Routing Method and System for Flying Ad Hoc Network Based on Graph Neural Network and Mobile Sensing

Abstract

The present invention discloses a routing method and system for flying ad hoc networks based on graph neural networks and mobile sensing, belonging to the field of wireless communication technology. The method includes: drones periodically upload their own node information to a high-altitude computing platform using an information collection protocol; the high-altitude computing platform constructs a network status graph using the uploaded node information as the input for training a graph neural network model, and trains the graph neural network model to predict the network status graph at the next moment; when a node makes a routing request, the high-altitude computing platform issues a command to collect all node information to form a network status graph; the network status graph is input into the trained graph neural network model to obtain the network status graph at the next moment, and then the shortest path algorithm is used to select the best route and send it to the drone making the routing request. By predicting the future network status graph through the graph neural network model, the accuracy and intelligence level of routing selection are improved, the packet arrival rate is increased, and the network throughput is optimized.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communication, and particularly relates to a routing method and system for flying ad-hoc networks based on graph neural networks and mobile sensing. Background Art

[0002] Recently, unmanned aerial vehicles (UAVs) or unmanned aircraft have become increasingly popular in military and civilian applications, such as aerial photography, communication, agriculture, and search and rescue. The use of UAVs in these applications requires reliable and fast wireless communication, which has led to the development of flying ad-hoc networks, in which multiple UAVs work together to achieve greater capabilities, such as cost-effectiveness, scalability, survivability, and higher speeds. Flying ad-hoc networks (FANETs) have unique routing challenges because FANETs involve highly dynamic and three-dimensional movement, which poses challenges in maintaining reliable and efficient routing paths as UAVs often change positions and encounter different network conditions.

[0003] In flying ad-hoc networks (FANETs), most existing FANET routing protocols are designed based on mobility or topological information, ignoring the traffic distribution in the network. Therefore, nodes located in severely congested areas may cause significant end-to-end delays, untimely path adjustment, and packet loss. However, although existing prediction-based FANET routing protocols perform congestion control, they often do not well consider the topological relationship between nodes, resulting in inconsistent prediction results with the actual situation.

[0004] Therefore, when facing congestion control, the present invention proposes a routing method for flying ad-hoc networks based on graph neural networks and mobile sensing to solve the above technical problems. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the purpose of the present invention is to provide a routing method and system for flying ad-hoc networks based on graph neural networks and mobile sensing, which solves the problems in the prior art.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] A routing method for flying ad-hoc networks based on graph neural networks and mobile sensing includes the following steps:

[0008] UAVs periodically upload their own node information to the high-altitude computing platform using an information collection protocol;

[0009] The high-altitude computing platform uses the uploaded node information to construct a network condition graph as the input for training the graph neural network model, trains the graph neural network model, and updates the model parameters for predicting the network condition graph at the next moment;

[0010] When a node makes a routing request, the high-altitude computing platform issues a command to collect information of all nodes and form a network status map. The network status map is input into the trained graph neural network model to obtain the network status map of the next moment, and then the shortest path algorithm is used to select the best route and send it to the drone making the routing request.

[0011] Furthermore, the flying ad-hoc network includes: users, drones, and a high-altitude computing platform;

[0012] Each user is associated with the drone with the strongest received signal strength. When a user has a communication requirement, the user will upload the data packet to the associated drone. Finding the route of the user's data packet means finding the route of the associated drone, and the drone is regarded as a node;

[0013] After receiving the data packet, the drone sends a routing request to the high-altitude computing platform. The high-altitude computing platform sends an information collection command to the drone. After collecting the drone information, it preprocesses to form a network status map. The network status map is input into the trained graph neural network model to obtain the network status map of the next moment, and then the shortest path algorithm is used to select the best route and send it to the drone making the routing request. After receiving the best route, the drone forwards the data packet sent by the user.

[0014] Furthermore, the steps for the drone to upload its own node information to the high-altitude computing platform are as follows:

[0015] All nodes regularly generate and send NLIP messages;

[0016] After receiving the NLIP message, the neighbor node checks the validity of the message and determines whether it is a duplicate message. If it is not a duplicate message, it forwards it to other neighbor nodes and increments according to the hop count field in the message to indicate an increase in the hop count of the path;

[0017] All nodes broadcast the NLIP message until the high-altitude computing platform collects complete information of all nodes in the flying ad-hoc network. The high-altitude computing platform selects multi-point relay nodes based on the collected network information;

[0018] Each node regularly sends source announcement data to the multi-point relay nodes according to the information collection protocol and updates the node routing table according to the network status. If a neighbor node loses connection or does not receive the source announcement data, the neighbor node is removed from the node routing table.

