Dynamic routing method and device for delay tolerant network, equipment and storage medium

Abstract

The invention provides a dynamic routing method, device and equipment for a delay-tolerant network and a storage medium, and the method comprises the steps: splicing node feature information in a target delay-tolerant network with target data packet information to obtain node feature input, and constructing an adjacent matrix of the target delay-tolerant network; the node feature input and the adjacent matrix are jointly used as input of a trained graph neural network model, and the probability of each node corresponding to the target data packet is obtained; forwarding the target data packet based on the probability that each node corresponds to the target data packet; wherein the graph neural network model is obtained by training on the basis of global information of a sample delay tolerant network and information of a sample data packet with a forwarding node label. Therefore, under the condition that the network structure of the delay tolerant network changes dynamically, the optimal forwarding node can be selected in the data packet forwarding process, and the success rate of the routing method is improved.

Description

Technical Field

[0001] The present invention relates to the field of wireless communication technology, and in particular to a dynamic routing method, device, equipment and storage medium for a delay tolerant network. Background Art

[0002] In traditional networks, communication between nodes requires that the source node and the destination node maintain a continuous communication link. However, in some extreme environments, due to the sparsity of nodes, the rapid movement of nodes, and other reasons, the network topology is in frequent changes, and it is difficult to have a definite path between end-to-end. Traditional networks usually perform poorly in this situation, and Delay Tolerant Network (DTN) is an effective solution in this scenario. Delay Tolerant Network works in a "carry-store-forward" manner. Each node caches the received data packets and forwards the data packets when it meets other nodes, forwarding the data packets to the destination node hop by hop.

[0003] Existing routing methods for delay-tolerant networks usually only focus on the energy consumption and contribution rate of nodes to select routes, but do not take into account that in actual situations, the rapid movement of nodes will cause the network structure to be constantly changing. This dynamic change of the network structure will have a great impact on the choice of routes, resulting in a low success rate of routing methods. Summary of the invention

[0004] The present invention provides a dynamic routing method, device, equipment and storage medium for a delay tolerant network, which are used to solve the defect in the prior art that the dynamic change of the network structure affects the success rate of the routing method, and improve the success rate of the routing method in the delay tolerant network.

[0005] The present invention provides a dynamic routing method for a delay tolerant network, comprising the following steps: splicing node feature information in a target delay tolerant network with target data packet information to obtain node feature input, and constructing an adjacency matrix of the target delay tolerant network; The node feature input and the adjacency matrix are used together as inputs of the trained graph neural network model to obtain the probability that each node corresponds to the target data packet; forwarding the target data packet based on the probability that each node corresponds to the target data packet; The graph neural network model is trained based on the global information of the sample delay-tolerant network and the sample data packet information with forwarding node labels.

[0006] According to a dynamic routing method for a delay tolerant network provided by the present invention, the node feature information includes: The moving speed information of each node; Data transmission rate information of each node; Buffer occupancy information of each node; The location information of each node.

[0007] According to a dynamic routing method for a delay-tolerant network provided by the present invention, the target data packet information includes: Source node identification information of the target data packet; Destination node identification information of the target data packet; The size information of the target data packet; The remaining lifetime information of the target data packet.

[0008] According to a dynamic routing method for a delay tolerant network provided by the present invention, forwarding the target data packet based on the probability that each node corresponds to the target data packet specifically includes: Based on the probability that each node corresponds to the target data packet, determining the probability that the nodes in the neighbor set of the current node correspond to the target data packet; The probability that the nodes in the neighbor set of the current node correspond to the target data packet is arranged in descending priority order, and the target data packet is forwarded.

[0009] According to a dynamic routing method of a delay tolerant network provided by the present invention, the node feature information in the target delay tolerant network is obtained based on the feature information of the current node and the node feature set stored by the current node; The adjacency matrix is ​​constructed by depth-first traversing each neighboring node connected to the current node.

[0010] According to a dynamic routing method for a delay-tolerant network provided by the present invention, the graph neural network model consists of multiple graph convolutional network GCN layers, a global average pooling layer and a linear layer.

