An automatic vending system based on pressure sensors

Abstract

The application discloses an automatic vending system based on a pressure sensor, and relates to the technical field of Internet of Things.The application comprises the following steps: deploying a flexible pressure sensor array on a shelf and connecting a control unit; the control unit identifies the taking and placing actions of goods based on the changes of the flexible pressure sensor array; the control unit identifies the names, quantities and taking and placing time of the goods through a dynamic pressure distribution mapping technology; the control unit automatically calculates the total price of the goods according to the identification results of the goods; the control unit updates the inventory record after a customer confirms and completes payment; the control unit predicts the replenishment demand based on historical sales data and automatically generates a replenishment list; the application utilizes the flexible pressure sensor array and advanced data analysis technology to realize accurate identification and automatic vending management of goods, significantly improves the shopping experience of customers, optimizes inventory control, effectively reduces operating costs and promotes the transformation and upgrading of the retail industry towards intelligence and unmanned direction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Internet of Things, and particularly relates to an automatic vending system based on a pressure sensor. BACKGROUND

[0002] In the field of automatic vending systems, with the rapid development of Internet of Things technology, traditional mechanical automatic vending machines are gradually transforming towards intelligence, aiming to improve user experience and operational efficiency. At present, the common automatic vending machines on the market mostly rely on RFID (Radio Frequency Identification), visual recognition or weighing sensor technology to monitor the access situation of goods. However, these methods each have certain limitations. The cost of RFID tags and their potential pollution problem to the environment cannot be ignored; the visual recognition technology can provide rich information, but its accuracy and processing speed are greatly affected by light conditions, obstructions and the complexity of the goods packaging; and the traditional weighing sensor can only detect weight changes and cannot accurately identify specific goods categories and quantities, which is not satisfactory in a vending environment with a large variety of goods. SUMMARY

[0003] In view of the above existing problems, the present application is proposed.

[0004] Therefore, the present application provides an automatic vending system based on a pressure sensor, which solves the limitations of automatic vending systems based on traditional visual and RFID technologies in terms of goods identification accuracy, real-time performance and cost-effectiveness, and improves the accuracy of inventory management and restocking prediction.

[0005] To solve the above technical problems, the present application provides the following technical solutions:

[0006] In a first aspect, the present application embodiment provides an automatic vending system based on a pressure sensor, which comprises a sensor deployment module for deploying a flexible pressure sensor array on a shelf and connecting a control unit;

[0007] A pick-and-place action recognition module is used to identify the pick-and-place action of goods based on the changes of the flexible pressure sensor array by the control unit;

[0008] A goods identification module is used to identify the name, quantity and pick-and-place time of goods through a dynamic pressure distribution mapping technology;

[0009] A price calculation module is used to automatically calculate the total price of goods according to the goods identification result;

[0010] An inventory management module is used to update the inventory record by the control unit after the customer confirms and completes payment;

[0011] An intelligent restocking prediction module is used to predict the restocking demand based on historical sales data and automatically generate a restocking list.

[0012] As a preferred solution of the automatic vending system based on pressure sensor, wherein: the adaptive pressure baseline refers to the state without taking and placing actions on the shelf, the control unit continuously reads and records the pressure values at multiple time points, uses an algorithm to remove outliers, calculates the adaptive pressure baseline B(t), and stores it in the control unit for pressure change comparison analysis;

[0013] The adaptive pressure baseline B(t) is represented as,

[0014] B(t) = μ + φ(P(t-1)-B(t-1)) + θ∈(t-1) + ∈(t);

[0015] Wherein, μ represents the intercept term, φ represents the autoregressive coefficient, θ represents the moving average coefficient, ∈(t) represents the white noise error term, P(t-1) represents the pressure value at the previous time, B(t-1) represents the baseline value at the previous time, and t represents the time.

