Intelligent networking method and system for wireless ad hoc network

Abstract

The invention relates to the technical field of communication, and provides an intelligent networking method and system for a wireless ad hoc network, and the method comprises the steps: firstly collecting the position, signal strength, bandwidth load and other state information of each ad hoc network node, and then transmitting the state information to a distributed edge calculation module through a wireless module; the method comprises the following steps: establishing a network topological graph reflecting link signal attenuation and packet loss rate, transmitting ad-hoc network node state information and the network topological graph to a cloud computing platform for comprehensive analysis to generate a current optimal transmission path, networking based on the current optimal transmission path, and finally continuously monitoring the moving state of the ad-hoc network node and updating the information of the ad-hoc network node. And repeating the steps to dynamically optimize the networking strategy. Therefore, the transmission efficiency and the stability of the wireless ad hoc network can be effectively improved.

Description

Technical Field

[0001] This application relates to the field of communication technology, and more specifically, to an intelligent networking method and system for wireless ad hoc networks. Background Technology

[0002] In marine environments, applications such as ocean monitoring, disaster early warning, and remote control require high-bandwidth, low-latency data transmission. For example, real-time video surveillance, meteorological data transmission, and emergency rescue command all require efficient and stable network support.

[0003] Due to the complex and ever-changing marine environment, traditional routing algorithms, typically designed based on fixed network topologies, cannot effectively adapt to dynamically changing node locations and link states. Specifically, these algorithms assume that nodes and links in the network are static and stable. However, in practical applications, especially in ad hoc maritime networks, the frequent changes in node locations (such as on survey vessels) lead to continuous shifts in network topology, impacting wireless communication quality. Furthermore, link quality can become unstable due to signal attenuation, interference, and other factors. In this dynamic environment, traditional routing algorithms exhibit slow response times, time-consuming path recalculation, and are prone to data transmission delays and packet loss, failing to meet real-time and reliability requirements.

[0004] Therefore, an optimized intelligent networking solution for wireless ad hoc networks is needed. Summary of the Invention

[0005] This application addresses the shortcomings of existing technologies by providing an intelligent networking method and system for wireless ad hoc networks.

[0006] According to one aspect of this application, an intelligent networking method for wireless ad hoc networks is provided, comprising:

[0007] S1: Collect status information of multiple self-organizing network nodes, including location information, signal strength, and bandwidth load;

[0008] S2: The status information of the multiple self-organizing network nodes is transmitted to the distributed edge computing module through the wireless module. The distributed edge computing module is used to construct a network topology map based on the status information of the multiple self-organizing network nodes. The edges of the network topology map are the link status parameters of each link in the network topology map. The link status parameters include signal attenuation value and packet loss rate.

[0009] S3: Transmit the status information of the multiple self-organizing network nodes and the network topology diagram to the cloud computing platform;

[0010] S4: On the cloud computing platform, the status information of the multiple self-organizing network nodes and the network topology are comprehensively analyzed to generate the current optimal transmission path;

[0011] S5: Form a network based on the current optimal transmission path;

[0012] S6: Monitor the movement status of the multiple self-organizing network nodes, update the status information of the multiple self-organizing network nodes, and repeat steps S2-S4 to update the current optimal transmission path for network formation.

[0013] According to another aspect of this application, an intelligent networking system for wireless ad hoc networks is provided, comprising:

[0014] The node status information collection module is used to collect status information of multiple self-organizing network nodes, including location information, signal strength, and bandwidth load.

[0015] The node status information transmission module is used to transmit the status information of the multiple self-organizing network nodes to the distributed edge computing module through the wireless module. The distributed edge computing module is used to construct a network topology map based on the status information of the multiple self-organizing network nodes. The edges of the network topology map are the link status parameters of each link in the network topology map. The link status parameters include signal attenuation value and packet loss rate.

[0016] The node status network topology information transmission module is used to transmit the status information of the multiple self-organizing network nodes and the network topology map to the cloud computing platform.

[0017] The current optimal transmission path generation module is used to perform comprehensive analysis on the cloud computing platform on the status information of the multiple self-organizing network nodes and the network topology to generate the current optimal transmission path.

[0018] The networking module is used to form a network based on the current optimal transmission path;

[0019] The network update module is used to monitor the movement status of the multiple self-organizing network nodes and update the status information of the multiple self-organizing network nodes. It repeatedly executes the node status information transmission module - current optimal transmission path generation module to update the current optimal transmission path for network formation.

[0020] This application, by adopting the above technical solution, has significant technical effects:

[0021] The intelligent networking method and system for wireless ad hoc networks provided in this application first collects status information such as the location, signal strength, and bandwidth load of each ad hoc network node. Then, this status information is transmitted via a wireless module to a distributed edge computing module to construct a network topology map reflecting link signal attenuation and packet loss rate. Next, the ad hoc network node status information and the network topology map are transmitted to a cloud computing platform for comprehensive analysis to generate the current optimal transmission path. Networking is then performed based on this optimal transmission path. Finally, the movement status of ad hoc network nodes is continuously monitored and their information is updated, and the above steps are repeated to dynamically optimize the networking strategy. This effectively improves the transmission efficiency and stability of the wireless ad hoc network. Attached Figure Description

[0022] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0023] Figure 1 This is a flowchart of an intelligent networking method for wireless ad hoc networks according to an embodiment of this application.

[0024] Figure 2 This is a flowchart of S4 in the intelligent networking method for wireless ad hoc networks according to an embodiment of this application.

[0025] Figure 3 This is a flowchart of step S41 in the intelligent networking method for wireless ad hoc networks according to an embodiment of this application.

[0026] Figure 4 This is a flowchart of S43 in the intelligent networking method for wireless ad hoc networks according to an embodiment of this application.

[0027] Figure 5 This is a system block diagram of an intelligent networking system for wireless ad hoc networks according to an embodiment of this application. Detailed Implementation

[0028] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0029] In maritime applications such as marine monitoring, disaster early warning, and remote control, efficient and stable data transmission is crucial, especially for real-time video surveillance, meteorological data transmission, and emergency rescue command. However, the complex and ever-changing marine environment, with frequent changes in node (e.g., survey vessels) locations, leads to unstable wireless communication quality. Traditional routing algorithms, based on fixed network topology designs, struggle to adapt to these dynamically changing node locations and link states. These algorithms assume that network nodes and links are static and stable, but in the ad hoc maritime network environment, frequent changes in node locations cause continuous changes in the network topology. Combined with signal attenuation and interference, this affects link quality, resulting in slow response times, time-consuming path recalculation, and data transmission delays and packet loss, failing to meet real-time and reliability requirements.

