Instant retail intelligent distribution scheduling method based on space-time clustering, medium and system

Abstract

The application discloses a kind of instant retail intelligent distribution scheduling methods, medium and system based on space-time clustering, and the geographical information of hierarchical geographic unit is obtained by electronic fence and the order data to be distributed is collected, and the space-time clustering information is generated by using time window prediction model to calculate distribution time window, according to which the orders in the same geographic unit with overlapping distribution time window are aggregated into the same address task package, then based on the real-time state of rider, the task package is dynamically allocated by matching algorithm, and the algorithm parameters are updated according to the distribution feedback. The application significantly improves the accuracy of order aggregation and the rationality of task allocation, realizes the fine scheduling of transport capacity resources and the optimization and improvement of overall distribution efficiency.

Description

Technical Field

[0001] This invention relates to the field of intelligent logistics and distribution technology, specifically to a real-time intelligent retail delivery scheduling method, medium, and system based on spatiotemporal clustering. Background Technology

[0002] Instant retail delivery is a crucial component of e-commerce, with its core objective being the efficient delivery of goods within the user's desired timeframe. Currently, the industry's commonly used delivery scheduling methods rely primarily on riders accepting orders based on proximity or simple geographical partitioning, lacking effective utilization of the spatiotemporal distribution characteristics of orders. Due to the randomness and suddenness of order generation, especially during peak hours, a large number of dispersed orders can lead to multiple delivery personnel repeatedly traveling the same routes within the same area, wasting transportation resources and impacting overall delivery efficiency and timeliness. Existing technologies struggle to effectively integrate orders from nearby geographical areas within a short period and lack mechanisms for dynamic task allocation based on real-time delivery status, resulting in system lag, low resource utilization, and an inability to meet the ever-increasing demands for delivery agility and cost-effectiveness in instant retail scenarios. Therefore, a new delivery method that comprehensively optimizes order aggregation and dynamic scheduling capabilities is urgently needed to improve overall operational efficiency and enhance user experience. Summary of the Invention

[0003] In view of the above problems, the present invention provides a real-time retail intelligent delivery scheduling method, medium and system based on spatiotemporal clustering. By clustering orders with hierarchical geographical units and dynamic time windows, it generates co-location task packages and intelligently matches riders, solving the problems of wasted transportation capacity and low delivery efficiency caused by scattered orders during peak hours.

[0004] To achieve the above objectives, in a first aspect, this application provides a real-time intelligent retail delivery scheduling method based on spatiotemporal clustering, comprising:

[0005] Geographic information of hierarchical geographic units within the target area is obtained based on electronic fences. The geographic information includes community information, building groups and building levels, and delivery order data within each hierarchical geographic unit is collected in real time.

[0006] Historical delivery data is acquired, and based on the order information in the historical delivery data and the order information in the order data to be delivered, the delivery time window of each order information is calculated through the time window prediction model to obtain spatiotemporal clustering information. The order information includes the purchased goods information and the user's expected delivery time.

[0007] Based on hierarchical geographic units and spatiotemporal clustering information, a same-address task package is generated. The same-address task package is configured to include multiple orders with overlapping delivery time windows within the same geographic unit area.

[0008] Real-time acquisition of information on multiple riders, with each rider's information including location and task status;

[0009] The matching algorithm dynamically assigns delivery packages with the same address to riders who meet the preset delivery conditions. The matching algorithm is configured to comprehensively consider the rider's current location, estimated arrival time, and delivery route of the delivery package.

[0010] In addition, it acquires delivery feedback information, maps and stores the delivery feedback information with rider information, and updates the configuration parameters of the matching algorithm based on the delivery feedback information.

[0011] In some embodiments, based on order information from historical delivery data and pending order data, a time window prediction model is used to calculate the delivery time window for each order to obtain spatiotemporal clustering information, including:

[0012] Based on historical delivery data, a delivery time feature database is constructed for each community. The database includes the path time from the store to each hierarchical geographical unit and the delivery time within the building group at different time periods.

[0013] Extract product feature vectors and user feature vectors from the order data to be delivered. Product feature vectors include product category combinations and freshness requirements, while user feature vectors include historical order time preferences and address features.

[0014] The delivery time feature library, along with product feature vectors and user feature vectors, is input into the time window prediction model. The optimal departure time window and the estimated delivery time interval for each order are calculated using a multi-objective optimization algorithm. The optimal departure time window and the estimated delivery time interval are then organized into a delivery time window.

[0015] For order sets that are located in the same hierarchical geographical unit and whose time window overlap exceeds a preset overlap threshold, spatiotemporal labeling is performed to generate spatiotemporally related clustered order groups, which are denoted as spatiotemporal clustering information.

[0016] In some embodiments, the delivery time feature library, along with product feature vectors and user feature vectors, is input into a time window prediction model. A multi-objective optimization algorithm is used to calculate the optimal departure time window and the estimated delivery time interval for each order. The optimal departure time window and the estimated delivery time interval are then organized into a delivery time window, including:

[0017] Based on path time, the real-time delivery baseline time for each hierarchical geographical unit under the current traffic conditions is calculated using a dynamic path planning algorithm.

[0018] Based on the freshness requirement parameters in the product feature vector and the time preference parameters in the user feature vector, a constraint satisfaction algorithm is used to personalize the real-time delivery baseline duration and generate an initial time window set.

[0019] The initial time window set is processed using a multi-objective optimization algorithm, which simultaneously optimizes the delivery efficiency objective, the product freshness objective, and the user satisfaction objective, and outputs the optimized time window parameters for each order.

[0020] The upper limit of departure time and the lower limit of delivery time in the optimized time window parameters are combined and matched to generate the best departure time window and the expected delivery time range with a flexible range. The flexible range includes a time buffer threshold that allows rider task allocation.

[0021] The optimal departure time window and the estimated delivery time range are organized into a delivery time window;

[0022] In addition, conflict detection is performed on the generated delivery time windows. When overlapping conflicts are detected among the delivery time windows of the same rider, a time window adjustment algorithm is used to reallocate the priority of the delivery time windows.

[0023] In some embodiments, generating a co-location task package based on hierarchical geographic units and spatiotemporal clustering information includes:

[0024] Based on building groups and building levels, a spatial proximity assessment model is established to calculate the geospatial aggregation degree between each order.

[0025] Based on the degree of geographic spatial aggregation and the degree of overlap of delivery time windows, a set of orders that meet the preset aggregation conditions is selected through a spatiotemporal coupling algorithm, and denoted as the aggregated order set.

[0026] A product compatibility analysis is performed on the selected aggregated order set, and the optimal product grouping scheme is generated based on the characteristics of product category, storage temperature and packaging specifications using a combinatorial optimization algorithm.

[0027] Based on the rider's carrying capacity parameters and delivery tool type, the optimal product grouping scheme is adjusted for load balancing to form task package units that meet the actual delivery conditions;

[0028] Assign a unique identifier to each task package unit, and sort the orders within the task package unit according to the delivery route optimization algorithm to generate a same-address task package containing delivery order suggestions;

[0029] In addition, the status changes of task package units are monitored in real time, and when new orders or order cancellations occur, the affected task packages at the same address are updated through a dynamic reorganization algorithm.

[0030] In some embodiments, a matching algorithm is used to dynamically assign delivery packages with the same address to riders who meet preset delivery conditions, including:

[0031] Establish a multi-dimensional evaluation model for riders and calculate the dynamic matching index of each rider in real time. The dynamic matching index includes the path distance between the current location and the pickup point of the same address task package, the remaining delivery time of the current task, the completion efficiency of similar historical tasks, and the vehicle compatibility.

