, including task addition instructions, task cancellation instructions, and equipment scheduling instructions issued by the warehouse management system; the unit identifies the type of the received incremental data and marks the category to which the data belongs and the associated core identifier.

[0096] Incremental data and basic data fusion: An incremental update algorithm is used to fuse incremental data with the fusion-verified warehouse dynamic data; newly added data is directly appended to the corresponding dataset; updated data is overwritten with the corresponding field in the original data based on the core identifier; deleted data is marked as invalid based on the core identifier; the entire fusion process only operates on the data that has changed and does not touch the original data that has not changed.

[0097] Data version management: A unique version number is assigned to the warehouse dynamic data after each incremental update; the version number includes two main elements: update time and update batch; the system automatically records the changes in each version of data, including the values ​​of the changed fields before and after the change; all versions of data are stored in the historical database, supporting data backtracking and querying by version number;

[0098] Final data output: The warehouse dynamic data that has completed incremental updates and version marking is defined as the final updated warehouse dynamic data; this data is pushed to the data output unit in real time;

[0099] Data output unit:

[0100] Data format adaptation: Based on the interface requirements of the adaptive decision engine, the format of the finally updated warehouse dynamic data is converted; the task urgency-related data required by the rule optimization unit is converted into a simplified indicator-type data format; the detailed material and equipment data required by the strategy generation unit is converted into a complete detail-type data format; the format conversion process strictly follows the preset field mapping relationship to ensure that the data fields are fully matched with the engine requirements;

[0101] Real-time data push: Data is pushed to the adaptive decision engine using a publish-subscribe model; the module acts as the data publisher, and the adaptive decision engine acts as the data subscriber; when the warehouse dynamic data is updated, the module immediately pushes a data update notification to the engine; after receiving the notification, the engine pulls the data in the corresponding format as needed; a heartbeat mechanism is used to maintain the communication link during the push process to ensure the stability of data transmission;

[0102] Push status monitoring and feedback: Real-time monitoring of data push status, including three statuses: push success, push failure, and push timeout; successfully pushed data records the push time and engine confirmation information; data that fails to push or times out automatically triggers a re-push mechanism, and data that fails to push after three re-pushes is marked as abnormal push data and recorded in the push log library; push status information is synchronously fed back to the dynamic update unit for optimizing subsequent data update strategies.

[0103] Furthermore, the multi-source heterogeneous data acquisition technology is based on distributed sensor networks and system interface integration technology. The distributed sensor network consists of RFID sensors, position sensors, and status sensors. Various sensors collect data from different dimensions according to their respective functions, and the data is transmitted in real time through wireless sensing protocols. The system interface integration technology adopts standardized interfaces to connect with the warehouse management system. The interface supports two modes: periodic data retrieval and event-triggered data push. The data acquired by the two acquisition methods have differences in structure and format. The core technology lies in achieving preliminary aggregation of heterogeneous data through a unified caching mechanism, laying the foundation for subsequent preprocessing operations.

[0104] Structured data cleaning technology is based on statistical methods and rule-based validation methods. Statistical methods are mainly used to handle outliers in numerical data, with the 3σ principle being the core algorithm. This principle assumes that the data follows a normal distribution, and data exceeding the mean plus or minus three standard deviations are considered outliers. Rule-based validation methods are mainly used to handle non-numerical data and data logical contradictions. By using a pre-defined list of valid values ​​and logical relationship rules, invalid data is identified and removed. The core of the technology lies in using differentiated cleaning strategies for different types of data to maximize data accuracy while ensuring data integrity.

[0105] Multi-source data fusion technology is based on weighted fusion algorithms and correlation indexing algorithms. The weighted fusion algorithm assigns different weight coefficients to multi-source data of the same dimension according to the reliability of the data source; the more reliable the data source, the larger the weight coefficient, and finally obtains the optimal data for that dimension through weighted calculation. The correlation indexing algorithm establishes the relationship between data based on core identification fields for heterogeneous data of different dimensions. The core identification fields include material identification, equipment number, and task number. The core technology lies in eliminating redundancy and conflicts between multi-source data through fusion operations, forming a unified data view that can comprehensively reflect the warehouse operation status.

[0106] Incremental data update technology is based on a difference detection algorithm. The algorithm first compares the core identifier fields of incremental data and basic data to identify three types of data changes: addition, update, and deletion. Differentiated update operations are performed for different types of changes to avoid the system resource consumption caused by full data updates. The core technology is to process only the data that has changed, which ensures the real-time nature of the data and improves the efficiency of data updates. At the same time, it is combined with version management technology to achieve full lifecycle traceability of data.

[0107] The publish-subscribe data push technology is based on a message queue mechanism. The module acts as the publisher, publishing data update events to the message queue. The adaptive decision engine acts as the subscriber, listening for update events in the message queue. When an update event occurs, the engine pulls the corresponding data on demand. The heartbeat mechanism is crucial for maintaining the communication link, with the module and engine periodically sending heartbeat packets to confirm each other's status. The core technology lies in realizing on-demand data push, reducing meaningless data transmission, and lowering the system communication load.

[0108] In this embodiment, the adaptive decision engine is the intelligent decision-making core of the warehouse classification management system. It receives real-time data input from the dynamic feature perception module and outputs material classification rules and integrated operation strategies that are adapted to the real-time operation status of the warehouse through hierarchical calculations of the rule optimization unit and the strategy generation unit. It is the key hub connecting data perception and instruction execution and directly determines the efficiency of warehouse operations and the rationality of resource allocation.