[0019] Furthermore, the graph neural network model includes multiple spatio-temporal convolutional blocks, and each spatio-temporal convolutional block includes two gated sequence convolutional layers and a spatial graph convolutional layer located in the middle of the two gated sequence convolutional layers.

[0020] Furthermore, the training process of the graph neural network model is as follows:

[0021] Preprocess the node features through MinMaxScaler normalization and construct the adjacency matrix according to the FANET topology structure;

[0022] Initialize the weights and parameters of the graph neural network model. Use the preprocessed node features and adjacency matrix to train the graph neural network model for neural network training. During the training process, define the loss function as the mean squared error, and add L1 penalty to control the complexity of the graph neural network model. At the same time, determine the adaptive learning rate using the Adam optimizer.

[0023] Furthermore, the specific steps to select the best route and send it to the drones with routing requests are as follows:

[0024] The service connection node requests transmission, and the node reports the service type and routing request to the high-altitude computing platform;

[0025] The high-altitude computing platform issues commands through relay nodes to let all nodes upload the current node parameter information, construct the node feature matrix of FANET, and then weight the connection matrix of the FANET according to the link maintenance time collected in the FANET network, which is expressed as the adjacency matrix to obtain the network status graph; then input the network status graph into the trained graph neural network model for prediction to obtain the network status graph at the next moment;

[0026] The high-altitude computing platform obtains the predicted network status graph at the next moment, normalizes each metric factor of the node, multiplies it by the influence factor of each metric for weighting to obtain the node cost, takes the cost of the node with the larger cost value on both sides of the link as the link cost, obtains the network cost graph, and uses the shortest algorithm to solve for the minimum cost, that is, the best route;

[0027] The high-altitude computing platform issues the best routing table to the routing request node to update the node routing table.

[0028] A routing system for flying ad-hoc networks based on graph neural networks and mobile sensing, including:

[0029] Routing discovery module: The drones periodically upload their own node information to the high-altitude computing platform using the information collection protocol;

[0030] Model training module: The high-altitude computing platform uses the uploaded node information to construct the network status graph as the input for training the graph neural network model, trains the graph neural network model, and updates the model parameters for predicting the network status graph at the next moment;

[0031] And, a routing optimization module: When a node makes a routing request, the high-altitude computing platform issues a command to collect information of all nodes to form a network status graph. The network status graph is input into the trained graph neural network model to obtain the network status graph at the next moment, and then the shortest path algorithm is used to select the best route and send it to the drone making the routing request.

[0032] A computer storage medium stores a readable program that, when running, can execute the above-mentioned routing method for flying ad-hoc networks based on graph neural networks and mobile sensing.

[0033] An electronic device includes: a processor, a memory, a communication interface, and a communication bus. The processor, the memory, and the communication interface complete communication with each other through the communication bus;

[0034] The memory is used to store at least one executable instruction, and the executable instruction causes the processor to execute the operations corresponding to the above-mentioned routing method for flying ad-hoc networks based on graph neural networks and mobile sensing.

[0035] A computer program product includes computer instructions that direct a computing device to execute the operations corresponding to the above-mentioned routing method for flying ad-hoc networks based on graph neural networks and mobile sensing.

[0036] Advantages of the present invention:

[0037] 1. The present invention defines a new node and link information flow data packet for network information transmission, routing discovery, and network formation. When the network topology changes, the data packet structure and content are dynamically adjusted, and the information transmission path can be adaptively optimized according to the real-time network status, thereby significantly improving the efficiency and reliability of information transmission.

[0038] 2. Based on the traditional network routing maintenance, the present invention introduces mobile sensing and congestion control based on graph neural networks. By integrating network performance evaluation parameters and combining the learning ability of graph neural networks, the system can predict the network status graph at the next moment to build a more comprehensive and intelligent network performance evaluation model, improving the adaptability and intelligent level of the network.

[0039] 3. By considering the remaining energy of nodes, the average link maintenance time of nodes, and the remaining buffer size of nodes as metrics for routing decisions, the present invention constructs a more comprehensive and accurate network performance evaluation model, providing an important basis for improving the reliability, efficiency, and intelligent management of the network.

