Method and system for logistics route planning

A graph neural network-based method learns node and edge embeddings in a logistics network, addressing inefficiencies in existing methods by providing real-time, adaptable, and scalable optimal path determination.

CN119067273APending Publication Date: 2024-12-03SIEMENS AG
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202410687061.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2023-05-30
Filing Date
2024-05-30
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

Existing methods for logistics route planning face challenges such as high computational complexity, inefficiency in handling real-time events, and inability to adapt to dynamic changes in logistics networks, leading to suboptimal path determination and reliance on outdated models.

Method used

A graph neural network-based approach that learns node and edge embeddings using a knowledge graph, incorporating real-time events, allowing for flexible adaptation to network changes without retraining, and providing scalable, automated path optimization.

Benefits of technology

Enables efficient, real-time determination of optimal logistics paths with reduced computational complexity, accommodating dynamic changes, and eliminating the need for domain knowledge, thus enhancing planning accuracy and flexibility.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067273A_ABST
    Figure CN119067273A_ABST
Patent Text Reader

Abstract

The invention discloses a method and a system for logistics route planning. A graph database stores a knowledge graph having nodes representing a current state of a logistics network and directed edges, where the nodes include: entity nodes, each entity node representing an entity in the logistics network; and event nodes, each event node representing an event, and wherein the directed edges include a delivery edge and an influence edge. In order to determine an optimal path between a sender and a receiver, a graph neural network model provides node and edge embedding. The neural network receives the embedding and calculates the probability that the entity nodes and the delivery edges are located on the optimal path for the corresponding entity nodes and the delivery edges. A greedy way-finding algorithm uses the calculated probability to calculate an optimal path. The method takes into account the effects of real-time events and serves as decision support for production planners and logistics personnel to identify optimal and alternative paths and make more intelligent plan choices. It also provides scalability because it has constant computational complexity, and the computational time is independent of hop counts, making it suitable for large maps.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] Independent of the use of the grammatical term, individuals with either masculine or feminine identities are included within the term.

[0002] In logistics networks, finding the optimal path between senders (e.g. suppliers) and receivers (e.g. production sites) is crucial for reliable planning, accurate delivery time forecasts, and flexible production. Flexible rerouting is necessary in the event of events that negatively affect transport flows. These events may include short-term events such as traffic jams, but also events such as natural disasters that may have a lasting impact on the logistics infrastructure.

[0003] Production planners often rely on delivery dates given by their suppliers without their own verification. Logistics simulations can support them with specific forecasts, but simulations require domain knowledge and predefined models that often only consider average values, such as those from processing times at a warehouse, without considering the impact of real-time events.

[0004] Generally, finding the optimal (or shortest) path between a sender and a receiver has been addressed by various works.

[0005] First, the shortest path algorithm: A variety of existing algorithms solve the problem of finding the shortest path between two nodes in a graph, for example, Dijkstra, EW, A Note on Two Problems in Connexion with Graphs, Numerische Mathematik 1, pp. 269-271, 1959; Bang-Jensen, J. and Gutin, G., Section 2.3.4: The Bellman-Ford-Moore Algorithm, in Digraphs: Theory, Algorithms and Applications (First ed.), pp. 55-58, 2000; and Dechter, R. and Pearl, J., Generalized Best-First Search Strategies and the Optimality of A*, Journal of the Association for Computing Machinery, Vol. 32, No. 3, July 1985, pp. 505-536. However, these methods are based on graph traversal and exhibit high computational complexity, making them unsuitable for large graphs.

[0006] Second, domain knowledge and human intuition: Given domain knowledge, humans can well identify approximate optimal paths on maps, as disclosed in Bongiorno, C. et al., Vector-based Pedestrian Navigation in Cities, Nature Computational Science 1(10), pp. 678-685, 2021. However, the disadvantages are that domain knowledge needs to be available, there is no guarantee that the optimal path can be found, and manual pathfinding is time-consuming.

[0007] Third, logistics simulation: Supply chains and transportation networks can be simulated by creating corresponding digital twins. Such simulators can investigate different scenarios to identify optimal or alternative paths and predict outcomes, such as delivery times. However, these simulations are only as good as their underlying models and rules. If the model does not reflect real-world conditions, the results may not make sense.

