Waybill delivery method and device, and electronic device

Abstract

The application discloses a way of delivery order dispatching, belonging to the field of computer technology, which helps to improve the efficiency of delivery order dispatching. The method comprises: obtaining a set of delivery capacity and a set of delivery orders to be dispatched; using a parallel assignment scheme to perform delivery order matching, and determining the matching relationship between each delivery order in the set of delivery orders to be dispatched and each delivery capacity in the set of delivery capacity; according to the matching relationship, determining the corresponding candidate delivery order of each delivery capacity in the current round of delivery order matching; then, for each delivery capacity corresponding to at least two candidate delivery orders, predicting the candidate delivery order selected by the delivery capacity in the current round of delivery order matching through a pre-trained delivery order sequence prediction model; according to the candidate delivery order selected by each delivery capacity in the current round of delivery order matching, determining the result of the current round of delivery order matching; in response to the completion of the dispatching of the set of delivery orders to be dispatched, performing the dispatching operation of the set of delivery orders to be dispatched to the set of delivery capacity according to the result of each round of delivery order matching.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a waybill dispatching method, apparatus, electronic device, and computer-readable storage medium. Background Technology

[0002] With the development of delivery services, they have become closely intertwined with people's lives. For example, express delivery, bulk logistics, and food delivery services have brought convenience to people's lives while also saving social transportation resources and improving capacity utilization efficiency. The matching relationship between waybills and delivery capacity is crucial to delivery efficiency and significantly impacts user experience. Therefore, optimizing the matching relationship between delivery capacity and waybills is a key issue in delivery scheduling systems. Especially in food delivery scenarios, the scheduling system needs to process the matching relationship between over 20 million waybills and over 700,000 delivery capacity every 30 seconds on average. The system needs to provide the matching relationship within a very short time, placing extremely high demands on its performance in this area. In existing technologies, to maximize the matching performance, a parallel assignment scheme is typically used to establish the matching relationship between waybills and riders.

[0003] However, the existing waybill dispatch schemes, by sacrificing solution quality for solution performance, do not yield the optimal matching relationship between the overall capacity and the waybills, and therefore require improvement. Summary of the Invention

[0004] This application provides a waybill dispatch method that helps improve waybill dispatch efficiency.

[0005] To address the aforementioned problems, firstly, embodiments of this application provide a waybill dispatch method, including:

[0006] Obtain the capacity set and the set of waybills to be dispatched;

[0007] A parallel assignment scheme is used to perform waybill matching, and the matching relationship between each waybill in the set of waybills to be dispatched and each capacity in the set of capacity is determined;

[0008] Based on the matching relationship, a candidate waybill corresponding to each capacity in the current round of waybill matching is determined, and the candidate waybill is selected from the waybill set to be dispatched;

[0009] For each capacity corresponding to at least two candidate waybills, the candidate waybill selected by the capacity in the current round of waybill matching is predicted by a pre-trained waybill sequence prediction model;

[0010] The current round waybill matching result is determined based on the candidate waybills selected by each of the aforementioned transport capacities in the current round waybill matching;

[0011] In response to the completion of the dispatch of the set of waybills to be dispatched, the dispatch operation of the set of waybills to be dispatched to the set of transport capacity is executed according to the matching results of each round of waybills.

[0012] Secondly, embodiments of this application provide a waybill dispatching device, comprising:

[0013] The module for obtaining the capacity set and the waybill set is used to obtain the capacity set and the waybill set to be dispatched;

[0014] The parallel matching module is used to perform waybill matching using a parallel assignment scheme to determine the matching relationship between each waybill in the set of waybills to be dispatched and each capacity in the set of transport capacity.

[0015] The candidate waybill determination module is used to determine, based on the matching relationship, the candidate waybill corresponding to each capacity in the current round of waybill matching, wherein the candidate waybill is selected from the waybill set to be dispatched;

[0016] The candidate waybill selection module is used to predict, for each capacity corresponding to at least two candidate waybills, the candidate waybill to be selected in the current round of waybill matching by using a pre-trained waybill sequence prediction model;

[0017] The waybill matching result determination module is used to determine the waybill matching result of the current round based on the candidate waybills selected in the waybill matching of each of the transport capacities in the current round;

[0018] The waybill dispatch module is used to respond to the completion of the dispatch of the waybill set to be dispatched, and to perform the dispatch operation of the waybill set to be dispatched to the capacity set according to the waybill matching results of each round.

[0019] Thirdly, embodiments of this application also disclose an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the waybill dispatching method described in embodiments of this application.

[0020] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, represents the steps of the waybill dispatching method disclosed in embodiments of this application.

[0021] The waybill dispatching method disclosed in this application involves obtaining a capacity set and a set of waybills to be dispatched; performing waybill matching using a parallel dispatch scheme to determine the matching relationship between each waybill in the set of waybills to be dispatched and each capacity in the capacity set; determining candidate waybills corresponding to each capacity in the current round of waybill matching based on the matching relationship, wherein the candidate waybills are selected from the waybills in the set of waybills to be dispatched; for each capacity corresponding to at least two candidate waybills, predicting the candidate waybills selected by the capacity in the current round of waybill matching using a pre-trained waybill sequence prediction model; determining the waybill matching result for the current round based on the candidate waybills selected by each capacity in the current round of waybill matching; and, in response to the completion of the dispatching of the set of waybills to be dispatched, performing a dispatching operation from the set of waybills to be dispatched to the capacity set based on the waybill matching result of each round, which helps to improve the efficiency of waybill dispatching.

[0022] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0023] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] Figure 1 This is a flowchart of the waybill dispatching method according to Embodiment 1 of this application;

[0025] Figure 2 This is a schematic diagram of the waybill sequence prediction model structure in Embodiment 1 of this application;

[0026] Figure 3 This is another flowchart of the waybill dispatching method according to Embodiment 1 of this application;

[0027] Figure 4 This is a schematic diagram of the selection path of transport capacity for waybills in Embodiment 1 of this application;

[0028] Figure 5 This is one of the schematic diagrams of the waybill dispatching device in Embodiment 2 of this application;

[0029] Figure 6 This is the second schematic diagram of the waybill dispatching device in Embodiment 2 of this application;

[0030] Figure 7 A block diagram schematically illustrates an electronic device for performing the method according to this application; and

[0031] Figure 8 A storage unit for holding or carrying program code implementing the method according to this application is illustrated schematically. Detailed Implementation

[0032] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0033] Example 1

[0034] This application discloses a waybill dispatching method, such as... Figure 1 As shown, the method includes steps 110 to 160.

[0035] Step 110: Obtain the capacity set and the set of waybills to be dispatched.

