A dynamic ad hoc network positioning method and system based on multi-unmanned aerial vehicle cooperation

Abstract

This application provides a dynamic ad hoc network positioning method and system based on multi-UAV cooperation, belonging to the field of UAV communication and positioning technology. First, this application acquires the arrival angle spectrum and bit error rate data of the communication link, and corrects the spatial covariance matrix using a weighting factor mapped by the bit error rate. Then, it determines the relative azimuth vector and constructs the observation matrix through eigenvalue decomposition and orthogonal projection. Next, it maps the observation matrix to a geometric configuration constraint space to establish a topological cost function, and uses gradient iteration to update and solve for the target configuration. Finally, it combines the reference position information to complete the coordinate transformation, achieving global positioning of the UAV swarm. This application can effectively improve the accuracy of UAV swarm cooperative positioning in dynamic environments by introducing bit error rate-weighted correction of angle estimation and combining it with geometric manifold constraint optimization.

Description

Technical Field

[0001] This application belongs to the field of UAV communication and positioning technology, and in particular relates to a dynamic self-organizing network positioning method and system based on multi-UAV cooperation. Background Technology

[0002] Dynamic self-organizing network positioning technology based on multi-UAV collaboration utilizes communication links and relative measurement information between UAVs to achieve high-precision position awareness without relying on external satellite navigation. This technology has broad application prospects in swarm formation flight, reconnaissance in complex environments, and emergency communication and rescue, and is a key foundation for realizing intelligent collaborative operations of UAV swarms.

[0003] Existing technologies typically utilize measurement parameters such as received signal strength, time of arrival, or angle of arrival, combined with traditional least squares or Kalman filtering algorithms, to calculate node positions. Some methods use array antennas to acquire signal characteristics, but these often rely on ideal channel assumptions to construct observation models and deduce the relative coordinates of the UAV through simple geometric polygonal positioning algorithms.

[0004] However, in dynamic and complex environments, communication link quality is often inconsistent, and directly using measurement data heavily affected by noise to build a model can lead to severe error propagation. Furthermore, traditional geometric solution methods lack effective utilization of the rigid constraints of the overall cluster topology, making them prone to getting trapped in local optima during nonlinear optimization, resulting in unstable solution results. Therefore, existing technologies suffer from insufficient positioning accuracy due to the weak anti-interference capability of positioning algorithms. Summary of the Invention

[0005] The purpose of this application is to provide a dynamic self-organizing network positioning method and system based on multi-UAV cooperation to solve the problem of insufficient positioning accuracy in the prior art.

[0006] To address the aforementioned technical problems, in a first aspect, this application provides a dynamic ad hoc network positioning method based on multi-UAV cooperation, comprising:

[0007] Acquire the arrival angle spectrum data of the signal transmission between nodes and neighboring nodes in the UAV cluster, and obtain the bit error rate distribution data generated by statistically analyzing the bit error rate of data transmission between communication links;

[0008] A spatial covariance matrix is ​​constructed based on the angle of arrival spectrum data, and the bit error rate distribution data is mapped to the channel weight factor. The covariance matrix is ​​then corrected using the channel weight factor to generate the target covariance matrix.

[0009] The target covariance matrix is ​​subjected to eigenvalue decomposition to determine the noise feature space. Orthogonal projection is performed on the noise feature space using the preset array manifold vector to determine the relative orientation vector between each node and its neighboring nodes. The observation matrix of the node is then constructed based on the relative orientation vector.

[0010] The observation matrices of all nodes are mapped to a preset geometric configuration constraint space, which is then transformed into configuration coordinate data representing the spatial geometric configuration of the UAV swarm. Based on the configuration coordinate data, a topological cost function including formation geometric constraints is constructed.

[0011] Calculate the gradient of the topological cost function in the geometric configuration constraint space, and iteratively update the position coordinates of each neighboring node in the opposite direction of the gradient until the topological cost function converges to obtain the target configuration coordinates.

[0012] The system acquires the reference position information of the target nodes in the UAV cluster, and uses the reference position information to perform coordinate transformation on the target configuration coordinates to obtain the global position coordinates of each node in the UAV cluster, thus completing the dynamic self-organizing network positioning.

[0013] Optionally, the method further includes:

[0014] The array antenna unit on the node is used to collect the radio frequency sampling signals transmitted by neighboring nodes, and the arrival angle spectrum data is obtained by calculating the spatial phase difference of the radio frequency sampling signals.

[0015] The total number of transmitted bits and the total number of error bits between the statistical node and its neighboring nodes during the data exchange process are counted, and the ratio of the total number of error bits to the total number of transmitted bits is calculated to obtain the bit error rate distribution data.

[0016] Extract the node number information and acquisition time information that are commonly associated from the angle of arrival spectrum data and the bit error rate distribution data, and establish the spatiotemporal correlation between the angle of arrival spectrum data and the bit error rate distribution data based on the node number information and acquisition time information.

[0017] Optionally, a spatial covariance matrix is ​​constructed based on the angle of arrival spectrum data, and the bit error rate distribution data is mapped to a channel weight factor. The covariance matrix is ​​then corrected using the channel weight factor to generate the target covariance matrix, including:

[0018] The spatial covariance matrix is ​​obtained by multiplying the vector dimensions by the conjugate transposes of all arrival angle spectrum data and taking the statistical mean.

[0019] The channel weight factor is obtained by numerically transforming the bit error rate distribution data using a preset mapping function.

[0020] The spatial covariance matrix is ​​weighted element-wise according to the channel weight factor to adjust the distribution intensity of each feature component in the spatial covariance matrix and generate the target covariance matrix.

