A logistics scheduling method and system for electronic commerce trade

Abstract

The application provides a logistics scheduling method and system for electronic commerce trade, and relates to the technical field of computers.The method comprises the following steps: obtaining remote package logistics information, and obtaining at least one lift sharing reservation information corresponding to a remote area as a travel destination; performing preliminary screening on each lift sharing reservation information by using package attributes, package expected arrival time, package delivery address and distribution station address to determine at least one candidate lift sharing reservation information matched; determining target lift sharing reservation information from each candidate lift sharing reservation information based on a logistics scheduling matching model; updating the driving path in the target lift sharing reservation information according to the distribution station address and the package delivery address, and sending a logistics scheduling instruction to a corresponding car owner client. Thus, the package in a remote area is transported by using lift sharing collaborative scheduling, which not only greatly shortens the distribution time, but also greatly reduces the transportation cost.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a logistics scheduling method and system for e-commerce trade. Background Technology

[0002] With the rapid development of e-commerce, the logistics and distribution system has gradually become one of the core supports for the efficient operation of e-commerce businesses. However, in the entire logistics and distribution process, the "last mile" delivery link is particularly critical, as it directly determines the quality of user experience and the success of logistics services.

[0003] Currently, for "last-mile" delivery, delivery stations are set up within a certain area. Packages are centrally delivered to these stations, and then delivery personnel deliver them door-to-door. Alternatively, smart parcel lockers are deployed in densely populated residential areas, allowing users to pick up their packages themselves. These models are highly effective in densely populated urban environments because they leverage economies of scale to reduce per-package delivery costs.

[0004] However, compared to cities, remote areas have dispersed populations and a limited number of parcels, making it difficult to achieve economies of scale. The delivery distance for a single parcel is long, and some areas have complex terrain (such as mountainous or hilly areas), resulting in long delivery times and high transportation costs. Setting up dedicated delivery stations or dispatching dedicated delivery personnel to remote areas is prohibitively expensive and unsustainable in the long run. To reduce costs, logistics companies typically adopt periodic delivery (such as weekly concentrated deliveries), significantly extending users' waiting times.

[0005] Currently, the industry has not proposed a better technical solution to the above problems. Summary of the Invention

[0006] This invention provides a logistics scheduling method and system for e-commerce trade, which at least solves the problems of long delivery times and high transportation costs in remote areas in the prior art.

[0007] In a first aspect, embodiments of the present invention provide a logistics scheduling method for e-commerce trade, comprising: acquiring remote parcel logistics information, the remote parcel logistics information including parcel delivery address, parcel attributes, timeliness urgency, and expected arrival time of the parcel in a corresponding remote area; acquiring at least one ridesharing reservation information with the corresponding trip destination in the remote area, the ridesharing reservation information including basic vehicle information, driving route, departure time, and logistics scheduling historical order records; performing preliminary screening on each of the ridesharing reservation information using the parcel attributes, the expected arrival time of the parcel, the parcel delivery address, and the delivery station address to determine at least one matching candidate ridesharing reservation information; inputting each of the candidate ridesharing reservation information and the remote parcel logistics information into a logistics scheduling matching model to determine a target ridesharing reservation information from the candidate ridesharing reservation information; updating the driving route in the target ridesharing reservation information according to the delivery station address and the parcel delivery address, and sending a logistics scheduling instruction to the corresponding driver's client.

[0008] Secondly, embodiments of the present invention provide a logistics scheduling system for e-commerce trade, comprising: a first acquisition unit, configured to acquire remote parcel logistics information, the remote parcel logistics information including the parcel delivery address, parcel attributes, timeliness urgency, and expected arrival time of the parcel in a corresponding remote area; a second acquisition unit, configured to acquire at least one ridesharing reservation information corresponding to the remote area as the destination of the trip, the ridesharing reservation information including basic vehicle information, driving route, departure time, and logistics scheduling historical order records; a reservation preliminary screening unit, configured to perform preliminary screening on each of the ridesharing reservation information based on the parcel attributes, the expected arrival time of the parcel, the parcel delivery address, and the delivery station address, to determine at least one matching candidate ridesharing reservation information; a scheduling matching unit, configured to input each of the candidate ridesharing reservation information and the remote parcel logistics information into a logistics scheduling matching model to determine the target ridesharing reservation information from the candidate ridesharing reservation information; and a scheduling execution unit, configured to update the driving route in the target ridesharing reservation information according to the delivery station address and the parcel delivery address, and send a logistics scheduling instruction to the corresponding driver client.

[0009] Thirdly, embodiments of the present invention provide an electronic device comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the above-described method.

[0010] Fourthly, embodiments of the present invention provide a storage medium storing one or more programs including execution instructions, the execution instructions being readable and executable by electronic devices (including but not limited to computers, servers, or network devices, etc.) to perform the steps of the method described above.

[0011] Fifthly, embodiments of the present invention also provide a computer program product, the computer program product including a computer program stored on a storage medium, the computer program including program instructions, which, when executed by a computer, cause the computer to perform the steps of the above-described method.

[0012] The above technical solution has at least the following advantages compared with the existing technology:

[0013] In this embodiment of the invention, by combining ride-sharing scheduling with traditional delivery models, the limitations of traditional models, such as the dispersed population in remote areas and the limited number of parcels, are overcome. By acquiring real-time information on remote parcels and ride-sharing bookings, precise filtering can be performed based on factors such as parcel attributes, expected arrival time, and delivery address. Idle ride-sharing vehicles are matched with suitable parcels, minimizing delivery costs and improving delivery efficiency. Therefore, compared to traditional methods of dispatching dedicated delivery personnel or periodic deliveries, using ride-sharing collaborative scheduling not only significantly shortens delivery time but also substantially reduces transportation costs. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 A flowchart illustrating an example of a logistics scheduling method for e-commerce trade according to an embodiment of the present invention is shown.

[0016] Figure 2 A flowchart illustrating an example of a logistics scheduling method for e-commerce trade according to an embodiment of the present invention is shown.

[0017] Figure 3 This diagram illustrates an example of a process flow chart for initial screening of various ridesharing booking information according to an embodiment of the present invention.