[0036] The delivery capacity described in this application embodiment can be the delivery capacity of riders, delivery robots, delivery vehicles, etc.

[0037] In existing technologies, when dispatching waybills, capacity scheduling systems typically use time slices as a dispatch cycle, dispatching waybills within each time slice and scheduling the capacity within that time slice. For example, waybill dispatch may occur every minute. During waybill dispatch, the capacity scheduling system creates a set of waybills to be dispatched within the current time slice and a set of available capacity (e.g., online capacity that is carrying waybills but has not reached its capacity limit) within the current time slice. Then, waybill dispatch is performed between the created capacity set and the set of waybills to be dispatched.

[0038] Typically, the capacity set includes multiple transport capacities, and the set of waybills to be dispatched includes multiple waybills to be dispatched.

[0039] In existing technologies, the capacity and waybill matching problem is transformed into an online parallel waybill assignment problem. When solving the parallel waybill assignment problem, the following principle is defined: each waybill can only be assigned to one capacity, but each capacity can be assigned to multiple waybills. The waybill set N is... j Allocate to transport capacity r j Scheduling score f(r) j N j Minimum, for each transport capacity r jThe number of orders assigned cannot exceed b. j The problem is defined as minimizing the scheduling score. This is a heuristic scheme for parallel solution, with each round consisting of three steps: finding the rider with the minimum scheduling score for each waybill, conflict resolution, and updating the waybills to be dispatched. The waybill set N is... j Allocate to transport capacity r j The scheduling score f(r) j N j The calculation can be performed using existing methods, and this embodiment does not limit the scope of the calculation.

[0040] The following section details the solution process for each round.

[0041] Step 120: Perform waybill matching using a parallel assignment scheme to determine the matching relationship between each waybill in the set of waybills to be dispatched and each capacity in the set of transport capacity.

[0042] Next, the transportation capacity scheduling system can adopt the existing waybill dispatch strategy to dispatch each waybill in the set of waybills to be dispatched to the transportation capacity in the set of transportation capacity in parallel, that is, to determine the matching relationship between each waybill in the set of waybills to be dispatched and each transportation capacity in the set of transportation capacity.

[0043] In some embodiments of this application, the matching relationship between waybills and transport capacity can be represented as the scheduling score of a waybill and a transport capacity after a waybill is dispatched to a certain transport capacity. The scheduling score is calculated according to a preset evaluation strategy.

[0044] In this embodiment of the application, the specific scheme for performing waybill matching using a parallel assignment scheme to determine the matching relationship between each waybill in the set of waybills to be dispatched and each capacity in the set of capacity is not limited.

[0045] Step 130: Based on the matching relationship, determine the candidate waybill corresponding to each capacity in the current round waybill matching, wherein the candidate waybill is selected from the waybill set to be dispatched.

[0046] After determining the matching relationship between each waybill and the transport capacity, for each waybill, based on the matching relationship, the transport capacity with the highest matching degree with the waybill can be determined (e.g., for each waybill, the transport capacity with the lowest scheduling score with the waybill can be taken as the transport capacity with the highest matching degree with the waybill). In this way, after one round of parallel matching, each transport capacity may not be matched with any waybill, or it may be matched with one or more waybills, which are taken as the candidate waybills corresponding to the transport capacity in the current round of parallel matching.

[0047] Step 140: For each capacity corresponding to at least two candidate waybills, predict the candidate waybill selected by the capacity in the current round of waybill matching using a pre-trained waybill sequence prediction model.

[0048] To improve the overall scheduling performance of the capacity scheduling system, the system will perform multiple rounds of parallel matching. In each round of parallel matching, only the best-matched waybill for each capacity will be determined as the last waybill to be dispatched to that capacity. Therefore, for a capacity that matches two or more waybills in a round of parallel matching (e.g., capacity 1 matches waybill 5 and waybill 7), conflict handling needs to be performed. That is, a candidate waybill will be selected from the multiple candidate waybills that match the capacity as the waybill determined for that capacity in this round of parallel matching.

[0049] In current technology, for a capacity that matches two or more waybills in a round of parallel matching (such as capacity 1 mentioned above), a waybill is selected from the waybills matched with that capacity according to a specified strategy (such as selecting the waybill with the lowest scheduling score) as the final waybill matched with that capacity in this round (such as waybill 5). This conflict resolution method in the prior art does not take into account the global information of the capacity when selecting waybills, which leads to a decrease in the overall capacity scheduling efficiency of the capacity scheduling system.

[0050] After studying existing parallel assignment schemes, the inventors discovered that each round of parallel matching can be divided into two steps: a candidate set generation step and a candidate set selection step. The candidate set generation step uses parallel matching to find the optimal rider for each order, resulting in a candidate order set for each rider. The candidate set selection step uses a strategy to select the optimal order from each rider's candidate order set and assign it to that rider in that round. In each round of parallel matching, each rider faces a selection node: choosing a specific order from the candidate order set. Thus, after multiple rounds of parallel matching (i.e., multiple selections), the optimal global match between orders and riders is achieved.

[0051] Therefore, the inventors modeled the entire delivery order problem as a sequence decision problem. That is, solving how multiple riders select orders at sequence nodes to achieve a globally optimal order-rider matching relationship. Due to online performance limitations, this application uses a machine learning approach to quickly prune heuristic results for faster verification of the optimal path. Simultaneously, since the path provided by parallel matching is not the optimal path, a better operations research algorithm is needed to provide a better path during the solution process, resulting in a more optimal overall path. Therefore, this application uses offline neighborhood search for operations research optimization to obtain the optimal order selection path for riders. Then, based on the obtained path, a delivery order sequence prediction model is trained for online prediction.

[0052] In some embodiments of this application, a time-series-based waybill sequence prediction model is used to predict the waybill that best matches the capacity in the current round of parallel matching.

[0053] The waybill sequence prediction model described in this application embodiment is built based on the Long Short-Term Memory (LSTM) model. For example... Figure 2 As shown, the waybill sequence prediction model includes an encoding network 210 and a decoding network 220. The decoding network 220 is used to encode and map the data input to the encoding network 210, outputting an encoding vector and a context vector. For each time t, the decoding network 220 encodes the context vector and data input to the decoding network 220 at time t to obtain the output for that time t. The context vector input to the decoding network 220 at time t can be either the context vector output by the decoding network 220 at time t-1 or the context vector output by the encoding network 210.

[0054] The waybill sequence prediction model described in this application embodiment is trained offline. For example... Figure 3 As shown, before predicting the candidate waybill selected by the transport capacity in the current round of waybill matching using a pre-trained waybill sequence prediction model, the method further includes: step 100.