[0109] The core function of the adaptive decision engine is to dynamically optimize material classification rules and intelligently generate operational strategies. It receives material characteristic data, in-transit task queue status data, and real-time equipment execution status data pushed by the dynamic feature perception module. First, the rule optimization unit calculates the optimal material classification threshold or rule parameters based on task urgency and equipment load status, classifying materials into two categories: Category 1 materials suitable for overall handling and Category 2 materials suitable for independent handling. Then, the strategy generation unit accurately matches components from a pre-set strategy component library and assembles them into an integrated operational strategy based on the material classification results and real-time equipment status. This engine possesses data-driven self-optimization capabilities; its decision results dynamically adjust according to the warehousing operation status. It also supports closed-loop system scheduling, continuously iterating and optimizing the decision model by receiving equipment execution feedback data, thus achieving intelligent full-process control of warehousing operations. The adaptive decision engine includes:

[0110] Rule optimization unit:

[0111] The rule optimization unit is the core of the adaptive decision engine's rule output. Its core objective is to output material classification rules that maximize operational efficiency based on real-time warehouse operation data. The specific execution process is detailed below:

[0112] Data Analysis and Quantification:

[0113] The system receives real-time status data of the in-transit task queue and real-time status data of the device execution layer from the dynamic feature perception module, and performs structured parsing and numerical quantization on the two types of data.

[0114] Based on the task queue status data, core fields such as task priority, remaining execution time, and quantity of associated materials are extracted to calculate the task urgency index. The calculation formula is as follows: ,in, For the first The urgency value of each task. For the first The priority coefficient of each task. The system's preset maximum task execution time. For the first The remaining execution time for each task;

[0115] Based on the equipment status data, core fields such as the current number of tasks, maximum capacity of tasks, and real-time power consumption of each handling device are extracted. The equipment load balancing index is then calculated using the following formula: ,in, This represents the equipment load balancing value. For the first The current number of tasks on the device. The average number of tasks across all devices. Total number of devices;

[0116] Based on task urgency indicators and equipment load balancing indicators, a quantitative indicator system is constructed to provide a quantitative basis for warehouse status assessment.

[0117] Warehouse status assessment:

[0118] Based on the established quantitative indicator system and combined with preset weights, the overall warehouse operation efficiency score is calculated using the following formula: ,in, To score overall efficiency, The weight for the task urgency indicator is 0.6. The average urgency value for all tasks. This represents the weight of the equipment load balancing index, with a value of 0.4. This represents the equipment load balancing value.

[0119] Simultaneously assess the real-time capacity margin of each handling device. The capacity margin calculation formula is as follows: ,in, For the first Capacity margin of the equipment For the first The maximum load of the equipment. For the first The current load of the device;

[0120] The final result is a warehouse status assessment that includes a comprehensive efficiency score, equipment capacity margin distribution, and task urgency distribution.

[0121] Candidate parameter extraction:

[0122] Based on the warehouse status assessment results, multiple sets of candidate classification rule parameters are extracted from the preset parameter space; the parameter space includes the core thresholds for material classification, such as the minimum volume threshold, minimum weight threshold, and minimum order batch threshold for the first type of material;

[0123] When the overall efficiency score is lower than the preset threshold, candidate parameters that are biased towards improving equipment utilization are extracted, such as lowering the volume threshold of the first type of material and increasing the total number of materials handled.

[0124] When the load balancing of the equipment is lower than the preset threshold, candidate parameters that bias towards balancing the load of the equipment are extracted. For example, the classification threshold is adjusted to tilt the workload towards the low-load equipment. After extraction, a set of candidate rule parameters is formed. The size of the set is controlled between 5 and 10 groups to ensure the efficiency of subsequent inference.

[0125] Optimization and Selection:

[0126] The pre-defined reinforcement learning optimization decision model is invoked to simulate and deduce each set of parameters in the candidate rule parameter set. The model builds a simulation environment based on historical operation data. After inputting candidate parameters, it simulates the material classification process and operation execution process, and outputs the corresponding efficiency indicators such as operation completion time, equipment utilization rate, and task delay rate for each set of parameters.

[0127] The comprehensive benefit value for each set of parameters is calculated using the following formula: ,in, For comprehensive benefit value, Assigning weights based on task time. To simulate the task completion time, The baseline task completion time, As a weight for equipment utilization rate, To simulate the average utilization rate of equipment, As a weight for the on-time rate of the task, To simulate task delay rates;

[0128] Select the candidate parameter with the highest comprehensive benefit value as the current optimal material classification threshold or rule parameter;

[0129] Parameter encapsulation and archiving:

[0130] The optimal material classification threshold or rule parameters are encapsulated in a preset format. The encapsulated content includes information such as parameter name, parameter value, applicable storage conditions, and effective time, and is output to the strategy generation unit.

[0131] At the same time, the quantitative indicator system, warehouse status assessment results, candidate parameter set, and comprehensive benefit value calculation process of this decision will be archived into the historical rule base. The data in the historical rule base will be used to optimize the iterative training of the decision model. The iteration cycle will be triggered after the system has accumulated 100 hours of operation or processed 1000 tasks.

[0132] Strategy generation unit:

[0133] The strategy generation unit is the core of the adaptive decision engine's strategy output. Its core objective is to generate an integrated operation strategy that can be directly executed based on the optimal classification rules and the real-time status of the equipment. The specific execution process is detailed as follows:

[0134] Material classification and identification:

[0135] Receive material feature data output from the dynamic feature perception module, including information such as material volume, weight, order attribution, and storage location, and combine it with the optimal classification parameters output by the rule optimization unit to perform material classification and discrimination;

[0136] When the volume of a material exceeds the volume threshold of the first category of materials, the weight exceeds the weight threshold, and the order batch exceeds the batch threshold, it is classified into the first category of material identifier set.