[0016] As a preferred solution of the automatic vending system based on pressure sensor, wherein: the control unit identifies the taking and placing actions of the goods based on the changes of the flexible pressure sensor array, and the specific steps are as follows:

[0017] Based on the weight of the goods, set the dynamic change threshold B d (t), represented as,

[0018]

[0019] Wherein, B d (t) represents the dynamic change threshold at time point t, α represents the adjustment parameter of the average weight of the goods to the dynamic change threshold, W a represents the average weight of the goods on the shelf, β represents the adjustment parameter of the historical pressure change volatility to the dynamic change threshold, m represents the number of historical data points, w k represents the weight of the kth historical time point, k represents the historical time point, P(t-k) represents the pressure value at time point t-k, and B(t-k) represents the baseline value at time point t-k.

[0020] The control unit receives continuous data stream from the flexible pressure sensor array, and processes the sensor data in real time;

[0021] Compare the difference between the processed sensor pressure value P(t) and the adaptive pressure baseline B(t) with the dynamic change threshold B d (t), and identify the taking and placing actions;

[0022] When |P(t)-B(t)|>B d(t) means that the pick-and-place action occurs and is marked;

[0023] When |P(t)-B(t)|≤B d (t) means that there is no pick-and-place action.

[0024] As a preferred solution of the automatic vending system based on pressure sensor according to the present application, wherein: a product pressure feature library is constructed, and the specific steps are as follows:

[0025] List all the planned products, and assign a unique code to each product;

[0026] The control unit records the pressure change data of the sensor before and after the product is placed;

[0027] Use the scikit-learn library in Python for feature engineering and machine learning, and extract the feature vector f i from the pressure change data;

[0028] Arrange the feature vector of each product into the library and associate it with the product code to form a product pressure feature library K, wherein K contains the jth feature vector f ij of the product and the product code c i .

[0029] As a preferred solution of the automatic vending system based on pressure sensor according to the present application, wherein: through dynamic pressure distribution mapping technology, the product name is identified, and the specific steps are as follows:

[0030] The control unit receives the real-time data stream P(t) from the sensor, calculates the dynamic pressure distribution D(t) using P(t) and the adaptive pressure baseline B(t), which is represented as,

[0031] D(t) = P(t)-B(t);

[0032] Based on the dynamic pressure distribution D(t) and the product pressure feature library K, a product recognition function is designed, which is represented as,

[0033]

[0034] Wherein, Ω(D(t), K) represents the product recognition function, and the output is the product code d j (t) represents the jth feature extracted from D(t), α j represents the jth feature weight, and β j represents the similarity decay rate;

[0035] Based on the recognized product code c i , the name of the product taken is determined.

[0036] As a preferred scheme of the automatic vending system based on pressure sensor, the specific steps of identifying the number of goods and the taking and placing time are as follows:

[0037] D(t) is divided into a plurality of local regions, for each divided region, a local pressure maximum value is detected, the distribution characteristics around each local maximum value are analyzed, and the number of goods in each region is calculated according to the pressure distribution characteristics of each region and the matching result with the goods feature library;

[0038] Based on the labeled taking and placing action, the time series analysis method is used to determine when the goods are taken out by detecting the mutation point in the pressure signal, and the taking and placing time is identified;

[0039] If the goods identification fails, the goods pressure feature library is re-identified.

[0040] As a preferred scheme of the automatic vending system based on pressure sensor, the specific steps of automatically calculating the total price of goods according to the goods identification result are as follows:

[0041] The name of the goods is used as a keyword to search for the corresponding goods information in the goods database;

[0042] If the goods information search fails, the goods database is re-searched;

[0043] Whenever a piece of goods is successfully identified and confirmed to be added to the shopping cart, the price of the goods is immediately read from the goods database and added to the current total amount of the shopping cart, and the total price of the goods is automatically calculated.

[0044] As a preferred scheme of the automatic vending system based on pressure sensor, the specific steps of the customer confirming the goods and the total price are as follows:

[0045] The list of goods purchased by the customer and the cumulative total price are displayed in real time through an electronic display screen;

[0046] If the customer puts the goods back on the shelf, the total price of the goods is immediately updated, and the control unit updates the goods information;

[0047] If the customer adds the type and quantity of goods, the displayed total price is immediately updated, so that the customer can know the status of the shopping cart at any time;

[0048] The encryption technology is used to protect the shopping records and personal information of the customer;

[0049] After the customer confirms that the goods and the total price are correct, the customer is provided with a variety of payment methods to choose from, and the payment is completed.

