Time-space-spectrum three-domain collaborative bandwidth prediction method based on high-dynamic airborne network

Abstract

The invention relates to a time-space-spectrum three-domain collaborative bandwidth prediction method based on a high-dynamic airborne network. The method comprises the following steps: acquiring airborne multi-dimensional original data through a node of a preset airborne network; processing the multi-dimensional original data to obtain three-domain feature data; based on the three-domain feature data, constructing a time-space-spectrum three-domain collaborative prediction model, and performing three-domain prediction; constructing a judgment matrix through an analytic hierarchy process based on the three-domain prediction result, calculating a weight vector of a corresponding domain, performing consistency check, obtaining a weight distribution result, and performing weighted summation to generate a bandwidth prediction value; obtaining a bandwidth allocation scheme through a linear programming algorithm, and updating a flow table based on an SDN controller to obtain differentiated bandwidth resource allocation; and the deviation between the actual bandwidth of the network and the predicted value is monitored in real time, and the bandwidth is adjusted in real time by using a dynamic adjustment mechanism.

Description

Technical Field

[0001] This invention belongs to the field of airborne network communication technology, specifically relating to a time-space-spectrum three-domain collaborative bandwidth prediction method based on highly dynamic airborne networks. Background Technology

[0002] Currently, the cooperative bandwidth prediction for high-dynamic airborne networks still has the following areas for improvement: High-dynamic airborne networks, as the core support for aviation communications, face complex challenges such as high-speed equipment movement (flight speeds reaching Mach 6), frequent link topology switching, severe air-to-ground channel attenuation, and intense competition for spectrum resources. Existing bandwidth prediction methods generally suffer from three major limitations: First, they often focus on a single time or spatial dimension, ignoring the impact of dynamic changes in spectral resources on bandwidth, resulting in insufficient coordination across the three domains and difficulty adapting to the time-space-spectral coupling characteristics of airborne networks. Second, the data processing stage lacks dedicated optimization for airborne scenarios, failing to effectively filter random noise, spatial outliers, and spectral interference signals in the raw data, affecting feature extraction accuracy. Third, prediction models rely on fixed algorithms or single weight allocations, unable to dynamically adapt to high-dynamic scenarios such as airborne equipment movement and link switching, leading to large prediction errors and resulting in unbalanced bandwidth allocation, making it difficult to guarantee the QoS requirements of high-priority services. Furthermore, traditional bandwidth allocation uses a static configuration model, lacking dynamic linkage with prediction results, resulting in low resource utilization and an inability to quickly respond to bandwidth fluctuations, severely restricting the communication reliability and service quality of airborne networks. Summary of the Invention

[0003] To address the aforementioned problems in the existing technology, this invention provides a time-space-spectrum three-domain collaborative bandwidth prediction method based on highly dynamic airborne networks. The objective of this invention can be achieved through the following technical solutions: S1: Obtain multi-dimensional raw data of the airborne network through the nodes of the preset airborne network; S2: The multi-dimensional raw data is filtered and denoised in real time through edge computing nodes. The data is logically isolated by dual encapsulation, and redundant transmission is performed based on the result of the logical isolation. At the same time, a time-space-spectrum three-domain data matrix is ​​constructed, and the three-domain data matrix is ​​standardized to obtain three-domain feature data. S3: Based on the three-domain feature data, a spatiotemporal-spectral three-domain collaborative prediction model is constructed. The prediction model performs three-domain bandwidth prediction through statistical learning and analytical calculation. Based on the three-domain prediction results, a judgment matrix is ​​constructed using the analytic hierarchy process (AHP), the weight vectors of the corresponding domains are calculated, and a consistency check is performed to obtain the weight allocation results. Based on the weight allocation results, the three-domain prediction results are weighted and summed to generate the bandwidth prediction value. S4: Based on the bandwidth prediction value and the airborne network service priority division results, a multi-objective optimization scheduling model is constructed; the multi-objective optimization scheduling model uses preset corresponding indicators as constraints, obtains the bandwidth allocation scheme through linear programming algorithm, and updates the flow table to obtain differentiated bandwidth resource allocation; based on the bandwidth resource allocation, the bandwidth is adjusted in real time by monitoring the deviation between the actual network bandwidth and the prediction value and using a dynamic adjustment mechanism.

[0004] As a preferred embodiment of the present invention, the acquisition of multi-dimensional raw data of the airborne network includes: acquiring network operation status data based on the communication interface and data transmission link of the corresponding node of the airborne network; acquiring spatiotemporal dynamic data based on the positioning and motion sensing components carried by the airborne equipment; and acquiring spectral resource data based on spectral resource monitoring and signal detection. The network operation status data, spatiotemporal dynamic data and spectral resource data are integrated to obtain multi-dimensional raw data.

[0005] Specifically, the real-time filtering and denoising includes: using a sliding window threshold filtering method based on the multi-dimensional raw data to obtain the screening results of outliers; smoothing the random noise of the screening results; identifying and correcting isolated points in the spatial domain data through a neighborhood comparison verification mechanism; and using an adaptive filtering algorithm to remove noise data caused by interference frequency bands in combination with the frequency band characteristics of the spectral data, thereby obtaining standardized multi-dimensional raw data for real-time filtering and denoising.

