Visual modeling method and system for dense library based on artificial intelligence

Abstract

The invention provides a dense warehouse visual modeling method and system based on artificial intelligence, and the method comprises the steps: building a five-dimensional feature system, obtaining corresponding cargo feature data in real time according to the five-dimensional feature system, and generating a standardized cargo feature vector; carrying out dynamic clustering on the cargo feature vectors by adopting an improved deep clustering algorithm, automatically generating a twin model adaptive to different cargo types, and embedding logic parameters related to cargo operation rules into the twin model to form a two-dimensional model library; the method comprises the following steps: acquiring cargo constraints, space constraints and equipment constraints of a dense warehouse in real time, allocating weights based on an AHP (Analytical Hierarchy Process), and then determining an optimal model matched with cargoes in a two-dimensional model library by adopting a fusion algorithm; and constructing a three-dimensional twinborn model of the dense warehouse, collecting three-dimensional space coordinates of the goods in real time, mapping the optimal model to the three-dimensional twinborn model in real time according to the three-dimensional space coordinates, and finally realizing accurate dynamic visual display of the real-time condition of the dense warehouse.

Description

Technical Field

[0001] This invention belongs to the field of visualization modeling technology, and specifically relates to a dense library visualization modeling method and system based on artificial intelligence. Background Technology

[0002] In the field of warehouse management, dense storage warehouse is a warehousing model with the core design goal of "maximizing the use of warehouse space and increasing the storage capacity per unit area". Its core logic is to reduce the space waste caused by aisles and gaps in traditional warehousing by optimizing the layout of shelves, storage equipment or operation processes, so as to achieve a higher level of "space density" (i.e. the amount of goods that can be stored per unit area / volume).

[0003] Currently, the visualization modeling of dense warehouses is only based on static model libraries built on fixed cargo types (such as "boxes / pallets"). This cannot adapt to the characteristics of logistics cargo with variable sizes, large density differences, and irregular packaging shapes. Furthermore, when retrieving the corresponding model from the static model library, the model is usually retrieved through a single cargo type label, which easily leads to the problem of the model not matching the actual scenario. Even further, after retrieving the model, existing technologies mostly use a visualization method of static map + cargo icon, which cannot intuitively show the spatial relationship between cargo and the structure and equipment of the dense warehouse, resulting in a poor user experience. Summary of the Invention

[0004] Based on this, the present invention provides a method and system for visual modeling of dense libraries based on artificial intelligence, which aims to accurately and dynamically visualize the real-time status of dense libraries.

[0005] A first aspect of this invention provides a method for visual modeling dense libraries based on artificial intelligence, the method comprising: A five-dimensional feature system is constructed. Based on the five-dimensional feature system, the corresponding cargo feature data is acquired in real time, and a standardized cargo feature vector is generated. The five-dimensional feature system includes physical features, packaging features, logistics features, safety features, and dense warehouse adaptation features. An improved deep clustering algorithm is used to dynamically cluster the cargo feature vectors, automatically generating twin models adapted to different cargo types, and embedding logical parameters about cargo operation rules into the twin models to form a two-dimensional model library. The improved deep clustering algorithm is the DBSCAN algorithm based on dense library structure constraints. The system collects cargo constraints, spatial constraints, and equipment constraints of the dense warehouse in real time, assigns weights based on the AHP (Analytic Hierarchy Process), and then uses a fusion algorithm to determine the optimal model that matches the cargo in a two-dimensional model library. A three-dimensional twin model of the dense warehouse is constructed, the three-dimensional spatial coordinates of the goods are collected in real time, and the optimal model is mapped to the three-dimensional twin model in real time based on the three-dimensional spatial coordinates.

[0006] Furthermore, the physical characteristics include size, weight, and density; the packaging characteristics include material, hardness, and sealing; the logistics characteristics include turnover frequency, storage duration, and loading and unloading requirements; the safety characteristics include fragility, corrosivity, and explosion-proof rating; and the compact warehouse adaptation characteristics include rack load-bearing capacity matching and compatibility with storage and retrieval robotic arms.

[0007] Furthermore, in the step of using an improved deep clustering algorithm to dynamically cluster the cargo feature vectors, automatically generating twin models adapted to different cargo types, and embedding logical parameters about cargo operation rules into the twin models to form a two-dimensional model library, the cargo feature vectors are dynamically clustered using shelf structure parameters and logistics equipment parameters as constraints. The shelf structure parameters include layer height, load-bearing capacity, and aisle width, while the logistics equipment parameters include the robotic arm's grasping range and the AGV's carrying capacity.

[0008] Furthermore, in the step of using an improved deep clustering algorithm to dynamically cluster the cargo feature vectors, automatically generating twin models adapted to different cargo types, and embedding logical parameters about cargo operation rules into the twin models to form a two-dimensional model library, at least logical parameters about cargo storage location rules and access action rules are embedded in the twin models.

