Transportation scheduling method and device, electronic equipment and storage medium

Abstract

The present disclosure provides a transportation scheduling method and device, electronic equipment and storage medium, which relates to the technical field of artificial intelligence such as deep learning and intelligent scheduling. The specific implementation scheme is: based on the attribute information of a plurality of goods, the attribute information of a plurality of transportation tools, a target function and a constraint, a machine learning model is used to determine the first confidence of at least one good placed in each transportation tool under the condition that the constraint is met and the target function is minimized, to determine the corresponding relationship between the plurality of goods and the plurality of transportation tools based on the first confidence of at least one good placed in each transportation tool, and the target function represents the imbalance of the plurality of transportation tools in resources; in the case that the target function value of the corresponding relationship meets the set threshold, the corresponding relationship is determined as the target transportation relationship between the plurality of goods and the plurality of transportation tools. It can guarantee that the imbalance of the plurality of transportation tools in resources is minimized, thereby reducing the transportation cost in the transportation process.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, specifically to the fields of deep learning and intelligent scheduling technology, and particularly to transportation scheduling methods, devices, electronic devices and storage media. Background Technology

[0002] In recent years, with the continuous development of society, transportation has become an important link in the circulation of goods. When transporting goods using multiple means of transport, such as vehicles or ships, it is crucial to reduce transportation costs during the process. Summary of the Invention

[0003] This disclosure provides a transportation scheduling method, apparatus, electronic device, and storage medium.

[0004] According to one aspect of this disclosure, a transportation scheduling method is provided, the method comprising: acquiring attribute information of multiple items, attribute information of multiple transportation vehicles, an objective function, and constraints; wherein the objective function represents the resource imbalance of the multiple transportation vehicles; based on the attribute information of the multiple items, the attribute information of the multiple transportation vehicles, the objective function, and the constraints, employing a machine learning model to determine a first confidence level for placing at least one item in each of the transportation vehicles while satisfying the constraints and minimizing the objective function, thereby determining a correspondence between the multiple items and the multiple transportation vehicles based on the first confidence level of placing at least one item in each of the transportation vehicles; and determining the correspondence as a target transportation relationship between the multiple items and the multiple transportation vehicles when the objective function value of the correspondence satisfies a set threshold.

[0005] According to another aspect of this disclosure, a method for training a machine learning model for transportation scheduling is provided. The method includes: acquiring multiple training samples, wherein at least one of the training samples includes attribute information of multiple sample items, attribute information of multiple sample transportation vehicles, an objective function, and constraints; the objective function represents the resource imbalance of the multiple sample transportation vehicles; for at least one training sample, based on the attribute information of the multiple sample items, the attribute information of the multiple sample transportation vehicles, the objective function, and the constraints, using a machine learning model to determine a third confidence level for placing at least one sample item in each of the sample transportation vehicles while satisfying the constraints and minimizing the objective function; and training the machine learning model based on the third confidence level of placing at least one sample item in each of the sample transportation vehicles in at least one training sample.

[0006] According to another aspect of this disclosure, a transportation scheduling device is provided, the device comprising: a first acquisition module, configured to acquire attribute information of multiple items, attribute information of multiple transportation vehicles, an objective function, and constraints; wherein the objective function represents the resource imbalance of the multiple transportation vehicles; a first determination module, configured to, based on the attribute information of the multiple items, the attribute information of the multiple transportation vehicles, the objective function, and the constraints, use a machine learning model to determine a first confidence level for placing at least one of the items in each of the transportation vehicles while satisfying the constraints and minimizing the objective function, so as to determine the correspondence between the multiple items and the multiple transportation vehicles based on the first confidence level of placing at least one of the items in each of the transportation vehicles; and a second determination module, configured to, when the objective function value of the correspondence satisfies a set threshold, determine the correspondence as a target transportation relationship between the multiple items and the multiple transportation vehicles.

[0007] According to another aspect of this disclosure, a training apparatus for a machine learning model for transportation scheduling is provided. The apparatus includes: a second acquisition module for acquiring multiple training samples, wherein at least one of the training samples includes attribute information of multiple sample items, attribute information of multiple sample transportation vehicles, an objective function, and constraints; the objective function represents the resource imbalance of the multiple sample transportation vehicles; a third determination module for determining, for at least one training sample, a third confidence level for placing at least one sample item in each of the sample transportation vehicles while satisfying the constraints and minimizing the objective function, based on the attribute information of the multiple sample items, the attribute information of the multiple sample transportation vehicles, the objective function, and the constraints, using a machine learning model; and a training module for training the machine learning model based on the third confidence level of placing at least one sample item in each of the sample transportation vehicles in at least one training sample.

[0008] According to another aspect of this disclosure, an electronic device is provided, 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 transportation scheduling method of this disclosure, or to perform a training method for a machine learning model for transportation scheduling of this disclosure.

[0009] According to another aspect of this disclosure, a non-transitory computer-readable storage medium is provided that stores computer instructions for causing the computer to execute the transportation scheduling method disclosed in embodiments of this disclosure, or to execute the training method for a machine learning model for transportation scheduling disclosed in embodiments of this disclosure.

[0010] According to another aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the transportation scheduling method of this disclosure, or the steps of the training method of the machine learning model for transportation scheduling of this disclosure.

[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0012] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0013] Figure 1 This is a schematic flowchart of a transportation scheduling method according to the first embodiment of this disclosure;

[0014] Figure 2 This is a flowchart illustrating the transportation scheduling method according to the second embodiment of this disclosure;

[0015] Figure 3 This is an architecture diagram of a transportation scheduling method according to a second embodiment of the present disclosure;

[0016] Figure 4 This is a flowchart illustrating a transportation scheduling method according to a third embodiment of the present disclosure;

[0017] Figure 5 This is a flowchart illustrating a method for training a machine learning model for transportation scheduling according to a fourth embodiment of the present disclosure.

[0018] Figure 6 This is a flowchart illustrating a method for training a machine learning model for transportation scheduling according to a fifth embodiment of this disclosure.

[0019] Figure 7 This is a schematic diagram of the structure of a transportation dispatching device according to the sixth embodiment of this disclosure;

[0020] Figure 8 This is a schematic diagram of the structure of a training apparatus for a machine learning model for transportation scheduling according to the seventh embodiment of this disclosure;

[0021] Figure 9 This is a block diagram of an electronic device used to implement the transportation scheduling method or the training method of a machine learning model for transportation scheduling according to the embodiments of this disclosure. Detailed Implementation

[0022] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0023] It is understandable that means of transportation, such as vehicles or ships, possess various resources, including weight resources (the weight the means of transportation can bear), spatial resources (the space the means of transportation has), and other types of resources. Goods also possess various resources, including weight resources (the weight of the goods), spatial resources (the space the goods occupy), and other types of resources.

[0024] Taking weight as an example, suppose the means of transportation are ships, and there are two ships, each with a weight resource of 10 tons. All the goods to be transported also have a weight resource of 10 tons. When transporting multiple goods using multiple means of transport, if the overall resources of the various means of transport are unbalanced—for example, one ship carries 9 tons of goods while another carries 1 ton—the ship carrying 1 ton will have low resource utilization and may be running empty, while the ship carrying 9 tons will consume fuel too quickly during transport and incur additional unexpected costs such as needing to switch ships or requiring rescue, resulting in high transportation costs.

[0025] To reduce transportation costs, resources need to be allocated rationally to ensure the overall balance of resources across all modes of transport. This requires deciding which mode of transport to load each item onto when transporting multiple goods using multiple modes of transport, in order to minimize the imbalance of resources across all modes of transport.

[0026] Related technologies, to determine which transportation vehicle to load each item onto, can use solvers such as SCIP (Solving Constraint Integer Programming) to solve this problem. Users can input the necessary information into the solver and set its parameters. The solver can then automatically invoke internal heuristic algorithms, such as nearest neighbor search, to find feasible solutions until the problem is solved. This method is time-consuming. To improve efficiency, for each item and each transportation vehicle, a binary classification model can be used to predict the probability of whether the item is placed on that vehicle, and then these probabilities can be used to quickly determine the correspondence between items and transportation vehicles. However, this binary classification model predicts the probability of an item being placed on a particular vehicle. When the input features are similar, the predicted probability can easily hover around 0.5, leading to predictions that are close to random guesses and resulting in inaccurate final correspondences between multiple items and multiple transportation vehicles.

[0027] This disclosure provides a transportation scheduling method and a training method for a machine learning model used for transportation scheduling, which can rationally allocate resources, ensure the overall resource balance of various transportation tools, and thus reduce transportation costs during the transportation process. The transportation scheduling method includes: acquiring attribute information of multiple items, attribute information of multiple transportation tools, an objective function, and constraints; wherein the objective function represents the resource imbalance among multiple transportation tools; based on the attribute information of multiple items, the attribute information of multiple transportation tools, the objective function, and the constraints, using a machine learning model to determine a first confidence level for placing at least one item in each transportation tool while satisfying the constraints and minimizing the objective function; determining the correspondence between multiple items and multiple transportation tools based on the first confidence level of placing at least one item in each transportation tool; and determining the correspondence as the target transportation relationship between multiple items and multiple transportation tools when the objective function value of the correspondence meets a set threshold. This ensures that the resource imbalance among multiple transportation tools is minimized, thereby reducing transportation costs during the transportation process.

[0028] The transportation scheduling method, training method for machine learning models used in transportation scheduling, apparatus, electronic equipment, non-transitory computer-readable storage medium, and computer program products disclosed herein relate to the field of artificial intelligence technology, specifically the fields of deep learning and intelligent scheduling technology.

[0029] Artificial intelligence (AI) is the study of enabling computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, and planning). It encompasses both hardware and software technologies. AI hardware technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, and big data processing. AI software technologies primarily include computer vision, speech recognition, natural language processing, machine learning / deep learning, big data processing, and knowledge graph technologies.

[0030] Deep learning is a new research direction in the field of machine learning. It has been introduced into machine learning to bring it closer to its original goal—artificial intelligence. Deep learning learns the inherent patterns and hierarchical representations of sample data. The information gained during this learning process greatly helps in interpreting data such as text, images, and sound. Its ultimate goal is to enable machines to have analytical and learning capabilities like humans, and to be able to recognize data such as text, images, and sound.

[0031] The following description, with reference to the accompanying drawings, outlines embodiments of a transportation scheduling method, a training method for a machine learning model for transportation scheduling, an apparatus, an electronic device, a non-transitory computer-readable storage medium, and a computer program product.

[0032] It should be noted that the acquisition, storage, and application of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0033] First, the transportation scheduling method provided in the embodiments of this disclosure will be described.

[0034] Figure 1 This is a flowchart illustrating the transportation scheduling method according to the first embodiment of this disclosure. It should be noted that the transportation scheduling method of this embodiment is executed by a transportation scheduling device, which can be implemented by software and / or hardware. The transportation scheduling device can be configured in an electronic device, which may include, but is not limited to, terminal devices, servers, etc. This embodiment does not specifically limit the type of electronic device.