[0018] Figure 4 The following is an example of an operation flowchart illustrating how a target ride-sharing booking information is determined from various candidate ride-sharing booking information based on a logistics scheduling matching model according to an embodiment of the present invention.

[0019] Figure 5 A structural block diagram of an example logistics scheduling system for e-commerce transactions according to an embodiment of the present invention is shown.

[0020] Figure 6 This is a schematic diagram of the structure of an embodiment of the electronic device of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0022] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms “first,” “second,” and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms “an,” “a,” or “the,” and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms “comprising,” “including,” or “including,” and similar terms mean that the element or object preceding the word encompasses the element or object listed following the word and its equivalents, without excluding other elements or objects. The terms “connected,” “linked,” or “connected,” and similar terms are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect.

[0023] It should be noted that the terms "up", "down", "left", "right", "front", and "back" used in this invention are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0024] Figure 1 A flowchart illustrating an example of a logistics scheduling method for e-commerce trade according to an embodiment of the present invention is shown.

[0025] Regarding the execution entity of the method in this embodiment of the invention, it can be any controller or processor with computing or processing capabilities to intelligently match the delivery needs of remote parcels with existing ridesharing booking information. Through intelligent scheduling, based on factors such as the urgency of the parcel's delivery time and expected arrival time, the most suitable ridesharing vehicle is selected for delivery. In the current context of increasingly affordable and widespread car ownership, by combining ridesharing resources with the logistics and delivery needs of remote areas, the logistics and delivery costs and delivery time in remote areas are effectively reduced.

[0026] In some examples, it can be integrated and configured in the e-government system server through software, hardware, or a combination of both, and the type of e-government system server can be diverse, such as mobile phones, tablets, or desktop computers, etc.

[0027] like Figure 1 As shown, in step S110, remote parcel logistics information is obtained, which includes the parcel delivery address, parcel attributes, timeliness urgency, and expected arrival time of the parcel in the corresponding remote area.

[0028] In some implementations, real-time logistics information about remote parcels is obtained by connecting to the API interface of e-commerce platforms or logistics management systems. Specifically, the parcel delivery address is the geographical coordinates or address of the user or the delivery location; parcel attributes include characteristics such as the parcel's volume, weight, and whether it is fragile, which helps in subsequent scheduling to consider delivery methods and cargo requirements; the urgency level can be based on the delivery method selected by the user or the priority set by the merchant, indicating the parcel's delivery time requirements, such as normal or expedited; the expected arrival time of the parcel is the time when the parcel is expected to arrive at the designated delivery station, which can be generated by the e-commerce platform or logistics service provider when confirming the delivery plan.

[0029] Preferably, the acquired package information is formatted and verified, and then stored in a database or memory cache to facilitate subsequent processing by the scheduling algorithm.

[0030] In step S120, at least one ride-sharing booking information with the destination being a remote area is obtained. The ride-sharing booking information includes basic vehicle information, driving route, departure time, and logistics dispatch history.

[0031] In some implementations, ride-sharing booking information is obtained from the ride-sharing platform (or related vehicle management system). Based on the delivery demand in the target remote area, ride-sharing booking information related to that area is queried and retrieved. Specifically, basic vehicle information may include owner information and vehicle type (size, load capacity); the driving route is the ride-sharing driver's planned route, including the origin and destination, as well as the main road segments along the way, and can be obtained through map APIs or trip information uploaded by the owner; the departure time is the planned departure time of the vehicle, used to determine the availability and scheduling of ride-sharing; the logistics dispatch history is a record of the vehicle's past collaborative logistics dispatching, including the number of completed orders, service evaluations, etc., which helps to assess the reliability and efficiency of ride-sharing dispatching.

[0032] In step S130, the package attributes, expected arrival time, delivery address, and delivery station address are used to initially screen each ridesharing booking information to determine at least one matching candidate ridesharing booking information.

[0033] In some implementations, rule-based matching algorithms are used to initially screen each ridesharing booking, removing vehicles that clearly do not meet the criteria. For example, various package dimensions are matched with ridesharing booking information, and the results of each match are combined for initial screening, retaining the ridesharing bookings that meet the criteria as a candidate list.

[0034] For example, based on the package's size and weight, rideshares capable of carrying the package are selected. For instance, if the package is large or heavy, vehicles with high load capacity or that are empty are chosen. Rideshares with departure times later than or close to the package's expected arrival time are selected to ensure that rideshares can be dispatched. The package's delivery address and delivery station address are matched with each rideshare's route, filtering out rideshares whose routes deviate significantly from the package's delivery address.

[0035] In step S140, the information of each candidate ridesharing reservation and the remote parcel logistics information are input into the logistics scheduling matching model to determine the target ridesharing reservation information from the information of each candidate ridesharing reservation.

[0036] It should be noted that the logistics scheduling and matching model can employ various types of machine learning algorithms to optimize scheduling decisions, such as reinforcement learning, linear programming, integer programming, and genetic algorithms, to automatically select the best rideshare. No restrictions are imposed here. Preferably, a multi-objective optimization function can be designed to comprehensively consider factors such as cost, timeliness, and route optimization. This function considers delivery costs (e.g., shortest travel distance or time), delivery timeliness (e.g., whether the package arrives on time as expected), and rideshare scheduling reliability, balancing multiple objectives to ensure that the selected rideshare can both collaboratively complete logistics transportation and minimize transportation costs.

[0037] In step S150, the driving route in the target ride-sharing reservation information is updated according to the delivery station address and the package delivery address, and a logistics dispatch instruction is sent to the corresponding driver's client.

[0038] In some implementations, based on the delivery station address of the target rideshare and the package delivery address, map APIs (such as Baidu Maps, Gaode Maps, etc.) are used to calculate the optimal route through GIS and dynamic traffic data, thereby replanning the driving route and avoiding congested or impassable road sections. Then, logistics dispatch instructions are generated, including the vehicle's driving route, the scheduled delivery time, package information, etc., and forwarded to the corresponding driver's client via driver information or the rideshare platform API. Preferably, during the delivery process, the driving status of the rideshare is monitored in real time, such as route deviations or traffic congestion, and the corresponding vehicle monitoring information is synchronized to the logistics tracking information.