[0009] Furthermore, in the step of real-time collection of cargo constraints, spatial constraints, and equipment constraints of the dense warehouse, and weighting based on the AHP (Analytic Hierarchy Process) method, followed by the use of a fusion algorithm to determine the optimal model matching the cargo in the dual-dimensional model library, the cargo constraints include real-time collected cargo feature vectors, the spatial constraints include at least the shelf load-bearing capacity, layer height, and aisle width of the target storage area, and the equipment constraints include at least the gripping force of the robotic arm responsible for storing and retrieving the corresponding cargo and the AGV carrying capacity.

[0010] Furthermore, the step of using a fusion algorithm to determine the optimal model matching the goods in the dual-dimensional model library includes: Based on the cosine similarity algorithm, the similarity between the real-time collected cargo feature vector and the cargo feature vector of each twin model in the dual-dimensional model library is calculated, and candidate models are determined based on the similarity. The optimal model is determined based on the verification results of the storage location rules and access action rules of the candidate models with the corresponding spatial constraints and device constraints.

[0011] Furthermore, the step of constructing a five-dimensional feature system, acquiring corresponding cargo feature data in real time based on the five-dimensional feature system, and generating standardized cargo feature vectors includes: Environmental features are introduced, and the environmental features, physical features, and logistics features are fused together. A feature embedding space is constructed through dynamic weighted comparison learning. Based on the LSTM-GARCH hybrid model and feature vectors in the feature embedding space, combined with the trend of cargo type change and logistics link adjustment, potential drift features are predicted, and standardized cargo feature vectors are generated according to the potential drift features. Based on the potential drift characteristics, and using the health feature sample library as a benchmark, the feature distribution is adjusted by adaptive maximum mean difference so that the drifted features meet the requirements for generating standardized cargo feature vectors. The health feature sample library contains feature samples obtained through manual annotation.

[0012] A second aspect of this invention provides an artificial intelligence-based dense library visualization modeling system for implementing the artificial intelligence-based dense library visualization modeling method described in the first aspect, the system comprising: The first construction module is used to construct a five-dimensional feature system. Based on the five-dimensional feature system, the corresponding cargo feature data is acquired in real time, and a standardized cargo feature vector is generated. The five-dimensional feature system includes physical features, packaging features, logistics features, safety features, and dense warehouse adaptation features. The clustering module is used to dynamically cluster the cargo feature vectors using an improved deep clustering algorithm, automatically generate twin models adapted to different cargo types, and embed logical parameters about cargo operation rules into the twin models to form a two-dimensional model library. The improved deep clustering algorithm is the DBSCAN algorithm based on dense library structure constraints. The model determination module is used to collect the cargo constraints, spatial constraints, and equipment constraints of the dense warehouse in real time, and allocate weights based on the AHP (Analytic Hierarchy Process). Then, a fusion algorithm is used to determine the optimal model that matches the cargo in the two-dimensional model library. The second construction module is used to construct a three-dimensional twin model of the dense warehouse, collect the three-dimensional spatial coordinates of the goods in real time, and map the optimal model to the three-dimensional twin model in real time based on the three-dimensional spatial coordinates.

[0013] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the artificial intelligence-based dense library visualization modeling method provided in the first aspect.

[0014] A fourth aspect of the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the artificial intelligence-based dense library visualization modeling method provided in the first aspect.

[0015] This invention provides an AI-based visualization modeling method and system for dense warehouses. It constructs a five-dimensional feature system, acquires corresponding cargo feature data in real time based on this system, and generates standardized cargo feature vectors. An improved deep clustering algorithm is used to dynamically cluster the cargo feature vectors, automatically generating twin models adapted to different cargo types. Logical parameters regarding cargo operation rules are embedded in these twin models, forming a two-dimensional model library. Cargo constraints, spatial constraints, and equipment constraints of the dense warehouse are collected in real time, and weights are allocated based on the Analytic Hierarchy Process (AHP). A fusion algorithm is then used to determine the optimal model matching the cargo in the two-dimensional model library. A three-dimensional twin model of the dense warehouse is constructed, and the three-dimensional spatial coordinates of the cargo are collected in real time. Based on these coordinates, the optimal model is mapped to the three-dimensional twin model in real time. Specifically, by focusing on the core process of "model library construction - dynamic retrieval - visualization," it overcomes the limitations of the existing technology's "static modeling - manual matching - isolated display." Attached Figure Description

[0016] Figure 1 The flowchart illustrates the implementation of an artificial intelligence-based dense library visualization modeling method provided in Embodiment 1 of the present invention. Figure 2 This is a structural block diagram of an artificial intelligence-based dense library visualization modeling system provided in Embodiment 3 of the present invention; Figure 3 This is a structural block diagram of an electronic device provided in Embodiment 4 of the present invention. Detailed Implementation

[0017] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0018] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0020] Example 1 According to an embodiment of the present invention, an embodiment of a dense library visualization modeling method based on artificial intelligence is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0021] This first embodiment provides an artificial intelligence-based method for dense library visualization modeling, which can be used in electronic devices, such as computers. Please refer to... Figure 1 , Figure 1 The flowchart of an implementation method for a dense library visualization modeling method based on artificial intelligence provided in Embodiment 1 of the present invention is shown, specifically including steps S01 to S04.