[0030] Based on this, this application proposes an intelligent networking method for wireless ad hoc networks, which combines distributed edge computing and cloud computing platforms and utilizes advanced technologies such as deep learning to effectively address challenges such as dynamic environments and link instability in maritime ad hoc networks. Figure 1 This is a flowchart of an intelligent networking method for wireless ad hoc networks according to an embodiment of this application. Figure 1 As shown, the intelligent networking method for wireless ad hoc networks according to an embodiment of this application includes: S1: collecting status information of multiple ad hoc network nodes, the status information including location information, signal strength, and bandwidth load; S2: transmitting the status information of the multiple ad hoc network nodes to a distributed edge computing module via a wireless module, the distributed edge computing module being used to construct a network topology map based on the status information of the multiple ad hoc network nodes, the edges of the network topology map being link status parameters of each link in the network topology map, the link status parameters including signal attenuation value and packet loss rate; S3: transmitting the status information of the multiple ad hoc network nodes and the network topology map to a cloud computing platform; S4: on the cloud computing platform, performing comprehensive analysis on the status information of the multiple ad hoc network nodes and the network topology map to generate the current optimal transmission path; S5: networking based on the current optimal transmission path; S6: monitoring the movement status of the multiple ad hoc network nodes and updating the status information of the multiple ad hoc network nodes, repeating steps S2-S4 to update the current optimal transmission path for networking.

[0031] In S1, status information of multiple ad hoc network nodes is collected, including location information, signal strength, and bandwidth load. Specifically, these multiple ad hoc network nodes include fixed nodes and dynamic nodes; fixed nodes (including ocean stations and buoys) and dynamic nodes (research vessels). It should be understood that, considering the complex and ever-changing marine ad hoc network environment, the status of both fixed and dynamic nodes in the network is constantly changing. Location information, signal strength, and bandwidth load can intuitively reflect the working status of each node. For example, during navigation, the position of a research vessel constantly changes, which alters its signal strength and available bandwidth load with other nodes; while fixed nodes, although their position remains constant, may still have their signal strength and bandwidth load affected by surrounding environmental factors. Therefore, collecting this information can reflect the real-time operating status of the entire ad hoc network, helping to construct a more accurate network topology map and select the optimal path for data transmission to ensure high network stability and low latency.

[0032] The following is a detailed description of a specific implementation process for "collecting status information of multiple ad hoc network nodes, including location information, signal strength, and bandwidth load":

[0033] First, suitable hardware devices must be deployed on each ad hoc network node to collect relevant information. To obtain location information, a Global Positioning System (GPS) module will be installed. Similar to nodes on oceanographic survey vessels or buoys, the GPS module can accurately obtain latitude and longitude coordinates in real time, thus determining the node's location within the network. For signal strength information collection, dedicated wireless signal strength detection equipment will be used to monitor the strength of the wireless signals emitted and received by the node itself. Regarding bandwidth load monitoring, traffic monitoring chips or software tools need to be deployed at the node's network interface to statistically analyze the data traffic sent and received by the node per unit time, thereby measuring bandwidth usage.

[0034] Determining the data acquisition frequency is a crucial step, and it's set based on the application scenario of the ad hoc network and the degree of network dynamic change. In scenarios like marine monitoring, where dynamic nodes such as survey vessels move rapidly and the network topology changes frequently, the acquisition frequency of location information, signal strength, and bandwidth load needs to be increased, perhaps every few seconds, to promptly capture changes in node status. For relatively stable fixed nodes such as shore-based base stations or specific marine stations, the acquisition frequency can be appropriately reduced, for example, every few minutes, which effectively acquires information while minimizing resource consumption.

[0035] After hardware deployment and frequency setting are completed, the data acquisition and preprocessing phase begins. Hardware devices on each node operate at the set frequency: the GPS module continuously acquires location coordinate data, the wireless signal strength detection device monitors signal strength in real time, and the bandwidth monitoring tool calculates bandwidth load data. However, during actual data acquisition, environmental interference and other factors may cause the acquired data to contain noise or outliers. Therefore, data preprocessing is necessary. For location information, filtering algorithms are used to remove abnormal coordinate points caused by GPS signal obstruction or interference; for signal strength data, data that significantly deviates from the normal range is removed by setting a reasonable threshold range; and for bandwidth load data, smoothing algorithms are used to process data with large fluctuations, thereby ensuring data accuracy and reliability.

[0036] The preprocessed data needs to be stored. Each self-organizing network node will be equipped with a local storage unit to temporarily store the collected status information. The storage unit can be a small flash memory chip or a memory area. When the data volume is small, the data is directly stored in the local cache and awaits transmission. If the data volume is large or the collection frequency is high, a circular storage method is adopted. That is, when the storage unit is full, the new data overwrites the oldest stored data to ensure that the latest status information is always retained.

[0037] Before transmitting the collected status information to the distributed edge computing module, data transmission preparation is required. Location information, signal strength, and bandwidth load data are packaged according to a specific format, such as JSON or XML, to ensure a unified data structure for easy transmission and parsing. Simultaneously, a node identifier and timestamp are added to each data packet. The node identifier distinguishes different ad hoc network nodes, and the timestamp records the specific time of data collection, facilitating subsequent analysis by understanding the data's timeliness and the sequence of node status changes. This provides reliable data support for subsequent network analysis and decision-making.

[0038] In S2, the status information of the multiple self-organizing network nodes is transmitted to the distributed edge computing module via a wireless module. The distributed edge computing module constructs a network topology graph based on the status information of the multiple self-organizing network nodes. The edges of the network topology graph represent the link status parameters of each link, including signal attenuation and packet loss rate. Correspondingly, considering the large number of self-organizing network nodes at sea and the resulting massive amount of status information, directly transmitting all information to the core cloud computing platform for processing would place a huge burden on the core network's data transmission and processing, leading to network congestion and increased latency. Therefore, transmitting the status information of the multiple self-organizing network nodes to the distributed edge computing module via a wireless module allows for preliminary data processing closer to the data source, reducing the data transmission volume and processing burden on the core network and ensuring its efficient operation. Furthermore, in the maritime environment, node status changes frequently; for example, the movement of a survey vessel can alter its link status with other nodes in real time. The distributed edge computing module can promptly acquire this changing information, significantly shortening processing time compared to transmitting data to a remote core platform for processing, thus meeting real-time requirements. It's also worth mentioning that the distributed architecture of the distributed edge computing module means that even if a node or part of the network fails, other nodes can continue to operate and maintain basic functions, improving the overall reliability and fault tolerance of the system. Next, considering that signal attenuation and packet loss rate are key indicators for measuring network link performance, in a maritime ad hoc network environment, signals weaken during transmission due to factors such as distance and obstacles. By measuring and recording signal attenuation, the actual communication quality of each link can be accurately understood. Packet loss rate reflects the proportion of data packets lost during transmission and is an important indicator for measuring link stability. A high packet loss rate usually indicates a link problem or excessive load. Therefore, constructing a network topology map using these parameters allows for timely capture of changes in network status, making the topology map more consistent with actual network conditions, and enabling more accurate assessment of link quality, thus allowing for more precise selection of the optimal path for data transmission.