[0008] Fourth, graph representation: Neural network methods, such as the method disclosed in Osanlou, K. et al., Constrained shortest path search with graph convolutional neural networks, arXiv preprint arXiv:2108.00978 v1 [cs. AI], 2021, available on the Internet at https: / / doi.org / 10.48550 / arXiv.2108.00978 on May 8, 2023, directly use the graph as input for path optimization, where the input is usually in the form of an adjacency matrix. However, such methods can only be applied to fixed graphs, because any changes in the graph structure require retraining the model.

[0009] The purpose of the present invention is to identify problems in the prior art and to find technical solutions to the problems.

[0010] The object of the invention is solved by the independent claims. Further advantageous arrangements and embodiments of the invention are set forth in the respective dependent claims.

[0011] According to the method for logistics route planning, the following operations are performed by a component, wherein the component is a hardware component and / or a software component executed by one or more processors:

[0012] -Storing a knowledge graph with nodes and directed edges representing the current state of the logistics network through a graph database, where the nodes include

[0013] - entity nodes, each of which represents an entity in the logistics network, in particular a supplier, distribution center, port, airport or production site, and

[0014] - event nodes, each event node representing an event, wherein each event is an event that negatively affects at least one of the entities,

[0015] And the directed edges include

[0016] - delivery edges between entity nodes, and

[0017] - impact edges between each event node and one or more entity nodes,

[0018] - receiving, through a user interface, a selection of a sender node and a receiver node among the entity nodes,

[0019] - Determine at least one optimal path option between the sender node and the receiver node by:

[0020] -Provide node embeddings for each entity node based on the corresponding node embeddings and edge embeddings for each delivery edge through a graph neural network model that receives the knowledge graph as input,

[0021] - calculating, for each entity node, a probability that the entity node is located on the optimal path by a first neural network that receives the corresponding node embedding as input,

[0022] - calculating for each delivery edge the probability that the delivery edge is on the optimal path by a second neural network that receives as input the corresponding edge embedding,

[0023] - using the calculated probabilities to calculate at least one optimal path option between the sender node and the receiver node,

[0024] - outputting at least one optimal path option through a user interface.

[0025] The system for logistics route planning includes the following components:

[0026] - A graph database configured to store a knowledge graph having nodes and directed edges representing a current state of a logistics network, wherein the nodes include

[0027] - entity nodes, each of which represents an entity in the logistics network, in particular a supplier, distribution center, port, airport or production site, and

[0028] - event nodes, each event node representing an event, wherein each event is an event that negatively affects at least one of the entities,

[0029] And the directed edges include

[0030] - delivery edges between entity nodes, and

[0031] - influence edges between each event node and one or more entity nodes,

[0032] - A user interface configured for

[0033] - receiving a selection of a sender node and a receiver node among the entity nodes, and

[0034] - output at least one optimal path option,

[0035] - the following component, wherein the component is a hardware component and / or a software component executed by one or more processors, and wherein the component is configured to determine at least one optimal path option between a sender node and a receiver node:

[0036] - a graph neural network model configured to receive a knowledge graph as input, and to provide a node embedding for each entity node and an edge embedding for each delivery edge based on a corresponding node embedding,

[0037] - a first neural network configured to receive each node embedding as input, and to calculate, for a corresponding entity node, a probability that the entity node is located on the optimal path,

[0038] a second neural network configured to receive each edge embedding as input, and to calculate, for a corresponding delivery edge, a probability that the delivery edge is on the optimal path,

[0039] - a component configured to calculate at least one optimal path option between a sender node and a receiver node using the calculated probabilities.

[0040] The following advantages and explanations are not necessarily the result of the object of the independent claims. Rather, they may be advantages and explanations that apply only to certain embodiments or variants.

[0041] The term "computer" should be interpreted as broadly as possible and in particular covers all electronic devices with data processing properties. Thus, a computer may be, for example, a personal computer, a server, a client, a programmable logic controller (PLC), a handheld computer system, a pocket PC device, a mobile radio device, a smartphone, or any other communication device that can process data with the aid of a computer, such as a processor and other electronic devices for data processing. A computer may in particular comprise one or more processors and a memory unit.

[0042] In conjunction with the present invention, "memory", "memory unit" or "memory module" and the like may refer to volatile memory, for example in the form of random access memory (RAM), or permanent memory such as a hard disk, solid state drive or magnetic disk.