[0032] A matching matrix of same-address task packages and riders is constructed based on dynamic matching degree index and reinforcement learning algorithm. The matching weight value between each same-address task package and available riders is calculated in real time and updated to the same-address task package-rider matching matrix. The matching weight value is set according to the urgency of the task, the overlap of the delivery route and the rider's skill rating.

[0033] A distributed task allocation algorithm is used to globally optimize the same-address task package-rider matching matrix, resulting in an optimized same-address task package-rider matching matrix.

[0034] Select riders who meet the matching threshold from the same address task package-rider matching matrix, bind the same address task package to the rider's corresponding delivery task, and change the matching status of the current same address task package to match successfully.

[0035] For each successfully matched task package with the same address, a dynamic delivery route plan is generated. The dynamic delivery route plan includes pickup route planning based on real-time traffic conditions, delivery order optimization of multiple orders within the task package, and marking of expected time nodes.

[0036] In addition, when a rider is detected to have deviated from the planned route or when a delivery delay occurs, the task reassignment process is automatically triggered, and a backup rider is re-matched through a nearest scheduling algorithm.

[0037] In some embodiments, a multidimensional rider evaluation model is established to calculate the dynamic matching degree index of each rider in real time, including:

[0038] By collecting riders' dynamic location data in real time through rider terminals, and combining the road network topology, the optimal path distance between the rider's current location and the pickup point of the same address is calculated, and the actual travel time is predicted based on real-time traffic flow data.

[0039] Extract the current delivery task information of each rider from the order management system, and combine the geographical distribution of the tasks and the remaining delivery volume to calculate the remaining delivery time of the rider's current task using the time estimation method.

[0040] Analyze the historical delivery records of each rider under similar spatiotemporal conditions, and extract historical efficiency indicators for similar tasks, including on-time delivery rate and exception handling efficiency.

[0041] Based on the cargo volume, weight, and special requirements of the same address task package, and combined with the capacity parameters of the vehicles associated with each rider, the vehicle compatibility with the vehicle of the same address task package is calculated.

[0042] Based on the characteristics of the current delivery scenario, an adaptive algorithm is used to dynamically adjust the weight ratios of four dimensions: optimal path distance, remaining delivery time for the current task, efficiency indicators of historical similar tasks, and vehicle adaptability.

[0043] The optimal path distance, remaining delivery time for the current task, historical efficiency indicators of similar tasks, and vehicle suitability are weighted according to a multi-dimensional fusion algorithm to output the rider's dynamic matching index.

[0044] In some embodiments, a matching matrix for same-address task packages and riders is constructed based on a dynamic matching degree index combined with a reinforcement learning algorithm. The matching weight values ​​between each same-address task package and available riders are calculated in real time, and these weight values ​​are then updated in the same-address task package-rider matching matrix. This includes:

[0045] The characteristics of the same address task package, rider status and environmental variables are encoded into a multi-dimensional state vector. The state vector includes the spatiotemporal distribution characteristics of the same address task package, the rider's real-time location and load status, road network traffic status and weather influencing factors.

[0046] The matching relationship between riders and task packages at the same address is abstracted into a discrete set of actions. The optimal matching strategy is learned through a deep Q-network algorithm, and the matching weight value and matching strategy of each candidate matching pair are output.

[0047] In some embodiments, obtaining delivery feedback information, mapping and storing the delivery feedback information with rider information, and updating the configuration parameters of the matching algorithm based on the delivery feedback information include:

[0048] A spatiotemporal alignment algorithm is used to update delivery feedback information to the current rider's historical delivery records, generating a rider profile that includes historical delivery performance, abnormal event records, and capability assessments;

[0049] In addition, feedback features including route deviation rate, timeliness achievement, service rating fluctuation and exception handling timeliness are extracted from delivery feedback information;

[0050] Optimize the weight coefficients, priority thresholds, and path planning parameters in the deep Q-network algorithm based on feedback characteristics.

[0051] In a second aspect, the present invention also provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the method described in the first aspect.

[0052] In a third aspect, the present invention also provides an instant retail intelligent delivery scheduling system based on spatiotemporal clustering, which is applicable to the method described in the first aspect.

[0053] Unlike existing technologies, the above-mentioned technical solution provides a method, medium, and system for real-time intelligent retail delivery scheduling based on spatiotemporal clustering. It acquires geographic information of hierarchical geographic units through electronic fences and collects order data to be delivered. A time window prediction model is used to calculate delivery time windows to generate spatiotemporal clustering information. Based on this, orders with overlapping delivery time windows within the same geographic unit are aggregated into task packages with the same address. Then, based on the rider's real-time status, a matching algorithm dynamically allocates task packages, and the algorithm parameters are updated based on delivery feedback. This invention significantly improves the accuracy of order aggregation and the rationality of task allocation, achieving refined scheduling of transportation resources and optimizing overall delivery efficiency.

[0054] The above description of the invention is merely an overview of the technical solution of this application. In order to enable those skilled in the art to better understand the technical solution of this application and to implement it based on the description and drawings, and to make the above-mentioned objectives and other objectives, features and advantages of this application easier to understand, the following description is provided in conjunction with the specific embodiments and drawings of this application. Attached Figure Description

[0055] The accompanying drawings are only used to illustrate the principles, implementation methods, applications, features, and effects of specific embodiments of the present invention and other related contents, and should not be considered as limitations on this application.

[0056] In the accompanying drawings of the instruction manual:

[0057] Figure 1 This is a flowchart illustrating steps S101 to S105 of the scheduling method described in a specific implementation.

[0058] Figure 2 This is a flowchart illustrating steps S201 to S204 of the scheduling method described in a specific implementation.

[0059] Figure 3 This is a flowchart illustrating steps S301 to S305 of the scheduling method described in a specific implementation.

[0060] Figure 4 This is a flowchart illustrating steps S401 to S405 of the scheduling method described in a specific implementation.

[0061] Figure 5 The method steps S501 to S505 of the scheduling method described in the specific implementation are shown in the figure. Detailed Implementation

[0062] To illustrate the possible application scenarios, technical principles, implementable specific solutions, and achievable objectives and effects of this application in detail, the following description, in conjunction with the listed specific embodiments and accompanying drawings, provides a detailed explanation. The embodiments described herein are merely illustrative of the technical solutions of this application and are therefore intended to limit the scope of protection of this application.

[0063] In this document, the term "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The term "embodiment" appearing in various places throughout the specification does not necessarily refer to the same embodiment, nor does it specifically limit its independence or connection with other embodiments. In principle, in this application, as long as there are no technical contradictions or conflicts, the technical features mentioned in each embodiment can be combined in any way to form corresponding implementable technical solutions.

[0064] Unless otherwise defined, the technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the use of related terms herein is merely for the purpose of describing particular embodiments and is not intended to limit this application.

[0065] In the description of this application, the term "and / or" is used to describe the logical relationship between objects, indicating that three relationships can exist. For example, A and / or B means: A exists, B exists, and A and B exist simultaneously. Additionally, the character " / " in this document generally indicates that the preceding and following objects have an "or" logical relationship.

[0066] In this application, terms such as “first” and “second” are used only to distinguish one entity or operation from another, and do not necessarily require or imply any actual quantity, hierarchy or order relationship between these entities or operations.