[0035] like Figure 1 As shown, the transportation scheduling method may include:

[0036] Step 101: Obtain the attribute information of multiple items, the attribute information of multiple transportation tools, the objective function, and the constraints; wherein, the objective function represents the imbalance of resources among multiple transportation tools.

[0037] The term "item" refers to any item that needs to be transported, such as a box containing food or a bag containing clothing. This embodiment does not limit the scope of the invention.

[0038] The attribute information of an item may include any attribute information related to the item, such as the quantity of the item, the size of the space occupied by the item, and the weight of the item, etc. This disclosure does not limit this.

[0039] The means of transport can be any tool capable of transporting goods, such as a ship or vehicle, and this disclosure does not limit this.

[0040] The attribute information of the means of transport may include any attribute information related to the means of transport, such as the quantity of means of transport, the space and weight capacity of the means of transport, etc. This disclosure does not limit this.

[0041] The objective function represents the functional relationship between the desired outcome and relevant factors when placing multiple items onto multiple transportation vehicles. In this embodiment, the objective function represents the resource imbalance among the multiple transportation vehicles and can be used to indicate the optimal solution for the target transportation relationship between multiple items and multiple transportation vehicles. For example, the solution that minimizes the objective function value is the optimal solution. The target transportation relationship between multiple items and multiple transportation vehicles indicates which transportation vehicle each item should be loaded onto for transportation.

[0042] Constraints are restrictions that must be followed when placing multiple items onto multiple modes of transport. For example, each item must be placed on only one mode of transport, and the total amount of resources carried by each mode of transport cannot exceed the total amount of resources that mode of transport can carry.

[0043] Step 102: Based on the attribute information of multiple items, the attribute information of multiple transportation vehicles, the objective function, and the constraints, a machine learning model is used to determine the first confidence level of at least one item placed in each transportation vehicle while satisfying the constraints and minimizing the objective function. Based on the first confidence level of at least one item placed in each transportation vehicle, the correspondence between multiple items and multiple transportation vehicles is determined.

[0044] The first confidence level represents the likelihood that an item will be placed on a means of transport.

[0045] The machine learning model can be any neural network model capable of multi-class classification, and this disclosure does not impose any restrictions on it.

[0046] In an example embodiment, for at least one item, feature information can be extracted from the attribute information of multiple items, the attribute information of multiple means of transport, the objective function, and constraints. The feature information is then input into a machine learning model. The machine learning model, based on the attribute information of multiple items, the attribute information of multiple means of transport, the objective function, and constraints, predicts the first confidence level of placing the item in each means of transport when the constraints are satisfied and the objective function is minimized.

[0047] In the example embodiment, after determining the first confidence level for placing at least one item into each mode of transport, for any one of the at least one items, the mode of transport with the highest first confidence level that is greater than a preset confidence threshold can be identified as the target mode of transport for that item, thereby determining the correspondence between the at least one item and multiple modes of transport. Thus, when the quantity of the at least one item is equal to the total quantity of all items, the correspondence between multiple items and multiple modes of transport can be determined. Furthermore, when the quantity of the at least one item is less than the total quantity of all items, for the other items besides the at least one item, the target modes of transport for those other items can be determined using a solver, based on experience, or through other means; this disclosure does not impose any limitations on this.

[0048] Step 103: If the objective function of the correspondence satisfies the set threshold, the correspondence is determined as the target transportation relationship between multiple items and multiple means of transport.

[0049] The threshold setting is a pre-defined threshold that the objective function needs to satisfy, and it can be set arbitrarily as needed. This embodiment of the present disclosure does not impose any restrictions on it.

[0050] The objective function is defined as the imbalance of resources among multiple modes of transportation.

[0051] In the example embodiment, after determining the correspondence between multiple items and multiple transportation vehicles based on a first confidence level of at least one item being placed in each transportation vehicle, the objective function value of this correspondence can be determined. If the objective function value is less than a set threshold, this correspondence can be determined as the target transportation relationship between multiple items and multiple transportation vehicles. Thus, the target transportation relationship between multiple items and multiple transportation vehicles can be determined under the condition that the constraints are satisfied, the objective function is minimized, and the objective function value satisfies the set threshold.

[0052] The transportation scheduling method provided in this disclosure employs a machine learning model to determine the first confidence level of at least one item being placed in each means of transport, satisfying constraints and minimizing the objective function. Based on this first confidence level, the method determines the correspondence between multiple items and multiple means of transport. When the objective function of the correspondence meets a set threshold, the correspondence is determined as the target transportation relationship between multiple items and multiple means of transport. Compared to related technologies that use a binary classification model to predict the probability of each item being placed in each means of transport to determine the transportation relationship, this method allows the machine learning model to avoid lazy random guessing, truly learning the internal knowledge of the problem, improving the model's predictive ability, and obtaining a more accurate target transportation relationship. This maximizes the balance of overall resources of each means of transport during the transportation process and reduces transportation costs. Furthermore, the transportation scheduling method provided in this embodiment, compared to solving the problem through a solver, does not require automatically calling internal heuristic algorithms such as nearest neighbor search to solve feasible solutions to the problem based on user settings until the solution is completed. It can quickly determine the target transportation relationship between multiple items and multiple transportation vehicles, which is highly efficient. Moreover, the correspondence between multiple items and multiple transportation vehicles obtained based on the machine learning model is more accurate.

[0053] In summary, the transportation scheduling method provided in this disclosure acquires attribute information of multiple items, attribute information of multiple transportation vehicles, an objective function, and constraints. The objective function represents the resource imbalance among the multiple transportation vehicles. Based on the attribute information of the multiple items, the attribute information of the multiple transportation vehicles, the objective function, and the constraints, a machine learning model is used to determine a first confidence level for placing at least one item in each transportation vehicle while satisfying the constraints and minimizing the objective function. Based on the first confidence level of placing at least one item in each transportation vehicle, the correspondence between the multiple items and the multiple transportation vehicles is determined. When the objective function value of the correspondence meets a set threshold, the correspondence is determined as the target transportation relationship between the multiple items and the multiple transportation vehicles. Therefore, the resource imbalance among the multiple transportation vehicles can be minimized, thereby reducing transportation costs during the transportation process.

[0054] The following is combined with Figure 2 The present disclosure provides a further explanation of the process by which, based on the attribute information of multiple items, the attribute information of multiple transportation vehicles, the objective function, and constraints, a machine learning model is used to determine the first confidence level of placing at least one item into each transportation vehicle while satisfying the constraints and minimizing the objective function, and the process by which, based on the first confidence level of placing at least one item into each transportation vehicle, the correspondence between multiple items and multiple transportation vehicles is determined.

[0055] Figure 2 This is a flowchart illustrating the transportation scheduling method according to the second embodiment of this disclosure. Figure 2 As shown, the transportation scheduling method may include the following steps:

[0056] Step 201: Obtain the attribute information of multiple items, the attribute information of multiple transportation tools, the objective function, and the constraints; wherein, the objective function represents the imbalance of resources among multiple transportation tools.

[0057] The relevant explanations for step 201 can be found in step 101, and will not be repeated here.

[0058] Step 202: Input the attribute information of multiple items, the attribute information of multiple means of transportation, the objective function, and the constraints into the solver, so as to obtain multiple first variables X through the solver. i,b The corresponding first feature vector, multiple second variables Y b,r The corresponding second eigenvector and the constraint corresponding third eigenvector.

[0059] Among them, the first variable X i,b The second variable Y represents whether to put the i-th item into the b-th transport vehicle. b,r Let i represent the imbalance value of the b-th means of transport on the r-th resource, where i is an integer between 1 and I, b is an integer between 1 and B, r is an integer between 1 and R, I is the quantity of goods, B is the quantity of means of transport, and R is the quantity of different types of resources. I, B, and R are integers greater than or equal to 1.

[0060] Among them, the first variable X i,b Let X be a binary variable. When the first variable is 1, it indicates that the i-th item is placed in the b-th transport vehicle; when the first variable is 0, it indicates that the i-th item is not placed in the b-th transport vehicle. That is, for any i and b, the first variable X... i,b The value can be either 0 or 1.

[0061] Second variable Y b,r Y can represent the normalized imbalance value of the b-th mode of transport on the r-th resource. That is, for any b and r, the second variable Y b,r The value of can be a number greater than or equal to 0 and less than or equal to 1.

[0062] It should be noted that after obtaining the second variable, the third variable Z can also be obtained based on the second feature vector of the second variable. r The corresponding eigenvectors. Among them, the third variable Z... rThis can represent the maximum normalized imbalance value of the r-th resource across all modes of transport. Correspondingly, the third variable can take values ​​greater than or equal to 0 and less than or equal to 1.

[0063] The constraint of placing multiple items into multiple modes of transport can include the following four constraints:

[0064] 1. All items must be placed on only one means of transport:

[0065] Right now

[0066] 2. The resources of the goods loaded on a means of transport shall not exceed the resource capacity of that means of transport:

[0067]

[0068] Among them, Size i,r Capacity represents the amount of the r-th resource required for the i-th item, and is a constant. b,r Let r represent the quantity of the r-th resource of the b-th means of transport, which is a constant.

[0069] 3. The imbalance value of the b-th means of transport on the r-th resource satisfies the following formula (3):

[0070]

[0071] IV. Each imbalance value is less than the maximum imbalance value:

[0072]

[0073] The objective function can be:

[0074] B×R×∑ r Z r +∑ b,r Y b,r (5)

[0075] The first eigenvector represents the characteristic information of the first variable; the second eigenvector represents the characteristic information of the second variable; and the third eigenvector represents the characteristic information of the constraints. The characteristic information may include, for example, the values ​​on the left side of the equals sign in the constraints, the parameters in the constraints, and the values ​​on the right side of the equals sign.

[0076] In the example embodiment, the attribute information of multiple items, the attribute information of multiple means of transportation, the objective function, and the constraints can be input into the solver. The solver then performs multiple iterations to solve for the correspondence between the multiple items and the multiple means of transportation. During the solver's solution process, initialization can be performed by extracting features from the attribute information of the multiple items, the attribute information of the multiple means of transportation, the objective function, and the constraints, thereby obtaining multiple first variables X. i,b The corresponding first feature vector, multiple second variables Y b,r The corresponding second eigenvector and the constraint's corresponding third eigenvector. Furthermore, to enrich the feature information of the first variable, the second variable, and the constraint, multiple first variables X can be obtained by combining this correspondence with the extracted features after the solver obtains a correspondence between multiple items and multiple means of transport in one iteration. i,b The corresponding first feature vector, multiple second variables Y b,r The corresponding second eigenvector and the constraint corresponding third eigenvector.

[0077] The number of third eigenvectors constrained can be one or more, and this embodiment does not limit this.