[0050] As a preferred solution of the automatic vending system based on pressure sensors, after the customer completes the payment, the control unit updates the inventory record, and the specific steps are as follows:

[0051] After the payment is completed, the system communicates with the payment platform through the API interface to verify whether the payment is successful;

[0052] When the payment is not completed, the control unit prompts the customer to confirm the payment information again and re-pays;

[0053] The control unit starts the inventory adjustment program immediately after receiving the confirmation signal of successful payment to update the inventory record.

[0054] As a preferred solution of the automatic vending system based on pressure sensors, based on historical sales data, the demand for replenishment is predicted, and a replenishment list is automatically generated, and the specific steps are as follows:

[0055] Analyze historical sales data to identify hot-selling goods;

[0056] Use a time series model to predict future sales based on past sales data of hot-selling goods;

[0057] Real-time monitoring of inventory levels to identify goods below the safety stock;

[0058] Based on the predicted sales and current inventory status, calculate the replenishment demand of each product;

[0059] Automatically generate a replenishment list listing the ID, predicted demand, recommended replenishment amount, and priority of each product.

[0060] The present application has the following advantages: by deploying a flexible pressure sensor array on the shelf and using innovative dynamic pressure distribution mapping technology, the present application realizes accurate recognition of product picking and placing actions, automatic judgment of product type and quantity, instant total price calculation, real-time update of inventory records, and intelligent replenishment prediction based on historical data analysis. This series of steps not only significantly improves the shopping experience of customers, reduces queuing time, ensures sufficient goods, but also greatly optimizes the inventory management of retailers, reduces operating costs, improves overall operational efficiency and supply chain response speed, and ultimately achieves a highly intelligent, efficient, and low-cost retail environment, creating a win-win situation for customers and businesses. BRIEF DESCRIPTION OF DRAWINGS

[0061] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0062] Figure 1 Flowchart for the pressure sensor-based vending system in Example 1.

[0063] Figure 2 Logic diagram for customer automatic purchase in Example 1. DETAILED DESCRIPTION

[0064] In order to make the above objectives, features and advantages of the present application more apparent, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0065] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details given herein. In other instances, well-known methods have not been described in detail in order to avoid unnecessarily obscuring the present application. Therefore, the specific embodiments of the present application as set forth are not intended as being exhaustive or limiting of the present application.

[0066] Secondly, the "one embodiment" or "an embodiment" referred to herein means that a specific feature, structure, or characteristic described can be included in at least one implementation of the present application. The "in one embodiment" appearing in various places in the specification does not all refer to the same embodiment, nor is it necessarily a separate or alternative implementation of other embodiments.

[0067] Example 1, with reference to Figure 1 and Figure 2 , is the first embodiment of the present application, which provides a pressure sensor-based vending system, including the following steps:

[0068] S1, deploy a flexible pressure sensor array on the shelf and connect a control unit, the specific steps are as follows:

[0069] Select a flexible pressure sensor with high sensitivity and wide range to ensure accurate capture of weight changes of goods, while having good durability and environmental adaptability;

[0070] According to the size of the shelf and the density of the goods, design the layout of the sensor array to ensure that each goods placement area is covered with enough sensors to achieve accurate pressure change detection;

[0071] Firmly install the sensor array on the shelf using special fixing materials or structures to ensure that the sensors are not easily displaced or damaged in daily use;

[0072] Use special signal cables or wireless modules to connect the sensor array with the central control unit to ensure the stability and security of data transmission;

[0073] The communication parameters of the sensor array, such as baud rate, address, data format, etc., are configured on the control unit to ensure that the sensor data can be correctly parsed and processed.

[0074] It should be noted that by deploying a flexible pressure sensor array on the shelf, the state of the goods on the shelf can be monitored in real time, including the taking and placing actions of the goods. This deployment enables the system to perceive the slight movement of the goods, thereby timely capturing the purchase behavior of the customers, recording the dynamic changes of the goods without manual intervention, improving the accuracy of goods tracking, reducing human errors, and also providing basic data support for subsequent goods identification and inventory management.

