An unmanned intelligent supermarket system and its implementation method

Abstract

The present invention discloses an unmanned intelligent supermarket system and an implementation method thereof, belonging to the field of unmanned retail technology, comprising: dividing a supermarket area into N storage areas and constructing a grid node diagram; utilizing shelf sensors to collect the types and remaining quantities of goods at each grid node in real time, and determining the seasonal status of the goods in combination with a seasonal goods database; then using the remaining quantity intervals of the goods as nodes, constructing a storage map based on the correlation of the goods, and forming a commodity association network through partition connection; constructing a commodity association sales model based on the commodity association network, formulating a commodity migration strategy based on the seasonal status and remaining quantity, and then establishing a differential replenishment model based on the migration results and the real-time remaining quantity to realize dynamic replenishment; collecting equipment operation data through sensors, and using machine learning algorithms to establish a fault prediction model to timely warn equipment with fault risks; thus, intelligent management of goods in unmanned retail stores and efficient operation and maintenance of equipment are realized, thereby improving operational efficiency.

Description

Technical Field

[0001] The present invention belongs to the field of unmanned retail technology, and specifically relates to an unmanned intelligent supermarket system and an implementation method thereof. Background Art

[0002] With the continuous development of technology, unmanned retail stores have gradually emerged, providing consumers with a more convenient shopping experience. However, in the operation of unmanned retail stores, product replenishment and fault warning are crucial links. Traditional retail stores' replenishment and product display mainly rely on manual experience, which is inefficient and difficult to dynamically respond to sales changes. Although unmanned retail stores have reduced labor costs through automated equipment, they still lack intelligent solutions for dynamic product allocation, seasonal demand adaptation, and fault warning. In existing technologies, product replenishment strategies are usually based on fixed rules or simple historical product remaining quantity statistics. They cannot be optimized and adjusted based on real-time sales data, regional heat distribution, and product relevance. This leads to the accumulation of slow-selling products or the shortage of popular products. In addition, the passive handling mode of equipment failure also affects operational efficiency. Summary of the Invention

[0003] In response to the shortcomings of the existing technology, the present invention proposes an unmanned intelligent supermarket system and its implementation method, which divides the supermarket area into N storage areas and constructs a grid node diagram; uses shelf sensors to collect the types and remaining quantities of goods at each grid node in real time, and determines the seasonal status of the goods in combination with the seasonal goods database; then uses the remaining quantity intervals of the goods as nodes, constructs a storage map based on the correlation of the goods, and forms a commodity association network through partition connection; constructs a commodity association sales model based on the commodity association network, formulates a commodity migration strategy based on the seasonal status and remaining quantity, and then establishes a differential replenishment model based on the migration results and the real-time remaining quantity to achieve dynamic replenishment; collects equipment operation data through sensors, and uses machine learning algorithms to establish a fault prediction model to timely warn equipment with failure risks; realizes intelligent management of goods in unmanned retail stores and efficient operation and maintenance of equipment, thereby improving operational efficiency.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A method for implementing an unmanned intelligent supermarket system includes: executing an unmanned retail store intelligent replenishment and fault warning method to realize the unmanned intelligent supermarket system's unmanned function, including:

[0006] The supermarket area of ​​the unmanned retail store is divided into storage areas. A grid node diagram is constructed for each product storage area. Sensors are used to obtain real-time information on the corresponding product type and remaining quantity, as well as the seasonal status of the products at each grid node. The remaining quantity intervals of the products are used as nodes. Combined with the product association, a product storage map is constructed, and the product association network is formed by connecting the partitions.

[0007] Based on the product association network, a product association sales model is constructed, and a product migration strategy is formulated based on the seasonal status and the remaining quantity of the product. At the same time, a differentiated replenishment model is established based on the migration results and the real-time remaining quantity of the product to determine the replenishment demand of each product and formulate a dynamic replenishment strategy;

[0008] By deploying sensors on the equipment to collect equipment operation data in real time, and using machine learning algorithms to establish an equipment failure prediction model, early warning information will be issued in a timely manner when the risk of equipment failure is predicted.

[0009] Specifically, the specific steps of constructing a grid node graph for each product storage area include:

[0010] A1: Obtain N divided storage areas. Based on the actual physical dimensions of each storage area, use 3D modeling software to digitally model each area.

[0011] A2: Divide each storage area into grids based on the modeling results. After the grid division is completed, use the Cartesian coordinate system with the lower left corner of the storage area as the coordinate origin. , determine the three-dimensional coordinates of each grid node according to the position of the grid in three-dimensional space , and assign a unique number to each grid node in the form of area code + row and column layer number;

[0012] A3: Define the basic attributes of each grid node and establish a node attribute database; the basic attributes include but are not limited to the corresponding product type, the current remaining quantity of the product, the node's load-bearing limit, lighting requirements, and temperature and humidity requirements;

[0013] A4: Analyze the correlation between products. Using graph theory principles, construct the connection relationship between grid nodes based on the degree of correlation between products, and form a grid node graph with product correlation information; the correlation includes complementary relationships, substitution relationships, and collocation habits; the connection relationship includes strong connections and weak connections.

[0014] Specifically, the specific steps of A4 include:

[0015] A4.1: Extract inventory data for each product at different time points over a month from the unmanned retail store system, and collect customer behavior data within the store to form retail business data. This behavior data includes the time customers spend at the shelf, and the records of when they pick up and put down items.

[0016] A4.2: Convert the retail business data into a transaction dataset, scan the transaction dataset, count the number of occurrences of each individual item, and generate frequent 1-itemsets. Each transaction in the transaction dataset represents an order and includes all items purchased in the order. The frequent 1-itemsets are sets of individual items.

[0017] A4.3: Based on the frequent 1-itemset, generate a candidate 2-itemset through a concatenation operation; the candidate 2-itemset is a product combination consisting of two individual items;

[0018] A4.4: Perform a second scan of the transaction dataset for candidate 2-itemsets and calculate the support of each candidate 2-itemset by calculating the ratio of the number of times any two items A and B appear in orders at the same time to the total number of orders. ;

[0019] A4.5: Delete candidate 2-itemsets whose support is less than the preset minimum support threshold to obtain frequent 2-itemsets;

[0020] A4.6: Based on frequent 2-itemsets, iteratively generate frequent N-itemsets by performing concatenation and pruning operations.

