Warehouse operation method and system based on storage location and picking path integrated optimization

Abstract

The invention relates to the technical field of intelligent warehouse management, in particular to a warehouse operation method and system based on storage location and picking path integrated optimization, and the method comprises the steps: building a dynamic storage location distribution optimization model with the minimization of the total space waste value of a whole warehouse as an optimization target based on an initialized storage layout parameter set; performing iterative solution on the dynamic storage location allocation optimization model to obtain an initial tray storage location allocation scheme; according to the real-time order information and the initial tray storage location distribution scheme, generating an optimal order picking path of each order; and carrying out storage location secondary allocation on the remaining material storage location according to the warehouse remaining material storage state after the optimal order picking path is executed to obtain a storage location reallocation scheme, thereby generating a storage location-path integrated operation instruction, and controlling the warehouse management system to operate through the storage location-path integrated operation instruction. According to the invention, through collaborative optimization of the storage location allocation and the picking path, integrated collaborative optimization of the storage layout, the storage location allocation and the picking path is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent warehouse management, and in particular to a warehouse operation method and system based on integrated optimization of storage locations and picking paths. Background Art

[0002] In the field of warehouse management, the rapid development of the Internet of Things (IoT), artificial intelligence (AI), and automation technologies are driving the innovation of warehouse management models. However, existing technologies still have some problems in warehouse management.

[0003] Traditional warehouse management mostly adopts a fixed storage mode, that is, each inventory unit (SKU, Stock Keeping Units are assigned specific storage locations. This approach results in low space utilization, slow response, and difficulty adapting to dynamic market demands. Although the concept of random storage has improved space utilization and access speed, random storage requires a powerful information system to track and manage each inventory unit in the warehouse and lacks efficient dynamic optimization methods. This is particularly evident in the coordinated optimization of storage location allocation and picking paths. However, storage location allocation and picking path planning are often treated separately, lacking an integrated optimization model. Existing methods typically optimize storage space utilization or picking path length independently, ignoring the dynamic relationship between the two. This leads to frequent idle space and repeated path intersections in actual operations. Secondly, while random storage can improve space utilization, the initially optimized storage location layout gradually becomes ineffective as inventory changes dynamically. Existing relocation strategies often focus on local adjustments and lack a global, periodic optimization mechanism, resulting in cumulative efficiency losses. Furthermore, routes generated by traditional path planning algorithms (such as the traveling salesman problem (TSP) solution) often include complex turns and reverse movements, which increases the cognitive load on pickers and reduces actual operational efficiency.

[0004] In summary, existing technologies have difficulty balancing storage density and operational fluency in warehouse management, and are unable to meet the real-time response requirements of large-scale warehouses in actual applications, becoming a bottleneck restricting the development of intelligent warehousing. Therefore, there is an urgent need for new solutions that can deeply integrate dynamic storage optimization and humanized path planning to systematically improve the overall efficiency of warehousing operations. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention provides a warehouse operation method and system based on the integrated optimization of storage locations and picking paths.

[0006] In a first aspect, the present invention provides a warehouse operation method based on integrated optimization of storage locations and picking paths, the method comprising the following steps: Initialize the warehouse layout according to the warehouse physical structure data and pallet attribute data, and generate the initialization warehouse layout parameter set; Based on the initialized warehouse layout parameter set, a dynamic storage space allocation optimization model is constructed with minimizing the total space waste value of the entire warehouse as the optimization goal; Iteratively solving the dynamic storage location allocation optimization model to obtain an initial pallet storage location allocation plan under the current warehouse layout; In response to the real-time order information, an optimal picking path for each order is generated using a picking path optimization algorithm based on attention learning according to the real-time order information and the initial pallet storage allocation plan; According to the storage status of the remaining materials in the warehouse after the execution of the optimal picking path, the layout data of the remaining materials to be optimized is evaluated, and the storage locations of the remaining materials are secondary allocated according to the layout data of the remaining materials to be optimized to obtain a storage location reallocation plan; A storage location-path integrated operation instruction is generated according to the optimal picking path and the storage location reallocation plan, and the operation of the warehouse management system is controlled by the storage location-path integrated operation instruction.

[0007] In a further embodiment, the step of constructing a dynamic storage space allocation optimization model with minimizing the total space waste value of the entire warehouse as the optimization goal based on the initialized storage layout parameter set includes: Define a pallet stacking identification variable and a single pallet storage location allocation variable; wherein the single pallet storage location allocation variable is used to indicate that a single pallet is stored in a storage location alone, and the pallet stacking identification variable is used to indicate that two different pallets are stored in the same storage location; According to the initialized warehouse layout parameter set and the predetermined waste value measurement rule, the total warehouse space waste value of all pallet storage location storage combinations is calculated using the pallet stacking identification variable and the single pallet storage location allocation variable; Taking minimizing the total space waste value of the entire warehouse as the optimization goal, a dynamic storage space allocation optimization model is constructed.

[0008] In a further embodiment, the calculation process of the total space waste value of the entire warehouse is specifically as follows: When two pallets to be stored are stored in the same storage location, the space waste generated when the pallets to be stored are stacked in the same storage location is calculated based on the sum of the heights of the pallets to be stored in the same storage location, and the double pallet stacking waste value of all pallets to be stored is obtained; When a single pallet to be stored is stored in a single pallet storage location, if it is detected that the single pallet to be stored does not allow another pallet to be stored to be placed above it, a restrictive space waste value is obtained according to the sum of the differences between the heights of all single pallet storage locations and the height of the single pallet to be stored; Based on the pre-acquired single pallet space waste coefficient, the single pallet space waste value is measured when a pallet to be stored is already stored at a position where two pallets can be placed and the remaining height is too low to accommodate another pallet to be stored. The weighted sum of the double pallet stacking waste value, the restrictive space waste value and the single pallet space waste value is calculated using the pallet stacking identification variable and the single pallet storage location allocation variable to obtain the total space waste value of the entire warehouse.

[0009] In a further embodiment, the constraints of the dynamic storage location allocation optimization model include storage restriction constraints for each storage location, pallet storage feasibility constraints, space restriction constraints, level restriction constraints, and stacking rule restriction constraints; The storage restriction constraint of each storage location is specifically that the number of pallets stored in each storage location does not exceed its maximum capacity limit; The pallet storage feasibility constraint is specifically that each pallet is stored in only one storage location; The space limitation constraint is specifically that the total height of all pallets stored in each storage location does not exceed the preset maximum storage location height; The layer restriction constraint specifically means that the number of pallet storage layers does not exceed the maximum number of layers allowed; The stacking rule restriction specifically states that if two pallets are not allowed to be placed on top of another pallet, the two pallets cannot be stored in the same storage location.

[0010] In a further embodiment, the step of iteratively solving the dynamic storage location allocation optimization model to obtain an initial pallet storage location allocation plan under the current warehouse layout includes: Scan the warehouse layer by layer from the bottom to the top. During the scanning process, determine the set of pallets to be stored according to the initialized warehouse layout parameter set, and set the current level to the bottom of the warehouse. The current level is used as the level to be processed, and the pallets in the level to be processed are selected from the set of pallets to be stored and sorted in descending order by pallet height to build a list of pallets to be processed; Filter out all vacant storage locations in the level to be processed and sort them in descending order by storage location height to generate a list of available locations in the level to be processed; Based on the list of pending pallets and the list of available locations, the dynamic storage location allocation optimization model is solved using a pallet location allocation algorithm to obtain a primary storage location allocation solution; The primary storage location allocation solution is optimized for pallet location reallocation based on a local search strategy to obtain an initial pallet storage location allocation solution under the current warehouse layout.

[0011] In further embodiments, the step of solving the dynamic storage allocation optimization model based on the list of to-be-processed pallets and the list of available locations using a pallet location allocation algorithm to obtain a primary storage allocation solution comprises: initializing an un-stowed pallet set as an empty set, and when the list of to-be-processed pallets is not empty, performing a pallet allocation operation according to the constraint conditions of the dynamic storage allocation optimization model, and sequentially detecting the pallets in the list of to-be-processed pallets; if it is detected that there is a pallet in the list of to-be-processed pallets whose height is greater than the height of the first storage location in the list of available locations, moving it from the list of to-be-processed pallets to the un-stowed pallet set, and continuing to detect the next pallet in the list of to-be-processed pallets; selecting the pallet with the maximum height from the list of to-be-processed pallets as a reference pallet, and searching for a pallet with the second maximum height from the list of to-be-processed pallets, if the sum of the heights of the reference pallet and the pallet with the second maximum height does not exceed the height of the first storage location in the list of available locations, and at least one pallet allows another pallet to be placed above it, then storing the reference pallet and the pallet with the second maximum height as a paired pallet in the lowest available storage location; otherwise, only storing the reference pallet in the lowest available storage location; if the selected lowest available storage location is a double-depth storage location, searching for other pallets or pallet pairs that allow stacking in the list of to-be-processed pallets to fill the remaining space of the double-depth storage location; deleting the selected paired pallets from the list of to-be-processed pallets and the set of to-be-stowed pallets, and removing the used available storage locations from the list of available locations; if the un-stowed pallet set is empty and the to-be-processed level is not the bottom level, then re-inserting the pallets in the un-stowed pallet set into the list of to-be-processed pallets in descending order of pallet height, adjusting the current level to the previous level, and regenerating the list of available locations; if the un-stowed pallet set is empty and the to-be-processed level is the bottom level, selecting the pallet with the minimum height from the un-stowed pallet set and storing it in the remaining empty storage locations at the current level or below; if the current level is less than the maximum number of levels and the un-stowed pallet set is empty, updating the to-be-processed level to the next level, and re-selecting the to-be-processed pallets at the current level and the list of available locations, and continuing the pallet allocation until all pallets satisfy the constraint conditions of the dynamic storage allocation optimization model, stopping the iteration, and outputting the primary storage allocation solution.