[0039] Through this invention, by acquiring ride-sharing booking information and matching it with package delivery needs, the maximum utilization of logistics resources (such as ride-sharing) is achieved. Because basic vehicle information, driving routes, departure times, and other factors are precisely scheduled, waste of vehicle resources or unnecessary empty runs are effectively avoided, optimizing the logistics process and improving the utilization rate of social resources. Therefore, combined with the dynamic scheduling of ride-sharing, delivery plans can be flexibly adjusted according to the actual driving routes of vehicles. In complex terrains in remote areas (such as mountains and hills), ride-sharing drivers are generally local residents; their familiarity with the terrain and their driving flexibility can effectively improve the timeliness and success rate of delivery.

[0040] Figure 2 A flowchart illustrating an example of a logistics scheduling method for e-commerce trade according to an embodiment of the present invention is shown.

[0041] like Figure 2 As shown, in step S210, a logistics dispatch instruction is sent to the corresponding vehicle owner client.

[0042] For details on the implementation of step S210, please refer to the above text. Figure 1 The description will not be repeated here.

[0043] In step S220, it is detected whether the feedback for the logistics scheduling instruction is a confirmation feedback notification or a denial feedback notification.

[0044] In some implementations, detailed logistics dispatch instructions are sent to the selected target rideshare driver via an interface with the driver's client (such as a mobile application or in-vehicle terminal system), requiring the driver to confirm acceptance and provide feedback. The driver can view detailed task information on the client and click the "Confirm" or "Deny" option after confirmation.

[0045] In step S231, upon detecting a confirmation feedback notification from the vehicle owner's client regarding the logistics dispatch instruction, a logistics dispatch and delivery task is generated.

[0046] Specifically, all relevant information (vehicle information, route, package information, etc.) will be extracted from the dispatch instructions to formally generate a delivery task, and the task details will be transmitted to the driver's client. A task confirmation notification will be sent through the driver's client, including the task number, route, estimated arrival time, etc., and the driver will be required to carry out the task according to the specified route and time. At the same time, the delivery status will be updated on the logistics dispatch platform, and real-time tracking will begin to ensure visibility and management during the delivery process.

[0047] In step S233, if a denial feedback notification from the driver's client regarding the logistics dispatch instruction is detected, the remaining ride-sharing order information, delivery station address, and remote parcel logistics information (excluding the target ride-sharing order information) are input into the logistics dispatch matching model to re-determine the target ride-sharing order information.

[0048] Specifically, from the initially screened candidate ridesharing bookings, rejected ridesharing bookings are removed. The remaining ridesharing booking information, delivery station addresses, and remote parcel logistics information are re-entered into the logistics scheduling and matching model. The logistics scheduling and matching model then optimizes and filters the remaining ridesharing options based on factors such as parcel timeliness, delivery station addresses, ridesharing routes, and vehicle status, generating new target ridesharing bookings and updating the task list on the driver's end, while continuing to await feedback from other drivers.

[0049] Through the real-time feedback mechanism provided in this invention, the system can quickly generate and push tasks after the driver confirms acceptance of the order, thereby reducing task initiation time and improving the overall efficiency of logistics scheduling. Furthermore, when the driver refuses to accept the order, the scheduling platform can quickly adjust and re-optimize the scheduling plan, avoiding interruptions to delivery tasks and ensuring the continuity and efficiency of logistics delivery. Thus, through dynamic matching and real-time feedback, with driver confirmation of acceptance and real-time task tracking, both drivers and users can obtain a better service experience, ensuring timely package delivery and reducing delivery delays caused by scheduling failures.

[0050] Figure 3 A flowchart illustrating an example of initial screening of various ridesharing booking information according to an embodiment of the present invention is shown.

[0051] like Figure 3 As shown, in step S310, for each ridesharing booking information, based on the package attributes, expected arrival time of the package, delivery address of the package, and delivery station address, the package attribute matching degree, time matching degree, and space matching degree relative to the ridesharing booking information are calculated, and combined to obtain the corresponding order matching degree.

[0052] Specifically, based on various information about the package (such as package attributes, timeliness, delivery address, etc.), it is matched with data such as the trip and departure time of the ride-sharing service in order to assess the suitability of each ride-sharing service for package delivery needs.

[0053] It should be noted that the matching calculation methods for each dimension can be diverse and should not be limited here. For example, for the matching degree of package attributes, the difference value calculation method (such as Euclidean distance, cosine similarity, etc.) can be used; for the matching degree of time, the overlap metric between the ride-sharing trip time and the package delivery window can be calculated (such as intersection degree, time difference metric); for the matching degree of space, spatial matching can be performed based on the geographical information of the package delivery address, delivery station address and ride-sharing route, such as calculating Haversine distance, Manhattan distance, etc., to measure the spatial proximity between the delivery route and the ride-sharing route.

[0054] Then, the package attribute matching degree, time matching degree and spatial matching degree are summarized, such as by weighted average or other fusion methods (such as multi-attribute decision method), to obtain the comprehensive order matching degree of each ride-sharing and package delivery task.

[0055] In step S320, a preset number of ridesharing booking information with the highest matching degree are selected from each ridesharing booking information to determine at least one matching candidate ridesharing booking information.

[0056] In some implementations, orders are sorted from highest to lowest matching degree, and a preset number of top-ranked ridesharing vehicles (e.g., the first 3 or 5) are selected for subsequent dispatch. This number can be adjusted according to actual needs and business models. Preferably, if some ridesharing vehicles fail to accept orders due to force majeure (such as malfunctions, driver refusal, etc.), the next candidate ridesharing vehicle with a high matching degree is automatically selected for rescheduling.

[0057] Through the embodiments of this invention, the refined matching and evaluation of package attributes, time, and space enables more accurate scheduling decisions, ensuring that packages are delivered to the most suitable rideshare. Furthermore, by sorting and filtering rideshare booking information, the system maximizes the utilization of rideshare transportation capacity, improves resource allocation efficiency, and reduces wasted time on empty runs and inefficient routes. When faced with different types of packages or diverse rideshare orders, the system can flexibly adjust matching criteria to adapt to different logistics needs, ensuring the stability and efficiency of the delivery network.