[0039] The following is a detailed explanation of a specific implementation process for "constructing a network topology diagram based on the state information of the multiple self-organizing network nodes":

[0040] First, the distributed edge computing module acquires status information from multiple self-organizing network nodes, including location information, signal strength, and bandwidth load. To ensure the reliability and validity of the data, the raw data needs to be preprocessed to remove outliers and fill in missing values, so that the data can accurately reflect the actual status of the nodes and provide a solid data foundation for subsequent construction work.

[0041] Next, the nodes and their connections are determined. Based on the node location information, the location of each ad hoc network node is determined in the network topology diagram, with each node having a corresponding location identifier. Then, the relative positions and signal strength between the nodes are used to determine whether a connection exists between them. If the signal strength between two nodes exceeds a certain threshold, it indicates that a link connection can be established between them, thus determining the connection relationships between the nodes in the network topology diagram.

[0042] Next, signal attenuation needs to be calculated. For established node pairs, the signal attenuation during transmission needs to be calculated based on the signal propagation model and factors such as distance between them and intermediate obstacles. In marine environments, due to the absorption and refraction of signals by seawater, a free-space propagation model can be used, modified according to the actual situation. By comprehensively considering these factors, the signal attenuation value of each link can be calculated relatively accurately, thereby measuring the signal loss during transmission. Packet loss rate statistics are also essential. Over a period of time, by monitoring the data packets transmitted through the links, the total number of data packets sent and the number of lost data packets are counted, and the packet loss rate of each link is calculated. To obtain real-time and accurate packet loss rate data, a sliding window mechanism can be used to statistically analyze the sending and losing of data packets within continuous time windows. This can reflect the stability of the links in a timely manner and provide important link status information for the network topology diagram.

[0043] After completing the above steps, a network topology graph can be constructed. Based on the defined nodes and connections, the network topology structure is built, and the calculated signal attenuation and packet loss rates are labeled as link state parameters for the corresponding links in the graph. A graph data structure is used for storage and representation, with each node represented by a vertex, and links between nodes represented by edges. The attributes of the edges record link state parameters such as signal attenuation and packet loss rates, thus forming a complete network topology graph.

[0044] Due to the complex and ever-changing environment of maritime ad hoc networks, nodes are mobile, and environmental factors are constantly changing, leading to continuous changes in node and link states. Therefore, it is essential to monitor changes in node state information in real time, recalculate signal attenuation and packet loss rates when states change, and update link state parameters in the network topology diagram. This ensures that the network topology diagram accurately reflects the actual network state, providing a reliable basis for subsequent path planning and network optimization.

[0045] In S3, the status information of the multiple ad hoc network nodes and the network topology map are transmitted to the cloud computing platform. It should be understood that in wireless ad hoc networks, although distributed edge computing modules can quickly process local data and react instantly, the cloud platform has a greater advantage when facing more complex analysis tasks and long-term data storage needs. Specifically, maritime ad hoc networks cover a wide area, involve numerous nodes, and generate massive amounts of data. By transmitting the status information of the multiple ad hoc network nodes and the network topology map to the cloud computing platform, the platform's powerful data storage capabilities can be utilized to effectively store this large-scale data. Moreover, the cloud computing platform possesses massive computing resources, enabling complex and more refined analysis of the collected ad hoc network node status information and network topology map, thereby helping to formulate more scientific and reasonable routing strategies and significantly improving network performance.

[0046] In S4, the cloud computing platform performs a comprehensive analysis of the status information of the multiple ad hoc network nodes and the network topology to generate the current optimal transmission path. It should be understood that the operation of a maritime ad hoc network is affected by various factors. The location, signal strength, and bandwidth load in the node status information, as well as the link status parameters (signal attenuation and packet loss rate) in the network topology, all describe the local or overall characteristics of the network from different perspectives. Therefore, in order to select paths with good signal quality, low packet loss rate, and sufficient bandwidth to meet the high bandwidth and low latency communication requirements of the maritime ad hoc network, the technical concept of this application is to use deep learning-based data analysis and encoding techniques to perform feature encoding on the status information of the multiple ad hoc network nodes and the network topology. Then, the encoded attribute features of the multiple ad hoc network nodes and the semantic features of the network topology are comprehensively analyzed. Based on the ad hoc network link status representation containing network topology information, the path selection weight value corresponding to each filtered representation is automatically generated, and the transmission path corresponding to the maximum value is taken as the current optimal transmission path. This ensures that the selected path can provide optimal communication quality and stability while meeting the requirements for high bandwidth and low latency.

[0047] Specifically, Figure 2 This is a flowchart of step S4 in an intelligent networking method for wireless ad hoc networks according to an embodiment of this application. Figure 2As shown, step S4 includes: S41, performing feature encoding on the state information of the plurality of ad hoc network nodes and the network topology map to obtain multiple ad hoc network node attribute feature vectors and a network topology semantic feature map; S42, performing node attribute-network topology comprehensive analysis on the multiple ad hoc network node attribute feature vectors and the network topology semantic feature map to obtain multiple ad hoc network link state representation vectors containing network topology information; S43, performing feature filtering on the multiple ad hoc network link state representation vectors containing network topology information based on key clues of ad hoc network links to obtain multiple filtered ad hoc network path state representation vectors containing network topology information; S44, obtaining multiple path selection weight values ​​based on the multiple filtered ad hoc network path state representation vectors containing network topology information; S45, selecting the transmission path corresponding to the maximum value among the multiple path selection weight values ​​as the current optimal transmission path.

[0048] In S41, feature encoding is performed on the state information of the multiple ad hoc network nodes and the network topology map to obtain attribute feature vectors of multiple ad hoc network nodes and a semantic feature map of the network topology. Specifically, Figure 3 This is a flowchart of step S41 in an intelligent networking method for wireless ad hoc networks according to an embodiment of this application. Figure 3 As shown, S41 includes: S411, using a node state encoder based on a multilayer perceptron to encode the state information of each of the multiple self-organizing network nodes to obtain the attribute feature vector of the multiple self-organizing network nodes; S412, passing the network topology graph through a network topology feature extractor based on a DCNN model to obtain the network topology semantic feature graph.

[0049] In S411, a node state encoder based on a multilayer perceptron is used to encode the state information of each of the multiple self-organizing network nodes to obtain the attribute feature vectors of the multiple self-organizing network nodes. Accordingly, considering that the state information of self-organizing network nodes contains multi-dimensional data such as location, signal strength, and bandwidth load, and there are complex interrelationships between these dimensions, this application uses a node state encoder based on a multilayer perceptron to encode the state information of each of the multiple self-organizing network nodes to obtain the attribute feature vectors of the multiple self-organizing network nodes in order to capture and extract the correlations between the attributes of each node and extract deep features from the data. It is worth mentioning that the multilayer perceptron (MLP), as a feedforward neural network, has powerful nonlinear mapping capabilities and can effectively process such complex, multi-dimensional, and interrelated information. Through multiple neurons and hidden layers, the MLP can automatically learn the complex patterns and inherent laws in these data, thereby accurately encoding the state information of each self-organizing network node. At the same time, it can deeply analyze the data, extract the potential features in the original state information that are valuable for network transmission path selection, and transform them into more representative attribute feature vectors.