[0006] Specifically, the process of logically isolating the data includes: constructing a three-dimensional logical isolation framework based on the three attributes of the multi-dimensional original data—time domain, spatial domain, and spectral domain—and combining the processing requirements and interaction boundaries of the corresponding domain data; and assigning a unique logical identifier and an independent storage partition to the corresponding domain data based on the three-dimensional logical isolation framework, while establishing a data tag mapping mechanism to bind data information with logical identifiers to obtain data isolation at the logical level.

[0007] Specifically, the redundant transmission process includes: dividing the data according to the application priority of the logically isolated three-domain data using a preset division standard; transmitting the three-domain data in real time using a corresponding data transmission mechanism based on the data division results; simultaneously establishing a real-time transmission status monitoring mechanism to monitor the real-time transmission process of the three-domain data and applying corresponding data repair methods to repair the data of the corresponding priority; and based on the data transmission, monitoring, and repair process, redundantly transmitting the three-domain data through multi-link parallelism and data sharding with multiple replicas.

[0008] Specifically, the process of obtaining the three-domain feature data includes: based on the temporal characteristics of the time domain data, dividing the continuous time-series data into time periods with similar characteristics through time series segmentation, and using periodic feature decomposition to identify the periodic change pattern of the data; integrating the corresponding attributes of the time domain data to obtain the time domain feature vector; Based on the spatial correlation and dynamism of spatial domain data, the spatial distance correlation between corresponding devices is calculated through device location correlation analysis, and a spatial domain feature matrix is ​​constructed through spatial dynamic feature extraction. Based on the resource attributes and interference characteristics of spectral domain data, the usage priority of corresponding frequency bands is determined by prioritizing frequency band resources. At the same time, the degree of matching between frequency bands and bandwidth requirements is evaluated by spectral adaptability analysis. A spectral domain feature set is generated by extracting spectral resource adaptability features. Based on the feature processing capabilities of edge computing nodes, the time-domain feature vector, spatial-domain feature matrix, and spectral-domain feature set are dimensionally aligned and then normalized to obtain three-domain feature data with a unified format.

[0009] Specifically, the process of predicting bandwidth through statistical learning and analytical calculation includes: using the time-domain feature vector to analyze the time dimension change pattern of bandwidth demand based on the time-domain feature vector, combining the intra-domain feature matching mechanism of historical bandwidth data and current time-domain features to obtain the time-domain bandwidth demand trend parameter, and obtaining the time-domain bandwidth prediction value based on the trend parameter. Based on the spatial correlation weight parsing model, the device location and topology are quantified. Combined with the device motion trend coefficient obtained by spatial dynamic feature extraction, the space and bandwidth are adapted and calculated to obtain the spatial dimension bandwidth adjustment parameters. The spatial domain bandwidth prediction value is obtained through parameter derivation. The matching efficiency between spectrum resources and bandwidth requirements is evaluated based on the spectral resource adaptation analysis algorithm. The interference suppression coefficient obtained by extracting spectral resource adaptation features is combined with the calculation of spectrum and bandwidth to obtain the spectral dimension bandwidth coefficient. Based on the bandwidth coefficient, the predicted value of spectral domain bandwidth is predicted.

[0010] Specifically, the calculation of the corresponding domain weight vector includes: obtaining a preliminary ranking of the corresponding domains through a domain priority initial judgment mechanism; constructing a comparison judgment matrix based on the preliminary ranking result; and calculating the weight vector of the corresponding domain through an eigenvector solving mechanism; verifying the compliance of the weight vector based on a consistency verification algorithm; and adjusting the scale of the judgment matrix based on the verification result through a weight vector iterative optimization mechanism to obtain a compliant three-domain weight vector.

[0011] Specifically, generating the bandwidth prediction value includes: matching the elements of the three-domain prediction results and the three-domain weight vector through a domain prediction and weight matching mechanism to obtain the weight ratio of the prediction value, constructing a three-domain prediction weighted fusion formula based on the weight ratio, and obtaining the initial fusion calculation result; An initial fusion model is constructed based on the initial fusion calculation results. The predicted values ​​of the three domains are integrated and calculated to output the initial bandwidth prediction results. In combination with the actual situation of the airborne network, a bandwidth prediction rationality verification rule is constructed to obtain the verification results. Based on the verification results, the model is fine-tuned through the historical deviation correction mechanism. At the same time, combined with the airborne network bandwidth constraints, a bandwidth prediction value adapted to the actual operating scenario is obtained.

[0012] Specifically, the process of obtaining a bandwidth allocation scheme includes: obtaining the optimization direction of the corresponding model based on the bandwidth prediction value and the service priority division result; converting the corresponding restrictions and requirements into quantitative constraints by integrating the constraints to obtain a model constraint set; calculating the allocation result of the corresponding service attributes based on the objective function of the constraint set using a linear programming algorithm; and supplementing the corresponding information based on the allocation result through an allocation scheme integration mechanism to obtain a bandwidth allocation scheme.