[0012] In further embodiments, the step of performing pallet location re-allocation optimization on the primary storage allocation solution based on a local search strategy to obtain an initial pallet storage allocation scheme under the current warehouse layout comprises: Calculate the storage location waste value of each storage location in the primary storage location allocation solution according to the storage location type and the pallet stacking status, and sort all storage locations in the primary storage location allocation solution in descending order according to the storage location waste value to generate a list of storage locations to be optimized; Traversing the list of storage locations to be optimized, and taking each storage location in the list of storage locations to be optimized as a source location, detecting feasible moves in sequence according to the four pre-established movement type priorities; If a feasible move is detected, the feasible move is executed in descending order of waste reduction magnitude, and the pallet position and storage location status are updated in real time based on the storage location information of the feasible move; After each round of feasible moves, the total space waste value of the entire warehouse is recalculated. If the total space waste value of the entire warehouse is greater than the preset waste threshold, a new list of storage locations to be optimized is generated based on the updated storage location status, and the above steps are repeated, and the optimization is iterated continuously until the total space waste value of the entire warehouse is less than the preset waste threshold. The iterative optimization is terminated, and the initial pallet storage location allocation plan under the current warehouse layout is output.

[0013] In a further embodiment, the step of generating the optimal picking path for each order using an attention learning-based picking path optimization algorithm based on the real-time order information and the initial pallet storage allocation plan includes: Obtaining a list of pallets to be picked according to the real-time order information, and extracting the current warehouse layout and storage location information of the pallets to be picked from the initial pallet storage location allocation plan based on the list of pallets to be picked; The current warehouse layout is converted into a graph structure based on the storage location information of the to-be-picked pallets to obtain a warehouse topology graph; wherein the graph nodes of the warehouse topology graph include the storage location nodes of the to-be-picked pallets, the top and bottom nodes of the shelf aisles, and the warehouse entrance and exit nodes, and the edges represent the traversable path relationships between the graph nodes; An attention mechanism is used to learn the circuit vertical edges and the circuit horizontal edges from the warehouse topology graph, and a circuit graph is constructed according to the circuit vertical edges and the circuit horizontal edges in sequence; the attention mechanism includes an input layer, an embedding layer, a multi-head attention mechanism layer, a feedforward layer, a residual connection and a layer normalization layer, and a linear layer; All effective picking routes starting from the warehouse entrance node, passing through each to-be-picked pallet storage node and arriving at the warehouse exit node are extracted from the tour graph, and the effective picking route with the minimum total cost is selected as the optimal picking path.

[0014] In a further embodiment, the step of performing a secondary allocation of storage locations of the remaining materials according to the remaining material layout data to be optimized to obtain a storage location reallocation plan includes: Generate a storage movement list based on the remaining material layout data to be optimized, and define a maximum allowed execution time; The storage movement operations in the storage movement list are sequentially executed according to preset movement priorities, and a random number is generated during execution of each storage movement, and a number of storage movements equal to the random number are randomly selected from the storage movement list for deletion; The movement execution time of each storage movement is calculated during execution of each storage movement, and the movement execution time of each movement is accumulated, and when the movement execution time accumulation value reaches the maximum allowed execution time, the execution of the storage movement is stopped, and the currently executed storage movement operation constitutes a storage movement operation scheme; According to the total waste reduction amount of the storage movement operation scheme, the optimal storage space redistribution scheme is selected from all storage movement operation schemes.

[0015] In a second aspect, the present application provides a warehouse operation system based on storage space and picking path integrated optimization, the system comprises: A warehouse initialization module is configured to initialize the warehouse layout according to warehouse physical structure data and pallet attribute data, and generate an initialization warehouse layout parameter set; A storage space model construction module is configured to construct a dynamic storage space allocation optimization model with the optimization objective of minimizing the total warehouse space waste value based on the initialization warehouse layout parameter set; A storage space model solving module is configured to iteratively solve the dynamic storage space allocation optimization model to obtain an initial pallet storage space allocation scheme under the current warehouse layout; A picking path analysis module is configured to generate an optimal picking path for each order by using an attention learning-based picking path optimization algorithm according to real-time order information and the initial pallet storage space allocation scheme in response to the real-time order information; A storage space redistribution module is configured to evaluate to obtain remaining material layout data to be optimized according to the remaining material storage state of the warehouse after execution of the optimal picking path, perform secondary allocation of storage space for the remaining material storage location according to the remaining material layout data to be optimized, and obtain a storage space redistribution scheme; A warehouse operation module is configured to generate storage space-path integrated operation instructions according to the optimal picking path and the storage space redistribution scheme, and control the warehouse management system operation through the storage space-path integrated operation instructions.

[0016] The application provides a warehouse operation method and system based on integrated optimization of storage locations and picking paths. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 FIG. 1 is a flowchart of a warehouse operation method based on integrated optimization of storage locations and picking paths provided by an embodiment of the application; Figure 2 FIG. 3 is a schematic view of a shelf structure with three columns and four layers provided by an embodiment of the application; Figure 3 FIG. 4 is a side view of a shelf structure with three columns and four layers provided by an embodiment of the application; Figure 4 FIG. 6 is a process block diagram of a warehouse operation based on integrated optimization of storage locations and picking paths provided by an embodiment of the application; Figure 5 FIG. 7 is a schematic view of a warehouse layout after storage location allocation provided by an embodiment of the application; Figure 6 FIG. 8 is a schematic view of a picking route of an S-shaped picking strategy provided by an embodiment of the application; Figure 7 FIG. 9 is a schematic view of a picking route of a return picking strategy provided by an embodiment of the application; Figure 8 FIG. 10 is a schematic view of a picking route of a maximum gap heuristic strategy provided by an embodiment of the application; Figure 9 FIG. 11 is a schematic view of a picking route of a composite picking strategy provided by an embodiment of the application; Figure 10is a schematic diagram of an attention mechanism structure provided by an embodiment of the present application. Figure 11 is a warehouse operation system block diagram based on integrated optimization of storage locations and picking paths provided by an embodiment of the present application.

[0018] Label explanation: 101, warehouse initialization module; 102, storage location model construction module; 103, storage location model solving module; 104, picking path analysis module; 105, warehouse location reallocation module; 106, warehouse operation module. DETAILED DESCRIPTION

[0019] The embodiments of the present application will be described in detail below with reference to the accompanying drawings. The embodiments are given only for illustrative purposes and cannot be understood as limiting the present application. The accompanying drawings are only for reference and illustration and do not constitute a limitation on the scope of patent protection of the present application, because many changes can be made to the present application without departing from the spirit and scope of the present application.

[0020] Figure 1 is a warehouse operation method flowchart based on integrated optimization of storage locations and picking paths provided by an embodiment of the present application. The present application provides a warehouse operation method based on integrated optimization of storage locations and picking paths, as shown in the figure, the method comprises the following steps: Figure 1 S1. Initialize the warehouse layout according to the warehouse physical structure data and the pallet attribute data, and generate an initialization warehouse layout parameter set.

[0021] For the storage needs of multiple categories of materials in a flexible warehouse, the present embodiment can use standardized pallets as cargo carriers for warehouse-in and warehouse-out operations. The bottom surface size of all pallets is uniformly set to 150 cm x 150 cm, but the pallet height varies. The warehouse physical structure is composed of multiple columns of storage locations, each column uses single-depth or double-depth shelving configuration, and the bottom surface size of each storage location is the same as the bottom surface size of the pallet and leaves a few centimeters of operation allowance, for example, the operation allowance can be set to 3 cm, the specific value is determined according to the actual operation space requirement and operation convenience, to ensure that the operator has enough space to operate during the pallet storage and removal process. Therefore, the width of all storage locations is the same, but the height is designed differently according to the actual storage requirements and pallet height characteristics, to achieve efficient use of storage space.

[0022] ​In this embodiment, the storage locations are identified by column-layer two-dimensional coordinates, supporting the mixed placement of multiple categories of goods. At the same time, the storage locations are divided into single-depth storage locations and double-depth storage locations according to the depth design of the shelf. The single-depth storage location is installed on the single-depth shelf. The single-depth storage location can store up to two pallets, which are stacked up and down and are defined as upper position 2 and lower position 1 respectively; the double-depth storage location is installed on the double-depth shelf. The double-depth shelf is formed by installing one shelf behind another shelf without a backboard. The double-depth storage location can store up to four pallets in a front-to-back double-layer stacking structure, with the four pallets being position 1 on the front lower layer, position 2 on the front upper layer, position 3 on the rear lower layer, and position 4 on the rear upper layer.

[0023] For shelves viewed from the front, Figure 2 This is a schematic diagram of a shelf structure with three columns and four layers provided by an embodiment of the present invention; for back-to-back double-depth shelves, Figure 3 This is a side view of a shelf structure with three columns and four layers provided by an embodiment of the present invention. The shelf has four layers, and each storage position can accommodate four pallets. On this basis, this embodiment can determine the correspondence between the storage positions and the shelf layout, clarify the shelf, column and layer information to which each storage position belongs, and at the same time, determine the operating rules for accessing the pallets stored in the rear positions 3 and 4. When the pallet stored in the front position 1 is not empty, the pallets stored in positions 1 and 2 must be emptied first before the pallets in positions 3 and 4 can be accessed; similarly, if a pallet is to be placed in the rear position 3 or position 4, the pallet in the front position 1 or 2 must be emptied first.