[0055] Step 100: Train the waybill sequence prediction model.

[0056] The training of the waybill sequence prediction model includes: constructing training samples corresponding to each capacity in the corresponding historical time slice based on the waybill set and capacity set of each historical time slice; and training the waybill sequence prediction model based on the training samples.

[0057] The training samples are labeled with the optimal waybill selection path for each capacity, obtained by solving the set of waybills and capacity based on the corresponding historical time slices. The optimal waybill selection path is the one that maximizes the scheduling performance of the capacity scheduling system. For example, the optimal waybill selection path is the one with the lowest score after scoring several waybill selection paths according to a preset path planning evaluation strategy.

[0058] First, when constructing training samples, training samples are built for each rider in the capacity set of each historical time slice, based on the order set and capacity set of each historical time slice. Taking riders as the capacity and a time slice of 1 minute in the capacity scheduling system as an example, assuming that the rider set of a certain historical time slice includes 15 riders and the order set includes 10 orders, the process of constructing training samples is the process of constructing training samples for each rider. That is, based on the rider set and capacity set of that historical time slice, 15 training samples are constructed. The sample data of each training sample represents the features that need to be considered when dispatching orders, and the sample label represents the optimal order selection path for that rider when faced with the aforementioned 10 orders.

[0059] In some embodiments of this application, the sample data includes, but is not limited to, one or more of the following data features: features of each waybill in the waybill set within the time slice (e.g., one or more features such as pickup and delivery location, pickup and delivery time, etc.), rider profile features (e.g., one or more features such as rider delivery speed, experience level, etc.), features of waybills carried by the rider (e.g., one or more features such as the number of waybills carried by the rider, pickup and delivery locations of each waybill, pickup and delivery time of each waybill, etc.), rider location information, and waybill delivery scenario features (e.g., features such as weather, weekdays, or holidays, etc.). Other features may also be selected for the sample features of each training sample in the embodiments of this application, which will not be listed individually in this embodiment.

[0060] In some embodiments of this application, the optimal waybill selection path, used as a sample label, is the optimal path for riders to select waybills, solved using a neighborhood search method based on an operations research optimization strategy. The waybill selection path can be represented by a waybill representation sequence.

[0061] In some embodiments of this application, solving for the optimal waybill selection path for a corresponding capacity based on the waybill set and capacity set of the corresponding historical time slice includes: for the waybill set and capacity set of a specified historical time slice, determining the waybill subset and capacity subset to be solved in the current round; for the waybill subset and capacity subset to be solved in the current round, solving for the first matching relationship between the capacity in the capacity subset and the waybill in the waybill subset by performing an iterative neighborhood search on the matching relationship between waybill and capacity; in response to the fact that the waybill in the waybill set and the capacity in the capacity set have not been matched, based on the waybill subset and capacity subset to be solved in the current round... The first matching relationship is obtained by solving the subset problem. The current round's order subset and capacity subset are then redefined. Next, the process jumps to the step of performing an iterative neighborhood search on the order and capacity subsets for the current round, solving for the first matching relationship between capacity in the capacity subset and order in the order subset. In response to the completion of the matching between order and capacity in the order subset, the optimal order selection path for each capacity in the capacity subset is determined based on the sequentially solved first matching relationships between capacity in the capacity subset and order in the order subset. Here, the first matching relationship is the optimal matching relationship that satisfies a preset operational optimization objective; for example, the first matching relationship between order and capacity is the order selection path that maximizes the overall scheduling performance of the capacity scheduling system.

[0062] Using the aforementioned set of waybills and capacity as the basis, multiple rounds of matching are performed on these sets, with each round determining one or more waybills for capacity selection. In the first round of matching, the waybill and capacity sets can be pruned (e.g., narrowing the matching range based on scheduling regions) to determine candidate sets of waybills and capacity (e.g., determining candidate capacity for each waybill and candidate waybills for each capacity). Based on the pruning results, a subset of waybills and a subset of capacity are constructed. Then, the first round of waybill and capacity matching is performed within these subsets, selecting the waybill with the highest matching degree for each capacity. By modifying the search results within the candidate sets and changing the combination of candidate sets, search quality can be effectively improved and search costs reduced.

[0063] In the embodiments of this application, the order selected by each delivery capacity is determined by performing multiple rounds of iterative neighborhood search, that is, determining the order with the highest matching degree for each delivery capacity. After multiple rounds of iterative neighborhood search, each delivery capacity in the delivery capacity subset will either select one order or not select one order; that is, each delivery capacity in the delivery capacity subset may match one order that meets the preset conditions, or it may not match any order. If not all the orders in the order subset are selected by the rider, the next round of the solution process needs to be performed to solve for the orders that were not selected by the rider. That is, the order subset is reconstructed from the orders that were not selected by the rider, and through another round of the solution process, the order with the highest matching degree for each delivery capacity is determined.

[0064] Assuming that each rider can be matched with an order in each round of the solution process, after multiple rounds, each rider in the capacity set will be matched with an order sequence, and the order at each sequence node is guaranteed to be the globally optimal solution. For example, after three rounds of the solution process, all orders in the order set are matched with the optimal capacity. Then, for a certain capacity R1 in the capacity set, if it selects an order W1 in the first round, does not select an order in the second round, and selects an order W3 in the third round, then the optimal order selection path for capacity R1 can be represented as {W1, None, W3}. Following this method, the optimal order selection path for each capacity in the capacity set can be determined.

[0065] In some embodiments of this application, the matching between the waybill set and the capacity set is determined after all waybills in the waybill set have been matched with capacity. In other embodiments of this application, the matching between the waybill set and the capacity set is determined after a preset number of rounds of solving.

[0066] The solution process will be explained in more detail below.

[0067] In some embodiments of this application, the step of solving the first matching relationship between the capacity in the capacity subset and the waybills in the waybill subset by performing an iterative neighborhood search on the matching relationship between waybills and capacity includes: determining an initial solution set for the capacity subset and the waybill subset, wherein the solutions in the initial solution set are used to indicate the initial matching relationship between each waybill in the waybill subset and each capacity in the capacity subset; performing a first search operator on the solutions in the initial solution set to determine a first neighborhood solution set, wherein the first search operator is used to change the matching relationship between waybills and capacity; and, in response to the existence of a solution in the first neighborhood solution set that satisfies a preset optimization condition, replacing the solution in the initial solution set targeted by the first search operator with the solution in the first neighborhood solution set that satisfies the preset optimization condition, and jumping to iterative execution of the first search operator on the solutions in the initial solution set. The steps include: determining a first neighborhood solution set; responding to the absence of a solution satisfying a preset optimization condition in the first neighborhood solution set, executing a second search operator on the solutions in the initial solution set to determine a second neighborhood solution set, wherein the second search operator is used to exchange the matching relationship between waybills and transport capacity; responding to the presence of a solution satisfying the preset optimization condition in the second neighborhood solution set, replacing the solution in the initial solution set targeted by the second search operator with the solution satisfying the preset optimization condition in the second neighborhood solution set, and determining a first matching relationship between transport capacity in the transport capacity subset and waybills in the waybill subset based on the solution in the initial solution set obtained after the most recent replacement; and responding to the absence of a solution satisfying the preset optimization condition in the second neighborhood solution set, jumping to the iterative execution of the step of executing the second search operator on the solutions in the initial solution set to determine the second neighborhood solution set.