[0021] Optionally, the target covariance matrix is ​​subjected to eigenvalue decomposition to determine the noise feature space. An orthogonal projection process is then performed on the noise feature space using a predefined array manifold vector to determine the relative orientation vector between each node and its neighboring nodes. Based on these relative orientation vectors, the observation matrix of each node is constructed, including:

[0022] Eigenvalue decomposition is performed on the target covariance matrix, and a preset number of feature vectors are extracted in order of eigenvalue values ​​from low to high to generate noise feature vectors.

[0023] A noise feature space is constructed using noise feature vectors, and a preset array manifold vector is orthogonally projected onto the noise feature space to generate spatial pseudospectral data representing the distribution of signal energy with spatial angle.

[0024] Extract the numerical maxima of the spatial pseudospectral data within the preset angle search area, and determine the relative orientation vector of each neighboring node relative to the node.

[0025] The relative orientation vectors belonging to the same node are combined according to the preset node numbering order to obtain the observation matrix of each node.

[0026] Optionally, the observation matrices of all nodes are mapped to a preset geometric configuration constraint space, transforming them into configuration coordinate data representing the spatial geometric configuration of the UAV swarm. A topological cost function, including formation geometric constraints, is then constructed based on the configuration coordinate data, including:

[0027] Orthogonalization is performed on the observation matrices corresponding to all nodes to extract orthogonal basis data;

[0028] By mapping orthogonal basis data to a geometric configuration constraint space, and determining the subspace configuration point corresponding to each node within the geometric configuration constraint space, configuration coordinate data representing the geometric distribution of the cluster are obtained.

[0029] The formation topology data is constructed based on the relative positional relationship between each node in the formation geometric constraints, and the formation topology data is transformed into reference configuration data located in the geometric configuration constraint space through orthogonalization mapping.

[0030] The projection deviation values ​​between the configuration coordinate data and the reference configuration data in the geometric configuration constraint space are calculated. The projection deviation values ​​are squared and accumulated to generate error scalar data, and the error scalar data is determined as the topological cost function.

[0031] Optionally, the gradient of the topological cost function within the geometric configuration constraint space is calculated, and the position coordinates of each neighboring node are iteratively updated along the opposite direction of the gradient until the topological cost function converges, yielding the target configuration coordinates, including:

[0032] By performing partial derivative operations on the topological cost function, the initial gradient data of the configuration coordinate data at the current position is determined;

[0033] The initial gradient data is projected onto the local constraint surface in the geometric constraint space to obtain the tangential projection component located in the local constraint surface. The tangential projection component is then determined as the changing gradient and used as the direction of iterative search.

[0034] The iterative search direction is multiplied by the preset step size coefficient to obtain the position displacement vector of each neighbor node in the local constraint plane. The position displacement vector is then mapped to the geometric constraint space using a preset nonlinear mapping operator to update the iterative position coordinates of each neighbor node in the geometric constraint space.

[0035] Using the updated iteration position coordinates as input, the process of iterating from calculating the initial gradient data to updating the iteration position coordinates is executed repeatedly until the numerical change of the topological cost function between two adjacent iterations is less than the preset convergence threshold, thus generating the target configuration coordinates.

[0036] Optionally, the reference position information of the target node in the UAV swarm is obtained, and the target configuration coordinates are transformed using the reference position information to obtain the global position coordinates of each node in the UAV swarm, thus completing dynamic ad hoc network positioning, including:

[0037] Perform a geometric space reconstruction mapping on the target configuration coordinates to obtain a local configuration matrix representing the relative distribution relationship of nodes in the UAV swarm;

[0038] Extract the relative coordinates of the corresponding target nodes in the local configuration matrix, calculate the rotation and translation features between the relative coordinates of the nodes and the reference position information, and determine the coordinate transformation parameters;

[0039] By performing an affine transformation on the local configuration matrix using coordinate transformation parameters, the UAV swarm is mapped to the global geographic coordinate system, thus obtaining the global position coordinates of each node in the UAV swarm.

[0040] Secondly, this application provides a dynamic self-organizing network positioning system based on multi-UAV cooperation, comprising:

[0041] The acquisition module is used to acquire the arrival angle spectrum data of the signal transmission between the nodes and neighboring nodes in the UAV cluster, and to acquire the bit error rate distribution data generated by statistically analyzing the bit error rate of data transmission between the communication links.

[0042] The generation module is used to construct a spatial covariance matrix based on the angle of arrival spectrum data, map the bit error rate distribution data into a channel weight factor, and correct the covariance matrix through the channel weight factor to generate the target covariance matrix.

[0043] The construction module is used to perform eigenvalue decomposition on the target covariance matrix to determine the noise feature space, and to perform orthogonal projection on the noise feature space using a preset array manifold vector to determine the relative orientation vector between each node and its neighboring nodes, and to construct the observation matrix of the node based on the relative orientation vector.

[0044] The construction module is also used to map the observation matrix of all nodes to a preset geometric configuration constraint space, transform it into configuration coordinate data representing the spatial geometric configuration in the UAV cluster, and construct a topological cost function including formation geometric constraints based on the configuration coordinate data;

[0045] The generation module is also used to calculate the gradient of the topological cost function in the geometric configuration constraint space, and iteratively update the position coordinates of each neighboring node in the opposite direction of the gradient until the topological cost function converges to obtain the target configuration coordinates.

[0046] The conversion module is used to obtain the reference position information of the target node in the UAV cluster, and use the reference position information to perform coordinate transformation on the target configuration coordinates to obtain the global position coordinates of each node in the UAV cluster, thus completing the dynamic self-organizing network positioning.

[0047] Thirdly, this application provides an electronic device, comprising:

[0048] Memory, used to store computer programs;

[0049] A processor, configured to execute the computer program to implement the steps of the dynamic self-organizing network positioning method based on multi-UAV cooperation as described in the first aspect above.

[0050] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the dynamic self-organizing network positioning method based on multi-UAV cooperation as described in the first aspect above.