[0011] The present invention also provides a dynamic routing device for a delay tolerant network, comprising the following modules: An acquisition module, used for concatenating node feature information in a target delay tolerant network with target data packet information to obtain node feature input, and constructing an adjacency matrix of the target delay tolerant network; An input module, used to use the node feature input and the adjacency matrix as inputs of the trained graph neural network model to obtain the probability that each node corresponds to the target data packet; A forwarding module, configured to forward the target data packet based on the probability that each node corresponds to the target data packet; The graph neural network model is trained based on the global information of the sample delay-tolerant network and the sample data packet information with forwarding node labels.

[0012] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the dynamic routing method of the delay-tolerant network as described above is implemented.

[0013] The present invention also provides a non-transitory computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the dynamic routing method of the delay-tolerant network as described above is implemented.

[0014] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the dynamic routing method for a delay tolerant network as described above is implemented.

[0015] The dynamic routing method, apparatus, device and storage medium of a delay tolerant network provided by the present invention trains a graph neural network model based on the global information of a sample delay tolerant network and sample data packet information with a forwarding node label, then concatenates the node feature information in the target delay tolerant network with the target data packet information to obtain a node feature input, and constructs an adjacency matrix of the target delay tolerant network. The node feature input and the adjacency matrix are used together as inputs of the trained graph neural network model to obtain the probability that each node corresponds to the target data packet, and then the target data packet is forwarded according to the probability that each node corresponds to the target data packet. In this way, when the network structure of the delay tolerant network changes dynamically, the optimal forwarding node can be selected during the data packet forwarding process, thereby improving the success rate of the routing method. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 It is a flow chart of the dynamic routing method of the delay tolerant network provided by the present invention.

[0018] Figure 2 The figure is a flow chart of the operation of the routing method provided by the present invention.

[0019] Figure 3 It is a structural schematic diagram of a dynamic routing device for a delay tolerant network provided by the present invention.

[0020] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] Combine the following Figure 1-Figure 4 The invention describes a dynamic routing method, device, equipment and storage medium for a delay tolerant network.

[0023] Figure 1 is a flow chart of a dynamic routing method for a delay tolerant network provided by the present invention, such as Figure 1 As shown, the method comprises the following steps: Step 100: Concatenate node feature information in a target delay tolerant network with target data packet information to obtain node feature input, and construct an adjacency matrix of the target delay tolerant network.

[0024] Step 101: Use the node feature input and the adjacency matrix as inputs to the trained graph neural network model to obtain the probability that each node corresponds to the target data packet.

[0025] Step 102: forward the target data packet based on the probability that each node corresponds to the target data packet.

[0026] Among them, the graph neural network model is trained based on the global information of the sample delay-tolerant network and the sample data packet information with forwarding node labels.

[0027] Specifically, the target data packet is any data packet that needs to be sent by a service on a node in the target delay-tolerant network. For each data packet, there are multiple possible transmission paths from the source node to the target node. The probability of successful transmission of each path is the product of the probability of encountering all nodes on the path. The probability of successful transmission of a data packet is the sum of the probabilities of successful transmission of all possible paths. The purpose of the solution provided by the present invention is to improve the probability of successful transmission of the target data packet.

[0028] First, the graph neural network model can be trained based on the global information of the sample delay tolerant network and the sample data packet information with forwarding node labels, wherein the global information of the sample delay tolerant network includes the node feature information of all nodes in the sample delay tolerant network and the connection relationship between each node.

[0029] The model input of the graph neural network model includes two parts: node feature input and adjacency matrix. For the target delay-tolerant network, it can be represented as a dynamic graph , represents the set of nodes in the network, is a set of time-related edges, indicating that The edge that exists in the network at any time. Since the nodes in the network are constantly moving, when two nodes move within the communication range, the two nodes can transmit data, and at this time there is an edge in the graph.

[0030] The node feature information in the target delay-tolerant network can be concatenated with the target data packet information to obtain the node feature input. , there is one The node features of the entire graph can be represented as a matrix ,in Is a node The target data packet information can then be added to the feature vector of each node to obtain the node feature input.