[0022] Step S01: Construct a five-dimensional feature system. Based on the five-dimensional feature system, acquire the corresponding cargo feature data in real time and generate a standardized cargo feature vector. The five-dimensional feature system includes physical features, packaging features, logistics features, safety features, and dense warehouse adaptation features.

[0023] In this embodiment of the invention, the traditional binary feature of "size + weight" is broken through, and a five-dimensional feature system is constructed, consisting of physical features, packaging features, logistics features, safety features, and high-density warehouse adaptation features. The physical features include size, weight, and density; the packaging features include material, hardness, and sealing performance; the logistics features include turnover frequency, storage duration, and loading and unloading requirements; the safety features include fragility, corrosivity, and explosion-proof rating; and the high-density warehouse adaptation features include rack load-bearing capacity matching and compatibility with storage and retrieval robotic arms.

[0024] Specifically, cargo feature data can be collected in real time through laser contour scanning, millimeter-wave density detection, and an IoT sensor array to generate standardized cargo feature vectors. For example, [1.2m × 0.8m × 0.5m, 50kg, 0.8g / cm³]. 3 Cardboard hard boxes, monthly turnover 3 times, general cargo, shelf load-bearing capacity B).

[0025] Step S02: The improved deep clustering algorithm is used to dynamically cluster the cargo feature vectors, automatically generating twin models adapted to different cargo types, and embedding logical parameters about cargo operation rules into the twin models to form a two-dimensional model library. The improved deep clustering algorithm is the DBSCAN algorithm based on dense library structure constraints.

[0026] It should be noted that the cargo feature vector is dynamically clustered based on the shelf structure parameters and logistics equipment parameters as constraints. The shelf structure parameters include shelf height, load-bearing capacity, and aisle width, while the logistics equipment parameters include the robotic arm's gripping range and the AGV's (Automated Guided Vehicle) carrying capacity. In addition, logical parameters regarding cargo storage location rules and access operation rules are embedded in the twin model. For example, storage location rules may prioritize lower shelves for high-density goods, and access operation rules may involve low-speed gripping of fragile goods by the robotic arm.

[0027] In this embodiment of the invention, all constraint parameters in the constraint conditions are converted into numerical formats comparable to the cargo feature vector. For example, if the "cargo height" in the cargo feature vector is h_goods, then h_goods≤h_i (i is the target shelf layer). If the "cargo weight" in the cargo feature vector is w_goods, then w_goods≤w_i (shelf load-bearing capacity) and w_goods≤w_max (robotic arm gripping) and w_goods≤w_agv (AGV carrying capacity) must be satisfied. Finally, a constraint threshold matrix is ​​formed, which serves as a pre-screening condition for subsequent clustering (only cargo feature vectors that meet the constraints can enter the clustering stage).

[0028] Based on multimodal sensing devices (laser scanners, weight sensors, material recognition cameras, etc.) in intelligent dense warehouses, cargo feature data is collected and an initial feature vector is constructed. Then, outlier removal and numerical standardization are performed. The standardized feature vectors are compared one by one with the constraint threshold matrix, and only vectors that satisfy all constraints are retained. Furthermore, the traditional DBSCAN relies on two core parameters: neighborhood radius ε and minimum number of points MinPts. The improved version associates these parameters with dense warehouse constraints and dynamically calculates their values. Specifically, the neighborhood radius ε, i.e., the feature similarity threshold, is determined based on the difference in shelf load capacity and shelf height. Understandably, if the load capacity difference in a shelf area is ≤50kg and the shelf height difference is ≤0.3m, then ε is set to 0.2 (representing feature similarity ≥80% in the standardized feature space). If the hardware parameter differences in the area are large (e.g., a load capacity difference of 100kg), then ε is reduced to 0.1 (similarity ≥90%) to ensure that goods within the same cluster are compatible with the same type of hardware. As for the minimum number of points MinPts, i.e., the minimum number of samples to form a cluster, it is determined based on the number of shelf shelves. Understandably, if an area has 20 shelves, then MinPts is set to 5 (ensuring that each cluster corresponds to at least 5 shelves for adaptation), avoiding redundancy in the model template due to insufficient cluster samples.

[0029] Furthermore, the traditional DBSCAN only determines whether a sample belongs to the same cluster based on feature distance. The improved version adds hardware constraint verification at each step of cluster merging. Specifically, it calculates the feature distance (such as Euclidean distance) between sample point p and its neighboring sample point q. If the distance is ≤ ε, it is initially determined to be potentially of the same cluster. A second verification is performed to check whether the features of p and q are compatible with the same set of hardware parameters. For example, p's weight of 300kg is compatible with the first shelf of area A (load capacity 500kg), and q's weight of 200kg is compatible with the second shelf of area A (load capacity 300kg). Although the feature distance is ≤ ε, they are determined to be of different clusters because they are compatible with different shelf levels, and their clusters are not merged. If p and q meet both the feature distance requirement and are compatible with the same set of hardware parameters, they are merged into the same cluster, gradually forming hardware-compatible clusters.

