Migration method and device for cross-warehouse scheduling, equipment and medium
By constructing environmental fingerprint vectors and policy gene libraries in cross-warehouse scheduling, the problems of lack of flexibility and verification in cross-warehouse scheduling migration methods are solved, and efficient and secure policy migration and scheduling optimization are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHENZHEN TODAY INT SOFTWARE TECH CO LTD
- Filing Date
- 2026-04-13
- Publication Date
- 2026-06-23
Smart Images
Figure CN122048248B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of warehouse scheduling technology, and in particular to a method, apparatus, equipment and medium for cross-warehouse scheduling migration. Background Technology
[0002] With the rapid development of the smart warehousing industry, the number and scale of automated warehouses continue to expand. When enterprises deploy multiple warehouses in different regions, the scheduling system for each warehouse often needs to be optimized from scratch in terms of parameters and trained in terms of strategies. This results in long deployment cycles, high debugging costs, and low early operational efficiency for new warehouses. At the same time, the extensive successful scheduling experience accumulated by existing warehouses over long-term operation is difficult to systematically migrate to new warehouses, creating a serious "experience silo" problem. Here, "experience silos" refers to the phenomenon where the scheduling knowledge of different warehouses is isolated from each other and cannot be shared or reused.
[0003] In existing technologies, some solutions attempt to transfer model parameters from one scenario to another through transfer learning (a machine learning method that applies model knowledge from a learned task to a new task) or model fine-tuning. However, these methods have the following shortcomings: First, the granularity of the transfer is the entire model parameters rather than independently reusable policy units, lacking flexibility. In other words, "full migration" cannot achieve selective reuse of individual scheduling policies. Second, there is a lack of a systematic description and matching mechanism for the physical and business differences between the source and target scenarios, making it impossible to quantify the similarity between the two warehouse environments. Third, there is a lack of security verification and progressive adaptation processes for policies in the target environment. Directly migrating unverified policies carries the risk of scheduling failures or even security incidents due to differences in equipment or business.
[0004] Therefore, existing warehouse scheduling migration methods lack flexibility, cannot quantify warehouse similarity, and lack verification, resulting in low migration efficiency. Summary of the Invention
[0005] This invention provides a method, apparatus, device, and medium for cross-warehouse scheduling migration, aiming to solve the problems of low migration efficiency caused by the lack of flexibility, inability to quantify warehouse similarity, and lack of verification in existing cross-warehouse scheduling migration methods.
[0006] To address the aforementioned problems, in a first aspect, embodiments of the present invention provide a migration method for cross-warehouse scheduling, the method comprising:
[0007] Obtain multidimensional collected data from the source warehouse, and mine the multidimensional collected data to obtain a candidate strategy gene set;
[0008] The candidate strategy gene set is evaluated for fitness based on the evaluation strategy to obtain a validated strategy gene library.
[0009] Obtain the physical parameters and business characteristic parameters of the target warehouse, and construct an environmental fingerprint from the physical parameters and the business characteristic parameters to obtain the target environmental fingerprint vector;
[0010] The target environment fingerprint vector and the validated strategy gene library are subjected to inheritance screening using inheritance screening rules to obtain a candidate inheritance gene set.
[0011] The localized scheduling strategy set is obtained by adjusting the candidate inheritance gene set based on the adjustment strategy.
[0012] In the target warehouse, migration control instructions are generated and executed according to the localized scheduling policy set.
[0013] Secondly, embodiments of this application provide a migration device for cross-warehouse scheduling, the device comprising:
[0014] The mining unit is used to acquire multidimensional collected data from the source warehouse and mine the multidimensional collected data to obtain a candidate strategy gene set.
[0015] An evaluation unit is used to evaluate the fitness of the candidate strategy gene set based on an evaluation strategy, and obtain a validated strategy gene library.
[0016] The construction unit is used to obtain the physical parameters and business feature parameters of the target warehouse, and to construct an environmental fingerprint of the physical parameters and the business feature parameters to obtain the target environmental fingerprint vector.
[0017] A screening unit is used to perform inheritance screening on the target environment fingerprint vector and the verified strategy gene library using inheritance screening rules to obtain a candidate inheritance gene set.
[0018] An adjustment unit is used to adjust the candidate inheritance gene set based on an adjustment strategy to obtain a localized scheduling strategy set;
[0019] The migration unit is used to generate and execute migration control instructions in the target warehouse according to the localized scheduling policy set.
[0020] Thirdly, embodiments of this application provide a computer device, the computer device including a memory and a processor connected to the memory; the memory is used to store a computer program, and the processor is used to run the computer program stored in the memory to perform the method described in the first aspect above.
[0021] Fourthly, embodiments of this application provide a storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, implement the method described in the first aspect above.
[0022] This invention provides a method, apparatus, device, and medium for cross-warehouse scheduling migration. The method includes: acquiring multi-dimensional data from a source warehouse and mining the data to obtain a candidate strategy gene set; evaluating the fitness of the candidate strategy gene set based on an evaluation strategy to obtain a validated strategy gene library; acquiring physical parameters and business characteristic parameters of a target warehouse and constructing an environmental fingerprint from the physical parameters and business characteristic parameters to obtain a target environmental fingerprint vector; performing inheritance screening on the target environmental fingerprint vector and the validated strategy gene library using inheritance screening rules to obtain a candidate inheritance gene set; adjusting the candidate inheritance gene set based on an adjustment strategy to obtain a localized scheduling strategy set; and generating and executing migration control instructions in the target warehouse according to the localized scheduling strategy set. Therefore, this embodiment of the invention obtains a validated strategy gene library by mining and evaluating the source repository, constructs an environmental fingerprint of the parameters of the target repository and the validated strategy gene library, and performs inheritance screening to obtain a candidate inheritance gene set. Based on the adjustment strategy, the candidate inheritance gene set is adjusted to obtain a localized scheduling strategy set. In the target repository, migration control instructions are generated and executed according to the localized scheduling strategy set, thereby achieving flexible, quantifiable, and screenable migration, thus improving migration efficiency. Attached Figure Description
[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0024] Figure 1 A flowchart illustrating the migration method for cross-warehouse scheduling provided in an embodiment of the present invention;
[0025] Figure 2 A schematic block diagram of a cross-warehouse scheduling migration device provided in an embodiment of the present invention;
[0026] Figure 3 A schematic block diagram of a computer device provided for an embodiment of the present invention. Detailed Implementation
[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0028] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.
