A method for allocating storage locations in an automated storage and retrieval system (AS / RS) driven by digital twins.

Abstract

This invention provides a method for allocating storage locations in an automated three-dimensional warehouse using digital twin technology. The method constructs a warehouse management system through five dimensions: a physical layer, a twin layer, a data layer, a service layer, and a connectivity layer. The physical layer represents the actual warehousing operation scenario; the twin layer maps the equipment and environment of the physical layer to a virtual space, constructing a digital model of the warehouse; the data layer is responsible for the collection, processing, and storage of data from the physical layer, providing accurate input for the service and twin layers; the service layer provides optimization decision-making and human-computer interaction functions for users and system management; and the connectivity layer is responsible for the data and control transmission between the layers, ensuring the system's synergy and real-time performance. By supporting multi-objective optimization algorithms through digital twin technology, intelligent and precise storage location allocation is achieved, providing a novel solution for digital logistics management.

Description

Technical Field

[0001] This invention belongs to the field of automated warehouse location allocation, specifically relating to an automated three-dimensional warehouse location allocation method driven by digital twins. Background Technology

[0002] The global manufacturing industry is undergoing a major transformation driven by integration, intelligence, and digitalization. In this process, warehousing, as a crucial link in enterprise operations, also faces the challenge of transitioning from traditional industry to intelligent manufacturing. With the gradual increase in logistics storage costs and increasingly stringent requirements for warehousing efficiency, Automated Storage and Retrieval Systems (AS / RS) are gradually replacing traditional warehouses due to their advantages such as high space utilization, large storage capacity, and low labor costs, and are widely used in manufacturing, retail, e-commerce, and other fields. To improve warehouse operational efficiency and resource utilization, optimizing warehouse location allocation has become an important research topic. In complex warehouse environments, due to the diversity of goods, frequent changes in demand, and the complexity of warehouse layout, traditional location allocation methods have the following shortcomings:

[0003] 1. Lack of intelligent decision support: Location allocation needs to balance multiple objectives such as picking efficiency, shelf balance, and equipment energy consumption, which traditional single-objective or simple multi-objective optimization algorithms are inadequate for. Furthermore, even when multiple location allocation schemes are generated using traditional methods, the decision-making process still heavily relies on the subjective judgment of managers, especially when multiple objectives conflict. This experience-driven decision-making approach can lead to inconsistencies and decision biases.

[0004] 2. Poor dynamic response capability: Existing methods are mostly based on static rules or historical data, making it difficult to respond in real time to changes in actual inventory demand, environmental conditions, or system status. When demand changes, traditional methods cannot adjust the storage location allocation plan in a timely manner, resulting in decreased allocation efficiency.

[0005] 3. Lack of systematic digital support: Traditional methods cannot make full use of real-time data provided by sensors and IoT devices, lack effective simulation tools to predict the actual effect of storage location allocation schemes, and managers find it difficult to adjust allocation strategies in a timely manner to adapt to complex warehouse environments. Summary of the Invention

[0006] Objective of the Invention: The technical problem to be solved by this invention is to address the shortcomings of existing technologies by providing a method for allocating storage locations in an automated three-dimensional warehouse using digital twin-driven technology. This method optimizes storage location allocation and improves warehouse operational efficiency by deeply integrating the physical environment of the actual warehouse with its digital virtual model.

[0007] The method includes: constructing a five-dimensional digital twin model, including a physical layer, a twin layer, a data layer, a service layer, and a connectivity layer;

[0008] The physical layer is the actual location for goods storage and retrieval operations, representing the actual physical entities and environment in the digital twin warehouse management system. The physical entities include shelves, stacker cranes, inbound and outbound platforms, turnover boxes, and goods. Data from the physical layer is uploaded to the data layer through real-time interaction, while simultaneously receiving optimization instructions from the service layer.

[0009] The twin layer is a digital mirror of the physical space. By constructing a warehouse digital twin model that maps the virtual and real warehouse environments, it provides simulation support for location allocation and warehouse operations. The warehouse digital twin model relies on 3D modeling technology. In Blender software, it constructs accurate 3D models based on the shape, color, and material characteristics of physical entities. High-quality rendering ensures a high degree of consistency between the virtual model and the real warehouse environment. Simultaneously, it sets a model scene tree based on the parent-child relationship information between 3D models. The model data is output as 3D model files in .gltf or .glb format. The Three.js engine loads and parses the .gltf or .glb format 3D models on the web. Combined with environmental data collected by IoT devices, it achieves real-time mapping of warehouse layout, equipment status, and environmental parameters. The Vue3 framework dynamically binds warehouse data to ensure real-time updates of shelf inventory, stacker crane location, and operating status. Integrating all geometric models, dynamic data, and rendering scenes, a complete warehouse digital twin system is constructed.

[0010] The data layer, as the data support layer connecting the physical layer and the twin layer, provides input to the service layer by acquiring dynamic and static data, cleaning and storing the data. The dynamic data is used to reflect the real-time operation of the automated warehouse, including cargo flow data, equipment status, environmental data, etc., while the static data includes basic data such as warehouse geometric dimensions, equipment physical constraints, and cargo material properties.

[0011] The service layer provides users with a human-computer interaction interface and decision support, realizing a combination of intelligent and manual decision-making in cargo location allocation;

[0012] The connection includes a data connection and a control connection. The data connection is used to collect physical layer data to the data layer through IoT (Internet of Things) devices and network devices. The information processed by the data layer is input to the twin layer for modeling and simulation, and input to the service layer for generating a cargo location allocation scheme. The control connection is used to feed back cargo location allocation instructions to the physical layer to guide actual operation, and to feed them back to the twin layer for virtual operation.

[0013] The service layer performs the following steps:

[0014] Step 1: Obtain the processed location status, cargo characteristics, and equipment status information from the data layer;

[0015] Step 2: Use the intelligent improvement algorithm SA-NSGA-Ⅱ to generate a set of optional storage location allocation schemes. ,in This represents the I-th intelligent storage location allocation scheme;

[0016] Step 3: Using a decision-making method combining fuzzy hierarchical analysis (FAHP) and the topology-topic solution ranking method (TOPSIS), the set of storage location allocation schemes is analyzed. Zhongjueyou selects the best intelligent storage location allocation solution ,in Indicates the first The coordinates of the storage location to which each item is assigned;

[0017] Step 4: Display the visualized results of the intelligent storage location allocation scheme on the human-computer interaction interface, including the location and storage location number of each goods task, and display the storage location utilization and equipment load through dynamic charts, allowing users to judge the intelligent storage location allocation scheme. Does it need adjustment? If so, the modified solution set is as follows: ,in Indicates the adjusted number Output the storage location allocation scheme for each item; otherwise, directly output the intelligent allocation scheme. ;

[0018] Step 5, according to the final plan This generates operation instructions for physical layer devices, which are then used by the corresponding warehousing equipment to perform the task of allocating storage locations.