[0067] Without further limitations, the use of terms such as “comprising,” “including,” “having,” or other similar open-ended expressions in this application is intended to cover non-exclusive inclusion, which does not exclude the presence of additional elements in a process, method, or product that includes the stated elements, such that a process, method, or product that includes a list of elements may include not only those defined elements but also other elements not expressly listed, or elements inherent to such a process, method, or product.

[0068] As understood in the Examination Guidelines, in this application, expressions such as "greater than," "less than," and "exceeding" are understood to exclude the stated number; expressions such as "above," "below," and "within" are understood to include the stated number. Furthermore, in the description of the embodiments in this application, "multiple" means two or more (including two), and similar expressions related to "multiple" are also understood in this way, such as "multiple groups" and "multiple times," unless otherwise explicitly specified.

[0069] In the description of the embodiments of this application, the space-related expressions used, such as "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "vertical," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential," indicate the orientation or positional relationship based on the orientation or positional relationship shown in the specific embodiments or drawings. They are only for the purpose of describing the specific embodiments of this application or for the reader's understanding, and do not indicate or imply that the device or component referred to must have a specific position, a specific orientation, or be constructed or operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of this application.

[0070] The processor described in the embodiments of this application can be implemented by hardware, firmware, software, or a combination thereof. It can be a circuit, one or more of an application-specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field-programmable gate array (FPGA), a central processing unit (CPU), a controller, a microcontroller, or a microprocessor. It also includes other physical, biological, or chemical structures that can implement the same or equivalent functions as the processors listed above, such as biological neurons, quantum computing units, DNA computing units, etc., so that the processor can execute some or all of the steps in the computer program or method involved in the various embodiments of this application, or any combination of the steps mentioned therein.

[0071] Please see Figure 1 In a first aspect, this embodiment provides a real-time intelligent retail delivery scheduling method based on spatiotemporal clustering, including:

[0072] S101. Obtain geographic information of hierarchical geographic units within the target area based on electronic fences. The geographic information includes community information, building groups and building levels, and collect the order data to be delivered in each hierarchical geographic unit in real time.

[0073] S102. Obtain historical delivery data, and calculate the delivery time window for each order based on the order information in the historical delivery data and the order information to be delivered, using a time window prediction model to obtain spatiotemporal clustering information. The order information includes the purchased goods information and the user's expected delivery time.

[0074] S103. Generate a same-address task package based on hierarchical geographic units and spatiotemporal clustering information. The same-address task package is configured to include multiple orders with overlapping delivery time windows within the same geographic unit area.

[0075] S104. Real-time acquisition of information on multiple riders, including location information and task status for each rider.

[0076] S105. The matching algorithm dynamically assigns the same address task package to the rider who meets the preset delivery conditions for delivery. The matching algorithm is configured to comprehensively consider the rider's current location, estimated arrival time and task package delivery route.

[0077] In addition, it acquires delivery feedback information, maps and stores the delivery feedback information with rider information, and updates the configuration parameters of the matching algorithm based on the delivery feedback information.

[0078] In step S101, the electronic fence is a virtual geographic boundary pre-defined using Geographic Information System (GIS) technology, used to automatically identify and acquire geographic information of hierarchical geographic units within the target area. Hierarchical geographic units are delivery area units divided according to spatial granularity. The unit information includes the unit name and boundary coordinates, while the building group represents a set of buildings physically adjacent to each other. The building level is refined to specific building numbers and unit information. Order data to be delivered is acquired in real time through the order management system, containing key information such as user addresses and product lists, used for subsequent clustering and scheduling.

[0079] In step S102, the time window prediction model is a machine learning-based prediction algorithm that uses time-delivery records from historical delivery data and order information from orders to be delivered for training and inference. Historical delivery data includes information such as delivery routes and actual delivery times for past orders; order information includes details about purchased goods, storage conditions, etc., and the user's expected delivery time represents the user's specified time requirement. By analyzing this data, the model calculates a flexible delivery time window for each order, thereby obtaining spatiotemporal clustering information that reflects the spatiotemporal distribution characteristics of the orders.

[0080] In step S103, the same-address task package is an order collection unit created to improve batch delivery efficiency. Its generation depends on the spatial division of hierarchical geographical units and the temporal coupling of spatiotemporal clustering information. Specifically, the system selects orders located in the same geographical unit (such as the same building group) and whose delivery time windows overlap for aggregation and packaging. This process is achieved by calculating the overlap of time windows and comparing it with a preset overlap threshold to ensure that orders within the package can be delivered by the same rider in similar time periods, thereby reducing duplicate routes and waiting time.

[0081] In step S104, rider information is reported in real time via the rider's terminal APP, location information is provided by the GPS module, and task status includes whether the rider is currently available, the amount of tasks already accepted, and the estimated completion time. This information provides a real-time basis for subsequent dynamic task allocation.

[0082] In step S105, the matching algorithm is a real-time decision-making model that considers multiple factors to assign suitable riders to generated task packages with the same address. This algorithm comprehensively considers the rider's current location, the estimated time to reach the pickup point, the optimized delivery routes for multiple orders within the task package, and the rider's current workload. By calculating the matching degree between riders and task packages and dynamically binding task packages to the optimal rider, precise scheduling of transportation resources and overall optimization of delivery routes are achieved.

[0083] Delivery feedback is collected after an order is completed, including data such as actual delivery time, user rating, and route deviation. This feedback is linked to and stored with the corresponding rider information to build a rider performance profile. Furthermore, the feedback is used to iteratively optimize the matching algorithm's configuration parameters, such as adjusting route planning weights or time estimation coefficients, thereby enabling the algorithm to have self-learning capabilities and continuously adapt to changes in the actual delivery environment.

[0084] This embodiment acquires hierarchical geographic information through electronic fences and uses a time window prediction model to generate delivery time windows and spatiotemporal clustering information for orders, thereby aggregating and generating task packages with the same address. A matching algorithm dynamically assigns suitable riders to these task packages, and the algorithm parameters are continuously optimized based on delivery feedback. This method effectively improves the accuracy of order aggregation and rider scheduling, reduces duplicate routes and waiting times, and thus significantly improves overall delivery efficiency and resource utilization.

[0085] Please see Figure 2 In some embodiments, based on order information from historical delivery data and pending order data, a time window prediction model is used to calculate the delivery time window for each order, thereby obtaining spatiotemporal clustering information, including:

[0086] S201. Construct a delivery time feature library based on historical delivery data, with each community as a unit. The delivery time feature library includes the path time from the store to each hierarchical geographical unit and the delivery time within the building group at different time periods.

[0087] S202. Extract the product feature vector and user feature vector from the order data to be delivered. The product feature vector includes product category combination and freshness requirements, and the user feature vector includes historical order time preference and address features.

[0088] S203. Input the delivery time feature library, product feature vector, and user feature vector into the time window prediction model. Calculate the optimal departure time window and estimated delivery time interval for each order through a multi-objective optimization algorithm. Organize the optimal departure time window and estimated delivery time interval into a delivery time window.

[0089] S204. For order sets that are in the same hierarchical geographical unit and whose time window overlap exceeds a preset overlap threshold, perform spatiotemporal labeling to generate clustered order groups with spatiotemporal correlation, which are denoted as spatiotemporal clustering information.

[0090] In step S201, the delivery time feature database refers to a historical delivery timeliness database built with residential communities as the basic unit. It analyzes a large amount of historical delivery data to statistically analyze and store the standard route time from each store to a specified hierarchical geographical unit (such as a residential community or building group) within different time periods, as well as the typical time required to complete the last few hundred meters of delivery within a building group. This feature database provides an important benchmark for subsequent time window predictions.