[0031] The adjacency matrix of the target delay-tolerant network can be constructed based on the current connection status of the network. , where if there is an edge from node Connect to a Node but ,otherwise .

[0032] Optionally, the graph neural network model consists of multiple graph convolutional networks (GCN) layers, a global average pooling layer, and a linear layer.

[0033] Specifically, in the graph neural network model, multiple GCN layers can be used to update the feature information of the node first, and finally obtain the feature representation of the node. The calculation process of each GCN layer is as follows:

[0034] in The adjacency matrix of self-loops is added to each node. is the identity matrix. and is the trainable weight matrix of each layer. is an activation function, such as . is the activation matrix of layer l, .

[0035] Afterwards, a global average pooling layer can be used to aggregate node features into graph-level features. The calculation process is as follows:

[0036] in, The output node of the last GCN layer The feature representation of .

[0037] Finally, a linear layer can be used to map the graph features to the probability distribution of each forwarding node. The calculation process is as follows:

[0038] in, is the weight matrix of the output layer, is the output layer offset, is a normalization function that maps the probability of each node to between 0 and 1 and makes the sum of the probabilities of all nodes equal to 1.

[0039]

[0040] Optionally, during training, for each packet successfully transmitted from the source node to the destination node, a cross entropy loss function can be used to calculate the difference between the model predicted node and the actual forwarding node (i.e., the forwarding node label):

[0041] in, is the number of node categories, , when the node For the actual forwarding node ,otherwise is 0, Nodes representing model predictions In some implementations, the Adam algorithm can be used to update the model parameters. The goal of the model is to minimize .

[0042] After obtaining the node feature input and the adjacency matrix, the node feature input and the adjacency matrix can be used together as the input of the trained graph neural network model to obtain the probability of each node corresponding to the target data packet, and select the forwarding node with the highest corresponding probability according to the probability of each node corresponding to the target data packet to forward the target data packet, thereby increasing the probability of successful transmission of the target data packet.

[0043] It should be noted that in the embodiment of the present invention, the target data packet may be forwarded after determining the probability that each node corresponds to the target data packet, or the data packet may be forwarded after determining the probability that each node corresponds to all data packets in a certain service.

[0044] The dynamic routing method for a delay tolerant network provided by the present invention trains a graph neural network model based on the global information of a sample delay tolerant network and sample data packet information with a forwarding node label, then concatenates the node feature information in the target delay tolerant network with the target data packet information to obtain a node feature input, and constructs an adjacency matrix of the target delay tolerant network. The node feature input and the adjacency matrix are used together as inputs of the trained graph neural network model to obtain the probability that each node corresponds to the target data packet, and then the target data packet is forwarded according to the probability that each node corresponds to the target data packet. In this way, when the network structure of the delay tolerant network changes dynamically, the optimal forwarding node can be selected in the process of data packet forwarding, thereby improving the success rate of the routing method.

[0045] According to a dynamic routing method for a delay tolerant network provided by the present invention, the node feature information may include: (1) The moving speed information of each node. Each node in a delay-tolerant network may be in constant motion, so the node feature information can include the moving speed information of each node, thereby improving the success rate of the routing method.

[0046] (2) Data transmission rate information of each node. The data transmission rates of each node in a delay-tolerant network are not consistent. Therefore, the node characteristic information can include the data transmission rate information of each node, so as to determine the node with lower transmission delay by considering the data packet transmission delay.

[0047] (3) Buffer occupancy information of each node. The buffer occupancy of each node in a delay-tolerant network is not consistent. Therefore, the node feature information can include the buffer occupancy information of each node, so as to determine the node with more sufficient buffer resources by considering the buffer occupancy.

[0048] (4) Location information of each node. Each node in a delay-tolerant network may be in constant motion, so the node feature information can include the location information of each node, thereby improving the success rate of the routing method.

[0049] According to a dynamic routing method for a delay-tolerant network provided by the present invention, the target data packet information may include: (1) Source node identification information of the target data packet; (2) Destination node identification information of the target data packet; (3) The size information of the target data packet; (4) The remaining time to live (TTL) information of the target data packet.