[0043] The method and system, or at least some of their embodiments, provide a graph neural network-based approach that learns node and edge embeddings while taking into account the impact of real-time events. The probability of each node and edge being on an optimal path is obtained based on the learned embeddings. The method and system, or at least some of their embodiments, can be used as decision support for production planners and logisticians to identify optimal and alternative paths and make better informed planning choices.

[0044] The method and system, or at least some of their embodiments, provide scalability because they have constant computational complexity and the computation time does not depend on the number of hops. Compared to many shortest path algorithms and manual pathfinding, they are applicable to large graphs.

[0045] The method and system, or at least some of their embodiments, provide usability for non-experts, since no domain knowledge is needed to use them to find optimal paths in an automated manner. Unlike logistics simulations, they do not require pre-definition of a model that should reflect real-world situations.

[0046] An embodiment of the method comprises an additional step of adapting the knowledge graph due to current changes in the logistics network by removing and / or adding entity nodes and / or delivery edges, and / or removing and / or adding event nodes and / or impact edges before determining at least one optimal path option.

[0047] This embodiment provides flexibility because it can adapt to changes in the graph structure without retraining. The graph neural network model can handle new nodes / edges as well as removed nodes / edges. This is advantageous because in a logistics network, frequent changes may occur, for example, when a new event affects an airport, making it impossible to make a delivery to the airport that day.

[0048] In an embodiment of the method, the knowledge graph is adapted in real time to reflect the dynamically changing state of the logistics network.

[0049] According to this embodiment, the model can even give predictions in real time. Since the knowledge graph is updated in real time to reflect the dynamically changing state of the logistics network, the model can also be used to provide changing optimal path options in real time.

[0050] In another embodiment of the method, a greedy path-finding algorithm calculates at least one optimal path option between a sender node and a receiver node.

[0051] In another embodiment of the method, a greedy path-finding algorithm initializes each optimal path option with a sender node and greedily appends delivery edges and entity nodes with the highest probability to the optimal path option until a receiver node is reached.

[0052] In another embodiment of the method, the first neural network and the second neural network are multilayer perceptrons.

[0053] In an embodiment of the method, a graph neural network model is constructed to receive an initialization of each node embedding, wherein in particular, each node embedding is initialized with a one-hot encoded vector declaring whether the node represents a sender node, a receiver node, or another node, and each layer of the graph neural network model is iteratively applied to the node embeddings.

[0054] In another embodiment of the method, the graph neural network model is constructed to obtain node embedding by the following formula when applying each layer of the graph neural network model:

[0055]

[0056] in

[0057] -i identifies the node,

[0058] -z i represents node embedding,

[0059] -l identifies the layer,

[0060] -t j Indicates the node type,

[0061] -N(i) represents the neighbors of the node,

[0062] - is a scalar value,

[0063] - is a trainable parameter,

[0064] -σ1 is a nonlinear activation function, and

[0065] -M(·;t j ) is a multilayer perceptron depending on the node type.

[0066] This embodiment allows adding new logistics and event nodes to the knowledge graph without retraining the graph neural network model. Therefore, the graph neural network model can flexibly adapt to structural changes and / or new events in the logistics network.

[0067] In an embodiment of the method, each directed edge is weighted with a weight, and the scalar value depends on the weight and is calculated as follows:

[0068]

[0069] in

[0070] -w ij represents the weight,

[0071] - and is a trainable parameter, and

[0072] -σ2 is a non-linear activation function.

[0073] In another embodiment of the method, the weight of the delivery edge represents a key performance indicator, in particular delivery time, cost, emissions, reliability or any aggregation of these values, and the weight of the impact edge represents an impact level.

[0074] A computer program product having program instructions for executing the method according to one of the method claims.

[0075] Supply device for a computer program product according to the preceding claim, wherein the supply device stores and / or provides a computer program product.

[0076] The foregoing and other aspects of the present invention are best understood from the following detailed description when read in conjunction with the accompanying drawings. For the purpose of illustrating the present invention, presently preferred embodiments are shown in the accompanying drawings, however, it is to be understood that the present invention is not limited to the specific means disclosed. The embodiments may be combined with each other. In addition, the embodiments may be combined with any of the above-mentioned features. Unless otherwise stated, the same reference numerals between the drawings represent the same features or functionally the same elements. Included in the drawings are the following figures:

[0077] Figure 1 A first embodiment is shown,

[0078] Figure 2 Another embodiment is shown,

[0079] Figure 3 A subgraph SG representing a part of the logistics network is shown,

[0080] Figure 4 A flow chart of a possible exemplary embodiment of a method for logistics route planning is shown.

