Path optimization method and device for picking goods in unmanned warehouse

Abstract

The invention discloses a path optimization method and device for picking goods in an unmanned warehouse, and relates to the technical field of warehousing path planning. Determining a self-service store according to the target order; after it is determined that the target deliveryman arrives at the self-service store, if the identity verification information uploaded by the target deliveryman is verified to be correct, an opening instruction is sent to the self-service store, and a commodity type list is extracted from the target order; performing path planning based on the current position of the target deliveryman and the plurality of placement positions to obtain a path candidate set; candidate paths in the path candidate set are removed according to the total demand pressure value, and a path optimization set is obtained; selecting an optimized path with the minimum dynamic congestion score from the path optimization set as a target path; and sending the target path to the target deliveryman, so that the target deliveryman sorts and packages the commodities in the target order according to the target path, and delivers the packaged commodities. According to the technical scheme, the problem that multiple deliverymen interfere with one another when working at the same time is solved.

Description

Technical Field

[0001] This application relates to the field of warehouse route planning technology, specifically to a method and apparatus for optimizing picking routes in an unmanned warehouse. Background Technology

[0002] With the rapid development of new retail and the unmanned economy, on-demand delivery services have become a key link connecting online orders with offline physical inventory. By introducing unmanned vending systems, traditional stores can not only reduce the cost of manual supervision but also provide 24 / 7 merchandise sorting services, enabling delivery personnel to autonomously complete merchandise sorting and delivery tasks, which has become an important means of improving store operational efficiency.

[0003] Currently, stores primarily utilize technologies such as intelligent shelf management systems and electronic tags to support merchandise sorting. These systems provide delivery personnel with product location information and sorting guidance through functions like product positioning and inventory management. However, existing merchandise sorting systems still have significant shortcomings in sorting route planning. Due to the lack of intelligent route planning methods, delivery personnel need to determine sorting routes based on experience. This extensive sorting method is not only inefficient but also prone to causing congestion within the store. Especially during peak order periods, when multiple delivery personnel are working simultaneously, the lack of scientific route planning and scheduling mechanisms easily leads to mutual interference, resulting in low sorting efficiency. Summary of the Invention

[0004] This application provides a method and apparatus for optimizing picking routes in an unmanned warehouse. This method effectively solves the problem of mutual interference when multiple delivery personnel work at the same time, significantly improves sorting efficiency, and reduces the risk of channel congestion.

[0005] Firstly, this application provides a method for optimizing picking routes in an unmanned warehouse. The method includes: obtaining product information from a target order and determining the unmanned vending machine store based on the product information; sending the target order and the address of the unmanned vending machine store to a target delivery person so that the target delivery person can go to the unmanned vending machine store according to the target order; after determining that the target delivery person has arrived at the unmanned vending machine store, obtaining the identity verification information uploaded by the target delivery person; if the identity verification information is correct, sending an opening command to the unmanned vending machine store and extracting a list of product categories from the target order; retrieving the product layout map of the unmanned vending machine store, determining the placement position of each product in the product category list according to the product layout map, obtaining the current position of the target delivery person, performing path planning on the current position and multiple placement positions based on preset rules to obtain a path candidate set; and optimizing the picking route of the unmanned vending machine store within a preset time. The system calculates the demand heat value for each product type in multiple pending orders, including target orders. It then statistically analyzes the target product types corresponding to all channels of each candidate path in the candidate path set, summing the demand heat values ​​across multiple target product types to obtain the total demand pressure value. Each total demand pressure value corresponds to one candidate path. Based on the total demand pressure value, candidate paths in the candidate path set are eliminated to obtain an optimized path set. The system calculates the dynamic congestion score for each optimized path in the optimized path set based on personnel movement trajectories within a preset time window. The optimized path with the minimum dynamic congestion score is selected as the target path. Finally, the target path is sent to the target delivery person so that they can sort and pack the goods in the target order according to the target path and deliver the packed goods.

[0006] By adopting the above technical solutions, the unmanned vending store is opened after obtaining the product information of the target order and verifying identity, effectively ensuring the security of the sorting process. Preliminary path planning based on the product layout map provides basic sorting route guidance for the target delivery personnel, avoiding the blindness of traditional experience-based sorting. By calculating the demand heat value of each product type in the pending orders and combining it with the product types corresponding to the channels in the candidate paths to evaluate the total demand pressure value, intelligent identification and elimination of high-demand-pressure paths are achieved. Simultaneously, by calculating the dynamic congestion score to select the optimal path, the problem of mutual interference when multiple delivery personnel work simultaneously is effectively solved. This not only significantly improves sorting efficiency but also reduces the risk of channel congestion, allowing the target delivery personnel to sort and pack goods according to the planned optimal path, greatly improving sorting accuracy and delivery timeliness.

[0007] Optionally, based on the movement trajectories of people within a preset time window, the dynamic congestion score is calculated for each optimized path in the path optimization set. Specifically, this includes: obtaining the current state of other people in the unmanned vending store at the current moment, including the target position, movement speed, and movement direction; predicting the future movement trajectories of other people within the preset time window based on the movement direction and movement speed in the merchandise layout map, obtaining multiple predicted movement trajectories; calculating the estimated arrival time corresponding to each coordinate point on the target optimized path based on the movement speed, mapping each estimated arrival time to each coordinate point on the target optimized path, obtaining the planned movement trajectory, where the target optimized path is any optimized path in the path optimization set; obtaining the target predicted point and target planned point at the target time point from the target predicted movement trajectory and the planned movement trajectory, calculating the position and direction of the target planned point and the target predicted point, obtaining the instantaneous conflict risk value; and summing the instantaneous conflict risk values ​​in the target optimized path to obtain the dynamic congestion score corresponding to the target optimized path.

[0008] By employing the aforementioned technical solution, the estimated arrival time is mapped onto various coordinate points along the optimized path, forming a complete planned movement trajectory. This allows for accurate assessment of the spatiotemporal distribution of delivery personnel during the sorting process. By calculating the positional and directional relationships between the planned and predicted target points, an instantaneous conflict risk value is obtained, effectively quantifying the potential collision risk at different time points. The instantaneous conflict risk values ​​along the entire optimized path are summed to obtain a dynamic congestion score. Based on trajectory prediction, the dynamic obstacle avoidance method effectively prevents mutual interference among multiple delivery personnel during the sorting process, significantly reducing the risk of channel congestion and improving sorting efficiency.

[0009] Optionally, the position and direction of the target planned point and the target predicted point are calculated to obtain the instantaneous conflict risk value. Specifically, this includes: obtaining the first position coordinates and the first movement direction vector of the target planned point, and obtaining the second position point and the second movement direction vector of the target predicted point; calculating the distance between the first position coordinates and the second position coordinates to obtain the spatial distance; determining the maximum relative distance and the minimum relative distance based on the spatial distance; when the spatial distance is greater than or equal to the maximum relative distance, determining the distance risk component as 0; when the spatial distance is less than the maximum relative distance but greater than the minimum relative distance, dividing the difference between the maximum relative distance and the spatial distance by the difference between the maximum relative distance and the minimum relative distance to obtain the distance risk component; when the spatial distance is less than or equal to the minimum relative distance, determining the distance risk component as 1; calculating the dot product of the first movement direction vector and the second movement direction vector to obtain the direction dot product value; calculating the direction threat factor using a linear mapping function on the direction dot product value, where the direction threat factor corresponding to a negative direction dot product value is greater than the direction threat factor corresponding to a positive direction dot product value; and multiplying the distance risk component by the direction threat factor to obtain the instantaneous conflict risk value at the target time point.

[0010] By adopting the above technical solution, a dynamic calculation method for the maximum and minimum relative distances is first introduced, effectively considering practical factors such as delivery personnel's walking speed, safe reaction time, and the physical characteristics of the passageway, making the distance risk assessment more in line with the needs of actual scenarios. In particular, the calculation of the distance risk component employs a piecewise function design based on spatial distance, ensuring low-risk conditions outside the safe distance while accurately identifying high-risk situations within the danger distance. Simultaneously, a dot product operation is performed on the movement direction vectors, and a directional threat factor is introduced to quantify the degree of danger of personnel moving towards each other into a specific numerical indicator, thereby distinguishing the risk levels under different combinations of movement directions. By multiplying the distance risk component by the directional threat factor to obtain the instantaneous conflict risk value, a comprehensive assessment of potential collision risks is achieved, thus preventing mutual interference between delivery personnel.