[0068] First, for the current subset of transport capacity and subset of waybills to be solved, the matching relationship between the transport capacity in the transport capacity subset and each waybill in the waybill subset is initialized. In some embodiments of this application, the matching relationship between the transport capacity in the transport capacity subset and each waybill in the waybill subset can be initialized using historical waybill dispatch data. Then, an initial solution set for the transport capacity subset and the waybill subset is constructed based on the initialized matching relationship. In the embodiments of this application, the solution obtained by neighborhood search is the matching relationship between waybills and transport capacity.

[0069] In the waybill assignment process, both the search scope for waybills and riders undergo pruning, meaning both have candidate sets. Search results are then modified within these candidate sets. By altering the combination of candidate sets, search quality can be effectively improved and search costs reduced. This application defines two primitive semantic operators: a first search operator (e.g., named the shift operator) and a second search operator (e.g., named the double shift operator) to characterize more complex search actions. The first search operator specifies the best rider among the candidate waybills. Executing the first search operator results in a change in the number of waybills, such as transferring rider R1's waybill W1 to rider R3, which reduces rider R1's waybill count by 1 and increases rider R3's waybill count by 1. The second search operator is used to select a specific order from a candidate rider. By executing the second search operator, an order swap occurs, such as transferring order W1 from rider R1 to rider R3, and transferring order W2 from rider R3 to rider R1. The number of orders for riders R1 and R3 remains unchanged.

[0070] During the neighborhood search process, by executing the first search operator, the order matched by a certain rider is reassigned to other riders, changing the matching relationship between the order and the rider and the order. Then, it is determined whether the changed matching relationship meets the preset optimization conditions (for example, whether the changed matching relationship reduces the dispatch score of the rider and the order). If it does, the changed matching relationship can be considered a better solution (i.e., a better matching relationship). The changed matching relationship replaces the original matching relationship, and other matching relationships are tried to be changed until the changed matching relationship no longer meets the preset optimization conditions.

[0071] Next, the second search operator is executed to swap the orders matched between two riders, changing the matching relationship between orders and riders. Then, it is determined whether the changed matching relationship satisfies the preset optimization conditions (e.g., whether the scheduling score for riders and orders decreases). If it does, the changed matching relationship is considered a better solution (i.e., a superior matching relationship), and the changed matching relationship is output. If it does not, the order swapping between riders continues until a better matching relationship is found, and then the superior matching relationship is output.

[0072] Based on historical delivery data across multiple time slices, several training samples can be constructed. Then, a sequence prediction model is trained using these training samples. This model learns from the training samples the influence of the rider's current delivery orders and the order selected in the previous moment on the order selected in the next moment.

[0073] The following further explains the process of predicting riders' choice of waybills using the waybill sequence prediction model obtained through training.

[0074] During the online prediction process, the prediction of the candidate waybill selected by the capacity in the current round of waybill matching using a pre-trained waybill sequence prediction model includes: in response to the current round of waybill matching being the first round of waybill matching for the capacity set, predicting the candidate waybill selected by the capacity in the current round of waybill matching through the encoding network; and in response to the current round of waybill matching not being the first round of waybill matching for the capacity set, predicting the candidate waybill selected by the capacity in the current round of waybill matching through the decoding network. In the decision-making process of the entire path sequence, the first node has the greatest uncertainty and is the most difficult to predict. After the first node is determined, subsequent nodes depend on the determination of the previous node, resulting in lower uncertainty. Therefore, in this embodiment, the first waybill selected by the capacity is predicted separately by the encoding network 210 of the waybill sequence prediction model.

[0075] In some embodiments of this application, predicting the candidate waybill selected by the transport capacity in the current round of waybill matching through the encoding network includes: determining the first model input feature corresponding to the transport capacity in the first round of waybill matching, wherein the first model input feature includes one or more of the following: preset transport capacity profile features, preset first waybill features of the candidate waybill, waybill dispatch scenario features matched by the waybill set to be dispatched, and preset second waybill features of the transport capacity already carrying waybills; inputting the first model input feature into the encoding network of the waybill sequence prediction model, and performing classification mapping on the first model input feature through the encoding network; determining the waybill selected by the transport capacity in the current round of waybill matching based on the classification mapping result output by the encoding network. The preset first waybill feature may, for example, include one or more features among: pickup / delivery point features, pickup / delivery time features, and user information; the preset second waybill feature may, for example, include one or more features among: pickup / delivery point features, pickup / delivery time features, and user information.

[0076] For example, when rider R1 is already carrying two delivery orders w1 and w2, and faces the first choice under parallel assignment (such as choosing one delivery order from w3, w4, and w5), in predicting the first delivery order that rider R1 will choose during the decision-making process, the first model input features are determined based on rider R1's own information (such as rider R1's delivery speed, rider level, and geographical location), information on the delivery orders already carried (such as the pick-up and delivery points of delivery orders w1 and w2), information on the current delivery scenario (such as sunny day or Monday), and information on the delivery orders to be selected (such as the pick-up and delivery points of delivery orders w3, w4, and w5). Then, the determined first model input features are input into the encoding network 210 for classification mapping to estimate the probability of rider R1 choosing each of the delivery orders w3, w4, and w5. Finally, the delivery order with the highest probability is determined as the delivery order that rider R1 chooses when facing the first choice under parallel assignment (such as delivery order W3).

[0077] The output of the encoding network 210 includes classification results and context vectors. During the classification mapping process of the encoding network 210 based on the input features of the first model, the encoding network 210 generates a context vector (denoted as context1) based on current environmental variables such as preset capacity profile features, the dispatch scenario features matched by the set of waybills to be dispatched, and the preset second waybill features of the capacity already carrying waybills. Then, the context vector and the latent vector of the candidate waybill (e.g., the latent vector of the pick-up / delivery point) obtained by encoding the preset first waybill features of the candidate waybill are matched using a dot product. Then, a classification mapping is performed through an activation function, outputting the probability of each candidate waybill. Simultaneously, the encoding network 210 also outputs the generated context vector (e.g., context1). In the Long Short-Term Memory model, the context vectors output by the encoding and decoding networks are the latent vectors of the encoding network, representing the context information in a vector format. They carry the sequence information of the input sequence and are used as the encoding input for the next time step to improve prediction accuracy.