[0091] The product feature vector in step S202 is a numerical feature extracted from the order to be delivered to describe the attributes of the product. It mainly includes product category combinations (such as fresh produce, daily necessities, etc.) and corresponding preservation requirements (such as temperature control level, timeliness). The user feature vector is a set of features characterizing user preferences, including historical order time preferences derived from past order analysis (such as preference for midday delivery) and address features (such as building entrance location, whether upstairs is required, etc.). These feature vectors collectively provide input for personalized time window prediction.

[0092] In step S203, the time window prediction model receives a delivery time feature library, product feature vectors, and user feature vectors as input. The core of the model employs a multi-objective optimization algorithm, which simultaneously balances multiple objectives such as delivery efficiency, product freshness, and user satisfaction to calculate the optimal store departure time window (i.e., the suggested time range for riders to pick up goods from the store) and the estimated delivery time interval for each order. Finally, these two time intervals are combined to form a flexible delivery time window, providing a time-dimensional benchmark for order aggregation.

[0093] In step S204, the preset overlap threshold is a configurable system parameter used to determine whether the delivery time windows of two orders are close enough to be aggregated. The system calculates the degree of overlap between the delivery time windows of all orders located in the same hierarchical geographical unit (such as the same building group). When the overlap of a group of orders exceeds the preset threshold, they are considered highly correlated in time and space, and are thus marked as a clustered order group, recorded as spatiotemporal clustering information. This process achieves automatic grouping of orders in the spatiotemporal dimension, laying a solid foundation for generating same-address task packages.

[0094] This embodiment constructs a delivery time feature library and extracts product feature vectors and user feature vectors, which are then input into a time window prediction model. A multi-objective optimization algorithm is used to calculate the delivery time window for each order. The system automatically clusters orders with similar time and space based on a preset overlap threshold, generating spatiotemporal clustering information. This embodiment achieves accurate aggregation of orders in the spatiotemporal dimension, providing a core basis for subsequent packaging and scheduling, and effectively improving the efficiency of batch delivery and the rationality of route planning.

[0095] Please see Figure 3 In some embodiments, the delivery time feature library, along with product feature vectors and user feature vectors, is input into a time window prediction model. A multi-objective optimization algorithm is used to calculate the optimal departure time window and the estimated delivery time interval for each order. The optimal departure time window and the estimated delivery time interval are then organized into a delivery time window, including:

[0096] S301. Based on path consumption time, calculate the real-time delivery baseline time of each hierarchical geographical unit under the current traffic conditions using a dynamic path planning algorithm;

[0097] S302. Based on the freshness requirement parameters in the product feature vector and the time preference parameters in the user feature vector, a constraint satisfaction algorithm is used to personalize the real-time delivery baseline duration and generate an initial time window set.

[0098] S303. The initial time window set is processed using a multi-objective optimization algorithm. The multi-objective optimization algorithm simultaneously optimizes the delivery efficiency objective, the product freshness objective, and the user satisfaction objective, and outputs the optimized time window parameters for each order.

[0099] S304. Combine and match the upper limit of departure time and the lower limit of delivery time in the optimized time window parameters to generate the best departure time window and the expected delivery time range with a flexible range. The flexible range includes a time buffer threshold that allows rider task allocation.

[0100] S305: Organize the optimal departure time window and the estimated delivery time range into a delivery time window;

[0101] In addition, conflict detection is performed on the generated delivery time windows. When overlapping conflicts are detected among the delivery time windows of the same rider, a time window adjustment algorithm is used to reallocate the priority of the delivery time windows.

[0102] In step S301, the dynamic route planning algorithm is used to calculate the optimal route from the store to each hierarchical geographical unit in real time. It dynamically corrects the historical route time by accessing real-time traffic flow data, road closure events and weather conditions, thereby outputting the real-time delivery benchmark time under the current traffic conditions. This time provides an accurate and timely benchmark input for subsequent time window calculations.

[0103] In step S302, the constraint satisfaction algorithm is used to handle multiple constraints. Based on the freshness requirement parameters in the product feature vector (such as the maximum allowable delivery time for fresh products) and the time preference parameters in the user feature vector (such as the expected delivery time period specified by the user), it makes personalized adjustments to the real-time delivery benchmark time and generates an initial time window set that satisfies various constraints, ensuring that the time window conforms to both objective delivery conditions and user subjective needs.

[0104] In step S303, the multi-objective optimization algorithm simultaneously optimizes the delivery efficiency objective (e.g., minimizing total delivery time), the product freshness objective (e.g., ensuring priority and rapid delivery of perishable goods), and the user satisfaction objective (e.g., meeting user expectations as much as possible). The algorithm performs a trade-off on the initial time window set and outputs a set of optimized time window parameters, achieving an optimal balance among multiple objectives.

[0105] Step S304 combines and matches the upper limit of departure time and the lower limit of delivery time in the optimized time window parameters to generate an optimal departure time window and an estimated delivery time range with a flexible interval. This flexible interval includes a time buffer threshold that allows for rider task allocation. This threshold is a system preset value used to absorb minor fluctuations in the delivery process, providing riders with reasonable operational flexibility and enhancing the robustness of the plan to actual disturbances.

[0106] Finally, conflict detection is performed on the generated delivery time windows. When overlapping and conflicting delivery time windows are detected for tasks assigned to the same rider, a time window adjustment algorithm is used to reallocate the priority of the delivery time windows. The time window adjustment algorithm typically dynamically adjusts the order of conflicting time windows based on factors such as the urgency of the order, the value of the goods, or the user's level, thereby ensuring that the task packages ultimately assigned to the riders are feasible and efficient in terms of time.

[0107] This embodiment calculates the real-time delivery baseline time using a dynamic path planning algorithm and employs constraint satisfaction and multi-objective optimization algorithms to generate a flexible time window that integrates delivery efficiency, product freshness, and user preferences. The solution introduces a time buffer threshold to enhance robustness and ensures scheduling feasibility through conflict detection and priority adjustment. This embodiment significantly improves the accuracy, personalization, and practical feasibility of time window prediction, providing core support for efficient rider task allocation.

[0108] Please see Figure 4 In some embodiments, generating a co-address task package based on hierarchical geographic units and spatiotemporal clustering information includes:

[0109] S401. Based on building groups and building levels, establish a spatial proximity assessment model and calculate the geospatial aggregation degree between each order;

[0110] S402. Based on the degree of geospatial aggregation and the degree of overlap of delivery time windows, a set of orders that meet the preset aggregation conditions is selected through a spatiotemporal coupling algorithm and denoted as the aggregated order set.

[0111] S403. Perform product compatibility analysis on the selected aggregated order set, and generate the optimal product grouping scheme based on product category, storage temperature and packaging specifications using a combinatorial optimization algorithm.

[0112] S404. Based on the rider's carrying capacity parameters and delivery tool type, the optimal product grouping scheme is adjusted for load balancing to form task package units that meet the actual delivery conditions.

[0113] S405. Assign a unique identifier to each task package unit, and sort the orders within the task package unit according to the delivery route optimization algorithm to generate a same-address task package containing delivery order suggestions.

[0114] In addition, the status changes of task package units are monitored in real time, and when new orders or order cancellations occur, the affected task packages at the same address are updated through a dynamic reorganization algorithm.

[0115] In step S401, the spatial proximity assessment model is a calculation model used to quantify the degree of geographical proximity between different orders. It calculates the actual physical distance or path reachability distance between order delivery addresses through the geocoding information of building groups and building levels, thereby outputting a geospatial aggregation index. This index provides a quantitative basis for the spatial dimension of subsequent order aggregation.