[0011] Optionally, the demand heat value is calculated for each product type among multiple pending orders received by the unmanned vending store within a preset time. Specifically, this includes: calculating the target demand quantity for each product type in the multiple pending orders, with one target demand quantity for each product type; obtaining the weather and seasonal information of the unmanned vending store at the current moment, and searching for multiple historical orders with the same seasonal information from the historical order database; calculating the similarity between the historical weather information and the weather information of the multiple historical orders to obtain multiple similarity scores; selecting the historical order with the highest similarity score from the multiple historical orders as the target historical order, and obtaining the historical average demand quantity of the target historical order; using the similarity score of the target historical order as the historical data confidence score, multiplying the historical data confidence score by a preset scaling factor to obtain a first weight, and summing the first weight and the second weight to 1; adding the product of the historical average demand quantity and the first weight to the product of the target demand quantity and the second weight to obtain the demand heat value, with one demand heat value for each product type.

[0012] By employing the aforementioned technical solution, the real-time demand quantity of each product type in pending orders is statistically analyzed to determine the current basic demand data. Similarity calculations are performed on weather information from historical orders, and the most similar historical orders are selected as reference samples. This not only ensures the relevance of historical data but also provides reliable historical demand reference values. The similarity of historical orders is used as the confidence level for historical data, and a preset scaling factor ensures that the weight of historical data is always less than that of real-time data. This guarantees the predictive model's inheritance of historical patterns while highlighting the dominant role of current real-time data. Finally, the demand heat value obtained through weighted calculation integrates empirical data from historical orders and fully considers the real-time demand of current orders, making product demand prediction more accurate and reliable. This demand heat calculation method based on multi-dimensional data can effectively guide the optimization selection of sorting paths.

[0013] Optionally, the target product types corresponding to all channels of each candidate path in the statistical path candidate set are summed based on the demand heat value to obtain the total demand pressure value. Specifically, this includes: obtaining multiple channels traversed from the target candidate path; determining multiple target product types within the target channel based on the product layout map; the target candidate path being any candidate path in the path candidate set; and the target channel being any one of the multiple channels; obtaining the demand heat value corresponding to each target product type; summing the multiple demand heat values ​​to obtain the channel demand pressure value of the target channel; and obtaining the channel demand pressure value corresponding to each channel in the target candidate path; and summing the multiple channel demand pressure values ​​to obtain the total demand pressure value of the target candidate path.

[0014] By adopting the above technical solution, the system analyzes the channels traversed by the target candidate path based on the product layout map, identifies the product type distribution within each channel, establishes a spatial mapping relationship between channels and products, obtains and sums the demand heat values ​​corresponding to each product type, transforms product demand into channel pressure, and thus accurately assesses the potential congestion risk of each channel. Particularly in the path evaluation stage, the total demand pressure value is obtained by accumulating the demand pressure values ​​of each channel. This not only considers the local pressure of a single channel but also achieves a global evaluation of the entire candidate path, effectively improving the scientific nature and accuracy of sorting path planning.

[0015] Optionally, candidate paths in the path candidate set are eliminated based on the total demand pressure value to obtain a path optimization set. Specifically, this includes: obtaining the total demand pressure value corresponding to the target candidate path from the path candidate set; obtaining the channel width corresponding to each channel traversed by the target candidate path, and selecting the minimum channel width from multiple channel widths; dividing the total demand pressure value by the minimum channel width to obtain the comprehensive path risk score of the target candidate path; and eliminating target candidate paths in the path candidate set whose comprehensive path risk score is greater than a preset risk threshold to obtain the path optimization set.

[0016] By employing the aforementioned technical solution, a comprehensive path risk score is obtained by calculating the ratio of the total demand pressure value to the minimum aisle width. This scoring mechanism not only considers the overall demand pressure on the path but also incorporates the physical capacity of the aisle, making the risk assessment more aligned with real-world scenarios. By setting preset risk thresholds and eliminating high-risk paths accordingly, potentially congested paths can be proactively avoided, resulting in a more reliable set of optimized paths. This path selection method based on physical constraints avoids the problem of neglecting aisle capacity in traditional path planning while ensuring that the optimized paths have sufficient passage space, significantly improving the feasibility and efficiency of sorting operations.

[0017] Optionally, after sending the target route to the target delivery person so that the target delivery person can sort and pack the goods in the target order according to the target route, the method further includes: obtaining the sorting video of the target delivery person from the unmanned vending store, determining the actual movement route based on the sorting video and the product layout map; comparing the actual movement route with the target route; if the actual movement route is consistent with the target route, selecting multiple sorting points from the target route, determining the dwell time of the target delivery person at each sorting point based on the sorting video, mapping the dwell time to each sorting point on the target route to obtain the actual sorting route map, establishing the correspondence between the target order and the actual sorting route map, and storing the correspondence in the sorting database; if the actual movement route is inconsistent with the target route, marking the inconsistent abnormal sorting points, and obtaining the target dwell time of the target delivery person at the abnormal sorting points; when the target dwell time exceeds a preset dwell threshold, confirming that the actual movement route is in an abnormal sorting state, and sending a warning message to the terminal of the unmanned vending store based on the abnormal sorting state.

[0018] By adopting the above technical solution, a complete sorting behavior tracking system was constructed by acquiring sorting videos of target delivery personnel and analyzing their actual movement routes in conjunction with product layout maps. Under normal sorting conditions, the system records the dwell time of target delivery personnel at each sorting point and maps this time data onto the actual sorting route map, forming a detailed sorting behavior profile. For anomaly identification, the system compares the differences between the actual movement route and the planned path, and makes judgments based on the dwell time at abnormal sorting points. When abnormal dwelling behavior is detected, a warning message can be sent to the terminal in a timely manner, enabling real-time intervention in the sorting process. This not only allows for timely detection and handling of sorting anomalies to ensure sorting quality, but also enables continuous optimization of the sorting process through data accumulation, improving the operational efficiency and service level of unmanned vending stores.

[0019] A second aspect of this application provides a path optimization device for picking goods in an unmanned warehouse. The device includes an acquisition unit, a verification unit, a processing unit, and a sending unit. The acquisition unit acquires product information from a target order and determines the unmanned vending machine store based on the product information; it sends the target order and the address of the unmanned vending machine store to a target delivery person so that the target delivery person can go to the unmanned vending machine store according to the target order; after determining that the target delivery person has arrived at the unmanned vending machine store, it acquires the identity verification information uploaded by the target delivery person; the verification unit, if the identity verification information is correct, sends an opening command to the unmanned vending machine store and extracts a list of product categories from the target order; the processing unit retrieves the product layout map of the unmanned vending machine store, determines the placement position of each product in the product category list according to the product layout map, acquires the current position of the target delivery person, and performs path planning on the current position and multiple placement positions based on preset rules to obtain a path candidate. The process involves: selecting a path candidate set; calculating the demand heat value for each product type among multiple pending orders received by the unmanned vending store within a preset time period, including target orders; statistically analyzing the target product types corresponding to all channels of each candidate path in the path candidate set, summing the demand heat values ​​for multiple target product types to obtain the total demand pressure value, where each total demand pressure value corresponds to one candidate path; eliminating candidate paths from the path candidate set based on the total demand pressure value to obtain the path optimization set; calculating the dynamic congestion score for each optimized path in the path optimization set based on the personnel movement trajectory within a preset time window, and selecting the optimized path with the minimum dynamic congestion score from the path optimization set as the target path; and sending the target path to the target delivery person so that the target delivery person can sort and pack the goods in the target order according to the target path and deliver the packed goods.

[0020] In a third aspect, this application provides an electronic device including a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory, causing the electronic device to perform the method as described in any of the above-described methods of this application.

[0021] In a fourth aspect, this application provides a computer-readable storage medium storing instructions that, when executed, perform any of the methods described above in this application.

[0022] In summary, one or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. After obtaining the product information of the target order and verifying identity, the unmanned vending store is opened, effectively ensuring the security of the sorting process. Preliminary route planning is performed based on the product layout map, providing basic sorting route guidance for the target delivery personnel and avoiding the blindness of traditional experience-based sorting. By calculating the demand heat value of each product type in the pending orders and combining it with the product types corresponding to the channels in the candidate paths to assess the total demand pressure value, intelligent identification and elimination of high-demand-pressure paths are achieved. Simultaneously, by calculating the dynamic congestion score to select the optimal path, the problem of mutual interference when multiple delivery personnel work simultaneously is effectively solved. This not only significantly improves sorting efficiency but also reduces the risk of channel congestion, allowing the target delivery personnel to sort and pack goods according to the planned optimal path, greatly improving sorting accuracy and delivery timeliness.