[0051] The dynamic self-organizing network positioning method based on multi-UAV cooperation provided in this application first corrects the spatial covariance matrix by using the channel weight factor generated by mapping bit error rate distribution data. This can dynamically adjust the distribution intensity of feature components according to the link quality, effectively suppressing noise interference and error propagation introduced by high bit error rate communication links, thereby ensuring the estimation accuracy of the relative azimuth vector.

[0052] Secondly, the observation matrix is ​​mapped to the geometric configuration constraint space, and a topological cost function including formation geometric constraints is constructed. Through gradient iterative updates within the constraint space, not only is the overall topological structure of the cluster enhanced to strengthen the rigid constraints of the solution, but the shortcomings of traditional nonlinear optimization methods that easily get trapped in local optima are also avoided. Therefore, this application can effectively solve the problem of weak anti-interference capability of positioning algorithms in complex dynamic environments and significantly improve the positioning accuracy of multi-UAV cooperative positioning.

[0053] Furthermore, this application firstly acquires radio frequency signals and calculates spatial phase difference using an array antenna, enabling precise acquisition of the arrival angle spectrum reflecting the spatial orientation of nodes from the physical layer. Combined with real-time bit error rate data from the data link layer, it achieves dual perception of the spatial attributes of the communication link and channel quality. Secondly, by establishing a spatiotemporal correlation through matching node numbers with acquired information, it ensures strict synchronization of channel quality assessment data and spatial angle measurement data in time and object in scenarios where the topology changes rapidly due to the high-speed movement of UAVs. This avoids misjudging the reliability of observation data due to spatiotemporal misalignment of data, effectively reducing the cumulative positioning error in dynamic environments and further improving the accuracy of positioning results. Attached Figure Description

[0054] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 A flowchart illustrating a dynamic self-organizing network positioning method based on multi-UAV cooperation provided in this application embodiment;

[0056] Figure 2 A flowchart illustrating a method for constructing an observation matrix provided in an embodiment of this application;

[0057] Figure 3 A flowchart illustrating a method for generating target configuration coordinates provided in an embodiment of this application;

[0058] Figure 4 A schematic diagram of a dynamic self-organizing network positioning system based on multi-UAV cooperation provided in an embodiment of this application;

[0059] Figure 5 This is a schematic diagram of the hardware structure of an electronic device provided in one embodiment of this application. Detailed Implementation

[0060] In the field of multi-UAV cooperative ad hoc network positioning, while existing technologies can utilize signal characteristics for position calculation, they exhibit significant limitations when facing dynamically changing and complex environments. The main drawback is that traditional methods are often based on ideal channel assumptions, lacking awareness and response to the uneven quality of communication links. This results in highly interfered, high-error-rate data directly participating in the calculation, causing severe error propagation.

[0061] Furthermore, due to the lack of effective utilization of the rigid constraints of the overall cluster topology, the nonlinear optimization process that relies solely on geometric polygonal positioning is prone to getting trapped in local optima, making it difficult for the final positioning results to meet the requirements of high-reliability collaborative operation in terms of stability and accuracy. There is an urgent need for a positioning scheme that can take into account both channel quality awareness and global topology optimization.

[0062] To address the aforementioned challenges, this application proposes a dynamic ad hoc network localization method based on multi-UAV cooperation. The core logic of this method lies in introducing a dual optimization mechanism involving channel quality and geometric constraints. First, a channel weight factor is generated using the bit error rate distribution data of the communication links, adaptively correcting the spatial covariance matrix to suppress interference from high-noise links on angle estimation at the signal processing level. Then, the observation matrix is ​​mapped to the geometric constraint space, constructing a topological cost function that includes formation geometric constraints, and a global optimal solution is found through gradient iterative search.

[0063] This scheme abandons the dependence on ideal channels and, through the synergistic effect of weighted correction for anti-interference and configuration constraint for anti-trap, effectively blocks the error propagation caused by low-quality links and avoids the local convergence problem of traditional optimization algorithms, thereby achieving high-precision cooperative positioning in complex dynamic environments.

[0064] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0065] To address the problems of existing technologies, embodiments of this application provide a dynamic ad hoc network positioning method, apparatus, device, computer storage medium, and computer program product based on multi-UAV cooperation. The dynamic ad hoc network positioning method based on multi-UAV cooperation provided in this application embodiment will be described below first.

[0066] Figure 1 A flowchart illustrating a dynamic self-organizing network positioning method based on multi-UAV cooperation provided in one embodiment of this application is shown. Figure 1 As shown, the method includes:

[0067] S101. Obtain the arrival angle spectrum data of the signal transmission between the nodes and neighboring nodes in the UAV cluster, and obtain the bit error rate distribution data generated by statistically analyzing the bit error rate of data transmission between the communication links.

[0068] Angle of arrival (AHA) spectrum data refers to spatial characteristic information reflecting the direction of a signal incident from the transmitting node to the receiving node. This can include spatial phase difference vectors calculated from the radio frequency signals received by the array antenna, array snapshot data sequences, or characteristic vectors of signal energy distribution with angle. Bit error rate (BER) distribution data refers to statistical information quantifying the quality of communication channels between nodes. It can be a set of numerical values ​​representing the ratio of erroneous bits to the total number of transmitted bits within a specific time window, used to indicate the degree of noise or multipath interference in the current communication link.

[0069] During the implementation of the scheme, the array antenna units on the node are first used to collect radio frequency (RF) sampling signals transmitted by neighboring nodes. Spatial phase difference is calculated on the RF sampling signals to obtain arrival angle spectrum data, which includes spatial orientation features. Simultaneously, the total number of transmitted bits and the total number of error bits during data exchange between the node and its neighbors are counted, and the ratio of the total number of error bits to the total number of transmitted bits is calculated to obtain bit error rate (BER) distribution data. Next, the node ID information and acquisition time information that are commonly associated are extracted from the BER spectrum data and the BER distribution data. A spatiotemporal correlation is established between the two based on the node ID information and the acquisition time information to ensure strict alignment between physical layer measurement data and link layer quality data for the same object at the same time.