[0137] Materials that do not meet any of the above conditions are classified into the second category of material identifier set; after classification, a material classification result list is generated, which includes information such as material identifier, material category, order to which it belongs, and storage location, and simultaneously marks high-urgency task-related materials that need to be processed first;

[0138] Equipment status extraction:

[0139] The system receives real-time device status data from the dynamic feature perception module, performs in-depth analysis on the data, and extracts the core status indicators for two types of devices:

[0140] For the Automated Guided Vehicle (AGV), real-time location coordinates, remaining battery percentage, current task count, current load, last maintenance time, and fault records are extracted to calculate the equipment's task load rate. The formula is as follows: ,in, For the first The workload rate of the tandem automated guided vehicle. Number of current tasks This represents the maximum number of tasks that can be carried.

[0141] For a bin handling robot, extract indicators such as real-time availability, remaining battery percentage, current load, gripping mechanism status, and path unobstructedness to calculate the equipment load utilization rate. The formula is as follows: ,in, For the first Load utilization rate of the tabletop material handling robot For the current load, Maximum load;

[0142] Based on the above indicators, a structured equipment status description set is generated, clearly marking the availability and operational capabilities of each piece of equipment;

[0143] Strategy component matching:

[0144] Based on the material classification results and the structured equipment status description set, suitable strategy components are retrieved from the strategy component library. Component matching follows the principles of capability adaptation and efficiency priority.

[0145] The strategy components for matching the first type of materials include a hidden automated guided vehicle (AGV) scheduling component, a mobile rack docking path planning component, an overall handling task sequence generation component, and a multi-material stacking optimization component. The hidden AGV scheduling component prioritizes matching equipment with a task load rate of less than 30% and a battery level of more than 50%. The mobile rack docking path planning component plans the shortest docking path based on the real-time location of the equipment and the storage location of the materials.

[0146] The strategy components for matching the second type of materials include a bin handling robot assignment component, a standardized bin storage and retrieval operation component, an independent pick-and-place task sequence generation component, and a path conflict avoidance component. The bin handling robot assignment component prioritizes matching equipment with a load utilization rate of less than 20% and a path smoothness of more than 80%. The standardized bin storage and retrieval operation component matches the optimal bin size according to the material size.

[0147] Initial strategy assembly:

[0148] The matched strategy components are combined and parameter-bound according to a predefined logical architecture, which follows the process of task decomposition, device assignment, path planning, and timing orchestration.

[0149] The task decomposition component breaks down the handling tasks of the first type of materials into four sub-tasks: mobile shelf retrieval, material loading, material transportation, and material unloading; and breaks down the handling tasks of the second type of materials into four sub-tasks: bin retrieval, material grabbing, bin transportation, and material placement.

[0150] The equipment assignment component assigns specific equipment to each subtask based on the equipment status description set and binds the equipment's operating parameters, such as the driving speed of the tractor-guided vehicle and the gripping force of the bin handling robot.

[0151] The path planning component plans the operation path for each device and marks the key nodes and traffic rules in the path;

[0152] The timing orchestration component allocates execution time windows to subtasks to prevent different devices from working simultaneously in the same area; after combination, it generates an initial integrated job strategy, which includes four core parts: task list, device list, path list, and timing list.

[0153] Conflict detection and optimization:

[0154] Based on the latest structured device status description set, conflict detection is performed on the initial integration operation strategy. The detection types include three categories: device resource conflicts, path conflicts, and timing conflicts.

[0155] Device resource conflict detection: Check if the same device is assigned multiple tasks with overlapping time windows;

[0156] Path conflict detection: Checks whether the planned paths of multiple devices overlap within the same time period;

[0157] Timing conflict detection: Check if the execution time window of any subtask exceeds the available time of the device; for detected conflicts, perform resolution optimization:

[0158] Equipment resource conflict: Adjust the task timing window or reassign backup equipment;

[0159] Path conflict: Adjust the device's path planning or execution order to prioritize ensuring smooth paths for high-urgency tasks;

[0160] Timing conflicts: Extend the time window of related tasks or split the tasks; after optimization, perform conflict detection again until there are no conflicts, and output the final executable integrated job strategy;

[0161] Strategy encapsulation and output:

[0162] The final executable integrated operation strategy is encapsulated in the Industrial Internet of Things (IIoT) protocol format. The encapsulated content includes information such as strategy number, generation time, execution validity period, and device instruction set index.

[0163] The policy is pushed to the elastic scheduling execution module through a dedicated communication interface, and a policy backup file is generated and stored in the system database. The backup file is retained for 7 days for job traceability and anomaly analysis.

[0164] Furthermore, the dynamic rule optimization technology is based on the theoretical construction of the integration of reinforcement learning and simulation inference; the core of the technology is the reinforcement learning optimization decision model, which uses the overall efficiency score of warehousing and the equipment load balance as the reward function, the material classification parameters as the action space, and the real-time status data of warehousing as the state space.

[0165] The model is pre-trained using historical operational data, and the parameter selection strategy is continuously adjusted during the training process to maximize the reward function value of the model output parameters. In actual operation, the model combines real-time warehouse conditions and simulation results to dynamically adjust the parameter selection logic, thereby achieving real-time adaptation between classification rules and warehouse conditions.

[0166] Archived data from the historical rule base is used regularly for incremental training of the model. After each incremental training, the model's parameter prediction accuracy can be improved by 3% to 5%, ensuring the long-term effectiveness of the technology.

[0167] The dynamic assembly technology of strategy components is based on a component-based architecture and a state matching algorithm; the core of the technology is the construction of the strategy component library and the design of the state matching algorithm.

[0168] The strategy component library adopts a modular design, breaking down the work processes of different types of materials into reusable atomic components. Each component contains three core parts: input parameters, execution logic, and output results, and supports adding, deleting, and modifying components.