[0019] Step 6: By monitoring the status of goods and equipment in real time, the execution results of the physical layer and the twin layer are fed back. If the goods are not placed in the correct position, the alarm is recorded and triggered. After the solution is completed, the data layer information is updated.

[0020] Step 2, the intelligent improvement algorithm SA-NSGA-Ⅱ includes the following steps:

[0021] Step 2-1, randomly generate a population of size . Initialized population The first in the population Individual chromosomes The Middle One gene Representing a cargo allocation task, the three-dimensional coordinates of genes are encoded into one-dimensional integers based on the rasterization concept. , ;

[0022] Step 2-2: Perform a feasibility check on the genes in the individual chromosome to monitor whether the individual meets the constraints, which include:

[0023] Uniqueness constraint of storage location: ,

[0024] Shelf load-bearing capacity constraints: ,

[0025] Shelf capacity constraints: ,

[0026] in, , , These refer to the number of shelf rows, columns, and shelves; Indicates the first The coordinates of the location of each item on the shelf; Indicates the first The occupancy status of the storage space allocated to each item; This indicates the total weight of all goods on a single shelf. This indicates the total weight of all goods on a single row of shelves; , These represent the maximum load-bearing capacity of a single row of shelving and the maximum load-bearing capacity of a single layer of shelving, respectively. Represents the set of positive integers;

[0027] Steps 2-3: Set up gene repair operators for individual chromosomes. Gene values ​​that do not meet the constraints Repair is performed, and the repaired gene is ,in, This represents a gene repair operator, used to adjust the gene values ​​in a chromosome so that the gene values ​​in the chromosome meet the constraints.

[0028] Steps 2-4: Calculate the fitness of each individual in the population within the objective function space for cargo location allocation optimization, and establish the following multi-objective cargo location optimization mathematical model:

[0029] ,

[0030] in, This represents the overall objective of optimization, which is to minimize the values ​​of multiple sub-objectives; Indicates the balance of the shelf; This indicates the distance between similar items in a single row of shelves; Indicates the time of stacker crane entry and exit from the warehouse;

[0031] Establish the following criteria for measuring the balance of all shelves in an automated warehouse:

[0032] ,

[0033] in, The width of the storage space; Indicates the first The quality of each item;

[0034] Calculate the distance between similar items in a single row of shelving using Manhattan distance:

[0035] ,

[0036] in, For the height of the storage space; Indicates the first The coordinates of the location of the item on the shelf, the first item The goods and the first The goods are of the same type; Indicates the type of goods; Indicates the total number of goods of the same type; Indicates type The collection of all goods;

[0037] Establish the following objective function for the stacker crane's inbound and outbound times:

[0038] ,

[0039] in, The length of the storage space; , , These represent the stacker crane forklift speed, horizontal travel speed, and vertical travel speed, respectively. Indicates the first Goods turnover rate; Represents the coordinates of the inbound / outbound station corresponding to the single row of shelves where the i-th item is located;

[0040] Steps 2-5, for each chromosome individual in the population ,judge Dominance relationships with other chromosomes, if they satisfy In sub-target , , If a chromosome is not inferior to other chromosomes in any of its sub-targets, and is strictly superior to other chromosomes in at least one sub-target, then it is called a chromosome. Dominates other chromosomes, statistical dominance number of chromosomes and record Dominant set of chromosomes Find all chromosomes, The chromosomes are not dominated by any other cleavage, forming the first frontier. Remove from the population Continue processing the remaining chromosomes, repeating steps 2-5 until all chromosomes are assigned to their corresponding frontal zones. ;

[0041] Steps 2-6, in sub-targets , , Randomly select any specific target The non-dominated frontier, i.e., the first frontier The chromosomes in the selection are based on the chosen objective function. The values ​​are sorted, the boundary solutions of the target space are marked as extreme solutions, and an infinite crowding degree is assigned. For non-boundary solutions, calculate the sum of normalized distances across all targets, using the following formula:

[0042] ;

[0043] in, Represents individual chromosomes The degree of congestion; Describe the objective function In individuals The next adjacent individual The value; Describe the objective function In individuals The previous adjacent individual The value; , They represent the objective functions respectively. In the current non-dominant frontier The maximum and minimum values ​​in;

[0044] Steps 2-7, in the binary bidding selection, for each pair of chromosomes in the population Chromosomes with lower non-dominant rank are preferentially selected, i.e., the leading edge is preferred. middle Minimum frontier Individuals in the group, if and If they belong to the same non-dominant rank, select the individual with the higher crowding level;

[0045] Steps 2-8: Use the sequential crossover operator to randomly select two parent generations. , Determine the intersection point and ,in , will the father Central Genes are directly copied from one generation to the other. offspring The corresponding position within, from the parent generation The remaining genes are replicated to the offspring. The offspring fill the empty spaces within the parent generation and maintain the order of the parent genes. The generation process is as follows: the parent generation Central Genes are directly copied from one generation to the other. offspring The corresponding position within, from the parent generation The remaining genes are replicated to the offspring. Empty spaces within the gene pool, while maintaining the order of the parent gene sequence;

[0046] An adaptive strategy based on crowding and iteration count is used to calculate the crossover rate. :

[0047] ,

[0048] in, , , These represent the minimum, maximum, and average values ​​of a given crossover rate; , The first The crowding degree of each chromosome individual and the average crowding degree of its frontier; , These are the current iteration count and the maximum iteration count, respectively.

[0049] Steps 2-9 involve a combined mutation method that combines exchange mutation and single-point mutation. In exchange mutation, two gene positions are randomly selected from the current solution for exchange. In single-point mutation, the value of a randomly selected gene position is replaced with a new random value. An adaptive strategy based on crowding and iteration count is used to calculate the mutation rate. This enables adaptive evolution of the population and increases its convergence. ,

[0050] in, , , These represent the minimum, maximum, and average values ​​of a given rate of variation.

[0051] Steps 2-10 involve introducing the simulated annealing (SA) algorithm into the offspring population. During the local search of the simulated annealing (SA) algorithm, in the sub-target... , , Randomly select any specific target As a search direction, search generates new entities:

[0052] ,

[0053] in, For the newly generated individuals; This is the current solution; , They are respectively The maximum and minimum solutions within the population; for A random number that is uniformly distributed within the range; The sign function is determined based on the input value. The symbol of a function Used to control the direction of disturbance, when hour If positive, otherwise Negative; The initial temperature; This is the temperature decay coefficient; This represents the current iteration number; for A random number that is uniformly distributed within the range;

[0054] By comparing the advantages and disadvantages of the new and old solutions, the Metropolis criterion is applied to determine whether the new individual should be retained and the probability of retention. The calculation formula is:

[0055] ,

[0056] in, It is a natural exponential function; Indicates the newly generated individual In the Objective function values ​​for each sub-objective; Indicates the current solution In the Objective function values ​​for each sub-objective;

[0057] Step 2-11: Perform steps 2-2 and 2-3 on the offspring population optimized by the simulated annealing (SA) algorithm;

[0058] Step 2-12: Select the parent population from steps 2-5 and 2-6. and the offspring population produced by the genetic operations in steps 2-7, 2-8, and 2-9. Merge into a new candidate population ;

[0059] Step 2-13: Introduce crowding-based selection strategies into elite selection strategies. Non-dominant hierarchy Evolutionary ranking index :

[0060] ,

[0061] according to For candidate populations Sort and select the previous ones. Individuals with different chromosomes form a new generation of elite population. This allows the selection criteria for elites to be dynamically adjusted as the evolutionary process progresses.