[0116] In step S402, the spatiotemporal coupling algorithm is used to comprehensively evaluate the geospatial aggregation degree and the overlap of delivery time windows among orders. It filters orders by setting preset aggregation conditions (such as spatial distance being lower than a threshold and time window overlap being higher than a threshold), and marks the set of orders that simultaneously meet the spatiotemporal similarity conditions as an aggregated order set, thereby achieving preliminary clustering of orders in the spatiotemporal dimension.

[0117] In step S403, product compatibility analysis, based on characteristics such as product category, storage temperature, and packaging specifications, determines whether products from multiple orders are suitable for combined delivery. A combinatorial optimization algorithm is used to generate the optimal product grouping scheme. Its goal is to maximize the variety and quantity of products in a single delivery while meeting product storage requirements (such as avoiding cross-contamination of odors and temperature conflicts) and packaging limitations, thereby improving delivery consolidation efficiency.

[0118] Step S404 adjusts the aforementioned product grouping scheme for load balancing based on the rider's carrying capacity parameters (such as the volume of the electric vehicle compartment and maximum load capacity) and the type of delivery tool (such as the configuration of the insulated box). This step ensures that the final task package unit meets the actual delivery equipment's carrying capacity limits in terms of volume, weight, and temperature control, avoiding overloading or equipment incompatibility issues.

[0119] In step S405, a unique identifier is assigned to each task package unit for traceability management. A delivery route optimization algorithm is then used to sort the orders within the package, generating task packages with the same address that include delivery order suggestions. Preferably, the delivery order suggestions aim to achieve the shortest path or the fewest number of trips up and down stairs. Finally, a dynamic reorganization algorithm monitors the task package status in real time. When orders increase or decrease, the composition and order of the affected task packages are adjusted promptly to ensure the real-time effectiveness and executability of the task packages. The entire process implements a complete task package generation and dynamic update mechanism, from order clustering, product compatibility assessment, load adaptation to route optimization.

[0120] This embodiment achieves spatiotemporal clustering of orders through spatial proximity evaluation and spatiotemporal coupling algorithms. Combined with product compatibility analysis and load balancing adjustments, it generates task packages with the same address that meet actual delivery conditions. Unique identifiers and route optimization ensure delivery efficiency, and a dynamic reorganization algorithm is used to handle order changes. This embodiment effectively improves the accuracy of order aggregation and the executability of task packages, significantly reduces duplicate routes, optimizes capacity allocation, and lowers delivery costs.

[0121] Please see Figure 5 In some embodiments, a matching algorithm dynamically assigns delivery packages with the same address to riders who meet preset delivery conditions, including:

[0122] S501. Establish a multi-dimensional evaluation model for riders and calculate the dynamic matching degree index of each rider in real time. The dynamic matching degree index includes the path distance between the current location and the pickup point of the same address task package, the remaining delivery time of the current task, the completion efficiency of similar historical tasks, and the vehicle adaptability.

[0123] S502. Construct a matching matrix of same address task packages and riders based on dynamic matching degree index and reinforcement learning algorithm. Calculate the matching weight value between each same address task package and available riders in real time and update the matching weight value to the same address task package-rider matching matrix. The matching weight value is set according to the urgency of the task, the overlap of the delivery route and the rider's skill rating.

[0124] S503. A distributed task allocation algorithm is used to globally optimize the same-address task package-rider matching matrix to obtain the optimized same-address task package-rider matching matrix.

[0125] S504. Select riders who meet the matching threshold in the same address task package-rider matching matrix, bind the same address task package to the delivery task corresponding to the rider, and change the matching status of the current same address task package to match successfully.

[0126] S505. Generate a dynamic delivery route plan for each successfully matched same-address task package. The dynamic delivery route plan includes pickup route planning based on real-time traffic conditions, delivery order optimization of multiple orders within the task package, and marking of expected time nodes.

[0127] In addition, when a rider is detected to have deviated from the planned route or when a delivery delay occurs, the task reassignment process is automatically triggered, and a backup rider is re-matched through a nearest scheduling algorithm.

[0128] In step S501, the rider multidimensional evaluation model is a computational model used to comprehensively evaluate the rider's delivery capabilities. It collects information in real time, including the rider's current location, remaining task time, historical efficiency data, and vehicle status, and generates a dynamic matching degree index through weighted calculation. This index quantitatively reflects the rider's suitability for performing specific same-address task packages, providing an objective basis for task allocation.

[0129] In step S502, a reinforcement learning algorithm is used to construct and continuously optimize the same-address task package-rider matching matrix. The algorithm dynamically calculates matching weight values ​​based on task urgency (e.g., priority for fresh food orders), delivery route overlap (e.g., proximity to the route), and rider skill rating (e.g., familiarity with the area), and updates the matrix with the real-time calculation results, forming a multi-dimensional weighted matching relationship network.

[0130] In step S503, the distributed task allocation algorithm optimizes the global matching matrix. By processing the matching combination of multiple task packages and riders in parallel, it solves the optimal allocation scheme while satisfying the goal of maximizing overall delivery efficiency. This effectively avoids local optima and improves the computational efficiency and rationality of large-scale task allocation.

[0131] Step S504 filters qualified riders from the optimized matrix based on a preset matching threshold, binds task packages with the same address to their delivery tasks, and updates the task package status to "matched successfully." The matching threshold can be dynamically adjusted according to the supply and demand of delivery capacity; for example, the threshold can be appropriately relaxed during peak hours to improve the matching success rate.

[0132] Step S505 generates a dynamic delivery route plan for each successfully matched task package. This plan integrates real-time traffic prediction, order delivery sequence optimization within the package, and precise time node calculation to provide riders with full-process navigation guidance. When the system detects a rider deviating from the route or experiencing a delay via GPS positioning, it automatically triggers a task reassignment process. A nearest-neighbor scheduling algorithm is used to quickly match a replacement rider from the backup rider pool, ensuring efficient completion of delivery tasks and system robustness. The entire process achieves a complete closed loop of task allocation, from rider evaluation, intelligent matching, global optimization to dynamic scheduling.

[0133] This embodiment constructs a dynamic matching matrix using a multi-dimensional rider evaluation model and reinforcement learning algorithms, employs a distributed optimization algorithm to intelligently match task packages with riders, and generates dynamic delivery route plans. When delivery anomalies occur, the system automatically triggers a nearby reassignment mechanism. This embodiment significantly improves task allocation accuracy and delivery efficiency, ensuring optimal utilization of transportation resources, while a dynamic fault-tolerance mechanism guarantees the reliability and robustness of the delivery system.

[0134] In some embodiments, a multidimensional rider evaluation model is established to calculate the dynamic matching degree index of each rider in real time, including:

[0135] By collecting riders' dynamic location data in real time through rider terminals, and combining the road network topology, the optimal path distance between the rider's current location and the pickup point of the same address is calculated, and the actual travel time is predicted based on real-time traffic flow data.

[0136] Extract the current delivery task information of each rider from the order management system, and combine the geographical distribution of the tasks and the remaining delivery volume to calculate the remaining delivery time of the rider's current task using the time estimation method.

[0137] Analyze the historical delivery records of each rider under similar spatiotemporal conditions, and extract historical efficiency indicators for similar tasks, including on-time delivery rate and exception handling efficiency.