[0070] For example, in a communication scenario between UAV node A and its neighbor node B, assume that node A receives the signal from node B at time T1 using a four-element array antenna, obtaining a set of arrival angle spectrum vectors including four complex components. Meanwhile, the link layer counts that n error bits exist in the m data bits sent by node B at time T1, and calculates the bit error rate distribution data.

[0071] Optionally, the method further includes:

[0072] The array antenna unit on the node is used to collect the radio frequency sampling signals transmitted by neighboring nodes, and the arrival angle spectrum data is obtained by calculating the spatial phase difference of the radio frequency sampling signals.

[0073] Radio frequency (RF) sampling signals refer to the digital baseband or intermediate frequency (IF) signal sequences converted from received high-frequency analog electromagnetic waves by a node through RF front-end circuitry. These signals include amplitude and phase information and are the fundamental physical data for array signal processing. Spatial phase difference refers to the phase delay that occurs when the same signal wavefront arrives at different elements of the array antenna. This delay has a geometrical mapping relationship with the signal incident angle and the element spacing.

[0074] In the implementation process, firstly, the array antenna elements mounted on the nodes, such as uniform linear arrays or uniform circular arrays, receive wireless signals transmitted by neighboring nodes. The signals are then sampled using an analog-to-digital converter to obtain multiple digitized radio frequency (RF) sample signals. Secondly, for each RF sample signal, a fast Fourier transform is used to convert the time-domain signal to a frequency-domain signal, or complex-form snapshot data is directly extracted in the time domain. Next, the phase lag or lead of the received signals from different array elements relative to the reference array element is calculated to obtain the spatial phase difference.

[0075] For example, suppose node A is equipped with a 4-element uniform linear array. After receiving a signal transmitted by its neighbor node B, it collects a set of complex snapshot vectors. This vector includes the received components of four array elements. The component of the second array element has a phase difference relative to the component of the first array element. At this time, this group includes Complex vector of information This refers to the acquired arrival angle spectrum data.

[0076] The total number of transmitted bits and the total number of error bits between the statistical node and its neighboring nodes during the data exchange process are counted, and the ratio of the total number of error bits to the total number of transmitted bits is calculated to obtain the bit error rate distribution data.

[0077] The total number of transmitted bits refers to the total number of bits included in the physical layer data frame demodulated by the receiving node within a complete communication interaction or a specific time window, including payload and parity bits. The total number of error bits refers to the number of bits that have been flipped or lost, as determined by channel decoding or verification mechanisms.

[0078] During the implementation of the scheme, firstly, during data exchange between nodes and their neighboring nodes, the receiving end's communication protocol stack demodulates and decodes the received data packets. Secondly, by comparing known pilot sequences and utilizing cyclic redundancy check (CRC) mechanisms or synoptic calculations during forward error correction decoding, the number of erroneous bits during transmission is counted and recorded as the total number of erroneous bits. Simultaneously, the total number of transmitted bits within the data packet or statistical period is recorded. Finally, the ratio of the total number of erroneous bits to the total number of transmitted bits is calculated using a division operation to obtain the bit error rate distribution data reflecting the channel quality at the current moment.

[0079] For example, suppose node A is decoding a frame of data from node B, with a total of 1000 bits transmitted. After forward error correction decoding, 50 bits are found to be uncorrectable or have failed verification, and are marked as error bits. The calculated bit error rate distribution is then as follows: .

[0080] Extract the node number information and acquisition time information that are commonly associated from the angle of arrival spectrum data and the bit error rate distribution data, and establish the spatiotemporal correlation between the angle of arrival spectrum data and the bit error rate distribution data based on the node number information and acquisition time information.

[0081] Node identification information refers to a numerical tag or address code used to uniquely identify each member in a drone swarm, such as a MAC address or ad hoc network ID. Data acquisition time information refers to the timestamps used when data is received or processed, ensuring the synchronization of data across different dimensions on the timeline.

[0082] In the specific implementation process, the node number information of the transmitting end is first parsed from the metadata or physical layer header of the RF sampling signal, and the precise time of signal arrival at the RF front end is recorded as the acquisition time information. Simultaneously, the source node number is extracted from the data frame header demodulated from the link layer, and the timestamp when the bit error rate statistics are completed is recorded. Then, using the node number and acquisition time as index keys, a matching search is performed on the arrival angle spectrum data and bit error rate distribution data within a preset time tolerance window. When the node numbers of two sets of data match and the timestamp difference is less than a preset threshold, the two are bound as an associated data pair, establishing a spatiotemporal correlation.

[0083] For example, suppose the processing unit of node A buffers data from node B, i.e., ID 002, at time... Arrival angle spectrum data collected at all times At the same time, the link layer reported the time from node B. The bit error rate data. The processing unit will... Using ID=002 and bit error rate data Bind the data to form a complete input record.

[0084] This embodiment achieves dual perception of communication link spatial attributes and channel quality, ensuring accurate alignment of channel quality assessment and angle measurement data in time and object in high-speed UAV movement scenarios, avoiding misjudgment of observation reliability due to data misalignment.

[0085] S102. Construct a spatial covariance matrix based on the arrival angle spectrum data, and map the bit error rate distribution data to the channel weight factor. Correct the covariance matrix using the channel weight factor to generate the target covariance matrix.

[0086] Optionally, step S102, which involves constructing a spatial covariance matrix based on the angle of arrival spectrum data, mapping the bit error rate distribution data to a channel weight factor, and correcting the covariance matrix using the channel weight factor to generate the target covariance matrix, may specifically include:

[0087] S1021. By multiplying the vector dimensions of all arrival angle spectrum data by conjugate transpose and taking the statistical mean, the spatial covariance matrix is ​​obtained.