[0062] Step 2-14: Iterate through steps 2-2 to 2-13 until the maximum number of iterations is met. It outputs the optimal Pareto front after iteration and decodes the chromosome.

[0063] Steps 2-3 include the following steps:

[0064] Step 2-3-1: At the start of the repair, initialize an available set. Store all available storage locations;

[0065] Step 2-3-2: Examine each individual chromosome. Each gene ,judge Does it meet the constraints in step 2-2?

[0066] Step 2-3-3, the available set There are two categories, including placeholder sets. and empty space set ;

[0067] Steps 2-3-4, through statistics The frequency of genes is used to determine if there are duplicate values. If the gene... If the occurrence count of a gene is greater than 1, it indicates that there is a duplicate gene in the chromosome; otherwise, skip the repair and output the current chromosome directly.

[0068] Steps 2-3-5, for each duplicated gene Find the set of unoccupied empty spaces. any element in Used to replace the current duplicated gene Update the placeholder set after the replacement is complete. and empty space set ;

[0069] Step 2-3-6: If the repaired chromosome still has infeasible solutions or duplication problems, repeat steps 2-3-4 to 2-3-5 until the chromosome satisfies the constraints.

[0070] Steps 2-3-7: After the repair is complete, output the final feasible chromosome. Satisfy all genes All are unique values ​​and all are in the available set middle.

[0071] Step 3 includes the following steps:

[0072] Step 3-1: Obtain the Pareto front solution set obtained through iteration of the SA-NSGA-Ⅱ algorithm, and combine it with... , , Three optimization objectives constitute a multi-indicator evaluation system;

[0073] Step 3-2: Score each evaluation indicator pairwise, and construct a fuzzy judgment matrix based on the pairwise comparison criteria. ,in, It is an indicator relative to indicators The fuzzy evaluation value is represented by the triangular fuzzy number. ; , , Each represents the evaluator's assessment of the indicators. relative to indicators The most conservative estimate, the most likely estimate, and the most optimistic estimate;

[0074] Step 3-3: Calculate the index using the geometric mean method. Fuzzy weights ;

[0075] Steps 3-4: Use the centroid method to adjust the fuzzy and uncertain fuzzy weights. Defuzzification into a single, practical weight value directly used for decision-making. After normalization, the weight of each evaluation indicator is obtained. ;

[0076] Steps 3-5: Construct the decision matrix ,in Indicates the first The scheme (i.e., chromosome individual) in the 1st... The evaluation values ​​on each indicator, and the decision matrix Standardization is performed to obtain the normalized decision matrix. Among them, due to the evaluation indicators , , All of these are cost-based indicators, meaning the smaller the target value, the better. The first scheme is in the Normalized evaluation values ​​for each indicator ; , Indicators The maximum and minimum values;

[0077] Step 3-6: Utilize the weights obtained in step 3-4 The normalization obtained in steps 3-5 By weighting, we obtain the weighted normalized decision matrix. Among them, the first The first scheme is in the Weighted normalized evaluation value for each indicator ;

[0078] Steps 3-7, due to evaluation indicators , , Since all indicators are cost-related, the positive ideal solution is the minimum value among all possible solutions for that indicator, and the negative ideal solution is the maximum value among all possible solutions for that indicator. The positive ideal solution is then determined. and negative ideal solution ,in, This represents a set of cost-related indicators;

[0079] Steps 3-8: Calculate the positive ideal solution for each scheme. Euclidean distance Calculate the negative ideal solution for each solution. Euclidean distance ;

[0080] Steps 3-9: Calculate the relative proximity of each solution. The intelligent storage location allocation schemes are sorted according to their relative proximity. The greater the relative proximity, the better the scheme. The optimal storage location allocation scheme is then output.

[0081] In steps 3-8, the following formulas are used to calculate the approximate solution and the ideal solution for each scheme. Euclidean distance :

[0082] ,

[0083] in, Indicates for indicators This indicator is present in all the proposed solutions. The optimal value.

[0084] In steps 3-8, the following formulas are used to calculate the negative ideal solution for each scheme. Euclidean distance :

[0085] ,

[0086] in, Indicates for indicators This indicator is present in all the proposed solutions. The worst value.

[0087] In steps 3-9, the relative closeness of each scheme is calculated using the following formula. : .

[0088] The present invention also provides an electronic device, including a processor and a memory, the memory storing program code that, when executed by the processor, causes the processor to perform the steps of the method.

[0089] The present invention also provides a storage medium storing a computer program or instructions that, when the computer program or instructions are run on a computer, execute the steps of the method described.

[0090] Compared with the prior art, the beneficial effects of the present invention include: (1) The method can perform real-time dynamic optimization. Through digital twin technology, the optimization of warehouse location allocation can respond to changes in demand and environmental conditions in real time, and achieve dynamic adjustment and efficient decision-making. Compared with the traditional static optimization method, it greatly improves the flexibility and adaptability of warehouse operation.

[0091] (2) The SA-NSGA-II improved algorithm can optimize multiple objectives at the same time and output a variety of Pareto optimal solution sets. At the same time, the FAHP-TOPSIS method is used to quantitatively analyze the advantages and disadvantages of each optimization scheme, avoiding reliance on experience or subjective judgment.

[0092] (3) The visualization capabilities of the digital twin model make the optimization process of warehouse location allocation more intuitive and transparent, providing a solution for the digital and intelligent transformation of warehouse management. At the same time, this method is not only applicable to general automated storage and retrieval systems, but can also be extended to various scenarios such as multi-temperature storage, complex multi-layer structure warehouses, and unmanned warehouses. Attached Figure Description

[0093] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, and the advantages of the present invention in the above and / or other aspects will become clearer.

[0094] Figure 1 The digital twin framework architecture design diagram for the method provided by this invention.

[0095] Figure 2 Technical diagram of the cargo location allocation scheme provided by the present invention.

[0096] Figure 3The flowchart of the improved SA-NSGA-Ⅱ algorithm provided for this invention.

[0097] Figure 4 Chromosome encoding and decoding diagram for the method provided by this invention.

[0098] Figure 5 Cross operator graph for the method provided by the present invention.

[0099] Figure 6 A mutation operator diagram for the method provided by this invention.

[0100] Figure 7 A flowchart illustrating the gene repair process of the method provided by this invention.