[0024] It should be noted that the single-depth shelf storage scenario is suitable for high-frequency access and low-density storage scenarios. In high-frequency access scenarios, for pallets that are not allowed to be stacked, such as fragile items and high-value materials, the waste value of the difference between the position height and the pallet height is calculated to avoid invalid space occupation and ensure access efficiency; in low-density access scenarios, when pallets are allowed to be stacked but the remaining height is less than the minimum pallet height, the waste value is marked as 0, and the rule of not allowing stacking on top is followed to meet strict constraints such as fire prevention and load-bearing capacity; at the same time, the double-depth shelf storage scenario is suitable for high-density storage and mixed storage scenarios. In high-density storage scenarios, due to the large shelf depth, the double-depth position is split into independent single-depth positions to calculate the waste value. For double-depth storage positions, the front depth (position 1 and position 2) and the back depth (position 3 and position 4) are calculated separately. ) to maximize space utilization, while following the access order of the front and rear pallets, that is, the front-depth pallet must be emptied before accessing the rear-depth pallet. In mixed storage scenarios, if some pallets are allowed to be stacked and some are not, when different types of pallets are stored in the same double-depth position, the upper space waste value of each depth is calculated separately according to the single pallet rule to ensure storage compliance; in addition, general constraint scenarios cover layer height restrictions and compliance storage scenarios. In the layer height restriction scenario, over-limit storage is avoided by verifying that the product of the number of layers and the pallet height is not greater than the storage location height; in the compliance storage scenario, for pallets that are not allowed to be stacked due to batch requirements, the upper space is forcibly released through waste value calculation, that is, when such pallets are detected, the waste value of the space above them is calculated and they are marked as unavailable to avoid illegal stacking and meet storage needs.

[0025] Then, this embodiment collects the physical structure data, pallet attribute data and real-time storage status data of the warehouse in real time through the warehouse management system, and initializes the warehouse layout according to the physical structure data, pallet attribute data and real-time storage status data of the warehouse to generate an initialization warehouse layout parameter set. Figure 4This is a flowchart of the warehouse operation process based on the integrated optimization of storage locations and picking paths provided by an embodiment of the present invention. In this embodiment, the initialization warehouse layout parameter set includes the physical coordinates, storage capacity and initial occupancy status of each storage location; the warehouse physical structure data includes the unique storage location number assigned to each single-depth and double-depth location, the storage location height of the single-depth storage location and the double-depth storage location, the number of storage location layers and the type of shelf used to identify the single and double depth of each storage location, etc. The unique storage location number is used to identify each storage location; for the number of storage location layers, the single-depth storage location is fixed to two layers, and the double-depth storage location is fixed to two layers each in the front depth and the rear depth; the pallet attributes The data includes pallet number, pallet height, maximum allowable stacking layers, stacking limit mark and minimum pallet height, etc. The minimum pallet height is used to determine whether the remaining space can store another pallet; the real-time storage status data includes real-time storage relationship and space occupancy data, among which the real-time storage relationship includes the pallet storage position binary variable and the pallet stacking relationship binary variable. The storage position binary variable is used to track the binding relationship between the pallet and the storage location; the pallet stacking relationship binary variable is used to identify the stacking order within the same storage location; the space occupancy data includes the used height of the single-depth storage location and the used height of the front and rear depths of the double-depth storage location.

[0026] When a pallet is put on, taken off, or moved, the waste value of the corresponding storage location is triggered to be recalculated. For single-depth storage locations, the waste value is calculated according to whether one or two pallets are stored and whether stacking is allowed, according to the preset waste value measurement rule formula; for double-depth storage locations, the waste values ​​of the front and rear depths are calculated separately, and the sum of the two is used as the total waste value of the storage location. By calculating the waste value in real time, data support is provided for subsequent storage location allocation and picking path optimization.

[0027] S2. Based on the initialized warehouse layout parameter set, a dynamic storage location allocation optimization model is constructed with minimizing the total space waste value of the entire warehouse as the optimization goal.

[0028] In some embodiments, the step of constructing a dynamic storage space allocation optimization model with minimizing the total space waste value of the entire warehouse as an optimization goal based on the initialized storage layout parameter set includes: The step of constructing a dynamic storage space allocation optimization model based on the initialized storage layout parameter set with minimizing the total space waste value of the entire warehouse as the optimization goal includes: Define a pallet stacking identification variable and a single pallet storage location allocation variable; wherein the single pallet storage location allocation variable is used to indicate that a single pallet is stored in a storage location alone, and the pallet stacking identification variable is used to indicate that two different pallets are stored in the same storage location; According to the initialized warehouse layout parameter set and the predetermined waste value measurement rule, the total warehouse space waste value of all pallet storage location storage combinations is calculated using the pallet stacking identification variable and the single pallet storage location allocation variable; Taking minimizing the total space waste of the entire warehouse as the optimization goal, a dynamic storage space allocation optimization model is constructed.

[0029] In a specific embodiment, this embodiment determines a waste value measurement rule based on the storage location type and the pallet stacking state. The waste value measurement rule is specifically as follows: If the storage location is empty, the waste value is zero; If there is only one pallet stored in the storage location and another pallet is allowed to be placed above it, the waste value is zero when the difference between the storage location height and the height of the stored pallets is greater than or equal to the height of the lowest pallet; otherwise, the waste value is the storage location height minus the pallet height; If a single pallet i is stored in a storage location And it is not allowed to place another pallet on top of it, then the waste value is the storage height minus the pallet height; If two pallets i and k are stacked in the storage location In the example, the waste value is the storage location height minus the sum of the two pallet heights.

[0030] This embodiment treats each double-depth storage location as two independent single-depth storage locations to simplify the model complexity while retaining the access order constraints of the front and back depths. For any pallet and any location, this embodiment pre-calculates the single pallet space waste coefficient. , single pallet space waste coefficient Used to measure the waste when a pallet is already stored in a position that allows double pallets and there is not enough space to place another pallet. If the pallet does not allow another pallet to be placed above it or the pallet height plus the minimum pallet height is less than or equal to the storage height, the single pallet space waste coefficient is zero; otherwise, the single pallet space waste coefficient The space waste coefficient of a single pallet is the storage height minus the pallet height. The specific calculation is: =1 or +minc≤ ,but =0; otherwise = ,in, is the stacking limit identifier of pallet i, minc is the minimum pallet height, which represents the minimum value of all pallet heights and is used to determine whether the remaining space can store another pallet; is the height of tray i, For storage height.

[0031] This embodiment calculates the space waste generated when stacking pallets in the same storage location based on the sum of the heights of the pallets to be stored. Specifically, for any two pallets to be stored i and pallet to be stored k ( ), when the pallet i and the pallet k to be stored are stored together in the storage location When the space wasted by stacking the pallets to be stored is ,in, For storage height, is the height of the pallet i to be stored, is the height of the pallet k to be stored. This embodiment traverses all possible pallet combinations to obtain the double-pallet stacking waste value of each pallet to be stored. When a single pallet to be stored is stored in a single-pallet storage position, if it is detected that the pallet does not allow another pallet to be stored to be placed above it, the sum of the differences between the single-pallet storage position height and the single pallet to be stored height is calculated to obtain the restrictive space waste value. At the same time, this embodiment measures the single-pallet space waste value when a pallet to be stored is already stored in a position where double pallets are allowed to be placed and another pallet to be stored cannot be placed thereon based on the single-pallet space waste coefficient obtained in advance.

[0032] Then, this embodiment defines the pallet stacking identification variable and pallet location allocation variables Both are binary variables, among which the pallet stacking identification variable is used to indicate that two different pallets are stored in the same storage location. If pallet i and pallet k are stacked in the storage location , then the pallet stacking identification variable Take 1, otherwise the pallet stacking identification variable Take 0; the pallet storage allocation variable is used to indicate that a single pallet is stored in a storage location. If pallet i is stored in storage location , then the pallet storage allocation variable Take 1, otherwise the pallet storage allocation variable Take 0.

[0033] Next, this embodiment uses the pallet stacking identification variable and pallet location allocation variables The weighted sum of the space waste values ​​of the above three parts is performed to obtain the total space waste value of the entire warehouse. The dynamic storage space allocation optimization model is constructed with minimizing the total space waste value as the optimization goal. In this embodiment, the dynamic storage space allocation optimization model is modeled as an integer linear programming problem, and its objective function is as follows: Where, For storage The height of the storage space Vertical space capacity; is the height of pallet i, indicating the vertical space occupied by pallet i; is the height of pallet k, indicating the vertical space occupied by pallet k; is the pallet stacking identification variable, which is a binary variable. If pallet i and pallet k are stacked in the storage location In ,otherwise ; is the stacking limit flag of pallet i, which is a Boolean variable, where Indicates that pallet i is not allowed to be stacked on top of other pallets. Indicates that stacking is allowed; Assign variables to the pallet storage location, which are binary variables. If pallet i is stored alone in the storage location In ,otherwise ; is the single pallet space waste coefficient, which represents the storage space The waste value generated when pallet i is stored alone; m is the total number of storage locations, that is, the total number of all storage locations in the warehouse that can be used to store pallets; n is the total number of pallets; j is the storage location index; i is the pallet index; k is the pallet index, which usually represents another pallet stacked with pallet i, where .

[0034] The objective function of the dynamic storage location allocation optimization model is to minimize the total space waste value of the entire warehouse, where the first term is the sum of all double-pallet stacking waste values, the second term is the sum of all restrictive space waste values, and the third term is the sum of all single-pallet space waste values. The constraints of the dynamic storage location allocation optimization model include the storage restriction constraint of each storage location, the pallet storage feasibility constraint, the space restriction constraint, the hierarchical restriction constraint, and the stacking rule restriction constraint. The mathematical expression of the storage restriction constraint of each storage location is: In the embodiment, the storage limit constraint of each storage location is specifically that the number of trays stored in each storage location does not exceed the maximum capacity limit of the storage location, ensuring that each storage location stores at most one tray or a pair of stacked trays. Specifically, a single-depth storage location stores at most two trays, and the front and rear depths of a double-depth storage location are independently limited as single-depth storage locations.