[0169] The state matching algorithm calculates the matching degree between the material classification result and the component based on cosine similarity. The calculation formula is as follows: ,in, This is the matching degree value. This is the feature vector of the material classification result. For the component's adaptation feature vector, The number of feature dimensions;

[0170] The algorithm prioritizes components with a matching degree higher than 0.8 for assembly to ensure the adaptability of the strategy. This technology supports the expansion of warehousing equipment and material types. When a new equipment type is added, only the corresponding equipment scheduling component needs to be added, without reconstructing the entire strategy generation logic.

[0171] The workflow of the adaptive decision engine is as follows:

[0172] Initialization phase:

[0173] After the adaptive decision engine starts, it first performs a hardware self-test to check whether the communication interface with the dynamic feature perception module and the elastic scheduling execution module is smooth, and to check the integrity of the historical rule base and the strategy component base.

[0174] Load the pre-trained parameters of the reinforcement learning optimization decision model, load the latest version of the policy component library, and load the initial threshold of the material classification parameter space;

[0175] Establish a publish-subscribe communication link with the dynamic feature perception module to subscribe to three types of core data: material feature data, task queue status data, and equipment status data; establish an instruction issuance communication link with the elastic scheduling execution module to configure the protocol format and frequency of instruction push.

[0176] Data receiving stage:

[0177] It receives data pushed by the dynamic feature perception module in real time and performs integrity verification on the data. The verification includes whether the data fields are complete, whether the data format is correct, and whether the data values ​​are within a reasonable range.

[0178] Data that passes verification is stored in the engine's local cache. The cache uses a first-in, first-out storage strategy and has a capacity of 1,000 data records to prevent data overflow.

[0179] Data that fails the verification is marked as abnormal data, the reason for the abnormality is recorded and pushed to the system operation and maintenance platform, and the retransmission mechanism of the dynamic feature perception module is triggered at the same time.

[0180] Rule optimization phase:

[0181] When the cached data reaches a preset threshold or a preset time interval, the rule optimization process is triggered. The rule optimization unit extracts task queue status data and device status data from the cache.

[0182] The system sequentially performs data parsing and quantification, warehouse status assessment, candidate parameter extraction, optimization deduction and selection, parameter encapsulation and archiving operations, and outputs the current optimal material classification parameters.

[0183] Synchronize the optimal parameters to the policy generation unit and update the local parameter cache;

[0184] Strategy generation phase:

[0185] The strategy generation unit extracts material feature data from the cache, combines it with the optimal classification parameters to perform material classification and generate material classification results.

[0186] Extract equipment status data to generate a structured equipment status description set, and perform strategy component matching, initial strategy assembly, conflict detection and optimization operations based on classification results and equipment status description set;

[0187] The final integrated job strategy is encapsulated and pushed to the elastic scheduling execution module, and the strategy generation log is recorded.

[0188] Iterative learning phase:

[0189] Receive equipment operation execution feedback data pushed by the dynamic feature perception module, including indicators such as task completion time, equipment utilization rate, and task delay rate;

[0190] The feedback data is correlated with the parameters and strategies of this decision to calculate the actual benefit value of the decision;

[0191] When the cumulative number of associated data reaches 100 sets, incremental training of the optimized decision-making model is triggered. After training is completed, the model parameters are updated to improve the accuracy of subsequent decisions.

[0192] End phase:

[0193] When a system stop command is received, the adaptive decision engine stops receiving data and performing decision operations.

[0194] Save the current model parameters, strategy component library configuration, decision logs, and other data, and close the communication link with other modules;

[0195] Generate a module operation report, which includes information such as runtime, number of decisions, average strategy generation time, and job efficiency improvement rate, and push it to the system operation and maintenance platform.

[0196] In this embodiment, the elastic scheduling execution module is the instruction execution center of the warehouse classification management system. It transforms the abstract strategy into an executable driving instruction for the equipment by parsing, converting, timing, and distributing the integrated operation strategy, ensuring that the warehouse handling equipment completes the operation task accurately and efficiently. At the same time, it provides reliable data support for the closed-loop scheduling of the system through real-time feedback of the instruction execution status.

[0197] The elastic scheduling execution module is primarily responsible for receiving the integrated operation strategy output by the adaptive decision engine. Through a series of operations such as parsing, decomposing, transforming, adapting, and orchestrating the timing sequence, it generates a set of drive instructions matching the target equipment type. This module sequentially sends the instructions to the equipment execution layer according to a preset timing sequence and monitors the receipt and execution status of the instructions in real time. Finally, the module generates a structured instruction execution log, which is synchronously pushed to the dynamic feature perception module to ensure the collaborative and efficient operation of warehouse handling equipment. It also supports the closed-loop scheduling optimization of the system. The elastic scheduling execution module parses the integrated operation strategy into a specific set of equipment drive instructions and sends it to the equipment execution layer, including:

[0198] Instruction parsing and decomposition unit:

[0199] Policy reception and parsing: The system receives the integrated operation policy output by the policy generation unit in real time and performs structured parsing of the policy content according to the preset protocol format. The parsing process decomposes the policy's hierarchical architecture layer by layer to ensure that no core information is missed.

[0200] Core element extraction: From the parsed strategy content, three core elements are accurately extracted: equipment assignment information, job sequence information, and control parameter information. Equipment assignment information clarifies the type and number of the target equipment, job sequence information determines the execution steps and order of the tasks, and control parameter information includes key parameters such as the speed and load of the equipment.

[0201] Element structured integration: The three types of core elements extracted are integrated in a unified format to generate parsed strategy elements; the integrated element set has a clear mapping relationship and can directly provide data support for subsequent instruction conversion;

[0202] Instruction conversion and adaptation unit:

[0203] Equipment type identification: Based on the equipment assignment information in the strategy elements output by the instruction parsing and disassembly unit, the specific type of the target equipment is identified; the identification objects mainly include two types of core handling equipment: hidden automated guided vehicles and bin handling robots.