[0030] It should be noted that after clustering is completed, the validity of the results is evaluated using two indicators. First, the percentage of samples in each cluster that are compatible with the same hardware parameters is counted, and it must be ≥95% (if 30% of the samples in a cluster are compatible with shelves in area A and 70% are compatible with shelves in area B, then it is considered an invalid cluster) to evaluate the hardware compatibility rate. Second, the number of small clusters with a sample size < MinPts is counted. If the percentage is >10%, the ε and MinPts parameters are readjusted and the clusters are re-clustered to evaluate the cluster redundancy. Finally, a set of clusters that are "effectively compatible with hardware and have no redundancy" is output, with each cluster corresponding to a type of goods that are compatible with the hardware.

[0031] In this embodiment of the invention, a dual-dimensional model library is constructed from both geometric and logical dimensions. For each cluster, the average physical feature of all goods feature vectors within the cluster is extracted as the basic parameter of the 3D model. Based on the hardware parameters adapted to the cluster, logical rule scripts are automatically generated and embedded into the 3D model to supplement the logical dimension. Specifically, for storage location rules, rules are generated based on the shelf layers adapted to the weight and height of the goods (e.g., "If the weight of goods in cluster C1 is ≤300kg and the height is ≤1.5m, then the 2nd to 3rd shelf layers in area A will be allocated first"). For storage and retrieval action rules... Then: Based on the material hardness and fragility of the goods, the equipment action parameters are generated (e.g., "If the fragility of goods in cluster C2 is 1, then the robotic arm gripping speed is ≤0.2m / s and the gripping force is ≤50N; the AGV transport speed is ≤0.5m / s"). For rule conflicts, if a certain goods simultaneously meet multiple storage rules (e.g., it is suitable for both layer 2 of area A and layer 3 of area B), then the priority rule is embedded (e.g., "If the turnover frequency is >5 times / month, then layer 2 of area A, which is closer to the outbound port, is given priority"). Finally, each cluster corresponds to one logical rule package, which is bound to the 3D geometric model to form a twin model.

[0032] Step S03: Real-time collection of cargo constraints, spatial constraints, and equipment constraints of the dense warehouse, and weight allocation based on the AHP (Analytic Hierarchy Process). Then, a fusion algorithm is used to determine the optimal model matching the cargo in the dual-dimensional model library.

[0033] The system collects cargo constraints, spatial constraints, and equipment constraints in real time through sensors deployed within the dense warehouse. The cargo constraints include real-time collected cargo feature vectors. The spatial constraints include at least the shelf load-bearing capacity, layer height, and aisle width of the target storage area. The equipment constraints include at least the gripping force of the robotic arm responsible for storing and retrieving the corresponding goods and the carrying capacity of the AGV. The weights can be adjusted according to business needs.

[0034] Specifically, based on the cosine similarity algorithm, the similarity between the real-time collected cargo feature vector and the cargo feature vector of each twin model in the dual-dimensional model library is calculated, and candidate models are determined based on the similarity. Based on the verification results of the storage location rules and access action rules of the candidate models with the corresponding spatial constraints and device constraints, the optimal model is determined. It can be understood that a rule matching algorithm is used to perform a secondary screening of the candidate models, that is, to verify whether the "storage location rules" of the candidate models match the target area spatial constraints and whether the "access action rules" match the device constraints, and to eliminate models that do not meet the constraints.

[0035] Step S04: Construct a three-dimensional twin model of the dense warehouse, collect the three-dimensional spatial coordinates of the goods in real time, and map the optimal model to the three-dimensional twin model in real time based on the three-dimensional spatial coordinates.

[0036] Specifically, a full-scene 3D twin model of the dense warehouse can be constructed based on laser SLAM+BIM technology, including all physical elements such as shelves, robotic arms, AGVs, aisles, and entrances / exits, and is fully aligned with the spatial coordinate system of the actual dense warehouse (error ≤1cm), providing a high-precision scene carrier for visualization. At the same time, the 3D spatial coordinates (X / Y / Z axes) of the goods are collected in real time through UWB ultra-wideband positioning technology. Finally, based on the 3D spatial coordinates, the optimal model is mapped to the 3D twin model in real time for visualization.

[0037] In summary, the AI-based dense warehouse visualization modeling method in the above embodiments of the present invention constructs a five-dimensional feature system, acquires corresponding cargo feature data in real time based on the five-dimensional feature system, and generates standardized cargo feature vectors; it uses an improved deep clustering algorithm to dynamically cluster the cargo feature vectors, automatically generates twin models adapted to different cargo types, and embeds logical parameters about cargo operation rules into the twin models to form a two-dimensional model library; it collects cargo constraints, spatial constraints, and equipment constraints of the dense warehouse in real time, and assigns weights based on the AHP hierarchical analysis method, and then uses a fusion algorithm to determine the optimal model matching the cargo in the two-dimensional model library; it constructs a three-dimensional twin model of the dense warehouse, collects the three-dimensional spatial coordinates of the cargo in real time, and maps the optimal model to the three-dimensional twin model in real time based on the three-dimensional spatial coordinates. Specifically, by focusing on the core process of "model library construction - dynamic retrieval - visualization", it breaks through the limitations of "static modeling - manual matching - isolated display" in the prior art.