[0078] The solver can be any solver tool capable of solving mixed integer optimization problems, and this disclosure does not limit this.

[0079] Step 203: Based on each first feature vector, each second feature vector, and the third feature vector, a machine learning model is used to determine the first confidence level of placing at least one item into each means of transport.

[0080] In an example embodiment, the first feature vector corresponding to each first variable, the second feature vector corresponding to each second variable, and the third feature vector corresponding to the constraint can be input into a machine learning model to predict the first confidence level of at least one item being placed in each mode of transport. Alternatively, in another example embodiment, the first feature vector corresponding to each first variable, the second feature vector corresponding to each second variable, the feature vector corresponding to the third variable, and the third feature vector corresponding to the constraint can be input into a machine learning model to predict the first confidence level of at least one item being placed in each mode of transport.

[0081] In an example embodiment, the machine learning model may include a feature extraction module and a normalization module connected in sequence; correspondingly, step 203 can be implemented in the following way: input each first feature vector, each second feature vector, and the third feature vector into the feature extraction module, so that the feature extraction module fuses the first feature vector corresponding to each first variable with the other feature vectors in the feature vectors input to the feature extraction module except for the first feature vector, to obtain the fourth feature vector corresponding to each first variable, and combines the fourth feature vectors corresponding to each first variable in the target order to obtain the first vector matrix; input the first vector matrix into the normalization module, so that the normalization module determines the first confidence level of at least one item being placed in each means of transport.

[0082] The target order refers to the order in which the fourth eigenvectors corresponding to each first variable are combined to obtain the first vector matrix, and can be preset as needed. For example, the target order could be: first variable X... i,b The corresponding fourth eigenvector is the element in the i-th row and b-th column of the first vector matrix.

[0083] In an example embodiment, the feature extraction module may include an input layer, a graph convolutional neural network layer, a fully connected layer, and an activation layer connected in sequence. In this example embodiment, each first feature vector, each second feature vector, and each third feature vector can be input into the input layer. The input layer integrates these feature vectors, mapping them to the same dimension to obtain feature vectors of the same dimension corresponding to each first variable, each second variable, and each constraint. Next, the feature vectors of the same dimension corresponding to each first variable, each second variable, and each constraint output from the input layer can be input into the graph convolutional neural network layer. The graph convolutional neural network layer captures the features of relationships between variables and between variables and constraints based on these feature vectors. These captured features are then fused into the feature vector corresponding to the first variable. Furthermore, through the fully connected layer and the activation layer, non-linear features are extracted from the features of the first variable to obtain the fourth feature vector of each first variable. Finally, these fourth feature vectors can be combined in the target order to obtain the first vector matrix. After obtaining the first vector matrix, it can be input into the normalization module. The normalization module then determines the first confidence level of placing at least one item into each mode of transport based on the fourth feature vectors of the first variables representing whether the item is placed into each mode of transport. The activation function used in the activation layer can be ReLU (Linear Rectification function) or other activation functions; no restriction is imposed here.

[0084] Since for each first variable, the machine learning model combines the feature information of the first variable, the feature information of other first variables, the feature information of the second variable, and the feature information of the constraints to obtain the fourth feature vector corresponding to the first variable, it can obtain richer feature information of the first variable. Therefore, for at least one item, based on the fourth feature vector of each first variable representing whether the item is placed in each means of transport, the first confidence level of the item being placed in each means of transport can be accurately determined.

[0085] In an example embodiment, the feature extraction module may include a feature extraction layer and a matrix transformation layer connected in sequence. Correspondingly, each first feature vector, each second feature vector, and each third feature vector are input into the feature extraction module. The feature extraction module then fuses the first feature vector corresponding to each first variable with other feature vectors in the input feature extraction module's feature vectors, excluding the first feature vector, to obtain a fourth feature vector corresponding to each first variable. This may include: inputting each first feature vector, each second feature vector, and each third feature vector into the feature extraction layer, so that the feature extraction layer fuses the first feature vector corresponding to each first variable with other feature vectors in the input feature extraction module's feature vectors, excluding the first feature vector. Other feature vectors are fused to obtain the fourth feature vector corresponding to each first variable. The second feature vector corresponding to each second variable is then fused with the other feature vectors in the feature vectors input to the feature extraction module, excluding the second feature vector, to obtain the fifth feature vector corresponding to each second variable. The fourth feature vector corresponding to each first variable and the fifth feature vector corresponding to each second variable are then combined in an initial order to obtain the second vector matrix. A mapping table storing the position information of each fourth and fifth feature vector in the second vector matrix is ​​obtained. The second vector matrix and the mapping table are then input into the matrix transformation layer to obtain the fourth feature vector corresponding to each first variable from the second vector matrix based on the position information.

[0086] The initial order refers to the arrangement of the first eigenvectors corresponding to each first variable and the second eigenvectors corresponding to each second variable when the solver outputs them.

[0087] Among them, reference Figure 3 The machine learning model may include a feature extraction layer. This feature extraction layer may include, in sequence, an input sub-layer 31, a graph convolutional neural network sub-layer 32, a fully connected sub-layer 33, and an activation sub-layer 34. Additionally, the machine learning model also includes a matrix transformation layer 35 connected to the activation sub-layer 34, and a normalization module 36 connected to the matrix transformation layer 35.

[0088] In the example embodiment, the first feature vector, the second feature vector, and the third feature vector output by the solver 30 can be input into the input sub-layer 31. The input sub-layer 31 integrates the first feature vector, the second feature vector, and the third feature vector, and maps each feature vector to the same dimension to obtain the feature vectors of the same dimension corresponding to each first variable, each second variable, and each constraint. Next, the feature vectors of the same dimension corresponding to each first variable, each second variable, and each constraint output from input sublayer 31 can be input into graph convolutional neural network sublayer 31. Through graph convolutional neural network sublayer 31, based on the feature vectors of the same dimension corresponding to each first variable, each second variable, and each constraint, features of the relationships between variables and between variables and constraints are captured. These captured features are then fused into the feature vectors corresponding to the first variables and the second variables. Furthermore, through fully connected sublayer 33 and activation sublayer 34, non-linear features in the features of the first variables are extracted to obtain the fourth feature vectors of each first variable. Similarly, through fully connected sublayer 33 and activation sublayer 34, non-linear features in the features of the second variables are extracted to obtain the fifth feature vectors of each second variable. Finally, the fourth and fifth feature vectors can be combined in an initial order to obtain the second vector matrix.

[0089] Understandably, when the solver outputs the first eigenvectors corresponding to each first variable and the second eigenvectors corresponding to each second variable, the transportation scheduling device can obtain a mapping table from the solver that stores the order of the first and second eigenvectors. The order of the first and second eigenvectors in this mapping table indicates the position information of the fourth and fifth eigenvectors corresponding to each first variable in the second vector matrix. Then, after inputting the second vector matrix and the mapping table into the matrix transformation layer 35, the matrix transformation layer 35 can obtain the fourth eigenvectors corresponding to each first variable from the second vector matrix based on the position information of the fourth and fifth eigenvectors stored in the mapping table. Furthermore, it can combine the fourth eigenvectors in the target order to obtain the first vector matrix.

[0090] For example, suppose for a certain first variable X 2,5 According to the target order, the first variable X 2,5 The corresponding fourth eigenvector is the element in the 2nd row and 5th column of the first vector matrix. According to the position information in the mapping table, the first variable X... 2,5 The corresponding fourth eigenvector is the first element in the second vector matrix. Therefore, the first variable X can be obtained from the second vector matrix based on its position information. 2,5The corresponding fourth feature vector, and the first variable X 2,5 The corresponding fourth eigenvector is placed in the 2nd row and 5th column of the first vector matrix.

[0091] After obtaining the first vector matrix, the first vector matrix can be input into the normalization module 36 so that, for at least one item, the first confidence level of the item being placed in each means of transport can be determined by the normalization module 36 based on the fourth feature vector of each first variable representing whether the item is placed in each means of transport.

[0092] Therefore, when the fourth feature vectors corresponding to each first variable obtained by the feature extraction layer in the machine learning model are arranged in any order, the matrix transformation layer can be used to obtain the first vector matrix in the target order, which lays the foundation for the normalization module to accurately determine the first confidence level of at least one item to be placed in each means of transport based on the first vector matrix.

[0093] Step 204: Based on the first confidence level of placing at least one item into each means of transport, determine the first target means of transport corresponding to at least one target item among the at least one items.

[0094] In the example embodiment, a confidence threshold can be preset, and for any item among at least one item, if there is a first confidence level greater than the confidence threshold among the first confidence levels of the item being placed in each means of transport, the means of transport corresponding to the first confidence level greater than the confidence threshold is determined as the first target means of transport corresponding to the item, and the item is referred to as the target item.

[0095] For example, suppose the confidence threshold is 0.9 and the number of transportation vehicles is 4. For the third item, the first confidence level for placing it in the first transportation vehicle is 0.01, the first confidence level for placing it in the second transportation vehicle is 0.02, the first confidence level for placing it in the third transportation vehicle is 0.95, and the first confidence level for placing it in the fourth transportation vehicle is 0.02. Since 0.95 is greater than 0.9, the third item can be called the target item, and the third transportation vehicle can be identified as the first target transportation vehicle corresponding to the third item.

[0096] Step 205: Input the attribute information of multiple items, the attribute information of multiple transportation vehicles, the objective function, constraints, and the first target transportation vehicle corresponding to at least one target item into the solver, so as to determine the second target transportation vehicle corresponding to the other items among the multiple items except for at least one target item.

[0097] Step 206: Based on the first target transport vehicle corresponding to at least one target item and the second target transport vehicles corresponding to other items, determine the correspondence between multiple items and multiple transport vehicles.

[0098] In an example embodiment, the correspondence between some items and some transportation vehicles can be determined based on the first confidence level of at least one item placed in each transportation vehicle as determined by the machine learning model. Then, the correspondence between other items and other transportation vehicles can be determined by the solver, thereby obtaining the correspondence between multiple items and multiple transportation vehicles.

[0099] Specifically, for each target item, based on the first target transportation vehicle corresponding to that target item, the values ​​of multiple first variables corresponding to that target item can be determined. Then, the subproblems with these determined first variable values ​​are treated as entirely new mixed-integer optimization problems and input into the solver for solving. Since a large number of first variable values ​​are already determined, the search space of the subproblems is reduced, decreasing the problem size. Therefore, it becomes easier for the solver to solve the subproblems, leading to a faster and more feasible solution.

[0100] For example, suppose the number of vehicles is 4. For the 3rd item, the first variable corresponding to this item includes X. 3,1 X 3,2 X 3,3 X 3,4 , representing whether to place the third item in the first, second, third, and fourth transport vehicles, respectively. Assuming the third item corresponds to the third transport vehicle, then X can be... 3,3 The value of X is set to 1. 3,1 X 3,2 X 3,4 The value of is determined to be 0, thus the subproblem with the values ​​of these first variables already determined is treated as a brand new mixed integer optimization problem and input into the solver for solving.