[0101] Figure 8 The flowchart illustrates the method provided in this invention, which combines the Fuzzy Hierarchical Analysis (FAHP) and the Topology-Topology-Sorting Method (TOPSIS) to determine the optimal cargo location allocation scheme.

[0102] Figure 9 Three-dimensional comparison of the optimal Pareto front of the improved SA-NSGA-II algorithm with the traditional NSGA-II and MOPSO algorithms provided in this invention.

[0103] Figure 10 This is a schematic diagram of the storage location allocation results obtained by the MOPSO algorithm.

[0104] Figure 11 This is a schematic diagram of the cargo location allocation results obtained by the NSGA-II algorithm.

[0105] Figure 12 This is a schematic diagram of the cargo location allocation results obtained by the improved SA-NSGA-Ⅱ method of this invention.

[0106] Figure 13 An example diagram illustrating the implementation of the digital twin warehouse method provided by this invention. Detailed Implementation

[0107] Reference Figure 1 This invention provides a method for allocating storage locations in an automated three-dimensional warehouse using digital twins, comprising: constructing a five-dimensional digital twin model, including a physical layer, a twin layer, a data layer, a service layer, and connections, covering the entire process from physical data collection to virtual twin and then to decision feedback.

[0108] The physical layer is the actual location for goods storage and retrieval operations, representing the actual physical entities and environment in the system, including physical entities such as shelves, stacker cranes, and inbound / outbound platforms; data from the physical layer is uploaded to the data layer through real-time interaction, while simultaneously receiving optimization instructions from the service layer;

[0109] The digital twin layer is a digital mirror of the physical space. By constructing a virtual-to-real twin model that maps the physical warehouse, it provides simulation support for location allocation and warehouse operations. The warehouse digital twin model relies on 3D modeling technology. In Blender software, it constructs accurate 3D models based on the shape, color, and material characteristics of physical entities. High-quality rendering ensures a high degree of consistency between the virtual model and the real warehouse environment. Simultaneously, it sets the model scene tree based on the parent-child relationship information between 3D models. The model data is output as 3D model files in .gltf or .glb format. The Three.js engine loads and parses the .gltf or .glb format 3D models on the web. Combined with environmental data collected by IoT devices, it achieves real-time mapping of warehouse layout, equipment status, and environmental parameters. The Vue3 framework dynamically binds warehouse data to ensure real-time updates of shelf inventory, stacker crane location, and operating status. Integrating all geometric models, dynamic data, and rendering scenes, a complete warehouse digital twin system is constructed.

[0110] As the data support layer connecting the physical layer and the twin layer, the data layer acquires dynamic and static data, cleans and stores the data, and provides input to the service layer to ensure the real-time performance and accuracy of the system. Dynamic data is used to reflect the real-time operation of the automated warehouse, including cargo flow data, equipment status, environmental data, etc., while static data consists of basic data such as warehouse geometry, equipment physical constraints, and cargo material properties.

[0111] The service layer provides users with an intuitive operating interface and scientific decision support, realizing a combination of intelligent and human decision-making in cargo location allocation.

[0112] The connectivity includes data connectivity and control connectivity. Data connectivity is used to collect physical layer data to the data layer through IoT devices (Internet of Things) and network devices. The information processed by the data layer is input to the twin layer for modeling and simulation, and then input to the service layer to generate a storage location allocation plan. Control connectivity is used to feed back storage location allocation instructions to the physical layer to guide actual operations and to the twin layer for virtual operations.

[0113] Reference Figure 2 The specific steps of combining intelligent decision-making with human decision-making are as follows:

[0114] Step 1: Obtain the processed location status, cargo characteristics, and equipment status information from the data layer;

[0115] Step 2: Use the intelligent improvement algorithm SA-NSGA-Ⅱ to generate a set of optional storage location allocation schemes. ;

[0116] Step 3: Using a decision-making method combining fuzzy hierarchical analysis (FAHP) and the topology-topic solution ranking method (TOPSIS), the set of storage location allocation schemes is analyzed. The best solution for intelligent warehouse allocation was selected from among the top choices.

[0117] Step 4: Display the visualized results of the intelligent storage location allocation scheme on the human-computer interaction interface, including the location and storage location number of each goods task, and display the storage location utilization and equipment load through dynamic charts, allowing users to judge the intelligent storage location allocation scheme. Does it need adjustment? If so, the modified solution set is as follows: ,in Indicates the adjusted number The system outputs a storage location allocation plan for each type of cargo; otherwise, it directly outputs an intelligent allocation plan. ;

[0118] Step 5, according to the final plan This generates operation instructions for physical layer devices, which are then used by the corresponding warehousing equipment to perform the task of allocating storage locations.

[0119] Step 6: By monitoring the status of goods and equipment in real time, the execution results of the physical layer and the twin layer are fed back. If the goods are not placed in the correct position, the alarm is recorded and triggered. After the solution is completed, the data layer information is updated.

[0120] Reference Figure 3 To address the problems of significantly increased computational resource consumption, slow convergence speed, and poor population diversity maintenance in complex real-world applications, an improved SA-NSGA-II algorithm is proposed to optimize the cargo location allocation problem. Step 2 includes the following steps:

[0121] Step 2-1, randomly generate a population of size . Initialized population The first in the population Individual chromosomes The Middle One gene Representing a cargo allocation task, the three-dimensional coordinates of the gene are encoded into one-dimensional integers based on the concept of rasterization;

[0122] Step 2-2: Perform a feasibility check on the genes in the individual chromosome to monitor whether the individual meets the constraints, which include:

[0123] Uniqueness constraint of storage location: ,

[0124] Shelf load-bearing capacity constraints: ,

[0125] Shelf capacity constraints: ,

[0126] Steps 2-3: Set up gene repair operators for individual chromosomes. Gene values ​​that do not meet the constraints Repair is performed, and the repaired gene is ;

[0127] Steps 2-4: Calculate the fitness of each individual in the population within the objective function space for cargo location allocation optimization, and establish the following multi-objective cargo location optimization mathematical model:

[0128] ,

[0129] Establish the following criteria for measuring the balance of all shelves in an automated warehouse:

[0130] ,

[0131] Calculate the distance between similar items in a single row of shelving using Manhattan distance:

[0132] ,

[0133] Establish the following objective function for the stacker crane's inbound and outbound times:

[0134] ,

[0135] Steps 2-5, for each chromosome individual in the population ,judge Dominance relationships with other chromosomes, if they satisfy In sub-target , , If a chromosome is not inferior to other chromosomes in any of its sub-targets, and is strictly superior to other chromosomes in at least one sub-target, then it is called a chromosome. Dominates other chromosomes, statistical dominance number of chromosomes and record Dominant set of chromosomes Find all chromosomes, The chromosomes are not dominated by any other cleavage, forming the first frontier. Remove from the population Continue processing the remaining chromosomes, repeating the above process until all chromosomes have been assigned to their corresponding frontal zones. ;