[0040] 4. The present invention uniformly allocates routes by means of a high-altitude computing platform, realizing the function of routing on demand according to service types. Meanwhile, it also operates at the data link layer, transmitting routing information in the form of Ethernet frames, reducing the packet processing delay. Compared with the routing strategy based on the shortest path in the prior art, it considers multiple channel parameters and adaptively selects routes. Description of the Drawings

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0042] Figure 1 It is a schematic diagram of the flight ad hoc network structure of the present invention;

[0043] Figure 2 It is a schematic diagram of the protocol routing discovery process of the present invention;

[0044] Figure 3 It is a flowchart of the flight ad hoc network routing method based on graph neural network and mobile perception of the present invention;

[0045] Figure 4 It is a curve graph of the system packet transmission success rate and the number of unmanned aerial vehicles realized by the algorithm proposed by the present invention under the same packet arrival rate and rate of unmanned aerial vehicles;

[0046] Figure 5 It is a curve graph of the system packet transmission success rate and the packet arrival rate realized by the algorithm proposed by the present invention under the same number and rate of unmanned aerial vehicles;

[0047] Figure 6 It is a curve graph of the system packet transmission success rate and the rate of unmanned aerial vehicles realized by the proposed algorithm under the same number and packet arrival rate of unmanned aerial vehicles. Detailed Embodiments

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0049] Embodiment 1

[0050] As Figure 1 shown, the present invention proposes a flight ad hoc network (FANET) based on graph neural network and mobile perception, including: users, unmanned aerial vehicles, and high-altitude computing platforms;

[0051] Each user is associated with the drone with the strongest received signal strength (RSS). When a user has a communication need, the data packet will be uploaded to the associated communication drone. Since the user is associated with the drone, finding the route of the user's data packet is equivalent to finding the route of the drone. Therefore, the drone is directly regarded as a node to more conveniently find the data packet route.

[0052] After receiving the data packet, the communication drone sends a routing request to the high-altitude computing platform, and the high-altitude computing platform sends an information collection command to the drone.

[0053] In order to better implement the graph neural network model, the high-altitude computing platform defines a node and link information flow data packet (NLIP) that shares node / link information throughout the FANET. NLIP is defined to contain node and link information. Table 1 shows the structure of NLIP:

[0054] Table 1 Node and Link Information Flow Data Packet Format

[0055]

[0056] It can be packed into a UDP data packet. The check header field indicates that this is a routing message. The count field is used to avoid duplicate message transmission. It also helps to update the local mapping information stored in the buffer. The remaining fields contain two tables, which contain node and link information of the neighborhood.

[0057] The node information table contains the IP, buffer status, remaining energy, location information, and clustering information of each node. The link information table contains the start and end nodes of each link, the received signal strength, and the estimated duration.

[0058] After receiving the information collection command, the drone uploads the NLIP containing its own specific node and link information to the high-altitude computing platform.

[0059] After receiving the NLIP, the high-altitude computing platform collects the specific node and link information of the FANET, preprocesses the node features and adjacency matrix to form a network status graph; the network status graph is input into the trained graph neural network model to obtain the network status graph at the next moment, and then the shortest path algorithm is used to select the best route and send it down to the drone that sent the routing request. After receiving the best route, the drone forwards the data packet sent by the user.

[0060] Embodiment 2

[0061] As Figure 3 shown, the routing method for flying ad hoc networks based on graph neural networks and mobile sensing includes the following steps:

[0062] S1, The drone periodically uploads its own node information to the high-altitude computing platform using an information collection protocol, i.e., the routing discovery process;

[0063] The routing discovery process in this flying ad-hoc network is as Figure 2 shown and includes the following steps:

[0064] S11, Define a new node and link information flow data packet (NLIP), which contains information about the node itself and the links with neighbor nodes, and is used for network information transfer, routing discovery, and network formation;

[0065] S12, All nodes periodically generate and send NLIP (node and link information flow data packet) messages;

[0066] The NLIP message contains key information such as the unique identifier, sequence number, hop count, and time-to-live (TTL) of the sending node; the purpose of the NLIP message is to let other nodes in the network know the existence of the source node and provide a basis for the routing discovery process. The NLIP message is sent to surrounding nodes by broadcasting, and the propagation range is controlled by the TTL field. When the TTL is reduced to 0, the NLIP message will be discarded, limiting its maximum propagation range and avoiding network overload.