[0029] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.
[0030] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.
[0031] It should be noted that if any AI models, software tools, or components not belonging to the applicant appear in the embodiments of this application, they are merely illustrative examples and do not represent actual use. The user personal information involved in the embodiments of this application is obtained by an entity authorized (knowing and consenting) by the relevant parties or fully authorized by all parties through various legal and compliant means. The collection, storage, use, processing, transmission, provision, and disclosure of the information, data, and signals involved all comply with relevant laws and regulations and do not violate public order and good morals.
[0032] To facilitate understanding of this invention, the technical terms involved in this invention are explained as follows:
[0033] Strategy Gene: The smallest reusable scheduling decision unit encapsulated in a predefined template, containing structured information such as strategy intent, preconditions, execution actions, constraints, and validation metrics;
[0034] Environment Fingerprint Vector: A compact numerical vector obtained by encoding and dimensionality reduction of the physical layout and business characteristics of a warehouse. It is used to uniquely represent and quantify the operating environment characteristics of different warehouses.
[0035] Please see Figure 1 , Figure 1 This is a flowchart illustrating the migration method for cross-warehouse scheduling provided in an embodiment of the present invention. Figure 1 As shown, this invention provides a cross-warehouse scheduling migration method, applicable to a wide range of scenarios including logistics companies, manufacturing companies, and third-party warehousing operators with multiple warehouses. It solves the inefficiency of optimizing a scheduling system from scratch when a new warehouse goes live, enabling cross-warehouse strategy knowledge transfer and rapid reuse. Specifically, in multi-warehouse enterprise scenarios, when a company builds a new warehouse in a new region, it can directly inherit suitable scheduling strategies from the gene pool of mature warehouses, shortening the optimization cycle from several months to several days. In third-party warehousing operator scenarios, operators can establish cross-customer, cross-industry strategy gene pools to quickly customize scheduling solutions for different customers' warehouses. In intelligent warehousing equipment manufacturer scenarios, equipment manufacturers can deliver validated strategy gene pools with their equipment, improving the plug-and-play capability of the equipment and customer satisfaction.
[0036] The method includes the following steps S110-S160.
[0037] S110. Obtain multidimensional collected data from the source warehouse, and mine the multidimensional collected data to obtain a candidate strategy gene set.
[0038] In this embodiment, the source repository refers to a mature repository that has been operating and has accumulated rich scheduling experience. After obtaining the multidimensional collection data from the source repository, the multidimensional collection data can be mined to obtain a candidate strategy gene set.
[0039] In one embodiment, the step of acquiring multidimensional collected data from the source repository and mining the multidimensional collected data to obtain a candidate strategy gene set includes:
[0040] The multidimensional data is obtained by using multiple types of sensors to collect data from the source warehouse.
[0041] The multidimensional collected data is subjected to time axis alignment, outlier cleaning, and format standardization to obtain a standard log sequence;
[0042] The standard log sequence is segmented and extracted using a sliding window to obtain paired sequences;
[0043] Cluster analysis and high-frequency success pattern recognition are performed on the paired sequences to extract the candidate strategy gene set.
[0044] In this embodiment, the multi-type sensors include RFID readers, photoelectric sensors, vision cameras, weighing sensors, speed encoders, and vibration sensors. The multi-type sensors are deployed at key operational nodes of the source warehouse to continuously collect equipment status data, cargo flow data, and scheduling execution data of the source warehouse. That is, the multi-dimensional collected data includes equipment status data, cargo flow data, and scheduling execution data. The RFID reader is a radio frequency identification device used to read electronic tags on goods to obtain their identity information; the photoelectric sensor is a sensor that detects the passage of objects using the principle of beam interruption; the vision camera is an industrial camera used to acquire image information such as the appearance and size of the goods; the weighing sensor is a sensor used to measure the weight of the goods in real time; the vibration sensor is a sensor used to monitor the operating status of the equipment to assess its health; the speed encoder is a rotary encoding device used to measure the operating speed of the conveyor line or AGV (Automated Guided Vehicle); the equipment status data includes, for example, conveyor line speed, AGV position and power, stacker crane load status, etc.; the goods flow data includes goods location, weight, volume, SKU (Stock Keeping Unit) number, flow timestamp, etc.; the scheduling execution data includes scheduling instruction issuance time, actual execution time, execution result feedback, etc.
[0045] The process of time-axis alignment, outlier cleaning, and format standardization of the multi-dimensional collected data yields a standard log sequence. Specifically, time-axis alignment unifies data from different devices and sampling frequencies to a millisecond-level precision clock, ensuring strict temporal synchronization of different dimensions of data for the same event. Outlier cleaning removes noise data caused by sensor drift, communication packet loss, etc., through statistical tests and threshold determination methods, ensuring data quality. Format standardization converts proprietary data formats from different device manufacturers into a unified data model, eliminating format heterogeneity. The process of time-axis alignment, outlier cleaning, and format standardization of the multi-dimensional collected data yields a standard log sequence.
[0046] The standard log sequence is segmented and extracted using a sliding window to obtain paired sequences. Specifically, the continuous standard log sequence is divided into independent segments according to the scheduling period. The scheduling period is, for example, every 5 minutes or every 100 orders per window. For each independent segment, the scheduling decision context and execution performance indicators are extracted to form a decision-performance paired sequence. The scheduling decision context includes the current warehouse congestion status, equipment availability, and characteristics of the pending order queue. The execution performance indicators include throughput, average transit time, number of congestion occurrences, and equipment utilization within the window. Here, "decision-performance paired sequence" refers to binding the scheduling decisions made by the system within a certain period with the actual operational effects of those decisions into a set of records, facilitating subsequent analysis of which decisions produced excellent results under what conditions.
