Aviation airborne miniature 5G ad hoc network multimode communication terminal

Abstract

The invention relates to the technical field of communication, in particular to an aviation airborne miniature 5G ad hoc network multimode communication terminal, which comprises a flight state prediction module, a state vector and time step length linear transformation relation is constructed by adopting a kinematics equation based on real-time GPS position, velocity vector, acceleration and attitude angle data of the terminal and neighbor nodes, and the state vector and time step length linear transformation relation is obtained; and by using the residual error between the newly received kinematics data and the predicted value at the previous moment, weighting the correction state vector and the predicted covariance, circulating a recursive correction process, outputting three-dimensional space coordinates of multiple nodes under a plurality of time steps in the future, and obtaining a future position set of the nodes. According to the method, the basis of routing decision is changed by directly processing real-time GPS position, velocity vector and attitude angle data and adopting three-dimensional space coordinates of continuous nodes of a kinematics equation in a future time window. It no longer relies on instantaneous snapshots of the current network state, but is based on future predictions.

Description

Technical Field

[0001] The present invention relates to the field of communication technologies, and in particular to an aviation-borne micro 5G self-organizing network multi-mode communication terminal. Background Art

[0002] The field of communications technology is a branch of science and engineering that studies and applies information transmission and exchange. Its core is to effectively, reliably, and securely transmit information from the source to the destination through various transmission media.

[0003] Existing communication technologies exhibit design limitations when dealing with high-speed node mobility and complex electromagnetic environments. Core weaknesses lie in their passive response mechanisms and incomplete environmental perception. For routing selection, traditional mobile ad hoc network protocols generally rely on passive link quality detection, such as periodic signaling exchanges or monitoring the success or failure of data transmissions to determine link connectivity. This mechanism is essentially a "post-event" approach, initiating route reconstruction only when link quality deteriorates to the point of communication interruption. In scenarios like aeronautical ad hoc networks, where topology changes rapidly at the speed of seconds or even milliseconds, this response delay is critical. By the time a new route is calculated, some of the links that comprise it may already be on the verge of disruption, leading the network into a vicious cycle of frequent route reconstructions, resulting in signaling overhead and persistent communication interruptions. Therefore, improvements are needed. Summary of the Invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and propose an aviation airborne micro 5G self-organizing network multi-mode communication terminal.

[0005] In order to achieve the above-mentioned purpose, the present invention adopts the following technical solution: an aviation airborne micro 5G self-organizing network multi-mode communication terminal comprises: The flight state prediction module uses the kinematic equation to construct a linear transformation relationship between the state vector and the time step based on the real-time GPS position, velocity vector, acceleration, and attitude angle data of itself and its neighboring nodes. It then uses the residual between the newly received kinematic data and the predicted value at the previous moment to weightedly correct the state vector and prediction covariance. The correction process is then repeated recursively to output the three-dimensional spatial coordinates of multiple nodes at multiple future time steps, thereby obtaining a set of future node positions. The link potential field construction module calculates the Euclidean distance between nodes based on the future position set of the nodes, combines the preset airborne antenna gain and signal frequency, substitutes the path loss formula to obtain the signal reception strength, and then obtains the expected bandwidth by querying the mapping relationship table between signal strength and bandwidth. At the same time, the body attitude angle data is used to determine whether the line of sight vector between nodes intersects with the body three-dimensional model. If there is an intersection, the expected bandwidth is set to zero. Finally, the inverse of the expected bandwidth is used as the link potential energy value to construct the network topology potential field data.

[0006] Preferably, the terminal further includes: The spectrum feature perception module periodically scans the broadband spectrum through the RF front end, performs Fourier transform on the collected I / Q sample sequence, and multiplies the transform result with the frequency-shifted complex conjugate version of the transform result point by point and performs time-smoothing averaging to generate a two-dimensional matrix of cyclic frequency and spectrum frequency. The module searches for local maximum points whose amplitude exceeds the preset detection threshold on the two-dimensional matrix, combines the coordinates and amplitude of the maximum points into a fixed-length vector, and obtains the signal fingerprint feature vector. The communication avoidance planning module integrates multiple signal fingerprint feature vectors and the direction information indicated by the multi-antenna phase difference, determines the coordinates of the interference source by solving the hyperbolic equation group formed by the arrival time difference of multiple nodes, marks the interference source influence range and the high potential energy links in the network topology potential field data as prohibited, and starts from the source node and expands layer by layer to the adjacent node with the lowest cumulative potential energy value until it reaches the destination node to form a path. Then, each link in the path is sequentially assigned a channel number in the spectrum resource pool that is not occupied by neighboring links and not covered by interference, and an adaptive communication routing strategy is established.

[0007] Preferably, the flight status prediction module includes: The state vector construction submodule combines the position, velocity, and acceleration components into a state vector based on the real-time GPS position, velocity vector, acceleration, and attitude angle data of itself and its neighboring nodes. It also sets the state transfer matrix based on the constant acceleration motion assumption. The matrix linearly maps the current state vector to the predicted state of the next time step to obtain the initial motion state model. A prediction residual calculation submodule, based on the initial motion state model, obtains a measurement residual by subtracting the newly received kinematic data from the model prediction value, determines a gain factor by calculating the ratio of the prediction covariance to the measurement noise covariance, and then uses the gain factor to perform a weighted correction on the state vector and synchronously updates the prediction covariance matrix to obtain a corrected state covariance; The coordinate iteration prediction submodule uses the corrected state vector as the input for the next iteration based on the corrected state covariance, repeatedly executes the closed-loop calculation process of state prediction and measurement update, and generates a sequence of three-dimensional coordinates of multiple nodes at future moments by continuously extrapolating multiple time steps to obtain a set of future node positions.