[0023] 2. By mapping the estimated arrival time to various coordinate points on the optimized path, a complete planned movement trajectory is formed. This accurately grasps the spatiotemporal distribution of the target delivery personnel during the sorting process. By calculating the positional and directional relationship between the target planned point and the target predicted point, an instantaneous conflict risk value is obtained, effectively quantifying the potential collision risk at different time points. The instantaneous conflict risk values ​​along the entire optimized path are summed to obtain a dynamic congestion score. The dynamic obstacle avoidance method based on trajectory prediction can effectively prevent mutual interference among multiple delivery personnel during the sorting process, significantly reducing the risk of channel congestion and improving sorting efficiency. Attached Figure Description

[0024] Figure 1 This is a first flowchart illustrating a path optimization method for picking goods in an unmanned warehouse provided in an embodiment of this application; Figure 2 This is a second flowchart illustrating a path optimization method for picking goods in an unmanned warehouse, as provided in an embodiment of this application. Figure 3 This is a schematic diagram of the structure of a path optimization device for picking goods in an unmanned warehouse provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of this application.

[0025] Explanation of reference numerals in the attached drawings: 301, acquisition unit; 302, verification unit; 303, processing unit; 304, sending unit; 400, electronic device; 401, processor; 402, memory; 403, user interface; 404, network interface; 405, communication bus. Detailed Implementation

[0026] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0027] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0028] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0029] Currently, stores primarily utilize technologies such as intelligent shelf management systems and electronic tags to support merchandise sorting. These systems provide delivery personnel with product location information and sorting guidance through functions like product positioning and inventory management. However, existing merchandise sorting systems still have significant shortcomings in sorting route planning. Due to the lack of intelligent route planning methods, delivery personnel need to determine sorting routes based on experience. This extensive sorting method is not only inefficient but also prone to causing congestion within the store. Especially during peak order periods, when multiple delivery personnel are working simultaneously, the lack of scientific route planning and scheduling mechanisms easily leads to mutual interference, resulting in low sorting efficiency. Existing route planning methods mostly rely on static shortest path calculations, lacking prediction of the future movements of other personnel and failing to proactively avoid dynamic conflicts. Therefore, they are ineffective in scenarios involving multiple tasks running concurrently.

[0030] To address the problem of mutual interference when multiple delivery personnel work simultaneously, this application provides a method for optimizing picking routes in an unmanned warehouse. This method can be applied to a platform server that provides product sorting guidance services for the unmanned warehouse. For example, this method can run on a server deployed in the cloud, which communicates with the positioning system (such as UWB, Bluetooth AOA), video surveillance system, access control system, and the delivery personnel's mobile terminals within the unmanned warehouse. Figure 1This is a schematic diagram of the first process of a path optimization method for picking goods in an unmanned warehouse provided in an embodiment of this application. (Refer to...) Figure 1 The method includes the following steps S101-S105.

[0031] S101: Obtain product information from the target order, determine the unmanned vending store based on the product information; send the target order and the address of the unmanned vending store to the target delivery person so that the target delivery person can go to the unmanned vending store according to the target order.

[0032] In step S101 above, product information, such as product name and quantity, is first obtained from the target order. Then, the unmanned vending machine store with the corresponding product inventory is determined based on this information. After determining the unmanned vending machine store, the delivery address information is obtained from the target order, and a delivery person near the delivery address is located using an LBS positioning system. When multiple delivery persons are detected near the delivery address, the delivery person with the shortest straight-line distance is selected as the target delivery person by calculating the straight-line distance between each delivery person and the delivery address. This selection method can effectively shorten delivery time and improve delivery efficiency. The unmanned vending machine store mentioned in this embodiment refers to an unmanned warehouse, which can be a pre-positioned warehouse for instant delivery, a community unmanned convenience store, or other location with inventory and sorting functions.

[0033] After identifying the target delivery person, the target order information (including the product list, quantity, etc.) and the specific address of the unmanned vending store are sent to the target delivery person's mobile terminal. The target delivery person can then go to the corresponding location to sort the goods based on the received address. This method of assigning delivery people based on the nearest delivery address not only reduces the delivery person's empty runs but also improves delivery efficiency. Furthermore, by automatically assigning delivery tasks, it avoids delays that might occur due to manual scheduling, laying the foundation for subsequent intelligent sorting and timely delivery. As a new retail format without human intervention, unmanned vending stores, through intelligent systems for product management and sorting, can significantly reduce labor costs and provide 24 / 7 delivery service support.

[0034] S102: After confirming that the target delivery person has arrived at the unmanned vending store, obtain the identity verification information uploaded by the target delivery person.

[0035] In step S102 above, upon arrival of the target delivery person at the unmanned vending store, an identity verification process is initiated to ensure sorting security. The target delivery person scans a QR code posted at the vending store to access the corresponding page. Following the on-screen instructions, they capture their facial image using a camera and simultaneously enter their name, mobile phone number, and assigned order number on the verification interface. Upon receiving the identity verification information uploaded by the target delivery person, the system compares the facial image with pre-stored delivery person photos on the delivery platform, verifies that the entered name and mobile phone number match the delivery person information recorded in the order, and checks the validity of the order number.

[0036] Multi-factor authentication mechanisms not only effectively prevent unauthorized personnel from entering unmanned vending stores, but also establish a clear correlation between sorting activities and specific delivery personnel, facilitating subsequent tracking and accountability. By requiring target delivery personnel to upload complete identity verification information, delivery personnel identities can be accurately identified, ensuring that every sorting operation is traceable, significantly improving the security and traceability of unmanned vending stores.

[0037] S103: If the identity verification information is correct, send an opening command to the unmanned vending store and extract a list of product categories from the target order.

[0038] In step S103 above, after verifying the identity information of the target delivery person—that is, when the facial image, name, mobile phone number, and order number all match the preset information—the access control process of the unmanned vending store is triggered. First, an opening command is sent to the intelligent access control system of the unmanned vending store, opening the store door via an electronically controlled actuator. Simultaneously, the sensing system inside the store continuously monitors the movement of people in the entrance area. Once the target delivery person is detected to have entered the unmanned vending store, a closing command is immediately sent to the access control system to ensure the store door closes promptly. This intelligent access control mechanism effectively prevents the store door from remaining open for extended periods, avoiding unauthorized entry and thus ensuring the safe operation of the unmanned vending store. After ensuring the safe entry of the target delivery person, a list of product categories is automatically extracted from the target order to prepare for subsequent sorting route planning. This strict access management method not only enables accurate identification and control of sorting personnel but also prevents unauthorized entry by promptly closing the store door, significantly improving the security level of the unmanned vending store. At the same time, this precise access mechanism based on identity verification also provides a reliable foundation for subsequent sorting behavior tracking and security management, ensuring that the entire sorting process is carried out in a controllable environment.

[0039] Furthermore, when a mismatch is detected between the identity verification information uploaded by the target delivery person and the information stored in the database, the opening of the unmanned vending store will be immediately blocked. Simultaneously, an identity verification failure message will pop up on the target delivery person's mobile terminal, and the identity verification interface will be automatically reset. The server will resend the identity verification request to the target delivery person, guiding them to re-capture their facial image and re-enter their personal information. This instant feedback and retry mechanism helps delivery people promptly identify and correct input errors, improving verification efficiency. It also records the specific reasons for each verification failure, such as unclear facial images or incorrect information, and provides corresponding improvement suggestions when resending the verification request. The system counts the number of identity verification failures by the target delivery person in real time. When the number of failures reaches a preset threshold (e.g., three consecutive verification failures), it automatically escalates to a more stringent identity verification method. This involves either calling the fingerprint recognition module on the mobile terminal to require the target delivery person to perform fingerprint verification, or sending a verification code to the target delivery person's pre-registered mobile phone number via SMS, requiring the delivery person to enter the received verification code for identity confirmation. This progressive verification mechanism ensures security while also providing alternative verification methods for possible special situations (such as facial recognition being affected by lighting conditions). If the target delivery person still fails identity verification after completing the upgrade verification, the anomaly is automatically recorded, and an alert is sent to the delivery platform's security management system. Simultaneously, the delivery person's sorting privileges are temporarily locked, and the delivery platform's customer service personnel are notified for manual verification and handling. This multi-layered protection mechanism not only effectively prevents security risks such as identity theft but also provides a basis for subsequent security management optimization through data recording.

[0040] By implementing this hierarchical identity verification mechanism, both the rigor of identity verification and sufficient operational flexibility are ensured, effectively balancing the needs of security and convenience. At the same time, the comprehensive anomaly handling mechanism provides reliable assurance for the secure operation of unmanned vending stores, enhancing the security and reliability of the entire delivery system.

[0041] S104: Retrieve the product layout map of the unmanned vending store, determine the placement position of each product in the product category list based on the product layout map, obtain the current position of the target delivery person, and perform path planning on the current position and multiple placement positions based on preset rules to obtain a path candidate set.