[0047] The process involves clustering and identifying high-frequency success patterns in the paired sequences to extract the candidate strategy gene set. Specifically, clustering analysis is performed on the paired sequences to identify scheduling decision patterns that repeatedly occur and exhibit excellent performance under similar context conditions, and these patterns are extracted as candidate strategy genes. A set of several candidate strategy genes is then established. The clustering analysis can employ either the DBSCAN density clustering algorithm or the K-Means mean clustering algorithm. Each candidate strategy gene represents an independently reusable scheduling decision unit, such as a "narrow aisle multi-AGV collaborative avoidance strategy," a "peak-hour diversion point dynamic adjustment strategy," or a "palletizing order optimization strategy for specific SKU combinations."
[0048] S120. The candidate strategy gene set is evaluated for fitness based on the evaluation strategy to obtain the validated strategy gene library.
[0049] In this embodiment, after obtaining the candidate strategy gene set, the fitness of the candidate strategy gene set can be evaluated based on the evaluation strategy to obtain a validated strategy gene library.
[0050] In one embodiment, the fitness evaluation of the candidate strategy gene set based on the evaluation strategy to obtain a validated strategy gene library includes:
[0051] Each strategy gene in the candidate strategy gene set is structurally encapsulated according to a preset gene template to obtain a structured gene object set;
[0052] A hash calculation is performed on each gene object in the structured gene object set to obtain an identifiable gene object set;
[0053] The set of identifiable gene objects is replayed and tested on the historical data of the source repository to obtain a fitness evaluation vector set;
[0054] A comprehensive fitness score is obtained by performing multi-dimensional weighted scoring based on the fitness evaluation vector set, and the comprehensive fitness score is then filtered according to a preset fitness threshold to obtain the validated strategy gene library.
[0055] In this embodiment, the gene template defines a complete data structure of several strategy gene objects, which is the core carrier for realizing standardized strategy encapsulation. The encapsulated content in the structured gene object set includes a strategy intent description, a set of preconditions, a sequence of execution actions, a set of constraints, and a set of verification indicators. The strategy intent description records the goal of the strategy in natural language or structured label form, such as "reducing congestion at merging points" or "improving nighttime picking efficiency." The set of preconditions defines the prerequisites for the strategy to take effect, such as "merging point queue length > N," "current time period is nighttime," and "number of available AGVs ≥ M," etc., and is expressed in the form of logical expressions. The strategy will only be triggered when the real-time state meets the preconditions. The sequence of execution actions records the actions that the strategy needs to perform in the form of an ordered list. Specific control actions, such as "reducing the speed of segment S3 to 0.8 m / s", "opening the diversion port D2", and "raising the priority of pallet P to the highest", include an abstract device identifier, action type, parameter value, and time constraints for each action sequence. The abstract device identifier is not bound to the device number of a specific warehouse but uses a generic functional identifier for cross-warehouse mapping. The constraint set records the safety boundaries of the strategy execution, such as "the speed of segment S3 must not be lower than 0.5 m / s" and "the duration of opening diversion port D2 does not exceed 600 seconds", ensuring that the strategy will not cause equipment to exceed limits or pose safety hazards. The verification indicator set defines quantitative indicators for evaluating the effectiveness of strategy execution, such as "throughput increase ≥ 5%" and "congestion rate reduction ≥ 10%", providing standards for subsequent fitness assessments.
[0056] The process involves hashing each gene object in the structured gene object set to obtain an identifiable gene object set. This identifiable gene object set is then tested on historical data from the source repository to obtain a fitness evaluation vector set. Specifically, the hash calculation uses the SHA-256 hash algorithm. This algorithm is used to perform a digest calculation on the serialized representation of the gene content of each gene object and generate a unique gene identifier, thus obtaining an identifiable gene object set. This ensures that the same strategy content, packaged in different times and different repositories, obtains the same identifier, facilitating deduplication and version management. Simultaneously, historical data from the source repository is acquired, and the identifiable gene object set is tested on this historical data. The throughput increment, latency reduction rate, congestion avoidance rate, and energy consumption change rate of each gene object are calculated to obtain the fitness evaluation vector set. That is, the fitness evaluation vector set includes throughput increment, latency reduction rate, congestion avoidance rate, and energy consumption change rate.
[0057] The fitness evaluation vector set is used to perform multi-dimensional weighted scoring to obtain a comprehensive fitness score. This comprehensive fitness score is then filtered according to a preset fitness threshold to obtain the validated strategy gene library. Specifically, the fitness evaluation uses a preset multi-dimensional weighted scoring model, with the following formula: Where F(g) represents the overall fitness score of gene g, f i (g) represents the normalized score of the i-th evaluation dimension (such as throughput increment score, latency reduction rate score, etc.), w i This represents the weight coefficient for the corresponding dimension, where n represents the number of evaluation dimensions. Weights can be configured based on business priorities; for example, e-commerce warehousing can be assigned a higher weight to throughput increment, while pharmaceutical warehousing can be assigned a higher weight to accuracy. Only gene objects with a comprehensive fitness score exceeding a preset fitness threshold will be written into the gene library to obtain the validated strategy gene library, ensuring that all strategies stored in the gene library are validated and of high quality.
[0058] S130. Obtain the physical parameters and business characteristic parameters of the target warehouse, and construct an environmental fingerprint from the physical parameters and the business characteristic parameters to obtain the target environmental fingerprint vector.
[0059] In this embodiment, after obtaining the physical parameters and business feature parameters of the target warehouse, an environmental fingerprint can be constructed from the physical parameters and the business feature parameters to obtain the target environmental fingerprint vector.