[0101] Step 207: If the objective function of the correspondence satisfies the set threshold, the correspondence is determined as the target transportation relationship between multiple items and multiple means of transport.

[0102] In summary, the transportation scheduling method provided in this disclosure obtains attribute information of multiple items, attribute information of multiple transportation vehicles, an objective function, and constraints. The objective function represents the resource imbalance among the multiple transportation vehicles. The attribute information of the multiple items, the attribute information of the multiple transportation vehicles, the objective function, and the constraints are input into a solver to obtain multiple first variables X. i,b The corresponding first feature vector, multiple second variables Y b,rThe corresponding second feature vector and the third feature vector corresponding to the constraint, based on each first feature vector, each second feature vector and the third feature vector, use a machine learning model to determine the first confidence level of at least one item being placed in each means of transportation. This allows the machine learning model to make predictions based on the first feature vectors corresponding to multiple first variables, the second feature vectors corresponding to multiple second variables and the third feature vector corresponding to the constraint determined by the solver, avoiding lazy random guessing for prediction. This truly allows the model to learn the knowledge inside the problem, improves the model's predictive ability, and thus accurately determines the first confidence level of at least one item being placed in each means of transportation. By determining the first target transport vehicle corresponding to at least one target item among at least one item based on the first confidence level of placing at least one item into each transport vehicle, the attribute information of multiple items, the attribute information of multiple transport vehicles, the objective function, constraints, and the first target transport vehicle corresponding to at least one target item are input into the solver. The solver then determines the second target transport vehicles corresponding to the other items among the multiple items besides the at least one target item. Based on the first target transport vehicle corresponding to at least one target item and the second target transport vehicles corresponding to the other items, the correspondence between multiple items and multiple transport vehicles is determined. This reduces the problem size input into the solver, thereby quickly determining the correspondence between multiple items and multiple transport vehicles.

[0103] In one possible implementation, after determining the correspondence between multiple items and multiple means of transport, the objective function for this correspondence may not meet a set threshold. The following section will combine... Figure 4 In light of the above, the transportation scheduling method provided in this disclosure will be further explained.

[0104] Figure 4 This is a flowchart illustrating a transportation scheduling method according to a third embodiment of this disclosure. Figure 4 As shown, the transportation scheduling method may include the following steps:

[0105] Step 401: Obtain the attribute information of multiple items, the attribute information of multiple transportation tools, the objective function, and the constraints; wherein, the objective function represents the imbalance of resources among multiple transportation tools.

[0106] Step 402: Input the attribute information of multiple items, the attribute information of multiple means of transportation, the objective function, and the constraints into the solver, so as to obtain multiple first variables X through the solver. i,b The corresponding first feature vector, multiple second variables Y b,r The corresponding second eigenvector and the constraint corresponding third eigenvector.

[0107] Among them, the first variable X i,bThe second variable Y represents whether to put the i-th item into the b-th transport vehicle. b,r Let i represent the imbalance value of the b-th means of transport on the r-th resource, where i is an integer between 1 and I, b is an integer between 1 and B, r is an integer between 1 and R, I is the quantity of goods, B is the quantity of means of transport, and R is the quantity of different types of resources. I, B, and R are integers greater than or equal to 1.

[0108] Step 403: Based on each first feature vector, each second feature vector, and the third feature vector, a machine learning model is used to determine the first confidence level of placing at least one item into each means of transport.

[0109] Step 404: Based on the first confidence level of placing at least one item into each means of transport, determine the first target means of transport corresponding to at least one target item among the at least one items.

[0110] Step 405: Input the attribute information of multiple items, the attribute information of multiple transportation vehicles, the objective function, constraints, and the first target transportation vehicle corresponding to at least one target item into the solver, so as to determine the second target transportation vehicle corresponding to the other items among the multiple items except for at least one target item.

[0111] Step 406: Based on the first target transport vehicle corresponding to at least one target item and the second target transport vehicles corresponding to other items, determine the correspondence between multiple items and multiple transport vehicles.

[0112] The specific implementation process and principle of steps 401-406 can be referred to the description of the above embodiments, and will not be repeated here.

[0113] Step 407: If the objective function value of the correspondence does not meet the set threshold, perform at least one iteration until the objective function value of the correspondence meets the set threshold.

[0114] In the example embodiment, after determining the correspondence between multiple items and multiple means of transport in step 406, the objective function value of the correspondence can be determined. If the objective function value of the correspondence does not meet the set threshold, at least one round of iteration is performed. In each round of iteration, the correspondence between multiple items and multiple means of transport, as well as the objective function value of the correspondence, are determined until the objective function value of the correspondence meets the set threshold.

[0115] In at least one iteration, based on the correspondence between multiple items and multiple means of transport determined in the previous iteration, the first feature vector, the second feature vector, and the third feature vector obtained in the previous iteration are updated. Based on the updated first feature vector, the second feature vector, and the third feature vector, a machine learning model is used to determine the second confidence level of at least one item placed in each means of transport under the condition of satisfying the constraints and minimizing the objective function. Based on the second confidence level of at least one item placed in each means of transport, the correspondence between multiple items and multiple means of transport is determined.

[0116] The second confidence level represents the likelihood that an item will be placed on a means of transport.

[0117] In the example embodiment, the process of updating the first feature vector, second feature vector, and third feature vector obtained in the previous iteration based on the correspondence between multiple items and multiple means of transport determined in the historical iteration can be as follows: In the previous iteration, some features are extracted from the attribute information of multiple items, the attribute information of multiple means of transport, the objective function, constraints, and the first target means of transport corresponding to at least one target item by the solver, and combined with the correspondence between multiple items and multiple means of transport determined in the historical iteration, the updated first feature vector, second feature vector, and third feature vector are obtained.

[0118] In the example embodiment, the feature information of the corresponding variable represented by each first feature vector, or each second feature vector, or each first feature vector and each second feature vector may include at least one of the following: the value of the corresponding variable in the historical iteration, the number of times the value of the corresponding variable has changed in the historical iteration, the value of the corresponding variable with the most values ​​in the historical iteration, the maximum value of the corresponding variable in the historical iteration, the minimum value of the corresponding variable in the historical iteration, and whether the corresponding variable is a stable variable.

[0119] The above information can effectively represent the value characteristics of a variable, which is beneficial for machine learning models to capture deeper features and use them for confidence prediction. Therefore, by including the above information in the feature information of the corresponding variables represented by each first feature variable and / or each second feature variable, the feature information of the variables can be extracted in a targeted manner and enriched. Consequently, the machine learning model can make confidence predictions based on richer and deeper feature information, thereby improving the accuracy of the model's predictions.

[0120] Among them, stable variables are those whose values ​​remain unchanged or whose number of changes is less than a preset threshold in historical iterations.

[0121] It should be noted that in each iteration, based on the updated first feature vector, second feature vector, and third feature vector, a machine learning model is used to determine the second confidence level of at least one item placed in each means of transport while satisfying the constraints and minimizing the objective function. The process of determining the correspondence between multiple items and multiple means of transport based on the second confidence level of at least one item placed in each means of transport is similar to the process in steps 403-406, and will not be repeated here.

[0122] Step 408: If the objective function of the correspondence satisfies the set threshold, the correspondence is determined as the target transportation relationship between multiple items and multiple means of transport.

[0123] In summary, the transportation scheduling method provided in this disclosure, after determining the correspondence between multiple items and multiple transportation vehicles, performs at least one iteration when the objective function value of the correspondence does not meet a set threshold, until the objective function value of the correspondence meets the set threshold. When the objective function value of the correspondence meets the set threshold, the correspondence is determined as the target transportation relationship between multiple items and multiple transportation vehicles. This achieves the optimal solution for the target transportation relationship between multiple items and multiple transportation vehicles when the objective function value meets the set threshold through multiple iterations.

[0124] In an example embodiment, a method for training a machine learning model for transportation scheduling is also provided. Figure 5 This is a flowchart illustrating a training method for a machine learning model for transportation scheduling according to a fourth embodiment of the present disclosure.

[0125] It should be noted that the training method for the machine learning model for transportation scheduling provided in this embodiment is executed by a training device for the machine learning model for transportation scheduling, hereinafter referred to as the training device. This training device can be implemented by software and / or hardware, and can be configured in an electronic device, which may include, but is not limited to, terminal devices, servers, etc. This embodiment does not specifically limit the electronic device.

[0126] like Figure 5 As shown, the training method for a machine learning model used for transportation scheduling may include the following steps:

[0127] Step 501: Obtain multiple training samples, wherein at least one training sample includes attribute information of multiple sample items, attribute information of multiple sample transportation vehicles, objective function, and constraints; the objective function represents the resource imbalance of multiple sample transportation vehicles.

[0128] The sample item can be any item that needs to be transported, such as a box containing food or a bag containing clothing, etc. This disclosure does not limit this.

[0129] The attribute information of the sample items may include any attribute information related to the sample items, such as the quantity information of the sample items, the size of the space occupied by the sample items, and the weight of the sample items, etc. This disclosure does not limit this.

[0130] The sample transport vehicle can be any tool capable of transporting sample items, such as a ship or vehicle, and this disclosure does not limit this.

[0131] The attribute information of the sample transport vehicle may include any attribute information related to the sample transport vehicle, such as the quantity of sample transport vehicles, the space and weight capacity of the sample transport vehicle, etc. This disclosure does not limit this.

[0132] It should be noted that the number of sample items, the number of sample transportation vehicles, and other attribute information in the training samples may be the same as or different from the number of items, the number of transportation vehicles, and other attribute information in the application process after the machine learning model is trained, i.e., the transportation scheduling method of the above embodiments. This disclosure does not impose any restrictions on this.

[0133] The objective function represents the functional relationship between the desired outcome and relevant factors when placing multiple sample items onto multiple sample transport vehicles. In this embodiment, the objective function represents the resource imbalance among the multiple sample transport vehicles and can be used to indicate the optimal solution for the target transport relationship between the multiple sample items and the multiple sample transport vehicles. For example, the solution that minimizes the objective function value is the optimal solution. The target transport relationship between the multiple sample items and the multiple sample transport vehicles indicates which sample transport vehicle each sample item should be loaded onto for transport.

[0134] Constraints are the restrictions that must be followed when placing multiple sample items into multiple sample transport vehicles. For example, each sample item must be placed on only one sample transport vehicle, and the total amount of resources carried by each sample transport vehicle cannot exceed the total amount of resources that the sample transport vehicle can carry.

[0135] Step 502: For at least one training sample, based on the attribute information of multiple sample items, the attribute information of multiple sample transportation vehicles, the objective function, and the constraints, a machine learning model is used to determine the third confidence level of at least one sample item placed in each sample transportation vehicle while satisfying the constraints and minimizing the objective function.