[0008] Preferably, the link potential field construction module includes: The link bandwidth prediction submodule calculates the second norm of the difference between the three-dimensional coordinates of any two nodes based on the future node position set to obtain the Euclidean distance. It then calculates the free space path loss based on the preset airborne antenna gain and signal frequency parameters and subtracts the loss value to obtain the signal reception strength. It then uses the strength value as an index to search and output the corresponding bandwidth value in a preset signal strength and bandwidth mapping table to obtain the original link bandwidth data. The occlusion effect judgment submodule rotates and translates the three-dimensional mesh model of the body according to the attitude angle data based on the original link bandwidth data and the body attitude angle data, constructs the line of sight vector line segments between the nodes, and determines whether the line segments penetrate any triangular facet in the body model through the ray and triangular mesh intersection test. If penetration occurs, the corresponding expected bandwidth value is cleared to zero to obtain the corrected link bandwidth data; The potential link conversion submodule, based on the corrected link bandwidth data, traverses the expected bandwidth value of each link in the data, calculates the inverse and assigns it as the link potential energy value, aggregates all link potential energy values ​​into a weighted graph with nodes as vertices and links as edges, and constructs network topology potential field data.

[0009] Preferably, the spectrum feature perception module includes: The cyclic spectrum matrix generation submodule periodically scans the broadband spectrum through the RF front end, performs a short-time Fourier transform on the collected I / Q sample sequence, and multiplies the transformed result matrix element-by-element with its own frequency-shifted complex conjugate version. It then performs a sliding average along the time axis to generate a two-dimensional matrix of cyclic frequencies and spectrum frequencies, thereby obtaining a two-dimensional spectrum correlation matrix. The spectral peak feature extraction submodule sets an amplitude threshold based on the two-dimensional spectral correlation matrix and traverses each element in the matrix. If the amplitude of an element is greater than the amplitude threshold and also greater than the amplitudes of eight adjacent elements, the position coordinates and amplitude of the element are recorded to obtain the key spectral peak coordinate set.

[0010] Preferably, the spectrum feature perception module further includes: The fingerprint vector formatting submodule selects several points with the highest peak amplitude based on the key peak coordinate set, arranges the cyclic frequency coordinates, spectrum frequency coordinates and amplitude values ​​in sequence, and fills them into a vector of predefined length. If the number of peaks is insufficient, the remaining positions are filled with zeros to obtain the signal fingerprint feature vector.

[0011] Preferably, the communication avoidance planning module includes: The interference source positioning submodule integrates multiple signal fingerprint feature vectors and the direction information indicated by the multi-antenna phase difference, extracts the arrival timestamps of the same interference signal at different nodes, and uses the differences between multiple timestamps to construct a nonlinear hyperbolic equation system with the interference source location as the unknown variable. The system of equations is solved to obtain the spatial coordinate set of the interference source. The comprehensive routing cost calculation submodule, based on the interference source spatial coordinate set and network topology potential field data, marks the circular area with the interference source coordinates as the center and the influence radius as the boundary and the links with potential energy values ​​higher than the threshold as no-entry areas, uses an extended queue to start from the source node, iteratively expands to the node with the lowest cumulative potential energy among the adjacent nodes, and records the extended path until it reaches the destination node, thereby establishing the optimal communication path.

[0012] Preferably, the communication avoidance planning module further includes: The spectrum channel allocation submodule, based on the preferred communication path, allocates the first channel in the available channel list to the first link in the path, traverses subsequent links, and allocates a channel number to each link that does not conflict with the allocated channels on the path, the channels being used by neighboring nodes, and the channels covered by the interference area, thereby establishing an adaptive communication routing strategy.