[0078] In some embodiments of this application, predicting the candidate waybill selected by the capacity in the current round of waybill matching through the decoding network includes: determining a second model input feature corresponding to the capacity in the current round of waybill matching, wherein the second model input feature includes a preset third waybill feature of the candidate waybill; inputting the second model input feature into the decoding network of the waybill sequence prediction model, and performing classification mapping on the first model input feature and the context vector generated by the waybill sequence prediction model in the previous round of waybill matching through the decoding network; and determining the candidate waybill selected by the capacity in the current round of waybill matching based on the classification mapping result output by the decoding network. The context vector generated by the waybill sequence prediction model includes the context vector output by the encoding network in the first round of waybill matching, or the context vector output by the decoding network in non-first rounds of waybill matching. The preset third feature can be a pickup / delivery point feature.

[0079] Taking rider R1 as an example, if rider R1 is already carrying orders w1 and w2, and faces parallel assignment, rider R1's first choice is order w3. When rider R1 faces a second choice for parallel assignment (such as choosing an order from orders w4, w6, and w7), the process is based on rider R1's own information (such as delivery speed, rider level, and geographical location), information about the orders already carried (such as the pick-up and delivery points of orders w1 and w2), information about the current order delivery scenario (such as sunny day, Monday), and information about the orders to be selected (such as orders w4, w6, and w7). Information such as the pick-up and delivery point of w7 is used to determine the input features of the second model. Then, these determined input features are used as the feature input of the decoding network 220. Simultaneously, the context vector `context1` output by the encoding network 210 during the classification mapping of rider R1's first choice is used as the context vector of the decoding network 220. The decoding network 220 performs classification mapping based on the second model input features and the context vector, estimating the probability that rider R1 will select each of the three delivery orders w4, w6, and w7 for dispatch. It also outputs the context vector for classification mapping in the next time step. Furthermore, based on the probabilities of each delivery order output by the decoding network 220, the delivery order selected by rider R1 when facing the second choice under parallel dispatch can be determined.

[0080] Similarly, when rider R1 faces subsequent choices such as the third and fourth choices in parallel assignment, predictions are made through the decoding network.

[0081] In this way, for each round of parallel matching, when a matching conflict occurs, the optimal waybill that the transport capacity will select can be predicted through the waybill sequence prediction model.

[0082] Step 150: Determine the current vessel waybill matching result based on the candidate waybills selected by each of the transport capacities in the current vessel waybill matching.

[0083] For the capacity that is matched with a waybill in parallel assignment, after the conflict resolution steps mentioned above, each capacity will be matched with at most one waybill in this round of waybill matching.

[0084] Step 160: In response to the completion of the dispatch of the set of waybills to be dispatched, the dispatch operation of the set of waybills to be dispatched to the capacity set is executed according to the waybill matching results of each round.

[0085] After N rounds of waybill matching and conflict resolution (N being an integer greater than or equal to 1), all waybills in the set to be dispatched may be assigned to the matched transport capacity. At this point, based on the waybill matching results of each round, for each transport capacity, the waybills selected by that capacity in each round of waybill matching are used as the final waybills assigned to that capacity. For example... Figure 4 As shown, after three rounds of parallel assignment, rider R1 selected waybill W3 in the first round of waybill matching, waybill W7 in the second round, and did not select any waybill in the third round. Therefore, it can be determined that waybills W3 and W7 from the list of waybills to be dispatched will be assigned to rider R1. After three rounds of parallel assignment, rider R2 selected waybill W5 in the first round of waybill matching, waybill W6 in the second round, and waybill W8 in the third round. Therefore, it can be determined that waybills W5, W6, and W8 from the list of waybills to be dispatched will be assigned to rider R2.

[0086] In some other embodiments of this application, after multiple rounds of waybill matching, there may still be waybills that have not been matched with riders. To improve the efficiency of waybill delivery, an upper limit can be set for the number of parallel waybill matching rounds. When the upper limit is reached, the set of waybills to be delivered is considered to have been completed. Figure 3 As shown, after determining the current vessel waybill matching result based on the candidate waybills selected in the current vessel waybill matching according to each of the aforementioned transport capacities, the method further includes: step 170.

[0087] Step 170: In response to the fact that the set of waybills to be dispatched has not been dispatched, update the set of waybills to be dispatched according to the current round of waybill matching results. For the updated set of waybills to be dispatched, jump to the step of performing waybill matching using the parallel assignment scheme to determine the matching relationship between each waybill in the set of waybills to be dispatched and each capacity in the capacity set.

[0088] If, after the current round of parallel order matching and conflict resolution, no orders in the order set are yet to be dispatched, then each order matched with available capacity is removed from the order set. For orders that are not matched with available capacity (including orders matched with riders during parallel assignment but not ultimately selected by the riders), the next round of order matching continues. That is, jump to step 120 and repeat step 120 and its subsequent steps until all orders in the order set are matched with available capacity or until the number of executions of step 120 reaches the upper limit of the parallel order matching rounds.

[0089] The waybill dispatching method disclosed in this application involves obtaining a capacity set and a set of waybills to be dispatched; performing waybill matching using a parallel dispatch scheme to determine the matching relationship between each waybill in the set of waybills to be dispatched and each capacity in the capacity set; determining candidate waybills corresponding to each capacity in the current round of waybill matching based on the matching relationship, wherein the candidate waybills are selected from the waybills in the set of waybills to be dispatched; for each capacity corresponding to at least two candidate waybills, predicting the candidate waybills selected by the capacity in the current round of waybill matching using a pre-trained waybill sequence prediction model; determining the waybill matching result for the current round based on the candidate waybills selected by each capacity in the current round of waybill matching; and, in response to the completion of the dispatching of the set of waybills to be dispatched, performing a dispatching operation from the set of waybills to be dispatched to the capacity set based on the waybill matching result of each round, which helps to improve the efficiency of waybill dispatching.

[0090] The waybill dispatching method disclosed in this application modeles the waybill assignment problem as a sequential decision problem, uses a sequential model to predict the waybills selected for capacity selection, solves the waybill matching conflict problem caused by parallel assignment, replaces the use of a fixed waybill selection strategy to solve the waybill matching conflict problem caused by parallel assignment, and takes into account the global waybill matching result, thereby improving the waybill dispatching efficiency.