[0050] In step S412, the network topology graph is processed by a network topology feature extractor based on a DCNN model to obtain a network topology semantic feature map. Accordingly, considering that the network topology graph contains rich structural information, such as the connection relationships between nodes and the link quality relationships reflected by link state parameters (signal attenuation values ​​and packet loss rates), the technical solution of this application uses a network topology feature extractor based on a DCNN model to obtain a network topology semantic feature map. Specifically, the DCNN model has the characteristics of local connectivity and weight sharing, enabling it to automatically learn the spatial features and hierarchical structure information in the network topology graph. Specifically, through continuous feature extraction and abstraction via convolutional layers, from low-level pixel-level features to high-level semantic features, the deep features crucial for path planning and network analysis in the network topology graph are accurately extracted.

[0051] In step S42, a node attribute-network topology comprehensive analysis is performed on the plurality of ad hoc network node attribute feature vectors and the network topology semantic feature map to obtain a plurality of ad hoc network link state representation vectors containing network topology information. Specifically, in this embodiment, step S42 includes: passing the plurality of ad hoc network node attribute feature vectors and the network topology semantic feature map through a node attribute-network topology comprehensive analyzer based on graph attention networks to obtain the plurality of ad hoc network link state representation vectors containing network topology information. It should be understood that in maritime ad hoc networks, the interaction between nodes and the relationship between nodes and links have a significant impact on network performance. Therefore, in order to more accurately reflect the actual transmission quality of each link (such as signal attenuation value, packet loss rate, etc.) and provide a reliable basis for subsequent path selection, this application passes the plurality of ad hoc network node attribute feature vectors and the network topology semantic feature map through a node attribute-network topology comprehensive analyzer based on graph attention networks to obtain a plurality of ad hoc network link state representation vectors containing network topology information. It's worth noting that GAT's attention mechanism can focus on the relationships between different nodes and links, accurately capturing how changes in node state affect link state, and how link state in turn affects communication between nodes. For example, when the signal strength of a node changes, GAT can quickly locate the affected link through its attention mechanism and analyze its impact on the overall network link state.

[0052] In S43, feature filtering based on key clues of the ad hoc network links is performed on the multiple ad hoc network link state representation vectors containing network topology information to obtain multiple filtered ad hoc network path state representation vectors containing network topology information. Specifically, Figure 4 This is a flowchart of step S43 in an intelligent networking method for wireless ad hoc networks according to an embodiment of this application. Figure 3As shown, step S43 includes: S431, rearranging the plurality of ad hoc network link state representation vectors containing network topology information to obtain a sequence of rearranged ad hoc network link state representation vectors containing network topology information; S432, extracting the i-th rearranged ad hoc network link state representation vector containing network topology information from the sequence of rearranged ad hoc network link state representation vectors containing network topology information as the ad hoc network link state representation vector to be analyzed; S433, selecting the first i-1 rearranged ad hoc network link state representation vectors containing network topology information from the sequence of rearranged ad hoc network link state representation vectors containing network topology information. S434: Input the key clue capture network to obtain the key clue guidance vector of the pre-sequential distribution of the self-organizing network link state containing network topology information; S435: Input the key clue guidance vector of the pre-sequential distribution of the self-organizing network link state containing network topology information and the characterization vector of the self-organizing network link state containing network topology information into the clue weaving gating network to obtain the self-organizing network link state clue weaving closed-loop strength factor containing network topology information; S436: Based on the comparison between the self-organizing network link state clue weaving closed-loop strength factor containing network topology information and a preset threshold, determine whether to filter the characterization vector of the self-organizing network link state containing network topology information.

[0053] It is understandable that multiple ad hoc network link state representation vectors containing network topology information are high-dimensional data, which may contain redundant information and noise information generated by external factors. For example, in the complex electromagnetic environment at sea, signal transmission may be subject to various interferences, resulting in some unstable and abnormal data mixed into the link state representation vectors. Furthermore, different links play different roles in the network; some links are crucial to the overall network connectivity and data transmission efficiency, while others have a relatively smaller impact. To effectively remove this redundant and interfering information, extract the key features truly valuable for path planning and network analysis, and highlight key information, this application performs feature filtering based on key clues of ad hoc network links on the multiple ad hoc network link state representation vectors containing network topology information to obtain multiple filtered ad hoc network path state representation vectors containing network topology information. Specifically, this method, by constructing a closed-loop screening system that fuses saliency coarse descriptions and key clues, can significantly improve the performance of feature selection in terms of context adaptation and task relevance, ensuring that the obtained representation vectors retain key information closely related to optimal path planning, thus selecting the optimal path that truly meets the requirements of high bandwidth and low latency communication.

[0054] Specifically, in this embodiment, S431 includes: extracting the maximum value of each self-organizing network link state representation vector containing network topology information from the plurality of self-organizing network link state representation vectors containing network topology information as a significant coarse descriptor to obtain a sequence of significant coarse descriptors for self-organizing network link states. This process can be expressed by the following formula:

[0055] X = {x1, x2, ..., x} i ,...,x n}

[0056] S = [s1, s2, ..., s i ,...,s n ]

[0057] s i =max(x i )

[0058] Where X is the plurality of ad hoc network link state representation vectors containing network topology information, x1, x2, x3, x4, x5, x6, x7, x8, x9, x1, x1, x2, x3, x4, x9, x1, x1, x2, x3, x4, x5 ... i and x n Let be the 1st, 2nd, 1st, and 1st ad hoc link state representation vectors containing network topology information, respectively. max(·) represents the maximum value operation. S is a sequence of crude descriptors of the saliency of ad hoc link states, s1, s2, s3, s4, s5, s6, s7, s8, s9, s1 ... i and s n These are the 1st, 2nd, 1st, and nth crude descriptors of the saliency of the ad hoc network link state, respectively.

[0059] Based on the ascending order of the sequence of the crude descriptors of the ad hoc network link state saliency, the multiple ad hoc network link state representation vectors containing network topology information are rearranged to obtain the sequence of rearranged ad hoc network link state representation vectors containing network topology information. This process can be expressed by the following formula:

[0060]

[0061] Where σ(X;S) represents the rearrangement of X based on the ascending order of S. It is a sequence of ad hoc network link state rearrangement representation vectors containing network topology information. and These are the 1st, 2nd, 1st, and 2nd ad hoc link state rearrangement representation vectors containing network topology information in the sequence of ad hoc link state rearrangement representation vectors.