[0136] The third confidence level represents the likelihood that a sample item will be placed in a sample transport vehicle.

[0137] The specific implementation process and principle of step 502 can be found in the description of the embodiment of the transportation scheduling method, and will not be repeated here.

[0138] Step 503: Train the machine learning model based on the third confidence level of at least one sample item in at least one training sample being placed in each sample transport vehicle.

[0139] In the example embodiment, after determining the third confidence level of at least one sample item in each training sample being placed in each sample transport vehicle, the sample correspondence between multiple sample items and multiple sample transport vehicles can be determined based on the third confidence level of at least one sample item being placed in each sample transport vehicle, and the objective function value of the sample correspondence can be determined. The machine learning model is trained by minimizing the objective function value.

[0140] In an example embodiment, each training sample can be labeled with the sample correspondence between multiple sample items and multiple sample transportation vehicles. Based on the third confidence level of at least one sample item in each training sample being placed in each sample transportation vehicle and the sample correspondence, a loss value is determined, and the machine learning model is trained based on the loss value.

[0141] It should be noted that the trained machine learning model can be used to execute the aforementioned transportation scheduling method. The process of executing the aforementioned transportation scheduling method using the trained machine learning model can be found in the description of the embodiments of the transportation scheduling method described above, and will not be repeated here.

[0142] In summary, the training method for a machine learning model for transportation scheduling provided in this disclosure acquires multiple training samples. At least one training sample includes attribute information of multiple sample items, attribute information of multiple sample transportation vehicles, an objective function, and constraints. The objective function represents the resource imbalance among the multiple sample transportation vehicles. For at least one training sample, based on the attribute information of the multiple sample items, the attribute information of the multiple sample transportation vehicles, the objective function, and the constraints, a machine learning model is used to determine the third confidence level of placing at least one sample item into each sample transportation vehicle while satisfying the constraints and minimizing the objective function. Based on the third confidence level of placing at least one sample item into each sample transportation vehicle in at least one training sample, the machine learning model is trained, thus achieving the training of a machine learning model for transportation scheduling. By utilizing the trained machine learning model, the target transportation relationship between multiple items and multiple transportation vehicles can be determined, improving the accuracy of the determined target transportation relationship, ensuring that the resource imbalance among multiple transportation vehicles is minimized, thereby reducing transportation costs during the transportation process.

[0143] The following is combined with Figure 6 This further explains the training method for the machine learning model for transportation scheduling provided in this disclosure. Figure 6 This is a flowchart illustrating a method for training a machine learning model for transportation scheduling according to a fifth embodiment of this disclosure.

[0144] like Figure 6 As shown, the training method for a machine learning model used for transportation scheduling may include the following steps:

[0145] Step 601: Obtain multiple training samples, wherein at least one training sample includes attribute information of multiple sample items, attribute information of multiple sample transportation vehicles, objective function, and constraints; the objective function represents the resource imbalance of multiple sample transportation vehicles.

[0146] The specific implementation process and principle of step 601 can be referred to the description of the above embodiments, and will not be repeated here.

[0147] Step 602: For at least one training sample, input the attribute information of multiple sample items, the attribute information of multiple sample transportation vehicles, the objective function, and the constraints into the solver, and perform multiple iterations through the solver to obtain multiple third variables P output by the solver in at least one iteration. i’,b’ The corresponding sixth eigenvector and multiple fourth variables Q b’,r’ The corresponding seventh eigenvector and the constraint corresponding eighth eigenvector.

[0148] Among them, the third variable P i’,b’ Indicates whether to put the i'th sample item into the b'th sample transport vehicle, the fourth variable Q b’,r’ Let i' represent the imbalance value of the b'th sample transport vehicle on the r'th type of resource, where i' is an integer between 1 and I', b' is an integer between 1 and B', r' is an integer between 1 and R', I' is the quantity of sample items, B' is the quantity of sample transport vehicles, and R' is the number of types of resources. I', B', and R' are integers greater than or equal to 1.

[0149] The sixth eigenvector represents the feature information of the third variable; the seventh eigenvector represents the feature information of the fourth variable; and the eighth eigenvector represents the feature information of the constraints. This feature information may include, for example, the values ​​on the left side of the equals sign in the constraints, the parameters in the constraints, and the values ​​on the right side of the equals sign.

[0150] In the example embodiment, for each training sample, the attribute information of multiple sample items, the attribute information of multiple sample transportation vehicles, the objective function, and the constraints can be input into the solver. The solver then performs multiple iterations to solve the correspondence between the multiple sample items and the multiple sample transportation vehicles until the solver finishes solving or a preset time is reached. During the solver's solution process, in each iteration, features can be extracted from the attribute information of the multiple sample items, the attribute information of the multiple sample transportation vehicles, the objective function, and the constraints, thereby obtaining multiple third variables P. i’,b’ The corresponding sixth eigenvector and multiple fourth variables Q b’,r’ The corresponding seventh feature vector and the eighth feature vector corresponding to the constraint. Furthermore, to enrich the feature information of the third variable, fourth variable, and constraint, multiple third variables P can be obtained by combining this correspondence with the extracted features after each iteration of the solver, when a correspondence is obtained between multiple sample items and multiple sample transportation vehicles. i’,b’ The corresponding sixth eigenvector and multiple fourth variables Q b’,r’ The corresponding seventh eigenvector and the constraint corresponding eighth eigenvector.

[0151] In the example embodiment, when the solver obtains the first few correspondences, although these correspondences satisfy the constraints, the corresponding objective function values ​​are far from the set threshold. Therefore, in this embodiment, after the solver performs multiple iterations, the third variable P output by the solver in at least one iteration can be obtained. i’,b’ The corresponding sixth eigenvector and multiple fourth variables Q b’,r’ The corresponding seventh eigenvector and the constraint's corresponding eighth eigenvector, for example, after the solver obtains the fourth correspondence, the third variable P output by the solver in each subsequent iteration is then obtained. i’,b’ The corresponding sixth eigenvector and multiple fourth variables Q b’,r’ The corresponding seventh eigenvector and the constraint corresponding eighth eigenvector.

[0152] The characteristic information of the corresponding variable represented by each sixth feature vector, or each seventh feature vector, or the sixth feature vector and the seventh feature vector may include at least one of the following: the value of the corresponding variable in the historical iteration, the number of times the value of the corresponding variable has changed in the historical iteration, the value of the corresponding variable with the most occurrences in the historical iteration, the maximum value of the corresponding variable in the historical iteration, the minimum value of the corresponding variable in the historical iteration, and whether the corresponding variable is a stable variable.

[0153] The above information can effectively represent the value characteristics of a variable, which is beneficial for machine learning models to capture deeper features and use them for confidence prediction. Therefore, by including the above information in the feature information of the corresponding variables represented by each sixth and / or seventh feature variable, the feature information of the variables can be extracted in a targeted manner and enriched. Consequently, the machine learning model can be trained and learned based on richer and deeper feature information, thereby improving the accuracy of model prediction.

[0154] In addition, while the solver outputs the sixth eigenvector corresponding to the third variable Pi',b', the seventh eigenvector corresponding to multiple fourth variables Q b',r', and the eighth eigenvector corresponding to the constraints, the training device can also obtain a mapping table for matrix transformation during model training.

[0155] Step 603: Based on the sixth feature vector, the seventh feature vector, and the eighth feature vector corresponding to at least one round of iteration of at least one training sample, a machine learning model is used to determine the third confidence level of at least one sample item being placed in each sample transportation vehicle.

[0156] Step 604: Train the machine learning model based on the third confidence of at least one sample item placed into each sample transport vehicle corresponding to at least one round of iteration of at least one training sample.

[0157] In an example embodiment, the sixth feature vector, the seventh feature vector, and the eighth feature vector (or may also include a mapping table) corresponding to one iteration of a training sample can be used as a sub-training sample. Then, a machine learning model can be used to determine the third confidence level of at least one sample item in the sub-training sample being placed in each sample transportation vehicle, and the machine learning model can be trained based on the third confidence level of at least one sample item in each sub-training sample being placed in each sample transportation vehicle.

[0158] Because the solver can optimize the sample correspondence between multiple sample items and multiple sample transportation vehicles determined in previous rounds through multiple iterations, it can ultimately determine the optimal solution for the sample correspondence between multiple sample items and multiple sample transportation vehicles. Therefore, the machine learning model can be trained based on the sixth feature vector corresponding to multiple third variables, the seventh feature vector corresponding to multiple fourth variables, and the eighth feature vector corresponding to the constraints determined by the solver in multiple iterations. This allows the machine learning model to be trained on more accurate training data, thereby improving the predictive ability of the machine learning model.

[0159] In an exemplary embodiment, for at least one training sample, the attribute information of multiple sample items, the attribute information of multiple sample transportation vehicles, the objective function, and the constraints are input into the solver. After multiple iterations by the solver, the values ​​of each third variable output by the solver in at least one iteration can be obtained. The values ​​of each third variable can indicate the sample correspondence between multiple sample items and multiple sample transportation vehicles in one iteration. In this embodiment, the values ​​of each third variable output by the solver in each iteration of each training sample can be used as labels to train the machine learning model.

[0160] That is, in step 604, training the machine learning model based on the third confidence of at least one sample item placed in each sample transportation vehicle corresponding to at least one round of iteration of at least one training sample can include: determining the cross-entropy loss based on the third confidence of at least one sample item placed in each sample transportation vehicle corresponding to at least one round of iteration of at least one training sample and the value of each third variable; and training the machine learning model based on the cross-entropy loss.

[0161] Therefore, by determining the third confidence level of at least one sample item placed in each sample transportation vehicle based on at least one round of iteration corresponding to at least one training sample and the corresponding values ​​of each third variable, the cross-entropy loss can be determined. Based on the cross-entropy loss, the machine learning model can be trained, thereby achieving supervised training of the machine learning model and improving its predictive ability.

[0162] It should be noted that the machine learning model in this embodiment is a multi-class classification model. The number of predicted categories is strongly correlated with the number of sample transportation vehicles. However, the number of predicted categories can be adaptively adjusted by replacing the mapping table according to the actual situation. That is, the number of categories in multi-class classification is adaptive. Therefore, the machine learning model in this embodiment has good generalization ability when applied to transportation scheduling to determine the correspondence between multiple items and multiple transportation vehicles. For example, when the number of transportation vehicles is 10, the machine learning model is a 10-class classification model. After training the 10-class machine learning model, most of the parameters of this model can be reused to handle the transportation scheduling problem of 20 transportation vehicles. Only the variable mapping table used by the model needs to be replaced to convert the matrix transformation layer of the model to a 20-class classification model. Based on this, the parameters of other parts of the machine learning model in this embodiment, such as graph convolutional neural network layers or graph convolutional neural network sublayers, fully connected layers or fully connected sublayers, can be reused, resulting in good generalization. In addition, based on the 10-class classification model, when facing the need for 20-class classification in the transportation scheduling process, the graph convolutional neural network layer or graph convolutional neural network sub-layer, fully connected layer or fully connected sub-layer, etc. can be continuously trained to improve the model's representation ability and alleviate the overfitting problem.