[0013] Specifically, the differentiated bandwidth resource allocation includes: based on the service priority in the bandwidth allocation scheme, implementing differentiated bandwidth allocation for corresponding priority services through a matching mechanism; converting the preset differentiated allocation rules into flow table information and distributing it to the corresponding nodes of the airborne network through the SDN controller; and simultaneously monitoring the actual bandwidth usage data of the corresponding services in real time to perform differentiated allocation of bandwidth resources.

[0014] Specifically, the process of adjusting bandwidth in real time using a dynamic adjustment mechanism includes: collecting actual bandwidth data of corresponding links and services in the airborne network based on a real-time network bandwidth monitoring mechanism; calculating the deviation rate based on the actual bandwidth data and the bandwidth prediction value; setting a deviation warning threshold and an adjustment threshold in combination with historical network deviation patterns to obtain a deviation judgment standard; performing differential processing based on the comparison results of the deviation judgment standard and the deviation rate; generating a corresponding allocation scheme and updating the flow table through the SDN controller to adjust the bandwidth in real time.

[0015] The beneficial effects of this invention are as follows: By integrating the temporal-spatial-spectral three-domain collaborative prediction framework, the temporal-domain time series patterns, spatial-domain dynamic characteristics, and spectral-domain resource attributes are integrated, breaking through the limitations of traditional single-domain prediction; combined with statistical learning and analytical calculation methods such as sliding window weighted averaging, spatial attenuation analysis, and spectral adaptation optimization, and with the analytic hierarchy process (AHP) to dynamically allocate the weights of the three domains, the prediction error rate is controlled, and the requirements of high-dynamic scenarios of airborne networks are accurately matched. Enhance data processing reliability: Employ multi-dimensional denoising strategies such as sliding window threshold filtering and adaptive filtering to effectively remove outliers, noise, and interference signals from the original data; Implement logical isolation and redundant transmission of data across three domains through VLAN+VXLAN dual encapsulation to ensure the security and integrity of data transmission and provide high-quality data support for feature extraction. Optimize bandwidth resource scheduling efficiency: Based on the prediction results, a multi-objective optimization scheduling model is constructed. Through the collaboration of linear programming algorithm and SDN controller, differentiated allocation of bandwidth slices is achieved, improving resource utilization by more than 30%. Combined with a dynamic adjustment mechanism, when the bandwidth deviation exceeds 15%, an adaptation strategy is quickly triggered to ensure service continuity and meet the QoS requirements of services with different priorities.

[0016] Achieve closed-loop optimization across the entire process: By adapting the feedback mechanism of execution data, dynamically adjust the parameters of the sensing probe and the feature extraction process to form a closed-loop system of data acquisition, processing, prediction, scheduling and feedback, continuously optimize the performance of the method, and significantly improve the communication stability and anti-interference capability of the airborne network in highly dynamic scenarios. Attached Figure Description

[0017] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0018] Figure 1 This is a flowchart illustrating the temporal-spatial-spectral three-domain collaborative bandwidth prediction method based on a highly dynamic airborne network according to the present invention. Figure 2 This is a schematic diagram of the collaborative prediction module in this invention. Detailed Implementation

[0019] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0020] Please see Figure 1-2 The temporal-spatial-spectral three-domain collaborative bandwidth prediction method based on highly dynamic airborne networks includes: S1: Obtain multi-dimensional raw data of the airborne network through the nodes of the preset airborne network; S2: The multi-dimensional raw data is filtered and denoised in real time through edge computing nodes. The data is logically isolated by dual encapsulation, and redundant transmission is performed based on the result of the logical isolation. At the same time, a time-space-spectrum three-domain data matrix is ​​constructed, and the three-domain data matrix is ​​standardized to obtain three-domain feature data. S3: Based on the three-domain feature data, a spatiotemporal-spectral three-domain collaborative prediction model is constructed. The prediction model performs three-domain bandwidth prediction through statistical learning and analytical calculation. Based on the three-domain prediction results, a judgment matrix is ​​constructed using the analytic hierarchy process (AHP), the weight vectors of the corresponding domains are calculated, and a consistency check is performed to obtain the weight allocation results. Based on the weight allocation results, the three-domain prediction results are weighted and summed to generate the bandwidth prediction value. S4: Based on the bandwidth prediction value and the airborne network service priority division results, a multi-objective optimization scheduling model is constructed; the multi-objective optimization scheduling model uses preset corresponding indicators as constraints, obtains the bandwidth allocation scheme through linear programming algorithm, and updates the flow table to obtain differentiated bandwidth resource allocation; based on the bandwidth resource allocation, the bandwidth is adjusted in real time by monitoring the deviation between the actual network bandwidth and the prediction value and using a dynamic adjustment mechanism.

[0021] As a preferred embodiment of the present invention, the acquisition of multi-dimensional raw data of the airborne network includes: acquiring network operation status data based on the communication interface and data transmission link of the corresponding node of the airborne network; acquiring spatiotemporal dynamic data based on the positioning and motion sensing components carried by the airborne equipment; and acquiring spectral resource data based on spectral resource monitoring and signal detection. The network operation status data, spatiotemporal dynamic data and spectral resource data are integrated to obtain multi-dimensional raw data.