[0067] S13, After receiving the NLIP message, neighbor nodes check the validity of the message to determine whether it is a duplicate message. If it is not a duplicate message, it is forwarded to other neighbor nodes, and the hop count field in the message is incremented to indicate an increase in the hop count of the path;

[0068] When checking the validity of the message, especially by using the sequence number of the NLIP to determine whether it is a duplicate message; if the message is new (i.e., the sequence number is received for the first time), the node forwards it to other neighbor nodes and appropriately increments it according to the hop count field in the message to indicate an increase in the hop count of the path.

[0069] S14, All nodes broadcast NLIP messages until the high-altitude computing platform collects complete information about all nodes in the flying ad-hoc network; the high-altitude computing platform selects multi-point relay nodes based on the collected network information to reduce subsequent broadcast control overhead;

[0070] S15, Each node periodically sends source announcement data to the multi-point relay nodes according to the information collection protocol, and updates the node routing table according to the network status. If a neighbor node loses connection or does not receive the source announcement data, the neighbor node is removed from the node routing table;

[0071] Each node gradually constructs and updates its own routing table based on the NLIP messages received from neighboring nodes. This incremental forwarding method of messages enables NLIP to traverse the entire network, and each node only needs to know its directly connected neighboring nodes without having to understand the entire network topology.

[0072] S2. The high-altitude computing platform uses the uploaded node information to construct a network status graph as the input for training the graph neural network model, trains the graph neural network model, and updates the model parameters for predicting the network status graph at the next moment;

[0073] Among them, the graph neural network model includes multiple spatio-temporal convolutional blocks. Each spatio-temporal convolutional block (ST-Conv block) forms a "sandwich" structure, which includes two gated sequence convolutional layers and a spatial graph convolutional layer in the middle;

[0074] The middle spatial graph convolutional layer is a bridge connecting the two gated sequence convolutional layers, which can achieve fast propagation from graph convolution to spatial state through temporal convolution. The "sandwich" structure also helps the network to fully apply the bottleneck strategy, downscale and upscale the channel C through the graph convolutional layer to achieve scale compression and feature compression. In addition, layer normalization is used within each ST-Conv block to prevent overfitting. The input and output of the ST-Conv block are both three-dimensional tensors.

[0075] The graph neural network model includes two spatio-temporal convolutional blocks (ST-Conv blocks) and a fully connected output layer. Each ST-Conv block contains two temporal gated convolutional layers and a spatial graph convolutional layer in the middle. Residual connections and bottleneck strategies are applied inside each block. The input is uniformly processed by the ST-Conv blocks to coherently explore spatial and temporal dependencies. The comprehensive features are integrated through an output layer to generate the final prediction result .

[0076] It takes multiple network status graphs at time instants (t - T, t - T + 1,..., t - 1, t) as input. In this embodiment, sampling is performed at intervals of 1 s, and the output is the network status graph predicted at time (t + 1). Therefore, the graph neural network model can capture the features of high-dimensional data;

[0077] The specific steps of S2 are as follows:

[0078] S21. Receive the information of nodes and links through the high-altitude computing platform and construct a network status graph;

[0079] S22. The high-altitude computing platform processes the constructed network condition map, performs spatio-temporal modeling using a graph neural network, models the spatio-temporal correlation in the FANET through multiple spatio-temporal convolutional blocks, extracts the evolution pattern from a series of network condition maps, and then predicts the network condition map at the next moment;

[0080] The specific construction process of the training process of the graph neural network model in S22 includes:

[0081] S221. Preprocess the node features (such as network traffic, speed, etc.) through MinMaxScaler normalization, and construct an adjacency matrix according to the FANET network topology structure;

[0082] S222. Initialize the weights and parameters of the graph neural network model, use the preprocessed node features and adjacency matrix to train the graph neural network model for neural network training. During the training process, define the loss function as the mean squared error, and add L1 penalty to control the complexity of the graph neural network model and reduce overfitting. At the same time, the adaptive learning rate is determined using the Adam optimizer, and iterative training is performed until the specified number of iterations is reached or the specified threshold error is satisfied;

[0083] S223. After training, obtain more accurate graph neural network model parameters to predict the network condition map at the next moment.

[0084] The output of the graph neural network model is usually a prediction result. In the network prediction task, the output of the graph neural network model is usually the traffic state (such as traffic flow, speed, etc.) at a future time step or multiple time steps, expressed as , where T ′ is the future time step, C ′ is the predicted feature.