[0050] According to a dynamic routing method of a delay tolerant network provided by the present invention, a target data packet is forwarded based on the probability that each node corresponds to the target data packet, specifically comprising: Based on the probability of each node corresponding to the target data packet, determine the probability of the nodes in the neighbor set of the current node corresponding to the target data packet; The probability of the nodes in the current node's neighbor set corresponding to the target data packet is ranked in descending priority order, and the target data packet is forwarded.

[0051] Specifically, in the process of forwarding the target data packet based on the probability of each node corresponding to the target data packet, the neighbor set of the current node can be determined first. It can be understood that the target data packet can only be forwarded from the current node to the nodes in the neighbor set of the current node.

[0052] Therefore, we can first determine the probability that the nodes in the current node neighbor set correspond to the target data packet, then prioritize the probabilities of the nodes in the current node neighbor set corresponding to the target data packet from large to small, and finally forward the target data packet according to the priority order.

[0053] Optionally, the node feature information in the target delay-tolerant network is obtained based on feature information of the current node and a node feature set stored by the current node.

[0054] Specifically, in a delay-tolerant network, each node will record its own information and use a timestamp to indicate the time when the information was last updated. Each node will also use a set to save the features of other nodes it has encountered. When nodes meet, they will exchange each other's saved node feature sets and update the newer node features according to the timestamp. In an embodiment of the present invention, the node feature information in the target delay-tolerant network can be obtained based on the feature information of the current node and the node feature set saved by the current node.

[0055] The embodiment of the present invention uses global information containing the characteristics of all nodes in the network during training, while in an actual environment a node can only obtain local information in the network. In this way, as nodes meet each other, each node can have more approximately global node information.

[0056] Optionally, the adjacency matrix in the embodiment of the present invention is constructed by depth-first traversal of each neighboring node connected to the current node.

[0057] The dynamic routing method for a delay tolerant network provided by the present invention is further explained below through embodiments in specific application scenarios.

[0058] The purpose of this embodiment is to propose a dynamic routing method for a delay-tolerant network based on a graph neural network. This embodiment first models the network scenario and the routing problems therein in view of the dynamically changing network structure and limited resources in the delay-tolerant network; then designs a dynamic decision routing method based on a graph neural network, which takes the node characteristics and network structure at the current moment as input, and dynamically generates the optimal forwarding node for each data packet by comprehensively considering resource constraints and network structure, thereby improving the success rate of the routing method and reducing latency. The specific scheme of this embodiment is as follows: 1. Network model

[0059] A delay-tolerant network can be represented as a dynamic graph , represents the set of nodes in the network, is a set of time-related edges, indicating that The edge that exists in the network at any time. Since the nodes in the network are constantly moving, when two nodes move within the communication range, the two nodes can transmit data, and at this time there is an edge in the graph.

[0060] 1.1 Node Model

[0061] Assuming that nodes move in a random manner and meet other nodes, the meeting time between nodes can be regarded as a random variable that conforms to a certain probability distribution. This probability distribution may be related to the actual scenario. Here, it is assumed that the nodes and nodes The probability of encountering is .

[0062] 1.2 Business Flow Model

[0063] Assume that there are several services on each node in the network, and each service has The data packets of each service are of the same size and have a fixed generation interval. If all data packets of a service are in If the packet reaches the destination within 10 seconds, it is considered to be sent successfully.

[0064] Represents a service in the network. Indicates business The generated Assuming the service Data packet collection , each data packet has the following attributes: , represents the source node; , represents the destination node; , indicating the packet size; , represents the survival time; , indicating the number of data packet copies.

[0065] 1.3 Routing Model

[0066] The nodes transmit data packets in a hop-by-hop manner. Assume that the routing strategy adopted by the node is expressed as a function , given the current node , the data packet needs to be transmitted and the current moment , The output is the optimal next hop node .

[0067] 1.4 Resource Constraints

[0068] This embodiment considers the two most common constraints in DTN network scenarios: limited buffers of nodes. This embodiment assumes that nodes The buffer size is . And the bandwidth during the node contact is limited. This embodiment assumes that the node and nodes The data transmission rate during the encounter is denoted as .