[0138] Based on the cargo volume, weight, and special requirements of the same address task package, and combined with the capacity parameters of the vehicles associated with each rider, the vehicle compatibility with the vehicle of the same address task package is calculated.

[0139] Based on the characteristics of the current delivery scenario, an adaptive algorithm is used to dynamically adjust the weight ratios of four dimensions: optimal path distance, remaining delivery time for the current task, efficiency indicators of historical similar tasks, and vehicle adaptability.

[0140] The optimal path distance, remaining delivery time for the current task, historical efficiency indicators of similar tasks, and vehicle suitability are weighted according to a multi-dimensional fusion algorithm to output the rider's dynamic matching index.

[0141] In this embodiment, the optimal path distance is the actual reachable distance from the rider's current location to the pickup point, calculated based on real-time road network topology and traffic flow data. This calculation utilizes a real-time traffic API from a third-party map service, combined with parameters such as road grade and congestion coefficient, to perform dynamic route planning, thereby accurately reflecting the rider's actual travel cost to the pickup point.

[0142] The remaining delivery time for the current task is an indicator derived from a comprehensive analysis of the rider's current workload using a time-prediction method. This method takes into account the geographical distribution density of tasks, historical data on the average delivery time per order, and current traffic conditions. It uses a regression model to predict the time required for the rider to complete the current task, providing a time dimension for assessing the rider's order-taking ability.

[0143] Historical performance metrics for similar tasks are performance data extracted from riders' historical delivery records. These metrics mainly include key indicators such as on-time delivery rate and timeliness of handling abnormal orders. Data mining techniques are used to analyze riders' historical performance under similar weather, time, and regional conditions to form an objective evaluation of their delivery capabilities.

[0144] Vehicle compatibility is a matching index derived by comparing the characteristics of the goods in the task package with the parameters of the rider's vehicle. Based on the total volume, weight, and temperature control requirements of the goods recorded in the task package, the matching degree is calculated with the box volume, load limit, and equipment configuration of the rider's registered vehicle to ensure that the vehicle's capabilities match the delivery needs.

[0145] The adaptive weight adjustment algorithm dynamically optimizes the weight ratio of each dimension based on the characteristics of the real-time delivery scenario. By monitoring environmental factors such as the current supply and demand ratio of transportation capacity, weather conditions, and regional order density, it automatically adjusts the contribution of each dimension in the matching degree calculation using the fuzzy logic control principle, enabling the evaluation model to have environmental adaptability.

[0146] The multi-dimensional fusion algorithm synthesizes the evaluation values ​​of the four dimensions into a final dynamic matching degree index through a weighted summation method. After standardization, the data of each dimension are linearly weighted according to the weight ratio determined by the adaptive algorithm, and finally output a matching degree score in the range of 0-1, providing a quantitative decision basis for rider selection.

[0147] This embodiment collects multi-dimensional data such as rider location, task load, historical performance, and vehicle parameters in real time, and uses an adaptive weighting algorithm to dynamically adjust the contribution ratio of each dimension. Finally, it outputs the rider's dynamic matching degree index through a multi-dimensional fusion algorithm, which realizes a precise quantitative assessment of the rider's delivery ability, significantly improves the accuracy and efficiency of task allocation, and ensures the optimal matching of transportation resources and delivery needs.

[0148] In some embodiments, a matching matrix for same-address task packages and riders is constructed based on a dynamic matching degree index combined with a reinforcement learning algorithm. The matching weight values ​​between each same-address task package and available riders are calculated in real time, and these weight values ​​are then updated in the same-address task package-rider matching matrix. This includes:

[0149] The characteristics of the same address task package, rider status and environmental variables are encoded into a multi-dimensional state vector. The state vector includes the spatiotemporal distribution characteristics of the same address task package, the rider's real-time location and load status, road network traffic status and weather influencing factors.

[0150] The matching relationship between riders and task packages at the same address is abstracted into a discrete set of actions. The optimal matching strategy is learned through a deep Q-network algorithm, and the matching weight value and matching strategy of each candidate matching pair are output.

[0151] In this embodiment, the multidimensional state vector is a feature set formed by digitally encoding the characteristics of the same address task package, rider status, and environmental variables. Specifically, it includes the geographical distribution density of the task package, time urgency, rider's real-time location coordinates, current number of tasks, road congestion index, and weather influencing factors such as temperature and humidity. Through feature engineering methods, heterogeneous data is transformed into standardized numerical features, providing a comprehensive state representation for matching decisions.

[0152] The discrete action set defines the matching relationship between each available rider and the task package at the same address as an independent action that can be selected. By arranging and combining the candidate rider list with the task package to be assigned, all possible matching pairs are generated as the candidate action space, thus transforming the matching problem into the action selection problem in reinforcement learning.

[0153] The Deep Q-Network algorithm is a reinforcement learning method that uses deep neural networks to approximate the Q-value function. This network takes a multi-dimensional state vector as input and calculates the long-term expected reward value (matching weight) for each candidate action (matching pair) through multiple layers of nonlinear transformations. The training process is stabilized through experience replay and a target network mechanism, ultimately outputting the optimal matching strategy and its corresponding matching weight value. By continuously collecting feedback data on actual delivery performance, using improved delivery efficiency and user satisfaction as reward signals, the network parameters are continuously optimized, enabling the matching matrix to have a continuous self-evolutionary capability.

[0154] This embodiment models the delivery matching problem as a reinforcement learning task and uses a deep Q-network algorithm to dynamically calculate the optimal matching weight between riders and task packages. It can adapt to complex and ever-changing delivery environments, achieve continuous self-optimization of matching strategies, significantly improve the allocation efficiency of task packages with the same address and the overall system response speed, and reduce delivery costs and improve user satisfaction through intelligent matching.

[0155] In some embodiments, obtaining delivery feedback information, mapping and storing the delivery feedback information with rider information, and updating the configuration parameters of the matching algorithm based on the delivery feedback information include:

[0156] A spatiotemporal alignment algorithm is used to update delivery feedback information to the current rider's historical delivery records, generating a rider profile that includes historical delivery performance, abnormal event records, and capability assessments;

[0157] In addition, feedback features including route deviation rate, timeliness achievement, service rating fluctuation and exception handling timeliness are extracted from delivery feedback information;

[0158] Optimize the weight coefficients, priority thresholds, and path planning parameters in the deep Q-network algorithm based on feedback characteristics.

[0159] In this embodiment, the spatiotemporal alignment algorithm accurately associates and matches delivery feedback information with rider trajectory data based on timestamps and geographic location information. By unifying the time base and coordinate transformation, it binds feedback data such as delivery completion time, route trajectory points, and user reviews with rider identity identifiers to ensure the spatiotemporal consistency of data mapping.

[0160] Rider profiles refer to digital profiles of riders' capabilities formed by aggregating historical delivery data. These profiles include historical performance data such as delivery on-time rate, timeliness of handling abnormal orders, and service rating trends, as well as records of abnormal events such as equipment failures and traffic control. Through the integration of multi-dimensional data, a comprehensive evaluation system for rider delivery capabilities is formed.

[0161] Feedback features refer to quantitative evaluation indicators extracted from delivery feedback information. Among them, the route deviation rate is calculated by comparing the difference between the actual trajectory and the planned route; the timeliness achievement rate is determined based on the ratio of the actual delivery time to the estimated time; the service rating fluctuation is calculated by using a sliding window to statistically determine the standard deviation of user ratings; and the anomaly handling timeliness records the time interval from the occurrence of an anomaly to its resolution. These features are processed through data cleaning and standardization to form a quantifiable evaluation vector.