[0081] In the following description, various aspects of the present invention and its embodiments will be described. However, it will be appreciated by those skilled in the art that the embodiments may be implemented with only some or all of them. For purposes of explanation, specific numbers and configurations are set forth to provide a thorough understanding. However, it will also be clear to those skilled in the art that the embodiments may be implemented without these specific details.

[0082] The components described may each be a hardware component or a software component. For example, a software component may be a software module such as a software library; a separate procedure, subroutine or function; or, depending on the programming paradigm, any other portion of a software code that implements the functionality of a software component. In particular, a combination of hardware components and software components may occur if some of the effects according to the invention are preferably implemented exclusively by special hardware (e.g., a processor in the form of an ASIC or FPGA), while some other parts are implemented by software.

[0083] Figure 1 A sample structure of a computer implementation of the present invention is shown, comprising:

[0084] (101) Computer system

[0085] (102) Processor

[0086] (103) Memory

[0087] (104) Computer program (product)

[0088] (105) User Interface

[0089] In this embodiment of the invention, the computer program 104 includes program instructions for performing the invention. The computer program 104 is stored in the memory 103, which, among other things, provides a supply device for the computer program 104 for the memory 103 and / or its associated computer system 101. The computer system 101 can perform the invention by executing the program instructions of the computer program 104 by the processor 102. The results of the invention can be presented on the user interface 105. Alternatively, they can be stored in the memory 103 or in another suitable device for storing data.

[0090] Figure 2 Another sample structure of a computer implementation of the present invention is shown, comprising:

[0091] (201)Supply of equipment

[0092] (202) Computer program (product)

[0093] (203) Computer Network / Internet

[0094] (204) Computer system

[0095] (205) Mobile devices / smart phones

[0096] In this embodiment, a supply device 201 stores a computer program 202, which includes program instructions for performing the present invention. The supply device 201 provides the computer program 202 via a computer network / Internet 203. As an example, a computer system 204 or a mobile device / smartphone 205 can load the computer program 202 and perform the present invention by executing the program instructions of the computer program 202.

[0097] In a variation of this embodiment, the supply device 201 is a computer-readable storage medium, such as an SD card, storing a computer program 202 and is directly connected to a computer system 204 or a mobile device / smartphone 205 so that it loads the computer program 202 and performs the present invention by executing the program instructions of the computer program 202.

[0098] Preferably, Figure 3 to Figure 4 The embodiment shown in FIG. Figure 1 or Figure 2 This is achieved using the structure shown in .

[0099] The following embodiments describe a method based on a graph neural network that learns node and edge embeddings of a knowledge graph while taking into account the impact of real-time events. The probability of each node and edge being on the optimal path is obtained based on the learned embeddings. Therefore, the following embodiments can be used as a decision support system for production planners and logistics personnel to identify optimal and alternative logistics paths and make more informed planning choices.

[0100] The following embodiment provides a method for finding the optimal route between a sender (eg, a supplier) and a receiver (eg, a production site), which takes into account the impact of current events. Note that the following embodiment assumes that there are infinite resources along the logistics path.

[0101] Knowledge Graph Definition

[0102] Let G = (V, E, T V ,T E ) is a knowledge graph, where V represents the set of nodes, E represents the set of edges, and T V represents a set of node types, and T E Represents a collection of edge types.

[0103] T V The node types in include entities in the logistics network, such as suppliers, distribution centers, ports, airports, and production sites, as well as events (events that negatively affect entities in the logistics network).

[0104] T E The edge types in include "deliver to" connecting entities in the logistics network and "influence" connecting events with one or more entities in the logistics network.

[0105] An edge is a tuple (i, j, t e ,w) description, where the directed edge between i∈V and j∈V has type t e ∈T E . Edge (i,j,t e ,w) has weight It can be interpreted as t e = Impact level of "impact". e = "delivered to", the weight may represent a KPI, such as delivery time, cost, emissions, reliability, or any aggregation of values ​​of interest. It is possible that there are several edges with different weights between two nodes, for example when there are several possible delivery paths.