[0163] The following is combined with Figure 7 This document describes the transportation dispatching device provided in this disclosure.

[0164] Figure 7 This is a schematic diagram of the structure of a transportation dispatching device according to the sixth embodiment of this disclosure.

[0165] like Figure 7 As shown, the transportation scheduling device 700 provided in this disclosure includes: a first acquisition module 701, a first determination module 702, and a second determination module 703.

[0166] The first acquisition module 701 is used to acquire attribute information of multiple items, attribute information of multiple transportation tools, objective function, and constraints; wherein the objective function represents the imbalance of resources among multiple transportation tools.

[0167] The first determining module 702 is used to determine, based on the attribute information of multiple items, the attribute information of multiple means of transport, the objective function and constraints, the first confidence level of at least one item placed in each means of transport under the condition of satisfying the constraints and minimizing the objective function, and to determine the correspondence between multiple items and multiple means of transport based on the first confidence level of at least one item placed in each means of transport.

[0168] The second determining module 703 is used to determine the correspondence as a target transportation relationship between multiple items and multiple means of transport when the objective function of the correspondence meets a set threshold.

[0169] It should be noted that the transportation scheduling device 700 provided in this embodiment can execute the transportation scheduling method of the aforementioned embodiment. The transportation scheduling device 700 can be implemented by software and / or hardware, and can be configured in an electronic device, which may include, but is not limited to, terminal devices, servers, etc. This embodiment does not specifically limit the electronic device.

[0170] In an example embodiment, the first determining module 702 may include:

[0171] The first acquisition submodule is used to input the attribute information of multiple items, the attribute information of multiple means of transportation, the objective function, and the constraints into the solver, so as to obtain multiple first variables X through the solver. i,b The corresponding first feature vector, multiple second variables Y b,r The corresponding second eigenvector and the constraint-related third eigenvector; where the first variable X i,b The second variable Y represents whether to put the i-th item into the b-th transport vehicle. b,r Let i represent the imbalance value of the b-th means of transport on the r-th resource, where i is an integer between 1 and I, b is an integer between 1 and B, r is an integer between 1 and R, I is the quantity of goods, B is the quantity of means of transport, and R is the quantity of different types of resources, and I, B, and R are integers greater than or equal to 1.

[0172] The first determination submodule is used to determine the first confidence level of placing at least one item into each means of transport based on each first feature vector, each second feature vector, and the third feature vector using a machine learning model.

[0173] In an example embodiment, the machine learning model includes a feature extraction module and a normalization module connected in sequence; the first determining submodule includes:

[0174] The processing unit is used to input each first feature vector, each second feature vector, and the third feature vector into the feature extraction module, so that the feature extraction module can fuse the first feature vector corresponding to each first variable with the other feature vectors in the feature vectors input to the feature extraction module except for the first feature vector to obtain the fourth feature vector corresponding to each first variable, and combine the fourth feature vectors corresponding to each first variable in the target order to obtain the first vector matrix.

[0175] A determining unit is used to input the first vector matrix into the normalization module to determine the first confidence level of placing at least one item into each means of transport by the normalization module.

[0176] In an example embodiment, the feature extraction module includes a feature extraction layer and a matrix transformation layer connected in sequence; the processing unit includes:

[0177] The processing subunit is used to input each first feature vector, each second feature vector, and the third feature vector into the feature extraction layer. Through the feature extraction layer, the first feature vector corresponding to each first variable is fused with other feature vectors in the feature vectors input to the feature extraction module, except for the first feature vector, to obtain the fourth feature vector corresponding to each first variable. The second feature vector corresponding to each second variable is fused with other feature vectors in the feature vectors input to the feature extraction module, except for the second feature vector, to obtain the fifth feature vector corresponding to each second variable. The fourth feature vector corresponding to each first variable and the fifth feature vector corresponding to each second variable are combined in an initial order to obtain the second vector matrix.

[0178] The first acquisition subunit is used to acquire a mapping table that stores the position information of each fourth feature vector and each fifth feature vector in the second vector matrix;

[0179] The second acquisition subunit is used to input the second vector matrix and the mapping relationship table into the matrix transformation layer so as to obtain the fourth feature vector corresponding to each first variable from the second vector matrix based on the position information.

[0180] In an example embodiment, the first determining module 702 includes:

[0181] The second determination submodule is used to determine the first target transport vehicle corresponding to at least one target item among at least one item, based on a first confidence level of at least one item being placed in each transport vehicle;

[0182] The third determination submodule is used to input the attribute information of multiple items, the attribute information of multiple transportation vehicles, the objective function, constraints, and the first target transportation vehicle corresponding to at least one target item into the solver, so as to determine the second target transportation vehicle corresponding to the other items among the multiple items except for at least one target item through the solver.

[0183] The fourth determination submodule is used to determine the correspondence between multiple items and multiple transportation vehicles based on the first target transportation vehicle corresponding to at least one target item and the second target transportation vehicles corresponding to other items.

[0184] In an example embodiment, the transportation scheduling device further includes:

[0185] The processing module is used to perform at least one iteration when the objective function value of the correspondence does not meet the set threshold, until the objective function value of the correspondence meets the set threshold.

[0186] In at least one iteration, based on the correspondence between multiple items and multiple means of transport determined in the previous iteration, the first feature vector, the second feature vector, and the third feature vector obtained in the previous iteration are updated. Based on the updated first feature vector, the second feature vector, and the third feature vector, a machine learning model is used to determine the second confidence level of at least one item placed in each means of transport under the condition of satisfying the constraints and minimizing the objective function. Based on the second confidence level of at least one item placed in each means of transport, the correspondence between multiple items and multiple means of transport is determined.

[0187] In the example embodiment, the feature information of the corresponding variable represented by each first feature vector and / or each second feature vector includes at least one of the following: the value of the corresponding variable in the historical iteration, the number of times the value of the corresponding variable has changed in the historical iteration, the value of the corresponding variable with the most occurrences in the historical iteration, the maximum value of the corresponding variable in the historical iteration, the minimum value of the corresponding variable in the historical iteration, and whether the corresponding variable is a stable variable.

[0188] It should be noted that the foregoing description of the embodiments of the transportation scheduling method also applies to the transportation scheduling device provided in this disclosure, and will not be repeated here.

[0189] The transportation scheduling device provided in this embodiment acquires attribute information of multiple items, attribute information of multiple transportation vehicles, an objective function, and constraints. The objective function represents the resource imbalance among the multiple transportation vehicles. Based on the attribute information of the multiple items, the attribute information of the multiple transportation vehicles, the objective function, and the constraints, a machine learning model is used to determine a first confidence level for placing at least one item in each transportation vehicle while satisfying the constraints and minimizing the objective function. Based on the first confidence level of placing at least one item in each transportation vehicle, a correspondence between multiple items and multiple transportation vehicles is determined. If the objective function value of the correspondence meets a set threshold, the correspondence is determined as the target transportation relationship between the multiple items and multiple transportation vehicles. This ensures that the resource imbalance among the multiple transportation vehicles is minimized, thereby reducing transportation costs during the transportation process.

[0190] In an example embodiment, a training apparatus for a machine learning model for transportation scheduling is also provided. The following is in conjunction with... Figure 8 The training apparatus for the machine learning model for transportation scheduling provided in this disclosure will be described.

[0191] Figure 8 This is a schematic diagram of the structure of a training apparatus for a machine learning model for transportation scheduling according to the seventh embodiment of the present disclosure.

[0192] like Figure 8As shown, the training device 800 for a machine learning model for transportation scheduling provided in this disclosure includes: a second acquisition module 801, a third determination module 802, and a training module 803.

[0193] The second acquisition module 801 is used to acquire multiple training samples, wherein at least one training sample includes attribute information of multiple sample items, attribute information of multiple sample transportation vehicles, objective function, and constraints; the objective function represents the resource imbalance of multiple sample transportation vehicles.

[0194] The third determination module 802 is used to determine, for at least one training sample, the third confidence level of placing at least one sample item into each sample transportation vehicle under the condition of satisfying the constraints and minimizing the objective function, based on the attribute information of multiple sample items, the attribute information of multiple sample transportation vehicles, the objective function, and the constraints, using a machine learning model.

[0195] Training module 803 is used to train a machine learning model based on the third confidence of at least one sample item in at least one training sample being placed in each sample transport vehicle.

[0196] It should be noted that the training device 800 for the machine learning model for transportation scheduling provided in this embodiment, hereinafter referred to as the training device, can execute the training method for the machine learning model for transportation scheduling in the aforementioned embodiment. The training device can be implemented by software and / or hardware, and can be configured on an electronic device, which may include, but is not limited to, terminal devices, servers, etc. This embodiment does not specifically limit the electronic device.

[0197] In an example embodiment, the third determining module 802 includes:

[0198] The second acquisition submodule is used to, for at least one training sample, input the attribute information of multiple sample items, the attribute information of multiple sample transportation vehicles, the objective function, and the constraints into the solver, and perform multiple iterations through the solver to obtain multiple third variables P output by the solver in at least one iteration. i’,b’ The corresponding sixth eigenvector and multiple fourth variables Q b’,r’ The corresponding seventh eigenvector and the constraint-related eighth eigenvector; where the third variable P i’,b’ Indicates whether to put the i'th sample item into the b'th sample transport vehicle, the fourth variable Q b’,r’Let i' represent the imbalance value of the b'th sample means of transport on the r'th type of resource, where i' is an integer between 1 and I', b' is an integer between 1 and B', r' is an integer between 1 and R', I' is the quantity of sample items, B' is the quantity of sample means of transport, and R' is the number of types of resources, and I', B', and R' are integers greater than or equal to 1.

[0199] The fifth determination submodule is used to determine the third confidence level of at least one sample item being placed in each sample transport vehicle based on the sixth feature vector, the seventh feature vector, and the eighth feature vector corresponding to at least one round of iteration of at least one training sample, using a machine learning model.

[0200] In the example embodiment, the third determining module 802 further includes:

[0201] The third acquisition submodule is used to acquire the values ​​of each third variable output by the solver in at least one iteration.

[0202] Training module 803 includes:

[0203] The sixth determination submodule is used to determine the cross-entropy loss based on the third confidence of at least one sample item placed in each sample transportation vehicle corresponding to at least one round of iteration of at least one training sample and the value of each third variable.

[0204] The training submodule is used to train machine learning models based on cross-entropy loss.

[0205] It should be noted that the foregoing description of the embodiment of the training method for the machine learning model for transportation scheduling also applies to the training apparatus for the machine learning model for transportation scheduling provided in this disclosure, and will not be repeated here.