[0075] S2, the control unit identifies the taking and placing actions of the goods based on the changes of the flexible pressure sensor array, and the specific steps are as follows:

[0076] S2.1, the adaptive pressure baseline refers to the state of no taking and placing action on the shelf, the control unit continuously reads and records the pressure values at multiple time points, uses an algorithm to remove outliers, calculates the adaptive pressure baseline B(t), and stores it in the control unit for pressure change comparison and analysis;

[0077] The adaptive pressure baseline B(t) is represented as,

[0078] B(t) = μ + φ(P(t-1) - B(t-1)) + θ ∈(t-1) + ∈(t);

[0079] Where μ represents the intercept term, which represents the standard value of the pressure baseline without external influence. This value usually reflects the average output of the sensor in the empty or stable state of the shelf, which can be understood as the static bias of the sensor. φ represents the autoregressive coefficient, which is used to capture the autocorrelation of time series data. θ represents the moving average coefficient, which is used to reflect the influence of past observation errors on the current baseline. ∈(t) represents the white noise error term, which represents the unpredictable instantaneous disturbance or measurement error. P(t-1) represents the pressure value at the previous time, B(t-1) represents the baseline value at the previous time, and t represents the time.

[0080] It should be noted that the adaptive pressure baseline B(t) is designed to dynamically adjust and calibrate the zero reference of the pressure sensor in a changing environment, thereby eliminating the influence of environmental factors (such as temperature changes, sensor drift, etc.) and long-term trends on measurement accuracy, ensuring that the system can accurately identify the subtle pressure changes caused by the taking and placing actions of the goods.

[0081] S2.2, based on the weight of the goods, set a dynamic change threshold B d (t), represented as,

[0082] B d (t) = αWa + βσ(t) ;

[0083] where B d (t) represents the dynamic change threshold at time point t, used to identify whether the pressure change caused by the picking or placing action of the goods on the shelf is significant enough to trigger the system response, a represents the adjustment parameter of the average weight of the goods to the dynamic change threshold, W a represents the average weight of the goods on the shelf, reflecting the general weight level of the goods on the shelf, and β represents the adjustment parameter of the historical pressure change volatility to the dynamic change threshold, by adjusting β, the importance of the historical data volatility in the threshold calculation can be controlled, and σ(t) represents the weighted standard deviation of the historical pressure change;

[0084]

[0085] The integration obtains the complete dynamic change threshold B d (t), which is represented as,

[0086]

[0087] where m represents the number of historical data points, w k represents the weight of the kth historical time point, k represents the historical time point, P(t-k) represents the pressure value at time point t-k, i.e. the pressure reading k time units ago, and B(t-k) represents the baseline value at time point t-k, i.e. the baseline reading k time units ago.

[0088] It should be noted that the dynamic change threshold B d (t) is designed to ensure that the system can not only sensitively detect significant pressure changes caused by picking or placing actions of goods on the shelf, but also exclude minor environmental disturbances or false positives, so as to accurately trigger the system response, while taking into account the average weight of the goods and the historical pressure change volatility.

[0089] S2.3, the control unit receives a continuous data stream from the flexible pressure sensor array, and processes the data of the sensor in real time;

[0090] The difference between the processed pressure value P(t) of the sensor and the adaptive pressure baseline B(t) is compared with the dynamic change threshold B d (t), and the picking or placing action is identified;

[0091] When |P(t)-B(t)|>B d (t), it means that the picking or placing action occurs, and the picking or placing action is marked;

[0092] When |P(t)-B(t)|≤B d (t), it means that there is no picking or placing action.

[0093] Further, the mutation point in the sensor data can also be identified by a differential algorithm, representing the taking and placing action of the commodity; a positioning algorithm is applied to determine the specific location of the change based on the pressure distribution characteristics of the change point;

[0094] The detailed data collection mode is started, and the continuous pressure data in a period of time before and after the change is recorded, including timestamp, pressure value, change rate, etc., for subsequent analysis. If an abnormally large pressure change is detected, it may be a system failure or abnormal operation, which should be immediately recorded and marked.

[0095] It should be noted that the control unit identifies the customer's taking and placing behavior of the commodity by analyzing the change pattern of the sensor data. This identification process is the core of the automated vending system, which can distinguish different operations of customers, lay a foundation for subsequent commodity name, quantity identification and price calculation, ensure the accuracy and fairness of the transaction, and reduce the waiting time of customers and improve the shopping experience.