[0022] In this embodiment, in a highly dynamic airborne network environment, the acquisition of multi-dimensional raw data needs to revolve around three core dimensions: network operation, spatiotemporal dynamics, and spectral resources, to adapt to the characteristics of scenarios involving high-speed device movement and frequent changes in link topology. From the perspective of network operation status data, it is necessary to rely on the communication interfaces and data transmission links of each node in the airborne network to collect key information reflecting the network operation status in real time, including data transmission rate, link packet loss rate, link latency, and node load rate. These data directly reflect the current carrying capacity, transmission reliability, and node processing saturation of the network, and are the network-level basis for subsequent bandwidth prediction. From the perspective of spatiotemporal dynamic data, it is necessary to continuously acquire data using the positioning and motion sensing components carried by the airborne equipment. The device's real-time geographic location coordinates and motion parameters; these data can accurately depict the device's motion trajectory and position change trends, and the device's spatiotemporal dynamic characteristics directly affect the link topology and signal transmission quality, making them the core raw data for spatial domain analysis; from the perspective of spectral resource data, it is necessary to collect information such as the frequency range of available spectrum, signal strength of each frequency band, interference signal strength, and frequency band occupancy status through spectral resource monitoring equipment and signal detection tools; the dynamic changes of spectral resources are directly related to the actual available bandwidth capacity, and this type of data is an indispensable input for spectral domain analysis; through the data integration module, the format differences of the three types of data are processed, standardized, and associated with a unified timestamp and spatial coordinate system to form a complete multi-dimensional raw dataset.

[0023] Specifically, the real-time filtering and denoising includes: using a sliding window threshold filtering method based on the multi-dimensional raw data to obtain the screening results of outliers; smoothing the random noise of the screening results; identifying and correcting isolated points in the spatial domain data through a neighborhood comparison verification mechanism; and using an adaptive filtering algorithm to remove noise data caused by interference frequency bands in combination with the frequency band characteristics of the spectral data, thereby obtaining standardized multi-dimensional raw data for real-time filtering and denoising.

[0024] In this embodiment, the high dynamic characteristics of airborne networks can easily lead to a large amount of interference information mixed into the original data, such as signal fluctuations, random noise, and spectral interference. Data quality needs to be ensured through multi-stage processing. A sliding window threshold filtering method is used to screen outliers. Based on historical data and typical scenario characteristics, normal threshold ranges are set for each data type. Data is monitored in real time in units of preset windows, and values ​​exceeding the threshold are marked. Continuous anomaly judgment rules are used to reduce false screening and ensure the reliability of the screening results. Random noise in the filtered data is processed using moving average or exponential weighted smoothing methods. The former weakens fluctuations by calculating the data mean, while the latter gives higher weight to recent data, smoothing noise while preserving the temporal trend, adapting to the dynamic needs of the network. A neighborhood comparison verification mechanism is used to identify and correct isolated points in the spatial domain. A device location association network is constructed to determine the neighborhood range of nodes. The data of the target node and neighboring nodes are compared. If the difference exceeds the spatial consistency threshold, it is judged as an isolated point. Correction is made based on the statistical characteristics of the neighborhood data to ensure the continuity of spatial data. Combining the frequency band characteristics of spectral data, an adaptive filtering algorithm is used to remove interference noise. First, the frequency band distribution is analyzed to identify the intervals susceptible to interference. Then, the filtering parameters are dynamically adjusted according to the characteristics of the interference signal to accurately filter the interference frequency band data, retain the effective signal information, and obtain standardized multi-dimensional original data.

[0025] Sliding window threshold filtering outlier formula: Let the sliding window length be N, and the i-th original data in the window be x. i (i=1,2,...,N), the mean of the data within the window is The standard deviation is The normal range of data within the window is: (k is the confidence coefficient, set according to the characteristics of airborne network data), the specific formula is as follows: , , If data x j (j) [1,N]) satisfies ( ) or( If x is determined, then x is determined. j This is an outlier.

[0026] Let the original data at time t be x. t The smoothed data is s t The smoothing coefficient is ,but: , s0=x0, where x0 is the original data at the initial moment.

[0027] Specifically, the process of logically isolating the data includes: constructing a three-dimensional logical isolation framework based on the three attributes of the multi-dimensional original data—time domain, spatial domain, and spectral domain—and combining the processing requirements and interaction boundaries of the corresponding domain data; and assigning a unique logical identifier and an independent storage partition to the corresponding domain data based on the three-dimensional logical isolation framework, while establishing a data tag mapping mechanism to bind data information with logical identifiers to obtain data isolation at the logical level.

[0028] In this embodiment, a three-dimensional logical isolation framework is constructed. Using domain attributes, processing requirements, and interaction boundaries as coordinate axes, the framework structure is designed based on the characteristics of data in each domain: temporal domain data emphasizes temporal continuity and real-time performance, requiring avoidance of temporal breaks caused by processing delays; spatial domain data emphasizes spatial correlation and dynamism, requiring association with temporal domain data but clearly defining interaction boundaries with the spectral domain; spectral domain data focuses on resource exclusivity and interference sensitivity, only interacting with other domains to a limited extent during resource adaptation, thereby clarifying the positioning and constraints of each domain.