[0058] Regarding the implementation details of step S310 above, in some embodiments, it can be implemented in the following ways:

[0059]

[0060] In the formula, M total M represents the order matching degree. attr M time and M space These represent attribute matching degree, time matching degree, and spatial matching degree, respectively, and α1, α2, and α3 represent the weight factors of each corresponding matching dimension.

[0061] By employing a weighted product approach, attribute matching degree, temporal matching degree, and spatial matching degree are integrated, and a normalized exponent is used. Ensure the final match score is within a reasonable range. Compared to a simple linear weighted summation, the weighted product form better reflects the synergistic effect between dimensions. For example, if the match score of a certain dimension is 0 (no match at all), the total match score is immediately 0.

[0062] It should be noted that the weights α1, α2, and α3 can be dynamically adjusted according to business needs or business scenarios. For example, for different business scenarios (such as emergency package delivery), the time matching weight α2 can be increased to prioritize timeliness.

[0063] M attr =ω1·f1(w,W max )+ω2·f2(v,V max )+ω3·f3(t,T support Equation (2)

[0064]

[0065]

[0066]

[0067] In the formula, f1(w,W) max f2(v,V) max f3(t,T) and f3(t,T) support ) represent the weight matching function, volume matching function, and type matching function, respectively; ω1, ω2, and ω3 represent the weights of the corresponding attribute dimensions; w represents the package weight, W max V represents the maximum weight that a rideshare can carry, and v represents the volume of the package. max T represents the maximum capacity of the vehicle; t represents the type of package. support This represents the set of types that the vehicle supports.

[0068] Here, the attribute matching degree calculation comprehensively considers the weight matching function, volume matching function, and type matching function. In the weight matching function and volume matching function, the matching degree is highest when the weight w approaches the vehicle's load-bearing limit W. max It decreases slowly over time, rather than in a nonlinear form. This avoids the insensitivity of linear descent and ensures the vehicle can effectively carry packages. In the type matching function, only packages of type t belonging to the vehicle's supported type set T are matched. support If the match is 1, the match score is 1; otherwise, it is 0, ensuring that the package type (such as cold chain or fragile) matches the vehicle's support capabilities.

[0069] Furthermore, the attribute matching degree flexibly adjusts the importance of different attributes through the weighting mechanism ω1, ω2, ω3. For example, for fragile items, the weight of ω3 can be increased.

[0070]

[0071] In the formula, T deliver Indicates the expected arrival time of the package, T depart This represents the departure time of the ridesharing, and k represents the time sensitivity factor.

[0072] Here, the exponential decay function is used to handle the time difference T. deliver -T depart It reflects the effect of time non-linearly. When the vehicle departs at time T... depart Approaching the expected arrival time of the package T deliver When the time difference is large, the matching degree is relatively high; when the time difference is large, the matching degree drops rapidly. By using a time sensitivity factor k, time sensitivity can be controlled. For example, the greater the urgency of the matter, the higher the corresponding k value, making the impact of the time difference on the matching degree more significant.

[0073] M space =f station (L route ,L station )+f delivery (L route ,L package Equation (7)

[0074]

[0075]

[0076] In the formula, f station and f delivery Let L represent the fit functions between the ride-sharing route and the delivery station address and the package delivery address, respectively; route Indicates the route of the rideshare, L station Indicates the address of the delivery station, L package Indicates the package delivery address; dist(L station ,L route ) represents the distance between the delivery station address and the nearest point on the ride-sharing route, dist(L) package ,L route) represents the distance between the package delivery address and the nearest point on the ride-sharing route, and σ1 and σ2 represent the tolerance parameters for the corresponding distances.

[0077] Here, an exponential decay function is used to process the distance between the delivery station address and the parcel delivery address and the nearest point on the ride-sharing route; the closer the distance, the higher the matching degree. σ1 and σ2 are used to control the sensitivity of the matching degree to distance changes to meet the configuration requirements of different delivery scenarios. By calculating path coverage in the spatial matching degree, ride-sharing routes that cover the delivery station and delivery address are prioritized, thereby reducing additional detour costs.

[0078] This invention introduces a comprehensive evaluation method that integrates attribute matching, temporal matching, and spatial matching to ensure more accurate matching of packages and ridesharing services, significantly reducing the probability of inefficient matching. Furthermore, by utilizing nonlinear formulas such as exponential decay, the matching degree becomes more sensitive under boundary conditions (such as time constraints or large spatial deviations), avoiding the limitations of traditional linear models. Additionally, operators can dynamically adjust weight settings according to actual scenario needs, achieving intelligent and flexible scheduling to meet the personalized scheduling requirements of different types of remote areas.

[0079] Figure 4 The following is an example of an operation flowchart illustrating how a target ride-sharing booking information is determined from various candidate ride-sharing booking information based on a logistics scheduling matching model according to an embodiment of the present invention.

[0080] like Figure 4 As shown, in step S410, a logistics scheduling diagram structure is constructed based on remote parcel logistics information and the information of each candidate ridesharing reservation.

[0081] Here, the logistics scheduling graph structure contains multiple graph nodes and edge connections. The multiple graph nodes include parcel nodes, ride-sharing nodes, and global nodes. The node features of parcel nodes are defined based on remote parcel logistics information, and the node features of ride-sharing nodes are defined based on candidate ride-sharing reservation information. Global nodes connect to each ride-sharing node and are used to summarize global features. The edge weight of the parcel-ride-sharing edge is defined based on the order matching degree corresponding to the connected node.

[0082] Specifically, the logistics scheduling graph structure consists of package nodes, multiple ride-sharing nodes, and a global node. The edge weights of the package-ride-sharing edges determine the strength of the relationship between each package and a ride-sharing vehicle. Each node in the logistics scheduling graph structure carries a feature vector, while each edge carries a weight value representing the degree of matching between the package and the ride-sharing vehicle. The global node's role is to aggregate the feature information of the entire scheduling system; it can summarize the features of all ride-sharing nodes, providing global information for the graph neural network.

[0083] In step S420, the logistics scheduling graph structure is input into the graph neural network to update the node features of each ride-sharing node and the global node according to the package features and edge weights, thereby determining the matching score corresponding to each ride-sharing node.

[0084] It should be noted that graph neural networks are a type of deep learning model specifically designed for processing graph data. They can update node features by propagating information through edges between nodes. In a logistics scheduling graph, the task of a graph neural network is to iteratively update the feature vectors of nodes based on package features, ride-sharing node features, and edge weights, so that they more accurately represent the relationships between nodes.