[0096] S3, identify the commodity name, quantity and taking and placing time through dynamic pressure distribution mapping technology, the specific steps are as follows:

[0097] S3.1, list all the planned commodities, and assign a unique code to each commodity;

[0098] The control unit records the pressure change data of the sensor before and after the commodity is placed;

[0099] Use the scikit-learn library in Python for feature engineering and machine learning to extract feature vectors f i from pressure change data;

[0100] Organize the feature vectors of each commodity into a database, associate them with the commodity code, and form a commodity pressure feature library K, where K contains the jth feature vector f ij of the commodity and the commodity code c i .

[0101] S3.2, the control unit receives real-time data stream P(t) from the sensor, uses P(t) and adaptive pressure baseline B(t) to calculate dynamic pressure distribution D(t), represented as,

[0102] D(t) = P(t) - B(t);

[0103] Based on the dynamic pressure distribution D(t) and the commodity pressure feature library K, design a commodity recognition function, represented as,

[0104]

[0105] where Ω(D(t), K) represents the product recognition function, outputting the product code, d j (t) represents the jth feature extracted from D(t), α j represents the feature weight, which is a non-negative numerical value, indicating the importance of the jth feature in the recognition process, β j represents the similarity decay rate, controlling the influence of feature value difference on similarity score;

[0106] It should be noted that the basis for designing the product recognition function is to efficiently and accurately identify the product from the dynamic pressure distribution D(t), by comparing the real-time pressure features with the patterns in the product feature library K, using the feature weight α j and the similarity decay rate β j to adjust the relative importance and difference sensitivity of different features, thereby accurately matching the current taken product name among numerous products.

[0107] Based on the recognized product code c i , the taken product name is determined.

[0108] S3.3, divide D(t) into multiple local regions, which can be defined based on spatial location or continuity of pressure change, the purpose of segmentation is to identify the spatial range occupied by a single product, for each segmented region, detect the local pressure maximum, which may correspond to the center or main stress point of the product, analyze the distribution characteristics around each local maximum, such as the shape, area, and intensity of the pressure distribution, according to the pressure distribution characteristics of each region and the matching results with the product feature library, calculate the number of products in the region;

[0109] Introduce error correction mechanisms such as historical data verification, multi-round measurement averaging, etc., to improve the accuracy of quantity calculation;

[0110] Based on the labeled taking and placing actions, use time series analysis method to determine when the product is taken out by detecting the mutation points in the pressure signal, and identify the taking and placing time;

[0111] If the product recognition fails, re-identify in the product pressure feature library.

[0112] It should be noted that the role of dynamic pressure distribution mapping technology in this step is to analyze the real-time data of the flexible pressure sensor array on the shelf, and then identify the type, quantity and taking and placing time of the product, in order to realize automatic product tracking and management.

[0113] It should be noted that by using dynamic pressure distribution mapping technology, the system can analyze the pressure distribution characteristics of different goods on the shelf, and then identify the types and quantities of goods. This technology can automatically distinguish different goods, even if they are similar in appearance, and can identify them through subtle differences in pressure distribution, greatly improving the accuracy of goods identification and reducing disputes caused by goods identification errors. At the same time, it provides a basis for subsequent inventory management and replenishment decisions.

[0114] S4, according to the identification result of the article, automatically calculate the total price of the goods, the specific steps are as follows:

[0115] Use the name of the goods as a keyword to search for the corresponding goods information in the goods database, including unit price, specification, and brand;

[0116] If the goods information search fails, re-search in the goods database;

[0117] Whenever a piece of goods is successfully identified and confirmed to be added to the shopping cart, immediately read the price of the goods from the goods database and add it to the current shopping cart total amount, automatically calculate the total price of the goods.

[0118] It should be noted that the system automatically retrieves the price of the goods from the database based on the identification result of the goods, and calculates the total price of the goods selected by the customer. This process realizes unmanned settlement, and the customer can immediately see the total price of the selected goods, which is convenient and fast, speeds up the settlement process, reduces the waiting time in line, and improves customer satisfaction and store operation efficiency.