[0035] The mathematical expression of the tray storage feasibility constraint is: The tray storage feasibility constraint is specifically that each tray is stored in only one storage location, that is, each tray must be and can only be assigned to one storage location, ensuring that all trays can be reasonably stored.

[0036] The mathematical expression of the space limit constraint is: The space limit constraint is specifically that the total height of all trays stored in each storage location does not exceed the preset maximum storage location height, ensuring that the tray storage does not exceed the storage location space.

[0037] The mathematical expression of the level limit constraint is: The level limit constraint is specifically that the number of storage levels of a tray does not exceed the maximum number of allowed storage levels , complying with the storage level requirements of the tray.

[0038] The mathematical expression of the stacking rule limit constraint is: The stacking rule limit constraint is specifically that if two trays are not allowed to have another tray placed above them, the two trays cannot be stored in the same storage location, ensuring that the tray stacking complies with the rules.

[0039] The mathematical expression of the variable definition constraint is: In the formula, is the number of storage levels of the storage location , wherein a single-depth storage location is fixed at 2 levels, and a double-depth storage location has 2 levels in front and back, i.e., a double-depth storage location has 4 levels in total; is the maximum number of storage levels allowed for tray i, indicating the highest level at which the tray can be stacked above other trays in the storage location; is the stacking limit identifier of tray k.

[0040] S3. Iteratively solve the dynamic storage allocation optimization model to obtain an initial tray storage allocation scheme under the current warehouse layout.

[0041] In some embodiments, the step of iteratively solving the dynamic storage location allocation optimization model to obtain an initial pallet storage location allocation solution under the current warehouse layout includes: Scan the warehouse layer by layer from the bottom to the top. During the scanning process, determine the set of pallets to be stored according to the initialized warehouse layout parameter set, and set the current level to the bottom of the warehouse. The current level is used as the level to be processed, and the pallets in the level to be processed are selected from the set of pallets to be stored and sorted in descending order by pallet height to build a list of pallets to be processed; Filter out all vacant storage locations in the level to be processed and sort them in descending order by storage location height to generate a list of available locations in the level to be processed; Based on the list of pending pallets and the list of available locations, the dynamic storage location allocation optimization model is solved using a pallet location allocation algorithm to obtain a primary storage location allocation solution; The primary storage location allocation solution is optimized for pallet location reallocation based on a local search strategy to obtain an initial pallet storage location allocation solution under the current warehouse layout.

[0042] In some embodiments, the step of solving the dynamic storage location allocation optimization model using a pallet location allocation algorithm based on the list of pallets to be processed and the list of available locations to obtain a primary storage location allocation solution includes: Initializing the set of unstored pallets to be an empty set, and when the list of pallets to be processed is not empty, performing a pallet allocation operation according to the constraint conditions of the dynamic storage allocation optimization model, and sequentially detecting the pallets in the list of pallets to be processed; If it is detected that there is a pallet in the list of pallets to be processed whose pallet height is greater than the height of the first storage location in the list of available locations, move the pallet from the list of pallets to be processed to the set of unstored pallets, and continue to detect the next pallet in the list of pallets to be processed; Select the pallet with the largest height from the list of pallets to be processed as the reference pallet, and search for the pallet with the second largest height from the list of pallets to be processed. If the sum of the heights of the reference pallet and the pallet with the second largest height does not exceed the height of the first storage location in the list of available locations, and at least one pallet can be placed above another pallet, then the reference pallet and the pallet with the second largest height are stored as paired pallets in the available storage location with the lowest height; otherwise, only the reference pallet is stored in the available storage location with the lowest height. If the selected available storage location with the lowest height is a double-depth storage location, searching for other pallets or pallet pairs that are allowed to be stacked in the list of pallets to be processed to fill the remaining space of the double-depth storage location; Deleting the selected paired pallet from the list of pallets to be processed and the set of pallets to be stored, and removing the used available storage locations from the list of available locations; If the unstored pallet set is empty and the to-be-processed level is not the bottom, the pallets in the unstored pallet set are reinserted into the to-be-processed pallet list in descending order of pallet height, the current level is adjusted to the previous level, and the available position list is regenerated; If the unstored pallet set is empty and the level to be processed is the lowest, the pallet with the smallest height is selected from the unstored pallet set and stored in the remaining vacant storage space at the current level or the level below it; If the current level is less than the maximum number of levels and the set of unstored pallets is empty, the level to be processed is updated to the next level, and the list of pallets to be processed and available locations at the current level is re-screened, and pallet allocation continues until all pallets meet the constraints of the dynamic storage allocation optimization model. The iteration is stopped and the primary storage allocation solution is output.

[0043] The pallet location allocation algorithm of this embodiment reserves low-level space for the storage of pallets with hierarchical restrictions by constructing an initial solution. To solve this problem, the construction process adopts a strategy of traversing layer by layer from bottom to top. For each given layer, this embodiment will give priority to arranging pallets that must be stored at this layer. Normally, pallets without hierarchical restrictions will be placed on the highest layer. During the hierarchical research process, in order to minimize space waste, this embodiment can only consider empty positions. Specifically, the empty positions in the current layer will be sorted from high to low according to height, and the pallets are also sorted according to this standard. When processing pallet storage, the first pallet in the sorted pallet list is first selected. The pallet location allocation algorithm will search for another pallet in this list. The pallet must meet two key conditions: First, the total height of the two pallets does not exceed the first empty position in the list. The algorithm then tries to find a pallet that has not yet been stacked, and the algorithm then tries to find another pallet that has not yet been stacked.

[0044] In a specific embodiment, the present embodiment first performs an initialization operation, initializing the set of pallets to be stored Notstored as the set of incoming pallets Incoming Pallets, setting the current level setting as the bottom level of the warehouse (current = 1), and setting the level variable h equal to the current level current, then the present embodiment constructs a pallet list to be processed PalletsList, which contains the pallets in the set of pallets to be stored Notstored with the maximum storage level h, and the pallets in the list are arranged in descending order of height, and the set of un-stored pallets Imp is initialized as an empty set, then the present embodiment constructs an available location list LocationsList, which contains all empty locations at level h, and the locations in the list are arranged in descending order of height, and then enters the loop processing pallet list phase, when the pallet list to be processed PalletsList is not empty, the following operations are performed: If the height of a pallet in the pallet list PalletsList is greater than the height of the first location in the location list LocationsList, the pallet is moved from the pallet list PalletsList to the not-stored pallet set Imp, and the first pallet in the pallet list PalletsList is selected. If there is a next pallet in the pallet list PalletsList such that the sum of the heights of the two pallets does not exceed the height of the first location in the location list LocationsList, and at least one of the pallets allows another pallet to be stacked above it, the next pallet is selected. The selected pallets (or pairs of pallets) are removed from the pallet list PalletsList and the not-stored pallet set Notstored. The selected pallets (or pairs of pallets) are stored in the empty locations in the location list LocationsList that have lower heights and can accommodate them. If the selected location supports the storage of four pallets, appropriate pallets or pairs of pallets are selected from the pallet list PalletsList and placed in the location to achieve the minimum waste of space utilization, and the selected pallets are removed from the pallet list PalletsList and the not-stored pallet set Notstored. The locations of the stored pallets are removed from the location list LocationsList, and the not-stored pallets are then processed. If the not-stored pallet set Imp is an empty set and the current processing level h is greater than 1, the pallets in the not-stored pallet set Imp are inserted into the appropriate locations of the pallet list PalletsList, the level variable h is reduced by 1 to move the level up and continue processing, and the location list is reconstructed. The last pallets are processed. If the not-stored pallet set Imp is an empty set and the current processing level reaches the bottom layer (h = 1), a pallet is selected from the not-stored pallet set Imp and attempts are made to store it in partially empty locations in the current level or lower levels to minimize space waste. The level variable h is set to the value of (current + 1) to continue processing by moving the level down, and the current variable is increased by 1. If the level variable h is less than the maximum number of layers, the pallet list is reconstructed. When all pallets satisfy the constraint conditions of the dynamic storage allocation optimization model (including height restrictions, stacking rules, etc.), the iteration is terminated, and the primary storage allocation solution is obtained through the above steps. The primary storage allocation solution includes but is not limited to the initial feasible solution of pallet location and stacking relationship. The embodiment effectively generates a primary storage allocation scheme that satisfies the minimum space waste minimization objective under the premise of ensuring storage compliance through systematic level scanning and dynamic allocation mechanisms, and provides a high-quality initial solution for subsequent optimization.

[0045] In some embodiments, the step of performing pallet location reallocation optimization on the primary storage location allocation solution based on a local search strategy to obtain an initial pallet storage location allocation solution under the current warehouse layout includes: Calculate the storage location waste value of each storage location in the primary storage location allocation solution according to the storage location type and the pallet stacking status, and sort all storage locations in the primary storage location allocation solution in descending order according to the storage location waste value to generate a list of storage locations to be optimized; Traversing the list of storage locations to be optimized, and taking each storage location in the list of storage locations to be optimized as a source location, detecting feasible moves in sequence according to the four pre-established movement type priorities; If a feasible move is detected, the feasible move is executed in descending order of waste reduction magnitude, and the pallet position and storage location status are updated in real time based on the storage location information of the feasible move; After each round of feasible moves, the total space waste value of the entire warehouse is recalculated. If the total space waste value of the entire warehouse is greater than the preset waste threshold, a new list of storage locations to be optimized is generated based on the updated storage location status, and the above steps are repeated, and the optimization is iterated continuously until the total space waste value of the entire warehouse is less than the preset waste threshold. The iterative optimization is terminated, and the initial pallet storage location allocation plan under the current warehouse layout is output.