[0060] In one embodiment, the step of obtaining the physical parameters and business feature parameters of the target warehouse, and constructing an environmental fingerprint from the physical parameters and the business feature parameters to obtain a target environmental fingerprint vector, includes:
[0061] Obtain the physical parameters of the target warehouse, which include at least the warehouse area, floor height, aisle width, rack type and quantity, conveyor line topology and equipment configuration list;
[0062] Obtain the business characteristic parameters of the target warehouse, which include at least the number of product categories, category distribution, average daily order volume, order peak-to-valley ratio, picking mode, and operation cycle requirements;
[0063] The physical parameters are topologically encoded to obtain a physical topological feature vector;
[0064] The business feature parameters are normalized and statistical features are extracted to obtain a business statistical feature vector;
[0065] The physical topology feature vector and the business statistical feature vector are concatenated and dimensionality-reduced to obtain the target environment fingerprint vector.
[0066] In this embodiment, the physical parameters include total warehouse area, floor height, aisle width and number, rack type and its arrangement density, conveyor topology, AGV / AMR (Autonomous Mobile Robot System) quantity and model, stacker crane parameters, etc.; rack type can include different storage structures such as beam racks, drive-in racks, and shuttle automated warehouses; conveyor topology includes the connection relationship and layout of belt conveyors, roller conveyors, sorting machines, and elevators. The business characteristic parameters include SKU category quantity and ABC classification distribution, average daily order volume and fluctuation coefficient, order type ratio, operation time window and cycle time requirements, temperature control requirements, etc.; the category quantity refers to the number of SKU categories; the classification distribution refers to the ABC classification distribution, which is an inventory management method that divides goods into three levels according to sales volume or value: A-class high frequency, B-class medium frequency, and C-class low frequency; the order type ratio is the proportion of order types, such as the ratio of full-case picking to partial-case picking.
[0067] The physical parameters are topologically encoded to obtain physical topological feature vectors. Specifically, the physical parameters are obtained to determine the physical layout of the warehouse. This physical layout is then converted into a node-edge structure graph feature vector, modeling the warehouse physical layout as a directed graph G=(V,E). Nodes V include key locations such as workstations, sorting ports, merging points, and charging stations, while edges E represent physical connections between nodes, such as conveyor lines and AGV channels. Each node is associated with an attribute vector (e.g., workstation type, equipment capacity, area occupancy), and each edge is associated with an attribute vector (e.g., conveyor line type, rated speed, capacity limit, length). Graph neural networks (GNNs) (a type of deep learning model specifically designed for processing graph-structured data, capable of extracting feature representations from graph structures through message passing mechanisms), such as GraphSAGE or GCN (Graph Convolutional Network), are used to encode the graph structure, obtaining the fixed-dimensional physical topological feature vectors.
[0068] The business feature parameters are normalized and statistical features are extracted to obtain a business statistical feature vector. Specifically, the business statistical feature vector is extracted by methods such as min-max normalization (a standardization method that scales the values to the [0,1] interval), quantile statistics, and coefficient of variation calculation of the business feature parameters.
[0069] The target environment fingerprint vector is obtained by concatenating and dimensionality-reducing the physical topology feature vector and the business statistical feature vector. Specifically, the physical topology feature vector and the business statistical feature vector are concatenated, and dimensionality reduction is performed using an autoencoder or PCA (Principal Component Analysis) to obtain a compact target environment fingerprint vector. The dimension is typically set to 64 or 128. In a preferred embodiment, the physical topology feature vector has a dimension of 64, the business statistical feature vector has a dimension of 32, and after concatenation, the dimension is reduced to 64 using an autoencoder as the final target environment fingerprint vector. The number of message passing layers in the graph neural network is set to 3, with a hidden dimension of 128 in each layer, configured according to the actual scenario.
[0070] S140. Use inheritance screening rules to perform inheritance screening on the target environment fingerprint vector and the verified strategy gene library to obtain a candidate inheritance gene set.
[0071] In this embodiment, after determining the target environment fingerprint vector and the verified strategy gene, inheritance screening rules can be used to perform inheritance screening on the target environment fingerprint vector and the verified strategy gene library to obtain a candidate inheritance gene set.
[0072] In one embodiment, the inheritance screening of the target environment fingerprint vector and the validated strategy gene library using inheritance screening rules to obtain a candidate inheritance gene set includes:
[0073] Extract the source environment fingerprint vector of each gene object in the verified strategy gene library to obtain the source environment fingerprint set;
[0074] Calculate the cosine similarity between the target environment fingerprint vector and each vector in the source environment fingerprint set to obtain a similarity ranking list;
[0075] Based on the similarity ranking list, a Top-K screening is performed to obtain the initial gene set;
[0076] Based on the verification conditions, a feasible gene set is obtained by performing a feasibility verification between the precondition set of each gene object in the initial screening gene set and the equipment capability list of the target warehouse.
[0077] A pre-defined model is used to score the correlation between the strategic intent description of each gene in the feasible gene set and the business objective description of the target warehouse to obtain a comprehensive inheritance score. The genes are then sorted and truncated according to the comprehensive inheritance score to obtain the candidate inheritance gene set.
[0078] In this embodiment, a three-level screening mechanism is adopted to ensure that the inherited strategy is highly adapted to the target environment. The screening and matching process includes: First, extracting the source environment fingerprint vector recorded during the encapsulation of each gene object in the gene library, and calculating the cosine similarity between the target environment fingerprint vector and each source environment fingerprint vector to obtain the similarity ranking list; Second, performing Top-K screening (i.e., selecting the top K results with the highest similarity, where the K value can be dynamically configured according to the complexity of the target repository, usually 10~50) to retain the K gene objects with the highest similarity as the initial screening gene set, and performing a precondition feasibility check on the initial screening gene set to obtain the feasible gene set; Third, using the preset model to score the similarity, feasibility, and semantic relevance between the strategy intent description and the target repository business objectives to obtain a comprehensive inheritance score. The preset model mentioned here is a large language model (LLM), which generally refers to a pre-trained language model with natural language understanding and semantic reasoning capabilities. This includes, but is not limited to, models based on the Transformer architecture such as the GPT series, BERT series, LLaMA series, Tongyi Qianwen, and DeepSeek. Industry-specific language models that have been fine-tuned with knowledge from the warehousing and logistics field can also be used.
[0079] The formula for calculating cosine similarity is as follows: ; where v s v represents the source environment fingerprint vector. t This represents the fingerprint vector of the target environment.