[0062] Correspondingly, considering that multiple ad hoc network link-state representation vectors containing network topology information are typically high-dimensional data containing a large amount of feature information, directly performing complex analysis on this high-dimensional data in the initial stage would face problems such as large computational load and long processing time. By extracting the maximum value of each ad hoc network link-state representation vector containing network topology information as a significant coarse descriptor, the high-dimensional feature vectors can be compressed into a low-dimensional feature representation. This reduces data complexity, lowers the difficulty of subsequent processing, and improves processing efficiency. Furthermore, the maximum value of a vector often represents, to some extent, a certain prominent feature or key attribute of the information contained in the vector. By capturing the maximum value, the representational effect of the polar distribution in the vector can be utilized to provide an intuitive and discriminative reference for subsequent feature analysis and ranking. That is, the obtained sequence of significant coarse descriptors of network link states retains the most prominent feature information of each corresponding vector. Although this description is coarse, it can represent the main features of the vector to a certain extent, providing a concise and effective data foundation for subsequent analysis.

[0063] It's understandable that rearrangement is essentially a deep reconstruction of data. The original ad hoc network link state representation vector may be unordered, with unclear relationships between features. However, after rearranging according to the ascending order of significance factors, the data has a new organizational form, which endows the features with "distribution order semantics." For example, in the link analysis of maritime ad hoc networks, the originally chaotic signal strength, bandwidth load, and other features, after rearrangement, each feature's position in the sequence represents its relative significance. This semantic information makes the data more logical and interpretable, allowing subsequent processing steps to more clearly understand the meaning implied by the data. Specifically, placing features with low significance at the beginning of the sequence effectively avoids assigning excessive weight to these features in subsequent calculations. This allows computing resources to be more rationally focused on the truly critical ad hoc network link state features, improving computational efficiency and analytical accuracy. Meanwhile, highly significant features are pushed to the end of the sequence, ensuring that features that have a significant impact on link state and network performance are fully considered. This allows for a more rational and efficient flow of information during the processing. It is also worth mentioning that this ranking method based on saliency-based descriptive factors is beneficial for enhancing contextual consistency. Enhanced contextual consistency ensures the coherence and logic of the entire feature analysis process, avoiding information loss or misinterpretation caused by disordered feature order. This improves the overall effectiveness of ad hoc network link status analysis and provides strong support for the stable and efficient operation of maritime ad hoc networks.

[0064] Next, the i-th reordered representation vector of the ad hoc network link state containing network topology information is extracted from the sequence of such vectors as the representation vector to be analyzed. It should be understood that each link in an ad hoc network has different characteristics and states. By extracting the i-th reordered representation vector of the ad hoc network link state containing network topology information as the representation vector to be analyzed, a deeper analysis of the corresponding link can be performed. Furthermore, ad hoc networks typically contain a large number of links, and processing the state information of all links simultaneously would impose a huge computational burden. Extracting the state representation vector of a single link for analysis can significantly reduce the amount of data processing. It should be noted that the i-th reordered representation vector of the ad hoc network link state containing network topology information is merely an illustrative method and does not refer to a specific representation vector.

[0065] Specifically, in this embodiment, S433 includes: calculating the saliency distribution representation value of each of the first i-1 ad hoc link state rearrangement representation vectors containing network topology information to obtain the saliency distribution representation value of the first i-1 ad hoc link states. This process can be expressed by the formula:

[0066]

[0067] in, Let μ be the k-th ad hoc link state rearrangement representation vector containing network topology information among the first i-1 ad hoc link state rearrangement representation vectors containing network topology information. k and σ k 2 They are respectively The mean and variance, σ is the marginal value, e k It is the k-th saliency distribution representation of the ad hoc network link state among the first i-1 saliency distribution representations;

[0068] The weighted normalization process is performed on each of the first i-1 ad hoc link state saliency distribution representations to obtain the first i-1 normalized ad hoc link state saliency distribution representations. This process can be expressed by the formula:

[0069]

[0070] Among them, e k It is the k-th saliency distribution representation of the ad hoc network link state among the first i-1 saliency distribution representations, a kIt is the k-th normalized ad hoc network link state saliency distribution representation value among the first i-1 normalized ad hoc network link state saliency distribution representation values;

[0071] Using the first i-1 normalized ad hoc network link state saliency distribution representation values ​​as a sequence of weights, the weighted sum of the first i-1 ad hoc network link state rearrangement representation vectors containing network topology information is calculated to obtain the ad hoc network link state preorder distribution key clue guidance vector containing network topology information. This process can be expressed by the formula:

[0072]

[0073] in, Let a be the k-th ad hoc link state rearrangement representation vector containing network topology information in the sequence of ad hoc link state rearrangement representation vectors containing network topology information. k It is the k-th normalized ad hoc network link state saliency distribution representation value among the first i-1 normalized ad hoc network link state saliency distribution representation values, V key It is a key clue guiding vector containing network topology information and the preceding distribution of link states in an ad hoc network.

[0074] Specifically, in the embodiments of this application, the saliency distribution representation value of the ad hoc network link state rearrangement representation vector containing network topology information is related to the maximum value, mean, variance, and marginal value of the ad hoc network link state rearrangement representation vector containing network topology information. The marginal value is used to measure the global autonomous edge coordination of the ad hoc network link state and improve its contribution to the overall self-consistent mode.

[0075] It is understandable that in a maritime ad hoc network, the state of a single link is not isolated but interconnected with surrounding links. By inputting the rearranged representation vectors of the first i-1 ad hoc network links containing network topology information into the key clue capture network, hidden connections between these links can be uncovered. For example, in a maritime ad hoc network, the signal strength changes of adjacent links may exhibit similar trends, and the bandwidth usage of preceding links may affect the available resources of the link currently being analyzed. The key clue capture network can infer and extract these contextual patterns, providing more comprehensive background information for the rearranged representation vector of the i-th ad hoc network link containing network topology information, allowing the analysis to move beyond the isolated state of a single link. In other words, the generated highly abstract key clue guidance vector of the ad hoc network link state preceding distribution containing network topology information condenses the important information of preceding links. This allows for a better assessment of the link's position and role in the entire network when selecting the optimal transmission path, based not only on the characteristics of the link being analyzed but also on the key clues of the preceding link distribution. This enables a more accurate planning of transmission paths that conform to the actual situation.

[0076] Specifically, the key clue capturing network, based on the statistical characteristics of ad hoc network link states, uses the marginal saliency factor of the extreme values ​​of features relative to the overall self-consistency trend, the marginal-global consistency factor of feature volatility and the overall self-consistency level, and the marginal correlation factor of feature volatility range based on the overall self-consistency relationship to measure the global saliency self-consistency edge synergy. This means performing global self-consistency relationship synergy based on edge saliency, so as to model the consistency and synergy of edge distributions globally within the saliency distribution pattern of the input model. In this way, it can ensure that the distribution of extreme edge data is consistent with the overall global self-consistency trend, while also ensuring that the data volatility based on edge distribution is self-consistent with the global whole. Therefore, the key clue guidance vector of the ad hoc network link state preceding distribution containing network topology information can essentially be a feature with certain edge volatility characteristics for contextual saliency. This can be improved by edge-global self-consistency re-inference to enhance its contribution to the overall self-consistency pattern, thus providing a reference index for global self-consistency fusion based on edge saliency synergy for potential patterns of sequence distribution.