[0021] A4.7: For each frequent item set generated, calculate The ratio of the support of item A to the confidence of all possible association rules is obtained ,in, Indicates an association relationship;

[0022] A4.8: By calculation The ratio of the support of item B to the support of item B is used to obtain the improvement of all possible association rules. ;

[0023] A4.9: Filter out the product association rules that meet the preset minimum support threshold and minimum confidence threshold to obtain the preliminary association relationship between the products.

[0024] Specifically, the specific steps of A4 further include:

[0025] A4.10: Based on the relationships between products and the remaining quantity of each product in the basic attributes, the sales attributes of each product are extracted. The products are then clustered according to their sales attributes and the preset number of clusters to form product cluster groups. The sales attributes include quantity and frequency.

[0026] A4.11: Create a graph with the grid nodes corresponding to each product as vertices. Each vertex contains basic product information. Create an empty graph structure and, based on the product clustering results, add all product nodes that have undergone cluster analysis to the graph to complete the initial construction of the graph structure. The basic information includes product type and current inventory quantity.

[0027] A4.12: Build edges between nodes based on the association rules and association levels between items. The association levels are measured based on the confidence level in the association rules, including strong and weak associations.

[0028] If the confidence of the association rule of the product node is greater than the preset maximum confidence value, it means there is a strong association, and the confidence value is used as the weight of the edge to establish a strong connection edge;

[0029] If the confidence of the association rule of the product node is greater than or equal to the preset minimum confidence value and less than or equal to the preset maximum confidence value, then the connection is made through the intermediate node and the weight of the edge is calculated based on the association transitivity;

[0030] If the confidence of the association rule of the product node is less than the preset minimum confidence value, it indicates that there is a weak association, and the node related to the product and all its connecting edges are deleted, and the association information of the product is removed from the graph structure to obtain the constructed grid node graph;

[0031] A4.13: Visualize the constructed grid node graph.

[0032] Specifically, the specific steps of A4.12 include:

[0033] A4.12.1: Traverse the selected product association rules, for each rule , find the corresponding vertices a and b in the graph structure according to the unique identifier of the product;

[0034] If both vertices exist, an edge is constructed between them;

[0035] A4.12.2: For constructed edges, evaluate the degree of association based on the confidence metric in the association rule and set the maximum and minimum confidence values.

[0036] If the confidence of the association rule is greater than the maximum confidence value, the confidence value is used as the weight of the edge to establish a strong connection edge;

[0037] If the confidence of the association rule is greater than or equal to the minimum confidence value and less than or equal to the maximum confidence value, then find the vertex d in the graph structure that corresponds to the intermediate product D that has a strong association with any two products A and B, build edges between a and d, and d and b respectively, and calculate according to the steps in A4.7 and , by calculating and The product of gets the weight of the corresponding edge;

[0038] If the confidence of the association rule is less than the minimum confidence value, the node related to the product and all its connecting edges are deleted, and the association information of the product is removed from the graph structure.

[0039] Specifically, the steps of constructing a product storage map and forming a product association network through partition connection include:

[0040] B1: Sensors on shelves are used to obtain real-time information about the product category and remaining quantity at each grid node. Furthermore, environmental sensors, combined with meteorological data and a known database of product characteristics, determine the seasonality of the products at each grid node.

[0041] B2: Integrate the locations and relationships of nodes in the grid node graph with the real-time data collected by sensors on the shelves and environmental sensors to form a comprehensive product dataset;

[0042] B3: Based on the product's historical remaining quantity data and inventory management strategy, the remaining quantity of the product is divided into M intervals, and each remaining quantity interval is used as a new node. Each new node contains information about the product type, remaining quantity interval range, and seasonal status attributes.

[0043] B4: Based on the comprehensive product dataset, re-evaluate the correlation between products according to steps A4.2-A4.9. Based on the degree of correlation between products, construct edges in a graph with product remaining quantity intervals as nodes.

[0044] B5: Build a product storage graph using the remaining quantity intervals of products as nodes and the associated edges between products as connections. During the graph construction process, the node attribute information is updated in real time.

[0045] B6: Divide the product storage map into Q partitions based on product categories and storage areas, analyze the product associations between different partitions, and establish connecting edges between the partitions; the weights of the connecting edges are determined based on the degree of association between the products in the partitions;

[0046] B7: Through partition connection, the product storage maps of each partition are integrated together to form a complete product association network.

[0047] Specifically, based on the product association network, a product association sales model is constructed, and a product migration strategy is formulated based on the seasonal status and the remaining quantity of the product. At the same time, a differentiated replenishment model is established based on the migration results and the real-time remaining quantity of the product to determine the replenishment demand of each product and formulate a dynamic replenishment strategy, including:

[0048] C1: Collect product association data in the product association network; the product association data includes but is not limited to product association degree, attribute information, and behavior data;

[0049] C2: Use feature extraction and one-hot encoding methods to extract features and encode the product-related data to obtain processed product-related data;

[0050] C3: Load the pre-built random forest regression model and train it using the processed product-related data to obtain a product-related sales model that can predict the remaining quantity of the product based on the input features.

[0051] C4: Based on the seasonal status information of each product, quantitatively evaluate the seasonal status of each product. At the same time, analyze the remaining quantity of each product and calculate the inventory turnover rate based on the historical remaining quantity of each product to determine the inventory backlog or shortage of the product.

[0052] C5: Formulate a product migration strategy based on the product remaining quantity predicted by the product-related sales model, the quantitative assessment results of seasonal status, and the remaining quantity analysis.

[0053] C6: Integrate product migration results data, real-time remaining quantity data, historical product remaining quantity data, and supplier supply information to determine factors influencing replenishment and construct a differentiated replenishment model using dynamic programming. This differentiated replenishment model aims to minimize inventory costs and out-of-stock losses. It inputs replenishment influencing factor data and outputs the optimal replenishment quantity and replenishment time for each product.

[0054] C7: Generate a replenishment plan based on the calculation results of the differential replenishment model, combined with the actual inventory situation and the required remaining quantity of the goods. The replenishment plan includes a replenishment product list, replenishment quantity, replenishment time, and supplier selection.

[0055] C8: Monitor the remaining quantity of goods and changes in inventory data in real time, and adjust the replenishment plan.

[0056] Specifically, each grid node corresponds to a shelf location and records the type of goods and the remaining quantity of goods at the shelf location; the equipment operation data includes the working status, operating time, and number of failures of the equipment.