[0091] On the other hand, by combining sequence learning with operations research optimization, the optimal path for selecting waybills is searched offline, and a waybill sequence prediction model is trained based on the optimal waybill selection path, which significantly improves the online waybill matching effect, further improves the waybill dispatch efficiency, and saves the online scheduling resources of the capacity scheduling system.

[0092] Example 2

[0093] This application discloses a waybill dispatching device, such as... Figure 5 As shown, the device includes:

[0094] The capacity set and waybill set acquisition module 510 is used to acquire the capacity set and the waybill set to be dispatched;

[0095] The parallel matching module 520 is used to perform waybill matching using a parallel assignment scheme to determine the matching relationship between each waybill in the set of waybills to be dispatched and each capacity in the set of transport capacity.

[0096] The candidate waybill determination module 530 is used to determine, according to the matching relationship, the candidate waybill corresponding to each capacity in the current round of waybill matching, wherein the candidate waybill is selected from the waybill set to be dispatched;

[0097] The candidate waybill selection module 540 is used to predict, for each capacity corresponding to at least two candidate waybills, the candidate waybill to be selected in the current round of waybill matching by using a pre-trained waybill sequence prediction model;

[0098] The waybill matching result determination module 550 is used to determine the waybill matching result of the current round based on the candidate waybills selected in the current round waybill matching for each of the transport capacities;

[0099] The waybill dispatch module 560 is used to respond to the completion of the dispatch of the waybill set to be dispatched, and to perform the dispatch operation of the waybill set to be dispatched to the transportation capacity set according to the waybill matching results of each round.

[0100] In some embodiments of this application, such as Figure 6 As shown, the device further includes:

[0101] The iterative judgment module 570 is used to update the set of waybills to be dispatched according to the current round of waybill matching results in response to the fact that the set of waybills to be dispatched has not been completed. For the updated set of waybills to be dispatched, the module jumps to execute the parallel matching module 520.

[0102] In some embodiments of this application, the waybill sequence prediction model includes: an encoding network and a decoding network, wherein the decoding network makes predictions based on the context output by the encoding network and the current input; such as Figure 6 As shown, the candidate waybill selection module 540 further includes:

[0103] The first candidate waybill selection submodule 5401 is used to predict the candidate waybill selected by the capacity in the current round of waybill matching in response to the current round of waybill matching being the first round of waybill matching for the capacity set;

[0104] The second candidate waybill selection submodule 5402 is used to predict the candidate waybill selected by the capacity in the current round of waybill matching in response to the current round of waybill matching not being the first round of waybill matching for the capacity set, through the decoding network.

[0105] In some embodiments of this application, the first candidate waybill selection submodule 5401 is further configured to:

[0106] The first model input features corresponding to the transport capacity in the first round of waybill matching are determined, wherein the first model input features include one or more of the following: preset transport capacity profile features, preset first waybill features of the candidate waybill, waybill dispatch scenario features matched by the set of waybills to be dispatched, and preset second waybill features of the transport capacity already carrying waybills.

[0107] The first model input features are input into the encoding network of the waybill sequence prediction model, and the first model input features are classified and mapped through the encoding network;

[0108] Based on the classification mapping results output by the coding network, the waybill selected by the transport capacity in the current round of waybill matching is determined.

[0109] In some embodiments of this application, the second candidate waybill selection submodule 5402 is further configured to:

[0110] Determine the second model input features corresponding to the transport capacity in the current round of waybill matching, wherein the second model input features include: the preset third waybill features of the candidate waybill;

[0111] The second model input features are input into the decoding network of the waybill sequence prediction model, and the decoding network performs classification mapping on the first model input features and the context vector generated by the waybill sequence prediction model in the previous round of waybill matching;

[0112] Based on the classification mapping results output by the decoding network, the candidate waybill selected by the transport capacity in the current round waybill matching is determined.

[0113] In some embodiments of this application, such as Figure 6 As shown, the device further includes:

[0114] The model training module 500 is used to construct training samples corresponding to each capacity in the corresponding historical time slice based on the order set and capacity set of each historical time slice. The sample label of the training sample is the optimal order selection path of the corresponding capacity obtained by solving based on the order set and capacity set of the corresponding historical time slice.

[0115] The model training module 500 is also used to train the waybill sequence prediction model based on the training samples.

[0116] In some embodiments of this application, the optimal route selection for a given capacity is determined based on the set of waybills and the set of capacity for the corresponding historical time slice, including:

[0117] For a given set of waybills and a set of capacity in a specified historical time slice, determine the subset of waybills and a subset of capacity to be solved in the current round;

[0118] For the current set of waybills and capacity subsets to be solved, an iterative neighborhood search is performed on the matching relationship between waybills and capacity to solve the first matching relationship between capacity in the capacity subset and waybills in the waybill subset;

[0119] In response to the fact that the waybills in the waybill set and the capacity in the capacity set have not been matched, the first matching relationship is obtained by solving the waybill subset and capacity subset to be solved for the current round. The waybill subset and capacity subset to be solved for the current round are re-determined. Then, the process jumps to the step of solving the first matching relationship between the capacity in the capacity subset and the waybill in the waybill set by performing an iterative neighborhood search on the matching relationship between the waybill and the capacity for the current round.

[0120] In response to the completion of the matching between the waybills in the waybill set and the capacity in the capacity set, the optimal waybill selection path for each capacity in the capacity set is determined based on the first matching relationship between the capacity in the capacity subset and the waybills in the waybill subset obtained by sequential solution.

[0121] In some embodiments of this application, the step of solving the first matching relationship between the capacity in the capacity subset and the waybills in the waybill subset by performing an iterative neighborhood search on the matching relationship between waybills and capacity includes:

[0122] Determine an initial solution set for the capacity subset and the waybill subset, wherein the solutions in the initial solution set are used to indicate the initial matching relationship between each waybill in the waybill subset and each capacity in the capacity subset;

[0123] A first search operator is executed on the solutions in the initial solution set to determine a first neighborhood solution set, wherein the first search operator is used to change the matching relationship between waybills and transport capacity;

[0124] In response to the existence of a solution that satisfies a preset optimization condition in the first neighborhood solution set, the solution in the initial solution set targeted by the first search operator is replaced by the solution in the first neighborhood solution set that satisfies the preset optimization condition, and the process jumps to the iterative execution step of executing the first search operator on the solutions in the initial solution set to determine the first neighborhood solution set.