[0077] Then, the preorder distribution key clue guidance vector of the ad hoc network link state containing network topology information and the characterization vector of the ad hoc network link state containing network topology information are input into the clue weaving gating network to obtain the closed-loop strength factor of the ad hoc network link state clue weaving containing network topology information. The above process can be expressed by the formula:

[0078]

[0079] in, V is the i-th ad hoc link state rearrangement representation vector containing network topology information in the sequence of ad hoc link state rearrangement representation vectors containing network topology information. key It is a key clue guiding vector containing network topology information about the preceding distribution of link states in an ad hoc network, (·). T For the transpose operation, S iV -1 yes and V key The inverse of the covariance matrix, where |·| is the absolute value operation. yes The corresponding self-organizing network link state clue weaving closed-loop strength factor containing network topology information.

[0080] It can be understood that the preceding key clue guidance vector of the ad hoc network link state containing network topology information carries the contextual background information of the link state, such as the state trends of surrounding links and the overall network layout characteristics; while the ad hoc network link state characterization vector containing network topology information focuses on the real-time state details of the link to be analyzed. The thread-weaving gated network, through a design similar to GRU, achieves deep information fusion between the two while simultaneously suppressing the interference of noise features. For example, in the maritime ad hoc network scenario, it can combine the historical bandwidth usage trend information of preceding links with the real-time bandwidth load information of the link to be analyzed, comprehensively presenting the actual situation of the link in the network and avoiding the one-sidedness of analysis based solely on a single vector. The generated ad hoc network link state thread weaving closed-loop strength factor containing network topology information serves as a comprehensive indicator, capable of fully quantifying the consistency, significance, and task relevance between the characteristic to be analyzed and its context link characteristics. Specifically, in terms of consistency, it measures the degree of matching between the current link state and the states of surrounding links, determining whether the link is in a coordinated and stable state within the network. Salience highlights the importance of the link within the entire network, such as its critical contribution to data transmission. Task relevance focuses on whether the link state meets the requirements of a specific communication task, such as the need for low-latency links in real-time video transmission. Through these quantitative indicators, the model can more intuitively understand the comprehensive value of each link, thus providing a clear basis for subsequent decision-making.

[0081] Finally, based on the comparison between the closed-loop strength factor of the ad hoc network link state cues containing network topology information and a preset threshold, it is determined whether to filter the ad hoc network link state characterization vector containing network topology information. The above process can be expressed by the following formula:

[0082]

[0083] in, yes The corresponding weaving closed-loop strength factor for ad hoc network link state clues containing network topology information. Yes Perform feature filtering operation, where θ is a preset threshold.

[0084] It is understandable that ad hoc network link states contain a wealth of complex information, and not all features are equally valuable for network analysis and path planning. The ad hoc network link state cues weaving closed-loop strength factor, containing network topology information, comprehensively considers the context and local saliency of the link being analyzed, fully and accurately reflecting the importance of the link state features. Comparing this with preset thresholds allows for final feature selection. Specifically, in maritime ad hoc network scenarios, links may be affected by various factors such as weather and electromagnetic interference, and some features may only exhibit temporary abnormal fluctuations, not truly reflecting the key attributes of the link. By comparing with thresholds, these interfering or unimportant features can be eliminated, retaining only the link state features that play a crucial role in network performance and data transmission, thereby improving the accuracy and reliability of subsequent link selection analysis results. Furthermore, the environment in which ad hoc networks operate is complex and ever-changing, with link states constantly evolving. For example, at sea, the movement of dynamic nodes such as survey vessels can alter the signal strength and bandwidth load of links, and weather changes can also interfere with link communication quality. By dynamically adjusting the feature importance judgment criteria through context support and local saliency analysis, and then combining this with preset thresholds for selection, the system can respond to these changes in a timely manner. When link states change, the weaving closed-loop strength factor of the ad hoc network link state clues, which contains network topology information, will change accordingly. After comparing it with a preset threshold, the system can quickly decide whether to retain or filter the new link state features. This ensures that the system can always maintain a high-quality, highly adaptable feature set based on the real-time network conditions, enabling the ad hoc network to operate stably in complex and ever-changing environments and meet the network requirements of various application scenarios.

[0085] In step S44, multiple path selection weight values ​​are obtained based on the multiple filtered self-organizing network path state representation vectors containing network topology information. Specifically, in this embodiment, step S44 includes: passing each of the multiple filtered self-organizing network path state representation vectors containing network topology information through a classifier-based path selector to obtain the multiple path selection weight values. That is, multiple filtered self-organizing network path state representation vectors containing network topology information, obtained after filtering key clues using multiple self-organizing network link state representation vectors containing network topology information, are classified to automatically generate multiple path selection weight values. In particular, the classifier-based path selector can convert these complex vector information into specific numerical values, i.e., path selection weight values. Through this quantification method, the relative merits of each path can be measured more intuitively and accurately, providing a clear numerical basis for path selection. In one specific embodiment of this application, each of the multiple filtered ad hoc network path state representation vectors containing network topology information is passed through a classifier-based path selector to obtain the multiple path selection weight values. This includes: using the fully connected layer of the classifier to perform fully connected encoding on the filtered ad hoc network path state representation vectors containing network topology information to obtain a fully connected encoded representation vector of the filtered ad hoc network path state; inputting the fully connected encoded representation vector of the filtered ad hoc network path state into the Softmax classification function of the classifier to obtain the probability value of the filtered ad hoc network path state representation vector belonging to each path selection; and determining the maximum probability value among the probability values ​​of each path selection as the path selection weight value.

[0086] In step S45, the transmission path corresponding to the maximum value among the multiple path selection weights is selected as the current optimal transmission path. By selecting the transmission path corresponding to the maximum value as the current optimal transmission path, the utilization efficiency of network resources can be improved, maximizing the satisfaction of the high bandwidth and low latency communication requirements of the maritime ad hoc network. This path features good signal quality, low packet loss rate, and sufficient bandwidth, ensuring the fast and accurate transmission of large data packets such as hydrological, meteorological, and monitoring video data, avoiding data loss and delay, and providing reliable data transmission guarantees for marine monitoring and other services.

[0087] In summary, S4 is clearly explained. It uses deep learning-based data analysis and encoding techniques to encode the state information and network topology of multiple ad hoc network nodes. Then, it comprehensively analyzes the encoded attribute features of the multiple ad hoc network nodes and the semantic features of the network topology. Based on the state representation of each ad hoc network link containing network topology information, it automatically generates path selection weight values ​​corresponding to each filtered representation after filtering by the ad hoc network link critical line. The transmission path corresponding to the maximum value is then selected as the current optimal transmission path. This ensures in real-time that the selected path provides optimal communication quality and stability while meeting high bandwidth and low latency requirements.

