An intelligent warehouse warehousing management method and device for furniture products

Abstract

The application discloses a kind of intelligent warehouse warehousing management method and device for furniture product, it is related to warehouse data management technical field, comprising: the real-time digital twin model of synchronization update is built with physical warehouse, the feature vector of each product to be warehoused batch is obtained;In digital twin model, with the minimum predicted completion time as the goal of on-shelf, dynamic path planning is carried out, and the optimal expected path is generated;The idle space of transport vehicle is combined with the optimal expected path to generate initial loading scheme;Matched from batch is retrieved from several batches to be dispatched in real time, and the initial loading scheme is optimized and adjusted according to the location information of from batch, to generate optimized loading scheme;Optimized loading scheme is assigned to transport vehicle, and transport vehicle is monitored and dynamically adjusted in real time through digital twin model. Through the product information of multiple batches to be warehoused is combined to carry in the path planning, the convenience of product warehousing management is improved, so as to improve production efficiency.

Description

Technical Field

[0001] This invention relates to the field of warehouse data management technology, specifically to an intelligent warehouse inbound management method and apparatus for furniture products. Background Technology

[0002] In the furniture product warehousing process, traditional goods identification and positioning methods often rely on manual operation or simple barcode systems, which are prone to errors and inefficient. Meanwhile, the complex environmental factors within warehouses, such as obstructions and signal interference, make accurately locating large furniture items a challenge. Regarding the collaborative operation of multiple transport vehicles, existing path planning and task scheduling algorithms struggle to effectively handle dynamic obstacles and real-time task changes. Furthermore, current single transport tasks primarily address the warehousing needs of the current batch of products, proceeding sequentially by batch. This results in insufficient space utilization of transport vehicles during each transport, impacting the overall efficiency of product warehousing and transportation. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides an intelligent warehouse inbound management method and device for furniture products. By constructing a digital twin model and coordinating route planning, it enables the combined transportation of multiple batches of products, improves the rationality of product transportation allocation for a single inbound shipment, and thus improves the work efficiency of product inbound transportation.

[0004] This invention provides an intelligent warehouse inbound management method for furniture products, the management method comprising:

[0005] Construct a real-time digital twin model that is updated synchronously with the physical warehouse, extract multi-dimensional features of each batch of products to be put into the warehouse, and encode physical attribute features, storage strategy features, operation features and context information features into feature vectors input to the real-time digital twin model.

[0006] Based on the product data of the current main batch of goods entering the warehouse, several target storage locations are determined. Based on the target storage locations and the feature vector, dynamic path planning is performed in the digital twin model with the goal of minimizing the expected shelf completion time, and the optimal expected path is generated.

[0007] Obtain the physical attributes of the main batch of products, calculate the available space of the transport vehicle based on the rated capacity of the transport vehicle, and generate an initial loading plan based on the physical attributes of the main batch of products, the available space of the transport vehicle, and the optimal expected path.

[0008] Using the optimal expected path and the available space of the transport vehicle as constraints, matching sub-batches are retrieved in real time from several batches to be scheduled. The sub-batches are dynamically added to the current transport task to form a merged task group. The optimal expected path and the initial loading plan are optimized and adjusted according to the cargo location information of the sub-batch to generate an optimized loading plan.

[0009] The optimized loading scheme is assigned to the transport vehicle, and the transport vehicle is monitored and dynamically adjusted in real time through the digital twin model during the task execution.

[0010] Furthermore, the construction of a real-time digital twin model that is synchronized with the physical warehouse includes:

[0011] Obtain static information such as the coordinates of the storage shelves, road network, and site locations of the physical warehouse, and construct a static layout layer of the physical warehouse based on the static information;

[0012] The system monitors and updates the inventory status of the physical warehouse, traffic flow of the road network, equipment operating status, and temporary obstacle information in real time, and constructs a dynamic layout layer for the physical warehouse.

[0013] A real-time digital twin model of the physical warehouse is constructed based on the static layout layer and the dynamic layout layer.

[0014] Furthermore, the physical attribute characteristics include the product's size data, volume data, and weight data;

[0015] The storage strategy features include the product's target shelf, shelf type, and storage priority data;

[0016] The operational features include estimated operation duration and data on the transfer operation tools;

[0017] The contextual information features include product entry time and associated order information.

[0018] Furthermore, the process of determining several target storage locations based on the product data of the current main batch entering the warehouse, and generating the optimal expected path in the digital twin model based on the target storage locations and the feature vector, with the objective of minimizing the estimated shelf completion time, includes:

[0019] Obtain the product data of the main batch, and combine it with the rated transportation space data of the transport vehicle to set the product types and quantities for a single transport vehicle trip within the main batch;

[0020] Based on the product type and quantity of a single transport, several target cargo locations are determined, and based on the several target cargo locations and the feature vector, several preliminary transport routes are planned in the digital twin model;

[0021] The time-varying A* algorithm is used to divide the running time of a single product delivery into discrete time slices, and a time-varying cost function combining the baseline travel time, time slice congestion coefficient and obstacle penalty time is established for each path segment.

[0022] The optimal expected path is generated based on the time-varying cost function.

[0023] Furthermore, the step of retrieving matching sub-batches from several pending scheduling batches in real time, using the optimal expected path and the available space of the transport vehicle as constraints, includes:

[0024] Using the optimal expected path as the core, the storage areas along the route and adjacent to it are extracted to form a geographical search range;

[0025] Using the available space of the transport vehicle and the geographical search range as hard constraints, the system filters and matches from several batches to be scheduled according to preset priority rules, selects products that meet the requirements of the available space of the transport vehicle and the geographical search range, and marks the batch to be scheduled containing the product as a secondary batch.

[0026] Furthermore, the step of using the available space of the transport vehicle and the geographical search range as hard constraints, and selecting and matching from several batches to be scheduled according to a preset priority rule, includes:

[0027] Using the geographic search range as a filtering criterion, batches matching the path area where the target cargo location is located within the geographic search range are selected from the production data of several batches to be scheduled.