[0029] Each domain's data is assigned a unique logical identifier containing the domain type, source, and generation time for easy traceability and management. The storage system is divided into independent partitions for the three domains, with dedicated access control policies. Virtual isolation technology ensures the logical independence of each partition, preventing mutual impact from faults or anomalies. A data tag mapping mechanism is established; core information is extracted from each data entry and associated with the logical identifier to generate structured tags. A tag index is built to enable fast querying and reverse tracing. This mechanism ensures that data always carries the domain ownership identifier during storage, transmission, and processing, avoiding confusion, strengthening access control, and guaranteeing the effectiveness of logical isolation.

[0030] Specifically, the redundant transmission process includes: dividing the data according to the application priority of the logically isolated three-domain data using a preset division standard; transmitting the three-domain data in real time using a corresponding data transmission mechanism based on the data division results; simultaneously establishing a real-time transmission status monitoring mechanism to monitor the real-time transmission process of the three-domain data and applying corresponding data repair methods to repair the data of the corresponding priority; and based on the data transmission, monitoring, and repair process, redundantly transmitting the three-domain data through multi-link parallelism and data sharding with multiple replicas.

[0031] In this embodiment, prioritization rules are established based on the airborne network service scenario and the degree of dependence of services on bandwidth prediction: real-time time-series data, key motion trajectory data, and core parameter data of available frequency bands that affect the accuracy of short-term predictions are classified as high priority; non-real-time historical data, non-critical position deviation data, and non-core frequency band interference data are classified as medium and low priority, and the classification process is dynamically calibrated by the network management module in conjunction with service load; a more reliable transmission method is selected for high-priority data to ensure that the data arrives in real time and completely; an efficiency-first transmission method is adopted for medium and low priority data to save network resources while ensuring basic reliability; the data transmission status of the three domains is tracked in real time, and information such as transmission rate and packet loss is collected; if abnormal data transmission of high-priority data is detected, data repair measures are immediately initiated to quickly restore data integrity; for medium and low priority data, selective repair is performed according to service tolerance to balance reliability and transmission efficiency.

[0032] Specifically, the process of obtaining the three-domain feature data includes: based on the temporal characteristics of the time-domain data, dividing the continuous time-series data into time periods with similar characteristics through time series segmentation, and using periodic feature decomposition to identify the periodic change pattern of the data; integrating the corresponding attributes of the time-domain data to obtain the time-domain feature vector. Based on the spatial correlation and dynamism of spatial domain data, the spatial distance correlation between corresponding devices is calculated through device location correlation analysis, and a spatial domain feature matrix is ​​constructed through spatial dynamic feature extraction. Based on the resource attributes and interference characteristics of spectral domain data, the usage priority of corresponding frequency bands is determined by prioritizing frequency band resources. At the same time, the degree of matching between frequency bands and bandwidth requirements is evaluated by spectral adaptability analysis. A spectral domain feature set is generated by extracting spectral resource adaptability features. Based on the feature processing capabilities of edge computing nodes, the time-domain feature vector, spatial-domain feature matrix, and spectral-domain feature set are dimensionally aligned and then normalized to obtain three-domain feature data with a unified format.

[0033] In this embodiment, based on the temporal characteristics of time-domain data, time series segmentation technology is used to divide continuous data into time periods with similar characteristics. At the same time, periodic feature decomposition is used to identify the periodic change pattern of the data, and the time-series segment features and periodic patterns are integrated to form a time-domain feature vector that can reflect the changing trend of the time dimension.

[0034] Time-domain periodic eigenvalue decomposition formula (Fourier transform to extract period): , By analyzing the peak frequency f0 of X(f), the data period T can be obtained. Then, construct the time-domain feature vector. , The average value of the data within the period. This represents the standard deviation of the data within the period.

[0035] To address the correlation and dynamism of spatial domain data, the spatial distance correlation between devices is calculated through device location correlation analysis. Spatial dynamic features are extracted by combining device motion parameters. The correlation and dynamic features are integrated to construct a spatial domain feature matrix, which reflects the spatial distribution and dynamic relationship of devices.

[0036] Based on the resource attributes and interference characteristics of spectral domain data, the usage priority of each frequency band is determined by prioritizing frequency band resources. The degree of matching between frequency band and bandwidth requirements is evaluated by using spectrum adaptability analysis. Key parameters related to priority and adaptability are extracted to generate a spectral domain feature set that reflects the availability and adaptability of spectrum resources.

[0037] By leveraging the processing power of edge computing nodes, the three-domain features are dimensionally aligned, and the differences in data dimensions are eliminated through normalization, resulting in a unified three-domain feature data that can be directly used as model input.