[0080] The comprehensive inheritance score is a weighted composite of the similarity score, feasibility score, and semantic relevance score, as shown in the following formula: ;in, represents the overall inheritance score of gene g, and sim is the cosine similarity. The score is the feasibility verification score (1 for passing, 0 for failing). The LLM semantic relevance score (normalized to [0,1]) is used, where α, β, and γ are weight coefficients for each screening level and satisfy α+β+γ=1. In a preferred embodiment, α is set to 0.4, β to 0.3, γ to 0.3, and the inheritance threshold is set to 0.6. When the target warehouse and the source warehouse belong to the same industry type, the weight of α can be appropriately increased to 0.5 to strengthen the influence of environmental similarity. The candidate inheritance gene set is obtained by sorting and truncating according to the comprehensive inheritance score. Specifically, the system sorts the gene objects in descending order according to the comprehensive inheritance score and truncates the gene objects with scores higher than the inheritance threshold to form the candidate inheritance gene set.
[0081] S150. Adjust the candidate inheritance gene set based on the adjustment strategy to obtain a localized scheduling strategy set.
[0082] In this embodiment, after determining the candidate inheritance gene set, the candidate inheritance gene set can be adjusted based on the adjustment strategy to obtain a localized scheduling strategy set.
[0083] In one embodiment, adjusting the candidate inheritance gene set based on the adjustment strategy to obtain a localized scheduling strategy set includes:
[0084] Obtain environmental data of the target warehouse, and construct a digital twin simulation environment based on the environmental data to obtain an instance of the target simulation environment;
[0085] The execution action sequence of each gene object in the candidate inheritance gene set is mapped to the device identifier and address space of the target warehouse to obtain the mapped gene object set;
[0086] The mapped gene object set is sequentially subjected to accelerated playback simulation in the target simulation environment instance to obtain a simulation evaluation result set.
[0087] The parameters of the gene objects whose fitness did not meet the standards in the simulation evaluation results were fine-tuned to obtain the fine-tuned gene object set.
[0088] The fine-tuned gene object set is simulated and verified again, and iterated until the fitness reaches a preset threshold or a preset maximum number of iterations is reached. The verified gene object set is then output as the localized scheduling strategy set.
[0089] In this embodiment, the environmental data of the target warehouse is acquired, and a digital twin simulation environment is constructed based on the environmental data to obtain a target simulation environment instance. Specifically, the environmental data includes warehouse CAD drawings, equipment parameters, and business data. A digital twin simulation environment is constructed based on the warehouse CAD drawings, equipment parameters, and business data of the target warehouse to obtain a target simulation environment instance. This simulation environment can simulate core logistics operation processes such as conveyor line operation, AGV scheduling, and stacker crane operation.
[0090] The execution action sequence of each gene object in the candidate inheritance gene set is mapped to the device identifier and address space of the target warehouse to obtain a mapped gene object set. The mapped gene object set is then sequentially subjected to accelerated playback simulation in the target simulation environment instance to obtain a simulation evaluation result set. Specifically, after mapping the abstract device identifier in the candidate genes to the actual device number and address space of the target warehouse to obtain the mapped gene object set, the mapped gene object set is sequentially subjected to accelerated playback in the target simulation environment instance at a preset acceleration factor to evaluate the actual fitness of each gene in the target environment to obtain the simulation evaluation result set. The acceleration factor is typically 10 to 100 times the actual speed.
[0091] For gene objects whose fitness did not meet the target in the simulation evaluation result set, parameter fine-tuning is performed to obtain a fine-tuned gene object set. Specifically, for gene objects whose fitness did not reach the preset threshold, the system automatically performs parameter fine-tuning, including but not limited to: speed coefficient adjustment, threshold boundary recalibration, and priority weight remapping. Among them, the speed coefficient adjustment is to scale the speed setting of the source warehouse proportionally to the rated speed of the target warehouse conveyor line; the threshold boundary recalibration is to adjust the queue length threshold in the trigger condition according to the capacity ratio of the target warehouse; and the priority weight remapping is to redistribute the order urgency weight according to the business priority of the target warehouse.
[0092] The fine-tuned gene object set is then subjected to simulation verification again, iterating until the fitness reaches a preset threshold or a preset iteration limit is reached. The verified gene object set is then output as the localized scheduling strategy set. Specifically, after fine-tuning the gene object set, simulation verification is performed again, iterating until the fitness reaches the target or the iteration limit is reached. The verified gene object set is then output as the localized scheduling strategy set. The preset iteration limit can be 50 times, and the preset threshold can be 0.7.
[0093] The parameter fine-tuning process can be efficiently searched using the objective function of Bayesian optimization (a global optimization method based on a probabilistic surrogate model), which is as follows: Where θ is the vector of parameters to be fine-tuned. Let g(θ) be the parameter search space, and E be the parameterized gene object. t Let F represent the target simulation environment and F be the fitness evaluation function. Bayesian optimization establishes a surrogate model of the fitness function through Gaussian process regression and uses a sampling function (such as Expected Improvement) to guide the selection of parameter sampling points in the next round, thereby quickly converging to the optimal parameter combination within a finite number of simulations.
[0094] S160. In the target warehouse, a migration control instruction is generated and executed according to the localized scheduling strategy set.
[0095] In this embodiment, after determining the localized scheduling policy set, migration control instructions can be generated and executed in the target repository based on the localized scheduling policy set.
[0096] In one embodiment, generating and executing migration control instructions based on the localized scheduling policy set in the target warehouse includes:
[0097] The activation strategy queue is obtained by matching the trigger conditions of each strategy in the localized scheduling strategy set with the target warehouse.
[0098] The execution action sequence of each policy instance in the activation policy queue is segmented at the instruction granularity and mapped to the device address to generate migration control instructions;
[0099] The migration control command is sent to the corresponding device controller in the target warehouse for execution via a standardized communication interface.