[0028] In the path region matching batch, storage matching batches that are compatible with the storage strategy of the main batch products are selected based on the storage strategy of the main batch products.

[0029] Within batches that meet storage policy compatibility, the optimal batch for filling the available space of the transport vehicle is selected based on the available space of the transport vehicle.

[0030] Furthermore, the step of dynamically adding the batches to the current transportation task to form a merged task group, and optimizing and adjusting the optimal expected route and the initial loading plan based on the cargo location information of the batches to generate an optimized loading plan includes:

[0031] Based on the location information of the secondary batch and the location information of the primary batch, obtain the target location information for the current transportation task;

[0032] Obtain the target cargo location information corresponding to the arrangement order of the optimal expected path, and determine the loading order of all batches of products in the merged task group according to the principle of first-in-last-out and the arrangement order;

[0033] An adjusted optimized loading plan is generated based on the target cargo location information and the loading sequence.

[0034] Furthermore, the step of assigning the optimized loading scheme to the transport vehicle and monitoring and dynamically adjusting the transport vehicle in real time through the digital twin model during task execution includes:

[0035] Based on the type of transport vehicle, current location, battery level, and current load status, a multi-agent reinforcement learning model is used to calculate and generate assignment instructions.

[0036] The optimized loading scheme is assigned to the transport vehicle according to the assignment instruction;

[0037] The system acquires the real-time transportation status of the transport vehicle and the obstacle information of the optimal expected path, detects whether the transport vehicle has a task delay, and dynamically adjusts the transportation status of the transport vehicle if a task delay exists.

[0038] Furthermore, the dynamic adjustment of the transport vehicle's transport status includes:

[0039] If a delay in the transport vehicle is detected by the digital twin model during task execution, a local path replanning using the D*Lite algorithm is triggered.

[0040] The present invention also provides an intelligent warehouse inbound management device for furniture products, the device being used to execute any of the management methods described above, the device comprising:

[0041] Model building module: used to build a real-time digital twin model that is synchronized with the physical warehouse, extract multi-dimensional features of each batch of products to be put into the warehouse, and encode physical attribute features, storage strategy features, operation features and context information features into feature vectors input to the real-time digital twin model;

[0042] The path planning module is used to determine several target storage locations based on the product data of the current main batch of goods entering the warehouse. Based on the target storage locations and the feature vector, dynamic path planning is performed in the digital twin model with the goal of minimizing the expected shelf completion time, and the optimal expected path is generated.

[0043] Solution generation module: used to obtain the physical attributes of the main batch of products, calculate the free space of the transport vehicle in combination with the rated capacity of the transport vehicle, and generate an initial loading plan based on the physical attributes of the main batch of products, the free space of the transport vehicle, and the optimal expected path.

[0044] The scheme optimization module is used to retrieve matching sub-batch from several batches to be scheduled in real time, using the optimal expected path and the idle space of the transport vehicle as constraints, dynamically add the sub-batch to the current transport task to form a merged task group, and optimize and adjust the optimal expected path and the initial loading scheme according to the cargo location information of the sub-batch to generate an optimized loading scheme.

[0045] The scheme adjustment module is used to assign the optimized loading scheme to the transport vehicle and to monitor and dynamically adjust the transport vehicle in real time through the digital twin model during the task execution process.

[0046] This invention provides a smart warehouse inbound management method and device for furniture products. By constructing a digital twin model and coordinating route planning, it enables the combined transportation of multiple batches of products, improves the rationality of product transportation allocation for a single inbound shipment, and thus improves the work efficiency of product inbound transportation. Attached Figure Description

[0047] Figure 1 This is a flowchart of an intelligent warehouse inbound management method for furniture products in an embodiment of the present invention;

[0048] Figure 2 This is a schematic diagram of the inbound management of the real-time digital twin model of the intelligent warehouse in this embodiment of the invention;

[0049] Figure 3 This is a schematic diagram of the optimized and adjusted loading scheme in an embodiment of the present invention;

[0050] Figure 4 This is a schematic diagram of an intelligent warehouse inbound management device for furniture products in an embodiment of the present invention. Detailed Implementation

[0051] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0052] Example 1:

[0053] Figure 1 A flowchart of an intelligent warehouse inbound management method for furniture products is shown in an embodiment of the present invention; Figure 2 This diagram illustrates the inbound management of a real-time digital twin model of an intelligent warehouse in an embodiment of the present invention. The management method includes:

[0054] S11: Construct a real-time digital twin model that is updated synchronously with the physical warehouse, extract multi-dimensional features of each batch of products to be put into storage, and encode physical attribute features, storage strategy features, operation features, and contextual information features into feature vectors input to the real-time digital twin model.

[0055] Specifically, the real-time digital twin model refers to setting up a virtual mapping in the digital space corresponding to the physical warehouse based on the data information of the physical warehouse. In the real-time data twin model, information such as the layout, inventory, equipment status, and traffic flow of the physical warehouse can be synchronized in real time, providing the management system with a comprehensive, accurate, and dynamic warehouse view, thereby supporting decision-making and task execution.

[0056] The batch to be put into storage refers to the collection of furniture products waiting to be stored in the warehouse. According to the actual processing and production situation, each furniture product is transported to the warehouse for classification and storage along with the processing and production order batch.

[0057] Furthermore, the feature vector refers to the information obtained from multi-dimensional feature extraction, which is converted into a unified mathematical representation through encoding, so as to be analyzed and calculated in the real-time digital twin model and realize accurate operation of furniture product warehousing.

[0058] Specifically, the construction of a real-time digital twin model that is synchronized with the physical warehouse includes:

[0059] The static information of the storage rack coordinates, road network, and site locations of the physical warehouse is obtained, and a static layout layer of the physical warehouse is constructed based on the static information. The static layout layer is used to store the static information of the rack coordinates, road network, and site locations. Based on the static layout layer, a component of the real-time digital twin model is formed to store relatively fixed physical structure information in the warehouse, so as to improve the construction accuracy of the real-time digital twin model.