[0085] It should be understood that graph neural networks can be of various types, such as graph convolutional networks (GCN) or other types of graph neural networks (such as GraphSAGE, GAT, etc.), which implement information propagation and aggregation. The feature vector of each node is updated by the features of its neighboring nodes, thereby obtaining a more meaningful node representation.

[0086] At each layer of the network, nodes update their features based on information from their neighbors. Through propagation across multiple graph neural network layers, node features gradually incorporate more global and local structural information. After processing, the updated ride-sharing node features are used to calculate a matching score for each node, reflecting the overall matching degree between different ride-sharing services and packages.

[0087] In step S430, the target ride-sharing reservation information is determined based on the candidate ride-sharing reservation information corresponding to the ride-sharing node with the highest matching score.

[0088] In some implementations, the system selects the ride-sharing booking information corresponding to the ride-sharing node with the highest matching degree, confirms the dispatch, and then outputs the target ride-sharing booking information, including vehicle information, departure time, route, etc., as the final delivery plan. Subsequently, specific dispatch instructions are generated and the relevant driver's client is notified to execute the delivery task.

[0089] Through the embodiments of this invention, a graph structure can comprehensively express the multiple relationships between packages and ride-sharing services. By transmitting information and updating features, scheduling accuracy is improved, ensuring that each package finds the most suitable ride-sharing service for delivery. Furthermore, by introducing global nodes and using feature propagation in the graph neural network, optimization can be performed from a global perspective, ensuring the efficiency and stability of overall logistics scheduling. Thus, the graph neural network can handle dynamically updated scheduling information, allowing the scheduling platform to flexibly adjust to different needs and environmental changes, ensuring maximum scheduling efficiency.

[0090] Regarding the implementation details of step S420, in some implementations, the graph neural network adopts a heterogeneous graph neural network, which includes a cascaded message passing mechanism module, an edge feature update module, and an output module.

[0091] Here, a heterogeneous graph neural network (Hetero-GNN) is employed, which includes package nodes, ride-sharing nodes, and global nodes, suitable for describing complex relationships between different types of objects (packages, vehicles, system states). Based on layer-by-layer message passing, nodes pass messages and update features through edges, aggregating neighborhood information layer by layer, and finally obtaining a global representation of each node.

[0092] Graph neural networks update node embeddings through multiple rounds of message passing. The message passing mechanism module is used to perform the following operations:

[0093] Message passing from the parcel node to the ride-sharing node:

[0094]

[0095]

[0096]

[0097] In the formula, l represents the layer index of the graph neural network. Let p be the feature vector of node p at layer l. and These are the feature vectors output by the ride-sharing node c at layer l and layer l+1, respectively. The representation of the message transmitted from the package node p to the ride-sharing node c at layer l+1; Let the edge weight between package node p and ride-sharing node c be represented at layer l, defined by the corresponding order matching degree; and These represent the linear transformation weight matrices for package features, ride-sharing features, and edge features, respectively; ReLU is the ReLU nonlinear activation function; N(c) is the set of neighboring package nodes of ride-sharing node c. The feature vector of the neighbor node q at the l-th layer is used to wrap the feature vector of the neighbor node q. Let be the representation of the attention weights of the wrapping node p and the ride-sharing node c at layer l+1, where a represents the attention weight vector and ‖ represents the vector concatenation operation.

[0098] Here, the package node transmits logistics characteristics (such as urgency, weight, volume, etc.) to the ride-sharing node, and the message... It describes the interaction between the package and the ride-sharing service, combined with edge features. It dynamically reflects the current matching relationship.

[0099] By adopting a dynamic edge feature update mechanism, edge weights are adjusted in real time based on changes in the characteristics of package and ride-sharing nodes. This reflects factors such as path deviation distance and time window fit in real time, capturing personalized matching relationships between different packages and ride-sharing services. This ensures more accurate matching between packages and ride-sharing services and reduces resource waste caused by static rules.

[0100] Attention weight Based on the characteristics of parcels and ride-sharing, the model can distinguish the importance of different parcels, prioritize the delivery needs of high-time-sensitivity parcels, and make full use of ride-sharing resources to balance timeliness and resource utilization.

[0101] Messages are transmitted from the ride-sharing node to the global node to aggregate overall resource and timeliness information:

[0102]

[0103] In the formula, This is the representation of the message from rideshare node c to global node g at layer l+1. This is the weight matrix passed from the rideshare to the global node; and Let be the feature vectors of global node g at layers l+2 and l+1, respectively, and let N(g) represent the set of ride-sharing nodes connected to global node g. This represents the weight matrix for updating global node features.

[0104] Here, the ride-sharing node reports its comprehensive characteristics to the global node g, reflecting the current status of transportation resource utilization. The design of the global node incorporates system-level optimization information into the node characteristics, enhancing the global perspective of the scheduling model.

[0105] Global node g will send the updated global information via message Passed to each ridesharing node c:

[0106]

[0107]

[0108] In the formula, This represents the message from global node g to rideshare node c at layer l+2. This represents the weight matrix passed from the global node to the rideshare node. This is the feature vector output by the ride-sharing node c at layer l+3.

[0109] Here, the global node feeds back system optimization information to the ride-sharing nodes, guiding the adjustment of ride-sharing node characteristics and optimizing the overall scheduling strategy. The global node aggregates the status information of all ride-sharing nodes and guides the scheduling strategy of ride-sharing nodes through the feedback mechanism. While meeting the needs of individual packages, it optimizes the allocation of global resources and avoids the problems of local optimization and global imbalance.

[0110] The edge feature update module is used to update the features of the package-hitchhiking edge after each layer of message passing, so that it can dynamically adjust to adapt to changes in node features:

[0111]

[0112] In the formula, The edge weight between wrapper node p and rideshare node c is represented at layer l+1; MLP e It is a multilayer perceptron, used to dynamically adjust the nonlinear mapping function of edge features.

[0113] The feature fusion of package nodes and ride-sharing nodes is considered by multilayer perceptron (MLP), and the edge features are dynamically adjusted so that the edge features can reflect the changes in node features, and the changes in edge features can adapt to the real-time adjustment of node features.