[0088] In S5, networking is performed based on the current optimal transmission path. It should be understood that the current optimal transmission path is the result of comprehensively considering factors such as the status information of ad hoc network nodes, the network topology, and link status parameters. Compared to other paths, networking based on the current optimal transmission path has significant advantages in data transmission stability, speed, and reliability, and can well meet the high bandwidth and low latency communication requirements of maritime ad hoc networks.

[0089] The following is a detailed explanation of a specific implementation process for "networking based on the current optimal transmission path":

[0090] Once the optimal transmission path is obtained, it must first be analyzed in depth. This process requires identifying the identifiers of each ad hoc network node within the path, as these nodes will constitute the actual data transmission link. Taking a maritime monitoring ad hoc network as an example, suppose the optimal transmission path starts from the survey vessel node, passes through multiple buoy nodes, and finally reaches the coastal base station. At this point, based on the information of these nodes, the connection sequence and logical relationships between them must be determined; this forms the basic framework of the network.

[0091] Next, based on the parsed node information, each node participating in the network is configured and its parameters are set accordingly. Regarding wireless communication parameters, adjustments need to be made based on the node's environment and transmission requirements. For example, in areas with weak signals, nodes need to appropriately increase their transmission power to ensure stable signal transmission between nodes; simultaneously, to reduce signal collisions and interference, channels with less interference should be selected based on the surrounding electromagnetic environment. Setting up the routing table is also crucial; each node must clearly define the next-hop node address for data forwarding so that data can be transmitted accurately along the optimal path, avoiding transmission errors or delays.

[0092] After node configuration is complete, the connection establishment and link initialization phase begins. Using technologies such as dedicated wireless communication frequency bands, wireless links are established between adjacent nodes. During connection establishment, a handshake protocol is executed to confirm the identity and communication capabilities of both nodes. Key parameters such as link transmission rate and encoding method are also negotiated. These parameters must match the performance requirements of the optimal transmission path to ensure data is transmitted on the link in the best possible condition, laying the foundation for efficient and stable data transmission.

[0093] After network deployment, data transmission officially begins along the optimal path. During transmission, flow control and error control are crucial for ensuring accurate and efficient data transmission. By monitoring the bandwidth load of nodes in real time, once a node's load approaches or reaches a threshold, flow regulation measures are promptly implemented, such as reducing the data transmission rate or buffering some data, to prevent data congestion leading to packet loss and increased latency. Error control techniques such as Cyclic Redundancy Check (CRC) are employed to verify the transmitted data; if an error is detected, a retransmission is immediately requested to ensure data accuracy.

[0094] In step S6, the movement status of the multiple ad hoc network nodes is monitored, and their status information is updated. Steps S2-S4 are repeated to update the current optimal transmission path for network formation. In other words, the maritime ad hoc network environment is complex, and ad hoc network nodes exhibit mobility, especially dynamic nodes like survey vessels. Node movement alters their relative position to other nodes, signal strength, and bandwidth load, thus affecting the network topology and link status parameters. Changes in network status may cause the original optimal transmission path to become obsolete. Without timely updates, data transmission may face increased latency and packet loss rates, impacting overall network performance. Repeatedly executing the relevant steps to update the optimal transmission path allows the network to adjust according to real-time conditions, maintaining good communication performance and ensuring efficient and stable data transmission.

[0095] In summary, the intelligent networking method for wireless ad hoc networks based on the embodiments of this application is explained. First, it collects status information such as the location, signal strength, and bandwidth load of each ad hoc network node. Then, it transmits this status information to a distributed edge computing module via a wireless module to construct a network topology map reflecting link signal attenuation and packet loss rate. Next, it transmits the ad hoc network node status information and the network topology map to a cloud computing platform for comprehensive analysis to generate the current optimal transmission path. Networking is then performed based on this optimal transmission path. Finally, the movement status of ad hoc network nodes is continuously monitored and their information is updated, and the above steps are repeated to dynamically optimize the networking strategy. This effectively improves the transmission efficiency and stability of the wireless ad hoc network.

[0096] Figure 5This is a system block diagram of an intelligent networking system for wireless ad hoc networks according to an embodiment of this application. Figure 5 As shown, an intelligent networking system 100 for wireless ad hoc networks according to an embodiment of this application includes: a node status information collection module 110, used to collect status information of multiple ad hoc network nodes, the status information including location information, signal strength, and bandwidth load; a node status information transmission module 120, used to transmit the status information of the multiple ad hoc network nodes to a distributed edge computing module via a wireless module, the distributed edge computing module being used to construct a network topology graph based on the status information of the multiple ad hoc network nodes, the edges of the network topology graph being link status parameters of each link in the network topology graph, the link status parameters including signal attenuation value and packet loss rate; and a node status network topology... The system comprises: a data transmission module 130, used to transmit the status information of the multiple self-organizing network nodes and the network topology map to a cloud computing platform; a current optimal transmission path generation module 140, used on the cloud computing platform to perform comprehensive analysis on the status information of the multiple self-organizing network nodes and the network topology map to generate a current optimal transmission path; a networking module 150, used to form a network based on the current optimal transmission path; and a networking update module 160, used to monitor the movement status of the multiple self-organizing network nodes, update the status information of the multiple self-organizing network nodes, and repeatedly execute the node status information transmission module - current optimal transmission path generation module to update the current optimal transmission path for networking.

[0097] Here, those skilled in the art will understand that the specific functions and operations of each unit and module in the intelligent networking system 100 for wireless ad hoc networks described above have been referenced above. Figures 1 to 4 The intelligent networking method for wireless ad hoc networks is described in detail here, and therefore, its repeated description will be omitted.

[0098] In summary, the intelligent networking system 100 for wireless ad hoc networks based on the embodiments of this application is explained. It first collects status information such as the location, signal strength, and bandwidth load of each ad hoc network node. Then, it transmits this status information to a distributed edge computing module via a wireless module to construct a network topology map reflecting link signal attenuation and packet loss rate. Next, it transmits the ad hoc network node status information and the network topology map to a cloud computing platform for comprehensive analysis to generate the current optimal transmission path. Based on this optimal transmission path, it performs network formation. Finally, it continuously monitors the movement status of ad hoc network nodes and updates their information, repeating the above steps to dynamically optimize the networking strategy. This effectively improves the transmission efficiency and stability of the wireless ad hoc network.

Claims

1. A smart networking method for wireless ad hoc networks, characterized in that, include: S1: Collect status information of multiple self-organizing network nodes, including location information, signal strength, and bandwidth load; S2: The status information of the multiple self-organizing network nodes is transmitted to the distributed edge computing module through the wireless module. The distributed edge computing module is used to construct a network topology map based on the status information of the multiple self-organizing network nodes. The edges of the network topology map are the link status parameters of each link in the network topology map. The link status parameters include signal attenuation value and packet loss rate. S3: Transmit the status information of the multiple self-organizing network nodes and the network topology diagram to the cloud computing platform; S4: On the cloud computing platform, the status information of the multiple self-organizing network nodes and the network topology are comprehensively analyzed to generate the current optimal transmission path; S5: Form a network based on the current optimal transmission path; S6: Monitor the movement status of the multiple self-organizing network nodes, update the status information of the multiple self-organizing network nodes, and repeat steps S2-S4 to update the current optimal transmission path for network formation.