[0057] An unmanned intelligent supermarket system includes: a grid node construction module, an information acquisition and judgment module, a product storage map construction module, a product strategy formulation module, and a fault prediction module, which cooperate to perform unmanned operations;

[0058] The grid node construction module is used to reasonably divide the supermarket area of ​​the unmanned retail store and construct a grid node map frame;

[0059] The information acquisition and judgment module is used to obtain product information in real time and judge the seasonality of the product;

[0060] The product storage map construction module is used to construct a product storage map and a product association network;

[0061] The product strategy formulation module is used to formulate product migration strategies and dynamic replenishment strategies based on the product association network;

[0062] The fault prediction module is used to collect equipment operation data in real time, predict equipment failures and issue early warnings.

[0063] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of a method for implementing an unmanned intelligent supermarket system when executing the computer program.

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] 1. The present invention proposes a method for implementing an unmanned intelligent supermarket system. In terms of product management, by scientifically dividing the supermarket area and constructing a grid node diagram, combined with real-time sensor data collection and a seasonal product database, the product status can be accurately grasped. Based on this, the product-related sales model, product migration strategy, and differentiated replenishment model established can dynamically adjust the product layout and replenishment plan according to the seasonal status, remaining quantity, and sales correlation of the product, effectively avoiding inventory backlogs or stock-outs, improving product circulation efficiency, and optimizing shelf space utilization.

[0066] 2. The present invention proposes a method for implementing an unmanned intelligent supermarket system. In terms of equipment management, sensors are used to collect equipment operation data in real time. With the help of an equipment failure prediction model established by a machine learning algorithm, the system can perceive the potential risk of equipment failure in advance and issue early warning information in a timely manner. This allows equipment maintenance personnel to take preventive maintenance measures before a failure occurs, reducing equipment downtime and maintenance costs, ensuring the normal operation of unmanned retail stores, and improving the intelligence and reliability of operation management. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 This is a schematic diagram of a method for implementing an unmanned intelligent supermarket system according to the present invention;

[0068] Figure 2 This is a principle flow chart of an implementation method of an unmanned intelligent supermarket system of the present invention;

[0069] Figure 3 A flowchart of forming a commodity association network in a method for implementing an unmanned intelligent supermarket system according to the present invention;

[0070] Figure 4 This is an architecture diagram of an unmanned intelligent supermarket system of the present invention. DETAILED DESCRIPTION

[0071] Example 1

[0072] See also Figure 1-Figure 3 The present invention provides an embodiment of a method for implementing an unmanned intelligent supermarket system, comprising: executing an unmanned retail store intelligent replenishment and fault warning method to realize the unmanned intelligent supermarket system's unmanned function. The unmanned intelligent supermarket system's unmanned function refers to the use of technical means to achieve automated and intelligent management of the entire supermarket operation process, thereby maintaining normal business operations without direct human involvement in on-site operations. The method comprises steps S101 to S103, including the following steps:

[0073] S101: The supermarket area of ​​the unmanned retail store is divided into storage areas. A grid node diagram is constructed for each product storage area. Sensors are used to obtain real-time information on the corresponding product category and remaining quantity, as well as the seasonal status of the products at each grid node. A product storage map is constructed based on the product association, using the remaining quantity intervals as nodes. The product association network is formed by connecting the partitions.

[0074] Each grid node corresponds to a shelf location and records the type of goods and the remaining quantity of goods at the shelf location.

[0075] The division of the supermarket area of ​​the unmanned retail store into storage areas is based on the following: based on the function of the supermarket area, such as dividing the area into food area, daily necessities area, fresh food area, etc.; at the same time, combined with the product category, such as further subdividing the food area into snacks, beverages, grains and oils, etc. Based on these functions and categories, the entire supermarket area is preliminarily divided into large areas to obtain N product storage areas, which define the scope for the subsequent grid node diagram construction.

[0076] S102: Based on the product association network, a product association sales model is constructed. Product migration strategies are formulated based on seasonality and product remaining quantities. Furthermore, a differentiated replenishment model is established based on the migration results and the real-time product remaining quantities. This model determines the replenishment demand for each product and formulates a dynamic replenishment strategy.

[0077] S103: By deploying sensors on the equipment to collect equipment operation data in real time, a machine learning algorithm is used to establish an equipment failure prediction model. When the risk of equipment failure is predicted, early warning information is issued in a timely manner.

[0078] The equipment operation data includes the equipment's working status, operating time, and number of failures.

[0079] Furthermore, the equipment failure prediction model is configured based on a convolutional neural network model and is trained in combination with real-time collected operating data to obtain a trained equipment failure prediction model. The convolutional neural network model is the existing technology content in this field and is not the creative solution of this application, so it will not be elaborated here.

[0080] The specific steps of constructing a grid node graph for each product storage area in step S101 include:

[0081] A1: Obtain N divided product storage areas, and digitally model each product storage area using 3D modeling software based on the actual physical space dimensions of each product storage area. In the present invention, the 3D modeling software is AutoCAD;

[0082] It should be noted that when conducting digital modeling, it is necessary to set the basic parameters of grid division, including the shape, size and layout direction of the grid, to ensure that the grid division can not only meet the needs of product display, but also facilitate subsequent data collection and management. Among them, the shape of the grid includes square and rectangle; the size, for example, 50cm×50cm is a basic grid unit, which can be flexibly adjusted according to the size of the goods and shelf specifications; the grid layout direction includes horizontal, vertical or mixed layout.

[0083] A2: Divide each storage area into grids based on the modeling results. After the grid division is completed, use the Cartesian coordinate system with the lower left corner of the storage area as the coordinate origin. , determine the three-dimensional coordinates of each grid node according to the position of the grid in three-dimensional space , and assign a unique number to each grid node in the form of area code + row, column and layer number, such as "A - 03 - 05 - 02" represents the grid node in the 3rd row, 5th column and 2nd layer of area A, which facilitates subsequent data recording and query;

[0084] A3: Define the basic attributes of each grid node and establish a node attribute database; the basic attributes include but are not limited to the corresponding product type, the current remaining quantity of the product, the node's load-bearing limit, lighting requirements, and temperature and humidity requirements;

[0085] A4: Analyze the correlation between products. Using graph theory principles, build connection relationships between nodes based on the degree of correlation between products to form a grid node graph with product correlation information. The correlation includes complementary relationships, substitution relationships, and collocation habits. The connection relationship includes strong connections and weak connections.

[0086] For example, for the grid nodes corresponding to two products that are often purchased at the same time, a strong association connection is established; for the grid nodes corresponding to products that have a substitution relationship, a weak association connection is established.

[0087] The specific steps of A4 include:

[0088] A4.1: Extract inventory data for each product at different time points over a month from the unmanned retail store system, and collect customer behavior data within the store to form retail business data. This behavior data includes the time customers spend at the shelf, and the records of when they pick up and put down items.