[0013] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, by directly processing real-time GPS position, velocity vector and attitude angle data, and using the three-dimensional spatial coordinates of continuous nodes in the future time window of the kinematic equation, the basis of routing decisions is changed. It no longer relies on an instantaneous snapshot of the current network status, but is based on future predictions. In particular, the body attitude angle data is introduced into the calculation to determine the physical obstruction of the signal by the wing or fuselage. This makes the link quality assessment no longer limited to a single distance loss model, but constructs a network topology potential field data that comprehensively reflects the future connection stability and physical accessibility. This mechanism of actively avoiding links that are about to fail or are blocked greatly improves the routing stability and data transmission success rate under high-maneuverability flight conditions. At the same time, by deeply analyzing the I / Q sample sequence to generate a two-dimensional matrix of cyclic frequency and spectral frequency, and extracting the key spectral peak coordinate set from it to form a signal fingerprint feature vector, a refined "portrait" of the spectrum environment is achieved. It can identify the intrinsic modulation characteristics of the signal and distinguish different types of signal sources. This spectrum situation is combined with the spatial coordinate set of the interference source calculated through the arrival time difference of multiple nodes and incorporated into the routing planning, so that the terminal can not only predict physical topology changes but also intelligently avoid electromagnetic interference. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a terminal flow chart of the present invention. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0016] See also Figure 1 The present invention provides a technical solution: an aviation airborne micro 5G self-organizing network multi-mode communication terminal comprising: The flight state prediction module uses the kinematic equation to construct a linear transformation relationship between the state vector and the time step based on the real-time GPS position, velocity vector, acceleration, and attitude angle data of itself and its neighboring nodes. It then uses the residual between the newly received kinematic data and the predicted value at the previous moment to weightedly correct the state vector and prediction covariance. The correction process is then repeated recursively to output the three-dimensional spatial coordinates of multiple nodes at multiple future time steps, thereby obtaining a set of future node positions. The link potential field construction module calculates the Euclidean distance between nodes based on the set of future node positions. Combining the preset airborne antenna gain and signal frequency, it substitutes the path loss formula to obtain the signal reception strength. The module then obtains the expected bandwidth by querying the mapping table between signal strength and bandwidth. The module also uses the aircraft attitude angle data to determine whether the line of sight vectors between nodes intersect with the aircraft's three-dimensional model. If so, the expected bandwidth is set to zero. Finally, the inverse of the expected bandwidth is used as the link potential energy value to construct the network topology potential field data. The spectrum feature perception module periodically scans the broadband spectrum through the RF front end, performs Fourier transform on the collected I / Q sample sequence, and multiplies the transform result with the frequency-shifted complex conjugate version of the transform result point by point and performs time-smoothing averaging to generate a two-dimensional matrix of cyclic frequency and spectrum frequency. The module searches for local maximum points whose amplitude exceeds the preset detection threshold on the two-dimensional matrix, combines the coordinates and amplitude of the maximum points into a fixed-length vector, and obtains the signal fingerprint feature vector. The communication avoidance planning module integrates the directional information indicated by multiple signal fingerprint feature vectors and multi-antenna phase differences, determines the coordinates of the interference source by solving the hyperbolic equation group formed by the arrival time difference of multiple nodes, marks the interference source's impact range and high-potential energy links in the network topology potential field data as prohibited, and starts from the source node and expands layer by layer to the adjacent node with the lowest cumulative potential energy value until it reaches the destination node to form a path. Then, each link in the path is sequentially assigned a channel number in the spectrum resource pool that is not occupied by neighboring links and not covered by interference, and an adaptive communication routing strategy is established.

[0017] The flight status prediction module includes: The state vector construction submodule combines the position, velocity, and acceleration components into a state vector based on the real-time GPS position, velocity vector, acceleration, and attitude angle data of itself and its neighboring nodes, and sets a state transfer matrix based on the constant acceleration motion example. The matrix linearly maps the current state vector to the predicted state of the next time step to obtain the initial motion state model; The prediction residual calculation submodule, based on the initial motion state model, subtracts the newly received kinematic data from the model prediction value to obtain the measurement residual. The gain factor is determined by calculating the ratio of the prediction covariance to the measurement noise covariance. The gain factor is then used to perform a weighted correction on the state vector and simultaneously update the prediction covariance matrix to obtain the corrected state covariance. The coordinate iteration prediction submodule uses the corrected state vector as the input for the next iteration based on the corrected state covariance, repeatedly executes the closed-loop calculation process of state prediction and measurement update, and generates a sequence of three-dimensional coordinates of multiple nodes at future moments by continuously extrapolating multiple time steps to obtain the set of future node positions.

[0018] Specifically, based on the real-time GPS position, velocity vector, acceleration and attitude angle data of itself and its neighboring nodes, the position component of each node in the Cartesian coordinate system is first calculated. , velocity component and the acceleration component Combined into a nine-dimensional state vector , where the subscript Represents the current time step, and then constructs the state transfer matrix based on the basic motion of constant acceleration, for example , this matrix describes how the state vector changes from the current time step Evolving to the next time step , its specific form is determined by the kinematic integral formula, that is, the position at the next moment is the current position plus the current speed and the time step The product of the acceleration and the time step squared, plus half of the product of the acceleration and the time step squared, the speed at the next moment is the current speed plus the acceleration and the time step The product of the acceleration remains unchanged in this model, resulting in a 9x9 sparse matrix that converts the current state vector Through matrix multiplication In a linear manner, it is mapped to the predicted state of the next time step At the same time, in order to quantify the uncertainty of the model itself, that is, the acceleration in actual motion is not absolutely constant, it is also necessary to set a process noise covariance matrix The diagonal elements of the matrix represent the variance of the acceleration component that may change in each time step. Its value is set based on the statistical analysis of the historical flight data of the aircraft model or its maximum maneuverability performance index. For example, by analyzing a large number of similar flight mission data, the average standard deviation of the acceleration change is 0.2 , then you can The diagonal elements corresponding to the acceleration components in the matrix are set to the square of the value, that is, 0.04, and the remaining off-diagonal elements are set or simplified to zero according to the correlation between the components. Finally, the initial state vector , state transfer matrix and the process noise covariance matrix Together they constitute the initial motion state model.