[0119] S5, after the customer confirms and completes the payment, the control unit updates the inventory record, the specific steps are as follows:

[0120] Display the list of goods purchased by the customer and the cumulative total price in real time through the electronic display screen;

[0121] If the customer puts the goods back on the shelf, immediately update the total price of the goods, and the control unit will update the goods information;

[0122] If the customer adds the type and quantity of goods, immediately update the displayed total price, so that the customer can know the status of the shopping cart at any time;

[0123] If there are goods that cannot be identified, provide feedback to the customer in time and guide them to take the correct operation;

[0124] Use encryption technology to protect the shopping records and personal information of customers, and ensure the safety of all transaction data;

[0125] After the customer confirms that the goods and the total price are correct, provide multiple payment methods for the customer to choose from to ensure a safe and fast payment process and complete the payment.

[0126] After payment is completed, the system communicates with the payment platform through the API interface to verify whether the payment is successful;

[0127] When the payment is not completed, the control unit reminds the customer to confirm the payment information again and re-pays;

[0128] The control unit starts the inventory adjustment program immediately after receiving the confirmation signal of successful payment, and updates the inventory record;

[0129] Specifically, the detailed information of the sold goods is extracted from the shopping cart record, and the corresponding quantity of goods is deducted from the inventory database according to the goods information;

[0130] If the inventory of the goods falls below the minimum threshold A, the system automatically triggers the early warning mechanism to remind the replenishment and stop selling;

[0131] Save all details of the transaction, including product type, quantity, unit price, total price, purchase time, payment method, customer ID, etc. Encrypt the transaction data and back it up to the cloud or local server regularly to ensure data security and recoverability. Create a transaction log to record the complete life cycle of each transaction, including creation, modification and closed state, for auditing and tracking purposes.

[0132] It should be noted that at the same time when the customer completes the payment, the system automatically updates the inventory database to record the sales of the goods, ensuring the real-time and accuracy of the inventory data, providing decision basis for subsequent inventory management, avoiding overstock or out-of-stock situations, optimizing inventory turnover rate and reducing operating costs.

[0133] S6, based on historical sales data, predict replenishment demand, and automatically generate a replenishment list, the specific steps are as follows:

[0134] Analyze historical sales data to identify best-selling goods, seasonal sales trends, holiday effects and other patterns;

[0135] Use time series models to predict future sales based on past best-selling goods sales data;

[0136] Use machine learning algorithms to combine external factors (weather, economic indicators, social media trends, etc.) for prediction, use cross-validation to evaluate model performance, and select the best model;

[0137] Simulate different market events to test their impact on predictions;

[0138] Real-time monitoring of inventory levels to identify goods below safe inventory levels, taking into account in-transit inventory and goods ordered but not yet received, to avoid duplicate orders;

[0139] Based on the predicted sales volume and the current inventory status, the replenishment demand of each commodity is calculated, and the safety stock principle is applied to ensure that there is no shortage due to unexpected demand increases;

[0140] An automatic replenishment list is generated, listing the ID, predicted demand, recommended replenishment amount and priority of each commodity;

[0141] Further, the replenishment list can also be manually reviewed to adjust the replenishment amount and priority according to the actual situation, optimizing the overall inventory strategy.

[0142] It should be noted that the system analyzes historical sales data to predict future sales trends over a period of time, automatically generating a replenishment list. This prediction mechanism helps businesses prepare in advance to ensure adequate supply of popular commodities, reducing potential sales losses due to stockouts, while also avoiding excessive inventory and capital occupation problems, improving the overall efficiency of the supply chain.

[0143] In summary, the present application realizes accurate recognition of commodity picking and placing actions, automatic judgment of type and quantity, instant total price calculation, real-time update of inventory records, and intelligent replenishment prediction based on historical data analysis. This series of steps not only significantly improves the shopping experience of customers, reduces queuing time, and ensures adequate supply of commodities, but also greatly optimizes the inventory management of retailers, reduces operating costs, improves overall operational efficiency and supply chain response speed, ultimately achieving a highly intelligent, efficient and low-cost retail environment, creating a win-win situation for customers and businesses.