[0100] In this embodiment, the system deploys the verified localized scheduling strategy set to the scheduling engine of the target warehouse. During actual operation, the scheduling engine continuously monitors the real-time status of the warehouse. When it detects that the preconditions of a certain strategy are met, it activates the corresponding strategy instance, parses its execution action sequence into migration control instructions for the PLC controller, and sends them to the corresponding device controller for execution after device address mapping and message encoding. Here, "message encoding" refers to serializing and encoding the abstract migration control instructions according to the format specifications of industrial communication protocols (such as MQTT message queue transmission protocol, OPC UA open platform communication protocol, or Modbus protocol, etc.) to form a data frame that can be directly transmitted in the communication network.
[0101] Furthermore, in the target warehouse, after generating and executing migration control instructions based on the localized scheduling strategy set, the process further includes: collecting device feedback data after execution in the target warehouse, calculating strategy execution effect indicators, and generating feedback evaluation records; updating the fitness scores of executed strategy genes online based on the feedback evaluation records, and encapsulating high-performing new strategy patterns as new gene objects and writing them into the gene bank to achieve continuous evolution of the gene bank.
[0102] In this embodiment, after execution, the system collects device feedback data (such as actual speed, diversion status, job completion time, and abnormal alarm information), calculates the strategy execution effect index, and generates the feedback evaluation record. Based on the feedback evaluation record, the system updates the fitness score of the executed strategy gene online. The online fitness update formula is as follows: Among them, F (t+1) (g) represents the updated fitness score, F t (g) represents the current fitness score, f actual(g) represents the actual performance score of this execution, and η is the learning rate (ranging from (0,1), controlling the impact of new information on historical scores). In practical applications, the recommended value range for η is 0.05 to 0.3, with 0.1 being the typical default value. When the warehouse operating environment changes rapidly, η can be appropriately increased to accelerate adaptation, while η can be decreased to maintain score stability when the environment is stable. Strategies with consistently excellent performance have their selection priority increased; strategies with declining performance have their priority reduced or trigger readjustment. Simultaneously, the system continuously monitors new successful scheduling patterns emerging during the operation of the target warehouse, encapsulates them as new strategy gene objects, and writes them into the gene pool, achieving continuous evolution and expansion of the gene pool.
[0103] In summary, this invention discloses a method for acquiring multidimensional data from a source repository and mining this data to obtain a candidate strategy gene set; evaluating the fitness of the candidate strategy gene set based on an evaluation strategy to obtain a validated strategy gene library; acquiring the physical parameters and business feature parameters of the target repository and constructing an environmental fingerprint from the physical parameters and business feature parameters to obtain a target environmental fingerprint vector; using inheritance filtering rules to perform inheritance filtering on the target environmental fingerprint vector and the validated strategy gene library to obtain a candidate inheritance gene set; adjusting the candidate inheritance gene set based on an adjustment strategy to obtain a localized scheduling strategy set; and generating and executing migration control instructions in the target repository according to the localized scheduling strategy set. As can be seen, the embodiments of the present invention construct a complete closed-loop system of "mining-encapsulation-matching-inheritance-verification-execution-feedback-evolution". Specifically: First, by using standardized gene templates, scattered scheduling experience is encapsulated into structured, independently reusable strategy gene objects, solving the problems of fragmented scheduling knowledge and difficulty in transfer. Second, by using environmental fingerprint vectors, a compact representation of the physical and business characteristics of the warehouse is achieved, providing a computable metric basis for cross-warehouse similarity matching. Third, by using a three-level screening mechanism combining cosine similarity matching, feasibility verification of preconditions, and semantic scoring of LLM (Large Language Model), the inherited strategy is ensured to be highly adapted to the target environment. Fourth, by using digital twin simulation and Bayesian parameter optimization, the strategy is fully verified and automatically fine-tuned before going online, avoiding the security risks of direct migration. Fifth, by using an execution feedback-driven online fitness update and new gene discovery mechanism, the continuous evolution of the gene pool is achieved. Therefore, this invention obtains a validated strategy gene library by mining and evaluating the source repository, constructs an environmental fingerprint of the parameters of the target repository and the validated strategy gene library, and performs inheritance screening to obtain a candidate inheritance gene set. Based on the adjustment strategy, the candidate inheritance gene set is adjusted to obtain a localized scheduling strategy set. In the target repository, migration control instructions are generated and executed according to the localized scheduling strategy set, thereby achieving flexible and screened verification migration and improving migration efficiency.
[0104] Figure 2 This is a schematic block diagram of a cross-warehouse scheduling migration device provided in an embodiment of the present invention. Figure 2 As shown, this embodiment of the invention provides a migration device 700 for cross-warehouse scheduling using the method described above. It addresses a wide range of application scenarios, including logistics companies, manufacturing companies, and third-party warehousing operators with multiple warehouses. This solves the inefficiency of optimizing a scheduling system from scratch when a new warehouse goes live, enabling cross-warehouse strategy knowledge transfer and rapid reuse. Specifically, in multi-warehouse enterprise scenarios, when a company builds a new warehouse in a new region, it can directly inherit suitable scheduling strategies from the gene pool of mature warehouses, shortening the optimization cycle from several months to several days. In third-party warehousing operator scenarios, operators can establish cross-customer, cross-industry strategy gene pools to quickly customize scheduling solutions for different customers' warehouses. In intelligent warehousing equipment manufacturer scenarios, equipment manufacturers can deliver validated strategy gene pools with their equipment, improving the plug-and-play capability of the equipment and customer satisfaction.
[0105] Please see Figure 2 The cross-warehouse scheduling migration device 700 includes:
[0106] Mining unit 701 is used to acquire multidimensional collected data from the source warehouse and mine the multidimensional collected data to obtain a candidate strategy gene set.
[0107] Evaluation unit 702 is used to evaluate the fitness of the candidate strategy gene set based on the evaluation strategy to obtain a validated strategy gene library.
[0108] The construction unit 703 is used to obtain the physical parameters and business feature parameters of the target warehouse, and to construct an environmental fingerprint of the physical parameters and the business feature parameters to obtain a target environmental fingerprint vector.
[0109] The screening unit 704 is used to perform inheritance screening on the target environment fingerprint vector and the verified strategy gene library using inheritance screening rules to obtain a candidate inheritance gene set.