[0038] Example 2 Embodiment 2 of this invention also provides a method for visual modeling of dense warehouses based on artificial intelligence. The difference from Embodiment 1 lies in addressing feature drift in dynamic scenarios, providing high-quality feature input for subsequent clustering. It should be noted that features are the foundation of mapping and clustering. In dynamic scenarios (such as changes in cargo type, environmental temperature and humidity, and handling frequency in an intelligent dense warehouse), feature drift is prone to occur (e.g., shifts in cargo weight feature distribution, additions to packaging material feature dimensions). Therefore, the steps of constructing a five-dimensional feature system, acquiring corresponding cargo feature data in real time based on the five-dimensional feature system, and generating standardized cargo feature vectors include: Step 1: Introduce environmental features and fuse the environmental features, physical features, and logistics features. Construct a feature embedding space through dynamic weighted comparison learning.

[0039] The environmental characteristics include the temperature and humidity of the storage area (collected in real time by temperature and humidity sensors deployed in different locations within the warehouse, with a sampling frequency of once per minute) and vibration frequency (vibration sensors installed in transport channels and next to shelves to record the vibration of goods during handling and storage, measured in Hertz). After cleaning and standardizing the environmental, physical, and logistical characteristics, the initial weights of stable characteristics are determined through historical data statistical analysis combined with domain expert experience. For volatile characteristics, the weight adjustment coefficients are determined based on the magnitude of the fluctuation coefficients, and the initial weights of volatile characteristics are also determined. It is understandable that the priority of volatile characteristics, such as temporary labeling, will fluctuate with changes in business needs and logistics scenarios, and dynamic weights need to be calculated based on the fluctuation coefficients of the most recent three logistics cycles.

[0040] Furthermore, a subset of samples is randomly selected from the cleaned and standardized feature data as anchor samples. For each anchor sample, positive and negative samples are then selected. Positive samples are those belonging to the same cargo type as the anchor sample and having high feature similarity (samples with a similarity greater than 0.8 are selected by calculating the cosine similarity of feature vectors). Negative samples are those belonging to different cargo types as the anchor sample and having low feature similarity (samples with a similarity less than 0.3 are selected). The feature data of the anchor samples, positive samples, and negative samples are input into a neural network model, and the model calculates the feature embedding vectors for each sample. Based on the previously determined initial weights, the contributions of different features in the embedding process are weighted to construct a dynamic weighted contrastive loss function. The goal of this loss function is to minimize the distance between the feature embedding vectors of the anchor sample and the positive sample, and maximize the distance between the anchor sample and the negative sample. The parameters of the neural network model are continuously adjusted using the backpropagation algorithm to minimize the dynamic weighted contrastive loss function until the model converges, completing the construction of the initial feature embedding space.

[0041] Step 2: Based on the LSTM-GARCH hybrid model and feature vectors in the feature embedding space, combined with the trend of cargo type changes and logistics link adjustments, predict potential drift features, and generate standardized cargo feature vectors based on the potential drift features.

[0042] This approach allows for prediction before feature drift occurs, providing ample time for subsequent corrective measures. It's important to note that data on changes in cargo type trends can be collected through a warehouse management system that tracks daily inbound, outbound, and inventory quantities of different cargo types, calculating the changing proportions of each type within the total cargo volume. For example, before major e-commerce promotions, the percentage of inbound small consumer goods (such as snacks and daily necessities) can be analyzed to determine if an upward trend is emerging. Simultaneously, combining this with market demand forecast data (such as sales forecast reports from e-commerce platforms) reveals potential trends in cargo type changes over a future period. Logistics link adjustment data can focus on logistics companies' route planning, changes in transportation modes (such as switching from road to rail), and the addition or removal of transit stations. Through data integration with logistics partners, the specific details and implementation time of logistics link adjustments can be obtained in real time, such as the specific routes for newly added short-distance transshipment links, the locations of transshipment stations, and their operating hours.

[0043] In other embodiments of the present invention, other related factor data that may affect feature drift, such as seasonal variation data and policy and regulation adjustment data, may also be included. Relevant information is obtained through weather forecast data from meteorological departments, policy documents issued by government departments, and other channels, and is converted into quantifiable feature data (such as converting seasonal variation into quarterly codes, with spring as 1, summer as 2, autumn as 3, and winter as 4).

[0044] It should be noted that, firstly, a health characteristic sample library is constructed, selecting samples from historical characteristic data that accurately reflect the normal characteristic state of goods as initial health samples. The selection criteria include: the goods corresponding to the sample have not experienced any abnormalities during storage and transportation (such as damage, loss, delays, etc.); the sample's characteristic data are complete and accurate, with no outliers after preprocessing in step one; and the sample can represent the typical characteristics of different types and batches of goods. The sample set of the initial health characteristic sample library is determined through manual annotation and selection, or through a combination of manual annotation and model selection (such as using clustering algorithms to group samples with similar characteristics into one class and selecting the central sample from each class). Furthermore, to ensure that the health characteristic sample library reflects the current normal state of goods characteristics, the sample library is updated regularly.