[0088] The spatial covariance matrix is ​​a second-order statistical characteristic matrix constructed based on the received signals of the array antenna. It is used to reflect the correlation between the received signals of different array elements and the distribution of signal energy in the spatial dimension. Its diagonal elements usually represent the signal power received by each array element, and the off-diagonal elements represent the cross-correlation characteristics between array elements.

[0089] During the implementation of the solution, the first step is to obtain data belonging to the same communication link within a preset time window. The arrival angle spectrum data samples are multiple snapshot vectors. Next, for each snapshot vector, a vector outer product operation is performed, that is, the vector is multiplied by its own conjugate transpose. Then, this... The outer product matrices are summed and divided by the number of samples. This yields the initial spatial covariance matrix.

[0090] For example, suppose a set of data is collected including The dimension is The set of complex snapshot vectors By performing... on each vector in this set Calculate and average, where Representing the conjugate transpose, we get a complex matrix .

[0091] S1022. The bit error rate distribution data is numerically transformed using a preset mapping function to obtain the channel weight factor.

[0092] The channel weight factor is a dimensionless scalar whose value is usually between 0 and 1. It is used to quantify the reliability of the current communication link in the positioning calculation. The closer the value is to 1, the more reliable the data of the link is.

[0093] During the implementation of the scheme, the bit error rate distribution data obtained in step S101 is used as an input variable and substituted into a preset mapping function. This mapping function can be a negative exponential function, an inverse proportional function, or a piecewise linear function. The calculated output value is the channel weight factor. For example, assume the bit error rate data of the communication link between node A and node B is... The preset mapping function is ,in Channel weighting factor, This is a constant used to adjust sensitivity. The channel weighting factor is calculated by substituting the data into the formula. .

[0094] S1023. The spatial covariance matrix is ​​weighted element by element according to the channel weight factor to adjust the distribution intensity of each feature component in the spatial covariance matrix and generate the target covariance matrix.

[0095] The target covariance matrix refers to the covariance data that, after channel quality correction, can more accurately reflect the spatial characteristics of the effective signal and suppress the influence of interference components.

[0096] During the implementation of the scheme, the calculated channel weight factor is used to perform scalar multiplication on the spatial covariance matrix generated in S1021. Specifically, the complex value of each row and column of the matrix is ​​multiplied by the weight factor, thereby adjusting the overall energy distribution intensity of the matrix.

[0097] For example, using weighting factors Regarding the aforementioned The initial spatial covariance matrix Make corrections. Assume... The first in Line number Column elements The corrected target covariance matrix is ​​then expressed as follows. The element at the corresponding position becomes The final generated target covariance matrix for:

[0098] ;

[0099] This embodiment effectively suppresses noise interference and error propagation introduced by high bit error rate communication links, ensuring that the subsequent feature decomposition process focuses more on high-quality signals, thereby significantly improving the accuracy of relative orientation vector estimation in complex electromagnetic environments.

[0100] S103. Perform eigenvalue decomposition on the target covariance matrix to determine the noise feature space. Use the preset array manifold vector to perform orthogonal projection on the noise feature space to determine the relative orientation vector between each node and its neighboring nodes. Construct the observation matrix of the node based on the relative orientation vector.

[0101] Optionally, step S103, which involves performing eigenvalue decomposition on the target covariance matrix to determine the noise feature space, performing orthogonal projection on the noise feature space using a preset array manifold vector to determine the relative orientation vector between each node and its neighboring nodes, and constructing the observation matrix of the node based on the relative orientation vector, may specifically include:

[0102] Figure 2 A flowchart illustrating a method for constructing an observation matrix according to an embodiment of this application is shown. Figure 2 As shown, the method includes:

[0103] S1031. Perform eigenvalue decomposition on the target covariance matrix, extract a preset number of eigenvectors in order of eigenvalue values ​​from low to high, and generate noise eigenvectors.

[0104] Eigenvalue decomposition is a mathematical operation that decomposes a matrix into an eigenvector matrix and an eigenvalue diagonal matrix. It reveals the matrix's intrinsic structure, where the magnitude of the eigenvalues ​​reflects the energy intensity along the corresponding eigenvector direction. Noise eigenvectors are the eigenvectors associated with the smaller eigenvalues ​​in the eigenvalue decomposition results. The subspace spanned by these vectors is primarily composed of ambient noise and receiver thermal noise and is orthogonal to the signal subspace containing the valid signal.

[0105] In the implementation process, firstly, linear algebraic algorithms such as the QR algorithm or the Jacobi iteration method are used to perform eigenvalue decomposition on the target covariance matrix generated in step S102. Secondly, the calculated... The eigenvalues ​​are arranged in ascending order of their numerical values. Since signal energy is usually much greater than noise energy, larger eigenvalues ​​correspond to the signal subspace, while smaller eigenvalues ​​correspond to the noise subspace. Finally, the number of information sources is estimated in advance based on information theory criteria such as MDL and AIC. That is, the number of neighboring nodes, select the one with the smallest value. The eigenvectors corresponding to each eigenvalue are labeled as noise eigenvectors.

[0106] For example, suppose for Target covariance matrix Decomposition yields One eigenvalue: Assume it exists. There are 10 neighboring nodes. Therefore, extract the smaller values. eigenvalues Corresponding column vector This is determined as a noise feature vector.

[0107] S1032. Construct a noise feature space using noise feature vectors, and perform orthogonal projection processing on the preset array manifold vectors onto the noise feature space to generate spatial pseudospectral data representing the distribution of signal energy with spatial angle.

[0108] Noise feature space refers to the linear space spanned by all extracted noise feature vectors as a basis. Array manifold vector refers to the ideal phase response vector, pre-calculated based on the array geometry, describing the effect of a unit-energy plane wave from a specific direction on each array element. Spatial pseudospectral data refers to a set of function curves or values ​​constructed based on orthogonality metrics, whose peak positions indicate the true direction of signal arrival.