[0110] The adjustment unit 705 is used to adjust the candidate inheritance gene set based on the adjustment strategy to obtain a localized scheduling strategy set;
[0111] The migration unit 706 is used to generate and execute migration control instructions in the target warehouse according to the localized scheduling strategy set.
[0112] It should be noted that those skilled in the art can clearly understand that the specific implementation process of the above-mentioned device can be referred to the corresponding description in the foregoing method embodiments. For the sake of convenience and brevity, it will not be repeated here.
[0113] The above-described device can be implemented as a computer program, and the computer program can be implemented in, for example... Figure 3 It runs on the computer device shown.
[0114] Please see Figure 3 , Figure 3 This is a schematic block diagram of an electronic device provided in an embodiment of the present invention. The electronic device 800 can be a terminal or a server. The terminal can be an electronic device with communication functions. The server can be a standalone server or a server cluster composed of multiple servers.
[0115] See Figure 3 The electronic device 800 includes a processor 802, a memory, and a network interface 805 connected via a system bus 801. The memory may include a non-volatile storage medium 803 and internal memory 804.
[0116] The non-volatile storage medium 803 may store an operating system 8031 and a computer program 8032. The computer program 8032 includes program instructions that, when executed, cause the processor 802 to perform a migration method for cross-warehouse scheduling.
[0117] The processor 802 provides computing and control capabilities to support the operation of the entire electronic device 800.
[0118] The internal memory 804 provides an environment for the execution of the computer program 8032 in the non-volatile storage medium 803. When the computer program 8032 is executed by the processor 802, the processor 802 can perform a migration method for cross-warehouse scheduling.
[0119] This network interface 805 is used for network communication with other devices. Those skilled in the art will understand that... Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the electronic device 800 to which the present invention is applied. The specific electronic device 800 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0120] The processor 802 is used to run a computer program 8032 stored in the memory to perform the following steps:
[0121] Obtain multidimensional collected data from the source warehouse, and mine the multidimensional collected data to obtain a candidate strategy gene set;
[0122] The candidate strategy gene set is evaluated for fitness based on the evaluation strategy to obtain a validated strategy gene library.
[0123] Obtain the physical parameters and business characteristic parameters of the target warehouse, and construct an environmental fingerprint from the physical parameters and the business characteristic parameters to obtain the target environmental fingerprint vector;
[0124] The target environment fingerprint vector and the validated strategy gene library are subjected to inheritance screening using inheritance screening rules to obtain a candidate inheritance gene set.
[0125] The localized scheduling strategy set is obtained by adjusting the candidate inheritance gene set based on the adjustment strategy.
[0126] In the target warehouse, migration control instructions are generated and executed according to the localized scheduling policy set.
[0127] It should be understood that, in this embodiment of the invention, the processor 802 may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.
[0128] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program includes program instructions and can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.
[0129] Therefore, the present invention also provides a storage medium. This storage medium can be a computer-readable storage medium. The storage medium stores a computer program, wherein the computer program includes program instructions. When executed by a processor, the program instructions cause the processor to perform the following steps:
[0130] Obtain multidimensional collected data from the source warehouse, and mine the multidimensional collected data to obtain a candidate strategy gene set;
[0131] The candidate strategy gene set is evaluated for fitness based on the evaluation strategy to obtain a validated strategy gene library.
[0132] Obtain the physical parameters and business characteristic parameters of the target warehouse, and construct an environmental fingerprint from the physical parameters and the business characteristic parameters to obtain the target environmental fingerprint vector;
[0133] The target environment fingerprint vector and the validated strategy gene library are subjected to inheritance screening using inheritance screening rules to obtain a candidate inheritance gene set.
[0134] The localized scheduling strategy set is obtained by adjusting the candidate inheritance gene set based on the adjustment strategy.
[0135] In the target warehouse, migration control instructions are generated and executed according to the localized scheduling policy set.
[0136] The storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.
[0137] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.
[0138] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of each unit is merely a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.
[0139] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the device of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.
[0140] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause an electronic device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.
[0141] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims. The software tools, models, or components appearing in the embodiments of the present invention are merely illustrative examples and do not represent actual use.
Claims
1. A migration method for cross-warehouse scheduling, characterized in that, The method includes: Obtain multidimensional collected data from the source warehouse, and mine the multidimensional collected data to obtain a candidate strategy gene set; The candidate strategy gene set is evaluated for fitness based on the evaluation strategy to obtain a validated strategy gene library. Obtain the physical parameters and business characteristic parameters of the target warehouse, and construct an environmental fingerprint from the physical parameters and the business characteristic parameters to obtain the target environmental fingerprint vector; The target environment fingerprint vector and the validated strategy gene library are subjected to inheritance screening using inheritance screening rules to obtain a candidate inheritance gene set. The localized scheduling strategy set is obtained by adjusting the candidate inheritance gene set based on the adjustment strategy. In the target warehouse, migration control instructions are generated and executed according to the localized scheduling policy set; The inheritance screening process, which uses inheritance screening rules to perform inheritance screening on the target environment fingerprint vector and the validated strategy gene library, yields a candidate inheritance gene set, including: Extract the source environment fingerprint vectors of each gene object in the validated strategy gene library to obtain the source environment fingerprint set; calculate the cosine similarity between the target environment fingerprint vector and each vector in the source environment fingerprint set to obtain a similarity ranking list; perform Top-K screening based on the similarity ranking list to obtain the initial screening gene set; perform feasibility verification on the precondition set of each gene object in the initial screening gene set and the equipment capability list of the target warehouse according to the verification conditions to obtain the feasible gene set; use a preset model to perform a correlation score between the strategy intent description of each gene in the feasible gene set and the business objective description of the target warehouse to obtain a comprehensive inheritance score, and sort and truncate according to the comprehensive inheritance score to obtain the candidate inheritance gene set; The process of adjusting the candidate inheritance gene set based on the adjustment strategy to obtain a localized scheduling strategy set includes: The process involves: acquiring environmental data of the target warehouse; constructing a digital twin simulation environment based on the environmental data to obtain a target simulation environment instance; mapping the execution action sequences of each gene object in the candidate inheritance gene set to the device identifier and address space of the target warehouse to obtain a mapped gene object set; sequentially performing accelerated playback simulations on the target simulation environment instance using the mapped gene object set to obtain a simulation evaluation result set; fine-tuning the parameters of gene objects whose fitness in the simulation evaluation result set does not meet the standard to obtain a fine-tuned gene object set; performing simulation verification on the fine-tuned gene object set again, iterating until the fitness reaches a preset threshold or a preset iteration limit, and outputting the verified gene object set as the localized scheduling strategy set.