[0106] Figure 3 An exemplary subgraph SG of a knowledge graph G is shown. The subgraph contains entity nodes EN representing entities in a logistics network, event nodes IN representing events in the logistics network, delivery edges DE between entity nodes EN, and influence edges between each event node IN and one or more entity nodes EN. As mentioned above, the delivery edge DE has an edge type of "delivered to", and the influence edge has an edge type of "influence". Figure 3 Examples of the weights discussed above are also shown.

[0107] Node Embedding

[0108] Given a knowledge graph G, the current embodiment uses a graph neural network model with L layers to learn embeddings of all nodes.

[0109] The embedding of a node i is used to declare that the node represents a sender (sender node), a receiver (receiver node), or other (one-hot encoded vector). Then, the embedding of node i when applying layer l is obtained by

[0110]

[0111] Where N(i) represents the neighbors of node i, are trainable parameters of the model, and σ1 is a nonlinear activation function.

[0112] M(·;t j ) is a multilayer perceptron, which depends on the node type t j ∈T V, that is, all logistics nodes share the same parameters, and all event nodes share the same parameters. Due to this construction, new logistics and event nodes can be added to the graph without retraining the entire model, and the model can flexibly adapt to structural changes or new events in the logistics network.

[0113] Scalar value Depends on the edge weight w ij and is calculated as follows:

[0114]

[0115] in and is a trainable parameter, and σ2 is a nonlinear activation function.

[0116] After applying all graph neural network layers, the final node embedding of node i is given by Given.

[0117] Edge Embedding

[0118] Since the current embodiment is interested in the optimal delivery path and wants to know which edges between logistics entities belong to it, it defines its embedding for each edge (i, j, 'deliver to', w) as

[0119]

[0120] Optimal path probability

[0121] For logistics node i, the probability that i is part of the optimal path is given by the first multilayer perceptron Given.

[0122] Similarly, for edge (i,j,'delivered to',w), the probability that (i,j,'delivered to',w) is on the optimal path is given by the second multilayer perceptron Given.

[0123] For example, through greedy pathfinding, the optimal path options can be derived with respect to the nodes and edges with the highest probability.

[0124] Model Training

[0125] In this section, the term model refers to both the graph neural network model and the first multilayer perceptron f v and the second multilayer perceptron f e , as they are trained in tandem to produce useful embeddings and probabilities, respectively.

[0126] The model's weights are trained using a gradient descent algorithm to minimize the difference between the predictions and the pre-computed training labels. The training labels can be obtained using analytical methods such as greedy search. After the model is trained, the training labels are no longer needed.

[0127] The entire knowledge graph containing all available knowledge about the existing logistics network forms the training dataset. During training, a subgraph is randomly sampled from the training dataset and the sender node as well as the receiver node are selected from the nodes of the subgraph. The node embeddings are initialized based on the configuration of the problem. In other words, for each subgraph sampled from the knowledge graph, the embedding of each node is initialized with a one-hot encoded vector that declares whether the node is a sender node, a receiver node, or something else.

[0128] The model then performs a forward pass, predicting the probability of each node and edge of the subgraph being on the optimal path, comparing it with the pre-computed training labels to obtain a loss that guides the optimization of the model's parameters.

[0129] When constructing the training dataset, nodes and edges are randomly added or removed to simulate the occurrence or end of events in the logistics network. The generated subgraphs have different numbers of nodes and edges, thus representing the real-world complexity of logistics networks. In addition, different path lengths between sender nodes and receiver nodes are sampled to ensure that the model can be generalized to a wide range of scenarios.

[0130] In other words, the training aims to cover as many variations as possible by preparing subgraphs with different levels of information. For different subgraph configurations, there may be different optimal paths. The model is trained to be robust to the input configuration.

[0131] deploy

[0132] During deployment, the entire knowledge graph is fed into the graph neural network model for accurate predictions. The trained graph neural network model can be used to generate node and edge embeddings. Subsequently, the remaining v and the second multilayer perceptron f e A forward pass is performed to obtain node and edge probabilities. Starting from the source node, a candidate for the optimal path can then be constructed by greedily appending nodes and edges with the highest probability to the path until the target node is reached.

[0133] In case of dynamic changes in the logistics network (e.g., edges in the knowledge graph need to be removed due to failures in the transportation system), predictions are performed based on the updated topology of the knowledge graph. In other words, the knowledge graph is adapted by removing and / or adding entity nodes and / or delivery edges, and / or by removing and / or adding event nodes and / or impact edges. The updated knowledge graph is then fed into the graph neural network model for the next prediction.