[0046] In a specific embodiment, the pallet location allocation algorithm is mainly used to optimize the storage of pallets with hierarchical restrictions. However, when the pallet location allocation algorithm is used in this embodiment, the lower-level locations in the primary storage location allocation solution may not be fully utilized, resulting in space waste. To solve this problem, this embodiment uses a local search algorithm to continuously execute a set of movement operations selected from four specific types to reduce space waste. Each type must meet the target storage location capacity limit and pallet storage constraints. Three of the four movement types involve operations involving two pallets, and the remaining one involves an operation involving a single pallet. During the optimization process, all double-depth locations are treated as two independent single-depth locations. Specifically, the four movement types are: Movement type 1: Move two pallets stacked at a source location to a destination at the same location. Movement type 2: transfer a pallet from one source location to another destination location; Move type 3: Move two pallets at different locations to a new empty destination location and stack one pallet on top of the other. Movement type 4: A pallet that is stored alone at the source location and does not allow a pallet to be placed above it is exchanged with two pallets at the destination location that both allow a pallet to be placed above them.

[0047] When executing these movement operations, this embodiment only considers feasible movements that can change the position of the pallet in the incoming pallet set IncomingPallets, that is, the target position must have sufficient capacity to store the pallet and must meet the relevant restrictions on the pallet. At the same time, these movement operations will only be performed when they can reduce waste. The fourth type of movement also accepts movements that do not change the number of available centimeters, because such movements may make other types of improved movements possible in the future. In this embodiment, the local search algorithm first constructs a waste list of the source position and sorts it from high to low according to the degree of waste. If there are identical waste values, they are processed randomly. Subsequently, the local search algorithm The search algorithm loops through the list four times, searching for one type of move each time. When an improving move is identified for a source location, further research on that location is stopped and the move is stored. The location information involved in the move is marked and not further studied. The search for the target location is always performed on a list of possible target locations (which may be empty, fully occupied, or partially occupied) sorted in ascending order of height. If there are identical situations, they are also solved in a random manner. If improving moves are stored, they are executed and the search process is restarted. For each move type, the list of possible target locations consists of the unmarked locations that meet the move requirements. Specifically: For type 1 movement, the source position is selected from the first position in the list of possible source positions. After the source position is selected, the target position is selected from the first position in the list of possible target positions whose height is less than or equal to the source position; For type 2 moves, the source location is also selected from the first location in the list of possible source locations. If the selected source location has only one pallet, an empty target location is searched. If no improving move exists, the program continues to search for a location containing a single pallet. If the source location contains two pallets and no improving move is found, the same procedure is repeated for the other pallet. Type 3 moves are performed by matching each position in the list of possible source positions with the next position, where at least one pallet must allow another pallet to be placed above it, and the target position is the first position from the list of possible target positions that can accommodate two pallets; For a Type 4 move, the source location is the first location in the list of possible source locations that contains a single pallet that does not allow a pallet to be placed above it, and the target location is the first location in the list of possible target locations that contains two pallets that both allow a pallet to be placed above them.

[0048] During the entire local search, if the sum of the improvements of the stored moves is the same as the current total waste, the stored moves are executed once and the local search stops, because the waste will obviously be reduced to zero at this point.

[0049] In the process of performing pallet location reallocation optimization on the primary storage location allocation solution based on the local search strategy to obtain the initial pallet storage location allocation solution under the current warehouse layout, the specific steps are as follows: This embodiment first executes the location list construction operation, initializes and creates an empty list LocList as the storage location list to be optimized, and then traverses all locations in the warehouse layout, and for each storage location , calculate each storage location in the primary storage location allocation solution based on the storage location type (single depth or double depth) and the pallet stacking status The resulting storage waste value ,like , it means that there are wasted storage locations. All the wasted storage locations and their corresponding waste values ​​are added to the storage location list to be optimized LocList. After completing the traversal of all locations, the storage location list to be optimized LocList is sorted from high to low according to the storage location waste value. Then, this embodiment uses four priority-sorted move operation types to optimize the storage locations. The storage location list to be optimized LocList is traversed in order, and each unmarked storage location unmarkedele is used as the source location. The value range of q is initially set to , then enters a loop process, for each movement type q, searches for feasible optimized movements in the unmarked target positions. In this loop process, this embodiment analyzes whether there is an improved movement with unmarkedelem as the source position and another unmarked position as the target position. If such an improved movement exists, the found movement operation is stored in the movement list StoredMovements, and the position information involved in the movement operation is marked. At the same time, it is checked whether the improvement sum of the stored movement operation is the same as the total waste value of the current warehouse layout. If the two are the same, the subsequent step of executing the stored movement operation is entered. If there is no improved movement, unmarkedelem is updated to the next unmarked element in the list LocList, and the research operation is continued until all unmarked elements are traversed.

[0050] Finally, this embodiment executes the storage move operation. During the execution phase, all moves in the move list StoredMovements are traversed in descending order of waste reduction. For each move, the corresponding pallet position movement operation is executed, and the pallet position and storage location status are updated in real time. At the same time, the results of the move execution are recorded according to actual needs. After each round of movement is completed, the total space waste value of the entire warehouse is recalculated. If the total waste value still exceeds the preset threshold, the storage location list to be optimized LocList is regenerated based on the latest storage location status, and the above optimization process is repeated. The iteration continues until the total waste value falls below the threshold. After all storage move operations are executed, the move list StoredMovements is cleared to prepare for subsequent optimization operations. If, after executing these move operations, any move operation causes the storage layout waste value to be greater than 0, the process returns to the location list construction operation step and the optimization process is repeated. If the waste values ​​are all 0 after all move operations are executed or the preset optimization target is achieved, the entire pallet position reallocation optimization process ends. The pallet storage location allocation plan obtained at this time is the initial pallet storage location allocation plan under the current storage layout.

[0051] S4. In response to the real-time order information, an attention-learning-based picking path optimization algorithm is used to generate an optimal picking path for each order according to the real-time order information and the initial pallet storage location allocation plan.

[0052] In some embodiments, the step of generating an optimal picking path for each order using an attention learning-based picking path optimization algorithm based on the real-time order information and the initial pallet storage allocation plan includes: Obtaining a list of pallets to be picked according to the real-time order information, and extracting the current warehouse layout and storage location information of the pallets to be picked from the initial pallet storage location allocation plan based on the list of pallets to be picked; The current warehouse layout is converted into a graph structure based on the storage location information of the to-be-picked pallets to obtain a warehouse topology graph; wherein the graph nodes of the warehouse topology graph include the storage location nodes of the to-be-picked pallets, the top and bottom nodes of the shelf aisles, and the warehouse entrance and exit nodes, and the edges represent the traversable path relationships between the graph nodes; The attention mechanism is used to learn the circuit vertical edges and circuit horizontal edges from the warehouse topology graph, and the circuit graph is constructed according to the order of the circuit vertical edges and circuit horizontal edges. The attention mechanism includes the input layer, embedding layer, multi-head attention mechanism layer, feedforward layer, residual connection and layer normalization layer, and linear layer. All effective picking routes starting from the warehouse entrance node, passing through each to-be-picked pallet storage node and arriving at the warehouse exit node are extracted from the tour graph, and the effective picking route with the minimum total cost is selected as the optimal picking path.

[0053] In a specific embodiment, after completing the storage allocation, this embodiment needs to optimize the processing path after the order arrives. Figure 5 This is a schematic diagram of the warehouse layout after storage location allocation provided by an embodiment of the present invention. Based on the storage location allocation, the current warehouse layout consists of multiple parallel vertical aisles, with items stored on both sides of the aisles. A horizontal aisle (one above and one below) is provided at the top and bottom of each aisle to facilitate navigation between aisles by pickers. A single warehouse is also provided. Assuming that the aisle width is negligible, the horizontal distance within the aisle does not need to be considered. Based on this, this embodiment first constructs a graph representation after storage location allocation. It should be noted that the single-picker routing problem can be considered a variant of the traveling salesman problem, in which items and warehouses are treated as nodes. An effective picking route starts from the warehouse, passes each item exactly once, and then returns to the warehouse. The single-picker routing problem can also be expressed as the Steiner traveling salesman problem, in which items are nodes that can be visited multiple times, while the top and bottom of the aisles are nodes that do not need to be visited. However, to be consistent with existing deep reinforcement learning research results, the standard TSP model is used here for illustration. Four common heuristic picking route patterns are S-shaped, return, midpoint, maximum gap, and compound picking strategies. The following briefly describes these four heuristic picking route patterns: S-shaped picking strategy: The S-shaped heuristic algorithm, also known as the traversal method, is one of the simpler picking strategies. It requires the picker to enter each aisle that contains at least one item to be picked, completely cross the aisle and exit from the other end, repeat this process for all aisles containing items and then return to the warehouse. Figure 6 This is a schematic diagram of a picking route for an S-shaped picking strategy provided by an embodiment of the present invention; Return Picking Strategy: For each aisle containing the required items, the picker enters and exits from the aisle in a forward direction, collecting all items on the pick list. Figure 7 This is a schematic diagram of a picking route for a return picking strategy provided by an embodiment of the present invention; Maximum Gap Heuristic: The picker completely traverses the first and last aisles containing items. For each aisle in between, the picker enters and exits the aisle from the bottom and top in a specific pattern that maximizes the distance of the aisles not traveled while ensuring that all items are visited. Figure 8 This is a schematic diagram of a picking route using a maximum gap heuristic strategy provided by an embodiment of the present invention; Composite picking strategy: Combining the advantages of return picking strategy and S-shaped picking strategy, for all adjacent aisles containing items in the picking list, a heuristic algorithm is selected to minimize the moving distance between the farthest picks. Figure 9 This is a schematic diagram of a picking route for a composite picking strategy provided by an embodiment of the present invention; Optimal Picking Strategy: An optimal procedure for picker routing in a rectangular warehouse, linearly solvable with respect to the number of aisles and pick locations. The algorithm sequentially constructs a partial tour subgraph, where nodes consist of the top and bottom of each aisle and the location of each item to be picked.