[0144] Embodiment 2

[0145] Referring to Table 1, the second embodiment of the present application is given, and experimental simulation data of the automatic vending system based on pressure sensors is given to further verify the advancement of the present application.

[0146] To verify the performance of the automatic vending system based on pressure sensors, the experiment was conducted in a simulated supermarket environment. The environment is equipped with shelves equipped with flexible pressure sensor arrays that can monitor the picking and placing actions of commodities. The experiment is divided into several stages: preparation stage, running stage and data analysis stage.

[0147] The preparation stage involves installing sensors, setting up control units, defining adaptive pressure baseline algorithms and dynamic change threshold algorithms. In addition, a commodity pressure feature library is created, containing the coding identification of all commodities and their corresponding pressure change feature vectors. To ensure the accuracy and reliability of the data, standardized commodities were used in the experiment, including commodities of different weights and volumes.

[0148] During the operation phase, the control unit continuously receives data streams from the sensors, processes the data in real-time, and identifies the pick-and-place actions based on the pre-set algorithms. Simultaneously, the system automatically calculates the total price of the goods, updates the inventory records, and adjusts the inventory after the customer completes the payment. During the experiment, a large amount of data about the pick-and-place of goods, sales trends, inventory levels, and restocking needs were collected.

[0149] During the data analysis phase, the experimental data were compared with those of the traditional manual selling system to quantify the efficiency and accuracy of the new system. The data collection period was one month, covering peak sales periods on weekdays and weekends.

[0150] Specifically, as shown in Table 1:

[0151] Table 1: Experimental Record Table

[0152]

[0153] By analyzing the above table data, it can be observed that the automatic selling system based on pressure sensors exhibits significant advantages in product identification, inventory management, and restocking prediction. For example, the average pick-and-place frequency of apples is 1200 times, while the automatic identification success rate is as high as 98.5%, indicating that even in a high-traffic environment, the system can maintain high precision. The inventory update delay is between 0.1 and 0.3 seconds, much lower than the manual update time of traditional systems, improving operational efficiency.

[0154] The restocking prediction accuracy is also very good, averaging more than 96%, which means that the system can effectively avoid inventory accumulation and out-of-stock situations, reduce waste, and improve customer satisfaction. Compared with the traditional manual selling system, the automatic selling system has obvious improvements in identification speed, accuracy, real-time inventory updating, and prediction ability, reflecting its innovation and practicality.

[0155] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit it. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, which should be covered by the claims of the present application.

Claims

1. An automated vending system based on a pressure sensor, characterized in that: include, A sensor deployment module for deploying a flexible pressure sensor array on a shelf and connecting it to a control unit; The pick-and-place action recognition module is used in the control unit to identify the pick-and-place actions of goods based on changes in the flexible pressure sensor array. The specific steps are as follows: When there is no picking or placing on the shelf, the control unit continuously reads and records the pressure values ​​at multiple time points, uses an algorithm to remove outliers, and calculates an adaptive pressure baseline. The data is stored in the control unit for comparative analysis of pressure changes. Adaptive pressure baseline , is represented as , ； in, Represents the intercept term. Represents the autoregressive coefficient. Represents the moving average coefficient. This represents the white noise error term. This indicates the pressure value at the previous time. This represents the baseline value from the previous time period. Indicates time; Set dynamic change thresholds based on product weight. , is represented as , ； in, Indicates a point in time The dynamic change threshold, This parameter represents the adjustment parameter for the average weight of goods relative to a dynamic change threshold. This indicates the average weight of the goods on the shelf. This parameter represents the adjustment parameter of historical pressure variation volatility to the dynamic change threshold. This indicates the number of historical data points. Indicates the first The weight of each historical time point Indicates a historical point in time. Indicates at a point in time pressure value Indicates at a point in time The baseline value; The control unit receives a continuous data stream from the flexible pressure sensor array and processes the sensor data in real time. The processed sensor pressure value and adaptive pressure baseline The difference is compared with the dynamically changing threshold Bd(t) to identify the picking and placing actions; when When this occurs, it indicates that a pick-up or put-down action has taken place, and the pick-up or put-down action is marked. when When there is no action of picking up or putting down; The product identification module is used to identify the product name, quantity, and pickup / placement time through dynamic pressure distribution mapping technology; The price calculation module is used to automatically calculate the total price of the goods based on the product recognition results; The inventory management module is used to control the system to update inventory records after the customer confirms and completes payment; The intelligent replenishment forecasting module is used to predict replenishment needs based on historical sales data and automatically generate a replenishment list.