[0206] The training apparatus for a machine learning model for transportation scheduling provided in this embodiment acquires multiple training samples. At least one training sample includes attribute information of multiple sample items, attribute information of multiple sample transportation vehicles, an objective function, and constraints. The objective function represents the resource imbalance among the multiple sample transportation vehicles. For at least one training sample, based on the attribute information of the multiple sample items, the attribute information of the multiple sample transportation vehicles, the objective function, and the constraints, a machine learning model is used to determine the third confidence level of placing at least one sample item in each sample transportation vehicle while satisfying the constraints and minimizing the objective function. Based on the third confidence level of placing at least one sample item in each sample transportation vehicle in at least one training sample, the machine learning model is trained, thus realizing the training of a machine learning model for transportation scheduling. By utilizing the trained machine learning model, the target transportation relationship between multiple items and multiple transportation vehicles can be determined, improving the accuracy of the determined target transportation relationship, ensuring that the resource imbalance among multiple transportation vehicles is minimized, thereby reducing transportation costs during the transportation process.

[0207] Based on the above embodiments, this disclosure also provides an electronic device, including: 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 execute the transportation scheduling method of this disclosure, or to execute the training method of the machine learning model for transportation scheduling of this disclosure.

[0208] Based on the above embodiments, this disclosure also provides a non-transitory computer-readable storage medium storing computer instructions, the computer instructions being used to cause the computer to execute the transportation scheduling method disclosed in the embodiments of this disclosure, or to execute the training method for a machine learning model for transportation scheduling disclosed in the embodiments of this disclosure.

[0209] Based on the above embodiments, this disclosure also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the transportation scheduling method of this disclosure, or the steps of the training method of the machine learning model for transportation scheduling of this disclosure.

[0210] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0211] Figure 9A schematic block diagram of an example electronic device 900 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0212] like Figure 9 As shown, the electronic device 900 may include a computing unit 901, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 902 or a computer program loaded from a storage unit 908 into a random access memory (RAM) 903. The RAM 903 may also store various programs and data required for the operation of the device 900. The computing unit 901, ROM 902, and RAM 903 are interconnected via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0213] Multiple components in device 900 are connected to I / O interface 905, including: input unit 906, such as keyboard, mouse, etc.; output unit 907, such as various types of monitors, speakers, etc.; storage unit 908, such as disk, optical disk, etc.; and communication unit 909, such as network card, modem, wireless transceiver, etc. Communication unit 909 allows device 900 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0214] The computing unit 901 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 901 performs the various methods and processes described above, such as transportation scheduling methods or methods for training machine learning models for transportation scheduling. For example, in some embodiments, the transportation scheduling method or the method for training machine learning models for transportation scheduling can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 908. In some embodiments, part or all of the computer program can be loaded and / or installed on device 900 via ROM 902 and / or communication unit 909. When the computer program is loaded into RAM 903 and executed by the computing unit 901, one or more steps of the transportation scheduling method or the method for training machine learning models for transportation scheduling described above can be performed. Alternatively, in other embodiments, the computing unit 901 may be configured by any other suitable means (e.g., by means of firmware) to execute a transportation scheduling method or a training method for a machine learning model for transportation scheduling.

[0215] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0216] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0217] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0218] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0219] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), the Internet, and blockchain networks.

[0220] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service ecosystem, addressing the shortcomings of traditional physical hosts and VPS (Virtual Private Server, or simply "VPS") services, such as high management difficulty and weak business scalability. A server can be a cloud server, a server in a distributed system, or a server incorporating blockchain technology.

[0221] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0222] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A transportation scheduling method, wherein, The method includes: The system acquires attribute information of multiple items, attribute information of multiple transportation vehicles, an objective function, and constraints. The objective function represents the resource imbalance among the multiple transportation vehicles and indicates the optimal solution for the target transportation relationship between the multiple items and the multiple transportation vehicles. The target transportation relationship indicates each item and its corresponding transportation vehicle. The resources of the transportation vehicles include weight resources (the weight the vehicle can carry) or spatial resources (the space the vehicle has). The constraints include restrictions on the placement of the multiple items into the multiple transportation vehicles. Based on the attribute information of multiple items, the attribute information of multiple means of transport, the objective function, and the constraints, a machine learning model is used to determine the first confidence level of placing at least one item in each of the means of transport while satisfying the constraints and minimizing the objective function, so as to determine the correspondence between the multiple items and the multiple means of transport based on the first confidence level of placing at least one item in each of the means of transport. If the objective function of the correspondence satisfies a set threshold, the correspondence is determined as the target transportation relationship between the multiple items and the multiple means of transport. The method of determining, based on the attribute information of multiple items, the attribute information of multiple transportation vehicles, the objective function, and the constraints, using a machine learning model to determine the first confidence level of placing at least one of the items into each of the transportation vehicles while satisfying the constraints and minimizing the objective function, includes: The attribute information of multiple items, the attribute information of multiple transportation vehicles, the objective function, and the constraint are input into the solver to obtain a first feature vector corresponding to multiple first variables Xi,b, a second feature vector corresponding to multiple second variables Yb,r, and a third feature vector corresponding to the constraint. Here, the first variable Xi,b indicates whether the i-th item is placed in the b-th transportation vehicle, the second variable Yb,r indicates the imbalance value of the b-th transportation vehicle on the r-th resource, i is an integer between 1 and 1, b is an integer between 1 and B, r is an integer between 1 and R, 1 is the quantity of the item, B is the quantity of the transportation vehicle, and R is the number of types of the resource. I, B, and R are integers greater than or equal to 1. Based on each of the first feature vectors, each of the second feature vectors and the third feature vector, the machine learning model is used to determine the first confidence level of at least one of the items being placed in each of the means of transport. The step of determining the correspondence between the plurality of items and the plurality of transport vehicles based on a first confidence level of at least one of the items being placed in each of the transport vehicles includes: Based on a first confidence level of placing at least one of the items into each of the transport vehicles, a first target transport vehicle corresponding to at least one target item among the at least one of the items is determined; The attribute information of multiple items, the attribute information of multiple transportation vehicles, the objective function, the constraints, and the first target transportation vehicle corresponding to the at least one target item are input into the solver so that the solver can determine the second target transportation vehicle corresponding to the other items among the multiple items besides the at least one target item. Based on the first target transport vehicle corresponding to the at least one target item and the second target transport vehicle corresponding to the other items, the correspondence between the plurality of items and the plurality of transport vehicles is determined.

2. The method according to claim 1, wherein, The machine learning model includes a feature extraction module and a normalization module connected in sequence; the step of determining a first confidence level for at least one of the items to be placed in each of the transport vehicles based on each of the first feature vectors, each of the second feature vectors, and the third feature vector includes: Each first feature vector, each second feature vector, and the third feature vector are input into the feature extraction module. The feature extraction module then merges the first feature vector corresponding to each first variable with other feature vectors in the feature vectors input into the feature extraction module, excluding the first feature vector, to obtain the fourth feature vector corresponding to each first variable. The fourth feature vectors corresponding to each first variable are then combined in the target order to obtain the first vector matrix. The first vector matrix is ​​input into the normalization module to determine a first confidence level for placing at least one of the items into each of the transport vehicles.

3. The method according to claim 2, wherein, The feature extraction module includes a feature extraction layer and a matrix transformation layer connected in sequence; the step of inputting each first feature vector, each second feature vector, and the third feature vector into the feature extraction module, so that the feature extraction module fuses the first feature vector corresponding to each first variable with the other feature vectors in the feature vectors input to the feature extraction module besides the first feature vector, to obtain the fourth feature vector corresponding to each first variable, includes: Each first feature vector, each second feature vector, and the third feature vector are input into the feature extraction layer. The feature extraction layer then fuses the first feature vector corresponding to each first variable with other feature vectors in the feature vectors input to the feature extraction module, excluding the first feature vector, to obtain a fourth feature vector corresponding to each first variable. The second feature vector corresponding to each second variable is then fused with other feature vectors in the feature vectors input to the feature extraction module, excluding the second feature vector, to obtain a fifth feature vector corresponding to each second variable. Finally, the fourth feature vector corresponding to each first variable and the fifth feature vector corresponding to each second variable are combined in an initial order to obtain a second vector matrix. Obtain a mapping table that stores the position information of each of the fourth feature vectors and each of the fifth feature vectors in the second vector matrix; The second vector matrix and the mapping table are input into the matrix transformation layer to obtain the fourth feature vector corresponding to each of the first variables from the second vector matrix based on the position information.

4. The method according to claim 1, wherein, The method further includes: If the objective function value of the correspondence does not meet the set threshold, at least one iteration is performed until the objective function value of the correspondence meets the set threshold. In the at least one iteration, based on the correspondence between the plurality of items and the plurality of transportation vehicles determined in the previous iteration, each of the first feature vectors, each of the second feature vectors and the third feature vector obtained in the previous iteration are updated. Based on the updated first feature vectors, each of the second feature vectors and the third feature vector, a machine learning model is used to determine the second confidence level of at least one of the items being placed in each of the transportation vehicles while satisfying the constraints and minimizing the objective function. Based on the second confidence level of at least one item being placed in each of the transportation vehicles, the correspondence between the plurality of items and the plurality of transportation vehicles is determined.

5. The method according to claim 4, wherein, The feature information of the corresponding variable represented by each of the first feature vectors and / or each of the second feature vectors includes at least one of the following: the value of the corresponding variable in the historical iteration, the number of times the value of the corresponding variable has changed in the historical iteration, the value of the corresponding variable with the most occurrences in the historical iteration, the maximum value of the corresponding variable in the historical iteration, the minimum value of the corresponding variable in the historical iteration, and whether the corresponding variable is a stable variable.