[0060] Furthermore, the static layout layer can be constructed in various forms. In this embodiment, the static layout layer is constructed using a geographic information system database to store the geographic coordinates of the shelves, the topology of the road network, and the precise location information of each workstation, so as to improve data support for the precise warehousing planning of furniture products.

[0061] The system continuously monitors and updates the inventory status of the physical warehouse, traffic flow of the road network, equipment operating status, and temporary obstacle information to construct a dynamic layout layer for the physical warehouse. This dynamic layout layer is used to update the inventory status of each storage location, traffic flow of each path segment, equipment operating status, and temporary obstacle information in real time. Based on this dynamic layout layer, another component of the real-time digital twin model is formed, used to update the constantly changing operational status information in the warehouse in real time. The dynamic layout layer can employ various technologies to achieve real-time data updates. In this embodiment, a message queue system can be used to receive real-time data streams from various sensors and business systems and store them in an in-memory database or a time-series database to ensure low-latency access and high-concurrency processing capabilities, thereby achieving dynamic detection and real-time updates of physical warehouse information.

[0062] A real-time digital twin model of the physical warehouse is constructed based on the static layout layer and the dynamic layout layer. The static layer includes shelf coordinates, road network, and site locations; the dynamic layer updates the inventory status of each storage location, traffic flow along the path, equipment operating status, and temporary obstacle information in real time. Simultaneously, for each batch of goods to be received, its multi-dimensional features are collected through the Internet of Things and information systems and encoded into a unified feature vector. This vector specifically includes a physical attribute vector P, a storage strategy vector S, an operational feature vector O, and a context vector C.

[0063] Among them, the physical attribute vector P contains size, volume, and weight information; the storage strategy vector S contains target storage area, shelf type, environmental requirements, and priority information; the operation feature vector O contains estimated operation time and required tool type information; and the context vector C contains inbound time window and associated order information.

[0064] Furthermore, in this embodiment, the warehouse receiving time is set from 8:00 to 18:00. Among the batches to be received, furniture product A has dimensions of 2.2m*0.8m*0.5m, a volume of 0.88m³, and a weight of 80kg; furniture product B has dimensions of 2.1m*1.0m*0.6m, a volume of 1.26m³, and a weight of 75kg.

[0065] Based on the static layout layer and the dynamic layout layer, a real-time digital twin model of the physical warehouse is constructed, and feature vectors for the batches to be received are generated.

[0066] Physical attribute vector P: P=[[2.2, 0.8, 0.5, 0.88, 80], [2.1, 1.0, 0.6, 1.26, 75]], corresponding to the size, volume, and weight data of the furniture product;

[0067] Storage strategy vector S: S = ["BULK_STORAGE", "HEAVY_DUTY_RACK", "HIGH"], that is, the target area is "large item area", the shelf type is "heavy-duty shelf", and the priority is "high".

[0068] Operation feature vector O: O=[300, "FORKLIFT_ATTACHMENT"], that is, the estimated time for unloading operation is 300 seconds, and the operation tool is a forklift.

[0069] Context vector C: C = ["08:30", "10:00", "P20250502"], which means the batch entry time is 08:30 and the batch number is P20250502.

[0070] By dividing the real-time digital twin model into a static layout layer and a dynamic state layer, and combining this with multi-dimensional feature extraction for each batch of goods to be received, accurate and real-time environmental perception and decision support are provided for intelligent warehouse inbound management. Specifically, the static layout layer, as the stable foundation of the warehouse's physical structure, stores information that remains unchanged or changes slowly, such as shelf coordinates, road networks, and station locations. This ensures that there is always an accurate and reliable physical spatial reference during path planning, avoiding inconsistencies or errors in layout information caused by frequent updates. Meanwhile, the dynamic state layer focuses on capturing the ever-changing internal conditions of the warehouse in real time, such as the inventory status of each storage location, traffic flow in each path segment, equipment operating status, and information on temporary obstacles. This layered design enables the model to efficiently process massive amounts of data. Static data does not require frequent loading and updating, while dynamic data can be refreshed with extremely low latency, thus solving the problems of untimely information updates and insufficient accuracy in traditional models.

[0071] Furthermore, the static layout layer can be set as a PostgreSQL database containing a PostGIS extension. The coordinates of the physical warehouse shelves are stored as geometric objects, with each shelf record containing its unique ID, length, width, and height dimensions, and precise 3D coordinates on the warehouse floor plan. This allows the road network to be represented as a series of line segment objects, based on which the travel paths of transport vehicles can be defined.

[0072] Furthermore, the dynamic state layer can be specifically a message queue system based on Apache Kafka, combined with a Redis in-memory database. Various sensors within the warehouse, such as RFID readers installed on shelves, LiDAR and cameras mounted on transport vehicles, and the equipment's own operational status sensors, send real-time data streams to Kafka topics based on the sensor system. This real-time data, after being processed by Kafka, is updated in the Redis in-memory database to ensure that the digital twin model can reflect the latest warehouse dynamics with millisecond-level latency.

[0073] S12: Based on the product data of the current main batch of goods entering the warehouse, determine several target storage locations. Based on the target storage locations and the feature vector, perform dynamic path planning in the digital twin model with the goal of minimizing the expected shelf completion time, and generate the optimal expected path.

[0074] Specifically, step S12 includes:

[0075] Obtain the product data of the main batch, combine it with the rated transportation space data of the transport vehicle, set the product types and quantities for a single transport by the transport vehicle within the main batch, extract the product types involved in the main batch products based on the product data of the main batch, and generate matching data for the product types and quantities for a single transport by the transport vehicle based on the rated transportation space data of the transportation layer.

[0076] Based on the product type and quantity of a single transport, several target cargo locations are determined. Several preliminary transport routes are planned in the digital twin model based on these target cargo locations and the aforementioned feature vectors. According to historical transport operation requirements, a basic operation time for a single transport is set. Combining the product data of the main batch, the target cargo locations of the main batch are marked in the real-time digital twin model. The loading position of the transport vehicle is used as the start and end point of the transport route. Several transport routes for the transport vehicle are set within the basic operation time. The transport route that passes through the most target cargo locations is selected from these routes. Then, combining the rated transport space data of the transport vehicle and the product data of the main batch, a transport route that can transport the most products within the basic operation time is matched, thereby obtaining the preliminary transport route.