[0204] Command template matching: Retrieve the preset device command template library and match the corresponding command template according to the identified device type; command templates for different types of devices differ in terms of field definition parameter format, etc., and the matching process ensures that the template and device type correspond completely;

[0205] Command draft generation: Substitute the job sequence information and control parameter information from the strategy elements into the matched device command template to complete the parameter filling and format conversion; after conversion, generate a set of drive command drafts that match the target device type;

[0206] Timing and verification unit:

[0207] Real-time device status acquisition: The latest real-time device status data is retrieved from the dynamic feature perception module. The data covers key information such as device location, power, task load, etc. This data is an important basis for timing orchestration and can ensure the feasibility of command execution.

[0208] Instruction timing orchestration: Based on the driver instruction draft set and real-time device status data, the execution timing of each instruction is planned and orchestrated; the orchestration process follows the principles of efficiency priority and conflict avoidance, and allocates reasonable time windows for instructions of different devices;

[0209] Conflict pre-detection and optimization: Conflict detection is performed on the orchestrated instruction set. The detection types include two categories: device resource conflicts and path conflicts. Once a conflict is found, the execution timing of the instructions is immediately adjusted or device resources are reallocated to complete the optimization of the instruction set and generate a timing-arranged instruction set.

[0210] Instruction distribution and synchronization unit:

[0211] Timing rule confirmation: Double-check the execution sequence of each instruction in the instruction set after timing arrangement to confirm the rationality and accuracy of the timing plan; the confirmation process ensures that the order of instruction issuance is completely consistent with the logical order of equipment operation;

[0212] Commands are issued sequentially: Through the preset device communication interface, each drive command is sequentially issued to the corresponding target handling device in the device execution layer according to the execution sequence of the commands; the issuance process adopts a point-to-point communication method to ensure the accuracy of command transmission;

[0213] Distribution status synchronization: Records the distribution status of instructions in real time, including different statuses such as issued and pending issuance and issuance failure; the distribution status will be synchronized to the subsequent instruction confirmation and monitoring unit to provide basic data for instruction monitoring;

[0214] Command confirmation and monitoring unit:

[0215] Acknowledgment signal monitoring: After the command is issued, the system monitors the command reception acknowledgment signal returned by the target device in real time; if no acknowledgment signal is received within a preset time, the command is deemed to have failed to be issued, and the command retransmission mechanism is triggered.

[0216] Initial execution status tracking: Receives the initial execution status signal returned by the device and tracks whether the device starts the operation according to the instructions; the status signal includes information such as successful or failed device startup, and the tracking process ensures timely detection of device execution abnormalities;

[0217] Reporting of abnormal situations: When a failure to receive a command or an abnormality in the execution of a device is detected, the abnormal information shall be immediately reported to the system operation and maintenance platform; at the same time, the issuance of subsequent commands to the device shall be suspended and the operation shall be resumed after the abnormality is resolved.

[0218] Execution log generation unit:

[0219] Log data collection: Collect all data during the instruction issuance and execution process, including key information such as instruction content, instruction issuance time, receipt confirmation time, and initial execution status; the collection process ensures data integrity and timeliness;

[0220] Structured log generation: The collected log data is integrated according to a preset format to generate structured instruction execution logs; structured logs have unified field definitions, which facilitates subsequent querying and analysis;

[0221] Log synchronization push: The generated structured instruction execution logs are pushed to the dynamic feature perception module in real time; after the push is completed, the log data will be incorporated into the warehouse dynamic data system to provide data support for the closed-loop scheduling of the system;

[0222] Furthermore, the instruction template matching technology is based on a modular design concept, constructing an instruction template library covering different device types. Each template in the library is customized for the communication protocol and control logic of a specific device. The core technology is to achieve rapid conversion of policy elements into device instructions through accurate matching of device type identification results and template attributes. This technology can effectively reduce the complexity of instruction conversion and improve the efficiency and accuracy of instruction generation.

[0223] The timing conflict prediction technology is based on real-time equipment status data and instruction execution logic to build a conflict detection model. The model compares and analyzes the instruction execution time window and the equipment availability time window to identify potential resource conflicts and path conflicts. The core technology is to resolve conflicts through timing adjustment and resource reallocation to ensure the coordination of multi-device operations. This technology can effectively avoid interference during equipment operation and improve the overall efficiency of warehousing operations.

[0224] The instruction synchronization distribution technology is based on the Industrial Internet of Things (IIoT) communication protocol and constructs a point-to-point instruction transmission link. The core technology is to issue instructions sequentially according to a preset time sequence and ensure the reliability of instruction transmission through a reception confirmation mechanism. When an instruction issuance fails, the technology will trigger a retransmission mechanism and adjust the retransmission strategy to adapt to changes in the communication environment. This technology can ensure the accuracy and timeliness of instruction issuance and provide support for the orderly operation of equipment.

[0225] Furthermore, the state changes generated during the execution of tasks by the equipment execution layer are fed back to the dynamic feature perception module in real time, forming a closed-loop scheduling; including:

[0226] The status change acquisition unit is used to collect status change data generated by various handling equipment in the equipment execution layer during the execution of operations in real time. The status change data includes at least equipment position update data, task execution progress data, and equipment fault alarm data.

[0227] The state data preprocessing unit is connected in communication with the state change acquisition unit. It is used to clean, timestamp align and format standardize the state change data to generate preprocessed state data.

[0228] The feedback data encapsulation unit is communicatively connected to the state data preprocessing unit and is used to encapsulate the preprocessed state data into a feedback data packet with a unified protocol format.

[0229] The feedback transmission unit is communicatively connected to the feedback data encapsulation unit and is used to transmit the feedback data packet to the dynamic feature sensing module in real time via a wireless communication network.