2. The intelligent networking method for wireless ad hoc networks according to claim 1, characterized in that, The S4 includes: The status information of the multiple ad hoc network nodes and the network topology map are respectively feature encoded to obtain multiple ad hoc network node attribute feature vectors and network topology semantic feature maps; A node attribute-network topology comprehensive analysis is performed on the multiple ad hoc network node attribute feature vectors and the network topology semantic feature map to obtain multiple ad hoc network link state representation vectors containing network topology information. The multiple self-organizing network link state representation vectors containing network topology information are subjected to feature filtering based on key clues of self-organizing network links to obtain multiple filtered self-organizing network path state representation vectors containing network topology information. Based on the multiple filtered self-organizing network path state representation vectors containing network topology information, multiple path selection weight values ​​are obtained. The transmission path corresponding to the maximum value among the multiple path selection weight values ​​is selected as the current optimal transmission path.

3. The intelligent networking method for wireless ad hoc networks according to claim 2, characterized in that, The state information of the multiple ad hoc network nodes and the network topology graph are respectively feature-encoded to obtain multiple ad hoc network node attribute feature vectors and a network topology semantic feature graph, including: The state information of each self-organizing network node in the state information of the multiple self-organizing network nodes is encoded using a node state encoder based on a multilayer perceptron to obtain the attribute feature vector of the multiple self-organizing network nodes. The network topology graph is processed by a network topology feature extractor based on a DCNN model to obtain the network topology semantic feature graph.

4. The intelligent networking method for wireless ad hoc networks according to claim 3, characterized in that, Performing node attribute-network topology comprehensive analysis on the multiple ad hoc network node attribute feature vectors and the network topology semantic feature map to obtain multiple ad hoc network link state representation vectors containing network topology information includes: passing the multiple ad hoc network node attribute feature vectors and the network topology semantic feature map through a node attribute-network topology comprehensive analyzer based on graph attention networks to obtain the multiple ad hoc network link state representation vectors containing network topology information.

5. The intelligent networking method for wireless ad hoc networks according to claim 4, characterized in that, The multiple ad hoc network link state representation vectors containing network topology information are subjected to feature filtering based on key clues of ad hoc network links to obtain multiple filtered ad hoc network path state representation vectors containing network topology information, including: The plurality of ad hoc network link state representation vectors containing network topology information are rearranged to obtain a sequence of ad hoc network link state rearranged representation vectors containing network topology information; Extract the i-th reordered representation vector of the self-organizing network link state containing network topology information from the sequence of the self-organizing network link state reordered representation vectors containing network topology information as the representation vector of the self-organizing network link state to be analyzed. The first i-1 reordered representation vectors of the ad hoc link state containing network topology information are input into the key clue capture network to obtain the preorder distribution key clue guidance vector of the ad hoc link state containing network topology information. The key clue guidance vector of the preorder distribution of the self-organizing network link state containing network topology information and the characterization vector of the self-organizing network link state containing network topology information are input into the clue weaving gated network to obtain the self-organizing network link state clue weaving closed-loop strength factor containing network topology information. Based on the comparison between the closed-loop strength factor of the self-organizing network link state clue weaving containing network topology information and a preset threshold, it is determined whether to filter the self-organizing network link state characterization vector containing network topology information to be analyzed.

6. The intelligent networking method for wireless ad hoc networks according to claim 5, characterized in that, The plurality of ad hoc network link-state representation vectors containing network topology information are rearranged to obtain a sequence of rearranged ad hoc network link-state representation vectors containing network topology information, including: The maximum value of each self-organizing network link state representation vector containing network topology information is extracted from the plurality of self-organizing network link state representation vectors containing network topology information as a significant coarse descriptor to obtain a sequence of significant coarse descriptors of self-organizing network link state. Based on the ascending order of the sequence of the crude descriptors of the saliency of the ad hoc network link state, the multiple ad hoc network link state representation vectors containing network topology information are rearranged to obtain the sequence of rearranged representation vectors of the ad hoc network link state containing network topology information.

7. The intelligent networking method for wireless ad hoc networks according to claim 6, characterized in that, The first i-1 ad hoc link state rearrangement representation vectors containing network topology information are input into the key clue capture network to obtain the ad hoc link state preorder distribution key clue guidance vector containing network topology information, including: Calculate the saliency distribution representation value of each of the first i-1 ad hoc link state rearrangement representation vectors containing network topology information to obtain the saliency distribution representation value of the first i-1 ad hoc link states; Weighted normalization is performed on each of the first i-1 ad hoc link state saliency distribution representation values ​​to obtain the first i-1 normalized ad hoc link state saliency distribution representation values. Using the first i-1 normalized ad hoc network link state saliency distribution representation values ​​as a sequence of weight values, calculate the weighted sum of the first i-1 ad hoc network link state rearrangement representation vectors containing network topology information to obtain the ad hoc network link state preceding distribution key clue guidance vector containing network topology information.

8. The intelligent networking method for wireless ad hoc networks according to claim 7, characterized in that, The significance distribution of the ad hoc network link state rearrangement representation vector containing network topology information is related to the maximum value, mean, variance, and marginal value of the ad hoc network link state rearrangement representation vector containing network topology information. The marginal value is used to measure the global autonomous edge coordination of the ad hoc network link state and improve its contribution to the overall self-consistent mode.

9. The intelligent networking method for wireless ad hoc networks according to claim 8, characterized in that, Based on the multiple filtered self-organizing network path state representation vectors containing network topology information, multiple path selection weight values ​​are obtained, including: passing each of the multiple filtered self-organizing network path state representation vectors containing network topology information through a classifier-based path selector to obtain the multiple path selection weight values.

10. An intelligent networking system for wireless ad hoc networks, characterized in that, include: The node status information collection module is used to collect status information of multiple self-organizing network nodes, including location information, signal strength, and bandwidth load. The node status information transmission module is used to transmit the status information of the multiple self-organizing network nodes to the distributed edge computing module through the wireless module. The distributed edge computing module is used to construct a network topology map based on the status information of the multiple self-organizing network nodes. The edges of the network topology map are the link status parameters of each link in the network topology map. The link status parameters include signal attenuation value and packet loss rate. The node status network topology information transmission module is used to transmit the status information of the multiple self-organizing network nodes and the network topology map to the cloud computing platform. The current optimal transmission path generation module is used to perform comprehensive analysis on the cloud computing platform on the status information of the multiple self-organizing network nodes and the network topology to generate the current optimal transmission path. The networking module is used to form a network based on the current optimal transmission path; The network update module is used to monitor the movement status of the multiple self-organizing network nodes and update the status information of the multiple self-organizing network nodes. It repeatedly executes the node status information transmission module - current optimal transmission path generation module to update the current optimal transmission path for network formation.

Images