[0042] In step S104 above, a pre-established merchandise layout map of the unmanned vending store is retrieved first. This map contains complete spatial distribution information of the merchandise. Specifically, the layout map records the exact coordinates of each vending machine, its number, and detailed information about each shelf and compartment within the vending machine. It also includes a database of mapping relationships between merchandise and vending machines, indicating which shelf and location each type of merchandise is stored in. By querying this mapping relationship, the specific placement of each item in the merchandise category list can be quickly located. For example, when the list includes "mineral water," the specific spatial coordinates of the item being stored in "Vendor 3, shelf 2, compartment 1 in area A" are retrieved. Simultaneously, the current location coordinates of the target delivery person are obtained in real time through positioning devices within the store (such as Bluetooth beacons or visual positioning systems). After obtaining the delivery person's location and the locations of all target merchandise, multiple possible sorting paths are calculated based on preset path planning rules, such as the shortest distance principle and rules to avoid cross-tripping, forming a path candidate set. For example, preset rules may include using the A* algorithm or Dijkstra's algorithm to generate multiple candidate paths with lengths within a preset threshold.

[0043] This route planning method, based on precise spatial positioning, not only helps delivery personnel quickly locate the specific position of each item but also improves sorting efficiency through reasonable route arrangement. By transforming the abstract product list into a specific sequence of spatial coordinates, it provides clear execution guidance for subsequent intelligent sorting, significantly improving the accuracy and efficiency of sorting operations.

[0044] S105: Calculate the demand heat value for each product type among multiple pending orders received by the unmanned vending store within a preset time. The multiple pending orders include the target order.

[0045] In step S105 above, the target demand quantity for each product type in multiple pending orders is calculated, with one target demand quantity for each product type; the weather and seasonal information of the unmanned vending store at the current moment are obtained, and multiple historical orders with the same seasonal information are searched from the historical order database; the similarity between the historical weather information and the weather information of multiple historical orders is calculated to obtain multiple similarity scores; the historical order with the highest similarity score is selected as the target historical order, and the historical average demand quantity of the target historical order is obtained; the similarity score of the target historical order is used as the historical data confidence score, and the historical data confidence score is multiplied by a preset scaling factor to obtain the first weight, and the sum of the first weight and the second weight is 1; the product of the historical average demand quantity and the first weight is added to the product of the target demand quantity and the second weight to obtain the demand heat value, with one demand heat value for each product type.

[0046] Specifically, the system performs real-time analysis on all pending orders received by the unmanned vending store within a preset time period (e.g., the last 2 hours). First, it extracts all pending orders from the order management database, including currently being processed target orders and other orders yet to be processed. For each order, the products are categorized and statistically analyzed by type; for example, "Coca-Cola" and "Pepsi" are categorized as "carbonated beverages," and "X-Master instant noodles" and "X-One instant noodles" are categorized as "instant noodles." Then, based on the product list of each pending order, the demand quantities of the same product type are accumulated. For example, when multiple orders contain "mineral water," the quantities of "mineral water" in these orders are summed to obtain the target demand quantity for that product type. The system establishes a counter for each product type to ensure accurate recording of real-time demand for each type of product. Real-time weather information, including temperature, humidity, and weather conditions, is obtained from a meteorological data interface. Simultaneously, the system automatically determines the season based on the current date, such as classifying March to May as spring. This environmental information will serve as a key parameter for querying historical data. Historical order records matching the current season will be selected from the historical order database to create a preliminary historical dataset.

[0047] For the selected historical orders, the similarity between historical weather information and current weather information is calculated. Specifically, parameters such as temperature and humidity differences are weighted according to preset weights to obtain a similarity value between 0 and 1. For example, if the temperature in a historical order is 25℃ and the current temperature is 26℃, this similarity might receive a higher score. The similarity scores of each historical order are sorted, and the order with the highest similarity is selected as the target historical order. The historical average demand quantity for each product type is extracted from this target historical order. This historical average demand quantity reflects the level of product demand under similar environmental conditions. The similarity value of the target historical order is used as a reliability indicator of the historical data and multiplied by a preset scaling factor (e.g., 0.6) to obtain the first weight. Simultaneously, a second weight is automatically calculated to ensure that the sum of the two weights is 1. This weighting mechanism can appropriately control the influence of historical data while ensuring its reference value. The final demand heat value is obtained by multiplying the historical average demand quantity by the first weight and adding the product of the target demand quantity and the second weight. This calculation method takes into account both the reference value of historical data and the impact of real-time demand, so that the demand heat value can more accurately reflect the actual sorting pressure of goods.

[0048] By implementing this demand heat calculation method that comprehensively considers both historical and real-time data, the sorting demand for goods can be predicted more accurately, providing a more reliable basis for subsequent route planning. At the same time, this dynamic calculation method based on environmental factors also improves the accuracy of demand forecasting, effectively enhancing the overall intelligence level of the sorting system.

[0049] S106: Statistically calculate the target product types corresponding to all channels of each candidate path in the candidate path set, and sum the multiple target product types according to the demand heat value to obtain the total demand pressure value.

[0050] In S106 above, multiple channels passed through are obtained from the target candidate path, and multiple target product types within the target channel are determined according to the product layout map. The target candidate path is any candidate path in the path candidate set, and the target channel is any one of the multiple channels. The demand heat value corresponding to each target product type is obtained, and the multiple demand heat values ​​are summed to obtain the channel demand pressure value of the target channel. The channel demand pressure value corresponding to each channel in the target candidate path is obtained, and the multiple channel demand pressure values ​​are accumulated and summed to obtain the total demand pressure value of the target candidate path.

[0051] Specifically, a target candidate path is first selected from the path candidate set for analysis. Based on the spatial trajectory information of this path, all channels traversed by the path are identified. For each target channel, a pre-established product layout database is accessed, which records in detail the product arrangement of the shelves on both sides of each channel. By querying the layout map, information on the product types on all shelves within the target channel can be obtained. For example, when analyzing channel A, it may be identified that this channel contains multiple product types such as beverages, snacks, and daily necessities. This spatial location-based product type identification method can comprehensively grasp the product distribution of each channel. After obtaining the product type information of the target channel, the demand heat value corresponding to each product type is queried. These demand heat values ​​come from the results calculated based on historical data and real-time orders. The demand heat values ​​of all product types within the target channel are summed to obtain the channel demand pressure value. This pressure value comprehensively reflects the current demand situation of all products within the channel. For example, if the demand heat value of beverages in channel A is 0.8, snacks is 0.6, and daily necessities is 0.4, then the channel demand pressure value is 1.8. This calculation method accurately reflects the actual sorting pressure of each channel. After obtaining the channel demand pressure values ​​of all channels along the target candidate path, a final cumulative calculation is performed. The channel demand pressure values ​​of all channels traversed by the target candidate path are summed to obtain the total demand pressure value of the target candidate path. This total value comprehensively reflects the cumulative sorting pressure along the entire path.

[0052] For example, if a path sequentially passes through channel A (pressure value 1.8), channel B (pressure value 2.1), and channel C (pressure value 1.5), then the total demand pressure value for this path is 5.4. By implementing this hierarchical pressure calculation method, a complete pressure assessment system can be gradually built from product type, channel level, to the entire path. This refined calculation method not only considers the demand intensity of products but also fully integrates the spatial distribution characteristics of products, making the path assessment results more accurate and reliable. Simultaneously, since each candidate path receives a specific total demand pressure value, the system can intuitively compare the pressure distribution of different path options, providing important decision-making basis for subsequent path optimization and effectively improving the intelligence level of the sorting system.

[0053] S107: Eliminate candidate paths from the path candidate set based on the total demand pressure value to obtain the path optimization set.

[0054] In S107 above, the total demand pressure value corresponding to the target candidate path is obtained from the path candidate set; the channel width corresponding to each channel traversed by the target candidate path is obtained, and the minimum channel width of the path is selected from multiple channel widths; the total demand pressure value is divided by the minimum channel width of the path to obtain the comprehensive risk score of the target candidate path; the target candidate paths in the path candidate set whose comprehensive risk score is greater than the preset risk threshold are eliminated to obtain the path optimization set.