[0077] The time-varying A* algorithm divides the operation time of a single product delivery into discrete time slices, and establishes a time-varying cost function for each path segment, combining the baseline travel time, time slice congestion coefficient, and obstacle penalty time. The optimal expected path is then generated based on this time-varying cost function. Dividing warehouse operation time into discrete time slices means segmenting the continuous operational time axis into a series of fixed or variable-length time periods. This allows for refined modeling and prediction of the dynamic changes in the warehouse environment over time, enabling the path planning algorithm to perform cost assessment and decision-making based on the specific environmental conditions within each time slice. Under ideal conditions—no traffic congestion, no temporary obstacles, and normal equipment operation—the shortest time required for a transport vehicle to traverse a specific path segment is obtained. This shortest time is used as the initial reference value for calculating the time-varying cost of the path segment. The time slice congestion coefficient is an indicator used to quantify the degree of traffic congestion on a specific path segment within a specific time slice. Its function is to reflect the impact of traffic flow on vehicle speed and time, thus reflecting the additional delays caused by congestion in path planning.

[0078] Furthermore, this embodiment of the invention divides warehouse operation time into discrete time slices, enabling route planning to more precisely perceive and predict environmental conditions within different time periods. Based on this, a time-varying cost function is established for each route segment. This function comprehensively considers the baseline travel time, the time-slice congestion coefficient, and the obstacle penalty time. The baseline travel time provides the ideal traffic efficiency, the time-slice congestion coefficient dynamically adjusts the route cost based on real-time or predicted traffic flow from the digital twin model, and the obstacle penalty time quantifies the delays caused by temporary obstacles. In this way, the time-varying cost function provides a more realistic assessment of route segment travel costs that changes dynamically over time.

[0079] In actual operation, when path planning is required for the main batch to be scheduled, the system aims to minimize the estimated completion time of shelving. When running the time-varying A* algorithm within the digital twin model, it searches for an optimal expected path based on potential congestion and obstacles within future time slices. By searching on a time-spreading graph, the time-varying A* algorithm can predict and avoid potential future congestion areas or temporary obstacles, thus selecting the optimal path in the time dimension. Due to the close integration of the time-varying A* algorithm with the real-time digital twin model, the digital twin model continuously provides information on the inventory status of each storage location, traffic flow of each path segment, equipment operating status, and temporary obstacles, providing real-time and accurate data support for the calculation of the time-varying cost function. Therefore, this solution can predict and avoid future environmental changes, thereby generating more accurate and efficient paths, significantly improving the accuracy and real-time performance of dynamic path planning.

[0080] Furthermore, in this embodiment, assuming the warehouse operates from 08:00 to 18:00 daily, the system divides this into discrete time slices of 5 minutes each. When a main batch needs to be put into the warehouse, the system first obtains the current warehouse status from the real-time digital twin model, including the real-time traffic flow of each path segment, whether there are temporary obstacles, and their expected duration.

[0081] By calculating the time-varying cost of each transportation route, the baseline travel time for a given route segment is 30 seconds. If the digital twin model predicts a large number of vehicles passing through this route segment during the time slot of 9:00-9:05, its time-slot congestion coefficient may be 1.5. Simultaneously, if the digital twin model detects a temporary obstacle on this route segment between 9:10-9:20, a significant obstacle penalty time will be imposed during these time slots, potentially rendering the route segment impassable during those time slots. The time-varying cost function integrates these factors to calculate the actual travel cost of this route segment in different time slots.

[0082] The time-varying A* algorithm takes the target location and feature vector of the main batch as input and searches on a time-varying graph composed of these time-varying costs, aiming to minimize the estimated shelf completion time. The algorithm considers the cost of a vehicle traveling along different path segments in different time slices from its current location and predicts the total time to reach the target location. For example, the algorithm might find a slightly longer spatially defined path that avoids future congestion or obstacles, resulting in a shorter estimated total time. In this way, the time-varying A* algorithm can calculate an optimal expected path.

[0083] S13: Obtain the physical attributes of the main batch of products, calculate the available space of the transport vehicle in combination with the rated capacity of the transport vehicle, and generate an initial loading plan based on the physical attributes of the main batch of products, the available space of the transport vehicle, and the optimal expected path.

[0084] Specifically, Figure 3 The diagram illustrates an optimized loading scheme in an embodiment of the present invention. The initial loading scheme is an adjustment of the product loading plan of the transport vehicle based on the transport vehicle's demand for transporting the main batch of products and the rated transport space of the transport vehicle.

[0085] Specifically, the initial loading scheme can be constructed by digitally fusing calculations of the "physical attributes of the main batch of products" and the "rated capacity of the transport vehicle". The structured dataset of the physical attributes of the main batch of products is analyzed. This structured dataset includes a three-dimensional model of the precise dimensional geometry of the main batch of products, with the center position marked. For irregularly shaped products, a polyhedral model or point cloud data can be used.

[0086] Based on the rated capacity of the transport vehicle, a three-dimensional spatial grid model is constructed. The cargo space of the transport vehicle is discretized into fine three-dimensional cells. The structured data of the main batch of products is filled into the cargo space model of the transport vehicle. Based on the single transport quantity of the main batch of products determined by the transport task, the system will calculate the remaining unoccupied space set in real time and mark it as the free space of the transport vehicle.

[0087] Furthermore, by combining the quantity and size data of each shipment of the main batch of products, the total volume and weight of each shipment of the main batch of products are calculated, taking the shipment of 3 furniture products A and 2 furniture products B as an example:

[0088] Total volume = 3 * 0.88 + 2 * 1.26 = 5.16 m³;

[0089] Total weight = 3 * 80 + 2 * 75 = 390 kg;

[0090] Matching the transport vehicle and calculating the available space: Based on the transport vehicle's rated capacity of 8m³ and load capacity of 1.8 tons, the transportation needs can be met. The available capacity of the transport vehicle is: 8 - 5.16 = 2.84m³, and the remaining load capacity of the transport vehicle is 1800 - 340 = 1460kg.