[0085] S3. When a node makes a routing request, the high-altitude computing platform issues a command to collect all node information to form a network condition map; the network condition map is input into the trained graph neural network model to obtain the network condition map at the next moment, and then the shortest path algorithm is used to select the best route and send it to the drone making the routing request;

[0086] The specific steps for selecting the best route and sending it to the drone making the routing request in S3 are as follows:

[0087] S31. Connect the service nodes and request transmission. The nodes report the service type and routing request to the high-altitude computing platform;

[0088] S32. The high-altitude computing platform issues commands through relay nodes to make all nodes upload the current node parameter information, including the remaining energy of the node, buffer size, three-dimensional position information, and flight speed, to construct the node feature matrix of the FANET. Then, according to the link maintenance time of each link in the collected FANET network, the connection matrix of the FANET is weighted, expressed as an adjacency matrix, to obtain the network status graph. After that, the network status graph is input into the trained graph neural network model for prediction to obtain the network status graph at the next moment.

[0089] The steps to obtain the network status graph at the next moment include:

[0090] S321. The high-altitude computing platform receives the NLIP information packets of the nodes through relay nodes to obtain the node parameter information of the FANET network.

[0091] S322. Denote the three-dimensional position information of the UAV as and denote the speed of the UAV as . Denote the remaining energy of the UAV as and denote the buffer size of the UAV as . Form a feature vector, written as . Weight the connection matrix of the FANET according to the link duration (LET) of each link, expressed as an adjacency matrix ; ; ;

[0092] S323. The input data of the graph neural network model is usually a spatio-temporal graph, which provides the dynamic information of each node at different time steps, expressed as node features , where N is the number of FANET nodes; T is the time step, which can be customized; C is the number of features;

[0093] S324. Input the node features and the adjacency matrix

[0094] into the trained graph neural network model for prediction to obtain the network status graph at the next moment.

[0095] S33. The high-altitude computing platform obtains the predicted network status graph at the next moment, normalizes each metric factor of the nodes, multiplies them by the influence factors of each metric for weighting to obtain the node cost, takes the cost of the node with the larger cost value on both sides of the link as the link cost, obtains the network cost graph, and uses the shortest algorithm to solve for the minimum cost, that is, the optimal route.

[0095] Construct a spatio-temporal graph through a high-altitude computing platform, train a graph neural network model, and dynamically select the best path by jointly predicting network conditions to avoid network congestion and achieve traffic load balancing. At the same time, the prediction results help to identify peak traffic periods in advance, adjust routing strategies or bandwidth allocation, and prevent network congestion and latency. Based on joint prediction and decision optimization, the intelligence and reliability of routing selection can be improved, thereby enhancing the overall network stability and user experience. The steps to solve the optimal route are as follows:

[0096] S331, the metric factors are taken as the remaining energy, LET, and the available size of the buffer, and they are normalized and represented as follows:

[0097]

[0098]

[0099]

[0100] Among them, represents the energy cost value of the drone at time after normalization, represents the link duration cost value of the drone at time after normalization, represents the buffer cost value of the drone at time after normalization, , , respectively represent the initial energy of the drone, the maintenance time of the drone system, and the buffer size, represents the remaining energy value of the drone at time , represents the remaining average link maintenance time of the drone at time , represents the remaining buffer size of the drone at time .

[0101] S332, according to the service type, multiply the normalized metric factors of each node by the influence factors of each metric to obtain the node cost, take the cost of the node with the larger cost value on both sides of the link as the link cost, obtain the network cost graph, and use the shortest algorithm to solve for the minimum cost, which is the optimal route.

[0102] The calculation formula for the node cost is:

[0103]

[0104] In the formula, is the node cost, , , are the influence factors of the remaining energy of the node, LET, and the available size of the buffer, respectively.

[0105] The predicted FANET network status diagram is represented as a network cost diagram as follows:

[0106]

[0107]

[0108] Among them, is the directed graph of FANET, is the set containing all UAV nodes, represents the set of links, is the link from node to , is the link cost in this link, represents the set of link costs. In the above equation, the cost has a constant 1, which means that the cost increases each time the path crosses a new hop. This helps to find a short path (i.e., fewer hops). is the matrix storing the predicted UAV node cost values. is the UAV at time 's predicted node cost value.

[0109] The objective function of the path search process can be expressed as:

[0110]

[0111] Among them, is the best path, l is the candidate path, L is the set of candidate paths, M is the set of system UAVs, i is one end node of each link of the candidate path l , j is the other end node of each link of the candidate path l ; According to the service type, change , , 's values, and use the shortest algorithm to solve the above formula to obtain the best route.