[0089] Specifically, SQL query statements such as SELECT product_id, stock_quantity, timestamp FROM inventory_log WHERE timestamp BETWEEN start_date AND end_date are used to extract inventory data of different products at different time points within a month from the unmanned retail store system, covering detailed information such as inventory quantity, replenishment time, and inventory turnover rate. At the same time, with the help of sensor networks deployed in the store, such as infrared sensors, cameras, and behavior monitoring equipment, customer behavior data in the store is collected, including the time customers stay in front of the shelves, the number and order of picking up and putting down goods, and the shopping path. The inventory data and behavior data are imported into the data processing platform Spark for cleaning, deduplication, and unified format processing. Then, according to the time dimension of the data and the product identification, the two types of data are correlated and integrated to form retail business data, which serves as the basic data source for subsequent analysis.

[0090] A4.2: Convert the retail business data into a transaction dataset, scan the transaction dataset, count the number of occurrences of each individual item, and generate frequent 1-itemsets. Each transaction in the transaction dataset represents an order and includes all items purchased in the order. The frequent 1-itemsets are sets of individual items.

[0091] A4.3: Based on the frequent 1-itemset, generate a candidate 2-itemset through a concatenation operation; the candidate 2-itemset is a product combination consisting of two individual items;

[0092] A4.4: Perform a second scan of the transaction dataset for candidate 2-itemsets and calculate the support of each candidate 2-itemset by calculating the ratio of the number of times any two items A and B appear in orders at the same time to the total number of orders. ;

[0093] A4.5: Delete candidate 2-itemsets whose support is less than the preset minimum support threshold to obtain frequent 2-itemsets;

[0094] A4.6: Based on frequent 2-itemsets, iteratively generate frequent N-itemsets by performing concatenation and pruning operations.

[0095] Among them, when generating frequent N-item sets, frequent 3-item sets and frequent 4-item sets are iteratively generated through iterative repetition until higher-order frequent item sets cannot be generated, so as to obtain frequent N-item sets.

[0096] A4.7: For each frequent item set generated, calculate The ratio of the support of item A to the confidence of all possible association rules is obtained ,in, Indicates an association relationship;

[0097] A4.8: By calculation The ratio of the support of item B to the support of item B is used to obtain the improvement of all possible association rules. ;

[0098] A4.9: Filter out the product association rules that meet the preset minimum support threshold and minimum confidence threshold, such as , get the preliminary association relationship between goods;

[0099] A4.10: Based on the relationships between products and the remaining quantity of each product in the basic attributes, the sales attributes of each product are extracted. The products are then clustered according to their sales attributes and the preset number of clusters to form product cluster groups. The sales attributes include quantity and frequency.

[0100] It should be noted that due to the different dimensions of sales quantity and sales frequency, in order to avoid affecting the clustering results, the Z-Score standardization method is used to process the sales attribute data. Among them, the Z-Score standardization method is the existing technology content in this field and is not an inventive solution of this application, so it will not be elaborated here.

[0101] In the present invention, the K-Means clustering algorithm is used to cluster the standardized product sales attribute data, including:

[0102] (1) Determine the preset number of clusters K by the elbow rule;

[0103] (2) Calculate the intra-cluster sum of squared errors (SSE) under different K values;

[0104] (3) Select the K value at the inflection point of the SSE curve as the optimal number of clusters, randomly initialize K cluster centers, and calculate the Euclidean distance from each product data point to each cluster center. The Euclidean distance calculation formula is the existing technical content in this field and is not the inventive solution of this application, so it will not be described in detail here;

[0105] (4) Assign the goods to the cluster to which the nearest cluster center belongs;

[0106] (5) Recalculate the center position of each cluster and repeat the steps of allocating and calculating the center until the cluster center reaches the preset number of iterations, and finally form a cluster grouping of goods.

[0107] A4.11: Create a graph with the grid nodes corresponding to each product as vertices. Each vertex contains basic product information. Create an empty graph structure and, based on the product clustering results, add all product nodes that have undergone cluster analysis to the graph to complete the initial construction of the graph structure. The basic information includes product type and current inventory quantity.

[0108] Furthermore, the specific steps of A4.11 include:

[0109] (1) Create a corresponding grid node for each product as a vertex in the graph. Each vertex contains information about the product type, current inventory quantity, and the cluster category to which it belongs;

[0110] (2) Use the graph data structure library NetworkX to create an empty undirected graph structure;

[0111] (3) Traverse the product clustering results, create a vertex object for each product based on its information, and add the vertex object to the graph structure to complete the initial construction of the graph structure;

[0112] (4) Use the dictionary vertex_mapping = {product_id: Vertex} to establish a mapping relationship between the product’s unique identifier and the vertex object.

[0113] A4.12: Build edges between nodes based on the association rules and association levels between items. The association levels are measured based on the confidence level in the association rules, including strong and weak associations.

[0114] If the confidence of the association rule of the product node is greater than the preset maximum confidence value, it means there is a strong association, and the confidence value is used as the weight of the edge to establish a strong connection edge;

[0115] If the confidence of the association rule of the product node is greater than or equal to the preset minimum confidence value and less than or equal to the preset maximum confidence value, then the connection is made through the intermediate node and the weight of the edge is calculated based on the association transitivity;

[0116] If the confidence of the association rule of the product node is less than the preset minimum confidence value, it indicates that there is a weak association, and the node related to the product and all its connecting edges are deleted, and the association information of the product is removed from the graph structure to obtain the constructed grid node graph;

[0117] The specific steps of A4.12 include:

[0118] A4.12.1: Traverse the selected product association rules, for each rule , find the corresponding vertices a and b in the graph structure according to the unique identifier of the product;

[0119] If both vertices exist, an edge is constructed between them;

[0120] A4.12.2: For constructed edges, evaluate the degree of association based on the confidence metric in the association rule and set the maximum and minimum confidence values.

[0121] If the confidence of the association rule is greater than the maximum confidence value, the confidence value is used as the weight of the edge to establish a strong connection edge;

[0122] If the confidence of the association rule is greater than or equal to the minimum confidence value and less than or equal to the maximum confidence value, then find the vertex d in the graph structure that corresponds to the intermediate product D that has a strong association with any two products A and B, build edges between a and d, and d and b respectively, and calculate according to the steps in A4.7 and , by calculating and The product of gets the weight of the corresponding edge;

[0123] If the confidence of the association rule is less than the minimum confidence value, the node related to the product and all its connecting edges are deleted, and the association information of the product is removed from the graph structure.