[0134] Entity nodes and event nodes can be added to the knowledge graph as long as they share the same node type as the entity nodes and event nodes encountered by the graph neural network model during training. There is no need to retrain model parameters because the graph neural network model already has embeddings for such types of nodes.

[0135] The model can even give predictions in real time. If the knowledge graph is updated in real time to reflect the dynamic changing state of the logistics network, the model can be used to provide changing optimal path options in real time.

[0136] Figure 4 A flow chart of a possible exemplary embodiment of a method for logistics route planning is shown.

[0137] In the first step 1, the graph database stores a knowledge graph with nodes and directed edges representing the current state of the logistics network. The knowledge graph has been discussed in detail above, where Figure 3 A subgraph of the knowledge graph is depicted in .

[0138] In a second step 2, the user interface receives a selection of a sender node and a receiver node among the entity nodes.

[0139] Determining at least one optimal path option between a sender node and a receiver node begins in a third step 3, where a graph neural network model receives a knowledge graph as input and provides a node embedding for each entity node and an edge embedding for each delivery edge based on the corresponding node embedding.

[0140] In the fourth step 4, the first neural network receives each node embedding as input and calculates the probability of the corresponding entity node being on the optimal path.

[0141] In the fifth step 5, the second neural network receives the corresponding edge embedding as input and calculates, for each delivery edge, the probability that the delivery edge is on the optimal path.

[0142] A sixth step 6 consists of calculating at least one optimal path option between the sender node and the receiver node using the calculated probabilities.

[0143] In a seventh step 7 , the user interface outputs at least one optimal path option.

[0144] With appropriate software and / or hardware architecture, some of these steps can be performed simultaneously. For example, the first neural network and the second neural network can operate in parallel.

[0145] For example, the method may be performed by one or more processors. Examples of processors include a microcontroller or microprocessor, an application specific integrated circuit (ASIC), or a neuromorphic microchip, in particular a neuromorphic processor unit. The processor may be part of any kind of computer, including a mobile computing device such as a tablet, smartphone, or laptop, or may be part of a server in a control room or in the cloud.

[0146] The above method can be implemented via a computer program product, which includes one or more computer-readable storage media on which instructions executable by one or more processors of a computing system are stored. Execution of the instructions causes the computing system to perform operations corresponding to the actions of the above method.

[0147] Instructions for implementing the processes or methods described herein may be provided on non-temporary computer-readable storage media or memory (such as cache, buffer, RAM, flash memory, removable media, hard drive or other computer-readable storage media). Computer-readable storage media include various types of volatile and non-volatile storage media. Functions, actions or tasks illustrated in the figure or described herein may be performed in response to one or more instruction sets stored in or on a computer-readable storage medium. Functions, actions or tasks may be independent of a particular type of instruction set, storage medium, processor or processing strategy, and may be performed by software, hardware, integrated circuits, firmware, microcode and the like operating alone or in combination. Similarly, processing strategies may include multi-processing, multi-tasking, parallel processing and the like.

[0148] The present invention has been described in detail with reference to the embodiments and examples of the present invention. However, within the spirit and scope of the present invention covered by the claims, changes and modifications may be made. The phrase "at least one of A, B and C" as an alternative expression may specify that one or more of A, B and C may be used.

Claims

1. A computer-implemented method for logistics route planning, wherein the components perform the following operations, and in, A component is a hardware component and / or a software component executed by one or more processors: -Store (1) a knowledge graph having nodes and directed edges representing the current state of the logistics network through a graph database, wherein the nodes include - Entity Nodes (EN), each of which represents an entity in the logistics network, in particular a supplier, distribution center, port, airport or production site, and - event nodes (IN), each event node (IN) representing an event, wherein each event is an event that negatively affects at least one of the entities, And the directed edges include - Delivery edges (DE) between entity nodes (EN), and - influence edges (IE) between each event node (IN) and one or more entity nodes (EN), - receiving (2) a selection of a sender node and a receiver node among the entity nodes (EN) through a user interface, - Determine at least one optimal path option between the sender node and the receiver node by: - providing (3) node embeddings for each entity node (EN) and (3) edge embeddings for each delivery edge (DE) based on the corresponding node embeddings via a graph neural network model that receives the knowledge graph as input, - calculating (4) for each entity node (EN) a probability that the entity node (EN) is located on the optimal path by a first neural network that receives the corresponding node embedding as input, - calculating (5) for each delivery edge (DE) the probability that the delivery edge (DE) is on the optimal path by a second neural network that receives the corresponding edge embedding as input, - using the calculated probabilities to calculate (6) at least one optimal path option between the sender node and the receiver node, - Outputting (7) at least one optimal route option via a user interface.