[0054] This embodiment extracts a list of pallets to be picked based on real-time order information. This real-time order information includes the type and quantity of items to be picked, as well as their corresponding storage locations. The corresponding storage location coordinates are then obtained from the initial storage location allocation plan. The current warehouse layout consists of multiple parallel vertical aisles with storage locations located on both sides of the aisles. A horizontal connecting passage is provided at the top and bottom of each aisle to facilitate navigation between aisles by pickers. Aisle widths are assumed to be negligible, so horizontal distances within aisles do not need to be considered. This embodiment abstracts the physical warehouse layout into a warehouse topology graph. The graph nodes include nodes for pallet locations to be picked, nodes at the top and bottom of shelf aisles, and nodes at warehouse entrances and exits. Edges represent traversable paths between nodes, and distance weights reflect the actual movement cost. This embodiment transforms the single-picker routing problem into a modified variant of the traveling salesman problem. An efficient picking route starts from the warehouse, passes each item exactly once, and then returns to the warehouse. This embodiment considers the item to be picked and the warehouse as required nodes, and adds aisle ends as optional access points. Starting from the warehouse, traversing all item locations and returning to the warehouse creates the shortest total path.

[0055] This embodiment uses the attention mechanism to approximate the strategy, sequentially outputting the vertical and horizontal edges to be added to the tour graph until a complete tour graph is constructed. The nodes of the tour graph are composed of the top and bottom nodes of the shelf aisle and each storage location node of the pallet to be picked. In this embodiment, Figure 10This is a schematic diagram of the attention mechanism structure provided by an embodiment of the present invention. The structural composition of the attention mechanism mainly includes an input layer, an embedding layer (corresponding to input Embedding), a channel coding, a multi-head attention mechanism, a feedforward layer, a residual connection and layer normalization, and a linear layer. The input signal is first converted by the embedding layer coding, and then fused with the result of the channel coding (indicated by ⊕ in the figure), and then enters the multi-head attention mechanism module, so that the multi-head attention mechanism can extract features and model relationships of the input signal, and capture the dependencies between different nodes. The output of the multi-head attention mechanism is processed by the feedforward layer, the residual connection and the layer normalization in turn. The residual connection and layer normalization operations are performed once between the multi-head attention mechanism and the feedforward layer, and once between the feedforward layer and the subsequent layer. Finally, the linear layer maps the final features to the output space, generates an edge addition probability distribution, and dynamically selects the optimal tour vertical edge (movement within the aisle) and tour horizontal edge (transfer between aisles) based on the edge addition probability distribution. The partial tour subgraph is gradually expanded until all nodes are completely covered, and all valid path solutions are extracted from the generated tour graph. This embodiment calculates the total cost of each valid path solution, which is the sum of the basic cost and the penalty term, where the basic cost is the sum of the edge weights (reflecting the driving distance). The penalty term is the product of the number of turns and a preset weight coefficient, which is dynamically adjusted by the attention weight entropy. This embodiment selects the effective path solution with the minimum total cost as the final picking path. In this embodiment, the embedding layer converts the input data into a vector representation suitable for model processing; the channel encoding provides the model with prior information such as the warehouse aisle layout; the multi-head attention mechanism can capture the complex dependencies between different positions in the input sequence, enabling the model to better understand the relationship between the warehouse layout and the location of items; the feedforward layer performs a nonlinear transformation on the output of the attention mechanism to enhance the model's expressiveness; residual connections and layer normalization help alleviate the problems of vanishing and exploding gradients, accelerate model training, and improve model stability; the linear layer converts the model output into the final decision result. This method sequentially determines which edges to add to the tour graph, providing a different perspective on the problem. Compared with traditional methods that directly output node arrangement, this embodiment focuses on the key relationships between different picking points through the attention mechanism, gradually deciding whether to add vertical or horizontal connecting edges, and dynamically generating an efficient picking path. This is particularly suitable for scenarios such as the single-picker routing problem, where the connections between nodes are not limited to simple arrangement relationships and also need to consider practical factors such as warehouse layout.

[0056] S5. Based on the storage status of the remaining materials in the warehouse after the execution of the optimal picking path, evaluate and obtain the layout data of the remaining materials to be optimized, and perform a secondary allocation of the storage locations of the remaining materials based on the layout data of the remaining materials to be optimized to obtain a storage location reallocation plan.

[0057] In some embodiments, the step of performing the second allocation of the storage locations of the remaining goods according to the remaining goods layout data to be optimized to obtain a storage location reallocation scheme comprises: generating a storage movement list according to the remaining goods layout data to be optimized, and defining a maximum allowed execution time; performing the storage movement operations in the storage movement list in a preset movement priority, and generating a random number in the process of performing each storage movement, and deleting a number of storage movements equal to the random number from the storage movement list; calculating the movement execution time of each storage movement in the process of performing each storage movement, and accumulating the movement execution time of each storage movement, and stopping the execution of the storage movement when the accumulated value of the movement execution time reaches the maximum allowed execution time, and constructing a storage movement operation scheme with the storage movement operations that have been executed; screening the optimal storage location reallocation scheme from all storage movement operation schemes according to the total waste reduction amount of the storage movement operation scheme.

[0058] The SLAP (pallet location assignment) algorithm can also be used to determine a more optimal storage location for the stored pallets. The specific operation is to redistribute all existing pallets assuming that the warehouse is empty. At this time, the incoming pallet set IncomingPallets is equal to the stored pallets. By comparing the difference in space availability metrics before and after processing, the improvement gap that can be achieved is determined. The manager can decide whether to reduce the gap by relocating part of the pallets during the low workload period according to the importance of this gap. In general, it is not possible to relocate all pallets, so the manager may focus on the following three issues: 1. Minimize the number of relocation movements, that is, perform a fixed number of relocation actions to reduce the current space waste; 2. Maximize waste reduction under a fixed number of relocation, that is, after performing a fixed number of relocation, the maximum waste reduction is obtained; 3. Maximize waste reduction in a given time, that is, the maximum waste reduction effect is achieved within a certain time.

[0059] To solve the above three situations, this embodiment can apply a local search algorithm to use the current pallet position as the initial solution. It should be noted that the same item will not be relocated multiple times. If the goal is to minimize the number of relocation moves, the storage moves are performed in the order of decreasing improvement until the required improvement is achieved or all moves have been performed. If the goal is to maximize the waste reduction under a fixed number of relocations, only the best storage move needs to be performed. In addition, all pallets other than the fixed best move pallet can be used. These algorithms may place this pallet in a new position. The difference between the waste of the current solution and the best solution indicates the improvement that can be obtained after performing a maximum of several moves. Different rules for selecting fixed pallets may produce different solutions. For example, pallets located in positions with less waste or pallets with strict hierarchical restrictions can be selected.

[0060] If the goal is to maximize waste reduction within a given time, the execution time of the relocation move needs to be determined. The execution time of the relocation move is the sum of the fulfillment time, input and output time, and transfer time between two locations, where the transfer time is the time required to load and unload the pallet involved in the relocation. However, the travel time and input and output time between locations are negligible compared to the time required for a forklift to store or retrieve a unit load at a storage location, so the process only considers the transfer time. One solution method is to consider the storage moves of the local search algorithm and assign them to forklifts, ensuring that the execution time assigned to any vehicle does not exceed the set maximum time limit. To this end, this embodiment proposes three deterministic allocation rules: give priority to the move with the greatest improvement; give priority to the move with the shortest execution time; give priority to the move with the largest ratio of improvement to execution time.

[0061] In order to evaluate the performance, the results generated according to these rules are compared with the sampling algorithm. A fixed number of solutions are obtained in an iterative manner. The moves in the stored move list are calculated by performing local search. The specific steps are as follows: define nsolution as the required number of solutions and pre-set a rule. The stored move list is sorted according to this rule. In the iterative solution generation phase, for each iteration, the moves are executed in sequence from the sorted move list. The execution process ensures that the maximum time limit is not exceeded. Depending on the actual situation, some moves may be deleted from the list, while other moves are reintroduced into the list. After each successful generation of a new solution, the performance indicators of the solution are recorded. Finally, the solution obtained by the sampling algorithm and its performance evaluation results are output.

[0062] In a specific embodiment, in the process of performing secondary allocation of storage locations of surplus materials according to the surplus material layout data to be optimized, this embodiment generates a storage movement list according to the surplus material layout data to be optimized, and the storage movement list includes all possible storage movement operations. Considering the generated storage movement list, the total number of movement operations is counted and marked as nstored. At the same time, for each storage movement operation in the storage movement list, , this embodiment defines v( ) is assigned to the first The priority values ​​of the stored moves are sorted by priority to form an ordered move list L={ , ,..., ,..., }, where nstored represents the total number of move operations, and sets the maximum allowed execution time in the entire location reallocation process to T. The maximum allowed execution time is used to limit the execution time of the entire move operation. Then, this embodiment repeats the following operation nsolution times to generate multiple location reallocation schemes. In the initialization stage of each repetition, for q taking a value from 1 to 4, unmarkedelem is set to the first unmarked element in the storage location list to be optimized LocList to ensure that there is a marked state at the beginning of each iteration. Then, the inner layer repeats the steps and executes the storage move operations in sequence in the order of the ordered move list L until the cumulative execution time reaches the maximum allowed execution time T. In this process, regardless of whether the move has been actually executed, each checked move needs to be marked for subsequent processing.