[0125] In response to the absence of a solution satisfying the preset optimization conditions in the first neighborhood solution set, a second search operator is executed on the solutions in the initial solution set to determine the second neighborhood solution set, wherein the second search operator is used to exchange the matching relationship between waybills and transport capacity;

[0126] In response to the existence of a solution in the second neighborhood solution set that satisfies the preset optimization condition, the solution in the initial solution set targeted by the second search operator is replaced with the solution in the second neighborhood solution set that satisfies the preset optimization condition. Based on the solution in the initial solution set obtained after the most recent replacement, a first matching relationship between the transport capacity in the transport capacity subset and the waybill in the waybill subset is determined.

[0127] In response to the absence of a solution satisfying the preset optimization condition in the second neighborhood solution set, the process jumps to the step of iteratively executing the second search operator on the solutions in the initial solution set to determine the second neighborhood solution set.

[0128] The waybill dispatching device disclosed in this application is used to implement the waybill dispatching method described in Embodiment 1 of this application. The specific implementation methods of each module of the device will not be repeated here, but can be found in the specific implementation methods of the corresponding steps in the method embodiment.

[0129] The waybill dispatching device disclosed in this application acquires a capacity set and a set of waybills to be dispatched; performs waybill matching using a parallel dispatch scheme to determine the matching relationship between each waybill in the set of waybills to be dispatched and each capacity in the capacity set; determines candidate waybills corresponding to each capacity in the current round of waybill matching based on the matching relationship, wherein the candidate waybills are selected from the waybills in the set of waybills to be dispatched; for each capacity corresponding to at least two candidate waybills, predicts the candidate waybills selected by the capacity in the current round of waybill matching using a pre-trained waybill sequence prediction model; determines the waybill matching result for the current round based on the candidate waybills selected by each capacity in the current round of waybill matching; and, in response to the completion of the dispatch of the set of waybills to be dispatched, performs a dispatch operation from the set of waybills to be dispatched to the capacity set based on the waybill matching result of each round, thereby helping to improve the waybill dispatching efficiency.

[0130] The waybill dispatching device disclosed in this application modeles the waybill assignment problem as a sequential decision problem, uses a sequential model to predict the waybills selected for capacity selection, solves the waybill matching conflict problem caused by parallel assignment, replaces the use of a fixed waybill selection strategy to solve the waybill matching conflict problem caused by parallel assignment, and takes into account the global waybill matching result, thereby improving the waybill dispatching efficiency.

[0131] On the other hand, by combining sequence learning with operations research optimization, the optimal path for selecting waybills is searched offline, and a waybill sequence prediction model is trained based on the optimal waybill selection path, which significantly improves the online waybill matching effect, further improves the waybill dispatch efficiency, and saves the online scheduling resources of the capacity scheduling system.

[0132] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus embodiments, since they are substantially similar to the method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0133] The above provides a detailed description of a waybill dispatching method and apparatus provided by this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method of this application and its core idea. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the idea of ​​this application. Therefore, the content of this specification should not be construed as a limitation of this application.

[0134] 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. Those skilled in the art can understand and implement this without any creative effort.

[0135] The various component embodiments of this application can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some or all of the components in the electronic device according to the embodiments of this application. This application can also be implemented as a device or apparatus program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such a program implementing this application can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0136] For example, Figure 7An electronic device is shown that can implement the methods according to this application. The electronic device may be a PC, mobile terminal, personal digital assistant, tablet computer, etc. The electronic device conventionally includes a processor 710 and a memory 720, and program code 730 stored in the memory 720 and executable on the processor 710. When the processor 710 executes the program code 730, it implements the methods described in the above embodiments. The memory 720 may be a computer program product or a computer-readable medium. The memory 720 may be an electronic memory such as flash memory, EEPROM (Electrically Erasable Programmable Read-Only Memory), EPROM, hard disk, or ROM. The memory 720 has a storage space 7201 for the program code 730 of a computer program for performing any of the method steps described above. For example, the storage space 7201 for the program code 730 may include various computer programs for implementing the various steps in the above methods. The program code 730 is computer-readable code. These computer programs can be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, CDs, memory cards, or floppy disks. The computer program includes computer-readable code that, when executed on an electronic device, causes the electronic device to perform the method according to the above embodiments.

[0137] This application also discloses a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the waybill dispatching method as described in Embodiment 1 of this application.

[0138] Such a computer program product can be a computer-readable storage medium, which can have the same characteristics as... Figure 7 The memory 720 in the illustrated electronic device is similarly arranged with storage segments, storage spaces, etc. Program code can be stored, for example, in a compressed form on the computer-readable storage medium. The computer-readable storage medium is typically as shown in the reference... Figure 8 The portable or fixed storage unit is described above. Typically, the storage unit includes computer-readable code 730', which is code read by a processor and, when executed by the processor, implements the various steps of the method described above.

[0139] The terms "an embodiment," "embodiment," or "one or more embodiments" as used herein mean that a particular feature, structure, or characteristic described in connection with an embodiment is included in at least one embodiment of this application. Furthermore, please note that the examples of the phrase "in one embodiment" do not necessarily all refer to the same embodiment.

[0140] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of this application may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0141] In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. This application can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0142] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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. Such 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 this application.

Claims

1. A method for distributing waybills, characterized in that, include: Obtain the capacity set and the set of waybills to be dispatched; A parallel assignment scheme is used to perform waybill matching, and the matching relationship between each waybill in the set of waybills to be dispatched and each capacity in the set of capacity is determined; Based on the matching relationship, a candidate waybill corresponding to each capacity in the current round of waybill matching is determined, and the candidate waybill is selected from the waybill set to be dispatched; For each capacity corresponding to at least two candidate waybills, the candidate waybill selected by the capacity in the current round of waybill matching is predicted by a pre-trained waybill sequence prediction model; The current round waybill matching result is determined based on the candidate waybills selected by each of the aforementioned transport capacities in the current round waybill matching; In response to the completion of the dispatch of the set of waybills to be dispatched, the dispatch operation of the set of waybills to be dispatched to the set of transport capacity is executed according to the matching results of each round of waybills.

2. The method according to claim 1, characterized in that, After the step of determining the current vessel waybill matching result based on the candidate waybills selected in the current vessel waybill matching according to each of the said transport capacities, the method further includes: In response to the fact that the set of waybills to be dispatched has not been fully dispatched, the set of waybills to be dispatched is updated according to the current round of waybill matching results. For the updated set of waybills to be dispatched, the process jumps to the step of performing waybill matching using a parallel dispatch scheme to determine the matching relationship between each waybill in the set of waybills to be dispatched and each capacity in the capacity set.