[0230] The dynamic data update unit is connected to the feedback transmission unit and is used to receive feedback data packets in the dynamic feature perception module, parse them, and then merge and incrementally update them with the current warehouse dynamic data to generate real-time updated warehouse dynamic data.

[0231] The closed-loop triggering unit communicates with the dynamic data update unit. When real-time updated warehouse dynamic data is generated, it automatically triggers the adaptive decision engine to re-execute rule optimization and strategy generation based on the real-time updated warehouse dynamic data, so as to generate a new integrated operation strategy and send it to the equipment execution layer through the elastic scheduling execution module, thereby forming a closed-loop scheduling.

[0232] A warehouse classification management method, comprising the following steps:

[0233] S1: Real-time acquisition and updating of warehouse dynamic data:

[0234] Data collection scope defined: Three core collection objects for warehouse dynamic data are identified, namely material characteristic data, in-transit task queue status data, and real-time status data of various handling equipment at the equipment execution layer; the equipment execution layer is equipped with at least one lurking automated guided vehicle and at least one bin handling robot.

[0235] Multi-source data acquisition implementation: Capture material identification information, specification information, storage location information, and real-time status data such as equipment location, power, and load through a pre-set sensor group; periodically retrieve information such as the number, type, priority, and execution progress of tasks in transit through the warehouse management system interface;

[0236] Data preprocessing operations: The collected raw data is cleaned, outliers in numerical data are removed using the 3σ principle, missing fields are filled with the mean of the same category, and duplicate data is deleted based on unique identifiers; the cleaned data is converted into a unified structured format, and the logical consistency of the data and multi-source data fusion are completed.

[0237] Incremental data update execution: Combining operational feedback data from the equipment execution layer and external system instructions, incremental updates are performed on the warehouse dynamic data after fusion and verification; new data is directly appended, updated data overwrites the original fields, deleted data is marked as invalid, and a unique version number is assigned to each updated data to achieve full lifecycle traceability of data;

[0238] S2: Dynamically calculate the optimal material classification threshold or rule parameters:

[0239] Construction of a quantitative indicator system: Analyze and quantify the status data of the in-transit task queue and the real-time status data of the equipment obtained by S1; calculate the task urgency index;

[0240] Warehouse status assessment: Based on task urgency and equipment load balance indicators, combined with preset weights, a comprehensive warehouse efficiency score is calculated;

[0241] Optimal parameter selection execution: Based on the warehouse situation assessment results, extract 5 to 10 sets of candidate classification rule parameters from the preset parameter space; call the reinforcement learning optimization decision model to simulate and deduce the candidate parameters, and calculate the comprehensive benefit value of each set of parameters;

[0242] S3: Material Category Classification:

[0243] Classification criteria are defined as follows: the classification criteria for the first category of materials are that the volume is greater than the volume threshold, the weight is greater than the weight threshold, and the order batch size is greater than the batch size threshold. All three conditions must be met simultaneously.

[0244] Classification and discrimination execution: Match and discriminate the material feature data obtained by S1 with the optimal classification parameters output by S2; materials that meet the classification criteria are classified into the first category of material identifier set, and materials that do not meet any criteria are classified into the second category of material identifier set; generate a classification result list containing material identifier, category, order to which it belongs, and storage location;

[0245] S4: Dynamic assembly and generation of integrated job strategy:

[0246] Equipment status extraction: Extract indicators such as the location, power consumption, and task load rate of the hidden automated guided vehicle from the real-time equipment status data obtained from S1, and extract indicators such as the availability status and load utilization rate of the bin handling robot to generate a structured equipment status description set.

[0247] Strategy component matching: Based on the material classification results, the first category of materials is matched with a hidden automated guided vehicle scheduling component, a mobile shelf docking path planning component, and an overall handling task sequence generation component; the second category of materials is matched with a bin handling robot assignment component, a standardized bin storage and retrieval operation component, and an independent pick-and-place task sequence generation component; the component matching follows the capability adaptation principle, giving priority to low-load, high-power equipment.

[0248] Initial strategy assembly: Following the logical architecture of task decomposition, equipment assignment, path planning, and timing arrangement, the matched strategy components are combined and parameters are bound to generate an initial integrated operation strategy that includes equipment assignment, operation path, task sequence, and control parameters.

[0249] Conflict detection and optimization: The initial integration operation strategy is tested for equipment resource conflicts, path conflicts, and timing conflicts; for the detected conflicts, they are resolved by adjusting the task timing window, reassigning backup equipment, and optimizing path planning, so as to generate the final executable integration operation strategy.

[0250] S5: Parses the integrated operation strategy into a specific set of device driver instructions:

[0251] Strategy Element Extraction: The final integrated operation strategy generated by S4 is analyzed to extract three core elements: equipment assignment information, operation sequence information, and control parameter information.

[0252] Instruction template matching: Identify the target equipment type based on the equipment assignment information, retrieve the preset equipment instruction template library, substitute the operation sequence and control parameters into the template, and convert it into a set of drive instruction drafts that match the equipment type;

[0253] Timing orchestration and verification: Combine real-time device status data to orchestrate the timing of the driver instruction draft set and allocate reasonable time windows for instructions of different devices; perform conflict pre-check on the orchestrated instruction set to ensure that there are no execution conflicts before generating the final device driver instruction set.

[0254] S6: Send the device drive instruction set to the corresponding handling device in the device execution layer to execute the task.