[0055] Specifically, the system first selects target candidate paths from the path candidate set and reads their corresponding total demand pressure values. Simultaneously, it accesses the spatial layout database of the unmanned vending store, which records detailed physical parameter information for each aisle. For each aisle traversed by the target candidate path, its aisle width data is extracted. For example, aisle A has a width of 1.5 meters, aisle B has a width of 1.2 meters, and aisle C has a width of 1.8 meters. These aisle width data reflect the actual passage capacity of each aisle. Through comparative analysis, the system identifies the aisle with the smallest width among all aisles traversed by the path and records its width value as the minimum aisle width. This minimum width often becomes the physical bottleneck of the entire path. Dividing the previously obtained total demand pressure value by the minimum aisle width yields the overall path risk score for the target candidate path. This calculation method considers the dual impact of product demand pressure and physical space limitations. For example, if a path has a total demand pressure value of 5.4 and a minimum aisle width of 1.2 meters, its overall path risk score is 4.5. This risk score reflects both the product demand pressure on the path and the impact of aisle width on sorting efficiency. Narrower aisles lead to higher risk scores, highlighting the importance of physical space constraints. The overall risk score of each target candidate path is compared to a preset risk threshold. This threshold is a safety standard set based on historical sorting data and actual operational experience. When a path's risk score exceeds this threshold, it indicates that the path may experience congestion during actual sorting due to narrow aisles or concentrated demand for goods. These high-risk paths are removed from the path candidate set, and the remaining low-risk paths constitute the final optimized path set.

[0056] By implementing this path optimization method that incorporates physical constraints, the actual feasibility of each candidate path can be evaluated more comprehensively. This optimization approach not only considers the pressure from commodity demand but also incorporates the key physical factor of channel width into the evaluation system, making the path selection results more in line with actual operational needs. Simultaneously, by establishing a screening mechanism based on risk scores, it is possible to effectively identify and eliminate potentially high-risk path solutions, providing a more reliable candidate set for subsequent path selection.

[0057] S108: Calculate the dynamic congestion score for each optimized path in the path optimization set based on the movement trajectory of people within a preset time window, and select the optimized path with the minimum dynamic congestion score from the path optimization set as the target path.

[0058] In S108 above, after obtaining the path optimization set, the dynamic congestion score is calculated for each optimized path in the path optimization set based on the movement trajectory of people within a preset time window. Specifically, this includes: obtaining the current state of other people in the unmanned vending store at the current moment, including the target position, movement speed, and movement direction; predicting the future movement trajectory of other people within the preset time window based on the movement direction and movement speed in the merchandise layout map, using the target position as the starting point, and obtaining multiple predicted movement trajectories; calculating the estimated arrival time corresponding to each coordinate point on the target optimized path based on the movement speed, mapping each estimated arrival time to each coordinate point on the target optimized path, and obtaining the planned movement trajectory, where the target optimized path is any optimized path in the path optimization set; obtaining the target prediction point and target planned point at the target time point from the target predicted movement trajectory and the planned movement trajectory, calculating the position and direction of the target planned point and the target prediction point, and obtaining the instantaneous conflict risk value; and summing up the instantaneous conflict risk values ​​in the target optimized path to obtain the dynamic congestion score corresponding to the target optimized path.

[0059] Specifically, the positioning system within the unmanned vending store collects real-time status information of all other personnel (including store restocking staff and other delivery personnel). This includes obtaining each person's target location coordinates, current speed (e.g., 1.2 m / s), and direction of movement (e.g., the angle relative to the store's due north direction). Based on this real-time status data, trajectory prediction calculations are performed within the coordinate system of the merchandise layout map. A physical kinematics model is used, combining the person's current position, speed, and direction, to predict their movement trajectory within a preset time window (e.g., the next 10 minutes). The prediction process considers physical constraints in the merchandise layout map, such as passageway restrictions and corner areas, ensuring that the predicted trajectory conforms to actual walkable paths. This prediction method generates multiple possible movement trajectories for each person within the preset time window. For each target optimized path in the path optimization set, the delivery person's expected movement state on that path needs to be calculated. Based on the delivery person's standard moving speed (e.g., 1 m / s) and the spatial distance of the path, the estimated arrival time for the delivery person to reach each coordinate point on the path is calculated. This calculation process considers features such as turns and stops on the path to generate more accurate time estimates. These estimated arrival times are mapped to path coordinates to form a planned movement trajectory containing spatiotemporal information. This planned movement trajectory records in detail the delivery person's expected location and time information when performing sorting tasks. Within a preset time window, the predicted movement trajectory and the planned movement trajectory are sampled and analyzed at fixed time intervals (e.g., 1 second). At each target time point, the target predicted point location of other personnel and the target planned point location of the delivery person are obtained. The spatial distance and the angle between the movement directions of these two types of points are calculated to obtain the instantaneous conflict risk value.

[0060] Furthermore, the instantaneous conflict risk value is obtained by calculating the position and direction of the target planned point and the target predicted point. Specifically, this includes: obtaining the first position coordinates and the first movement direction vector of the target planned point, and obtaining the second position point and the second movement direction vector of the target predicted point; calculating the distance between the first position coordinates and the second position coordinates to obtain the spatial distance; determining the maximum relative distance and the minimum relative distance based on the spatial distance; determining the distance risk component as 0 when the spatial distance is greater than or equal to the maximum relative distance; obtaining the distance risk component by dividing the difference between the maximum relative distance and the spatial distance by the difference between the maximum relative distance and the minimum relative distance when the spatial distance is less than or equal to the minimum relative distance; determining the distance risk component as 1 when the spatial distance is less than or equal to the minimum relative distance; calculating the dot product of the first movement direction vector and the second movement direction vector to obtain the direction dot product value; calculating the direction threat factor by performing a linear mapping function on the direction dot product value, where the direction threat factor corresponding to a negative direction dot product value is greater than the direction threat factor corresponding to a positive direction dot product value; and multiplying the distance risk component by the direction threat factor to obtain the instantaneous conflict risk value at the target time point.

[0061] Specifically, the first position coordinates (x1, y1) and first movement direction vector v1 of the target planned point are obtained from the planned movement trajectory, while the second position coordinates (x2, y2) and second movement direction vector v2 of the target predicted point are obtained from the predicted movement trajectory. These coordinates and vectors are represented in the two-dimensional coordinate system of the merchandise layout diagram. The movement direction vector is normalized to ensure that its magnitude is 1, representing only directional information. For example, at a certain moment, the delivery person's planned point coordinates are (3, 4), and the movement direction vector is (0.707, 0.707), indicating that it moves along a 45-degree angle. The Euclidean distance formula is used to calculate the spatial distance between the two positions. Based on the actual operational experience of unmanned vending stores, two threshold parameters are preset: a maximum relative distance (e.g., 2 meters) and a minimum relative distance (e.g., 0.5 meters). These two parameters represent the distance at which congestion risk needs to be considered and the distance at which congestion will definitely occur, respectively. By comprehensively considering the characteristics of personnel movement and the features of the scene environment, the maximum and minimum relative distances are set: The actual walking speed data of the target delivery person is collected in real time using positioning devices, while the walking speed records of other personnel (including customers and other delivery personnel) are extracted from the historical database to calculate the average walking speed. For example, the target delivery person's walking speed is 1.2 m / s, and the average walking speed of other personnel is 1.0 m / s. These two speed values ​​are added together (totaling 2.2 m / s), reflecting the relative approach speed of the two moving individuals when moving towards each other. Considering that personnel reaction and avoidance require a certain buffer time, a preset safe reaction time (e.g., 1.5 seconds) is set. The speed sum is multiplied by the safe reaction time to obtain the initial distance (e.g., 2.2 × 1.5 = 3.3 meters). This initial distance represents the minimum distance required to ensure safe avoidance under ideal conditions. However, considering the actual environmental characteristics of unmanned vending stores, the initial distance needs to be corrected according to the passage conditions. The average length of the store passage (e.g., 4 meters) is obtained by measurement or preset values ​​and used as a correction reference value. When the initial distance exceeds a certain percentage (e.g., 75%) of the average aisle length, it is corrected to a reasonable value not exceeding the aisle length to ensure the calculated maximum relative distance is consistent with the actual scenario. The target width data of the picking equipment (e.g., shopping cart or trolley) used by the delivery person is obtained, and the average shoulder width parameter is extracted from an ergonomic database. For example, the target width of the picking equipment is 0.6 meters, and the average shoulder width is 0.5 meters. These two values ​​are compared, and the larger value is selected as the minimum relative distance. In this embodiment, since the width of the picking equipment is greater than the shoulder width of the human body, 0.6 meters is determined as the minimum relative distance. This distance represents the minimum safe distance that must be maintained between two moving individuals; a distance smaller than this will result in physical collision or serious obstruction. By implementing this distance parameter calculation method based on actual motion characteristics and environmental constraints, a more reasonable risk assessment standard can be established.The calculation of the maximum relative distance fully considers the movement speed and reaction time of personnel, and makes appropriate corrections based on actual site conditions to ensure the practicality of the parameters. The determination of the minimum relative distance is based on physical constraints, effectively preventing potential collision risks.