[0114] The output module is used to calculate the matching score based on the final node features of the ride-sharing node:

[0115]

[0116] In the formula, S c W represents the matching score of the ride-sharing node c. o and b o These represent the weight matrix and bias term of the output layer, respectively, σ o Let L be the activation function of the output layer, and L be the depth of the graph neural network. This is the feature vector output by the ride-sharing node c at the final Lth layer.

[0117] Here, the ride-sharing node After passing through multiple layers of messages, it integrates the characteristics of adjacent packages as well as global system optimization information, and has a global perspective, enabling more reasonable path planning and package allocation in actual scheduling.

[0118] In the graph neural network provided in this embodiment of the invention, message passing and feature updates at each layer are based on the neighborhood of nodes and edges, resulting in low computational complexity. In large-scale logistics scheduling scenarios, it can quickly generate scheduling decisions, meeting real-time requirements. Furthermore, by dynamically updating edge weights, it avoids repeatedly calculating the entire graph relationship, requiring only local adjustments, reducing computational costs and keeping computational resource consumption controllable.

[0119] Through the embodiments of this invention, based on the characteristics of parcel nodes and ride-sharing nodes, and through layer-by-layer information aggregation and dynamic edge feature updates, the real-time matching degree of each ride-sharing node can be accurately evaluated. This dynamically adapts to the real-time changes in parcel demand and ride-sharing resource status in logistics scenarios, improving the matching efficiency of parcels and ride-sharing resources. Furthermore, by introducing global nodes, the heterogeneous graph neural network can comprehensively consider system-level resource distribution, achieving optimized feedback for ride-sharing nodes, improving the overall resource utilization of the system, and avoiding the defects of local optimization caused by ignoring the global perspective.

[0120] In some examples of embodiments of the present invention, the heterogeneous graph neural network is optimized and trained based on a data sample set and a loss function. Specifically, the loss function of the heterogeneous graph neural network is:

[0121]

[0122]

[0123] L time (x i ) = max(0,T i deliver -T i real Equation (21)

[0124]

[0125] In the formula, L represents the total loss function of the heterogeneous graph neural network, N is the total number of samples in the data sample set, i is the sample index, and x... i Let y be the input feature of sample i. i L represents the true label of sample i; match (x i ,y i L time (x i ) and L space (x i ) represent the matching loss term, time-sensitive loss term, and spatial constraint loss term for sample i, respectively, and λ1, λ2, and λ3 are the corresponding weight parameters for each loss term; S i T represents the matching score for sample i output by the model; i deliver T represents the estimated arrival time of the package in sample i. i real P represents the actual arrival time of the package in sample i; i Let L represent the set of packages assigned to ride-sharing service c(i) in sample i. c(i) L represents the current location of the ride-sharing vehicle in sample i.p Dist(L) represents the delivery address of package p in sample i. c(i) ,L p ) represents L c(i) With L p The path distance between them, dist(L) c(i) ,L station ) represents L c(i) With L station The path distance between them.

[0126] In this embodiment of the invention, the loss function of the heterogeneous graph neural network integrates matching loss, time-sensitive loss, and spatial constraint loss.

[0127] Firstly, in logistics scheduling scenarios, accurately matching packages and ride-sharing nodes is the most basic optimization goal. Matching loss is used to measure the matching score S predicted by the model. i With real label y i To ensure consistency, the cross-entropy function is used to quantify the accuracy of the model's judgment on matching and non-matching.

[0128] Secondly, in logistics and delivery scenarios, time sensitivity is also an important optimization objective. By using time-sensitive loss, the delivery tasks of urgent packages can be prioritized to ensure they are completed within a specified time window. Specifically, a timeout penalty mechanism is designed, using the max(0,·) function to incur additional losses when the estimated arrival time exceeds the actual arrival time. Furthermore, the time sensitivity of each sample is calculated independently, automatically prioritizing urgent orders during model training, and dynamically optimizing the scheduling scheme based on the urgency of the time window.

[0129] Thirdly, in logistics and delivery scenarios, route optimization is a key aspect of controlling costs and improving efficiency. By introducing spatial constraint loss, the route planning of ride-sharing is constrained, minimizing route deviation and avoiding excessive ineffective travel distances, thereby improving resource utilization. Specifically, by calculating the deviation ratio of the delivery route, the route planning of ride-sharing can be dynamically adjusted to shorten the delivery route as much as possible. Spatial constraint loss allows the model to not only focus on matching accuracy but also optimize route consumption in actual transportation.

[0130] Through the embodiments of the present invention, the total loss function of the heterogeneous graph neural network integrates multi-objective constraints such as matching accuracy, time sensitivity and path planning. By dynamically adjusting the weights λ1, λ2 and λ3, it can simultaneously optimize accuracy, timeliness and path planning, achieve comprehensive performance optimization in different scenarios, and meet the actual needs of various logistics scheduling scenarios.

[0131] The logistics scheduling system for e-commerce trade provided by this invention will be described below. The logistics scheduling system for e-commerce trade described below can be referred to in correspondence with the logistics scheduling method for e-commerce trade described above.

[0132] Figure 5 A structural block diagram of an example logistics scheduling system for e-commerce trade according to an embodiment of the present invention is shown.

[0133] like Figure 5 As shown, the logistics scheduling system 500 for e-commerce trade includes a first acquisition unit 510, a second acquisition unit 520, a reservation order screening unit 530, a scheduling matching unit 540, and a scheduling execution unit 550.

[0134] The first acquisition unit 510 is used to acquire remote parcel logistics information, which includes the parcel delivery address, parcel attributes, timeliness urgency, and expected arrival time of the parcel in the corresponding remote area.

[0135] The second acquisition unit 520 is used to acquire at least one ridesharing reservation information whose corresponding trip destination is the remote area. The ridesharing reservation information includes basic vehicle information, driving route, departure time, and logistics dispatch history.

[0136] The reservation initial screening unit 530 is used to perform initial screening on each of the ride-sharing reservation information based on the package attributes, the expected arrival time of the package, the delivery address of the package, and the delivery station address, so as to determine at least one matching candidate ride-sharing reservation information.