[0124] A4.13: Visualize the constructed grid node graph.

[0125] In the present invention, the graph visualization tool Gephi is used to visualize the constructed grid node graph. In Gephi, graph structure data, such as node information and edge information, is imported into the software, node color is set according to cluster category, node size is adjusted according to inventory quantity, and edge color and thickness are adjusted according to association weight display, so as to intuitively display the association relationship and cluster distribution between goods. At the same time, interactive functions are added, such as hover prompts for nodes and edges, displaying information such as product name and association rules, and adding zoom and drag operations to facilitate user analysis and exploration.

[0126] The specific steps of constructing a product storage map and forming a product association network through partition connection include:

[0127] B1: Sensors on the shelves are used to obtain real-time information about the product category and remaining quantity at each grid node. Furthermore, environmental sensors, including temperature, humidity, and light sensors, are used to determine the seasonality of the products at each grid node, combined with meteorological data and a known database of product characteristics.

[0128] B2: Integrate the locations and relationships of nodes in the grid node graph with real-time data collected by sensors on shelves and environmental sensors to form a comprehensive product dataset; the comprehensive product dataset includes product type, remaining quantity, seasonality, and node association information;

[0129] B3: Based on the product's historical remaining quantity data and inventory management strategy, the remaining quantity of the product is divided into M intervals, and each remaining quantity interval is used as a new node. Each new node contains information about the product type, remaining quantity interval range, and seasonal status attributes.

[0130] For example, low inventory is 0-10 pieces, medium inventory is 11-50 pieces, and high inventory is 51 pieces and above;

[0131] B4: Based on the comprehensive product dataset, re-evaluate the correlation between products according to steps A4.2-A4.9. Based on the degree of correlation between products, construct edges in a graph with product remaining quantity intervals as nodes.

[0132] For example, if the confidence of the association rule between product A and product B is 0.9, an edge with a weight of 0.9 can be established between the remaining quantity interval nodes corresponding to product A and product B.

[0133] B5: Build a product storage graph using the remaining quantity intervals of products as nodes and the associated edges between products as connections. During the graph construction process, the node attribute information is updated in real time.

[0134] Among them, the goods storage map reflects the relationship and storage status of different goods in different remaining quantity ranges.

[0135] Furthermore, the specific steps of B3-B5 include:

[0136] (1) Collect and organize comprehensive product data sets, including product types, remaining quantities, seasonal status, grid node locations, and related relationship information, conduct statistical analysis on the remaining quantity data, and determine reasonable remaining quantity interval division standards based on historical sales data and inventory management strategies;

[0137] (2) Create a new node for each interval based on the remaining quantity of the goods. When each node is initialized, it is given the attribute information of the product type, the remaining quantity interval range, and the seasonal status. For example, for a certain product node in the low inventory interval, its corresponding product type, the interval range of 0-10 pieces, and the seasonal status determined based on the previous judgment, such as seasonal or non-seasonal, are recorded.

[0138] (3) Based on the comprehensive data set of goods, analyze the degree of association between goods according to the established methods in A4.2-A4.9. For goods with associated relationships, build connecting edges between the corresponding remaining quantity interval nodes, where the edge weight is determined according to the confidence index in the association rule;

[0139] (4) Integrate all created nodes and constructed edges to form the basic structure of the goods storage map, and use graph theory to ensure the logical consistency and structural rationality of the map. Graph theory is the existing technology in this field and is not the inventive solution of this application, so it will not be described in detail here;

[0140] (5) Start the real-time data monitoring program to continuously obtain the latest data on the types, remaining quantities, and seasonal status of goods collected by shelf sensors and environmental sensors. When it is detected that the remaining quantity of the goods has changed, resulting in a change in the remaining quantity interval to which it belongs, the remaining quantity interval range attribute of the corresponding node is updated in a timely manner. At the same time, if the seasonal status of the goods changes, the seasonal status attribute information of the node is also updated synchronously to ensure that the map always accurately reflects the actual status of the goods.

[0141] B6: Divide the product storage map into Q partitions based on product categories and storage areas, analyze the product associations between different partitions, and establish connecting edges between the partitions; the weights of the connecting edges are determined based on the degree of association between the products in the partitions;

[0142] Furthermore, the specific steps of B6 include:

[0143] (1) Determine the zoning basis and zoning rules, where the storage areas are divided according to the physical layout and functional area settings of the store. For example, the store is divided into shelf area, refrigerated area, frozen area, and warehouse area. The shelf area is further subdivided according to the floor and shelf number. The warehouse area is divided into different storage units according to the storage requirements of the goods, such as turnover rate and weight, and the location coordinates, space size, temperature and humidity conditions of each storage area are recorded. The zoning rules include: dividing goods of the same category and the same storage conditions into one partition, or dividing goods in the same storage area with strong sales correlation into one partition;

[0144] (2) Create Q empty partition containers, assign a unique identifier to each partition, and initialize partition attributes, such as partition name, category range, and storage area range;

[0145] (3) Traverse all nodes in the product storage graph, that is, the nodes of the remaining quantity interval of the product, and assign the nodes to the corresponding partitions according to the category of the product represented by each node and the actual storage area, according to the partitioning rules. For example, for a snack product node, if it is stored on the snack shelf in the shelf area, and according to the partitioning rules, snack products are uniformly divided into the food-snack partition, then add the node to the food-snack partition container;

[0146] (4) Check the number and distribution of nodes in each partition. If the number of nodes in any partition is found to be less than 1 / 3 of the average number of nodes or more than 3 times the average number of nodes, re-evaluate the partitioning rules and adjust the nodes to ensure the balance and rationality of the partitioning. At the same time, consider the dynamic changes of goods and optimize the partitioning every month.

[0147] (5) Extracting the associated information of goods in different partitions, for example, counting the number of times goods from different partitions appear in the same order;

[0148] (6) Using a collaborative filtering algorithm to calculate the similarity between partitions based on customers’ purchase preferences for products in different partitions, wherein the collaborative filtering algorithm is a prior art in this field and is not an inventive solution of this application, and is not described in detail here;

[0149] (7) Based on the similarity results, determine the partition pairs with associated relationships and create connecting edges in the product storage graph, where each edge connects two partition nodes and is given a unique identifier;

[0150] (8) Determine the edge weight based on the degree of association between partitions and set the edge attributes, where the edge attributes include the edge type and update time.

[0151] B7: Through partition connection, the product storage maps of each partition are integrated together to form a complete product association network.