[0255] Instructions are issued sequentially: according to the execution sequence of the instruction set, the drive instructions are sequentially issued to the corresponding steerable automated guided vehicles and bin handling robots through the device communication interface;

[0256] Differentiated operation execution: The steerable automated guided vehicle performs mobile shelf docking and overall loading, transportation and unloading of Category I materials; the bin handling robot performs standardized bin grabbing and independent picking and placing of Category II materials; the operation strictly follows the control parameters such as speed and grabbing force in the instructions;

[0257] Command execution monitoring: Receives command reception confirmation signals and initial execution status signals returned by the equipment in real time, monitors the equipment operating status during the operation process, and promptly detects and reports abnormal situations;

[0258] S7: Feedback updates form a closed-loop scheduling:

[0259] Status data acquisition: Collect status change data of the equipment during the operation process, including equipment location update data, task execution progress data, and equipment fault alarm data;

[0260] Data preprocessing and feedback: The collected state change data is cleaned, timestamp aligned, and format standardized, and then encapsulated into a feedback data packet with a unified protocol format, which is transmitted to the dynamic feature perception module through a wireless communication network;

[0261] Closed-loop iteration trigger: The dynamic feature perception module merges and updates the feedback data with the current warehouse dynamic data to generate real-time warehouse dynamic data; it automatically triggers the adaptive decision engine to re-execute steps S2 to S6, generate a new integrated operation strategy and issue it for execution, forming a complete closed-loop scheduling process.

[0262] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A warehouse classification management system, characterized in that: include: The dynamic feature perception module is used to acquire and update warehouse dynamic data in real time. The warehouse dynamic data includes at least: material feature data, in-transit task queue status data, and real-time status data of various handling equipment in the equipment execution layer. An adaptive decision engine is communicatively connected to the dynamic feature perception module; the adaptive decision engine includes: The rule optimization unit is used to dynamically calculate and output the current optimal material classification threshold or rule parameters based on real-time task queue status data and real-time equipment status data; the material classification includes at least a first category of materials suitable for overall handling and a second category of materials suitable for independent handling. The strategy generation unit is used to classify target materials into corresponding categories based on material characteristic data and the current optimal rule parameters output by the rule optimization unit, and dynamically assemble and generate corresponding integrated operation strategies from a pre-set strategy component library based on real-time equipment status data; wherein, the strategy for assembling the first type of materials specifies at least a hidden automated guided vehicle as the core handling equipment and a mobile shelf as the handling unit; the strategy for assembling the second type of materials specifies at least a bin handling robot as the core handling equipment and a standardized bin as the handling unit. The elastic scheduling execution module communicates with the adaptive decision engine and is used to parse the integrated operation strategy into a specific set of equipment driving instructions and send them to the equipment execution layer; the equipment execution layer includes at least one lurking automated guided vehicle and at least one bin handling robot. The state changes generated during the execution of tasks by the device execution layer are fed back to the dynamic feature perception module in real time, forming a closed-loop scheduling.

2. The warehouse classification management system according to claim 1, characterized in that: The dynamic feature perception module is used to acquire and update warehouse dynamic data in real time, including: The raw data acquisition unit is used to periodically capture raw warehouse data through a preset sensor group and warehouse management system interface. The raw warehouse data includes at least the initial characteristic information of the materials, the initial task queue information, and the initial status information of each handling equipment. The data preprocessing unit is connected to the raw data acquisition unit and is used to clean and format the raw warehouse data to obtain formatted warehouse data. The formatted warehouse data includes at least a structured material feature dataset, a structured task queue dataset, and a structured equipment status dataset. The data verification and fusion unit communicates with the data preprocessing unit and is used to perform logical consistency verification and multi-source data fusion on the formatted warehouse data to generate dynamic warehouse data after fusion and verification. The dynamic update unit communicates with the data verification and fusion unit and is used to incrementally update and manage the version of the fused and verified warehouse dynamic data based on the real-time operation feedback from the equipment execution layer and external system instructions, so as to form the final updated warehouse dynamic data. The data output unit communicates with the dynamic update unit and is used to push the final updated warehouse dynamic data to the adaptive decision engine in real time according to a preset format, so that the rule optimization unit and strategy generation unit can call it.

3. The warehouse classification management system according to claim 1, characterized in that: The rule optimization unit is used to dynamically calculate and output the current optimal material classification threshold or rule parameters based on real-time task queue status data and real-time equipment status data, including: Receive real-time task queue status data and real-time device status data, and parse and quantify the data to generate a quantitative indicator system that includes task urgency indicators and device load balancing indicators. Based on a quantitative indicator system, the overall operational efficiency of the current warehouse and the real-time capacity margin of each handling equipment are assessed, and a warehouse status assessment result including a comprehensive efficiency score and a description of equipment capacity status is generated. Based on the warehouse status assessment results, multiple sets of candidate classification rule parameters that match the current warehouse status are extracted from the preset parameter space to form a candidate rule parameter set; The preset optimization decision model is used to simulate and predict the benefits of each group of parameters in the candidate rule parameter set, and the candidate rule parameter with the highest comprehensive benefit prediction is selected as the current optimal material classification threshold or rule parameter. The current optimal material classification threshold or rule parameters are encapsulated and standardized in format, and output to the strategy generation unit for use. At the same time, the quantitative indicator system and warehouse status assessment results on which this decision is based are archived to the historical rule base for subsequent iterative learning of the optimization decision model.