[0019] Based on the predicted state contained in the initial motion state model and the forecast covariance matrix , first define a measurement matrix , which is used to map the nine-dimensional state vector space to the actual measurable six-dimensional (position and velocity) measurement space, that is, It is a 6x9 matrix, the corresponding position and velocity components are 1, and the rest are 0. The newly received kinematic data from GPS, that is, the measurement vector , and through the measurement matrix Transformed model prediction value Perform subtraction to obtain the measurement residual , the residual reflects the deviation between the prediction and the actual measurement. Then, a key gain factor needs to be determined, namely the Kalman gain , its calculation is not a simple ratio, but a comprehensive consideration of the uncertainty of the forecast (based on the forecast covariance matrix ) and the uncertainty of the measurement itself (reflected by the measurement noise covariance matrix embodied) to dynamically adjust the weights, where the measurement noise covariance matrix The value of must be set in advance according to the performance specifications of the GPS device used. For example, if the device manual indicates that its positioning accuracy standard deviation is 1.5 meters and its speed accuracy standard deviation is 0.05 meters per second, then The elements at the corresponding positions on the matrix diagonal are , the element corresponding to the velocity is , the remaining elements are zero. The specific calculation of the Kalman gain involves a series of matrix operations on the prediction covariance, measurement matrix and measurement noise covariance. Subsequently, the calculated Kalman gain is used Perform weighted correction on the previously predicted state vector. The correction process is to convert the predicted state Add the product of the gain and the measurement residual to get the updated state estimate , and synchronously use the gain value to update the prediction covariance matrix, that is, subtract an amount determined by the Kalman gain and the measurement matrix from the original prediction covariance to reduce the uncertainty of the estimate and obtain the corrected state covariance .

[0020] Based on the modified state covariance And the revised state vector updated synchronously , this corrected state vector As the initial input for the next iterative calculation, it starts a repeated closed-loop calculation process. This process strictly follows the two major steps of prediction and update of the Kalman filter. At the beginning of each time step, the prediction step is first performed, using the state transfer matrix The current corrected state vector Extrapolate one time step forward to calculate the predicted state at the next moment , and simultaneously update its uncertainty to obtain the predicted covariance When new GPS measurement data arrives, the update step is executed, and the residual calculation, gain determination and state correction process are repeated to form a continuously rolling prediction-correction loop. In order to obtain the future trajectory, the system will start from the current latest corrected state vector Start, continuously and uninterruptedly perform the prediction step, perform continuous extrapolation, and the number of extrapolated time steps It is a key parameter that is pre-set according to tactical requirements and system computing capabilities. The setting principle is that the prediction duration should cover a complete routing decision and link establishment cycle. For example, if the network reconstruction decision cycle is 2 seconds, and the system's internal time step is is 0.1 seconds, then the extrapolation steps will be set to step, by continuously applying the state transfer matrix times, that is (in From 1 to ), the system generates a future path for its own node and all neighboring nodes in the network The predicted sequence of the state at each time point is finally extracted from these sequences one by one to obtain the three-dimensional spatial coordinates of each future time point. Components, after aggregation, get the set of future node positions.

[0021] The link potential field building blocks include: The link bandwidth prediction submodule calculates the bimodal difference between the three-dimensional coordinates of any two nodes based on the set of future node positions to obtain the Euclidean distance. Combining the preset airborne antenna gain and signal frequency parameters, it calculates the free space path loss and subtracts the loss value to obtain the signal reception strength. The strength value is then used as an index to search and output the corresponding bandwidth value in a preset signal strength and bandwidth mapping table to obtain the original link bandwidth data. The occlusion effect judgment submodule, based on the original link bandwidth data and the body attitude angle data, rotates and translates the body's three-dimensional mesh model according to the attitude angle data, constructs the line of sight vector segments between nodes, and determines whether the segment penetrates any triangular facet in the body model through ray intersection testing with the triangular mesh. If penetration occurs, the corresponding expected bandwidth value is cleared to zero to obtain the corrected link bandwidth data; The potential link conversion submodule, based on the corrected link bandwidth data, traverses the expected bandwidth value of each link in the data, calculates the inverse and assigns it as the link potential energy value, aggregates all link potential energy values ​​into a weighted graph with nodes as vertices and links as edges, and constructs network topology potential field data.

[0022] Specifically, based on the node future position set, which contains the three-dimensional coordinate predictions of each node in the network at multiple discrete time points in the future, the system first constructs a fully connected potential link topology for each time point, and calculates the potential link topology of any two nodes (for example, node i and node j) at a certain future time point. The predicted three-dimensional coordinates of and , calculate the two norms of the difference between the two coordinates, that is, solve their spatial straight-line distance as the Euclidean distance , then, combined with the preset parameters of the communication system hardware, including the fixed airborne antenna transmission gain , receiving gain And the signal operating frequency For example, the antenna gain is uniformly set to 3 dBi, and the signal frequency uses the 3.5 GHz band of 5G communication. Then, based on these parameters, the free space path loss is calculated. The loss value is proportional to the logarithm of the distance and the logarithm of the frequency. After calculating the path loss, the loss is subtracted from a preset baseline transmit power value (for example, 20 dBm) to obtain the expected signal reception strength of the link. Subsequently, the signal receiving strength value is used as a query index to search in a preset signal strength and bandwidth mapping table. This mapping table is generated by performing performance calibration tests on wireless communication modules in a laboratory environment before system deployment. It directly maps different received signal strength (RSSI) intervals to specific expected data bandwidths. For example, the calibration results may show that when the RSSI is higher than -70 dBm, 100 Mbps bandwidth can be supported; when the RSSI is between -70 dBm and -85 dBm, the bandwidth is reduced to 50 Mbps; when the RSSI is lower than -95 dBm, the link is considered unreliable and the bandwidth is 0 Mbps. By repeating this process for all node pairs and all future time points, a set of raw link bandwidth data containing the expected bandwidth of all potential links at each future moment is finally output.