[0091] Specifically, based on the single transport task requirements of the transport vehicle, that is, the time required for the transport vehicle to complete the delivery of products into the warehouse and return to the loading position, and in combination with the actual transport operation, the loading requirement of the main batch of products in the transport vehicle is set to be more than 70%, that is, the loading amount of the main batch of products transported by the transport vehicle in a single trip reaches more than 70% of the rated capacity of the transport vehicle, so as to ensure that the transport vehicle can transport the main batch of products efficiently.

[0092] Furthermore, using a combined transportation scheme of 3 furniture products A and 2 furniture products B, the filling ratio of the main batch of products is 5.16 / 8*100%=64.5%, which does not meet the loading requirements. Therefore, the combined transportation scheme is adjusted: a new combination scheme is formed by 2 furniture products A and 3 furniture products B. The volume of the main batch of products transported in a single trip is calculated to be 2*0.88+3*1.26=5.54m³, and its filling ratio is 5.54 / 8*100%=69.25%, which accounts for approximately 70% of the rated capacity of the transport vehicle, thus meeting the transportation and filling requirements.

[0093] Furthermore, once the main batch of product data has been transported, the system will use the next batch of products awaiting warehousing as the main batch, and adjust the planning of the next transport vehicle's transport task based on the newly defined main batch, thereby improving the overall efficiency of the smart warehouse operation and enabling efficient coordination between production cycle and transportation efficiency.

[0094] S14: Using the optimal expected path and the available space of the transport vehicle as constraints, retrieve matching sub-batch from several batches to be scheduled in real time, dynamically add the sub-batch to the current transport task to form a merged task group, and optimize and adjust the optimal expected path and the initial loading plan according to the cargo location information of the sub-batch to generate an optimized loading plan.

[0095] Specifically, the step of retrieving matching sub-batches from several pending scheduling batches in real time, using the optimal expected path and the available space of the transport vehicle as constraints, includes:

[0096] Using the optimal expected path as the core, the storage areas along and adjacent to it are extracted to form a geographical search range. Based on the predetermined transportation path of the main batch, a limited geographical area related to the path is defined as the search space for potential secondary batches. By narrowing the search range from the entire warehouse to a local area closely related to the current task path, search efficiency is improved and unnecessary computational burden is reduced. Based on the warehouse layout information stored in the digital twin model, all shelf areas, aisle segments, or specific storage units traversed by the optimal expected path can be identified. For example, the system can analyze each node on the path and include its storage area or its directly adjacent storage area in the search range. Alternatively, the system can calculate the geometric boundary of the optimal expected path and extend a preset buffer distance outside this boundary, defining all storage locations falling within this buffer area as the geographical search range.

[0097] Using the available space of the transport vehicle and the geographical search range as hard constraints, the system filters and matches from several batches awaiting scheduling according to a preset priority rule, selecting products that meet the requirements of the available space of the transport vehicle and the geographical search range, and marking the batch containing these products as a secondary batch. Using the available space of the transport vehicle and the geographical search range as hard constraints, the system filters and matches from other batches awaiting scheduling according to a preset priority rule. This technical feature defines two basic and insurmountable conditions for selecting and matching secondary batches from batches awaiting scheduling and introduces a hierarchical filtering mechanism. The available space of the transport vehicle refers to the remaining usable capacity of the transport vehicle after loading the main batch, including volume, weight, or size limitations, ensuring that the secondary batch can be actually loaded. The geographical search range ensures that the target cargo location of the secondary batch is located near the main batch's transportation path to minimize additional detours. These two hard constraints together constitute the initial screening threshold; only batches that simultaneously meet both conditions are further considered. Based on this, the system sorts and selects batches that meet the hard constraints according to preset priority rules to find the most suitable secondary batch. For example, a multi-stage filtering mechanism can be used to first filter out batches that do not meet the hard constraints, and then apply a series of predefined scoring or sorting criteria to the remaining batches.

[0098] Specifically, the step of using the available space of the transport vehicle and the geographical search range as hard constraints, and selecting and matching from several batches to be scheduled according to a preset priority rule, includes:

[0099] Using the geographic search range as a filtering criterion, the system filters product data from several batches to be scheduled, selecting batches whose target storage locations fall within the geographic search range and whose paths match the target storage locations. Upon receiving the main batch to be scheduled and completing dynamic path planning based on a digital twin model, the system first uses this optimal expected path as the core, intelligently extracting the storage areas along its route and those adjacent to it, thereby dynamically constructing a geographic search range. This establishment of a geographic search range ensures that subsequent batch searches are no longer blindly conducted across the entire warehouse, but rather focus on local areas highly relevant to the main batch's task path, significantly reducing the search space.

[0100] Furthermore, by dividing the geographical search scope, batches of products that meet the optimal expected path are selected from several batches of products to be dispatched, thereby conducting preliminary screening among several batches of products to be dispatched.

[0101] Within the path region matching batches, storage matching batches compatible with the storage strategy of the main batch products are selected based on the storage strategy of the main batch products. Further screening is then performed on the path region matching batches that have completed the initial screening, ensuring that the storage strategy of the product data in the batch to be scheduled matches the storage strategy of the main batch. This effectively improves the efficiency of a single data entry operation.

[0102] For example, if the main batch and the secondary batch of furniture products use the same handling fixtures and are located near each other in the warehouse, the two types of furniture products can be handled using the same handling fixtures during the warehousing process, reducing the number of steps required to change handling fixtures and thus effectively improving warehousing efficiency.

[0103] Within batches that meet storage policy compatibility, the optimal batch for filling the available space of the transport vehicle is selected based on the available space of the transport vehicle.

[0104] Based on this, the system uses the available space of the main batch transport vehicle and the determined geographical search range as dual hard constraints. This ensures that batches whose target cargo location is within the geographical search range and whose physical attributes (such as size and weight) can be accommodated by the current vehicle's available space can be selected as secondary batches. This dual hard constraint mechanism ensures that the matched secondary batches are geographically compatible with the main batch's path and physically transportable by the same vehicle, thus guaranteeing the feasibility of merging task groups.