[0045] In this embodiment of the invention, the trend data of changes in cargo types and the data of adjustments to logistics links are fused with the feature data of the corresponding period in the health feature sample library to construct the input feature vector of the LSTM-GARCH hybrid model. The dimension of the input feature vector is determined according to the number of related factors. For example, if there are n related factors such as changes in the proportion of cargo types and the number of times logistics links are adjusted, then the dimension of the input feature vector is n. At the same time, features that have drifted in the historical feature data are used as the label data of the model (1 indicates that drift has occurred, and 0 indicates that no drift has occurred).

[0046] Furthermore, the prepared input feature vectors and label data are divided into a training set (70% of the total data), a validation set (20% of the total data), and a test set (10% of the total data). The LSTM-GARCH hybrid model is trained using the training set. The LSTM part is used to capture the time-series features of the associated factor data and predict the changing trends of cargo features; the GARCH part is used to handle heteroscedasticity in the feature data and improve the model's prediction accuracy for feature fluctuations. During training, the model's hyperparameters (such as the number of hidden layer nodes in the LSTM, the learning rate, and the order of the GARCH model) are continuously adjusted using the validation set. AIC (Akaike Information Criterion) and BIC (Bayesian Information Criterion) are used as evaluation metrics for model selection, and the model with the smallest AIC and BIC values ​​is selected as the optimal model.

[0047] Furthermore, the current correlation factor data is input into the trained LSTM-GARCH hybrid model, and the model outputs the probability of each feature drifting within a preset time period. A drift probability threshold is set. If the predicted drift probability of a feature is greater than the drift probability threshold, the feature is determined to be a potential drift feature, and feature drift warning information is generated, including the name of the potential drift feature, the predicted drift time, and the drift probability.

[0048] Step 3: Based on the potential drift features, and using the health feature sample library as a benchmark, adjust the feature distribution through adaptive maximum mean difference so that the drifted features meet the requirements for generating standardized cargo feature vectors. The health feature sample library contains feature samples obtained through manual annotation.

[0049] It should be noted that for features that are confirmed to have drifted, the adaptive maximum mean difference (AMMD) between the real-time data distribution of that feature and the distribution of that feature in the health feature sample library is calculated. AMMD can adaptively adjust the parameters of the kernel function to better capture the differences between different distributions. Its calculation formula is based on the expected difference of the kernel function, which is obtained by calculating the norm of the difference between the means of the two distributions in the regenerating kernel Hilbert space (RKHS).

[0050] The parameters for feature normalization dynamic calibration are determined based on the calculated AMMD value. A larger AMMD value indicates a greater difference between the real-time feature distribution and the healthy feature distribution, requiring a larger adjustment. Finally, the determined calibration coefficients are used to perform normalization dynamic calibration on the drifted feature real-time data. Through calibration, the distribution of the drifted feature data is made as close as possible to the feature distribution in the healthy feature sample library.

[0051] In other embodiments of the present invention, the KL divergence and JS divergence of the calibrated feature data with the corresponding features in the health feature sample library are recalculated to determine whether the calibrated feature distribution meets the requirements. If the requirements are met, it indicates that the calibration effect is good and the feature drift problem is solved. If the requirements are not met, adaptive maximum mean difference adjustment continues, the calibration coefficients are readjusted, and recalibration is performed until the feature distribution meets the requirements. The calibrated feature data is added as a new sample to the health feature sample library, and the updated sample library is used to retrain and optimize the LSTM-GARCH hybrid model, update the parameters of the LSTM-GARCH hybrid model, and improve the accuracy of the LSTM-GARCH hybrid model in predicting future feature drift. At the same time, according to the type and cause of this feature drift, the weights of the correlation factors input to the LSTM-GARCH hybrid model are adjusted so that the LSTM-GARCH hybrid model can better capture the key factors affecting feature drift.