[0112] S34, the high-altitude computing platform sends the optimal routing table to the routing request node to update the node routing table;

[0113] When the node is in the routing maintenance stage, each node broadcasts the NLIP information packet through multi-point relay every 1s, which can regularly update the neighbor routing table of the node and ensure the stability of the network.

[0114] Based on a similar inventive concept, an embodiment of the present invention further provides a computer storage medium storing a readable program, which can execute the above-mentioned routing method for flying ad hoc networks based on graph neural networks and mobile sensing when the program runs.

[0115] Based on a similar inventive concept, an embodiment of the present invention provides an electronic device, including: a processor, a memory, a communication interface, and a communication bus, and the processor, the memory, and the communication interface complete mutual communication through the communication bus;

[0116] The memory is used to store at least one executable instruction, and the executable instruction causes the processor to execute the operations corresponding to the above-mentioned routing method for flying ad hoc networks based on graph neural networks and mobile sensing.

[0117] Based on a similar inventive concept, an embodiment of the present invention further provides a computer program product including computer instructions, and the computer instructions instruct a computing device to execute the operations corresponding to the above-mentioned routing method for flying ad hoc networks based on graph neural networks and mobile sensing.

[0118] Embodiment 3

[0119] In this embodiment, a specific embodiment is proposed to elaborate on the solution of the present invention in detail. The specific embodiment is simulated using Python software; the specific parameters are set as shown in Table 2;

[0120] Table 2 Simulation parameters

[0121]

[0122] Figure 4 Shows the performance comparison results of different routing protocols in terms of packet delivery ratio (PDR). From Figure 4It can be seen that the flow success rates of DSDV and GREEDY routing protocols are less than 50% in most cases, which is mainly attributed to their inability to effectively cope with the dynamic changes in network topology caused by the high-speed movement of nodes in 3D space. In contrast, DSDV performs relatively better in terms of PDR and can successfully transmit an average of 40% of the data packets. However, the performances of OPAR and JP protocols are significantly better than other algorithms, and both achieve an average packet transmission success rate of 70%, showing significant advantages. This result indicates that in a dynamically changing 3D network environment, the design of routing protocols needs to fully consider the real-time change characteristics of network topology.

[0123] When analyzing the PDR performance, the simulation experiment set the average PAR (Packet Arrival Rate) to 10 KB / s, the AV (Average Velocity) to 20 m / s, and observed the changes in protocol performance by changing the number of UAVs, as Figure 4 shown. The PDR of the four protocols shows an obvious three-stage change characteristic. This change rule can be explained as follows: in the initial stage, increasing the number of UAVs will improve the network connectivity and the number of available paths, thus increasing the packet delivery rate; but as the number of UAVs continues to increase, the network load gradually increases, resulting in a decrease in the packet arrival rate; finally, it reaches a balanced state. This three-stage change reflects the complex balance relationship among network density, connectivity, and load, and different factors play a dominant role in different stages.

[0124] The reason why the JP protocol can show the best PDR performance is mainly due to its advanced routing decision-making mechanism. By predicting and evaluating the network state, this protocol can effectively avoid selecting relay links with high traffic, high mobility, and low persistence, and comprehensively consider the current network metrics and future possible change trends. This forward-looking routing strategy makes JP superior to other routing protocols in all test scenarios. In contrast, DSDV and GREEDY perform the worst due to the lack of full consideration of the network environment. This result verifies the importance of the need for routing protocols to have prediction and adaptation capabilities in a dynamic network environment.

[0125] From Figure 5It can be seen that under the condition of lower PAR, the performance difference between the JP and OPAR protocols is not significant. However, as PAR increases, the packet arrival rate of all protocols shows a downward trend. It is worth noting that although the packet success rate of both JP and OPAR decreases, the decrease rate of JP is significantly slower. This advantage mainly stems from the fact that the JP protocol considers the remaining capacity of the UAV buffer during routing decision-making and predicts the congestion situation of the next hop, thus effectively avoiding the occurrence of packet congestion. This design enables JP to maintain good performance under high-load conditions.

[0126] Figure 6 It shows the impact of UAV speed on PDR. As the UAV speed increases, the PDR of all protocols shows a downward trend. However, due to considering the link expiration time, the JP and OPAR protocols not only maintain a relatively high PDR in high-speed mobile scenarios, but also the rate of performance decline is slower. This result indicates that in a high-speed mobile network environment, a routing strategy considering link stability can significantly improve protocol performance.