[0062] Furthermore, after determining the maximum and minimum relative distances between the unmanned vending stores, a piecewise function is used to calculate the distance risk component based on the calculated spatial distance. When the spatial distance is greater than or equal to the maximum relative distance, it indicates that the two locations are far apart and there is no congestion risk; therefore, the distance risk component is 0. When the spatial distance is between the maximum and minimum relative distances, a linear interpolation formula is used: Distance risk component = (Maximum relative distance - Spatial distance) / (Maximum relative distance - Minimum relative distance), yielding a risk value between 0 and 1. When the spatial distance is less than or equal to the minimum relative distance, it indicates that the two locations are too close and there is a serious congestion risk; in this case, the distance risk component is 1. The dot product of the two movement direction vectors is calculated: Dot product = v1·v2. This dot product reflects the relative relationship between the two movement directions and ranges from [-1, 1]. An asymmetric linear mapping function was designed. When the dot product value is negative (indicating moving in opposite directions), it maps to a larger directional threat factor (e.g., the range of 0.8-1.0), while when the dot product value is positive (indicating moving in the same direction), it maps to a smaller directional threat factor (e.g., the range of 0.2-0.4). This asymmetric design reflects that moving in opposite directions carries a higher congestion risk than moving in the same direction. Multiplying the calculated distance risk component by the directional threat factor yields the instantaneous conflict risk value at the target time point. This calculation method considers both the influence of spatial location and the threat level of the movement direction, enabling a more accurate reflection of potential congestion risk. For example, when two points are close together and moving in opposite directions, a higher instantaneous conflict risk value is obtained. This accurately assesses the potential conflict risk between two moving individuals at any given time, effectively reducing the probability of congestion during the sorting process.

[0063] The dynamic congestion score of a target optimized path is obtained by summing the instantaneous conflict risk values ​​at all points within a preset time window. This score comprehensively reflects all congestion risks that the path may encounter during execution. A higher dynamic congestion score indicates that the path is more likely to encounter personnel congestion during actual execution. Implementing this trajectory prediction-based dynamic congestion assessment method can more accurately predict and evaluate the actual feasibility of each optimized path. This assessment method not only considers the current personnel distribution but also anticipates potential congestion risks through a trajectory prediction mechanism, making the path assessment results more forward-looking and reliable.

[0064] Furthermore, after obtaining the dynamic congestion scores for each optimized path in the path optimization set, the dynamic congestion scores of all optimized paths in the set are compared, and the optimized path with the lowest score is selected as the final target path. This path selection method based on real-time personnel distribution can effectively avoid densely populated areas and provide delivery personnel with a sorting path with the lowest actual congestion risk. Through this dynamic optimization mechanism, not only static product distribution and channel constraints are considered, but real-time personnel activity is also incorporated into the decision-making scope, significantly improving the practicality and reliability of path planning, effectively reducing the probability of congestion during the sorting process, and improving overall delivery efficiency.

[0065] S109: Send the target route to the target delivery person so that the target delivery person can sort and pack the goods in the target order according to the target route and deliver the packed goods.

[0066] In step S109 above, the target route is converted into an electronic map containing specific navigation guidance. This map clearly marks the sorting route, product locations, and specific operational instructions. This information is pushed to the target delivery personnel in real time via their mobile terminals (such as tablets or smartphones). The target delivery personnel view the navigation information of the target route through their terminal devices and proceed to each product shelf location according to the planned route. At each shelf location, the specific product information that needs to be sorted in the target order is automatically displayed, including product name, quantity, and shelf location. The target delivery personnel accurately retrieve and sort the products based on this information. After all products are sorted, the delivery personnel pack the products in a standardized manner in the packaging area designated by the system, automatically generating a packing list and delivery labels. After packing is completed, the target delivery personnel confirm that the order information is correct and deliver the packaged products to the designated location according to the provided delivery route. This intelligent route execution solution not only improves sorting efficiency but also significantly reduces the sorting error rate. Through real-time route guidance and product information prompts, it effectively reduces the time wasted by delivery personnel when searching for products, while ensuring the accuracy of the sorting process.

[0067] In one possible implementation, when the target delivery person sorts goods based on the target route, the sorting process is recorded in real time. Real-time video analysis is used to accurately assess the sorting behavior. Sorting videos of the target delivery person are obtained from the unmanned vending machine store. The actual movement route is determined based on the sorting video and the product layout map. The actual movement route is compared with the target route. If the actual movement route matches the target route, multiple sorting points are selected from the target route. The dwell time of the target delivery person at each sorting point is determined based on the sorting video. The dwell time is mapped to each sorting point on the target route to obtain the actual sorting route map. A correspondence between the target order and the actual sorting route map is established and stored in the sorting database. If the actual movement route does not match the target route, the inconsistent abnormal sorting points are marked, and the target dwell time of the target delivery person at the abnormal sorting point is obtained. When the target dwell time exceeds a preset dwell threshold, the actual movement route is confirmed as an abnormal sorting state, and an early warning message is sent to the terminal of the unmanned vending machine store based on the abnormal sorting state.

[0068] Specifically, video surveillance equipment within the unmanned vending store is used to collect real-time videos of delivery personnel sorting goods. This equipment includes multiple cameras covering all shelf areas and aisle spaces. Computer vision algorithms are used to process the video stream in real time, extracting the delivery personnel's movement trajectory information. Object detection and tracking algorithms identify the delivery personnel's real-time location coordinates on the merchandise layout map, and these coordinates are connected chronologically to form the actual movement route. Spatial mapping technology is used to convert the pixel coordinates in the video into actual physical coordinates on the merchandise layout map, ensuring the accuracy of route identification. The identified actual movement route is then precisely compared with the previously planned target path. The comparison process uses a dynamic time warping algorithm, setting a reasonable error tolerance range to account for minor deviations during actual execution. When the actual movement route matches the target path within the error range, the normal sorting analysis process begins. If the route consistency is confirmed, pre-set sorting points are extracted from the target path; these sorting points correspond to the shelf positions of each item in the target order. Video analytics technology is used to accurately calculate the duration of the delivery personnel's stay at each sorting point. Specifically, the entry and exit times of delivery personnel in the sorting area are detected, and the time difference is calculated as the dwell time. This dwell time data is mapped to the corresponding sorting points on the target path to generate an actual sorting route map containing time-related information. This route map not only includes spatial trajectories but also records the specific operation time at each sorting point. The correlation between target orders and the actual sorting route map is established, and this information is stored in the sorting database for subsequent efficiency analysis and optimization.

[0069] Furthermore, when a significant deviation between the actual movement route and the target path is detected, inconsistent and abnormal sorting points are immediately marked. The behavior of delivery personnel at these abnormal sorting points is closely monitored, and their target dwell time is precisely calculated. If the target dwell time exceeds a preset dwell threshold (e.g., 1.5 times the normal sorting time), the situation is determined to be an abnormal sorting state. This state may indicate that the delivery personnel are encountering difficulties in finding goods or that other abnormal situations have occurred. Once an abnormal sorting state is confirmed, an early warning message containing the abnormal location, abnormal type, and abnormal duration is immediately generated and pushed in real time through the management terminal of the unmanned vending store. This early warning message helps managers to promptly identify and handle problems in the sorting process, preventing potential order errors or inefficiencies. By implementing this comprehensive sorting monitoring and analysis method, precise management and timely intervention in the sorting process can be achieved. It not only provides detailed sorting behavior data but also significantly improves the controllability and optimizability of the sorting process through the abnormal early warning mechanism.

[0070] After receiving a target order, the server first parses it to obtain the delivery address and product information. Based on the product information, it determines the unmanned vending machine store and checks the distance to delivery personnel near the delivery address to identify the target delivery person. Then, it sends the target order and the unmanned vending machine store address to the target delivery person's terminal so that the delivery person can accept the order. Upon arriving at the unmanned vending machine store, the delivery person scans a QR code at the store to upload identity verification information. The uploaded identity verification information is verified. If the verification fails, the unmanned vending machine store is confirmed to be closed, and a verification failure message is sent to the delivery person so that they can rescan the QR code to verify their identity. The upload process is as follows: When the verification is successful, the unmanned vending store is confirmed to be open, and the product category list of the target order is displayed. Then, based on the product layout map of the unmanned vending store, the placement position of each product in the product category list is determined. A path candidate set is generated based on the placement position and the current location of the target delivery person. The demand pressure value and dynamic congestion score of each path channel in the path candidate set are calculated. The path candidate set is then filtered based on the demand pressure value and dynamic congestion score to obtain the target path. The target path is then sent to the target delivery person so that the target delivery person can sort and pack the products of the target order according to the prompts of the target path. After packing, the packed products are delivered.