[0038] Specifically, the process of predicting bandwidth through statistical learning and analytical calculation includes: using the time-domain feature vector to analyze the time dimension change pattern of bandwidth demand based on the time-domain feature vector, combining the intra-domain feature matching mechanism of historical bandwidth data and current time-domain features to obtain the time-domain bandwidth demand trend parameter, and obtaining the time-domain bandwidth prediction value based on the trend parameter. Based on the spatial correlation weight parsing model, the device location and topology are quantified. Combined with the device motion trend coefficient obtained by spatial dynamic feature extraction, the space and bandwidth are adapted and calculated to obtain the spatial dimension bandwidth adjustment parameters. The spatial domain bandwidth prediction value is obtained through parameter derivation. The matching efficiency between spectrum resources and bandwidth requirements is evaluated based on the spectral resource adaptation analysis algorithm. The interference suppression coefficient obtained by extracting spectral resource adaptation features is combined with the calculation of spectrum and bandwidth to obtain the spectral dimension bandwidth coefficient. Based on the bandwidth coefficient, the predicted value of spectral domain bandwidth is predicted.

[0039] In this embodiment, based on time-domain feature vectors, a time-series feature evolution mining algorithm is used to analyze the temporal dimension variation pattern of bandwidth demand. Intra-domain feature matching is performed by combining historical bandwidth data with current time-series features to determine bandwidth demand trend parameters. Based on these trend parameters, a time-domain bandwidth prediction value is derived, reflecting the bandwidth change trend over time. A spatial correlation weight analysis model quantifies device location and network topology. Combined with device motion trend coefficients extracted from spatial dynamic features, spatial parameters are adapted to bandwidth demand for calculation, obtaining spatial dimension bandwidth adjustment parameters. Spatial domain bandwidth prediction values ​​are derived through parameter derivation, reflecting the impact of spatial dynamics on bandwidth. A spectral resource adaptation analysis algorithm is used to evaluate the matching efficiency between spectrum resources and bandwidth demand. Combined with interference suppression coefficients extracted from spectral domain features, a spectral dimension bandwidth coefficient is obtained through calculation of spectrum parameters and bandwidth demand. Based on this bandwidth coefficient, a spectral domain bandwidth prediction value is generated, reflecting the spectrum resources' ability to support bandwidth.

[0040] Specifically, the calculation of the corresponding domain weight vector includes: obtaining a preliminary ranking of the corresponding domains through a domain priority initial judgment mechanism; constructing a comparison judgment matrix based on the preliminary ranking result; and calculating the weight vector of the corresponding domain through an eigenvector solving mechanism; verifying the compliance of the weight vector based on a consistency verification algorithm; and adjusting the scale of the judgment matrix based on the verification result through a weight vector iterative optimization mechanism to obtain a compliant three-domain weight vector.

[0041] Specifically, generating the bandwidth prediction value includes: matching the elements of the three-domain prediction results and the three-domain weight vector through a domain prediction and weight matching mechanism to obtain the weight ratio of the prediction value, constructing a three-domain prediction weighted fusion formula based on the weight ratio, and obtaining the initial fusion calculation result; An initial fusion model is constructed based on the initial fusion calculation results. The predicted values ​​of the three domains are integrated and calculated to output the initial bandwidth prediction results. In combination with the actual situation of the airborne network, a bandwidth prediction rationality verification rule is constructed to obtain the verification results. Based on the verification results, the model is fine-tuned through the historical deviation correction mechanism. At the same time, combined with the airborne network bandwidth constraints, a bandwidth prediction value adapted to the actual operating scenario is obtained.

[0042] In this embodiment, the generation of bandwidth prediction values ​​requires fusing the three-domain prediction results with the weight vector, and verifying and correcting them in conjunction with the actual scenario to ensure that the results fit the actual needs of the airborne network. Through the domain prediction and weight matching mechanism, the prediction results of the time domain, spatial domain, and spectral domain are matched one-to-one with the corresponding weight vector elements to clarify the weight ratio of each domain prediction value. Based on the weight ratio, a three-domain prediction weighted fusion formula is constructed, and the three-domain prediction results and weight vector are substituted into it to calculate the initial bandwidth prediction result. Combining the actual operation of the airborne network, a bandwidth prediction value rationality verification rule is constructed. The initial prediction result is substituted into the rule for verification. If it does not conform to the rule, the initial result is fine-tuned through the historical deviation correction mechanism. At the same time, the bandwidth constraints of the airborne network are taken into account to obtain a bandwidth prediction value that is adapted to the actual operation scenario and meets the business requirements.

[0043] Weighted fusion formula: , The predicted values ​​in the time domain, spatial domain, and spectral domain are B, respectively. t B s B p The corresponding weights are wt and w s w p (w) t +w s +w p =1), initial bandwidth prediction value B init ).

[0044] Specifically, the process of obtaining a bandwidth allocation scheme includes: obtaining the optimization direction of the corresponding model based on the bandwidth prediction value and the service priority division result; converting the corresponding restrictions and requirements into quantitative constraints by integrating the constraints to obtain a model constraint set; calculating the allocation result of the corresponding service attributes based on the objective function of the constraint set using a linear programming algorithm; and supplementing the corresponding information based on the allocation result through an allocation scheme integration mechanism to obtain a bandwidth allocation scheme.

[0045] Specifically, the differentiated bandwidth resource allocation includes: based on the service priority in the bandwidth allocation scheme, implementing differentiated bandwidth allocation for corresponding priority services through a matching mechanism; converting the preset differentiated allocation rules into flow table information and distributing it to the corresponding nodes of the airborne network through the SDN controller; and simultaneously monitoring the actual bandwidth usage data of the corresponding services in real time to perform differentiated allocation of bandwidth resources.