[0023] Based on the original link bandwidth data and the body attitude angle data of each node, a pre-existing body three-dimensional mesh model that is completely consistent with the aircraft model is first called up. The model is composed of a large number of tiny triangular facets and accurately describes the shape of the aircraft. For each node in the network (that is, each aircraft), its latest attitude angle data (that is, roll angle, pitch angle and yaw angle) and three-dimensional position data are used to construct a 4x4 rigid body transformation matrix, and this matrix is ​​applied to perform corresponding rotation and translation transformations on the body three-dimensional mesh model of the node, so that its position and attitude in the global coordinate system are completely matched with its real-time state. After completing the model transformation of all relevant nodes, for any potential communication link existing in the original link bandwidth data, the system constructs a line of sight vector line segment connecting the antenna installation positions of the two nodes, and then performs high-precision The intersection test between the ray and the triangle mesh is performed. Specifically, the Möller-Trumbore algorithm is used to treat the line of sight vector segment as a finite-length ray. The ray is checked one by one to see if it intersects with any triangle facet on the fuselage model of any aircraft that is a potential obstruction. The condition for determining intersection is that the calculated intersection point must be between the starting point and the end point of the ray. Once any valid penetration is detected, that is, the line of sight is blocked by any part of the fuselage (whether the sending or receiving fuselage), the system immediately forces the expected bandwidth value of the link at that moment to be cleared to zero. This clearing operation has the highest priority, regardless of the original predicted bandwidth. By executing this occlusion effect judgment process for all links at all future time points, a corrected link bandwidth data is obtained that eliminates all non-line-of-sight link bandwidths.

[0024] Based on the corrected link bandwidth data, the system begins to quantitatively evaluate each potential communication link. The process is to traverse the expected bandwidth value of each link recorded in the data at each future time point. , and calculate its reciprocal, and directly assign this reciprocal value as the link potential energy value of the link at that moment ,Right now The internal logic of this conversion is that high bandwidth corresponds to low communication "resistance", so the potential energy value is low. Conversely, low bandwidth corresponds to high "resistance" and high potential energy value. For links with zero bandwidth due to obstruction or excessive distance, in order to indicate that they are absolutely unavailable in subsequent calculations, their potential energy value is set to a maximum value representing infinity, for example To avoid division by zero errors in calculations, after completing the calculation of all link potential energy values, these data are structured and aggregated to construct a weighted graph model that changes dynamically over time. In this model, each communication terminal node is abstracted as a vertex of the graph, and each potential communication link between nodes is abstracted as an edge connecting the corresponding vertex, and the weight of the edge is precisely set to the potential energy value of the link at the corresponding moment. By generating such a weighted graph for each predicted time point in the future, we can eventually construct a series of snapshots that can reflect the topological connectivity and communication cost of the network at different moments in the future. These snapshots together constitute the network topology potential field data.

[0025] The spectrum feature perception module includes: The cyclic spectrum matrix generation submodule periodically scans the broadband spectrum through the RF front end, performs a short-time Fourier transform on the collected I / Q sample sequence, and multiplies the transformed result matrix element-by-element with its own frequency-shifted complex conjugate version. It then performs a sliding average along the time axis to generate a two-dimensional matrix of cyclic frequencies and spectrum frequencies, thereby obtaining a two-dimensional spectrum correlation matrix. The spectrum peak feature extraction submodule sets an amplitude threshold based on the two-dimensional spectrum correlation matrix and traverses each element in the matrix. If the amplitude of an element is greater than the amplitude threshold and also greater than the amplitudes of eight adjacent elements, the position coordinates and amplitude of the element are recorded to obtain the key spectrum peak coordinate set; The fingerprint vector formatting submodule selects several points with the highest peak amplitude based on the key peak coordinate set, arranges the cyclic frequency coordinates, spectrum frequency coordinates and amplitude values ​​in sequence, and fills them into a vector of predefined length. If the number of peaks is insufficient, the remaining positions are filled with zeros to obtain the signal fingerprint feature vector.

[0026] Specifically, the RF front end periodically scans the broadband spectrum, firstly, at a preset sampling frequency, such as 200 MHz, the broadband spectrum range, such as 3.4 GHz to 3.6 GHz, is sampled at high speed to obtain a continuous complex I / Q sample sequence. Then, a short-time Fourier transform is applied to the sequence. This process converts the one-dimensional time domain sequence into a two-dimensional time-frequency representation. In the specific implementation, a Hanning window function with a length of 1024 sample points is selected, and a 50% overlap rate is set, that is, a Fourier transform is performed each time 512 sample points are slid forward to generate a time-frequency distribution matrix. ,in is the central moment of the time window, is the spectrum frequency. Then, in order to reveal the periodic characteristics of the signal, it is necessary to calculate its spectrum correlation function. This process is done by converting the time-frequency distribution matrix Compute with a frequency-shifted and complex-conjugated version of itself, for each cyclic frequency to be detected , calculate the product sequence ,in Represents the complex conjugate, this step is for each cycle frequency and spectrum frequencies The combination of will generate a product sequence that changes with time, and then this product sequence is multiplied along the time axis. The sliding window length is set according to the coherence time of the signal, for example, it is set to the length of 100 time windows, and the sliding window length is set according to the coherence time of the signal. and spectrum frequencies Repeat this calculation to generate a cycle frequency is one-dimensional, the spectrum frequency is a two-dimensional complex matrix of another dimension, that is, the two-dimensional spectral correlation matrix.