[0127] Embodiment 4

[0128] Based on the routing method for flying ad hoc networks based on graph neural networks and motion sensing proposed in Embodiment 1, in this embodiment, a routing system for flying ad hoc networks based on graph neural networks and motion sensing is proposed, which specifically includes:

[0129] Routing discovery module: The UAV periodically uploads its own node information to the high-altitude computing platform using the information collection protocol;

[0130] Model training module: The high-altitude computing platform constructs a network status graph using the uploaded node information as the input for training the graph neural network model, trains the graph neural network model, and updates the model parameters for predicting the network status graph at the next moment;

[0131] And a routing optimization module: When a node makes a routing request, the high-altitude computing platform issues a command to collect all node information to form a network status graph; the network status graph is input into the trained graph neural network model to obtain the network status graph at the next moment, and then the shortest path algorithm is used to select the best route and send it to the UAV making the routing request;

[0132] The method of the present invention can be implemented in hardware, firmware, or as software or computer code that can be stored in a recording medium (such as a CDROM, RAM, floppy disk, hard disk, or magneto-optical disk), or as computer code that is originally stored in a remote recording medium or a non-transitory machine-readable medium and downloaded through a network and will be stored in a local recording medium, so that the method described herein can be stored in such software processing on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It will be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component (e.g., RAM, ROM, flash memory, etc.) that can store or receive software or computer code, and when the software or computer code is accessed and executed by the computer, the processor, or the hardware, the method described herein is implemented. In addition, when a general-purpose computer accesses the code for implementing the method shown herein, the execution of the code converts the general-purpose computer into a dedicated computer for executing the method shown herein. The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments, and what is described in the above embodiments and the specification is only to illustrate the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements fall within the scope of the present invention claimed.

Claims

1. A routing method for flying ad hoc networks based on graph neural networks and mobile sensing, characterized in that Including the following steps: The drone periodically uploads its own node information to the high-altitude computing platform using an information collection protocol; The high-altitude computing platform uses the uploaded node information to construct a network status map as the input for training the graph neural network model, trains the graph neural network model, and updates the model parameters for predicting the network status map at the next moment; When a node makes a routing request, the high-altitude computing platform issues a command to collect all node information to form a network status map; the network status map is input into the trained graph neural network model to obtain the network status map at the next moment, and then the shortest path algorithm is used to select the best route and send it to the drone making the routing request; The flying ad-hoc network includes: users, drones, and a high-altitude computing platform; Each user is associated with the drone with the strongest received signal strength. When a user has a communication requirement, it will upload the data packet to the associated drone. Finding the route for the user's data packet means finding the route of the associated drone, and the drone is regarded as a node; After receiving the data packet, the drone sends a routing request to the high-altitude computing platform. The high-altitude computing platform sends an information collection command to the drone. After collecting the drone information, it preprocesses to form a network status map; the network status map is input into the trained graph neural network model to obtain the network status map at the next moment, and then the shortest path algorithm is used to select the best route and send it to the drone making the routing request. After receiving the best route, the drone forwards the data packet sent by the user; The steps for the drone to upload its own node information to the high-altitude computing platform are: All nodes regularly generate and send NLIP messages; After receiving the NLIP message, the neighbor node checks the validity of the message to determine whether it is a duplicate message. If it is not a duplicate message, it forwards it to other neighbor nodes and increments according to the hop number field in the message to indicate an increase in the hop count of the path; All nodes broadcast NLIP messages until the high-altitude computing platform collects all node information of the entire flying ad-hoc network. The high-altitude computing platform selects multi-point relay nodes based on the collected network information; Each node regularly sends source announcement data to the multi-point relay nodes according to the information collection protocol, and updates the node routing table according to the network status. If a neighbor node loses connection or does not receive the source announcement data, it removes the neighbor node from the node routing table.

2. The routing method for mobile ad hoc networks based on graph neural networks and mobile sensing according to claim 1, wherein The graph neural network model includes multiple spatio-temporal convolutional blocks, and each spatio-temporal convolutional block includes two gated sequence convolutional layers and a spatial graph convolutional layer in the middle of the two gated sequence convolutional layers.