[0109] During the implementation of the scheme, the extracted noise feature vectors are first combined into a noise subspace matrix. Next, construct the angle-dependent... Changing array manifold vector This vector includes the theoretical phase delay at different angles. Using the idea of ​​a multiple signal classification algorithm, the spatial pseudospectral function is calculated. The formula can be This calculation process is essentially a detection. The degree of orthogonality with the noise space, when When pointing in the direction of the real signal, the denominator approaches zero. A maximum value appears.

[0110] For example, a matrix is ​​constructed using the aforementioned noise feature vectors. Assuming a search angle from arrive Step. For each Generate guiding vector and calculate The final result is a set of data sequences that vary with angle, i.e., spatial pseudospectral data.

[0111] S1033. Extract the numerical maxima of the spatial pseudospectral data within the preset angle search area, and determine the relative orientation vector of each neighboring node belonging to the same node relative to the node.

[0112] The preset angle search area refers to the angle scanning range set according to the array's field of view or prior information, such as omnidirectional scanning. The relative orientation vector is a unit direction vector generated based on the calculated angle, used to describe the geometric orientation of a neighboring node relative to its own current node.

[0113] During the implementation of the scheme, peak searches are performed on the spatial pseudospectral data generated by S1032 within a preset angle search area. Peaks are precisely located by comparing adjacent values ​​or using the derivative method. The angle value corresponding to the location of the peak is the arrival direction of the neighboring nodes. Then, trigonometric functions are used to convert this angle into a unit direction vector in a two-dimensional or three-dimensional coordinate system, which is then determined as the relative azimuth vector. For example, suppose an angle is found in spatial pseudospectral data... A significant peak exists at this location. The azimuth angle of the neighboring node relative to the current node is confirmed to be... Convert it into a relative orientation vector on a two-dimensional plane. .

[0114] S1034. According to the preset node numbering order, the relative orientation vectors belonging to the same node are combined to obtain the observation matrix of each node.

[0115] The observation matrix is ​​a matrix formed by concatenating the relative orientation vectors of all neighboring nodes measured by the current node in a specific column order, and it centrally expresses the local geometric and topological information perceived by the node.

[0116] During the implementation of the scheme, if the current node detects multiple neighboring nodes, the identity ID of each neighboring node is first obtained. Secondly, the relative orientation vectors corresponding to each neighboring node are sorted in ascending order of ID. Finally, the sorted relative orientation vectors are concatenated as column vectors to construct a multi-column matrix, obtaining the observation matrix of that node. For example, suppose node A detects two neighboring nodes B (ID=002) and C (ID=003) in a two-dimensional plane. The calculated relative orientation vectors are as follows: and By combining the observations in ID order, a 2×2 observation matrix is ​​obtained. .

[0117] This embodiment effectively separates the signal and noise subspaces, enabling accurate reconstruction of the local geometric topology around nodes without prior location information, thus providing high-precision directional constraint data for subsequent global configuration calculations.

[0118] S104. Map the observation matrices of all nodes to the preset geometric configuration constraint space, transform them into configuration coordinate data representing the spatial geometric configuration of the UAV cluster, and construct a topological cost function including formation geometric constraints based on the configuration coordinate data.

[0119] Optionally, step S104, which maps the observation matrices of all nodes to a preset geometric configuration constraint space, transforms them into configuration coordinate data representing the spatial geometric configuration of the UAV swarm, and constructs a topology cost function including formation geometric constraints based on the configuration coordinate data, may specifically include:

[0120] S1041. Perform standard orthogonalization on the observation matrices corresponding to all nodes to extract orthogonal basis data.

[0121] Standard orthogonalization refers to the process of transforming the original matrix into a matrix composed of a set of pairwise orthogonal basis vectors with a magnitude of 1, using mathematical transformation algorithms such as Schmidt orthogonalization or QR decomposition. This aims to eliminate redundant scale information in the original observation data to extract pure directional features. Orthogonal basis data refers to matrices that satisfy orthogonality constraints after processing; it serves as a mathematical bridge connecting the physical observation space and the abstract geometric configuration space.

[0122] During the implementation of the scheme, the observation matrix of each node generated in step S103 is first obtained. Next, the QR decomposition algorithm is used to decompose the observation matrix, extracting the Q-matrix portion, i.e., the orthogonal matrix, which is then truncated into matrices with the same dimension as the signal subspace, yielding orthogonal basis data. For example, let's first assume the observation matrix of node A... It is a multi-column complex or real matrix. Secondly, for... Perform QR decomposition, i.e. and extract This serves as orthogonal basis data for node A. At this point, satisfy That is, the identity matrix.

[0123] S1042. Map the orthogonal basis data to the geometric configuration constraint space. By determining the subspace configuration point corresponding to each node in the geometric configuration constraint space, obtain the configuration coordinate data representing the geometric distribution of the cluster.

[0124] The geometric configuration constraint space refers to the Riemannian manifold space as defined in Table 1. Each point in this space represents a possible local geometric structure and naturally satisfies orthogonality constraints. Configuration coordinate data refers to the specific coordinate points or matrix representations corresponding to the orthogonal basis data in this manifold space. The preset geometric configuration constraint spaces are shown in Table 1 below:

[0125] Table 1: Preset Geometric Configuration Constraint Space

[0126]

[0127] As shown in Table 1, the geometric configuration constraint space is a non-Euclidean space with a specific topological structure. Each point in this space actually represents a matrix that satisfies orthogonality constraints. The subspace spanned by this matrix corresponds to the local geometry observed by the UAV node. Performing operations within this space ensures that all updated coordinates always satisfy the geometric rigidity constraints, avoiding structural distortions that occur in traditional optimization.