[0136] Steps 2-6, in sub-targets , , Randomly select any specific target The non-dominated frontier, i.e., the first frontier The chromosomes in the selection are based on the chosen objective function. The values ​​are sorted, the boundary solutions of the target space are marked as extreme solutions, and an infinite crowding degree is assigned. For non-boundary solutions, calculate the sum of normalized distances across all targets, using the following formula:

[0137] ;

[0138] Steps 2-7, in the binary bidding selection, for each pair of chromosomes in the population Chromosomes with lower non-dominant rank are preferentially selected, i.e., the leading edge is preferred. middle Minimum frontier Individuals in the group, if and If they belong to the same non-dominant rank, select the individual with the higher crowding level;

[0139] Steps 2-8: Use the sequential crossover operator to randomly select two parent generations. , Determine the intersection point and ,in , will the father Central Genes are directly copied from one generation to the other. offspring The corresponding position within, from the parent generation The remaining genes are replicated to the offspring. The offspring fill the empty spaces within the parent generation and maintain the order of the parent genes. The generation process and Similarly, the parent generation Central Genes are directly copied from one generation to the other. offspring The corresponding position within, from the parent generation The remaining genes are replicated to the offspring. It fills the empty spaces within the gene pool and maintains the order of the parent genes.

[0140] An adaptive strategy based on crowding and iteration count is used to calculate the crossover rate. :

[0141] ,

[0142] Steps 2-9 involve a combined mutation method that combines exchange mutation and single-point mutation. In exchange mutation, two gene positions are randomly selected from the current solution for exchange. In single-point mutation, the value of a randomly selected gene position is replaced with a new random value. An adaptive strategy based on crowding and iteration count is used to calculate the mutation rate. This enables adaptive evolution of the population and increases its convergence. .

[0143] Steps 2-10 involve introducing the simulated annealing (SA) algorithm into the offspring population. During the local search of the simulated annealing (SA) algorithm, in the sub-target... , , Randomly select any specific target As a search direction, search generates new entities:

[0144] ,

[0145] By comparing the advantages and disadvantages of the new and old solutions, the Metropolis criterion is applied to determine whether the new individual should be retained and the probability of retention. for:

[0146] .

[0147] Step 2-11: Perform feasibility check in step 2-2 and gene repair in step 2-3 on the offspring population optimized by the simulated annealing SA algorithm;

[0148] Step 2-12: Select the parent population selected through the fast non-dominated sorting in Step 2-5 and the crowding calculation in Step 2-6. and the offspring population produced by the genetic operations of step 2-7 selection, step 2-8 mutation, and step 2-9 crossover. Merge into a new candidate population ;

[0149] Step 2-13: Introduce crowding-based selection strategies into elite selection strategies. Non-dominant hierarchy Evolutionary ranking index :

[0150] ,

[0151] according to For candidate populations Sort and select the previous ones. Individuals form a new generation of elite population This allows the selection criteria for elites to be dynamically adjusted as the evolutionary process progresses.

[0152] Step 2-14: Iterate through steps 2-2 to 2-13 until the maximum number of iterations is met. It outputs the optimal Pareto front after iteration and decodes the chromosome.

[0153] Reference Figure 4 In the encoding design, each integer represents a location number in a storage allocation scheme, and the chromosome length equals the number of tasks. During decoding, for... Row, List, If the shelf on the first floor, The location code assigned to each item is: The three-dimensional coordinates of the assigned cargo location are:

[0154] ,

[0155] in, For rounding, This is the modulo operator. Taking a 12-row, 12-column, 10-layer shelf as an example, there are 1440 available storage locations. If there are 8 goods waiting to be assigned to storage locations, each good will be randomly assigned to an available storage location. Good 1 is assigned to storage location number 47. After decoding, the three-dimensional coordinates of the storage location assigned to storage location 1 are (1, 11, 4), and so on.

[0156] Reference Figure 5 To preserve the superior genes of the parent generation while avoiding illegal combinations and ensuring the legitimacy of the offspring, a sequential crossover operator is used to crossover the task sequences allocated according to their storage locations. Taking 8 tasks as an example:

[0157] Randomly select a pair of chromosomes and determine the location of the crossover point. , :

[0158] Parent generation 1: 47, 234, 345, 149, 154, 122, 277, 93;

[0159] Parent generation 2: 88, 1, 47, 288, 349, 149, 52, 206;

[0160] Chromosome 1 from the father The genes from the parent chromosome 2 are directly copied to the corresponding positions on the offspring chromosome 1. The remaining genes from the parent chromosome 2 are copied to the empty positions on the offspring chromosome 1, maintaining the order of the parent genes. The same process is repeated on the offspring chromosome 2. The genes from the parent chromosome 1 are directly copied to the corresponding positions on chromosome 2 of the offspring, and the remaining genes from the parent chromosome 1 are copied to the empty positions on chromosome 2 of the offspring, while maintaining the order of the parent genes.

[0161] Offspring 1: 88, 1, 345, 149, 154, 47, 288, 349;

[0162] Offspring 2: 234, 345, 47, 288, 349, 149, 154, 122;

[0163] Reference Figure 6 To improve solution diversity and reduce the risk of getting trapped in local optima, a combined mutation method using exchange mutation and single-point mutation is employed. Taking the eight tasks in offspring 2 generated by crossover as an example:

[0164] First, randomly select gene locations from the current chromosome. , Exchange their genes:

[0165] Offspring 2: 88, 1, 345, 149, 154, 47, 288, 349;

[0166] Offspring 2 after the crossover mutation: 88, 47, 345, 149, 154, 1, 288, 349;

[0167] Then randomly select a gene location Replace its value with a new random value:

[0168] Offspring 2 after single-point mutation: 88, 47, 345, 286, 154, 1, 288, 349;

[0169] Reference Figure 7 The optimization algorithm involves randomness during the selection, crossover, and mutation processes, which may result in some individuals in the generated chromosomes not meeting the constraints. To ensure the feasibility of offspring, a gene repair operator is implemented. Steps 2-3 include the following steps:

[0170] Step 2-3-1: At the start of the repair, initialize an available set. Store all available storage locations;

[0171] Step 2-3-2: Examine each individual chromosome. Each gene ,judge Does it meet the constraints in step 2-2?

[0172] Step 2-3-3, the available set There are two categories, including placeholder sets. and empty space set ;

[0173] Steps 2-3-4, through statistics The frequency of genes is used to determine if there are duplicate values. If the gene... If the occurrence count of a gene is greater than 1, it indicates that there is a duplicate gene in the chromosome; otherwise, skip the repair and output the current chromosome directly.

[0174] Steps 2-3-5, for each duplicated gene Find the set of unoccupied empty spaces. any element in Used to replace the current duplicated gene Update the placeholder set after the replacement is complete. and empty space set ;

[0175] Step 2-3-6: If the repaired chromosome still has infeasible solutions or duplication problems, repeat steps 2-3-4 to 2-3-5 until the chromosome satisfies the constraints.