[0069] 2. Optimize the model

[0070] 2.1. Packet transmission probability

[0071] For Business For each data packet, there are multiple possible transmission paths from the source node to the destination node. The probability of successful transmission on each path is the product of the probability of encountering all nodes on the path. The probability of successful transmission of a data packet is the sum of the probabilities of successful transmission on all possible paths.

[0072] For data packets , a transmission path is recorded as , , , the nodes on the path are transmitted in order, and the next hop nodes are all calculated by the current node through the routing strategy, that is, , the transmission probability on this path is:

[0073] Data Pack The probability of successful transmission is:

[0074] in From the data packet The set of all possible paths from the source node to the destination node.

[0075] business The probability of successful transmission can be expressed as the product of the probabilities that all packets of the service are successfully transmitted:

[0076] 2.2 Data packet transmission delay

[0077] For a data packet ,The delay from the source node to the destination node can be divided into two parts, one part is the cache time in the ,node during the node encounter interval, and the other part is the data transmission ,time when the nodes meet.

[0078] Similarly, there are multiple possible transmission paths from the source node to the destination node. One transmission path is recorded as , the cache time on this path is recorded as the sum of the contact intervals of each pair of adjacent nodes:

[0079] The encounter interval is defined as the inverse of the encounter probability:

[0080] The transmission time is recorded as the sum of the transmission time of each encounter:

[0081] The total transmission delay on this path is:

[0082] The expected delay of a data packet is the expected sum of the delays on all possible paths:

[0083] The service delay can be expressed as the maximum delay of all data packets of the service:

[0084] 2.3 Optimization issues

[0085] The optimization problem of this embodiment can be expressed as the following formula:

[0086] The constraints are:

[0087] The optimization goal of this embodiment is to maximize the delivery rate of all services in the network; constraint (1) indicates that the transmission delay of each data packet of the service should be less than the TTL of the data packet; constraint (2) indicates that the sum of the data packet sizes in the network cannot exceed the size of the buffer; constraint (3) indicates that the number of copies of each data packet in the network should be greater than 0 and less than or equal to the number of nodes in the network, that is, for the same data packet, each node can have at most one copy of the data packet.

[0088] 3. Designed routing method

[0089] 3.1 Model Structure

[0090] The model input includes two parts: node features and adjacency matrix. , the figure has nodes, then , for each node , there is one The node features of the entire graph can be expressed as a matrix ,in Is a node The eigenvector of . The adjacency matrix , where if there is an edge from node Connect to a Node but ,otherwise Specifically, the features designed for each node include the node's moving speed, data transmission rate, buffer occupancy, node location information (including x and y), and the packet information includes the source node id, destination node id, packet size, and remaining time to live (TTL). The packet information is added to the features of each node. The adjacency matrix is ​​generated by the current connection status of the network.

[0091] In the model, this embodiment first uses multiple GCN layers to update the feature information of the node, and finally obtains the feature representation of the node. The calculation process of each GCN layer is as follows:

[0092] in The adjacency matrix of self-loops is added to each node. is the identity matrix. and is the trainable weight matrix of each layer. is an activation function, such as . is the activation matrix of layer l, .

[0093] Afterwards, this embodiment uses a global average pooling layer to aggregate node features into graph-level features. The calculation process is as follows:

[0094] in, The output node of the last GCN layer The feature representation of .

[0095] Finally, this embodiment uses a linear layer to map the graph features to the probability distribution of each forwarding node. The calculation process is as follows:

[0096] in, is the weight matrix of the output layer, is the output layer offset, is a normalization function that maps the probability of each node to between 0 and 1 and makes the sum of the probabilities of all nodes equal to 1.

[0097]

[0098] During the training process, for each packet successfully transmitted from the source node to the destination node, the cross entropy loss function is used to calculate the difference between the model's predicted node and the actual forwarding node:

[0099] in, is the number of node categories, , when the node For the actual forwarding node ,otherwise is 0, Nodes representing model predictions The Adam algorithm is used to update the model parameters. The goal of the model is to minimize .