[0152] Among them, the product association network reflects the comprehensive association relationship and storage status between goods in the entire unmanned retail store.

[0153] Based on the product association network, a product association sales model is constructed, and a product migration strategy is formulated based on the seasonal status and the remaining quantity of the product. At the same time, a differentiated replenishment model is established based on the migration results and the real-time remaining quantity of the product to determine the replenishment demand of each product and formulate a dynamic replenishment strategy, including:

[0154] C1: Collect product association data in the product association network; the product association data includes but is not limited to product association degree, attribute information, and behavior data;

[0155] C2: Using a feature extraction method and a one-hot encoding method to perform feature extraction and encoding processing on the product-related data to obtain processed product-related data, wherein the feature extraction method adopts a principal component analysis method. The principal component analysis method and the one-hot encoding method are prior art contents in this field and do not constitute the inventive solution of this application, and are not described in detail here;

[0156] C3: Load a pre-built random forest regression model and train it using the processed product-related data to obtain a product-related sales model that can predict the remaining quantity of the product based on the input features. The random forest regression model is prior art in this field and does not constitute the inventive solution of this application, so it will not be described in detail here.

[0157] C4: Based on the seasonal status information of each product, quantitatively evaluate the seasonal status of each product. At the same time, analyze the remaining quantity of each product and calculate the inventory turnover rate based on the historical remaining quantity of each product to determine the inventory backlog or shortage of the product.

[0158] Furthermore, the specific steps of C4 include:

[0159] (1) Determine the key dimensions of seasonal status assessment based on the changing patterns of remaining quantity of goods. For example, for food products, consider the impact of holidays and climate on consumer demand; for clothing products, consider seasonal changes, fashion trends, and other factors;

[0160] (2) Set specific quantitative indicators and scoring standards for each evaluation dimension. For example, the seasonal status can be divided into five levels, each corresponding to a different score range: strong seasonality is 8-10 points, relatively seasonality is 6-7 points, moderate seasonality is 4-5 points, relatively unseasonal is 2-3 points, and non-seasonal is 0-1 point;

[0161] (3) Use the weighted average method to construct an evaluation model, assign corresponding weights to different evaluation dimensions, and calculate the comprehensive score of each dimension according to the weight, so as to determine the quantitative value of the seasonal status of the goods;

[0162] (4) Extract the beginning inventory quantity, ending inventory quantity and remaining quantity of each product in a month from the inventory management system;

[0163] (5) Add the beginning inventory quantity to the ending inventory quantity and divide by 2 to obtain the average inventory quantity of each product;

[0164] (6) Subtract the remaining quantity of the product from the total quantity of the product and divide it by the average inventory to obtain the inventory turnover rate for each product;

[0165] (7) Refer to the industry average inventory turnover rate and set a reasonable range for inventory turnover rate. For example, for the fast-moving consumer goods industry, the reasonable range for inventory turnover rate is 8-12 times / month; for the durable goods industry, the reasonable range is 2-4 times / month;

[0166] (8) Compare the actual inventory turnover rate of each product with the set reasonable range. If the inventory turnover rate is lower than the lower limit of the reasonable range, it is judged that the product has an inventory backlog; if it is higher than the upper limit of the reasonable range, it is judged that the product may have a shortage risk.

[0167] C5: Formulate a product migration strategy based on the product remaining quantity predicted by the product-related sales model, the quantitative assessment results of seasonal status, and the remaining quantity analysis.

[0168] The product migration strategy includes:

[0169] Products with a seasonal status quantification value of 8 or more in the seasonal product database, whose sales forecast within 7 days exceeds 1.5 times the average daily sales of the product within 30 days as predicted by the product-related sales model, and whose current grid node has an average daily customer flow of less than 60% of the store's average customer flow, and whose remaining inventory is greater than 2 times the safety stock threshold, will be preferentially moved to the top 20% of high-traffic grid nodes within the store.

[0170] For products in the seasonal product database with a seasonal status quantification value of ≤3 points, an inventory turnover rate lower than the industry standard of 40%, and a remaining inventory quantity exceeding three times the average sales volume within 30 days, they will be moved to the promotion area or displayed adjacent to seasonal products with a seasonal status quantification value ≥7 points, a confidence level based on association rules ≥0.8, and a lift ≥1.2.

[0171] C6: Integrate product migration results data, real-time remaining quantity data, historical product remaining quantity data, and supplier supply information to determine factors influencing replenishment and construct a differentiated replenishment model using dynamic programming. This differentiated replenishment model aims to minimize inventory costs and out-of-stock losses. It inputs replenishment influencing factor data and outputs the optimal replenishment quantity and replenishment time for each product.

[0172] C7: Generate a replenishment plan based on the calculation results of the differential replenishment model, combined with the actual inventory situation and the required remaining quantity of the goods. The replenishment plan includes a replenishment product list, replenishment quantity, replenishment time, and supplier selection.

[0173] C8: Monitor the remaining quantity of goods and changes in inventory data in real time, and adjust the replenishment plan.

[0174] Example 2

[0175] See also Figure 4 Another embodiment provided by the present invention is an unmanned intelligent supermarket system, comprising:

[0176] Grid node construction module, information acquisition and judgment module, product storage map construction module, product strategy formulation module, and fault prediction module;

[0177] The grid node construction module is used to reasonably divide the supermarket area of ​​the unmanned retail store and build a grid node graph to provide a basic framework for subsequent management;

[0178] The information acquisition and judgment module is used to obtain product information in real time and determine the seasonality of products, providing a basis for product management and strategy formulation;

[0179] The product storage map construction module is used to build a product storage map and a product association network, clearly displaying the relationship between products for easy analysis and management;

[0180] The product strategy formulation module is used to formulate product migration strategies and dynamic replenishment strategies based on the product association network, and optimize product layout and inventory management;

[0181] The fault prediction module is used to collect equipment operation data in real time, predict equipment failures and issue early warnings to ensure the normal operation of the equipment.

[0182] The grid node construction module includes: a region division unit and a grid node graph construction unit;

[0183] An area division unit is used to divide the supermarket area of ​​the unmanned retail store according to function and product category to obtain N product storage areas;

[0184] The grid node graph construction unit is used to construct a grid node graph for each product storage area, where each grid node corresponds to a shelf location and records the type of product and the remaining quantity of the product at the shelf location.

[0185] The information acquisition and judgment module includes: a product information acquisition unit and a judgment unit;

[0186] The product information acquisition unit is used to obtain the corresponding product type information and the remaining quantity of the product in real time through the sensors on the shelves for each grid node;

[0187] The judgment unit is used to establish a seasonal commodity database and judge the seasonal status of the goods at each grid node.