3. The routing method for mobile ad hoc network based on graph neural network and mobile sensing according to claim 2, wherein The training process of the graph neural network model is: Preprocess the node features through MinMaxScaler normalization and construct an adjacency matrix according to the FANET topology structure; Initialize the weights and parameters of the graph neural network model, use the preprocessed node features and adjacency matrix to train the graph neural network model for neural network training. During the training process, define the loss function as the mean square error, and add an L1 penalty to control the complexity of the graph neural network model. At the same time, the adaptive learning rate is determined using the Adam optimizer.

4. The routing method for mobile ad hoc networks based on graph neural network and mobile sensing according to claim 1, wherein The specific steps for selecting the best route and sending it to the drone making the routing request are: The service connection node requests transmission and reports the service type and routing request to the high-altitude computing platform; The high-altitude computing platform issues commands through relay nodes to let all nodes upload the current node parameter information, constructs the node feature matrix of the FANET, and then weights the connection matrix of the FANET according to the link maintenance time of each link in the collected FANET network, expressed as an adjacency matrix, to obtain the network status graph; then inputs the network status graph into the trained graph neural network model for prediction to obtain the network status graph at the next moment; The high-altitude computing platform obtains the predicted network status graph at the next moment, normalizes each node metric factor and multiplies it by the influence factor of each metric for weighting to obtain the node cost, takes the cost of the node with the larger cost value on both sides of the link as the link cost, obtains the network cost graph, and uses the shortest algorithm to solve for the minimum cost, that is, the optimal route; The high-altitude computing platform issues the optimal routing table to the routing request node to update the node routing table.

5. A routing system for mobile ad hoc networks based on graph neural networks and mobile sensing, characterized in that, It includes: Routing discovery module: The unmanned aerial vehicle (UAV) periodically uploads its own node information to the high-altitude computing platform using the information collection protocol; Model training module: The high-altitude computing platform uses the uploaded node information to construct the network status graph as the input for training the graph neural network model, trains the graph neural network model, and updates the model parameters for predicting the network status graph at the next moment; And, routing optimization module: When a node makes a routing request, the high-altitude computing platform issues commands to collect all node information to form a network status graph; the network status graph is input into the trained graph neural network model to obtain the network status graph at the next moment, and then uses the shortest path algorithm to select the optimal route and issue it to the UAV making the routing request; The flying ad-hoc network includes: users, UAVs, and high-altitude computing platforms; Each user is associated with the UAV with the strongest received signal strength. When a user has a communication need, it will upload the data packet to the associated UAV. Finding the route for the user data packet is to find the route of the associated UAV, and the UAV is regarded as a node; After receiving the data packet, the UAV sends a routing request to the high-altitude computing platform. The high-altitude computing platform sends an information collection command to the UAV. After collecting the UAV information, it preprocesses to form a network status graph; the network status graph is input into the trained graph neural network model to obtain the network status graph at the next moment, and then uses the shortest path algorithm to select the optimal route and issue it to the UAV making the routing request. The UAV forwards the data packet sent by the user after receiving the optimal route; The steps for the UAV to upload its own node information to the high-altitude computing platform are: All nodes regularly generate and send NLIP messages; After receiving the NLIP message, the neighbor node checks the validity of the message to determine whether it is a duplicate message. If it is not a duplicate message, it forwards it to other neighbor nodes and increments according to the hop number field in the message to indicate an increase in the hop count of the path; All nodes broadcast NLIP messages until the high-altitude computing platform collects all node information of the complete flying ad-hoc network, and the high-altitude computing platform selects multi-point relay nodes based on the collected network information; Each node periodically sends source announcement data to the multi-point relay node according to the information collection protocol, and updates the node routing table according to the network conditions. If a neighbor node loses connection or does not receive the source announcement data, the neighbor node is removed from the node routing table.

6. A computer storage medium stores a readable program, characterized in that, When the program runs, the program can instruct the computing device to execute the routing method for flying ad hoc networks based on graph neural networks and mobile sensing according to any one of claims 1-4.

7. An electronic device, characterized in that, Including: A processor, a memory, a communication interface, and a communication bus. The processor, the memory, and the communication interface complete communication with each other through the communication bus; The memory is used to store at least one executable instruction, and the executable instruction causes the processor to execute the operations corresponding to the routing method for flying ad hoc networks based on graph neural networks and mobile sensing according to any one of claims 1-4.

8. A computer program product, comprising computer instructions, characterized in that, The computer instruction instructs the computing device to execute the operations corresponding to the routing method for flying ad hoc networks based on graph neural networks and mobile sensing according to any one of claims 1-4.

Images