[0052] Example 3 Please see Figure 2 , Figure 2 This is a structural block diagram of an AI-based dense library visualization modeling system 200 provided in Embodiment 3 of the present invention. This AI-based dense library visualization modeling system 200 is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0053] Specifically, the AI-based dense library visualization modeling system 200 includes: a first construction module 21, a clustering module 22, a model determination module 23, and a second construction module 24, wherein: The first construction module 21 is used to construct a five-dimensional feature system. Based on the five-dimensional feature system, the corresponding cargo feature data is acquired in real time, and a standardized cargo feature vector is generated. The five-dimensional feature system includes physical features, packaging features, logistics features, safety features, and high-density warehouse adaptation features. The physical features include size, weight, and density. The packaging features include material, hardness, and sealing performance. The logistics features include turnover frequency, storage duration, and loading and unloading requirements. The safety features include fragility, corrosivity, and explosion-proof rating. The high-density warehouse adaptation features include shelf load-bearing capacity matching and compatibility with storage and retrieval robotic arms. Clustering module 22 is used to dynamically cluster the cargo feature vectors using an improved deep clustering algorithm, automatically generate twin models adapted to different cargo types, and embed logical parameters about cargo operation rules into the twin models to form a two-dimensional model library. The improved deep clustering algorithm is the DBSCAN algorithm based on dense warehouse structure constraints, which uses shelf structure parameters and logistics equipment parameters as constraints to dynamically cluster the cargo feature vectors. The shelf structure parameters include layer height, load-bearing capacity, and aisle width, and the logistics equipment parameters include robotic arm gripping range and AGV carrying capacity. In addition, logical parameters about cargo storage location rules and access action rules are embedded in the twin models. The model determination module 23 is used to collect the cargo constraints, spatial constraints, and equipment constraints of the dense warehouse in real time, and allocate weights based on the AHP analytic hierarchy process. Then, a fusion algorithm is used to determine the optimal model that matches the cargo in the dual-dimensional model library. The cargo constraints include the cargo feature vectors collected in real time. The spatial constraints include at least the shelf load-bearing capacity, layer height, and aisle width of the target storage area. The equipment constraints include at least the gripping force of the robotic arm responsible for storing and retrieving the corresponding cargo and the AGV carrying capacity. The second construction module 24 is used to construct a three-dimensional twin model of the dense warehouse, collect the three-dimensional spatial coordinates of the goods in real time, and map the optimal model to the three-dimensional twin model in real time according to the three-dimensional spatial coordinates.

[0054] Furthermore, in some optional embodiments of the present invention, the model determination module 23 includes: The calculation unit is used to calculate the similarity between the real-time collected cargo feature vector and the cargo feature vector of each twin model in the dual-dimensional model library according to the cosine similarity algorithm, and to determine the candidate model based on the similarity. The determining unit is used to determine the optimal model based on the verification results of the storage location rules and access action rules of the candidate models with the corresponding spatial constraints and device constraints, respectively.

[0055] Furthermore, in some optional embodiments of the present invention, the first building module 21 includes: A construction unit is used to introduce environmental features and fuse the environmental features, physical features, and logistics features, and construct a feature embedding space through dynamic weighted comparison learning; The prediction unit is used to predict potential drift features based on the LSTM-GARCH hybrid model and feature vectors in the feature embedding space, combined with the trend of cargo type change and logistics link adjustment, and generate standardized cargo feature vectors based on the potential drift features. The adjustment unit is used to adjust the feature distribution based on the potential drift features and the health feature sample library, by adaptive maximum mean difference, so that the drifted features meet the requirements for generating standardized cargo feature vectors. The health feature sample library contains feature samples obtained through manual annotation.

[0056] Example 4 In another aspect, the present invention also proposes an electronic device, please refer to [link to relevant documentation]. Figure 3 The image shows an electronic device according to Embodiment 4 of the present invention, including a memory 20, a processor 10, and a computer program 30 stored in the memory and executable on the processor. When the processor 10 executes the computer program 30, it implements the artificial intelligence-based dense library visualization modeling method described above.

[0057] In some embodiments, the processor 10 may be a central processing unit (CPU), controller, microcontroller, microprocessor or other data processing chip, used to run program code stored in memory 20 or process data, such as executing access restriction programs.

[0058] The memory 20 includes at least one type of readable storage medium, such as flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 20 can be an internal storage unit of an electronic device, such as the hard disk of the electronic device. In other embodiments, the memory 20 can also be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. Furthermore, the memory 20 can include both internal and external storage units of the electronic device. The memory 20 can be used not only to store application software and various types of data of the electronic device, but also to temporarily store data that has been output or will be output.

[0059] It should be pointed out that, Figure 3 The structure shown does not constitute a limitation on the electronic device. In other embodiments, the electronic device may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0060] This invention also proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the artificial intelligence-based dense library visualization modeling method described above.

[0061] Those skilled in the art will understand that the logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0062] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0063] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0064] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0065] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A dense library visualization modeling method based on artificial intelligence, characterized in that, The method includes: A five-dimensional feature system is constructed. Based on the five-dimensional feature system, the corresponding cargo feature data is acquired in real time, and a standardized cargo feature vector is generated. The five-dimensional feature system includes physical features, packaging features, logistics features, safety features, and dense warehouse adaptation features. An improved deep clustering algorithm is used to dynamically cluster the cargo feature vectors, automatically generating twin models adapted to different cargo types, and embedding logical parameters about cargo operation rules into the twin models to form a two-dimensional model library. The improved deep clustering algorithm is the DBSCAN algorithm based on dense library structure constraints. The system collects cargo constraints, spatial constraints, and equipment constraints of the dense warehouse in real time, assigns weights based on the AHP (Analytic Hierarchy Process), and then uses a fusion algorithm to determine the optimal model that matches the cargo in a two-dimensional model library. A three-dimensional twin model of the dense warehouse is constructed, the three-dimensional spatial coordinates of the goods are collected in real time, and the optimal model is mapped to the three-dimensional twin model in real time based on the three-dimensional spatial coordinates.