[0063] At the same time, this embodiment generates a random number num in each iteration. Based on the generated random number num, storage moves equal to num are randomly selected from the storage moves checked in the most recent four iterations, and these selected storage moves are temporarily deleted from the ordered move list L. These deleted storage moves will be reintroduced into the ordered move list L after six iterations to ensure the diversity of solutions. In the process of generating solutions in each iteration, the total waste reduction of the solution will be recorded. Finally, from all the generated storage move operation plans, the optimal storage location reallocation plan is screened out based on the key indicator of total waste reduction. This method can fully consider the time limit and efficiency requirements of warehouse operations. Through this structured secondary allocation strategy, the optimal adjustment of the storage location of remaining materials can be achieved within a limited time, effectively improving the utilization rate of warehouse space and reducing operating costs. It is suitable for the dynamic optimization needs of large-scale warehousing systems.

[0064] S6. Generate a storage location-path integrated operation instruction based on the optimal picking path and the storage location reallocation plan, and control the warehouse management system operation through the storage location-path integrated operation instruction.

[0065] The storage location allocation problem (SLAP) aims to maximize the available storage space in a chaotic warehouse system. This type of warehouse system has dynamic storage characteristics, adopts a single command mode, uses pallets as storage units, and access operations are performed by forklifts. However, the storage location allocation problem only aims to maximize the available storage space. To solve this problem, this embodiment proposes a heuristic algorithm and a local search algorithm. On this basis, this embodiment further couples the picking path optimization method based on attention learning to achieve global optimization from warehouse layout to path picking. The key advantage of the method proposed in this embodiment is that it can reduce the perceived complexity of the path. In fact, pickers usually prefer simple routes rather than those that may be the optimal solution but contain difficult to understand and inconsistent Such complexity will require the picker to make additional efforts to identify the direction, which will not only reduce the picking speed but also increase the probability of errors. The attention mechanism of this embodiment can simply integrate such constraints by shielding related actions when constructing the tour path. This method not only helps to simplify the picker's workflow and improve work efficiency, but also effectively reduces the possibility of errors, thereby optimizing the entire picking process. This fully demonstrates that compared with the traditional traveling salesman problem solution, the method proposed in this embodiment has greater flexibility and practicality in practical applications, especially when dealing with specific domain restrictions.

[0066] An embodiment of the present invention provides a warehouse operation method based on integrated optimization of storage locations and picking paths, the method initializing the warehouse layout according to the warehouse physical structure data and the pallet attribute data, generating an initialized warehouse layout parameter set; based on the initialized warehouse layout parameter set, constructing a dynamic storage location allocation optimization model with minimizing the total space waste value of the entire warehouse as the optimization goal; iteratively solving the dynamic storage location allocation optimization model to obtain an initial pallet storage location allocation plan under the current warehouse layout; in response to real-time order information, according to the initial pallet storage location allocation plan, adopting an attention learning-based picking path optimization algorithm to generate the optimal picking path for each order; according to the storage status of the remaining materials in the warehouse after the execution of the optimal picking path, obtaining the layout data of the remaining materials to be optimized, performing a secondary allocation of the storage locations of the remaining materials according to the layout data of the remaining materials to be optimized, and obtaining a storage location reallocation plan; generating a storage location-path integrated operation instruction according to the optimal picking path and the storage location reallocation plan, and controlling the operation of the warehouse management system through the storage location-path integrated operation instruction. Compared with the existing technology, this method realizes the integrated collaborative optimization of warehouse layout, storage location allocation and picking path through the collaborative optimization of storage location allocation and picking path, effectively reduces the total space waste value of the entire warehouse, improves warehouse operation efficiency and resource utilization, and thus realizes the intelligent and integrated management of warehouse operations.

[0067] It should be noted that the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of this application.

[0068] In one embodiment, Figure 11 As shown, an embodiment of the present invention provides a warehouse operation system based on integrated optimization of storage locations and picking paths, the system comprising: The warehouse initialization module 101 is used to initialize the warehouse layout according to the warehouse physical structure data and the pallet attribute data, and generate an initialization warehouse layout parameter set; A storage location model building module 102 is configured to build a dynamic storage location allocation optimization model based on the initialized storage layout parameter set, with minimizing the total space waste value of the entire warehouse as an optimization goal; The storage location model solving module 103 is used to iteratively solve the dynamic storage location allocation optimization model to obtain an initial pallet storage location allocation plan under the current warehouse layout; A picking path analysis module 104 is configured to generate an optimal picking path for each order in response to real-time order information and based on the real-time order information and the initial pallet storage allocation plan using a picking path optimization algorithm based on attention learning; The warehouse reallocation module 105 is configured to evaluate the remaining material storage status after the optimal picking path is executed, obtain the remaining material layout data to be optimized, and perform a secondary allocation of the remaining material storage locations according to the remaining material layout data to obtain a warehouse reallocation plan; The warehouse operation module 106 is used to generate a storage location-path integrated operation instruction according to the optimal picking path and the storage location reallocation plan, and control the operation of the warehouse management system through the storage location-path integrated operation instruction.

[0069] For the specific definition of a warehouse operation system based on the integrated optimization of storage locations and picking paths, please refer to the above-mentioned definition of a warehouse operation method based on the integrated optimization of storage locations and picking paths, which will not be repeated here. Those of ordinary skill in the art will appreciate that the various modules and steps described in conjunction with the embodiments disclosed in this application can be implemented in hardware, software, or a combination of both. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0070] An embodiment of the present invention provides a warehouse operation system based on integrated optimization of storage locations and picking paths. The system's storage initialization module initializes the storage layout based on warehouse physical structure data and pallet attribute data, and generates an initialization storage layout parameter set. The storage location model construction module constructs a dynamic storage location allocation optimization model based on the initialized warehouse layout parameter set, with minimizing the total space waste value of the entire warehouse as the optimization goal. The storage location model solution module iteratively solves the dynamic storage location allocation optimization model to obtain an initial pallet storage location allocation plan under the current warehouse layout. The picking path analysis module responds to real-time order information and uses an attention learning-based picking path optimization algorithm to generate an optimal picking path for each order based on the initial pallet storage location allocation plan. The storage location reallocation module obtains the layout data of the remaining materials to be optimized based on the storage status of the remaining materials in the warehouse after the execution of the optimal picking path, and performs a secondary storage location allocation on the storage locations of the remaining materials based on the layout data of the remaining materials to be optimized to obtain a storage location reallocation plan. The storage operation module generates storage location-path integrated operation instructions based on the optimal picking path and the storage location reallocation plan, and controls the operation of the warehouse management system through the storage location-path integrated operation instructions. Compared with the existing technology, this system realizes the integrated collaborative optimization of warehouse layout, storage location allocation and picking path through the collaborative optimization of storage location allocation and picking path, effectively reducing the total space waste value of the entire warehouse, improving warehouse operation efficiency and resource utilization, and thus realizing the intelligent and integrated management of warehouse operations.

[0071] The above-described embodiments merely represent several preferred implementations of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent. It should be noted that a person skilled in the art could make several improvements and substitutions without departing from the technical principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be based on the scope of protection of the claims.

Claims

1. A warehouse operation method based on integrated optimization of storage locations and picking paths, characterized in that: The following steps are involved: Initialize the warehouse layout according to the warehouse physical structure data and pallet attribute data, and generate the initialization warehouse layout parameter set; Based on the initialized warehouse layout parameter set, a dynamic storage space allocation optimization model is constructed with minimizing the total space waste value of the entire warehouse as the optimization goal; Iteratively solving the dynamic storage location allocation optimization model to obtain an initial pallet storage location allocation plan under the current warehouse layout; In response to the real-time order information, an optimal picking path for each order is generated using a picking path optimization algorithm based on attention learning according to the real-time order information and the initial pallet storage allocation plan; According to the storage status of the remaining materials in the warehouse after the execution of the optimal picking path, the layout data of the remaining materials to be optimized is evaluated, and the storage locations of the remaining materials are secondary allocated according to the layout data of the remaining materials to be optimized to obtain a storage location reallocation plan; A storage location-path integrated operation instruction is generated according to the optimal picking path and the storage location reallocation plan, and the operation of the warehouse management system is controlled by the storage location-path integrated operation instruction.

2. A warehouse operation method based on integrated optimization of storage locations and picking paths according to claim 1, characterized in that: The step of constructing a dynamic storage space allocation optimization model based on the initialized storage layout parameter set with minimizing the total space waste value of the entire warehouse as the optimization goal includes: Define a pallet stacking identification variable and a single pallet storage location allocation variable; wherein the single pallet storage location allocation variable is used to indicate that a single pallet is stored in a storage location alone, and the pallet stacking identification variable is used to indicate that two different pallets are stored in the same storage location; According to the initialized warehouse layout parameter set and the predetermined waste value measurement rule, the total warehouse space waste value of all pallet storage location storage combinations is calculated using the pallet stacking identification variable and the single pallet storage location allocation variable; Taking minimizing the total space waste value of the entire warehouse as the optimization goal, a dynamic storage space allocation optimization model is constructed.