[0046] In this embodiment, based on the service priority ranking specified in the bandwidth allocation scheme, differentiated resources are allocated to services of different priorities through a priority and bandwidth matching mechanism: high-priority services (such as flight control and navigation data transmission) are allocated sufficient bandwidth to ensure real-time transmission and reliability, and some redundant bandwidth can be reserved to cope with sudden demands; medium and low-priority services (such as multimedia playback and non-critical monitoring data transmission) are flexibly allocated according to the remaining bandwidth while meeting basic requirements, so as to avoid occupying high-priority service resources.

[0047] The pre-defined differentiated allocation rules (e.g., bandwidth thresholds for high-priority services, bandwidth adjustment ranges for medium and low-priority services) are converted into flow table information (including service identifiers, bandwidth limits, forwarding rules, etc.) recognizable by the SDN controller. This flow table information is then distributed to the corresponding nodes in the onboard network (e.g., switches, routers, communication terminals) via the SDN controller's southbound interface, ensuring that each node executes bandwidth allocation according to the rules. Real-time monitoring and dynamic adjustment are implemented. The network monitoring module collects real-time data on actual bandwidth usage for each service (e.g., bandwidth utilization, traffic trends). If insufficient bandwidth for high-priority services or wasted bandwidth for medium and low-priority services is detected, this information is promptly fed back to the SDN controller. The controller then fine-tunes the flow table information based on the monitoring data, dynamically optimizing bandwidth allocation and ensuring the continued effectiveness of the differentiated strategy.

[0048] Specifically, the process of adjusting bandwidth in real time using a dynamic adjustment mechanism includes: collecting actual bandwidth data of corresponding links and services in the airborne network based on a real-time network bandwidth monitoring mechanism; calculating the deviation rate based on the actual bandwidth data and the bandwidth prediction value; setting a deviation warning threshold and an adjustment threshold in combination with historical network deviation patterns to obtain a deviation judgment standard; performing differential processing based on the comparison results of the deviation judgment standard and the deviation rate; generating a corresponding allocation scheme and updating the flow table through the SDN controller to adjust the bandwidth in real time.

[0049] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A time-space-spectral three-domain collaborative bandwidth prediction method based on highly dynamic airborne networks, characterized in that, include: S1: Obtain multi-dimensional raw data of the airborne network through the nodes of the preset airborne network; S2: The multi-dimensional raw data is filtered and denoised in real time through edge computing nodes. The data is logically isolated by dual encapsulation, and redundant transmission is performed based on the result of the logical isolation. At the same time, a time-space-spectrum three-domain data matrix is ​​constructed, and the three-domain data matrix is ​​standardized to obtain three-domain feature data. S3: Based on the three-domain feature data, construct a time-space-spectrum three-domain collaborative prediction model. The prediction model performs three-domain prediction of bandwidth through statistical learning and analytical calculation. Based on the three-domain prediction results, a judgment matrix is ​​constructed using the analytic hierarchy process (AHP), the weight vectors of the corresponding domains are calculated, and a consistency check is performed to obtain the weight allocation results. Based on the weight allocation result, the three-domain prediction results are weighted and summed to generate a bandwidth prediction value; S4: Based on the bandwidth prediction value and the airborne network service priority division results, construct a multi-objective optimization scheduling model; The multi-objective optimization scheduling model uses preset indicators as constraints, obtains bandwidth allocation schemes through linear programming algorithms, and updates flow tables to obtain differentiated bandwidth resource allocations. Based on the bandwidth resource allocation, the model monitors the deviation between the actual network bandwidth and the predicted value, and uses a dynamic adjustment mechanism to adjust the bandwidth in real time.

2. The method according to claim 1, characterized in that, The acquisition of multi-dimensional raw data from the airborne network includes: acquiring network operation status data based on the communication interfaces and data transmission links of the corresponding nodes in the airborne network; and acquiring spatiotemporal dynamic data based on the positioning and motion sensing components carried by the airborne equipment. Based on spectral resource monitoring and signal detection, spectral resource data is acquired; the network operation status data, spatiotemporal dynamic data and spectral resource data are integrated to obtain multi-dimensional raw data.

3. The method according to claim 1, characterized in that, The real-time filtering and denoising includes: using a sliding window threshold filtering method based on the multi-dimensional raw data to obtain the screening results of outliers; smoothing the random noise of the screening results; identifying and correcting isolated points in the spatial domain data through a neighborhood comparison verification mechanism; and using an adaptive filtering algorithm to remove noise data caused by interference frequency bands in combination with the frequency band characteristics of the spectral data, thereby obtaining standardized multi-dimensional raw data for real-time filtering and denoising.

4. The method according to claim 1, characterized in that, The specific process of logically isolating the data includes: constructing a three-dimensional logical isolation framework based on the three attributes of the multi-dimensional original data—time domain, spatial domain, and spectral domain—and combining the processing requirements and interaction boundaries of the corresponding domain data; based on the three-dimensional logical isolation framework, assigning a unique logical identifier and an independent storage partition to the corresponding domain data, and establishing a data tag mapping mechanism to bind data information with logical identifiers, thereby achieving data isolation at the logical level.