[0176] Steps 2-3-7: After the repair is complete, output the final feasible chromosome. Satisfy all genes All are unique values ​​and all are in the available set middle.

[0177] Reference Figure 8 After obtaining a set of optimal Pareto solutions, a decision-making method combining fuzzy hierarchical analysis (FAHP) and topological sorting method (TOPSIS) is used to determine a set of optimal location allocation schemes from multiple allocation schemes as the final scheme. Step 3 includes the following steps:

[0178] Step 3-1: Obtain the Pareto front solution set obtained through iteration of the SA-NSGA-Ⅱ algorithm, and combine it with... , , Three optimization objectives constitute a multi-indicator evaluation system;

[0179] Step 3-2: Score each evaluation indicator pairwise, and construct a fuzzy judgment matrix based on the pairwise comparison criteria. Triangular fuzzy number The scale values ​​1-9 and their corresponding meanings are shown in Table 1.

[0180] Table 1

[0181]

[0182] Step 3-3: Calculate the index using the geometric mean method. Fuzzy weights ;

[0183] Steps 3-4: Use the centroid method to adjust the fuzzy and uncertain fuzzy weights. Defuzzification into a single, practical weight value directly used for decision-making. After normalization, the weight of each evaluation indicator is obtained. ;

[0184] Steps 3-5: Construct the decision matrix For the decision matrix Standardization is performed to obtain the normalized decision matrix. ;

[0185] Step 3-6: Utilize the weights obtained in step 3-4 The normalization obtained in steps 3-5 By weighting, we obtain the weighted normalized decision matrix. ;

[0186] Steps 3-7, due to evaluation indicators , , Since all indicators are cost-related, the positive ideal solution is the minimum value among all possible solutions for that indicator, and the negative ideal solution is the maximum value among all possible solutions for that indicator. The positive ideal solution is then determined. and negative ideal solution ;

[0187] Steps 3-8: Calculate the positive ideal solution for each scheme. Euclidean distance Calculate the negative ideal solution for each solution. Euclidean distance ;

[0188] Steps 3-9: Calculate the relative proximity of each solution. The intelligent storage location allocation schemes are sorted according to their relative proximity. The greater the relative proximity, the better the scheme. The optimal storage location allocation scheme, Pareto, is then output.

[0189] Reference Figure 9 To verify the applicability and effectiveness of the improved SA-NSGA-Ⅱ algorithm, a simulation experiment was conducted using data from 600 finished fabrics entering the warehouse of a textile company in April 2024. The data parameters of the first 20 turnover box numbers are shown in Table 2 below.

[0190] Table 2

[0191]

[0192] The basic parameters of the automated storage and retrieval system are shown in Table 3 below.

[0193] Table 3

[0194]

[0195] Considering the problem size, the MOPSO, NSGA-II and improved SA-NSGA-II algorithms were used for optimization, and the parameter settings of the three algorithms are shown in Table 4.

[0196] Table 4

[0197]

[0198] The population distribution under different dimensions after iterations of the MOPSO, NSGA-II, and improved SA-NSGA-II optimization methods shows that, compared with the initial population, the MOPSO, NSGA-II, and improved SA-NGSA-II algorithms all significantly improve population quality, and the values ​​of the three objective functions are significantly smaller. Compared with the traditional NSGA-II and MOPSO algorithms, the improved SA-NSGA-II algorithm performs better on the three objective functions, and the Pareto solution is more competitive and has a better distribution uniformity, proving the effectiveness of the improved SA-NSGA-II algorithm.

[0199] Reference Figure 10 , Figure 11 and Figure 12 To determine the optimal cargo location allocation scheme, the fuzzy judgment matrix is ​​shown in Table 5 below.

[0200] Table 5

[0201]

[0202] The objective function was obtained by fuzzy hierarchical analysis (FAHP). , , The corresponding fuzzy weights are respectively , , After deblurring , , After normalization, the combined weights of the three objective functions are as follows: , , .

[0203] TOPSIS, a sorting method that approximates the ideal solution, was used to... Figure 9 The obtained Pareto solution sets are sorted in a comprehensive manner, where the positive ideal solutions are: The negative ideal solution is The optimal Pareto solution after comprehensive sorting is shown in Table 6 below.

[0204] Table 6

[0205]

[0206] To more intuitively demonstrate the results of the optimized storage location allocation, the optimal allocation results were visualized. As shown in the figure, the optimal Pareto solutions of the MOPSO and NSGA-II algorithms have obvious flaws, including unreasonable storage locations, which negatively impact the efficiency of stacker crane inbound and outbound operations. In contrast, the simulation results of the SA-NSGA-II algorithm are stable, further proving its superiority in storage location optimization problems.

[0207] Reference Figure 13 Based on the aforementioned method, a virtual-real synchronization simulation platform is constructed, enabling real-time communication and action synchronization between physical devices and twin models. The final allocation results are presented in the three-dimensional visualization interactive interface shown in the figure. The system intuitively displays multiple key indicators within the warehouse to operators through multi-dimensional visualization charts, effectively leveraging data application capabilities to improve warehouse operation efficiency and transparency.

[0208] This invention provides a method for allocating storage locations in an automated three-dimensional warehouse using digital twin-driven technology. Many methods and approaches exist for implementing this technical solution; the above description is merely a preferred embodiment of the invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications should also be considered within the scope of protection of this invention. All components not explicitly stated in this embodiment can be implemented using existing technologies.