[0027] Based on the two-dimensional spectral correlation matrix, we first need to set an amplitude threshold to distinguish the effective signal peak from the background noise. The setting of this threshold adopts the constant false alarm rate detection idea. The specific calculation process is as follows: First, select an area in the two-dimensional spectral correlation matrix where it is known that there is no signal, such as a spectrum frequency interval far away from the main communication frequency band, and calculate the mean amplitude of all elements in the area. and standard deviation , then set the amplitude threshold to the mean plus a multiple of the standard deviation, for example, amplitude threshold = , if the amplitude mean of the background noise is 0.01 and the standard deviation is 0.005 through historical data analysis, the amplitude threshold will be set to 0.01 + 6 * 0.005 = 0.04. After setting the threshold, the system traverses each element in the two-dimensional spectrum correlation matrix through a nested loop. For the coordinates The current element is first checked to see if its amplitude is greater than the set amplitude threshold of 0.04. If this condition is met, it is further checked to see if it is a local maximum point, that is, its amplitude is compared with the eight adjacent elements (located at arrive In the 3x3 neighborhood of and is the minimum resolution of the matrix) is compared. Only when the amplitude of the element is strictly greater than the amplitudes of all its eight neighbors can it be confirmed as a valid spectrum peak. Once confirmed, the system immediately records the complete information of the element, including its cyclic frequency coordinates , spectrum frequency coordinates and its corresponding amplitude value, summarize all the spectrum peak information that meets the conditions, and obtain the key spectrum peak coordinate set.

[0028] Based on the key peak coordinate set, which contains a series of peak information consisting of triplets (cyclic frequency, spectral frequency, amplitude), the system first determines the number of peaks to be selected based on the preset fingerprint vector dimension, and sets the number of selected peaks to 10. This number is determined after balancing the uniqueness of the fingerprint and the dimension of the control vector. Then, all the peaks in the key peak coordinate set are sorted in descending order according to the amplitude value, and the 10 peak points with the highest amplitude are selected from the sorted list. Then, the system creates a one-dimensional floating-point vector with a predefined length of 30, which is determined by the number of peaks (10) multiplied by the number of features of each peak (3, namely cyclic frequency, spectral frequency and amplitude). The system selects the 10 peaks in order. The eigenvalues ​​of the first spectral peaks are filled into the vector. The specific arrangement is: the cyclic frequency, spectral frequency, and amplitude value of the first spectral peak occupy the 0th to 2nd positions of the vector, the eigenvalues ​​of the second spectral peak occupy the 3rd to 5th positions, and so on, until the eigenvalues ​​of the tenth spectral peak occupy the 27th to 29th positions. In this process, if the number of spectral peaks actually detected is less than 10, for example, only 7 spectral peaks are found, the system will fill the 21 eigenvalues ​​of the first 7 spectral peaks into the first 21 positions of the vector in sequence, and the remaining 9 positions in the vector (from the 21st to the 29th position) are all filled with zeros. In this way, no matter how many spectral peaks are detected, the length of the final generated vector is always consistent, and a fixed-length signal fingerprint feature vector is obtained.

[0029] The communication avoidance planning module includes: The interference source positioning submodule integrates multiple signal fingerprint feature vectors and the direction information indicated by the multi-antenna phase difference, extracts the arrival timestamps of the same interference signal at different nodes, and uses the differences between multiple timestamps to construct a nonlinear hyperbolic equation system with the interference source location as the unknown variable. The system of equations is solved to obtain the spatial coordinate set of the interference source. The comprehensive routing cost calculation submodule, based on the interference source spatial coordinate set and network topology potential field data, marks the circular area centered on the interference source coordinates and bounded by the impact radius, as well as links with potential energy values ​​above the threshold, as no-go areas. It uses an expansion queue to iteratively expand from the source node to the node with the lowest cumulative potential energy among the adjacent nodes, and records the expanded path until it reaches the destination node, thus establishing the optimal communication path. The spectrum channel allocation submodule, based on the preferred communication path, allocates the first channel in the available channel list to the first link in the path, traverses subsequent links, and assigns a channel number to each link that does not conflict with the allocated channels on the path, the channels being used by neighboring nodes, and the channels covered by the interference area, thereby establishing an adaptive communication routing strategy.

[0030] Specifically, the system integrates multiple signal fingerprint feature vectors reported by different nodes and the direction information indicated by the phase difference measured by the multi-antenna array of each node. First, the common interference signal is identified by comparing the signal fingerprint feature vectors generated by each node. When the Euclidean distance between the signal fingerprint feature vectors reported by two or more nodes is less than a preset similarity threshold (for example, 0.1), it is determined that these nodes are monitoring the same interference source. After confirming that at least four nodes are observing the same interference source, the system extracts the precise arrival timestamp of the specific interference signal from the I / Q data acquisition link of each node, and uses these timestamps to calculate the arrival time difference (TDOA) between different node pairs. For example, for node 1 and node 2, the arrival time difference is , this time difference defines a set of spatial locations, the distance difference between the set and the two nodes is a constant (in The speed of light) forms a hyperbolic surface with two nodes as the focus in three-dimensional space. By selecting three independent node pairs (for example, taking node 1 as the reference, forming three node pairs (1,2), (1,3), and (1,4)), a three-dimensional coordinate system with the interference source as the focus can be constructed. is a nonlinear hyperbolic system of unknown variables, which has the following form: in, It is The known coordinates of the nodes are obtained. Subsequently, the iterative least squares method (such as the Taylor series expansion method) is used to solve the nonlinear equations. The initial solution can be set by the intersection point of the direction information indicated by the multi-antenna phase difference (i.e., the angle of arrival AoA). The correction amount is iteratively calculated until the solution converges, and the three-dimensional coordinates of the interference source are finally obtained. If there are multiple different interference signal fingerprints, this process is repeated for each fingerprint to obtain the spatial coordinate set of the interference source.