[0188] The goods storage map construction module includes: a map node determination unit and a relationship establishment unit;

[0189] A graph node determination unit, used to take the remaining quantity interval of each product as a node in the graph;

[0190] The relationship establishment unit is used to establish connection relationships between nodes based on the correlation between goods, form a goods storage map, and form a commodity association network through partition connection.

[0191] The product strategy formulation module includes: model building unit, strategy formulation unit, and replenishment model establishment unit;

[0192] A model building unit, used to build a product association sales model based on the product association network;

[0193] Strategy formulation unit, used to formulate product migration strategies based on seasonal status and remaining product quantity;

[0194] The replenishment model establishment unit is used to establish a replenishment model with different product quantities based on the product migration results and the real-time remaining quantity of the products, calculate the replenishment demand for each product, and formulate a dynamic replenishment strategy.

[0195] The fault prediction module includes: equipment operation data acquisition unit, fault prediction model establishment unit, and notification unit;

[0196] Equipment operation data collection unit, used to collect equipment operation data in real time through sensors, including equipment working status, operating time, and number of faults;

[0197] A fault prediction model building unit is used to build a device fault prediction model using a machine learning algorithm to predict the time and type of device failure;

[0198] The notification unit is used to issue early warning information in a timely manner when a failure risk of the equipment is predicted, and notify maintenance personnel to carry out maintenance.

[0199] Example 3

[0200] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the steps of a method for implementing an unmanned intelligent supermarket system. For details, please refer to the above method embodiment, which will not be repeated here.

[0201] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific embodiments. The above-mentioned specific embodiments are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also change, modify, replace and modify the above-mentioned embodiments without departing from the purpose and scope of protection of the present invention. These are all protected by the present invention.

[0202] If the technical solution disclosed herein involves personal information, the product using the technical solution disclosed herein has clearly informed the individual of the personal information processing rules and obtained the individual's voluntary consent before processing the personal information. If the technical solution disclosed herein involves sensitive personal information, the product using the technical solution disclosed herein has obtained the individual's separate consent before processing the sensitive personal information and at the same time meets the "explicit consent" requirement. For example, on personal information collection devices such as cameras, a clear and prominent sign is set to inform that the individual has entered the personal information collection scope and that personal information will be collected. If the individual voluntarily enters the collection scope, it is deemed that they agree to the collection of their personal information; or on the personal information processing device, when the personal information processing rules are notified by obvious signs / information, the individual's authorization is obtained through pop-up information or by asking the individual to upload their personal information. The personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the type of personal information processed.

Claims

1. A method for implementing an unmanned intelligent supermarket system, characterized in that: include: Executing the unmanned retail store intelligent replenishment and fault warning method to realize the unmanned intelligent supermarket system unmanned function, including: The supermarket area of ​​the unmanned retail store is divided into storage areas. A grid node diagram is constructed for each product storage area. Sensors are used to obtain real-time information on the corresponding product type and remaining quantity, as well as the seasonal status of the products at each grid node. The remaining quantity intervals of the products are used as nodes. Combined with the product association, a product storage map is constructed, and the product association network is formed by connecting the partitions. Based on the product association network, a product association sales model is constructed, and a product migration strategy is formulated based on the seasonal status and the remaining quantity of the product. At the same time, a differentiated replenishment model is established based on the migration results and the real-time remaining quantity of the product to determine the replenishment demand of each product and formulate a dynamic replenishment strategy; By deploying sensors on equipment to collect real-time equipment operation data, and using machine learning algorithms to establish equipment failure prediction models, early warning information is issued in a timely manner when equipment failure risks are predicted; The specific steps of constructing a grid node graph for each product storage area include: A1: Obtain N divided storage areas. Based on the actual physical dimensions of each storage area, use 3D modeling software to digitally model each area. A2: Divide each storage area into grids based on the modeling results. After the grid division is completed, use the Cartesian coordinate system with the lower left corner of the storage area as the coordinate origin. , determine the three-dimensional coordinates of each grid node according to the position of the grid in three-dimensional space , and assign a unique number to each grid node in the form of area code + row and column layer number; A3: Define the basic attributes of each grid node and establish a node attribute database; the basic attributes include but are not limited to the corresponding product type, the current remaining quantity of the product, the node's load-bearing limit, lighting requirements, and temperature and humidity requirements; A4: Analyze the associations between products. Using graph theory, construct connection relationships between grid nodes based on the degree of association between products, forming a grid node graph with product association information. The associations include complementary relationships, substitution relationships, and collocation habits. The connection relationships include strong connections and weak connections. Based on the product association network, a product association sales model is constructed, and a product migration strategy is formulated based on the seasonal status and the remaining quantity of the product. At the same time, a differentiated replenishment model is established based on the migration results and the real-time remaining quantity of the product to determine the replenishment demand of each product and formulate a dynamic replenishment strategy, including: C1: Collect product association data in the product association network; the product association data includes but is not limited to product association degree, attribute information, and behavior data; C2: Use feature extraction and one-hot encoding methods to extract features and encode the product-related data to obtain processed product-related data; C3: Load the pre-built random forest regression model and train it using the processed product-related data to obtain a product-related sales model that can predict the remaining quantity of the product based on the input features. C4: Based on the seasonal status information of each product, quantitatively evaluate the seasonal status of each product. At the same time, analyze the remaining quantity of each product and calculate the inventory turnover rate based on the historical remaining quantity of each product to determine the inventory backlog or shortage of the product. C5: Formulate a product migration strategy based on the product remaining quantity predicted by the product-related sales model, the quantitative assessment results of seasonal status, and the remaining quantity analysis. C6: Integrate product migration results data, real-time remaining quantity data, historical product remaining quantity data, and supplier supply information to determine factors influencing replenishment and construct a differentiated replenishment model using dynamic programming. This differentiated replenishment model aims to minimize inventory costs and out-of-stock losses. It inputs replenishment influencing factor data and outputs the optimal replenishment quantity and replenishment time for each product. C7: Generate a replenishment plan based on the calculation results of the differential replenishment model, combined with the actual inventory situation and the required remaining quantity of the goods. The replenishment plan includes a replenishment product list, replenishment quantity, replenishment time, and supplier selection. C8: Monitor the remaining quantity of goods and changes in inventory data in real time, and adjust the replenishment plan.