[0162] The weight coefficient optimization step dynamically adjusts the contribution weights of each state feature in the deep Q-network based on feedback characteristics. The feature weight gradient is calculated using the backpropagation algorithm, and parameters are updated with the objective functions of maximizing delivery efficiency and optimizing user satisfaction. Priority threshold adjustment sets matching priority boundaries based on historical delivery performance distribution, while route planning parameters are dynamically updated using the road network cost function based on actual route efficiency. Through continuous parameter optimization, the matching algorithm acquires online learning capabilities, continuously improving the overall performance of the delivery system.

[0163] This embodiment achieves refined evaluation of the delivery process by establishing rider profiles and a feedback feature extraction mechanism. It ensures data accuracy through a spatiotemporal alignment algorithm and enables the matching algorithm to adaptively evolve by continuously optimizing the configuration parameters of the deep Q-network. This embodiment significantly improves the learning efficiency and matching accuracy of the delivery system, effectively reduces the delivery anomaly rate, and optimizes the rider delivery experience and user service satisfaction through personalized parameter adjustments.

[0164] In a second aspect, this embodiment also provides a computer-readable storage medium storing computer program instructions thereon, which, when executed by a processor, implement the method described in the first aspect.

[0165] The computer program involved in this embodiment can be stored in a computer device readable storage medium, which includes, but is not limited to, disks, magnetic tapes, magnetic cards, floppy disks, flash memory, optical disks, optical cards, read-only memory (ROM), random access memory (RAM), erasable programmable ROM (EPROM), and electrically erasable programmable ROM (EEPROM), etc. It also includes other biological, physical, or chemical structures capable of performing similar or equivalent functions to the storage media listed above, such as DNA, RNA, proteins, and other units with information storage capabilities. In specific embodiments, the storage medium involved can be one of the above-mentioned media types or a combination of the above media types. In different embodiments, the computer program involved in the embodiment can be centrally stored in a single medium or distributed across multiple media. The memory containing the computer device readable storage medium can be non-volatile memory or random access memory. These computer device readable storage media can be built into the device or connected to the device involved in the embodiment as an external device or part of an external device. In some embodiments, the memory having a computer device readable storage medium is deployed locally; in other embodiments, the memory may be deployed remotely from the processor, for example, as a network-attached memory accessed via RF circuitry or an external port and a communication network, wherein the communication network may be the Internet, one or more intranets, a local area network (LAN), a wide area network (WLAN), a storage area network (SAN), or a suitable combination thereof, as long as computer device access to the memory is enabled. Furthermore, the computer program involved in the embodiments may be stored in plaintext / ciphertext form, or it may be designed as training data, integrated and recombined through model training and implicitly stored in the parameter states of a deep neural network or other machine learning model.

[0166] In a third aspect, this embodiment also provides an instant retail intelligent delivery scheduling system based on spatiotemporal clustering, which is applicable to the method described in the first aspect.

[0167] In this embodiment, the real-time retail intelligent delivery scheduling system based on spatiotemporal clustering is a comprehensive software and hardware platform that integrates order aggregation, route planning and dynamic matching functions. It is deployed on a cloud server cluster and achieves high-concurrency data processing through a distributed computing architecture. It includes an order access module, a geofencing engine, a time window calculation unit and an intelligent matching core.

[0168] The order access module receives and parses retail order data from merchants in real time, extracting key fields such as product information, delivery address, and expected delivery time. The geofencing engine, based on a pre-defined hierarchical geographic unit database, uses a polygon fence algorithm to spatially match delivery addresses, achieving precise three-level positioning: community-building group-building. The time window calculation unit dynamically calculates the optimal delivery time window for each order by analyzing historical fulfillment data and combining traffic conditions and product characteristics.

[0169] The intelligent matching core employs a multi-threaded parallel computing architecture, receiving rider location information and task package status in real time, and generating rider-task package matching schemes through operations research optimization algorithms. The system interacts with external map services and weather APIs via a data bus to ensure the real-time nature and accuracy of scheduling decisions. Optionally, the system supports users initiating group order invitations, inviting other users within the same geographical unit to place orders together. When the order volume reaches a preset threshold, the system automatically triggers the process of generating task packages at the same address.

[0170] The system described in this embodiment executes the method described in the first aspect, significantly reducing repeated route deliveries, improving the delivery efficiency of a single rider, reducing delivery costs while increasing user satisfaction, and achieving overall efficient utilization of transportation resources and improved quality of delivery services.

[0171] Unlike existing technologies, the above technical solutions offer the following advantages: They effectively address the core issues of dispersed delivery capacity and redundant routes in instant retail delivery through hierarchical geographical units and spatiotemporal clustering. Precise geolocation is achieved using electronic fences, and flexible delivery windows are generated by combining time window prediction models to intelligently aggregate orders with the same address into task packages. Dynamic matching algorithms comprehensively consider rider location, load status, and real-time road conditions to achieve optimal allocation of delivery resources. The system also possesses self-learning capabilities, continuously optimizing algorithm parameters to adapt to changes in the delivery environment. These technical solutions significantly improve the delivery efficiency of individual riders and order aggregation, reduce redundant routes and wasted delivery capacity, lower delivery costs while increasing user satisfaction, and form an efficient and sustainable intelligent delivery system.

[0172] Finally, it should be noted that although the above embodiments have been described in the text and drawings of this application, this should not limit the scope of patent protection of this application. Any technical solutions that are based on the essential concept of this application and utilize the content described in the text and drawings of this application, resulting in equivalent structural or procedural substitutions or modifications, as well as the direct or indirect application of the technical solutions of the above embodiments to other related technical fields, are all included within the scope of patent protection of this application.

Claims

1. A real-time intelligent retail delivery scheduling method based on spatiotemporal clustering, characterized in that, include: Geographic information of hierarchical geographic units within the target area is obtained based on electronic fences. The geographic information includes community information, building groups and building levels, and delivery order data within each hierarchical geographic unit is collected in real time. Historical delivery data is acquired, and based on the order information in the historical delivery data and the pending order data, the delivery time window for each order is calculated using a time window prediction model to obtain spatiotemporal clustering information. The order information includes purchased product information and the user's expected delivery time, including: Based on the historical delivery data, a delivery time feature database is constructed at the community level. The delivery time feature database includes the path time from the store to each hierarchical geographical unit and the delivery time within the building group at different time periods. Extract the product feature vector and user feature vector from the order data to be delivered. The product feature vector includes product category combination and freshness requirements, and the user feature vector includes historical order time preference and address features. The delivery time feature library, along with the product feature vector and user feature vector, are input into the time window prediction model. The optimal departure time window and the expected delivery time interval for each order are calculated using a multi-objective optimization algorithm. The optimal departure time window and the expected delivery time interval are then organized into a delivery time window. Order sets located in the same hierarchical geographical unit and whose time window overlap exceeds a preset overlap threshold are spatiotemporally labeled to generate spatiotemporally related clustered order groups, which are denoted as the spatiotemporal clustering information. Based on the hierarchical geographic units and spatiotemporal clustering information, a same-address task package is generated. The same-address task package is configured to include multiple orders with overlapping delivery time windows within the same geographic unit area. Real-time acquisition of information on multiple riders, each rider's information including location information and task status; The same address task package is dynamically assigned to riders who meet the preset delivery conditions through a matching algorithm. The matching algorithm is configured to comprehensively consider the rider's current location, estimated arrival time and task package delivery route. In addition, the system acquires delivery feedback information, maps and stores the delivery feedback information with the rider information, and updates the configuration parameters of the matching algorithm based on the delivery feedback information.