2. The automated vending system based on a pressure sensor as described in claim 1, characterized in that: The specific steps for building a product stress feature library are as follows: List all planned products and assign a unique code to each product; The control unit records pressure change data from sensors before and after the product is placed. Using the scikit-learn library in Python for feature engineering and machine learning, extract feature vector f from stress variation data. i ; The feature vectors of each product are organized and stored in a database, and associated with the product code to form a product stress feature database K, where K contains the j-th feature vector f of the product. ij and product code c i .

3. The automated vending system based on a pressure sensor as described in claim 2, characterized in that: The product name is identified using dynamic pressure distribution mapping technology. The specific steps are as follows: The control unit receives a real-time data stream P(t) from the sensor, and uses P(t) and the adaptive pressure baseline B(t) to calculate the dynamic pressure distribution D(t), denoted as... ； Based on the dynamic pressure distribution D(t) and the commodity pressure feature library K, a commodity identification function is designed, expressed as follows: ； in, This represents a product identification function, whose output is the product code, d. j (t) represents the j-th feature extracted from D(t), β represents the feature weights. j Indicates the similarity decay rate; Based on the identified product code c i Determine the name of the product to be taken.

4. The automated vending system based on a pressure sensor as described in claim 3, characterized in that: The specific steps for identifying the quantity and pickup / placement time of goods are as follows: D(t) is divided into multiple local regions. For each region, the local pressure maximum value is detected, and the distribution characteristics around each local maximum value are analyzed. Based on the pressure distribution characteristics of each region and the matching results with the product feature library, the number of products in the region is calculated. Based on the tag-based pick-up and put-down action, time series analysis is used to determine when the goods are taken out by detecting abrupt changes in the pressure signal, thus identifying the pick-up and put-down time. If product identification fails, it will be re-identified in the product stress feature library.

5. The automated vending system based on a pressure sensor as described in claim 4, characterized in that: Based on the product recognition results, the total price of the products will be automatically calculated. The specific steps are as follows: Use the product name as a keyword to search for relevant product information in the product database; If the product information search fails, the search will be repeated in the product database; Whenever an item is successfully identified and added to the shopping cart, its price is immediately retrieved from the product database and added to the current total amount in the shopping cart, automatically calculating the total price of the item.

6. The automated vending system based on a pressure sensor as described in claim 5, characterized in that: The customer confirms the goods and the total price. The specific steps are as follows: The electronic display screen shows the list of items purchased by the customer and the total price in real time. If a customer puts an item back on the shelf, the total price of the item will be updated immediately, and the control unit will update the item information. If a customer adds product types and quantities, the displayed total price will be updated immediately, allowing the customer to keep track of the shopping cart status at any time. Encryption technology is used to protect customers' shopping records and personal information; After the customer confirms that the goods and total price are correct, multiple payment methods are offered for the customer to choose from, and the payment is completed.

7. The automated vending system based on a pressure sensor as described in claim 6, characterized in that: After the customer completes payment, the control unit updates the inventory record. The specific steps are as follows: After payment is completed, the system communicates with the payment platform via API to verify whether the payment status is successful. If the payment verification is not completed, the control unit will remind the customer to confirm the payment information again and pay again. Upon receiving a confirmation signal of successful payment, the control unit immediately initiates the inventory adjustment procedure to update the inventory records.

8. The automated vending system based on a pressure sensor as described in claim 7, characterized in that: Based on historical sales data, replenishment needs are predicted, and a replenishment list is automatically generated. The specific steps are as follows: Analyze historical sales data to identify best-selling products; Using time series models, predict future sales volume based on past sales data of best-selling products; Monitor inventory levels in real time and identify items that are below safety stock levels; Calculate the replenishment requirement for each product based on the predicted sales volume and current inventory status. Automatically generate a replenishment list, listing the ID, predicted demand, suggested replenishment quantity, and priority of each product.

Images