2. The method for implementing an unmanned intelligent supermarket system according to claim 1, characterized in that: The specific steps of A4 include: A4.1: Extract inventory data for each product at different time points over a month from the unmanned retail store system, and collect customer behavior data within the store to form retail business data. This behavior data includes the time customers spend at the shelf, and the records of when they pick up and put down items. A4.2: Convert the retail business data into a transaction dataset, scan the transaction dataset, count the number of occurrences of each individual item, and generate frequent 1-itemsets. Each transaction in the transaction dataset represents an order and includes all items purchased in the order. The frequent 1-itemsets are sets of individual items. A4.3: Based on the frequent 1-itemset, generate a candidate 2-itemset through a concatenation operation; the candidate 2-itemset is a product combination consisting of two individual items; A4.4: Perform a second scan of the transaction dataset for candidate 2-itemsets and calculate the support of each candidate 2-itemset by calculating the ratio of the number of times any two items A and B appear in orders at the same time to the total number of orders. ; A4.5: Delete candidate 2-itemsets whose support is less than the preset minimum support threshold to obtain frequent 2-itemsets; A4.6: Based on frequent 2-itemsets, iteratively generate frequent N-itemsets by performing concatenation and pruning operations. A4.7: For each frequent item set generated, calculate The ratio of the support of item A to the confidence of all possible association rules is obtained ,in, Indicates an association relationship; A4.8: By calculation The ratio of the support of item B to the support of item B is used to obtain the improvement of all possible association rules. ; A4.9: Filter out the product association rules that meet the preset minimum support threshold and minimum confidence threshold to obtain the preliminary association relationship between the products.

3. The method for implementing an unmanned intelligent supermarket system according to claim 2, characterized in that: The specific steps of A4 also include: A4.10: Based on the relationships between products and the remaining quantity of each product in the basic attributes, the sales attributes of each product are extracted. The products are then clustered according to their sales attributes and the preset number of clusters to form product cluster groups. The sales attributes include quantity and frequency. A4.11: Create a graph with the grid nodes corresponding to each product as vertices. Each vertex contains basic product information. Create an empty graph structure and, based on the product clustering results, add all product nodes that have undergone cluster analysis to the graph to complete the initial construction of the graph structure. The basic information includes product type and current inventory quantity. A4.12: Build edges between nodes based on the association rules and association levels between items. The association levels are measured based on the confidence level in the association rules, including strong and weak associations. If the confidence of the association rule of the product node is greater than the preset maximum confidence value, it means there is a strong association, and the confidence value is used as the weight of the edge to establish a strong connection edge; If the confidence of the association rule of the product node is greater than or equal to the preset minimum confidence value and less than or equal to the preset maximum confidence value, then the connection is made through the intermediate node and the weight of the edge is calculated based on the association transitivity; If the confidence of the association rule of the product node is less than the preset minimum confidence value, it indicates that there is a weak association, and the node related to the product and all its connecting edges are deleted, and the association information of the product is removed from the graph structure to obtain the constructed grid node graph; A4.13: Visualize the constructed grid node graph.

4. The method for implementing an unmanned intelligent supermarket system according to claim 3, characterized in that: The specific steps of A4.12 include: A4.12.1: Traverse the selected product association rules, for each rule , find the corresponding vertices a and b in the graph structure according to the unique identifier of the product; If both vertices exist, an edge is constructed between them; A4.12.2: For constructed edges, evaluate the degree of association based on the confidence metric in the association rule and set the maximum and minimum confidence values. If the confidence of the association rule is greater than the maximum confidence value, the confidence value is used as the weight of the edge to establish a strong connection edge; If the confidence of the association rule is greater than or equal to the minimum confidence value and less than or equal to the maximum confidence value, then find the vertex d in the graph structure that corresponds to the intermediate product D that has a strong association with any two products A and B, build edges between a and d, and d and b respectively, and calculate according to the steps in A4.7 and , by calculating and The product of gets the weight of the corresponding edge; If the confidence of the association rule is less than the minimum confidence value, the node related to the product and all its connecting edges are deleted, and the association information of the product is removed from the graph structure.

5. The method for implementing an unmanned intelligent supermarket system according to claim 4, characterized in that: The specific steps of constructing a product storage map and forming a product association network through partition connection include: B1: Sensors on shelves are used to obtain real-time information about the product category and remaining quantity at each grid node. Furthermore, environmental sensors, combined with meteorological data and a known database of product characteristics, determine the seasonality of the products at each grid node. B2: Integrate the locations and relationships of nodes in the grid node graph with the real-time data collected by sensors on the shelves and environmental sensors to form a comprehensive product dataset; B3: Based on the product's historical remaining quantity data and inventory management strategy, the remaining quantity of the product is divided into M intervals, and each remaining quantity interval is used as a new node. Each new node contains information about the product type, remaining quantity interval range, and seasonal status attributes. B4: Based on the comprehensive product dataset, re-evaluate the correlation between products according to steps A4.2-A4.

9. Based on the degree of correlation between products, construct edges in a graph with product remaining quantity intervals as nodes. B5: Build a product storage graph using the remaining quantity intervals of products as nodes and the associated edges between products as connections. During the graph construction process, the node attribute information is updated in real time. B6: Divide the product storage map into Q partitions based on product categories and storage areas, analyze the product associations between different partitions, and establish connecting edges between the partitions; the weights of the connecting edges are determined based on the degree of association between the products in the partitions; B7: Through partition connection, the product storage maps of each partition are integrated together to form a complete product association network.

6. The method for implementing an unmanned intelligent supermarket system according to claim 5, characterized in that: Each grid node corresponds to a shelf location and records the type of goods and the remaining quantity of goods at the shelf location; the equipment operation data includes the working status, operating time, and number of failures of the equipment.

7. An unmanned intelligent supermarket system, which is used to implement the method for implementing an unmanned intelligent supermarket system according to any one of claims 1 to 6, characterized in that: include: Grid node construction module, information acquisition and judgment module, product storage map construction module, product strategy formulation module, and fault prediction module work together to implement unattended functions; The grid node construction module is used to reasonably divide the supermarket area of ​​the unmanned retail store and construct a grid node map frame; The information acquisition and judgment module is used to obtain product information in real time and judge the seasonality of the product; The product storage map construction module is used to construct a product storage map and a product association network; The product strategy formulation module is used to formulate product migration strategies and dynamic replenishment strategies based on the product association network; The fault prediction module is used to collect equipment operation data in real time, predict equipment failures and issue early warnings.

8. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the processor implements the steps of the method for implementing an unmanned intelligent supermarket system as described in any one of claims 1 to 6.

Images