[0105] Furthermore, based on the above design, under the initial transportation plan, the transport vehicle performs combined transportation of 2 furniture products A and 3 furniture products B, with a total transportation volume of 5.56 m³. The idle space of the transport vehicle is 8 m³ - 5.56 m³ = 2.44 m³. Using the 2.44 m³ of idle space and the geographical search range as dual hard constraints, target products with a volume smaller than 2.44 m³ are selected from the batch of products to meet the unloading operation requirements along the transportation route, thereby constructing a merged task group to improve the transportation efficiency of the transport vehicle.

[0106] Subsequently, the system performs refined filtering and sorting of the slave batches that meet the hard constraints according to preset priority rules. These priority rules can be configured according to actual business needs, such as prioritizing batches with high path overlap, batches with compatible storage strategies, batches that can optimally fill vehicle idle space, and batches with tight order deadlines. Through this hierarchical filtering and priority matching mechanism, the system can quickly and accurately identify the slave batches most suitable for merging with the main batch from numerous batches to be scheduled. The entire process is closely integrated with the steps of building a real-time digital twin model, performing dynamic path planning, and generating an initial loading plan, ensuring the rationality and efficiency of merging task groups, thereby optimizing the overall transportation task execution efficiency and resource utilization.

[0107] Specifically, the step of dynamically adding the batch from the current transportation task to form a merged task group, and optimizing and adjusting the optimal expected route and the initial loading plan based on the cargo location information of the batch to generate an optimized loading plan includes:

[0108] Based on the location information of the secondary batch and the location information of the primary batch, obtain the target location information for the current transportation task;

[0109] The target cargo location information is obtained, corresponding to the arrangement order on the optimal expected path. The loading order of all batches of products in the merged task group is determined based on the last-in-first-out (LIFO) principle and this arrangement order. The "last-in-first-out" principle is a cargo loading and unloading strategy where the last cargo loaded into the transport vehicle is unloaded first, and the first cargo loaded is unloaded last. Implementing this principle ensures that in multi-point delivery or multi-batch unloading scenarios, the target cargo can be easily retrieved without moving or rearranging other cargo. For example, this can be achieved by placing batches destined for later destinations deeper inside the vehicle during loading, while placing batches destined for earlier destinations outside the vehicle or in easily accessible locations; or, at the software level, by logically sorting the loading plan, explicitly instructing loading personnel or automated equipment to operate according to this principle.

[0110] By immediately optimizing and adjusting the initial loading plan after forming a merged task group and optimizing the route, close coordination between the loading plan and the actual transportation route is ensured. Specifically, the system first performs dynamic route planning in a real-time digital twin model based on the main batch to be scheduled, combined with its target cargo location and multi-dimensional feature vector, to generate the optimal expected route. Then, based on the physical attributes of the main batch and the rated capacity of the transport vehicle, an initial loading plan including vehicle free space is generated. Subsequently, using the optimal expected route and the free space of the transport vehicle as dual hard constraints, matching secondary batches are retrieved in real time from other batches to be scheduled. These secondary batches are dynamically added to the current transportation task to form a merged task group, and the optimal expected route and the initial loading plan are optimized and adjusted according to the cargo location information of the newly added batches. During this optimization and adjustment process, this application further clarifies that the optimization and adjustment of the initial loading plan is based on the principle of last-in, first-out (LIFO) to determine the loading order of all batches in the merged task group. The application of this principle allows the loading order to be directly mapped back to the unloading order, thereby ensuring that the unloading order completely matches the order of cargo locations along the optimized final route. In this way, when the transport vehicle arrives at each target cargo location along the optimized route, the corresponding batch is always located in the easiest position to unload, avoiding unnecessary cargo handling or reordering on site, which greatly improves unloading efficiency and operational smoothness.

[0111] S15: Assign the optimized loading scheme to the transport vehicle, and monitor and dynamically adjust the transport vehicle in real time through the digital twin model during the task execution.

[0112] Specifically, the step of assigning the optimized loading scheme to the transport vehicle and monitoring and dynamically adjusting the transport vehicle in real time through the digital twin model during task execution includes:

[0113] Based on the type of transport vehicle, current location, battery level, and current load status, a multi-agent reinforcement learning model is used to calculate and generate assignment instructions.

[0114] The optimized loading scheme is assigned to the transport vehicle according to the assignment instruction;

[0115] The system acquires the real-time transportation status of the transport vehicle and the obstacle information of the optimal expected path, detects whether the transport vehicle has a task delay, and dynamically adjusts the transportation status of the transport vehicle if a task delay exists.

[0116] Specifically, the dynamic adjustment of the transport vehicle's transport status includes:

[0117] If a delay in the transport vehicle is detected by the digital twin model during task execution, a local path replanning using the D*Lite algorithm is triggered.

[0118] First, during task execution, the system continuously monitors the physical warehouse's operational status in real time using a digital twin model. As a precise digital mapping of the physical warehouse, the digital twin model integrates information from multiple data sources, including sensors, transport vehicles, and the task management system, thereby updating the inventory status of each storage location, traffic flow on each path segment, equipment operating status, and information on temporary obstacles in real time. When the digital twin model detects pre-defined anomalies, such as a blockage on a path segment, a vehicle malfunctioning, or the actual progress of the task significantly lagging behind the planned schedule, the system immediately triggers a corresponding dynamic adjustment mechanism.

[0119] Specifically, if a path blockage is detected, the system will quickly initiate local path replanning based on the D*Lite algorithm. The advantage of the D*Lite algorithm lies in its incremental search capability, which can efficiently update the path when the environment changes dynamically, avoiding the time-consuming process of recalculating from scratch. Through local replanning, the transport vehicle can quickly bypass the congested area and continue to complete the task, thereby minimizing the time delay caused by path obstruction.