3. A warehouse operation method based on integrated optimization of storage locations and picking paths as described in claim 2, characterized in that: The calculation process of the total space waste value of the entire warehouse is as follows: When two pallets to be stored are stored in the same storage location, the space waste generated when each pallet to be stored is stacked in the same storage location is calculated based on the sum of the heights of each pallet to be stored in the same storage location, and the double pallet stacking waste value of all pallets to be stored is obtained; When a single pallet to be stored is stored in a single pallet storage location, if it is detected that the single pallet to be stored does not allow another pallet to be stored to be placed above it, a restrictive space waste value is obtained according to the sum of the differences between the heights of all single pallet storage locations and the height of the single pallet to be stored; Based on the pre-acquired single pallet space waste coefficient, the single pallet space waste value is measured when a pallet to be stored is already stored at a position where two pallets can be placed and the remaining height is too low to accommodate another pallet to be stored. The weighted sum of the double pallet stacking waste value, the restrictive space waste value and the single pallet space waste value is calculated using the pallet stacking identification variable and the single pallet storage location allocation variable to obtain the total space waste value of the entire warehouse.

4. The warehouse operation method based on integrated optimization of storage locations and picking paths according to claim 2, characterized in that: The constraints of the dynamic storage location allocation optimization model include storage restriction constraints of each storage location, pallet storage feasibility constraints, space restriction constraints, level restriction constraints and stacking rule restriction constraints; The storage restriction constraint of each storage location is specifically that the number of pallets stored in each storage location does not exceed its maximum capacity limit; The pallet storage feasibility constraint is specifically that each pallet is stored in only one storage location; The space limitation constraint is specifically that the total height of all pallets stored in each storage location does not exceed the preset maximum storage location height; The layer restriction constraint specifically means that the number of pallet storage layers does not exceed the maximum number of layers allowed; The stacking rule restriction specifically states that if two pallets are not allowed to be placed on top of another pallet, the two pallets cannot be stored in the same storage location.

5. The warehouse operation method based on integrated optimization of storage locations and picking paths according to claim 1, characterized in that: The step of iteratively solving the dynamic storage location allocation optimization model to obtain an initial pallet storage location allocation solution under the current warehouse layout includes: Scan the warehouse layer by layer from the bottom to the top. During the scanning process, determine the set of pallets to be stored according to the initialized warehouse layout parameter set, and set the current level to the bottom of the warehouse. The current level is used as the level to be processed, and the pallets in the level to be processed are selected from the set of pallets to be stored and sorted in descending order by pallet height to build a list of pallets to be processed; Filter out all vacant storage locations in the level to be processed and sort them in descending order by storage location height to generate a list of available locations in the level to be processed; Based on the list of pending pallets and the list of available locations, the dynamic storage location allocation optimization model is solved using a pallet location allocation algorithm to obtain a primary storage location allocation solution; The primary storage location allocation solution is optimized for pallet location reallocation based on a local search strategy to obtain an initial pallet storage location allocation solution under the current warehouse layout.

6. A warehouse operation method based on integrated optimization of storage locations and picking paths according to claim 5, characterized in that: The step of solving the dynamic storage location allocation optimization model using a pallet location allocation algorithm based on the list of pallets to be processed and the list of available locations to obtain a primary storage location allocation solution includes: Initializing the set of unstored pallets to be an empty set, and when the list of pallets to be processed is not empty, performing a pallet allocation operation according to the constraint conditions of the dynamic storage allocation optimization model, and sequentially detecting the pallets in the list of pallets to be processed; If it is detected that there is a pallet in the list of pallets to be processed whose height is greater than the height of the first storage location in the list of available locations, move the pallet from the list of pallets to be processed to the set of unstored pallets, and continue to detect the next pallet in the list of pallets to be processed; Select the pallet with the largest height from the list of pallets to be processed as the reference pallet, and search for the pallet with the second largest height from the list of pallets to be processed. If the sum of the heights of the reference pallet and the pallet with the second largest height does not exceed the height of the first storage location in the list of available locations, and at least one pallet can be placed above another pallet, then the reference pallet and the pallet with the second largest height are stored as paired pallets in the available storage location with the lowest height; otherwise, only the reference pallet is stored in the available storage location with the lowest height. If the selected available storage location with the lowest height is a double-depth storage location, searching for other pallets or pallet pairs that are allowed to be stacked in the list of pallets to be processed to fill the remaining space of the double-depth storage location; Deleting the selected paired pallet from the list of pallets to be processed and the set of pallets to be stored, and removing the used available storage locations from the list of available locations; If the unstored pallet set is empty and the to-be-processed level is not the bottom, the pallets in the unstored pallet set are reinserted into the to-be-processed pallet list in descending order of pallet height, the current level is adjusted to the previous level, and the available position list is regenerated; If the unstored pallet set is empty and the level to be processed is the lowest, the pallet with the smallest height is selected from the unstored pallet set and stored in the remaining vacant storage space at the current level or the level below it; If the current level is less than the maximum number of levels and the set of unstored pallets is empty, the level to be processed is updated to the next level, and the list of pallets to be processed and available locations at the current level is re-screened, and pallet allocation continues until all pallets meet the constraints of the dynamic storage allocation optimization model. The iteration is stopped and the primary storage allocation solution is output.

7. A warehouse operation method based on integrated optimization of storage locations and picking paths according to claim 5, characterized in that: The step of performing pallet position reallocation optimization on the primary storage location allocation solution based on the local search strategy to obtain an initial pallet storage location allocation solution under the current warehouse layout includes: Calculate the storage location waste value of each storage location in the primary storage location allocation solution according to the storage location type and the pallet stacking status, and sort all storage locations in the primary storage location allocation solution in descending order according to the storage location waste value to generate a list of storage locations to be optimized; Traversing the list of storage locations to be optimized, and taking each storage location in the list of storage locations to be optimized as a source location, detecting feasible moves in sequence according to the four pre-established movement type priorities; If a feasible move is detected, the feasible move is executed in descending order of waste reduction magnitude, and the pallet position and storage location status are updated in real time based on the storage location information of the feasible move; After each round of feasible moves, the total space waste value of the entire warehouse is recalculated. If the total space waste value of the entire warehouse is greater than the preset waste threshold, a new list of storage locations to be optimized is generated based on the updated storage location status, and the above steps are repeated, and the optimization is iterated continuously until the total space waste value of the entire warehouse is less than the preset waste threshold. The iterative optimization is terminated, and the initial pallet storage location allocation plan under the current warehouse layout is output.

8. The warehouse operation method based on integrated optimization of storage locations and picking paths according to claim 1, characterized in that: The step of generating the optimal picking path for each order using an attention learning-based picking path optimization algorithm according to the real-time order information and the initial pallet storage location allocation plan includes: Obtaining a list of pallets to be picked according to the real-time order information, and extracting the current warehouse layout and storage location information of the pallets to be picked from the initial pallet storage location allocation plan based on the list of pallets to be picked; The current warehouse layout is converted into a graph structure based on the storage location information of the to-be-picked pallets to obtain a warehouse topology graph; wherein the graph nodes of the warehouse topology graph include the storage location nodes of the to-be-picked pallets, the top and bottom nodes of the shelf aisles, and the warehouse entrance and exit nodes, and the edges represent the traversable path relationships between the graph nodes; An attention mechanism is used to learn the circuit vertical edges and the circuit horizontal edges from the warehouse topology graph, and a circuit graph is constructed according to the circuit vertical edges and the circuit horizontal edges in sequence; the attention mechanism includes an input layer, an embedding layer, a multi-head attention mechanism layer, a feedforward layer, a residual connection and a layer normalization layer, and a linear layer; All effective picking routes starting from the warehouse entrance node, passing through each to-be-picked pallet storage node and arriving at the warehouse exit node are extracted from the tour graph, and the effective picking route with the minimum total cost is selected as the optimal picking path.

9. The warehouse operation method based on integrated optimization of storage locations and picking paths according to claim 1, characterized in that: The step of performing a secondary allocation of storage locations of the remaining materials according to the remaining material layout data to be optimized to obtain a storage location reallocation plan includes: Generate a storage movement list based on the remaining material layout data to be optimized, and define a maximum allowed execution time; Executing the storage move operations in the storage move list in sequence according to a preset move priority, generating a random number during the execution of each storage move, and randomly selecting storage moves from the storage move list equal to the number of the random numbers for deletion; During the execution of each storage movement, the movement execution time of each storage movement is calculated and accumulated. When the accumulated movement execution time reaches the maximum allowed execution time, the storage movement is stopped and the currently executed storage movement operations are combined into a storage movement operation plan. According to the total waste reduction of the storage movement operation plan, the optimal storage location reallocation plan is screened out from all storage movement operation plans.

10. A warehouse operation system based on integrated optimization of storage locations and picking paths, characterized in that: The system comprises: The warehouse initialization module is used to initialize the warehouse layout according to the warehouse physical structure data and pallet attribute data, and generate the initialization warehouse layout parameter set; A storage location model construction module is used to construct a dynamic storage location allocation optimization model based on the initialized storage layout parameter set with minimizing the total space waste value of the entire warehouse as the optimization goal; A storage location model solving module is used to iteratively solve the dynamic storage location allocation optimization model to obtain an initial pallet storage location allocation plan under the current warehouse layout; a picking path analysis module, configured to generate an optimal picking path for each order in response to real-time order information and based on the real-time order information and the initial pallet storage allocation plan, using a picking path optimization algorithm based on attention learning; A storage location reallocation module is configured to evaluate the storage status of the remaining materials in the warehouse after the optimal picking path is executed to obtain the remaining material layout data to be optimized, and to perform a secondary allocation of the storage locations of the remaining materials based on the remaining material layout data to obtain a storage location reallocation plan; The warehouse operation module is used to generate a storage location-path integrated operation instruction according to the optimal picking path and the storage location reallocation plan, and control the operation of the warehouse management system through the storage location-path integrated operation instruction.

Images