[0071] This application also provides a path optimization device for picking goods in an unmanned warehouse. Figure 3 This is a schematic diagram of a path optimization device for picking goods in an unmanned warehouse, provided in an embodiment of this application. (Refer to...) Figure 3 The device includes an acquisition unit 301, a verification unit 302, a processing unit 303, and a sending unit 304; Unit 301 retrieves product information from the target order and determines the unmanned vending store based on the product information; it sends the target order and the address of the unmanned vending store to the target delivery person so that the target delivery person can go to the unmanned vending store according to the target order; after confirming that the target delivery person has arrived at the unmanned vending store, it retrieves the identity verification information uploaded by the target delivery person. If the identity verification information is correct, the verification unit 302 sends an opening command to the unmanned vending store and extracts a list of product categories from the target order. Processing unit 303 retrieves the merchandise layout map of the unmanned vending store, determines the placement of each item in the merchandise category list based on the merchandise layout map, obtains the current location of the target delivery person, and performs path planning based on the current location and multiple placement locations according to preset rules to obtain a path candidate set; calculates the demand heat value for each item type in multiple pending orders received by the unmanned vending store within a preset time, including the target order; counts the target item types corresponding to all channels of each candidate path in the path candidate set, sums the multiple target item types according to the demand heat value to obtain the total demand pressure value, where one total demand pressure value corresponds to one candidate path; eliminates candidate paths in the path candidate set according to the total demand pressure value to obtain a path optimization set; calculates the dynamic congestion score for each optimized path in the path optimization set based on the personnel movement trajectory within a preset time window, and selects the optimized path with the minimum dynamic congestion score from the path optimization set as the target path; Sending unit 304 sends the target route to the target delivery person so that the target delivery person can sort and pack the goods in the target order according to the target route and deliver the packed goods.

[0072] In one possible implementation, the acquisition unit 301 is used to acquire the current state of other personnel in the unmanned vending store at the current moment, including the target position, movement speed, and movement direction; the processing unit 303 is used to predict the future movement trajectory of other personnel within a preset time window based on the movement direction and movement speed in the merchandise layout map, starting from the target position, to obtain multiple predicted movement trajectories; calculate the estimated arrival time corresponding to each coordinate point on the target optimized path based on the movement speed, and map each estimated arrival time to each coordinate point on the target optimized path to obtain the planned movement trajectory, where the target optimized path is any optimized path in the path optimization set; obtain the target predicted point and target planned point at the target time point from the target predicted movement trajectory and the planned movement trajectory, calculate the position and direction of the target planned point and the target predicted point to obtain the instantaneous conflict risk value; and sum the instantaneous conflict risk values ​​in the target optimized path to obtain the dynamic congestion score corresponding to the target optimized path.

[0073] In one possible implementation, the acquisition unit 301 is used to acquire the first position coordinates and the first movement direction vector of the target planned point, and to acquire the second position point and the second movement direction vector of the target predicted point; the processing unit 303 is used to calculate the distance between the first position coordinates and the second position coordinates to obtain the spatial distance; determine the maximum relative distance and the minimum relative distance based on the spatial distance; when the spatial distance is greater than or equal to the maximum relative distance, determine that the distance risk component is 0; when the spatial distance is less than the maximum relative distance but greater than the minimum relative distance, divide the difference between the maximum relative distance and the spatial distance by the difference between the maximum relative distance and the minimum relative distance to obtain the distance risk component; when the spatial distance is less than or equal to the minimum relative distance, determine that the distance risk component is 1; perform a dot product calculation on the first movement direction vector and the second movement direction vector to obtain the direction dot product value; perform a linear mapping function calculation on the direction dot product value to obtain the direction threat factor, where the direction threat factor corresponding to a negative direction dot product value is greater than the direction threat factor corresponding to a positive direction dot product value; multiply the distance risk component by the direction threat factor to obtain the instantaneous conflict risk value at the target time point.

[0074] In one possible implementation, the processing unit 303 is used to count the target demand quantity corresponding to each product type in multiple pending orders, with one target demand quantity corresponding to one product type; the acquisition unit 301 is used to acquire the weather and seasonal information of the unmanned vending store at the current moment, and search for multiple historical orders with the same seasonal information from the historical order database; the processing unit 303 is used to calculate the similarity between the historical weather information and the weather information of multiple historical orders, and obtain multiple similarities; the historical order with the highest similarity is selected from the multiple historical orders as the target historical order, and the historical average demand quantity of the target historical order is obtained; the similarity of the target historical order is used as the historical data confidence score, and the historical data confidence score is multiplied by a preset scaling factor to obtain a first weight, and the sum of the first weight and the second weight is 1; the product of the historical average demand quantity and the first weight is added to the product of the target demand quantity and the second weight to obtain the demand heat value, with one demand heat value corresponding to one product type.

[0075] In one possible implementation, the acquisition unit 301 is used to acquire multiple channels traversed from the target candidate path, determine multiple target product types within the target channel according to the product layout map, the target candidate path is any candidate path in the path candidate set, and the target channel is any one of the multiple channels; acquire the demand heat value corresponding to each target product type, sum the multiple demand heat values ​​to obtain the channel demand pressure value of the target channel; acquire the channel demand pressure value corresponding to each channel in the target candidate path, and sum the multiple channel demand pressure values ​​to obtain the total demand pressure value of the target candidate path.

[0076] In one possible implementation, the acquisition unit 301 is used to acquire the total demand pressure value corresponding to the target candidate path from the path candidate set; acquire the channel width corresponding to each channel traversed by the target candidate path, and select the minimum channel width of the path from multiple channel widths; the processing unit 303 is used to divide the total demand pressure value by the minimum channel width of the path to obtain the comprehensive risk score of the target candidate path; and remove the target candidate paths in the path candidate set whose comprehensive risk score is greater than a preset risk threshold to obtain the path optimization set.

[0077] In one possible implementation, the acquisition unit 301 is used to acquire the sorting video of the target delivery person from the unmanned vending store, and determine the actual movement route based on the sorting video and the product layout map; the processing unit 303 is used to compare the actual movement route with the target path; if the actual movement route is consistent with the target path, multiple sorting points are selected from the target path, and the dwell time of the target delivery person at each sorting point is determined based on the sorting video, and the dwell time is mapped to each sorting point on the target path to obtain the actual sorting route map, establish the correspondence between the target order and the actual sorting route map, and store the correspondence in the sorting database; if the actual movement route is inconsistent with the target path, the inconsistent abnormal sorting points are marked, and the target dwell time of the target delivery person at the abnormal sorting point is acquired; when the target dwell time exceeds a preset dwell threshold, the sending unit 304 confirms that the actual movement route is in an abnormal sorting state, and sends a warning message to the terminal of the unmanned vending store based on the abnormal sorting state.

[0078] It should be noted that the above embodiments of the apparatus are only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0079] This application also discloses an electronic device. (See reference...) Figure 4 , Figure 4 This application provides a schematic diagram of the structure of an electronic device. The electronic device 400 may include: at least one processor 401, at least one network interface 404, a user interface 403, a memory 402, and at least one communication bus 405.

[0080] The communication bus 405 is used to enable communication between these components.

[0081] The user interface 403 may include a display screen and a camera. Optionally, the user interface 403 may also include a standard wired interface and a wireless interface.

[0082] The network interface 404 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0083] The processor 401 may include one or more processing cores. The processor 401 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 402, and by calling data stored in memory 402. Optionally, the processor 401 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 401 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and application requests; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 401 and may be implemented as a separate chip.

[0084] The memory 402 may include random access memory (RAM) or read-only memory. Optionally, the memory 402 may include a non-transitory computer-readable storage medium. The memory 402 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 402 may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch functionality, sound playback functionality, image playback functionality, etc.), instructions for implementing the various method embodiments described above, etc. The data storage area may store data involved in the various method embodiments described above. Optionally, the memory 402 may also be at least one storage device located remotely from the aforementioned processor 401.

[0085] like Figure 4 As shown, the memory 402, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application for optimizing picking routes in an unmanned warehouse.

[0086] exist Figure 4 In the electronic device 400 shown, the user interface 403 is mainly used to provide an interface for users to input data and obtain user input data; while the processor 401 can be used to call the application program for optimizing picking routes in the unmanned warehouse stored in the memory 402. When executed by one or more processors, the electronic device performs one or more of the methods described in the above embodiments.

[0087] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0088] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0089] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some service interfaces; indirect couplings or communication connections between devices or units may be electrical or other forms.

[0090] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0091] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0092] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0093] The above description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of other embodiments of this disclosure upon considering the specification and the disclosure of practical truths. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure.