[0128] During the implementation of the scheme, firstly, since the column vectors of the orthogonal basis data Q of each node span a subspace, this subspace includes the linear combination information of the relative orientations of all the nodes' neighbors, reflecting the local geometric view of the node. Therefore, the Q of each node is regarded as a point, i.e., a spatial element, within the geometric configuration constraint space described in Table 1. Secondly, since the orthogonal basis data itself satisfies... Due to constraints, this mapping process essentially assigns matrices in linear space to coordinate points in manifold space. For each node in the cluster, its corresponding position on the manifold is determined, and the set of all these positions constitutes the configuration coordinate data representing the geometric distribution of the cluster.

[0129] For example, suppose we take the orthogonal basis data of node A. Mapping to Grassmanifold Above, marked as points on the manifold ,in For subspace dimension, This represents the physical dimension of the relative orientation vector. Similarly, the points of node B are obtained. The manifold points corresponding to all nodes. The set of values ​​constitutes the configuration coordinate data representing the global geometric distribution of the cluster.

[0130] S1043. Construct formation topology data based on the relative positional relationship between each node in the formation geometric constraints, and transform the formation topology data into reference configuration data located in the geometric configuration constraint space through orthogonalization mapping.

[0131] Formation geometric constraints refer to pre-defined formation structure parameters that a drone swarm is expected to maintain, such as the desired distance or relative angle between nodes. Formation topology data is a relative position matrix under ideal conditions constructed based on these constraints. Reference configuration data refers to the target points obtained by mapping the ideal topology data to the geometric configuration constraint space, representing the geometric shape that perfectly satisfies the formation requirements.

[0132] During the implementation of the scheme, firstly, based on the formation set in the task, such as a square or triangular formation, the theoretically observable relative azimuth vectors of neighboring nodes for each node are calculated, constructing the ideal observation matrix, i.e., the formation topology data. Next, the ideal observation matrix is ​​subjected to the orthogonalization process described in S1041 to obtain the ideal orthogonal basis. Finally, the ideal orthogonal basis is mapped to the geometric configuration constraint space described in Table 1 to obtain the reference configuration data. For example, assume that node A forms an isosceles right triangle with B and C. The theoretically observable relative azimuth vectors are calculated based on the geometric relationship, and the ideal matrix is ​​constructed. .right QR decomposition yields It is determined as a reference point on the manifold space. .

[0133] S1044. Calculate the projection deviation values ​​between the configuration coordinate data and the reference configuration data in the geometric configuration constraint space, perform square accumulation processing on the projection deviation values ​​to generate error scalar data, and determine the error scalar data as the topological cost function.

[0134] Projection deviation refers to the distance metric between observed configuration points and reference configuration points in the geometrically constrained space, typically expressed using projected F-norm distance or chordal distance. Error scalar data is the sum of deviation values ​​for all nodes, used to quantify the overall difference between the current cluster formation and the desired formation.

[0135] During the implementation of the scheme, the difference between the configuration coordinates of each node and the reference configuration data is first calculated using the distance metric formula unique to manifold space. Specifically, the Frobenius norm of the difference between the projection matrices of the two subspaces is calculated. Secondly, the squared values ​​of this norm for all nodes are summed to obtain a non-negative scalar value, which is then used to construct the topological cost function.

[0136] For example, for nodes Assuming its configuration coordinate data is The reference configuration data is Calculate the squared projective distance between the two nodes on the manifold and sum it over all nodes to construct the topological cost function. As shown in the following formula (1):

[0137] (1)

[0138] in, The total number of nodes. and These are the projection operators for the actual observation subspace and the reference subspace, respectively. This represents the Frobenius norm.

[0139] This embodiment cleverly transforms complex nonlinear formation constraints into a distance minimization problem in manifold space. This method not only ensures strict satisfaction of geometric constraints during optimization but also leverages the invariance of projected distance to rotation and translation, improving the robustness of the positioning algorithm to initial errors.

[0140] S105. Calculate the gradient of the topological cost function in the geometric configuration constraint space, and iteratively update the position coordinates of each neighboring node in the opposite direction of the gradient until the topological cost function converges to obtain the target configuration coordinates.

[0141] Optionally, step S105, which calculates the gradient of the topological cost function within the geometric configuration constraint space and iteratively updates the position coordinates of each neighboring node along the opposite direction of the gradient until the topological cost function converges to obtain the target configuration coordinates, may specifically include:

[0142] Figure 3 A flowchart illustrating a method for generating target configuration coordinates according to an embodiment of this application is shown. Figure 3 As shown, this method first determines the initial gradient data of the configuration coordinates at the current position by performing partial derivative operations on the topological cost function. Then, the initial gradient data is projected onto the local constraint surface of the geometric configuration constraint space, and the tangential projection component is extracted as the iterative search direction.

[0143] Based on this, the position displacement vector is calculated using a preset step size coefficient, and a nonlinear mapping operator is used to map this vector back to the geometric configuration constraint space, thereby completing the iterative update of the position coordinates of neighboring nodes. After the update is completed, it is determined whether the numerical change of the topological cost function between two adjacent iterations is less than a preset convergence threshold. If the convergence condition is not met, the gradient calculation step is returned with the updated coordinates as input until the convergence condition is met, and finally the target configuration coordinates are output. This method includes:

[0144] S1051. By performing partial derivative operations on the topological cost function, the initial gradient data of the configuration coordinate data at the current position is determined.

[0145] Initial gradient data refers to the first-order partial derivative matrix of the topological cost function with respect to the configuration coordinate variables from the perspective of Euclidean space. It indicates the direction in which the function value grows the fastest at the current point.

[0146] During the implementation of the scheme, the topological cost function constructed in step S104 was applied using the matrix calculus rule. Find information about the variables The partial derivatives. Specifically, calculate... .because This is the norm square of the difference in the projection matrices, and its derivative typically involves the product of the error term and the coordinate matrix. For example, for the above formula (1), the nodes are calculated using the chain rule. The initial gradient data is , where the matrix Dimensions and same.