[0100] 3.2 Routing Method

[0101] Figure 2 The flowchart of the routing method provided by the present invention is as follows: Figure 2As shown, this embodiment first collects the feature information of the nodes and the connection information between the nodes as the input of the model. Each node in the network will record its own information, as well as a timestamp indicating the time when the information was last updated. In addition, the node will use a set to save the features of other nodes it has encountered. When the nodes meet, they will exchange the node feature sets saved by each other, and update the newer node features according to the timestamp. Because this embodiment uses global information containing all node features in the network during training, and in the actual environment a node can only obtain local information in the network, in this way, as the nodes meet, each node can have more approximately global node information.

[0102] When forwarding a data packet, the node feature input of the model is first constructed based on the saved feature set. For the node features that are not saved by the current node, this embodiment sets them to 0, and then splices them with the data packet information to obtain the final node feature input. Then this embodiment constructs an adjacency matrix by depth-first traversal of each neighbor node connected to the current node, and uses them together as the input of the model to obtain the probability of forwarding to each node. After that, this embodiment excludes those nodes that are not in the neighbor set of the current node, and sorts the remaining nodes from large to small according to probability to obtain the priority of forwarding to each node. Finally, the data packet is forwarded according to the priority order.

[0103] The dynamic routing device of a delay tolerant network provided by the present invention is described below. The dynamic routing device of the delay tolerant network described below and the dynamic routing method of the delay tolerant network described above can be referred to each other.

[0104] Figure 3 A structural diagram of a dynamic routing device for a delay tolerant network provided by the present invention is shown in FIG. Figure 3 As shown, the device includes the following modules: The acquisition module 300 is used to concatenate the node feature information in the target delay tolerant network with the target data packet information to obtain the node feature input, and to construct an adjacency matrix of the target delay tolerant network; An input module 310 is used to use the node feature input and the adjacency matrix as inputs to the trained graph neural network model to obtain the probability that each node corresponds to the target data packet; A forwarding module 320, configured to forward the target data packet based on the probability that each node corresponds to the target data packet; Among them, the graph neural network model is trained based on the global information of the sample delay-tolerant network and the sample data packet information with forwarding node labels.

[0105] According to a dynamic routing device for a delay tolerant network provided by the present invention, the node characteristic information includes: The moving speed information of each node; Data transmission rate information of each node; Buffer occupancy information of each node; The location information of each node.

[0106] According to a dynamic routing device for a delay tolerant network provided by the present invention, the target data packet information includes: Source node identification information of the target data packet; Destination node identification information of the target data packet; The size information of the target data packet; The remaining lifetime information of the target data packet.

[0107] According to a dynamic routing device of a delay tolerant network provided by the present invention, a target data packet is forwarded based on the probability that each node corresponds to the target data packet, specifically comprising: Based on the probability of each node corresponding to the target data packet, determine the probability of the nodes in the neighbor set of the current node corresponding to the target data packet; The probability of the nodes in the current node's neighbor set corresponding to the target data packet is ranked in descending priority order, and the target data packet is forwarded.

[0108] According to a dynamic routing device for a delay tolerant network provided by the present invention, node feature information in a target delay tolerant network is obtained based on feature information of a current node and a node feature set stored by the current node; The adjacency matrix is ​​constructed by depth-first traversing each neighbor node connected to the current node.

[0109] According to a dynamic routing device for a delay-tolerant network provided by the present invention, the graph neural network model consists of multiple graph convolutional network GCN layers, a global average pooling layer and a linear layer.

[0110] Figure 4 is a schematic diagram of the structure of the electronic device provided by the present invention, such as Figure 4 As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430 and a communication bus 440, wherein the processor 410, the communications interface 420 and the memory 430 communicate with each other through the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute the dynamic routing method of the delay-tolerant network provided by the above methods, and the method includes: splicing node feature information in the target delay tolerant network with target data packet information to obtain node feature input, and constructing an adjacency matrix of the target delay tolerant network; The node feature input and the adjacency matrix are used as the input of the trained graph neural network model to obtain the probability that each node corresponds to the target data packet; Forward the target data packet based on the probability that each node corresponds to the target data packet; Among them, the graph neural network model is trained based on the global information of the sample delay-tolerant network and the sample data packet information with forwarding node labels.