2. The method for dense library visualization modeling based on artificial intelligence according to claim 1, characterized in that, The physical characteristics include size, weight, and density; the packaging characteristics include material, hardness, and sealing; the logistics characteristics include turnover frequency, storage duration, and loading and unloading requirements; the safety characteristics include fragility, corrosivity, and explosion-proof rating; and the compact warehouse adaptation characteristics include rack load-bearing capacity matching and compatibility with storage and retrieval robotic arms.

3. The dense library visualization modeling method based on artificial intelligence according to claim 2, characterized in that, In the step of using an improved deep clustering algorithm to dynamically cluster the cargo feature vectors, automatically generating twin models adapted to different cargo types, and embedding logical parameters about cargo operation rules into the twin models to form a two-dimensional model library, the cargo feature vectors are dynamically clustered using shelf structure parameters and logistics equipment parameters as constraints. The shelf structure parameters include layer height, load-bearing capacity, and aisle width, while the logistics equipment parameters include the robotic arm's grasping range and the AGV's carrying capacity.

4. The dense library visualization modeling method based on artificial intelligence according to claim 3, characterized in that, In the step of using an improved deep clustering algorithm to dynamically cluster the cargo feature vectors, automatically generating twin models adapted to different cargo types, and embedding logical parameters about cargo operation rules into the twin models to form a two-dimensional model library, at least logical parameters about cargo storage location rules and access action rules are embedded in the twin models.

5. The dense library visualization modeling method based on artificial intelligence according to claim 4, characterized in that, In the step of real-time acquisition of cargo constraints, spatial constraints, and equipment constraints in the dense warehouse, and weighting based on the AHP (Analytic Hierarchy Process) method, followed by the use of a fusion algorithm to determine the optimal model matching the cargo in a two-dimensional model library, the cargo constraints include real-time acquired cargo feature vectors, the spatial constraints include at least the shelf load-bearing capacity, layer height, and aisle width of the target storage area, and the equipment constraints include at least the gripping force of the robotic arm responsible for storing and retrieving the corresponding cargo and the AGV carrying capacity.

6. The method for dense library visualization modeling based on artificial intelligence according to claim 5, characterized in that, The step of using a fusion algorithm to determine the optimal model matching the goods in a two-dimensional model library includes: Based on the cosine similarity algorithm, the similarity between the real-time collected cargo feature vector and the cargo feature vector of each twin model in the dual-dimensional model library is calculated, and candidate models are determined based on the similarity. The optimal model is determined based on the verification results of the storage location rules and access action rules of the candidate models with the corresponding spatial constraints and device constraints.

7. The method for dense library visualization modeling based on artificial intelligence according to claim 1, characterized in that, The steps of constructing a five-dimensional feature system, acquiring corresponding cargo feature data in real time based on the five-dimensional feature system, and generating standardized cargo feature vectors include: Environmental features are introduced, and the environmental features, physical features, and logistics features are fused together. A feature embedding space is constructed through dynamic weighted comparison learning. Based on the LSTM-GARCH hybrid model and feature vectors in the feature embedding space, combined with the trend of cargo type change and logistics link adjustment, potential drift features are predicted, and standardized cargo feature vectors are generated according to the potential drift features. Based on the potential drift characteristics, and using the health feature sample library as a benchmark, the feature distribution is adjusted by adaptive maximum mean difference so that the drifted features meet the requirements for generating standardized cargo feature vectors. The health feature sample library contains feature samples obtained through manual annotation.

8. A dense library visualization modeling system based on artificial intelligence, characterized in that, The system is used to implement the AI-based dense library visualization modeling method as described in any one of claims 1-7, the system comprising: The first construction module is used to construct a five-dimensional feature system. Based on the five-dimensional feature system, the corresponding cargo feature data is acquired in real time, and a standardized cargo feature vector is generated. The five-dimensional feature system includes physical features, packaging features, logistics features, safety features, and dense warehouse adaptation features. The clustering module is used to dynamically cluster the cargo feature vectors using an improved deep clustering algorithm, automatically generate twin models adapted to different cargo types, and embed logical parameters about cargo operation rules into the twin models to form a two-dimensional model library. The improved deep clustering algorithm is the DBSCAN algorithm based on dense library structure constraints. The model determination module is used to collect the cargo constraints, spatial constraints, and equipment constraints of the dense warehouse in real time, and allocate weights based on the AHP (Analytic Hierarchy Process). Then, a fusion algorithm is used to determine the optimal model that matches the cargo in the two-dimensional model library. The second construction module is used to construct a three-dimensional twin model of the dense warehouse, collect the three-dimensional spatial coordinates of the goods in real time, and map the optimal model to the three-dimensional twin model in real time based on the three-dimensional spatial coordinates.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the AI-based dense library visualization modeling method as described in any one of claims 1-7.

10. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the AI-based dense library visualization modeling method as described in any one of claims 1-7.

Images