3. The method according to claim 1 or 2, characterized in that, The waybill sequence prediction model includes: An encoding network and a decoding network, wherein the decoding network makes predictions based on the context of the output of the encoding network and the current input; The step of predicting the candidate waybills selected in the current round of waybill matching using a pre-trained waybill sequence prediction model includes: In response to the current round of waybill matching being the first round of waybill matching for the capacity set, the candidate waybill selected by the capacity in the current round of waybill matching is predicted through the coding network; In response to the fact that the current round of waybill matching is not the first round of waybill matching for the capacity set, the decoding network predicts the candidate waybill selected by the capacity in the current round of waybill matching.

4. The method according to claim 3, characterized in that, The step of predicting the candidate waybills selected in the current round of waybill matching using the coding network includes: The first model input features corresponding to the transport capacity in the first round of waybill matching are determined, wherein the first model input features include one or more of the following: preset transport capacity profile features, preset first waybill features of the candidate waybill, waybill dispatch scenario features matched by the set of waybills to be dispatched, and preset second waybill features of the transport capacity already carrying waybills. The first model input features are input into the encoding network of the waybill sequence prediction model, and the first model input features are classified and mapped through the encoding network; Based on the classification mapping results output by the coding network, the waybill selected by the transport capacity in the current round of waybill matching is determined.

5. The method according to claim 3, characterized in that, The step of predicting the candidate waybills selected by the capacity in the current round of waybill matching through the decoding network includes: Determine the second model input features corresponding to the transport capacity in the current shipment order matching, wherein the second model input features include: The preset third waybill features of the candidate waybill; The second model input features are input into the decoding network of the waybill sequence prediction model, and the decoding network performs classification mapping on the second model input features and the context vector generated by the waybill sequence prediction model in the previous round of waybill matching; Based on the classification mapping results output by the decoding network, the candidate waybill selected by the transport capacity in the current round waybill matching is determined.

6. The method according to claim 1 or 2, characterized in that, Before the step of predicting the candidate waybills selected in the current round of waybill matching using a pre-trained waybill sequence prediction model, the method further includes: Based on the order set and capacity set of each historical time slice, a training sample corresponding to each capacity in the corresponding historical time slice is constructed. The sample label of the training sample is the optimal order selection path of the corresponding capacity obtained by solving based on the order set and capacity set of the corresponding historical time slice. The waybill sequence prediction model is trained based on the training samples.

7. The method according to claim 6, characterized in that, Based on the set of waybills and the set of capacity for the corresponding historical time slice, the optimal waybill selection path for the corresponding capacity is solved, including: For a given set of waybills and a set of capacity in a specified historical time slice, determine the subset of waybills and a subset of capacity to be solved in the current round; For the current set of waybills and capacity subsets to be solved, an iterative neighborhood search is performed on the matching relationship between waybills and capacity to solve the first matching relationship between capacity in the capacity subset and waybills in the waybill subset; In response to the fact that the waybills in the waybill set and the capacity in the capacity set have not been matched, the first matching relationship is obtained by solving the waybill subset and capacity subset to be solved for the current round. The waybill subset and capacity subset to be solved for the current round are re-determined. Then, the process jumps to the step of solving the first matching relationship between the capacity in the capacity subset and the waybill in the waybill set by performing an iterative neighborhood search on the matching relationship between the waybill and the capacity for the current round. In response to the completion of the matching between the waybills in the waybill set and the capacity in the capacity set, the optimal waybill selection path for each capacity in the capacity set is determined based on the first matching relationship between the capacity in the capacity subset and the waybills in the waybill subset obtained by sequential solution.

8. The method according to claim 7, characterized in that, The step of performing an iterative neighborhood search on the matching relationship between waybills and transport capacity to solve the first matching relationship between the transport capacity subset and the waybills in the waybill subset includes: Determine an initial solution set for the capacity subset and the waybill subset, wherein the solutions in the initial solution set are used to indicate the initial matching relationship between each waybill in the waybill subset and each capacity in the capacity subset; A first search operator is executed on the solutions in the initial solution set to determine a first neighborhood solution set, wherein the first search operator is used to change the matching relationship between waybills and transport capacity; In response to the existence of a solution that satisfies a preset optimization condition in the first neighborhood solution set, the solution in the initial solution set targeted by the first search operator is replaced by the solution in the first neighborhood solution set that satisfies the preset optimization condition, and the process jumps to the iterative execution step of executing the first search operator on the solutions in the initial solution set to determine the first neighborhood solution set. In response to the absence of a solution satisfying the preset optimization conditions in the first neighborhood solution set, a second search operator is executed on the solutions in the initial solution set to determine the second neighborhood solution set, wherein the second search operator is used to exchange the matching relationship between waybills and transport capacity; In response to the existence of a solution in the second neighborhood solution set that satisfies the preset optimization condition, the solution in the initial solution set targeted by the second search operator is replaced with the solution in the second neighborhood solution set that satisfies the preset optimization condition. Based on the solution in the initial solution set obtained after the most recent replacement, a first matching relationship between the transport capacity in the transport capacity subset and the waybill in the waybill subset is determined. In response to the absence of a solution satisfying the preset optimization condition in the second neighborhood solution set, the process jumps to the step of iteratively executing the second search operator on the solutions in the initial solution set to determine the second neighborhood solution set.

9. A waybill dispatching device, characterized in that, include: The module for obtaining the capacity set and the waybill set is used to obtain the capacity set and the waybill set to be dispatched; The parallel matching module is used to perform waybill matching using a parallel assignment scheme to determine the matching relationship between each waybill in the set of waybills to be dispatched and each capacity in the set of transport capacity. The candidate waybill determination module is used to determine, based on the matching relationship, the candidate waybill corresponding to each capacity in the current round of waybill matching, wherein the candidate waybill is selected from the waybill set to be dispatched; The candidate waybill selection module is used to predict, for each capacity corresponding to at least two candidate waybills, the candidate waybill to be selected in the current round of waybill matching by using a pre-trained waybill sequence prediction model; The waybill matching result determination module is used to determine the waybill matching result of the current round based on the candidate waybills selected in the waybill matching of each of the transport capacities in the current round; The waybill dispatch module is used to respond to the completion of the dispatch of the waybill set to be dispatched, and to perform the dispatch operation of the waybill set to be dispatched to the capacity set according to the waybill matching results of each round.

10. An electronic device, comprising a memory, a processor, and program code stored in the memory and executable on the processor, characterized in that, When the processor executes the program code, it implements the waybill dispatching method according to any one of claims 1 to 8.

11. A computer-readable storage medium having program code stored thereon, characterized in that, When the program code is executed by the processor, it implements the steps of the waybill dispatching method according to any one of claims 1 to 8.

Images