Claims

1. A method for allocating storage locations in an automated three-dimensional warehouse using digital twin-driven technology, characterized in that, include: Construct a five-dimensional digital twin model, including the physical layer, twin layer, data layer, service layer, and connectivity layer; The physical layer is the actual location for goods storage and retrieval operations, representing the actual physical entities and environment. The physical entities include shelves, stacker cranes, inbound and outbound platforms, turnover boxes, and goods. Data from the physical layer is uploaded to the data layer in real time and simultaneously receives optimization instructions from the service layer. The twin layer is a digital mirror of the physical space. By constructing a digital twin model of the warehouse that maps the virtual and real physical warehouse, it provides simulation support for location allocation and warehouse operations. The data layer, as the data support layer connecting the physical layer and the twin layer, provides input to the service layer by acquiring dynamic and static data, cleaning and storing the data. The dynamic data is used to reflect the real-time operation of the automated warehouse, including cargo flow data, equipment status, and environmental data. The static data includes warehouse geometric dimensions, equipment physical constraints, and cargo material properties. The service layer provides users with a human-computer interaction interface and decision support, realizing a combination of intelligent and manual decision-making in cargo location allocation; The connection includes a data connection and a control connection. The data connection is used to collect physical layer data to the data layer through IoT devices and network devices. The information processed by the data layer is input to the twin layer for modeling and simulation, and input to the service layer for generating a storage location allocation scheme. The control connection is used to feed back storage location allocation instructions to the physical layer to guide actual operation, and feed them back to the twin layer for virtual operation. The service layer performs the following steps: Step 1: Obtain the processed location status, cargo characteristics, and equipment status information from the data layer; Step 2: Use the intelligent improvement algorithm SA-NSGA-Ⅱ to generate a set of optional storage location allocation schemes S. A ={S1,S2,...,S} I }, where S I This represents the I-th intelligent storage location allocation scheme; Step 3: Using a decision-making method combining fuzzy hierarchical analysis (FAHP) and the topology-topic solution ranking method (TOPSIS), the set of storage location allocation schemes S is used to determine the optimal allocation schemes. A Zhongjue selects the best intelligent storage location allocation solution S I ={s1,s2,...,s r }, where s r This represents the coordinates of the storage location to which the r-th item is assigned; Step 4: Display the visualized results of the intelligent storage location allocation scheme on the human-computer interaction interface, including the location and storage location number of each goods task. Dynamic charts will display the storage location utilization and equipment load, allowing the user to evaluate the intelligent storage location allocation scheme S. I Does it need adjustment? If so, the modified set of solutions is S. M ={s′1,s′2,...,s′ r }, where s r ′ represents the adjusted location allocation scheme for the r-th item; otherwise, directly output the intelligent allocation scheme S. I ; Step 5, according to the final solution S F ∈{S I ,S M } generates operation instructions for physical layer devices, which are then used by the warehousing equipment to perform the task of allocating storage locations; Step 6: By monitoring the status of goods and equipment in real time, the execution results of the physical layer and the twin layer are fed back. If the goods are not placed in the correct position, the alarm is recorded and triggered. After the solution is completed, the data layer information is updated.

2. The method according to claim 1, characterized in that, Step 2, the intelligent improvement algorithm SA-NSGA-Ⅱ includes the following steps: Step 2-1: Randomly generate an initial population P0 = {X1, X2, ..., X} with a population size of N. N }, the individual X with chromosome n in the population n ={χ1,χ2,...,χ L The i-th gene χ in} i Representing a cargo allocation task, the three-dimensional coordinates of the gene are encoded into one-dimensional integers based on the rasterization concept, n∈N, i∈L; Step 2-2: Perform a feasibility check on the genes in the individual chromosome to monitor whether the individual meets the constraints, which include: Uniqueness constraint of storage location: Shelf load-bearing capacity constraints: Shelf capacity constraints: Where a, b, and c are the row, column, and shelf number, respectively; (x i ,y i ,z i ) represents the coordinates of the i-th item on the shelf; O(x i ,y i ,z i ) represents the occupancy status of the storage location allocated to the i-th item; m y This represents the total weight of all goods on a single shelf; m yz Indicates the total weight of all goods on a single row of shelves; M, M j These represent the maximum load-bearing capacity of a single row of shelving and the maximum load-bearing capacity of a single layer of shelving, respectively; N * Represents the set of positive integers; Steps 2-3: Set up the gene repair operator for individual chromosome X. n Gene values ​​x that do not meet the constraints i The repair is performed, and the repaired gene is X′. n =R(X) n ), where R(·) represents the gene repair operator; Steps 2-4: Calculate the fitness of each individual in the population within the objective function space for cargo location allocation optimization, and establish the following multi-objective cargo location optimization mathematical model: F = min(f1,f2,f3), Where F represents the overall optimization objective; f1 represents the balance of the rack; f2 represents the distance between similar goods in a single row of racks; and f3 represents the stacker crane's inbound and outbound time. Establish the following criteria for measuring the balance of all shelves in an automated warehouse: Where W is the width of the storage space; m i Indicates the quality of the i-th item; Calculate the distance between similar items in a single row of shelving using Manhattan distance: Where H is the height of the storage location; (x j ,y j ,z j Let ) represent the coordinates of the j-th item on the shelf, where the j-th item is of the same type as the i-th item; S represents the type of goods; g represents the total number of goods of the same type; C(S) represents the set of all goods of type S. Establish the following objective function for the stacker crane's inbound and outbound times: Where L is the length of the storage location; v x v y v z These represent the stacker crane forklift speed, horizontal travel speed, and vertical travel speed, respectively; r i Represents the turnover rate of the i-th goods; (X i ,Y i Z i () represents the coordinates of the inbound / outbound station corresponding to the single row of shelves containing the i-th item; Steps 2-5, for each chromosome individual X′ in the population n Determine X′ n Dominance relationships with other chromosomes, if X′ n If a chromosome is not inferior to other chromosomes in sub-targets f1, f2, and f3, and is strictly superior to other chromosomes in at least one sub-target, then X′ is called X′. n It controls other chromosomes and statistically controls X′. n The number of chromosomes n(X′) n ), and record X′ n The dominant set of chromosomes S(X′) n Find all n(X′) n A chromosome with ) = 0, n(X′) n Chromosomes with a value of 0 are not dominated by any other solutions and constitute the first frontier F1. F1 chromosomes are removed from the population, and the remaining chromosomes are processed, repeating steps 2-5 until all chromosomes are assigned to their corresponding frontier F1 chromosomes. k ; Steps 2-6: Randomly select any specific target f from sub-targets f1, f2, and f3. m For m∈1,2,3, the chromosomes in the non-dominated front, i.e., the first front F1, are determined according to the selected objective function f. m The values ​​are sorted, the boundary solutions of the target space are marked as extreme solutions, and an infinite crowding degree is assigned. For non-boundary solutions, calculate the sum of normalized distances across all targets, using the following formula: in, Represents the individual chromosome X′ n The degree of congestion; f m (X′ n+1 ) represents the objective function f m In individual X′ n The next adjacent individual X′ n+1 The value of f; m (X′ n-1 ) represents the objective function f m In individual X′ n The previous neighboring individual X′ n-1 The value; Let f represent the objective function respectively. m The maximum and minimum values ​​in the current nondominated frontier F1; Steps 2-7, in the binary bidding selection, for each pair of chromosomes (X) in the population p ,X q For each chromosome, p, q ∈ N, chromosomes with lower non-dominant rank are preferentially selected, i.e., chromosomes at the leading edge F are preferentially selected. k Minimum frontier of k Individuals in the group, if X p and X q If they belong to the same non-dominant rank, select the individual with the higher crowding level; Step 2-8, using the order crossover operator, randomly select two parents P a , P b , determine the crossover points p1 and p2, where p1 < p2, and directly copy the genes of parent P a within [p1, p2] to the corresponding positions in the offspring C a of parent P a . Copy the remaining genes from parent P b to the empty positions in the offspring C a , and maintain the arrangement order of the parent genes. The generation process of the offspring C b is as follows: directly copy the genes of parent P b within [p1, p2] to the corresponding positions in the offspring C b of parent P b . Copy the remaining genes from parent P a to the empty positions in the offspring C b , and maintain the arrangement order of the parent genes; An adaptive strategy based on crowding and iteration count is used to calculate the crossover rate P. c : Among them, P cmin P cmax P cavg These represent the minimum, maximum, and average values ​​of a given crossover rate; Let be the crowding degree of the individual with chromosome n and the average crowding degree of its front edge, respectively; t, t max These are the current iteration count and the maximum iteration count, respectively. Steps 2-9 involve a combined mutation method that integrates exchange mutation and single-point mutation. In exchange mutation, two gene positions are randomly selected from the current solution for exchange. In single-point mutation, a gene position is randomly selected and its value is replaced with a new random value. An adaptive strategy based on crowding and iteration count is used to calculate the mutation rate P. m This enables adaptive evolution of the population and increases its convergence. Among them, P mmin P mmax P mavg These represent the minimum, maximum, and average values ​​of a given rate of variation. Step 2-10: Introduce the simulated annealing (SA) algorithm into the offspring population. During the local search of the simulated annealing (SA) algorithm, randomly select any specific target f from the sub-targets f1, f2, and f3. m As a search direction, search generates new entities: X new =X pre +sign(R-0.5)·(X max -X pre )·(T0·α k )·ξ, Among them, X new For the newly generated individual; X pre X is the current solution; max X min X pre The maximum and minimum solutions within the population; R is a uniformly distributed random number within [0,1]; sign(·) represents the sign function, the sign of which is determined by the input value, and sign(R-0.5) controls the direction of the perturbation. When R>0.5, sign(R-0.5) is positive, otherwise sign(R-0.5) is negative; T0 is the initial temperature; α is the temperature decay coefficient; k is the current iteration number; ξ is a uniformly distributed random number within [0,1]. By comparing the superiority and inferiority of the old and new solutions, the Metropolis criterion is applied to determine whether the new individual should be retained, with a retention probability P. save The calculation formula is: Where exp(·) is the natural exponential function; f m (X new ) represents the newly generated individual X new The objective function value on the m-th sub-objective; f m (X pre ) represents the current solution X pre The objective function value on the m-th sub-objective; Step 2-11: Perform steps 2-2 and 2-3 on the offspring population optimized by the simulated annealing (SA) algorithm; Step 2-12, select the parent population P selected in steps 2-5 and 2-6. t The offspring population Q generated by the genetic operations in steps 2-7, 2-8, and 2-9 t Merge into a new candidate population R t ; Step 2-13: Introduce crowding-based selection strategies into elite selection strategies. Non-dominant level F k The evolutionary ranking index S of (n) n : Press S n For candidate population R t Sort the individuals and select the top N individuals based on their chromosomes to form a new generation of elite population P. t+1 This allows the selection criteria for elites to be dynamically adjusted as the evolutionary process progresses. Step 2-14: Iterate through steps 2-2 to 2-13 until the maximum number of iterations t is met. max It outputs the optimal Pareto front after iteration and decodes the chromosome.