[0031] Based on the interference source spatial coordinate set and network topology potential field data, the network topology is first pruned, the interference source coordinates are marked as the center, and an influence radius is set according to the type of interference source and the estimated power. The basis for setting the radius is to ensure that the interference signal strength at the radius boundary is lower than the normal operating threshold of the receiver. For example, for a broadband interference source with an estimated transmission power of 10 watts, its influence radius can be calculated according to the free space loss formula or set to 5 kilometers based on the empirical library. Any node or link that falls completely within this spherical area is directly marked as prohibited. At the same time, all links in the network topology potential field data are screened and a potential energy threshold is set. The threshold is related to the acceptable minimum link bandwidth. If the minimum bandwidth is required to be no less than 2 Mbps, the potential energy threshold is , any link whose potential energy value is higher than this threshold is also marked as forbidden. After marking the forbidden area and forbidden links, the Dijkstra algorithm is used to find the optimal path from the specified source node to the destination node. The algorithm starts from the source node and maintains a priority queue of nodes to be expanded. The queue is sorted from small to large according to the cumulative potential energy value of the nodes. Initially, only the cumulative potential energy of the source node is 0, and the rest are infinite. The algorithm cyclically removes the node with the lowest cumulative potential energy from the queue and examines all its unvisited and non-forbidden adjacent nodes. For each adjacent node, the new cumulative potential energy value to reach it through the current node is calculated. If the new value is lower, the cumulative potential energy of the adjacent node is updated and added or updated to the queue. At the same time, the predecessor node of the path is recorded. This process is iterated until the destination node is removed from the queue. At this time, by tracing back from the destination node along the recorded predecessor nodes to the source node, a path with the lowest total cumulative potential energy can be constructed, which is the preferred communication path.

[0032] Based on the preferred communication path, which is an ordered sequence of nodes from the source node to the destination node, the system first allocates a channel for the first link in the path (i.e., the connection between the source node and the first neighbor node on its path) from a globally available channel list. The allocation principle is to select the first available number in the list, such as channel 1. Subsequently, the system traverses each subsequent link in the order of the path and allocates a channel number that meets the conflict-free condition for each link. The conflict-free condition includes three aspects: first, the selected channel cannot be the same as any previously allocated channel on the path to avoid self-interference within the path; second, the selected channel cannot be the same as any channel currently being used by the neighbor nodes (nodes not on this path) of the nodes at both ends of the current link; If there is any channel conflict, the system will query the neighbor lists of the nodes at both ends of the link, collect the channel sets currently occupied by these neighbors, and exclude them from the available channels; finally, the selected channel cannot be covered by the located interference source, and the system will check whether the geographical location of the current link is within the influence radius of any interference source. If so, the signal fingerprint feature vector of the interference source is queried to determine the spectrum frequency range it occupies, and the corresponding channel number is marked as unavailable. For each link to be allocated, the system will combine the above three constraints to generate a temporary list of available channels, and select the channel with the smallest number for allocation. All links in the path and their assigned channel numbers are combined to establish an adaptive communication routing strategy.

Claims

1. An aviation-borne micro 5G self-organizing network multi-mode communication terminal, characterized in that: The terminal includes: The flight state prediction module uses the kinematic equation to construct a linear transformation relationship between the state vector and the time step based on the real-time GPS position, velocity vector, acceleration, and attitude angle data of itself and its neighboring nodes. It then uses the residual between the newly received kinematic data and the predicted value at the previous moment to weightedly correct the state vector and prediction covariance. The correction process is then repeated recursively to output the three-dimensional spatial coordinates of multiple nodes at multiple future time steps, thereby obtaining a set of future node positions. The link potential field construction module calculates the Euclidean distance between nodes based on the future position set of the nodes, combines the preset airborne antenna gain and signal frequency, substitutes the path loss formula to obtain the signal reception strength, and then obtains the expected bandwidth by querying the mapping relationship table between signal strength and bandwidth. At the same time, the body attitude angle data is used to determine whether the line of sight vector between nodes intersects with the body three-dimensional model. If there is an intersection, the expected bandwidth is set to zero. Finally, the inverse of the expected bandwidth is used as the link potential energy value to construct the network topology potential field data.

2. The aviation airborne micro 5G ad hoc network multi-mode communication terminal according to claim 1, characterized in that: The terminal further includes: The spectrum feature perception module periodically scans the broadband spectrum through the RF front end, performs Fourier transform on the collected I / Q sample sequence, and multiplies the transform result with the frequency-shifted complex conjugate version of the transform result point by point and performs time-smoothing averaging to generate a two-dimensional matrix of cyclic frequency and spectrum frequency. The module searches for local maximum points whose amplitude exceeds the preset detection threshold on the two-dimensional matrix, combines the coordinates and amplitude of the maximum points into a fixed-length vector, and obtains the signal fingerprint feature vector. The communication avoidance planning module integrates multiple signal fingerprint feature vectors and the direction information indicated by the multi-antenna phase difference, determines the coordinates of the interference source by solving the hyperbolic equation group formed by the arrival time difference of multiple nodes, marks the interference source influence range and the high potential energy links in the network topology potential field data as prohibited, and starts from the source node and expands layer by layer to the adjacent node with the lowest cumulative potential energy value until it reaches the destination node to form a path. Then, each link in the path is sequentially assigned a channel number in the spectrum resource pool that is not occupied by neighboring links and not covered by interference, and an adaptive communication routing strategy is established.