6. A method for training a machine learning model for transportation scheduling, wherein, The method includes: Multiple training samples are acquired, wherein at least one of the training samples includes attribute information of multiple sample items, attribute information of multiple sample transportation vehicles, an objective function, and constraints; the objective function represents the resource imbalance of the multiple sample transportation vehicles and is used to indicate the optimal solution of the target transportation relationship between the multiple items and the multiple transportation vehicles; the target transportation relationship is used to indicate each item and its corresponding transportation vehicle; the resources of the transportation vehicle include weight resources, i.e., the weight that the transportation vehicle can carry, or spatial resources, i.e., the space that the transportation vehicle has; the constraints include the restrictions on the multiple items when they are placed in the multiple transportation vehicles. For at least one of the training samples, based on the attribute information of multiple sample items, the attribute information of multiple sample transportation vehicles, the objective function, and the constraints, a machine learning model is used to determine the third confidence level of placing at least one of the sample items into each of the sample transportation vehicles while satisfying the constraints and minimizing the objective function. The machine learning model is trained based on the third confidence level of at least one of the sample items in at least one of the training samples being placed in each of the sample transport vehicles; For at least one of the training samples, based on the attribute information of multiple sample items, the attribute information of multiple sample transportation vehicles, the objective function, and the constraints, a machine learning model is used to determine the third confidence level of placing at least one of the sample items into each of the sample transportation vehicles while satisfying the constraints and minimizing the objective function, including: For at least one of the training samples, the attribute information of multiple sample items, the attribute information of multiple sample transportation vehicles, the objective function, and the constraint input solver are used to perform multiple iterations through the solver to obtain the sixth feature vector corresponding to multiple third variables Pi',b', the seventh feature vector corresponding to multiple fourth variables Q b',r', and the eighth feature vector corresponding to the constraint in at least one iteration. Here, the third variable Pi',b' indicates whether the i'th sample item is placed in the b'th sample transportation vehicle, the fourth variable Q b',r' indicates the imbalance value of the b'th sample transportation vehicle on the r'th type of resource, i' is an integer between 1 and I', b' is an integer between 1 and B', r' is an integer between 1 and R', I' is the quantity of the sample items, B' is the quantity of the sample transportation vehicles, R' is the number of types of resources, and I', B', and R' are integers greater than or equal to 1. Based on the sixth feature vector, the seventh feature vector, and the eighth feature vector corresponding to at least one round of iteration of at least one of the training samples, the machine learning model is used to determine the third confidence level of at least one of the sample items being placed in each of the sample transportation vehicles. The step of taking the attribute information of multiple sample items, the attribute information of multiple sample transportation vehicles, the objective function, and the constraint input solver for at least one training sample, and then performing multiple iterations through the solver, further includes: Obtain the values ​​of each of the third variables output by the solver in at least one iteration; The step of training the machine learning model based on a third confidence level of at least one of the sample items in at least one of the training samples being placed in each of the sample transport vehicles includes: Based on the third confidence level of at least one of the sample items placed in each of the sample transportation vehicles corresponding to at least one round of iteration of at least one of the training samples and the value of each of the third variables, the cross-entropy loss is determined. The machine learning model is trained based on the cross-entropy loss.

7. A transportation dispatching device, wherein, The device includes: The first acquisition module is used to acquire attribute information of multiple items, attribute information of multiple transportation vehicles, objective function, and constraints. The objective function represents the resource imbalance of the multiple transportation vehicles and is used to indicate the optimal solution of the target transportation relationship between the multiple items and the multiple transportation vehicles. The target transportation relationship indicates each item and its corresponding transportation vehicle. The resources of the transportation vehicles include weight resources (i.e., the weight the transportation vehicle can carry) or spatial resources (i.e., the space the transportation vehicle has). The constraints include restrictions on the multiple items when they are placed into the multiple transportation vehicles. The first determining module is used to determine, based on the attribute information of multiple items, the attribute information of multiple means of transport, the objective function, and the constraints, a machine learning model is used to determine a first confidence level for placing at least one item in each of the means of transport while satisfying the constraints and minimizing the objective function, so as to determine the correspondence between the multiple items and the multiple means of transport based on the first confidence level of placing at least one item in each of the means of transport. The second determining module is used to determine the correspondence as a target transportation relationship between the plurality of items and the plurality of transportation vehicles when the objective function value of the correspondence meets a set threshold. The first determining module includes: The first acquisition submodule is used to input the attribute information of multiple items, the attribute information of multiple transportation vehicles, the objective function, and the constraint into a solver, so as to obtain a first feature vector corresponding to multiple first variables Xi,b, a second feature vector corresponding to multiple second variables Yb,r, and a third feature vector corresponding to the constraint through the solver; wherein, the first variable Xi,b indicates whether the i-th item is placed in the b-th transportation vehicle, the second variable Yb,r indicates the imbalance value of the b-th transportation vehicle on the r-th resource, i is an integer between 1 and I, b is an integer between 1 and B, r is an integer between 1 and R, I is the quantity of the item, B is the quantity of the transportation vehicle, R is the number of types of the resource, and I, B, and R are integers greater than or equal to 1; The first determining submodule is used to determine, based on each of the first feature vectors, each of the second feature vectors and the third feature vectors, the machine learning model, a first confidence level of at least one of the items being placed in each of the transport vehicles. The first determining module includes: The second determining submodule is used to determine the first target transport vehicle corresponding to at least one target item among the at least one of the items, based on a first confidence level of at least one of the items being placed in each of the transport vehicles; The third determination submodule is used to input the attribute information of multiple items, the attribute information of multiple transportation vehicles, the objective function, the constraints, and the first target transportation vehicle corresponding to the at least one target item into the solver, so as to determine the second target transportation vehicle corresponding to the other items among the multiple items besides the at least one target item through the solver. The fourth determining submodule is used to determine the correspondence between the plurality of items and the plurality of transport vehicles based on the first target transport vehicle corresponding to the at least one target item and the second target transport vehicle corresponding to the other items.

8. The apparatus according to claim 7, wherein, The machine learning model includes a feature extraction module and a normalization module connected in sequence; The first determining submodule includes: The processing unit is configured to input each of the first feature vectors, each of the second feature vectors, and the third feature vector into the feature extraction module, so that the feature extraction module can fuse the first feature vector corresponding to each of the first variables with the other feature vectors in the feature vectors input to the feature extraction module except for the first feature vector to obtain the fourth feature vector corresponding to each of the first variables, and combine the fourth feature vectors corresponding to each of the first variables in a target order to obtain the first vector matrix. A determining unit is configured to input the first vector matrix into the normalization module to determine a first confidence level for placing at least one of the items into each of the transport vehicles through the normalization module.

9. The apparatus according to claim 8, wherein, The feature extraction module includes a feature extraction layer and a matrix transformation layer connected in sequence; the processing unit includes: The processing subunit is configured to input each of the first feature vectors, each of the second feature vectors, and the third feature vector into the feature extraction layer, so that the first feature vector corresponding to each of the first variables is fused with other feature vectors in the feature vectors input to the feature extraction module except for the first feature vector to obtain a fourth feature vector corresponding to each of the first variables, and the second feature vector corresponding to each of the second variables is fused with other feature vectors in the feature vectors input to the feature extraction module except for the second feature vector to obtain a fifth feature vector corresponding to each of the second variables, and the fourth feature vector corresponding to each of the first variables and the fifth feature vector corresponding to each of the second variables are combined in an initial order to obtain a second vector matrix; The first acquisition subunit is used to acquire a mapping table storing the position information of each of the fourth feature vectors and each of the fifth feature vectors in the second vector matrix; The second acquisition subunit is used to input the second vector matrix and the mapping relationship table into the matrix transformation layer to obtain the fourth feature vector corresponding to each of the first variables from the second vector matrix based on the position information.

10. The apparatus according to claim 7, wherein, The device further includes: The processing module is configured to perform at least one iteration when the objective function value of the correspondence does not meet the set threshold, until the objective function value of the correspondence meets the set threshold. In the at least one iteration, based on the correspondence between the plurality of items and the plurality of transportation vehicles determined in the previous iteration, each of the first feature vectors, each of the second feature vectors and the third feature vector obtained in the previous iteration are updated. Based on the updated first feature vectors, each of the second feature vectors and the third feature vector, a machine learning model is used to determine the second confidence level of at least one of the items being placed in each of the transportation vehicles while satisfying the constraints and minimizing the objective function. Based on the second confidence level of at least one item being placed in each of the transportation vehicles, the correspondence between the plurality of items and the plurality of transportation vehicles is determined.

11. The apparatus according to claim 10, wherein, The feature information of the corresponding variable represented by each of the first feature vectors and / or each of the second feature vectors includes at least one of the following: the value of the corresponding variable in the historical iteration, the number of times the value of the corresponding variable has changed in the historical iteration, the value of the corresponding variable with the most occurrences in the historical iteration, the maximum value of the corresponding variable in the historical iteration, the minimum value of the corresponding variable in the historical iteration, and whether the corresponding variable is a stable variable.

12. A training apparatus for a machine learning model used in transportation scheduling, wherein, The device includes: The second acquisition module is used to acquire multiple training samples, wherein at least one of the training samples includes attribute information of multiple sample items, attribute information of multiple sample transportation vehicles, an objective function, and constraints; the objective function represents the resource imbalance of the multiple sample transportation vehicles and is used to indicate the optimal solution of the target transportation relationship between the multiple items and the multiple transportation vehicles; the target transportation relationship is used to indicate each item and the corresponding transportation vehicle; the resources of the transportation vehicle include weight resources, i.e., the weight that the transportation vehicle can carry, or spatial resources, i.e., the space that the transportation vehicle has; the constraints include the restrictions on the multiple items when they are placed in the multiple transportation vehicles. The third determining module is used to determine, for at least one of the training samples, a third confidence level of placing at least one of the sample items into each of the sample transportation vehicles, based on the attribute information of multiple sample items, the attribute information of multiple sample transportation vehicles, the objective function, and the constraints, using a machine learning model. A training module is used to train the machine learning model based on a third confidence level of at least one of the sample items in at least one of the training samples being placed in each of the sample transport vehicles; The third determining module includes: The second acquisition submodule is used to, for at least one of the training samples, input the attribute information of multiple sample items, the attribute information of multiple sample transportation vehicles, the objective function, and the constraint input solver, and perform multiple iterations through the solver to obtain the sixth feature vector corresponding to multiple third variables Pi',b', the seventh feature vector corresponding to multiple fourth variables Q b',r', and the eighth feature vector corresponding to the constraint output by the solver in at least one iteration; wherein, the third variable Pi',b' indicates whether the i'th sample item is placed in the b'th sample transportation vehicle, the fourth variable Q b',r' indicates the imbalance value of the b'th sample transportation vehicle on the r'th type of resource, i' is an integer between 1 and I', b' is an integer between 1 and B', r' is an integer between 1 and R', I' is the number of sample items, B' is the number of sample transportation vehicles, R' is the number of types of resources, and I', B', and R' are integers greater than or equal to 1; The fifth determining submodule is used to determine the third confidence level of placing at least one of the sample items into each of the sample transport vehicles based on the sixth feature vector, the seventh feature vector, and the eighth feature vector corresponding to at least one round of iteration of at least one of the training samples, using the machine learning model. The third determining module further includes: The third acquisition submodule is used to acquire the values ​​of each of the third variables output by the solver in at least one iteration. The training module includes: The sixth determining submodule is used to determine the cross-entropy loss based on the third confidence level of at least one of the sample items placed in each of the sample transport vehicles corresponding to at least one round of iteration of at least one of the training samples and the value of each of the third variables. The training submodule is used to train the machine learning model based on the cross-entropy loss.

13. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-5, or to perform the method of claim 6.

14. A non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the method of any one of claims 1-5, or to perform the method of claim 6.

15. A computer program product comprising a computer program that, when executed by a processor, implements the steps of the method of any one of claims 1-5, or implements the steps of the method of claim 6.

Images