3. The method according to claim 2, characterized in that, Steps 2-3 include the following steps: Step 2-3-1: At the start of the repair, initialize an available set A to store all available storage locations; Step 2-3-2: Examine each individual's X chromosome. n ={χ1,χ2,...,χ L Each gene χ in} i Determine χ i Does it meet the constraints in step 2-2? Step 2-3-3: Classify the available set A into two categories: the placeholder set U and the empty set L; Steps 2-3-4, through statistical analysis of X n The frequency of genes in the gene matrix is ​​used to determine if there are duplicate values. If the gene χ² value is repeated... i If the occurrence count of a gene is greater than 1, it indicates that there is a duplicate gene in the chromosome; otherwise, skip the repair and output the current chromosome directly. Steps 2-3-5, for each repeated gene χ i Find any element l in the set L of unoccupied empty spaces. i , to replace the current repetitive gene χ i After the replacement is complete, update the placeholder set U and the empty set L; Step 2-3-6: If the repaired chromosome still has infeasible solutions or duplication problems, repeat steps 2-3-4 to 2-3-5 until the chromosome satisfies the constraints. Steps 2-3-7: After the repair is complete, output the final feasible chromosome X′. n Satisfying all gene χ′ i All are unique values ​​and all are in the available set A.

4. The method according to claim 3, characterized in that, Step 3 includes the following steps: Step 3-1: Obtain the Pareto front solution set obtained by the SA-NSGA-Ⅱ algorithm iteration, and construct a multi-index evaluation system by combining the three optimization objectives f1, f2, and f3. Step 3-2: Score each evaluation indicator pairwise, and construct a fuzzy judgment matrix based on the pairwise comparison criteria. u,v∈1,2,3, where It is the fuzzy evaluation value of indicator u relative to indicator v, denoted as the triangular fuzzy number. l uv m uv u uv These represent the evaluator's most conservative, most likely, and most optimistic estimates of indicator u relative to indicator v, respectively. Step 3-3: Calculate the fuzzy weights of index u using the geometric mean method. Steps 3-4: Use the centroid method to adjust the fuzzy and uncertain fuzzy weights. Defuzzification into a single, practical weight value directly used for decision-making. The weight of each evaluation indicator is obtained after normalization. Steps 3-5: Construct the decision matrix F nu =(f nu ) N×3 , where f nu Let F represent the evaluation value of the nth option on the uth index, with respect to the decision matrix F. nu After standardization, the normalized decision matrix R is obtained. nu =(r nu ) N×3 ; where the normalized evaluation value of the nth scheme on the uth index. These are the maximum and minimum values ​​of index u, respectively; Step 3-6: Use the weights w obtained in step 3-4 u The normalized R obtained in steps 3-5 nu By weighting, we obtain the weighted normalized decision matrix V. nu =(v nu ) N×3 ; where the weighted normalized evaluation value v of the nth scheme on the uth index is nu =w u ′·r nu ; Steps 3-7: Determine the ideal solution and negative ideal solution Among them, U - This represents a set of cost-related indicators; Steps 3-8: Calculate the positive ideal solution A for each scheme. + Euclidean distance Calculate the negative ideal solution A for each solution. - Euclidean distance Steps 3-9: Calculate the relative proximity C of each scheme. n The intelligent storage location allocation schemes are sorted according to their relative proximity. The greater the relative proximity, the better the scheme. The optimal storage location allocation scheme is then output.

5. The method according to claim 4, characterized in that, In steps 3-8, the following formula is used to calculate the approximate solution A for each scheme. + Euclidean distance in, This represents the optimal value of index u among all possible solutions.

6. The method according to claim 5, characterized in that, In steps 3-8, the negative ideal solution A for each scheme is calculated using the following formula. - Euclidean distance in, This represents the worst value of index u among all possible solutions.

7. The method according to claim 6, characterized in that, In steps 3-9, the relative closeness C of each scheme is calculated using the following formula. n :

8. An electronic device, characterized in that, It includes a processor and a memory, the memory storing program code that, when executed by the processor, causes the processor to perform the steps of the method as described in any one of claims 1 to 7.

9. A storage medium, characterized in that, It stores a computer program or instructions that, when run on a computer, perform the steps of the method as described in any one of claims 1 to 7.

Images