5. The method according to claim 1, characterized in that, The specific process of redundant transmission includes: dividing the data according to the application priority of the logically isolated three-domain data using a preset division standard; transmitting the three-domain data in real time using the corresponding data transmission mechanism based on the data division results; simultaneously establishing a real-time transmission status monitoring mechanism to monitor the real-time transmission process of the three-domain data and applying corresponding data repair methods to repair the data of the corresponding priority; and based on the data transmission, monitoring, and repair process, redundantly transmitting the three-domain data through multi-link parallelism and data sharding with multiple replicas.

6. The method according to claim 1, characterized in that, The specific process of obtaining the three-domain feature data includes: dividing the continuous time series data into time periods with similar characteristics by time series segmentation, and at the same time using periodic feature decomposition to identify the periodic change pattern of the data to obtain the time domain feature vector. Based on the spatial correlation and dynamism of spatial domain data, the spatial distance correlation between corresponding devices is calculated through device location correlation analysis, and a spatial domain feature matrix is ​​constructed through spatial dynamic feature extraction. Frequency band usage priority is obtained by prioritizing frequency band resources, and the degree of matching between frequency bands and bandwidth requirements is evaluated by spectrum adaptability analysis. A spectral domain feature set is generated by extracting spectral resource adaptability features. The time-domain feature vector, spatial-domain feature matrix, and spectral-domain feature set are dimensionally aligned and normalized to obtain three-domain feature data with a unified format.

7. The method according to claim 1, characterized in that, The specific process of predicting bandwidth through statistical learning and analytical calculation includes: based on the time domain feature vector and combined with historical data and current time series features, obtaining time domain bandwidth demand trend parameters and predicting values ​​through a time series feature evolution mining algorithm; quantifying device location and topology through a spatial correlation weight analysis model, and combining device motion trend coefficient adaptation calculation to obtain spatial domain bandwidth prediction values; evaluating resource matching efficiency based on a spectral resource adaptation analysis algorithm, and predicting spectral domain bandwidth values ​​by combining interference suppression coefficient calculation.

8. The method according to claim 1, characterized in that, The calculation of the corresponding domain weight vector includes: obtaining a preliminary ranking of the corresponding domains through a domain priority initial judgment mechanism; constructing a comparison judgment matrix based on the preliminary ranking result; and calculating the weight vector of the corresponding domain through an eigenvector solving mechanism; verifying the compliance of the weight vector based on a consistency verification algorithm; and adjusting the scale of the judgment matrix based on the verification result through a weight vector iterative optimization mechanism to obtain a compliant three-domain weight vector.

9. The method according to claim 1, characterized in that, The generation of bandwidth prediction values ​​includes: matching the elements of the three-domain prediction results and the three-domain weight vector through a domain prediction and weight matching mechanism to obtain the weight ratio of the prediction values, constructing a three-domain prediction weighted fusion formula based on the weight ratio, and obtaining the initial fusion calculation results; An initial fusion model is constructed based on the initial fusion calculation results. The predicted values ​​of the three domains are integrated and calculated to output the initial bandwidth prediction results. In combination with the actual situation of the airborne network, a bandwidth prediction rationality verification rule is constructed to obtain the verification results. Based on the verification results, the model is fine-tuned through the historical deviation correction mechanism. At the same time, combined with the airborne network bandwidth constraints, a bandwidth prediction value adapted to the actual operating scenario is obtained.

10. The method according to claim 1, characterized in that, The process of obtaining a bandwidth allocation scheme includes: obtaining the optimization direction of the corresponding model based on the bandwidth prediction value and the service priority division result; converting the corresponding restrictions and requirements into quantitative constraints by integrating the constraints to obtain a model constraint set; calculating the allocation result of the corresponding service attributes based on the objective function of the constraint set using a linear programming algorithm; and supplementing the corresponding information based on the allocation result through an allocation scheme integration mechanism to obtain a bandwidth allocation scheme.

11. The method according to claim 1, characterized in that, The differentiated bandwidth resource allocation includes: based on the service priority in the bandwidth allocation scheme, implementing differentiated bandwidth allocation for corresponding priority services through a matching mechanism; converting the preset differentiated allocation rules into flow table information and distributing it to the corresponding nodes of the airborne network through the SDN controller; and simultaneously monitoring the actual bandwidth usage data of the corresponding services in real time to perform differentiated allocation of bandwidth resources.

12. The method according to claim 1, characterized in that, The specific process of adjusting bandwidth in real time using a dynamic adjustment mechanism includes: collecting actual bandwidth data of corresponding links and services in the airborne network based on a real-time network bandwidth monitoring mechanism; calculating the deviation rate based on the actual bandwidth data and the bandwidth prediction value; setting a deviation warning threshold and an adjustment threshold in combination with historical network deviation patterns to obtain a deviation judgment standard; performing differential processing based on the comparison results of the deviation judgment standard and the deviation rate; generating a corresponding allocation scheme and updating the flow table through the SDN controller to adjust the bandwidth in real time.

Images