2. The real-time intelligent retail delivery scheduling method based on spatiotemporal clustering according to claim 1, characterized in that, The delivery time feature library, along with the product feature vector and user feature vector, are input into a time window prediction model. A multi-objective optimization algorithm is used to calculate the optimal departure time window and estimated delivery time interval for each order. These optimal departure time windows and estimated delivery time intervals are then organized into a delivery time window, including: Based on path time, the real-time delivery baseline time for each hierarchical geographical unit under the current traffic conditions is calculated using a dynamic path planning algorithm. Based on the freshness requirement parameters in the product feature vector and the time preference parameters in the user feature vector, a constraint satisfaction algorithm is used to personalize the real-time delivery baseline duration and generate an initial time window set. The initial time window set is processed using a multi-objective optimization algorithm, which simultaneously optimizes the delivery efficiency objective, the product freshness objective, and the user satisfaction objective, and outputs the optimized time window parameters for each order. The upper limit of departure time and the lower limit of delivery time in the optimized time window parameters are combined and matched to generate an optimal departure time window and an expected delivery time range with a flexible range. The flexible range includes a time buffer threshold that allows rider task allocation. The optimal departure time window and the estimated delivery time interval are organized into a delivery time window; In addition, conflict detection is performed on the generated delivery time windows. When overlapping conflicts are detected among the delivery time windows of the same rider, a time window adjustment algorithm is used to reallocate the priority of the delivery time windows.

3. The real-time intelligent retail delivery scheduling method based on spatiotemporal clustering according to claim 1, characterized in that, Based on the hierarchical geographic units and spatiotemporal clustering information, a task package with the same address is generated, including: Based on building groups and building levels, a spatial proximity assessment model is established to calculate the geospatial aggregation degree between each order. Based on the degree of geographic spatial aggregation and the degree of overlap of delivery time windows, a set of orders that meet the preset aggregation conditions is selected through a spatiotemporal coupling algorithm, and denoted as the aggregated order set. A product compatibility analysis is performed on the selected aggregated order set, and the optimal product grouping scheme is generated based on the characteristics of product category, storage temperature and packaging specifications using a combinatorial optimization algorithm. Based on the rider's carrying capacity parameters and delivery tool type, the optimal product grouping scheme is adjusted for load balancing to form task package units that meet the actual delivery conditions. Assign a unique identifier to each task package unit, and sort the orders within the task package unit according to the delivery route optimization algorithm to generate a same-address task package containing delivery order suggestions; In addition, the status changes of task package units are monitored in real time, and when new orders or order cancellations occur, the affected task packages at the same address are updated through a dynamic reorganization algorithm.

4. The real-time intelligent retail delivery scheduling method based on spatiotemporal clustering according to claim 1, characterized in that, The matching algorithm dynamically assigns the same-address task packages to riders who meet preset delivery conditions for delivery, including: Establish a multi-dimensional evaluation model for riders and calculate the dynamic matching degree index of each rider in real time. The dynamic matching degree index includes the path distance between the current location and the pickup point of the same address task package, the remaining delivery time of the current task, the completion efficiency of similar historical tasks, and the vehicle adaptability. Based on the dynamic matching degree index and reinforcement learning algorithm, a same-address task package-rider matching matrix is ​​constructed. By calculating the matching weight value between each same-address task package and available riders in real time, the matching weight value is updated to the same-address task package-rider matching matrix. The matching weight value is set according to the urgency of the task, the overlap of the delivery route and the rider's skill rating. A distributed task allocation algorithm is used to globally optimize the same-address task package-rider matching matrix, resulting in an optimized same-address task package-rider matching matrix. Select riders who meet the matching threshold from the same address task package-rider matching matrix, bind the same address task package to the delivery task corresponding to the rider, and change the matching status of the current same address task package to match successfully. For each successfully matched task package with the same address, a dynamic delivery route plan is generated. The dynamic delivery route plan includes pickup route planning based on real-time traffic conditions, delivery order optimization of multiple orders within the task package, and marking of expected time nodes. In addition, when a rider is detected to have deviated from the planned route or when a delivery delay occurs, the task reassignment process is automatically triggered, and a backup rider is re-matched through a nearest scheduling algorithm.

5. The real-time intelligent retail delivery scheduling method based on spatiotemporal clustering according to claim 4, characterized in that, Establish a multi-dimensional rider evaluation model to calculate the dynamic matching index of each rider in real time, including: By collecting riders' dynamic location data in real time through rider terminals, and combining the road network topology, the optimal path distance between the rider's current location and the pickup point of the same address is calculated, and the actual travel time is predicted based on real-time traffic flow data. Extract the current delivery task information of each rider from the order management system, and combine the geographical distribution of the tasks and the remaining delivery volume to calculate the remaining delivery time of the rider's current task using the time estimation method. Analyze the historical delivery records of each rider under similar spatiotemporal conditions, and extract historical efficiency indicators for similar tasks, including on-time delivery rate and exception handling efficiency. Based on the cargo volume, weight, and special requirements of the same address task package, and combined with the capacity parameters of the vehicles associated with each rider, the vehicle compatibility with the vehicle of the same address task package is calculated. Based on the characteristics of the current delivery scenario, an adaptive algorithm is used to dynamically adjust the weight ratios of four dimensions: optimal path distance, remaining delivery time for the current task, efficiency indicators of historical similar tasks, and vehicle adaptability. The optimal path distance, remaining delivery time for the current task, historical efficiency indicators of similar tasks, and vehicle suitability are weighted according to a multi-dimensional fusion algorithm to output the rider's dynamic matching index.

6. The real-time intelligent retail delivery scheduling method based on spatiotemporal clustering according to claim 4, characterized in that, Based on the dynamic matching degree index and reinforcement learning algorithm, a same-address task package-rider matching matrix is ​​constructed. The matching weight values ​​between each same-address task package and available riders are calculated in real time, and these matching weight values ​​are updated in the same-address task package-rider matching matrix, including: The characteristics of the same address task package, rider status and environmental variables are encoded into a multi-dimensional state vector. The state vector includes the spatiotemporal distribution characteristics of the same address task package, the rider's real-time location and load status, road network traffic status and weather influencing factors. The matching relationship between riders and the same address task packets is abstracted into a discrete action set. The optimal matching strategy is learned through a deep Q-network algorithm, and the matching weight value and matching strategy of each candidate matching pair are output.

7. The real-time intelligent retail delivery scheduling method based on spatiotemporal clustering according to claim 6, characterized in that, Obtaining delivery feedback information, mapping and storing the delivery feedback information with the rider information, and updating the configuration parameters of the matching algorithm based on the delivery feedback information, including: A spatiotemporal alignment algorithm is used to update delivery feedback information to the current rider's historical delivery records, generating a rider profile that includes historical delivery performance, abnormal event records, and capability assessments; In addition, feedback features including route deviation rate, timeliness achievement, service rating fluctuation and anomaly handling timeliness are extracted from the delivery feedback information; Optimize the weight coefficients, priority thresholds, and path planning parameters in the deep Q-network algorithm based on feedback characteristics.

8. A computer-readable storage medium storing computer program instructions thereon, characterized in that, The computer program instructions, when executed by a processor, implement the method as described in any one of claims 1 to 7.

9. A real-time intelligent retail delivery scheduling system based on spatiotemporal clustering, characterized in that, The method applicable to any one of claims 1 to 7.

Images