[0120] On the other hand, if equipment failure or task delay is detected, the system will consider regrouping the current merged task group. The merged task group is formed by real-time matching of batches based on the optimal expected path of the main batch and vehicle availability. When equipment failure causes a task to be unable to proceed as planned, or when a task batch experiences a severe delay, the system will reassess the composition of the task group and resource allocation. For example, the affected batch can be separated from the current task group and reassigned to other available transport vehicles, or the entire task group can be split and completed collaboratively by multiple vehicles. This regrouping mechanism ensures that even in unforeseen circumstances, the system can flexibly adjust task allocation, optimize resource utilization, and prevent a single point of failure from causing the entire task chain to stall.

[0121] This solution tightly integrates real-time monitoring, intelligent detection, and dynamic adjustment, enabling the warehouse inbound management system to exhibit high adaptability and robustness in the face of complex and ever-changing real-world operating environments. The digital twin model provides global real-time situational awareness, the D*Lite algorithm offers efficient path correction capabilities, and task regrouping provides flexible task scheduling strategies. This organic combination of technologies allows the system to quickly respond to and effectively resolve unexpected problems during task execution, ensuring the continuity and efficiency of inbound tasks and significantly improving the overall operational efficiency and reliability of the warehouse.

[0122] Example 2:

[0123] Figure 4 A schematic diagram of an intelligent warehouse inbound management device for furniture products is shown in an embodiment of the present invention. The management device is used to execute the management method and includes:

[0124] Model building module 10: Used to build a real-time digital twin model that is updated synchronously with the physical warehouse, extract multi-dimensional features of each batch of products to be put into the warehouse, and encode physical attribute features, storage strategy features, operation features and context information features into feature vectors input to the real-time digital twin model;

[0125] Path planning module 20: is used to determine several target storage locations based on the product data of the current main batch of goods entering the warehouse, and to perform dynamic path planning in the digital twin model with the goal of minimizing the expected shelf completion time based on the target storage locations and the feature vector, so as to generate the optimal expected path;

[0126] Scheme generation module 30: used to obtain the physical attributes of the main batch of products, calculate the free space of the transport vehicle in combination with the rated capacity of the transport vehicle, and generate an initial loading scheme based on the physical attributes of the main batch of products and the free space of the transport vehicle in combination with the optimal expected path.

[0127] Scheme optimization module 40: Used as constraints, the optimal expected path and the free space of the transport vehicle are used to retrieve matching sub-batch from several batches to be scheduled in real time, dynamically add the sub-batch to the current transport task to form a merged task group, and optimize and adjust the optimal expected path and the initial loading scheme according to the cargo location information of the sub-batch to generate an optimized loading scheme.

[0128] Scheme adjustment module 50: used to allocate the optimized loading scheme to the transport vehicle, and to monitor and dynamically adjust the transport vehicle in real time through the digital twin model during the task execution process.

[0129] By constructing a real-time digital twin model that is synchronized with the physical warehouse and extracting multi-dimensional features from each batch to be received, physical attribute features, storage strategy features, operational features, and contextual information features are encoded into a unified feature vector. In practical applications, digital twin models can be constructed in various ways. For example, a basic two-dimensional or three-dimensional warehouse layout map can be built, including fixed shelf locations, road networks, and site information. For feature extraction of batches to be received, operators can manually input information such as batch size, weight, and target storage area, which is then recorded in a database. This information can then be easily combined into a single data record as a feature representation of the batch.

[0130] Secondly, for the main batch that needs to be scheduled, dynamic path planning is performed in the digital twin model based on its target location and feature vector, with the objective of minimizing the estimated shelving completion time. This generates the optimal expected path, and an initial loading plan including available vehicle space is generated based on the physical attributes of the main batch and the rated capacity of the transport vehicle. During dynamic path planning, the time-varying A* algorithm can be used to find the shortest path on a static warehouse map. This algorithm primarily considers the geometric length of the path when calculating path costs. When generating the initial loading plan, the space occupied by the main batch can be subtracted from the total capacity of the transport vehicle based on the volume and weight of the main batch, thus obtaining the remaining available space.

[0131] Secondly, using the optimal expected route and vehicle free space as dual hard constraints, matching secondary batches are retrieved in real time from other batches awaiting scheduling. These secondary batches are dynamically added to the current transportation task to form a merged task group. The optimal expected route and initial loading plan are then optimized and adjusted based on the cargo location information of the newly added batches. When retrieving matching secondary batches, dispatchers can manually view the current list of batches awaiting entry and, based on experience, determine which batches have target cargo locations near the main batch's route and whose physical attributes can accommodate vehicle free space. Once the secondary batches are identified, dispatchers can manually add them to the current transportation task. Subsequently, the routes of all batches in the merged task group can be recalculated, and the loading order adjusted according to the new batch combinations.

[0132] Finally, the merged task groups are assigned to transport vehicles. During task execution, a digital twin model is used for real-time monitoring and dynamic adjustments. After task completion, the model parameters for route planning, material matching, and task allocation are optimized in a closed-loop manner based on actual execution data. When assigning tasks, they can be simply assigned to the nearest available vehicle. During task execution, if the digital twin model detects path congestion or equipment failure, operators can manually intervene to replan the route or adjust the task. After task completion, the actual execution time can be compared with the planned time, and the differences can be manually analyzed to allow for empirical adjustments in future planning.

[0133] This invention provides an intelligent warehouse inbound management device for furniture products. By constructing a digital twin model and coordinating route planning, it enables the combined transportation of multiple batches of products, improving the rationality of product transportation allocation for a single inbound shipment, thereby increasing the efficiency of product inbound transportation.

[0134] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc.