2. The method according to claim 1, wherein before determining at least one optimal path option, - Adapt the knowledge graph due to current changes in the logistics network by: - remove and / or add entity nodes (EN) and / or delivery edges (DE), and / or -Remove and / or add event nodes (IN) and / or influence edges (IE).

3. The method according to claim 2, -Where the knowledge graph is adapted in real time to reflect the dynamically changing state of the logistics network.

4. The method according to any one of the preceding claims, - wherein the greedy path finding algorithm calculates at least one optimal path option between a sender node and a receiver node.

5. The method according to claim 4, -Where the greedy pathfinding algorithm initializes each optimal path option with a sender node and greedily appends delivery edges (DE) and entity nodes (EN) with the highest probability to the optimal path option until the receiver node is reached.

6. The method according to any one of the preceding claims, - wherein the first neural network and the second neural network are multilayer perceptrons.

7. The method according to any one of the preceding claims, - wherein the graph neural network model is constructed to - receive an initialization of each node embedding, wherein in particular, each node embedding is initialized with a one-hot encoded vector declaring whether the node represents a sender node, a receiver node, or another node, and -Iteratively apply each layer of the graph neural network model to node embeddings.

8. The method according to claim 7, -The graph neural network model is constructed to obtain node embedding by the following formula when applying each layer of the graph neural network model: in -i identifies the node, -z i represents node embedding, -l identifies the layer, -t j Indicates the node type, -N(i) represents the neighbors of the node, is a scalar value, is a trainable parameter, -σ 1 is a nonlinear activation function, and -M(·;t j ) is a multilayer perceptron depending on the node type.

9. The method according to claim 8, - where each directed edge is weighted by a weight, and - where the scalar value depends on the weight and is calculated as follows: in -w ij represents the weight, and is a trainable parameter, and -σ 2 is a non-linear activation function.

10. The method according to claim 9, wherein - the weight of the Delivery Edge (DE) represents a key performance indicator, in particular delivery time, cost, emissions, reliability or any aggregation of these values, and -The weight of the influence edge (IE) indicates the influence level.

11. A system for logistics route planning, include: - a graph database configured to store (1) a knowledge graph having nodes and directed edges representing a current state of a logistics network, wherein the nodes include - Entity Nodes (EN), each of which represents an entity in the logistics network, in particular a supplier, distribution center, port, airport or production site, and - event nodes (IN), each event node (IN) representing an event, wherein each event is an event that negatively affects at least one of the entities, And the directed edges include - Delivery edges (DE) between entity nodes (EN), and - influence edges (IE) between each event node (IN) and one or more entity nodes (EN), - A user interface configured for - receiving (2) a selection of a sender node and a receiver node among the entity nodes (EN), and - output (7) at least one optimal path option, - the following component, wherein the component is a hardware component and / or a software component executed by one or more processors, and wherein the component is configured to determine at least one optimal path option between a sender node and a receiver node: - a graph neural network model configured to receive a knowledge graph as input, and to provide (3) node embeddings for each entity node (EN) and (3) edge embeddings for each delivery edge (DE) based on the corresponding node embeddings, a first neural network configured to receive each node embedding as input, and to calculate (4) for a corresponding entity node (EN) a probability that the entity node (EN) is located on the optimal path, a second neural network configured to receive as input each edge embedding and to compute (5) for a corresponding delivery edge (DE) a probability that the delivery edge (DE) is on the optimal path, - a component configured to calculate (6) at least one optimal path option between a sender node and a receiver node using the calculated probabilities.

12. Computer program product having program instructions for executing the method according to one of the method claims.

13. Supply device for a computer program product according to the preceding claim, wherein the supply device stores and / or provides a computer program product.

Citation Information

Cited By

  • Product logistics management scheduling method based on block chain, medium and equipment

    CN120725554A

  • Blockchain-based product logistics management scheduling method, medium and device

    CN120725554B