[0137] The scheduling matching unit 540 is used to input the candidate ridesharing booking information and the remote parcel logistics information into the logistics scheduling matching model, so as to determine the target ridesharing booking information from the candidate ridesharing booking information.

[0138] The dispatch execution unit 550 is used to update the driving route in the target ride-sharing reservation information according to the delivery station address and the package delivery address, and send logistics dispatch instructions to the corresponding driver client.

[0139] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of combined actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, as some steps can be performed in other orders or simultaneously according to the present invention. Secondly, 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 the present invention. In the above embodiments, the descriptions of each embodiment have their own emphasis; for parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0140] In some embodiments, the present invention provides a non-volatile computer-readable storage medium storing one or more programs including execution instructions, which can be read and executed by electronic devices (including but not limited to computers, servers, or network devices) to perform the logistics scheduling method for e-commerce trade described above.

[0141] In some embodiments, the present invention also provides a computer program product, the computer program product including a computer program stored on a non-volatile computer-readable storage medium, the computer program including program instructions, which, when executed by a computer, cause the computer to perform the above-described logistics scheduling method for e-commerce trade.

[0142] In some embodiments, the present invention also provides an electronic device comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform a logistics scheduling method for e-commerce trade.

[0143] Figure 6 This is a schematic diagram of the hardware structure of an electronic device for executing a logistics scheduling method for e-commerce trade, as provided in another embodiment of the present invention. Figure 6 As shown, the device includes:

[0144] One or more processors 610 and memory 620, Figure 6 Take the 610 processor as an example.

[0145] The equipment for implementing the logistics scheduling method for e-commerce trade may further include: an input device 630 and an output device 640.

[0146] The processor 610, memory 620, input device 630, and output device 640 can be connected via a bus or other means. Figure 6 Taking the example of a connection between China and Israel via a bus.

[0147] The memory 620, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the logistics scheduling method for e-commerce trade in the embodiments of the present invention. The processor 610 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions, and modules stored in the memory 620, thereby implementing the logistics scheduling method for e-commerce trade in the above-described method embodiments.

[0148] The memory 620 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device. Furthermore, the memory 620 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 620 may optionally include memory remotely located relative to the processor 610, and these remote memories can be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0149] Input device 630 can receive input digital or character information and generate signals related to user settings and function control of the electronic device. Output device 640 may include display devices such as a display screen.

[0150] The one or more modules are stored in the memory 620, and when executed by the one or more processors 610, they execute the logistics scheduling method for e-commerce trade in any of the above method embodiments.

[0151] The above-described product can execute the method provided in the embodiments of the present invention, and has the corresponding functional modules and beneficial effects for executing the method. Technical details not described in detail in this embodiment can be found in the method provided in the embodiments of the present invention.

[0152] The electronic devices of this invention exist in various forms, including but not limited to:

[0153] (1) Mobile communication devices: These devices are characterized by their mobile communication capabilities and primarily aim to provide voice and data communication. These terminals include smartphones, multimedia phones, feature phones, and low-end phones.

[0154] (2) Ultra-mobile personal computer devices: These devices fall under the category of personal computers, possessing computing and processing capabilities, and generally also have mobile internet access features. These terminals include: PDAs, MIDs, and UMPCs, etc.

[0155] (3) Portable entertainment devices: These devices can display and play multimedia content. This category includes audio and video players, handheld game consoles, e-book readers, as well as smart toys and portable car navigation devices.

[0156] (4) Other airborne electronic devices with data interaction capabilities, such as vehicle-mounted systems installed on vehicles.

[0157] The device embodiments described above are merely illustrative. 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 modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0158] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0159] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A logistics scheduling method for e-commerce trade, comprising: Obtain remote parcel logistics information, which includes the parcel delivery address, parcel attributes, timeliness urgency, and expected arrival time of the parcel in the corresponding remote area; Obtain at least one ridesharing booking information with the corresponding trip destination being the remote area. The ridesharing booking information includes basic vehicle information, driving route, departure time, and logistics dispatch history. The package attributes, the expected arrival time of the package, the delivery address of the package, and the delivery station address are used to initially screen each of the ride-sharing booking information to determine at least one matching candidate ride-sharing booking information; The candidate ridesharing booking information and the remote parcel logistics information are input into the logistics scheduling matching model to determine the target ridesharing booking information from the candidate ridesharing booking information, including: Based on the remote parcel logistics information and the candidate ridesharing booking information, a logistics scheduling graph structure is constructed. This graph structure includes multiple graph nodes and edge connections. The multiple graph nodes include parcel nodes, ridesharing nodes, and a global node. The node features of the parcel nodes are defined based on the remote parcel logistics information, and the node features of the ridesharing nodes are defined based on the candidate ridesharing booking information. The global node connects to each ridesharing node and is used to summarize global features. The edge weight of the parcel-ridesharing edge is defined based on the order matching degree corresponding to the connected node. The logistics scheduling graph structure is input into a graph neural network to update the node features of each ridesharing node and the global node according to the package features and edge weights, thereby determining the matching score corresponding to each ridesharing node; and the target ridesharing reservation information is determined according to the candidate ridesharing reservation information corresponding to the ridesharing node with the highest matching score. The graph neural network is a heterogeneous graph neural network, which includes a cascaded message passing mechanism module, an edge feature update module, and an output module. The message passing mechanism module is used to perform the following operations: message passing from the package node to the ride-sharing node; message passing from the ride-sharing node to the global node; and the global node passing the updated global information to each ride-sharing node via message. The edge feature update module is used to update the features of the package-ride-sharing edge after each layer of message transmission, so that it can be dynamically adjusted to adapt to changes in node features. The output module is used to calculate the matching score based on the final node characteristics of the ride-sharing node; Update the driving route in the target ride-sharing booking information based on the delivery station address and the package delivery address, and send logistics dispatch instructions to the corresponding driver's client.

2. The method according to claim 1, wherein, After sending the logistics dispatch instruction to the corresponding vehicle owner's client, the method further includes: Upon detecting a confirmation notification from the vehicle owner's client regarding the logistics dispatch instruction, a logistics dispatch and delivery task is generated. If a denial feedback notification is detected from the car owner's client regarding the logistics dispatch instruction, all remaining ride-sharing order information (excluding the target ride-sharing order information), delivery station addresses, and remote parcel logistics information are input into the logistics dispatch matching model to re-determine the target ride-sharing order information.