[0135] Furthermore, the embodiments of the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. An intelligent warehouse entry management method for a furniture product, characterized by, The management method comprises: constructing a real-time digital twin model updated synchronously with a physical warehouse, extracting multi-dimensional features of each batch of products to be warehoused, and encoding physical attribute features, storage strategy features, operation features, and context information features into a feature vector input into the real-time digital twin model; determining a plurality of target storage locations according to product data of a main batch currently being warehoused, and performing dynamic path planning in the digital twin model based on the target storage locations and the feature vector, with the objective of minimizing an estimated completion time of shelving, to generate an optimal expected path; acquiring physical attributes of the main batch of products, calculating free space of a transport vehicle in combination with a rated capacity of the transport vehicle, and generating an initial loading scheme in combination with the optimal expected path based on the physical attributes of the main batch of products and the free space of the transport vehicle; taking the optimal expected path and the free space of the transport vehicle as constraint conditions, retrieving matching slave batches from a plurality of batches to be dispatched in real time, dynamically adding the slave batches to a current transport task to form a merged task group, and optimizing and adjusting the optimal expected path and the initial loading scheme based on storage location information of the slave batches to generate an optimized loading scheme; allocating the optimized loading scheme to the transport vehicle, and performing real-time monitoring and dynamic adjustment on the transport vehicle through the digital twin model during task execution; the method comprises: acquiring product data of the main batch, and setting product types and quantities of a single transport of the transport vehicle within the main batch in combination with transport space data of the transport vehicle; determining a plurality of target storage locations according to the product types and quantities of the single transport, and planning a plurality of preliminary transport paths in the digital twin model based on the plurality of target storage locations and the feature vector; By time-varying A The algorithm divides the running time of single-product delivery into discrete time slices, and establishes a time-varying cost function for each path segment, which combines the benchmark travel time, the time slice congestion coefficient and the obstacle penalty time. generating an optimal expected path based on the time-varying cost function; the method comprises: taking the optimal expected path as the core, extracting storage areas along and adjacent to the optimal expected path to form a geographical retrieval range; taking the free space of the transport vehicle and the geographical retrieval range as hard constraints, and selecting products meeting the free space of the transport vehicle and the geographical retrieval range from a plurality of batches to be dispatched according to a preset priority rule, and marking the batches to be dispatched in which the products are located as slave batches.

2. The management method according to claim 1, characterized in that, the method comprises: acquiring static information of storage rack coordinates, road networks, and site locations of the physical warehouse, and constructing a static layout layer of the physical warehouse based on the static information; real-time detecting and updating inventory status, traffic flow of road networks, device working status, and temporary obstacle information of the physical warehouse, and constructing a dynamic layout layer of the physical warehouse; construct a real-time digital twin model of the physical warehouse based on the static layout layer and the dynamic layout layer.

3. The management method according to claim 1, characterized by, The physical property features include size data, volume data, and weight data of the product; The storage strategy features include target shelf, shelf type, and storage priority data of the product; The operation features include estimated operation duration and transfer operation tool data; The context information features include product warehousing time and associated order information.

4. The management method according to claim 1, characterized by, The screening and matching from the several to-be-scheduled batches according to the idle space of the transport vehicle and the geographic search range as hard constraints includes: Taking the geographic search range as a screening standard, screening a path area matching batch in which a target shelf is located within the geographic search range from the several to-be-scheduled batches; In the path area matching batch, taking the storage strategy of the main batch product as a screening standard to screen a storage matching batch compatible with the storage strategy of the main batch product; In the batch that meets the storage strategy compatibility, taking the idle space of the transport vehicle as a screening standard to screen a slave batch that optimally fills the idle space of the transport vehicle.

5. The management method according to claim 1, characterized by, The dynamic addition of the slave batch to the current transport task to form a merged task group, and the optimization and adjustment of the optimal expected path and the initial loading scheme according to the shelf information of the slave batch to generate an optimized loading scheme include: According to the shelf information of the slave batch and the shelf information of the main batch, obtaining target shelf information of the current transport task; Obtaining the arrangement order of the target shelf information in the optimal expected path, and determining the loading order of all batch products in the merged task group according to the arrangement order and the principle of first-in first-out; According to the target shelf information and the loading order, an adjusted optimized loading scheme is generated.

6. The management method according to claim 1, characterized by, The assignment of the optimized loading scheme to the transport vehicle, and the real-time monitoring and dynamic adjustment of the transport vehicle through the digital twin model during task execution include: According to the transport vehicle type, the current position, the power, and the current load state, a multi-agent reinforcement learning model is used for calculation to generate a dispatching instruction; According to the dispatching instruction, the optimized loading scheme is assigned to the transport vehicle; Obtaining the real-time transport state of the transport vehicle and the obstacle information of the optimal expected path, detecting whether there is a task delay situation for the transport vehicle, and if there is a task delay situation, dynamically adjusting the transport state of the transport vehicle.

7. The management method according to claim 6, characterized in that, The dynamic adjustment of the transport state of the transport vehicle includes: During the task execution, if it is detected by the digital twin model that the transport vehicle is in a task delay situation, local path replanning by D Lite algorithm is triggered.

8. An intelligent warehouse entry management device for a furniture product, characterized by, The management device is used to execute the management method of any one of claims 1 to 7, and the management device includes: A model construction module is configured to construct a real-time digital twin model that is synchronously updated with a physical warehouse, extract multi-dimensional features of each to-be-warehoused batch product, and encode physical property features, storage strategy features, operation features, and context information features as feature vectors input to the real-time digital twin model. The path planning module is configured to determine a plurality of target storage locations according to product data of a current inbound main batch, perform dynamic path planning in the digital twin model based on the target storage locations and the feature vector, and generate an optimal expected path with a minimum expected shelving completion time as a target; The scheme generation module is configured to obtain physical attributes of the main batch of products, calculate free space of the transport vehicle based on a rated capacity of the transport vehicle, and generate an initial loading scheme based on the physical attributes of the main batch of products, the free space of the transport vehicle, and the optimal expected path; The scheme optimization module is configured to retrieve a matching slave batch from a plurality of batches to be dispatched in real time based on the optimal expected path and the free space of the transport vehicle as constraint conditions, dynamically add the slave batch to a current transport task to form a merged task group, and optimize and adjust the optimal expected path and the initial loading scheme based on storage location information of the slave batch to generate an optimized loading scheme; The scheme adjustment module is configured to assign the optimized loading scheme to the transport vehicle, and monitor and dynamically adjust the transport vehicle in real time based on the digital twin model during task execution.

Images