4. The warehouse classification management system according to claim 3, characterized in that: The strategy generation unit is used to classify target materials into corresponding categories based on material characteristic data and the current optimal rule parameters output by the rule optimization unit, and dynamically assemble and generate corresponding integrated operation strategies from a pre-set strategy component library based on real-time equipment status data, including: Receive material feature data and current optimal rule parameters, and match and judge the material feature data according to the classification threshold contained in the current optimal rule parameters, and output the material classification result containing the first type of material identifier set and the second type of material identifier set; Receive real-time equipment status data and extract real-time location information, power information, task load information of each hidden automated guided vehicle, as well as real-time availability status information and load status information of each bin handling robot, to generate a structured equipment status description set. Based on the material classification results and the structured equipment status description set, retrieve and select strategy components that match the current material category and equipment status from the pre-set strategy component library; among them, the strategy components for matching materials in the first type of material identifier set include at least a hidden automated guided vehicle scheduling component, a mobile shelf docking path planning component, and an overall handling task sequence generation component; the strategy components for matching materials in the second type of material identifier set include at least a bin handling robot assignment component, a standardized bin storage and retrieval operation component, and an independent pick-and-place task sequence generation component; The selected strategy components are combined and parameter-bound according to the predefined strategy logic architecture to generate an initial integrated operation strategy that includes specific equipment assignment, operation path, task sequence and control parameters. Based on the latest structured equipment status description set, conflict detection and efficiency evaluation are performed on the initial integration operation strategy, and the detected equipment resource conflicts or path conflicts are resolved and optimized to output the final executable integration operation strategy. The final executable integrated job strategy is encapsulated according to a predetermined protocol format and output to the elastic scheduling execution module.

5. The warehouse classification management system according to claim 1, characterized in that: The elastic scheduling execution module is used to parse the integrated job strategy into a specific set of device driver instructions and send it to the device execution layer, including: The instruction parsing and decomposition unit is used to receive the integrated operation strategy output by the strategy generation unit, parse the integrated operation strategy, extract the equipment assignment information, operation sequence information and control parameter information, and generate the parsed strategy elements. The instruction conversion and adaptation unit is connected in communication with the instruction parsing and decomposition unit. It is used to identify the target equipment type based on the equipment assignment information in the parsed strategy elements, and convert the job sequence information and control parameter information into a set of drive instruction drafts that match the target equipment type based on the preset equipment instruction template library. The timing arrangement and verification unit is connected in communication with the instruction conversion and adaptation unit. It is used to receive the driver instruction draft set and, in combination with the real-time device status data obtained from the dynamic feature perception module, arrange the execution timing of each instruction in the driver instruction draft set and perform conflict pre-check, and generate the timing arranged instruction set. The instruction distribution and synchronization unit is connected to the timing arrangement and verification unit. It is used to sequentially send each driving instruction to the corresponding target handling device in the device execution layer through the device communication interface according to the instruction execution timing of the instruction set after timing arrangement. The instruction confirmation and monitoring unit is connected to the instruction distribution and synchronization unit. It is used to monitor and receive instruction reception confirmation signals and initial execution status signals returned by each handling device in the equipment execution layer in real time after the instruction is issued. The execution log generation unit communicates with the instruction confirmation and monitoring unit. It generates a structured instruction execution log based on the instruction receipt confirmation signal, the initial execution status signal, and the content of the issued instruction, and pushes the structured instruction execution log synchronously to the dynamic feature perception module.

6. The warehouse classification management system according to claim 5, characterized in that: The state changes generated during the execution of tasks by the device execution layer are fed back to the dynamic feature perception module in real time, forming a closed-loop scheduling; including: The status change acquisition unit is used to collect status change data generated by various handling equipment in the equipment execution layer during the execution of operations in real time. The status change data includes at least equipment position update data, task execution progress data, and equipment fault alarm data. The state data preprocessing unit is connected in communication with the state change acquisition unit. It is used to clean, timestamp align and format standardize the state change data to generate preprocessed state data. The feedback data encapsulation unit is communicatively connected to the state data preprocessing unit and is used to encapsulate the preprocessed state data into a feedback data packet with a unified protocol format. The feedback transmission unit is communicatively connected to the feedback data encapsulation unit and is used to transmit the feedback data packet to the dynamic feature sensing module in real time via a wireless communication network. The dynamic data update unit is connected to the feedback transmission unit and is used to receive feedback data packets in the dynamic feature perception module, parse them, and then merge and incrementally update them with the current warehouse dynamic data to generate real-time updated warehouse dynamic data. The closed-loop triggering unit communicates with the dynamic data update unit. When real-time updated warehouse dynamic data is generated, it automatically triggers the adaptive decision engine to re-execute rule optimization and strategy generation based on the real-time updated warehouse dynamic data, so as to generate a new integrated operation strategy and send it to the equipment execution layer through the elastic scheduling execution module, thereby forming a closed-loop scheduling.

7. A warehouse classification management method, characterized in that: The method implements the warehouse classification management system according to any one of claims 1-6, and the method includes the following steps: S1: Real-time acquisition and updating of warehouse dynamic data; warehouse dynamic data includes at least: material characteristic data, in-transit task queue status data, and real-time status data of various handling equipment in the equipment execution layer; the equipment execution layer includes at least one lurking automated guided vehicle and at least one bin handling robot; S2: Based on real-time task queue status data and real-time equipment status data, dynamically calculate and output the current optimal material classification threshold or rule parameters; material classification includes at least a first category of materials suitable for overall handling and a second category of materials suitable for independent handling; S3: Based on the material characteristic data and the current optimal rule parameters, classify the target material into either the first category or the second category of materials; S4: Based on real-time equipment status data and the category of target materials, dynamically assemble and generate corresponding integrated operation strategies from the pre-set strategy component library; wherein, the strategy assembled for the first type of materials specifies at least a hidden automated guided vehicle as the core handling equipment and a mobile shelf as the handling unit; the strategy assembled for the second type of materials specifies at least a bin handling robot as the core handling equipment and a standardized bin as the handling unit. S5: Parse the integrated operation strategy into a specific set of device driver instructions; S6: Send the device drive instruction set to the corresponding handling device in the device execution layer to execute the operation; S7: Collect the status change data generated when the equipment execution layer performs the operation, and feed back the status change data and update it to the warehouse dynamic data, so as to re-execute the steps starting from S2 based on the updated warehouse dynamic data, forming a closed-loop scheduling.

Images