3. The aviation airborne micro 5G ad hoc network multi-mode communication terminal according to claim 1, characterized in that: The flight state prediction module includes: The state vector construction submodule combines the position, velocity, and acceleration components into a state vector based on the real-time GPS position, velocity vector, acceleration, and attitude angle data of itself and its neighboring nodes. It also sets the state transfer matrix based on the constant acceleration motion assumption. The matrix linearly maps the current state vector to the predicted state of the next time step to obtain the initial motion state model. A prediction residual calculation submodule, based on the initial motion state model, obtains a measurement residual by subtracting the newly received kinematic data from the model prediction value, determines a gain factor by calculating the ratio of the prediction covariance to the measurement noise covariance, and then uses the gain factor to perform a weighted correction on the state vector and synchronously updates the prediction covariance matrix to obtain a corrected state covariance; The coordinate iteration prediction submodule uses the corrected state vector as the input for the next iteration based on the corrected state covariance, repeatedly executes the closed-loop calculation process of state prediction and measurement update, and generates a sequence of three-dimensional coordinates of multiple nodes at future moments by continuously extrapolating multiple time steps to obtain a set of future node positions.

4. The aviation airborne micro 5G ad hoc network multi-mode communication terminal according to claim 1, characterized in that: The link potential field construction module includes: The link bandwidth prediction submodule calculates the second norm of the difference between the three-dimensional coordinates of any two nodes based on the future node position set to obtain the Euclidean distance. It then calculates the free space path loss based on the preset airborne antenna gain and signal frequency parameters and subtracts the loss value to obtain the signal reception strength. It then uses the strength value as an index to search and output the corresponding bandwidth value in a preset signal strength and bandwidth mapping table to obtain the original link bandwidth data. The occlusion effect judgment submodule rotates and translates the three-dimensional mesh model of the body according to the attitude angle data based on the original link bandwidth data and the body attitude angle data, constructs the line of sight vector line segments between the nodes, and determines whether the line segments penetrate any triangular facet in the body model through the ray and triangular mesh intersection test. If penetration occurs, the corresponding expected bandwidth value is cleared to zero to obtain the corrected link bandwidth data; The potential link conversion submodule, based on the corrected link bandwidth data, traverses the expected bandwidth value of each link in the data, calculates the inverse and assigns it as the link potential energy value, aggregates all link potential energy values ​​into a weighted graph with nodes as vertices and links as edges, and constructs network topology potential field data.

5. The aviation airborne micro 5G ad hoc network multi-mode communication terminal according to claim 2, characterized in that: The spectrum feature perception module includes: The cyclic spectrum matrix generation submodule periodically scans the broadband spectrum through the RF front end, performs a short-time Fourier transform on the collected I / Q sample sequence, and multiplies the transformed result matrix element-by-element with its own frequency-shifted complex conjugate version. It then performs a sliding average along the time axis to generate a two-dimensional matrix of cyclic frequencies and spectrum frequencies, thereby obtaining a two-dimensional spectrum correlation matrix. The spectral peak feature extraction submodule sets an amplitude threshold based on the two-dimensional spectral correlation matrix and traverses each element in the matrix. If the amplitude of an element is greater than the amplitude threshold and also greater than the amplitudes of eight adjacent elements, the position coordinates and amplitude of the element are recorded to obtain the key spectral peak coordinate set.

6. The aviation airborne micro 5G ad hoc network multi-mode communication terminal according to claim 5, characterized in that: The spectrum feature perception module also includes: The fingerprint vector formatting submodule selects several points with the highest peak amplitude based on the key peak coordinate set, arranges the cyclic frequency coordinates, spectrum frequency coordinates and amplitude values ​​in sequence, and fills them into a vector of predefined length. If the number of peaks is insufficient, the remaining positions are filled with zeros to obtain the signal fingerprint feature vector.

7. The aviation airborne micro 5G ad hoc network multi-mode communication terminal according to claim 2, characterized in that: The communication avoidance planning module includes: The interference source positioning submodule integrates multiple signal fingerprint feature vectors and the direction information indicated by the multi-antenna phase difference, extracts the arrival timestamps of the same interference signal at different nodes, and uses the differences between multiple timestamps to construct a nonlinear hyperbolic equation system with the interference source location as the unknown variable. The system of equations is solved to obtain the spatial coordinate set of the interference source. The comprehensive routing cost calculation submodule, based on the interference source spatial coordinate set and network topology potential field data, marks the circular area with the interference source coordinates as the center and the influence radius as the boundary and the links with potential energy values ​​higher than the threshold as no-entry areas, uses an extended queue to start from the source node, iteratively expands to the node with the lowest cumulative potential energy among the adjacent nodes, and records the extended path until it reaches the destination node, thereby establishing the optimal communication path.

8. The aviation airborne micro 5G ad hoc network multi-mode communication terminal according to claim 7, characterized in that: The communication avoidance planning module also includes: The spectrum channel allocation submodule, based on the preferred communication path, allocates the first channel in the available channel list to the first link in the path, traverses subsequent links, and allocates a channel number to each link that does not conflict with the allocated channels on the path, the channels being used by neighboring nodes, and the channels covered by the interference area, thereby establishing an adaptive communication routing strategy.

Images