3. The method according to claim 1, wherein, The initial screening of each ridesharing booking information based on the package attributes, the expected arrival time of the package, the delivery address of the package, and the delivery station address to determine at least one matching candidate ridesharing booking information includes: For each of the aforementioned ride-sharing booking information, based on the package attributes, the expected arrival time of the package, the delivery address of the package, and the delivery station address, the package attribute matching degree, time matching degree, and space matching degree relative to the ride-sharing booking information are calculated, and combined to obtain the corresponding order matching degree; From each of the aforementioned ridesharing booking information, a predetermined number of ridesharing booking information with the highest matching degree are selected to determine at least one matching candidate ridesharing booking information.

4. The method according to claim 3, wherein, For each of the aforementioned ridesharing booking information, based on the package attributes, the expected arrival time of the package, the delivery address of the package, and the delivery station address, the matching degree of package attributes, time matching degree, and spatial matching degree relative to the ridesharing booking information is calculated, and then combined to obtain the corresponding order matching degree, including: ， In the formula, Indicates the order matching degree. , and These represent attribute matching degree, temporal matching degree, and spatial matching degree, respectively. This represents the weight factor for each corresponding matching dimension. ， ， ， ， In the formula, , and These represent the weight matching function, volume matching function, and type matching function, respectively. , and This indicates the weight of each corresponding attribute dimension; Indicates the weight of the package. This indicates the maximum weight that a rideshare can carry. Indicates the volume of the package. Indicates the maximum capacity of the vehicle; Indicates the package type. This represents the set of types that the vehicle supports; ， In the formula, This indicates the expected arrival time of the package. Indicates the departure time of the rideshare. Indicates time-sensitive factor; ， ， ， In the formula, and These functions represent the degree of fit between the ride-sharing route and the delivery station address and the package delivery address, respectively. This indicates the route for the rideshare. Indicates the address of the delivery station. Indicates the package delivery address; This indicates the distance between the delivery station address and the nearest point on the ride-sharing route. This indicates the distance between the package's delivery address and the nearest point on the ride-sharing route. and These represent the tolerance parameters for the corresponding distances.

5. The method according to claim 1, wherein the message passing mechanism module is configured to perform the following operations: Message passing from the parcel node to the ride-sharing node: ， ， ， In the formula, This represents the layer index of a graph neural network. For package node In the The feature vector of the layer, and These are the ride-sharing nodes. In the Layer and first The feature vector output by the layer, To the package node To the ride-sharing node The message being delivered is in the Layer representation; For package node With ride-sharing nodes The edge weights between them are in the th... The layer representation is defined by the corresponding order matching degree; , and These are the linear transformation weight matrices for package features, ride-sharing features, and edge features, respectively. It is a ReLU nonlinear activation function; For ride-sharing nodes The set of neighboring package nodes, Package the neighbor node In the The feature vector of the layer; For package node With ride-sharing nodes Attention weights in the th Layer representation, Represents the attention weight vector. This represents a vector concatenation operation; Messages are transmitted from the ride-sharing node to the global node to aggregate overall resource and timeliness information: ， ， In the formula, To the ride-sharing node To global node The message in the Layer representation, This is the weight matrix passed from the rideshare to the global node; and These are global nodes. In the Layer and first The feature vector of the layer, Indicates the relationship with global nodes A set of connected ride-sharing nodes. The weight matrix representing the global node feature update; global nodes The updated global information will be sent via message. Transmitted to each ridesharing node : ， ， In the formula, Indicates from global node To the ride-sharing node The message in the Layer representation, This represents the weight matrix passed from the global node to the rideshare node. For ride-sharing nodes In the Feature vectors output by the layer; The edge feature update module is used to update the features of the package-rideshare edge after each layer of message passing, so that it can be dynamically adjusted to adapt to changes in node features: ， In the formula, For package node With ride-sharing nodes The edge weights between them are in the th... Layer representation; It is a multilayer perceptron, used to dynamically adjust the nonlinear mapping function of edge features; The output module is used to calculate the matching score based on the final node features of the ride-sharing node: ， In the formula, Indicates a ride-sharing node Match score, and These represent the weight matrix and bias term of the output layer, respectively. The activation function of the output layer. It is the depth of the graph neural network. For ride-sharing nodes In the final The feature vector output by the layer.

6. The method according to claim 5, wherein, The loss function of the heterogeneous graph neural network is: ， ， ， ， In the formula, This represents the total loss function of a heterogeneous graphical neural network. The total number of samples in the data sample set. For sample index, For the sample Input features, For the sample The true label; , and Representing samples respectively The matching loss term, time-sensitive loss term, and spatial constraint loss term, For each loss term, the corresponding weight parameters are defined. This represents the output of the model for the sample. Match score; Indicates sample The estimated arrival time of the package. Indicates sample The actual arrival time of the package; Indicates sample Distributed to ride-sharing A collection of packages, Indicates sample The current location of the ride-sharing service. Indicates sample China Parcel Delivery address, express and Path distance between express and The path distance between them.

7. A logistics scheduling system for e-commerce trade, used to implement the method of any one of claims 1-6, the system comprising: The first acquisition unit is used to acquire remote parcel logistics information, which includes the parcel delivery address, parcel attributes, timeliness and urgency of the parcel and the expected arrival time of the parcel in the corresponding remote area. The second acquisition unit is used to acquire at least one ridesharing booking information whose corresponding trip destination is the remote area. The ridesharing booking information includes basic vehicle information, driving route, departure time, and logistics dispatch history order records. The reservation order screening unit is used to screen each of the ride-sharing reservation order information based on the package attributes, the expected arrival time of the package, the delivery address of the package, and the delivery station address, so as to determine at least one matching candidate ride-sharing reservation order information. The scheduling and matching unit is used to input the candidate ridesharing reservation information and the remote parcel logistics information into the logistics scheduling and matching model, so as to determine the target ridesharing reservation information from the candidate ridesharing reservation information. The scheduling execution unit is used to update the driving route in the target ridesharing reservation information according to the delivery station address and the package delivery address, and send logistics scheduling instructions to the corresponding driver's client.

Images