2. The method according to claim 1, characterized in that, The process of acquiring multidimensional data from the source repository and mining the multidimensional data to obtain a candidate strategy gene set includes: The multidimensional data is obtained by using multiple types of sensors to collect data from the source warehouse. The multidimensional collected data is subjected to time axis alignment, outlier cleaning, and format standardization to obtain a standard log sequence; The standard log sequence is segmented and extracted using a sliding window to obtain paired sequences; Cluster analysis and high-frequency success pattern recognition are performed on the paired sequences to extract the candidate strategy gene set.
3. The method according to claim 1, characterized in that, The fitness evaluation of the candidate strategy gene set based on the evaluation strategy to obtain a validated strategy gene library includes: Each strategy gene in the candidate strategy gene set is structurally encapsulated according to a preset gene template to obtain a structured gene object set; A hash calculation is performed on each gene object in the structured gene object set to obtain an identifiable gene object set; The set of identifiable gene objects is replayed and tested on the historical data of the source repository to obtain a fitness evaluation vector set; A comprehensive fitness score is obtained by performing multi-dimensional weighted scoring based on the fitness evaluation vector set, and the comprehensive fitness score is then filtered according to a preset fitness threshold to obtain the validated strategy gene library.
4. The method according to claim 1, characterized in that, The process of obtaining the physical parameters and business characteristic parameters of the target warehouse, and constructing an environmental fingerprint from the physical parameters and business characteristic parameters to obtain a target environmental fingerprint vector, includes: Obtain the physical parameters of the target warehouse, which include at least the warehouse area, floor height, aisle width, rack type and quantity, conveyor line topology and equipment configuration list; Obtain the business characteristic parameters of the target warehouse, which include at least the number of product categories, category distribution, average daily order volume, order peak-to-valley ratio, picking mode, and operation cycle requirements; The physical parameters are topologically encoded to obtain a physical topological feature vector; The business feature parameters are normalized and statistical features are extracted to obtain a business statistical feature vector; The physical topology feature vector and the business statistical feature vector are concatenated and dimensionality-reduced to obtain the target environment fingerprint vector.
5. The method according to claim 1, characterized in that, The step of generating and executing migration control instructions based on the localized scheduling policy set in the target warehouse includes: The activation strategy queue is obtained by matching the trigger conditions of each strategy in the localized scheduling strategy set with the target warehouse. The execution action sequence of each policy instance in the activation policy queue is segmented at the instruction granularity and mapped to the device address to generate migration control instructions; The migration control command is sent to the corresponding device controller in the target warehouse for execution via a standardized communication interface.
6. A migration device for cross-warehouse scheduling, characterized in that, The device includes: The mining unit is used to acquire multidimensional collected data from the source warehouse and mine the multidimensional collected data to obtain a candidate strategy gene set. An evaluation unit is used to evaluate the fitness of the candidate strategy gene set based on an evaluation strategy, and obtain a validated strategy gene library. The construction unit is used to obtain the physical parameters and business feature parameters of the target warehouse, and to construct an environmental fingerprint of the physical parameters and the business feature parameters to obtain the target environmental fingerprint vector. A screening unit is used to perform inheritance screening on the target environment fingerprint vector and the verified strategy gene library using inheritance screening rules to obtain a candidate inheritance gene set. An adjustment unit is used to adjust the candidate inheritance gene set based on an adjustment strategy to obtain a localized scheduling strategy set; A migration unit is configured to generate and execute migration control instructions in the target warehouse based on the localized scheduling strategy set. The inheritance screening process, which uses inheritance screening rules to perform inheritance screening on the target environment fingerprint vector and the validated strategy gene library, yields a candidate inheritance gene set, including: Extract the source environment fingerprint vectors of each gene object in the validated strategy gene library to obtain the source environment fingerprint set; calculate the cosine similarity between the target environment fingerprint vector and each vector in the source environment fingerprint set to obtain a similarity ranking list; perform Top-K screening based on the similarity ranking list to obtain the initial screening gene set; perform feasibility verification on the precondition set of each gene object in the initial screening gene set and the equipment capability list of the target warehouse according to the verification conditions to obtain the feasible gene set; use a preset model to perform a correlation score between the strategy intent description of each gene in the feasible gene set and the business objective description of the target warehouse to obtain a comprehensive inheritance score, and sort and truncate according to the comprehensive inheritance score to obtain the candidate inheritance gene set; The process of adjusting the candidate inheritance gene set based on the adjustment strategy to obtain a localized scheduling strategy set includes: The process involves: acquiring environmental data of the target warehouse; constructing a digital twin simulation environment based on the environmental data to obtain a target simulation environment instance; mapping the execution action sequences of each gene object in the candidate inheritance gene set to the device identifier and address space of the target warehouse to obtain a mapped gene object set; sequentially performing accelerated playback simulations on the target simulation environment instance using the mapped gene object set to obtain a simulation evaluation result set; fine-tuning the parameters of gene objects whose fitness in the simulation evaluation result set does not meet the standard to obtain a fine-tuned gene object set; performing simulation verification on the fine-tuned gene object set again, iterating until the fitness reaches a preset threshold or a preset iteration limit, and outputting the verified gene object set as the localized scheduling strategy set.
7. A computer device, characterized in that, The computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method as described in any one of claims 1-5.
8. A storage medium, characterized in that, The storage medium stores a computer program, which includes program instructions that, when executed by a processor, can implement the method as described in any one of claims 1-5.