Claims

1. A method for optimizing a picking path in an unmanned warehouse, characterized in that, The method comprises: obtaining commodity information from a target order, determining an unmanned store according to the commodity information, and sending the target order and the address of the unmanned store to a target delivery person so that the target delivery person goes to the unmanned store according to the target order; after determining that the target delivery person arrives at the unmanned store, obtaining identity verification information uploaded by the target delivery person; if the identity verification information is verified correctly, sending an opening instruction to the unmanned store, and extracting a commodity category list from the target order; calling a commodity layout map of the unmanned store, determining a placement position corresponding to each commodity in the commodity category list according to the commodity layout map, obtaining a current position of the target delivery person, performing path planning on the current position and a plurality of placement positions based on a preset rule to obtain a path candidate set; calculating each commodity type in a plurality of to-be-processed orders received by the unmanned store within a preset time to obtain a demand heat value, wherein the plurality of to-be-processed orders include the target order; counting target commodity types corresponding to all channels of each candidate path in the path candidate set, summing a plurality of target commodity types according to the demand heat value to obtain a total demand pressure value, wherein one total demand pressure value corresponds to one candidate path; performing elimination on candidate paths in the path candidate set according to the total demand pressure value to obtain a path optimization set; calculating each optimization path in the path optimization set based on personnel motion trajectories within a preset time window to obtain a dynamic congestion score, and selecting an optimization path with a minimum dynamic congestion score from the path optimization set as a target path; sending the target path to the target delivery person so that the target delivery person sorts and packages commodities in the target order according to the target path, and delivers the packaged commodities.

2. The method of claim 1, wherein, The calculation of each optimization path in the path optimization set based on personnel motion trajectories within a preset time window to obtain a dynamic congestion score specifically comprises: obtaining a current state of other personnel in the unmanned store at a current time, wherein the current state includes a target position, a moving speed, and a moving direction; in the commodity layout map, taking the target position as a starting point, predicting future motion trajectories of the other personnel within the preset time window according to the moving direction and the moving speed to obtain a plurality of predicted motion trajectories; calculating an estimated arrival time of each coordinate point on a target optimization path based on the moving speed, mapping each estimated arrival time to each coordinate point on the target optimization path to obtain a planned motion trajectory, and the target optimization path is any optimization path in the path optimization set; obtaining a target predicted point and a target planned point at a target time point from the target predicted motion trajectory and the planned motion trajectory, calculating the target planned point and the target predicted point in position and direction to obtain an instantaneous conflict risk value; The instantaneous conflict risk values in the target optimization path are accumulated and summed up to obtain the dynamic congestion score corresponding to the target optimization path.

3. The method of claim 2, wherein, The position and direction calculation on the target planning point and the target prediction point obtains an instantaneous conflict risk value, and specifically includes: The first position coordinates and the first moving direction vector of the target planning point are obtained, and the second position coordinates and the second moving direction vector of the target prediction point are obtained; the distance calculation on the first position coordinates and the second position coordinates obtains a spatial distance; The maximum relative distance and the minimum relative distance are determined according to the spatial distance; When the spatial distance is greater than or equal to the maximum relative distance, the distance risk component is determined to be 0; When the spatial distance is less than the maximum relative distance and greater than the minimum relative distance, the difference between the maximum relative distance and the spatial distance is divided by the difference between the maximum relative distance and the minimum relative distance to obtain the distance risk component; when the spatial distance is less than or equal to the minimum relative distance, the distance risk component is determined to be 1; The dot product calculation on the first moving direction vector and the second moving direction vector obtains a direction dot product value; the direction threat factor is obtained by the linear mapping function calculation on the direction dot product value, and the direction threat factor corresponding to the negative direction dot product value is greater than the direction threat factor corresponding to the positive direction dot product value; The distance risk component and the direction threat factor are multiplied to obtain the instantaneous conflict risk value at the target time point.

4. The method of claim 1, wherein, The calculation on each commodity type in a plurality of to-be-processed orders received by the unmanned store within a preset time obtains a demand heat value, and specifically includes: The target demand quantity corresponding to each commodity type in a plurality of to-be-processed orders is counted, and one target demand quantity corresponds to one commodity type. The weather information and the season information of the unmanned store at the current time are obtained, and a plurality of historical orders with the same season information as the season information are searched from a historical order database. The similarity calculation on the historical weather information of a plurality of historical orders and the weather information obtains a plurality of similarities. The historical order with the highest similarity from a plurality of historical orders is obtained as a target historical order, and the historical average demand quantity of the target historical order is obtained. The similarity of the target historical order is taken as a historical data confidence, and the historical data confidence is multiplied by a preset scaling coefficient to obtain a first weight, and the sum of the first weight and the second weight is 1. The product of the historical average demand quantity and the first weight is added to the product of the target demand quantity and the second weight to obtain the demand heat value, and one demand heat value corresponds to one commodity type.

5. The method of claim 4, wherein, The counting of the target commodity types corresponding to all channels of each candidate path in the path candidate set, and the summation of a plurality of target commodity types according to the demand heat value obtains a total demand pressure value, and specifically includes: Obtaining a plurality of channels passed through from a target candidate path, determining a plurality of target commodity types in a target channel according to the commodity layout, the target candidate path being any candidate path in the path candidate set, and the target channel being any one of the plurality of channels; Obtaining the demand heat value corresponding to each target commodity type, summing a plurality of demand heat values to obtain a channel demand pressure value of the target channel; Obtaining the channel demand pressure value corresponding to each of the channels in the target candidate path, and accumulating and summing a plurality of channel demand pressure values to obtain a total demand pressure value of the target candidate path.

6. The method of claim 5, wherein, The total demand pressure value is used to eliminate candidate paths in the path candidate set to obtain a path optimization set, specifically including: Obtaining the total demand pressure value corresponding to the target candidate path from the path candidate set; Obtaining the channel width corresponding to each of all channels passed through by the target candidate path, and selecting a path minimum channel width from a plurality of channel widths; Dividing the total demand pressure value by the path minimum channel width to obtain a path comprehensive risk score of the target candidate path; Eliminating the target candidate path in the path candidate set whose path comprehensive risk score is greater than a preset risk threshold to obtain the path optimization set.

7. The method of claim 1, wherein, After the target path is sent to the target delivery person so that the target delivery person sorts and packages the commodities in the target order according to the target path, the method further includes: Obtaining a sorting video of the target delivery person from the unmanned vending store, and determining an actual moving route according to the sorting video combined with the commodity layout; Comparing the actual moving route with the target path; If the actual moving route is consistent with the target path, filtering a plurality of sorting points from the target path, determining the target delivery person's stay duration at each sorting point combined with the sorting video, mapping the stay duration to each sorting point of the target path to obtain an actual sorting route map, establishing a corresponding relationship between the target order and the actual sorting route map, and storing the corresponding relationship in a sorting database; If the actual moving route is inconsistent with the target path, marking an inconsistent abnormal sorting point, and obtaining a target stay duration of the target delivery person at the abnormal sorting point; When the target stay duration is greater than a preset stay threshold, it is confirmed that the actual moving route is in an abnormal sorting state, and a warning information is sent to a terminal of the unmanned vending store according to the abnormal sorting state.

8. A path optimization device for order picking in an unmanned warehouse, characterized by, The device includes an obtaining unit, a verification unit, a processing unit, and a sending unit, The acquisition unit acquires commodity information from a target order, determines an unmanned store according to the commodity information, and sends the target order and an address of the unmanned store to a target delivery person, so that the target delivery person goes to the unmanned store according to the target order; and determines the target delivery person after arriving at the unmanned store, and acquires identity verification information uploaded by the target delivery person; The verification unit sends an opening instruction to the unmanned store if the identity verification information is verified correctly, and extracts a commodity type list from the target order; The processing unit calls a commodity layout map of the unmanned store, determines a placement position corresponding to each commodity in the commodity type list according to the commodity layout map, acquires a current position of the target delivery person, performs path planning on the current position and a plurality of placement positions based on a preset rule to obtain a path candidate set, and calculates each commodity type in a plurality of to-be-processed orders received by the unmanned store within a preset time to obtain a demand heat value, wherein the to-be-processed orders include the target order. The demand heat values of a plurality of target commodity types corresponding to all channels of each candidate path in the path candidate set are summed to obtain a total demand pressure value, wherein one total demand pressure value corresponds to one candidate path; the candidate paths in the path candidate set are removed according to the total demand pressure value to obtain a path optimization set; and each optimization path in the path optimization set is calculated based on personnel motion trajectories within a preset time window to obtain a dynamic congestion score, and an optimization path with a minimum dynamic congestion score is selected from the path optimization set as a target path. The sending unit sends the target path to the target delivery person, so that the target delivery person sorts and packages commodities in the target order according to the target path, and delivers the packaged commodities.

9. An electronic device, comprising: The electronic device includes a processor, a memory, a user interface, and a network interface, the memory is configured to store instructions, the user interface and the network interface are configured to communicate with other devices, and the processor is configured to execute the instructions stored in the memory to enable the electronic device to perform the method of any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the method of any one of claims 1-7.

Images