[0111] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0112] On the other hand, the present invention further provides a computer program product, the computer program product comprising a computer program, the computer program can be stored on a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can execute the dynamic routing method of the delay-tolerant network provided by the above methods, the method comprising: splicing node feature information in the target delay tolerant network with target data packet information to obtain node feature input, and constructing an adjacency matrix of the target delay tolerant network; The node feature input and the adjacency matrix are used as the input of the trained graph neural network model to obtain the probability that each node corresponds to the target data packet; Forward the target data packet based on the probability that each node corresponds to the target data packet; Among them, the graph neural network model is trained based on the global information of the sample delay-tolerant network and the sample data packet information with forwarding node labels.

[0113] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for performing the dynamic routing method of a delay-tolerant network provided by the above methods is implemented, and the method comprises: splicing node feature information in the target delay tolerant network with target data packet information to obtain node feature input, and constructing an adjacency matrix of the target delay tolerant network; The node feature input and the adjacency matrix are used as the input of the trained graph neural network model to obtain the probability that each node corresponds to the target data packet; Forward the target data packet based on the probability that each node corresponds to the target data packet; Among them, the graph neural network model is trained based on the global information of the sample delay-tolerant network and the sample data packet information with forwarding node labels.

[0114] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0115] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A dynamic routing method for a delay tolerant network, characterized in that: include: splicing node feature information in a target delay tolerant network with target data packet information to obtain node feature input, and constructing an adjacency matrix of the target delay tolerant network; The node feature input and the adjacency matrix are used together as inputs of the trained graph neural network model to obtain the probability that each node corresponds to the target data packet; Forwarding the target data packet based on the probability that each node corresponds to the target data packet; The graph neural network model is trained based on the global information of the sample delay-tolerant network and the sample data packet information with forwarding node labels.

2. The dynamic routing method for a delay tolerant network according to claim 1, characterized in that: The node feature information includes: The moving speed information of each node; Data transmission rate information of each node; Buffer occupancy information of each node; The location information of each node.

3. The dynamic routing method for a delay tolerant network according to claim 1, characterized in that: The target data packet information includes: Source node identification information of the target data packet; Destination node identification information of the target data packet; The size information of the target data packet; The remaining lifetime information of the target data packet.

4. The dynamic routing method for a delay tolerant network according to any one of claims 1 to 3, characterized in that: The forwarding of the target data packet based on the probability that each node corresponds to the target data packet specifically includes: Based on the probability that each node corresponds to the target data packet, determining the probability that the nodes in the neighbor set of the current node correspond to the target data packet; The probability that the nodes in the neighbor set of the current node correspond to the target data packet is arranged in descending priority order, and the target data packet is forwarded.

5. The dynamic routing method for a delay tolerant network according to any one of claims 1 to 3, characterized in that: The node feature information in the target delay-tolerant network is obtained based on the feature information of the current node and the node feature set stored by the current node; The adjacency matrix is ​​constructed by depth-first traversing each neighboring node connected to the current node.

6. The dynamic routing method for a delay tolerant network according to claim 1, characterized in that: The graph neural network model consists of multiple graph convolutional network (GCN) layers, a global average pooling layer and a linear layer.

7. A dynamic routing device for a delay tolerant network, characterized in that: include: An acquisition module, used for concatenating node feature information in a target delay tolerant network with target data packet information to obtain node feature input, and constructing an adjacency matrix of the target delay tolerant network; An input module, used to use the node feature input and the adjacency matrix as inputs of the trained graph neural network model to obtain the probability that each node corresponds to the target data packet; A forwarding module, configured to forward the target data packet based on the probability that each node corresponds to the target data packet; The graph neural network model is trained based on the global information of the sample delay-tolerant network and the sample data packet information with forwarding node labels.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the dynamic routing method for a delay tolerant network according to any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the dynamic routing method for a delay tolerant network according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the dynamic routing method for a delay tolerant network according to any one of claims 1 to 6 is implemented.

Images