[0147] S1052. Project the initial gradient data onto the local constraint surface of the geometric configuration constraint space to obtain the tangential projection component located in the local constraint surface, and determine the tangential projection component as the changing gradient to serve as the iterative search direction.

[0148] The local constraint surface refers to the tangent space at the current configuration point in the manifold space. It is a linear Euclidean space tangent to the manifold, encompassing all small motion directions near the point that satisfy the geometric constraints. The tangential projection component is the vector obtained by orthogonally projecting the initial gradient data onto this tangent space, ensuring that movement along this direction does not disrupt the geometry of the manifold.

[0149] During the implementation of the scheme, the orthogonal projection operator specific to the manifold space is used to transform the initial gradient data obtained from S1051. Mapped to the current point tangent space Above. For Grassmann manifolds, varying gradients The calculation formula is usually as follows: For example, the node initial gradient Substituting into the projection formula, the gradient of change is calculated. , will The reverse direction is the determined iterative search direction.

[0150] S1053. Perform a product operation between the iterative search direction and the preset step size coefficient to obtain the position displacement vector of each neighbor node in the local constraint plane, and use the preset nonlinear mapping operator to map the position displacement vector to the geometric configuration constraint space to update the iterative position coordinates of each neighbor node in the geometric configuration constraint space.

[0151] The position displacement vector is the tangent vector obtained by moving a certain step along the negative gradient direction in the tangent space, representing the displacement under a local linear approximation. A nonlinear mapping operator is a mathematical operation that maps the tangent vector in the tangent space back to the surface of the manifold space, such as exponential mapping or contraction mapping based on QR decomposition, used to transform linear displacement into curvilinear motion that satisfies orthogonal constraints.

[0152] During the implementation of the scheme, the step size coefficient can first be fixed or determined through line search. Secondly, calculate the position displacement vector. Next, the coordinates are updated using nonlinear mapping operators. Commonly used operators are... ,in For the updated iterative position coordinates, This indicates taking the Q factor of the QR decomposition.

[0153] S1054. Using the updated iteration position coordinates as input, repeatedly execute the iterative process from calculating the initial gradient data to updating the iteration position coordinates until the numerical change of the topological cost function between two adjacent iterations is less than the preset convergence threshold, and generate the target configuration coordinates.

[0154] The target configuration coordinates refer to the configuration coordinate data that minimizes the topological cost function when the optimization algorithm converges. It represents the most likely relative geometric distribution of the UAV swarm under the current observations and constraints. The preset convergence thresholds are shown in Table 2 below.

[0155] Table 2: Preset Convergence Threshold Comparison Table

[0156]

[0157] As shown in Table 2, the preset convergence threshold is not just a single value, but a set of parameters used to comprehensively determine whether the optimization process has reached a steady state. When the state during the iteration process satisfies any one or a specific combination of conditions in the table, the algorithm is considered to have converged, and the computation is stopped to save resources.

[0158] During the implementation of the plan, firstly, the updated Substituting back to S1051, repeat the gradient, projection, and update steps. At the end of each iteration, calculate the current cost function value. Compared with the previous value The absolute value of the difference. Next, compare this difference with the function value change threshold defined in Table 2. Compare the values. If the difference is less than the threshold, or the gradient magnitude is less than... Or reach the maximum number of iterations. If the loop stops, then the loop will stop.

[0159] For example, suppose the function value changes by a threshold. for After 50 iterations, the calculation found that... This value is less than the threshold set in Table 2. At this point, the algorithm is considered to have converged. The output is the target configuration coordinates.

[0160] This embodiment cleverly solves the nonlinear optimization problem with orthogonal constraints, ensuring that the result of each iteration is strictly within the geometric configuration constraint space, avoiding the cumulative error caused by constraint violation, and thus quickly and stably converges to the globally optimal target configuration coordinates.

[0161] S106. Obtain the reference position information of the target node in the UAV cluster, and use the reference position information to perform coordinate transformation on the target configuration coordinates to obtain the global position coordinates of each node in the UAV cluster, thus completing the dynamic self-organizing network positioning.

[0162] Optionally, step S106, which involves obtaining the reference position information of the target nodes in the UAV cluster and using the reference position information to perform coordinate transformation on the target configuration coordinates to obtain the global position coordinates of each node in the UAV cluster, and completing the dynamic self-organizing network positioning process, may specifically include:

[0163] S1061. Perform geometric space restoration mapping on the target configuration coordinates to obtain a local configuration matrix representing the relative distribution relationship of nodes in the UAV cluster.

[0164] Geometric space reconstruction refers to the process of mapping configuration coordinate points located in manifold space, i.e., abstract geometric space, back to specific Euclidean physical space through specific inverse transformations or projection operations. The local configuration matrix is ​​the coordinate matrix obtained after reconstruction, describing the relative positions of all nodes in the cluster within a custom local coordinate system.

[0165] During the implementation of the scheme, the optimized orthogonal basis matrix is ​​first extracted from the target configuration coordinates output in step S105. Secondly, by utilizing the inverse transformation concept of multidimensional scaling or singular value decomposition, the... This is converted to Euclidean coordinates that include a distance scale. Specifically, this can be achieved by solving an optimization problem to find the coordinates of the coordinates. The corresponding Gram matrix is ​​then decomposed to obtain the coordinates. For example, assume the optimized target configuration coordinates are... Perform singular value decomposition or direct eigenvalue decomposition on it to obtain the local configuration matrix. = Each column of this matrix represents the relative coordinates of a node, such as... wait.

[0166] Specifically, due to the configuration coordinate points located within the geometric configuration constraint space Essentially, it is a representation of the observed subspace, including the geometric structural properties of the cluster topology, such as relative orientation relationships, but it does not yet possess the absolute length scale in physical space. Therefore, firstly, multidimensional scaling (MDS) or scaling based on... Derived Gram matrix By performing eigenvalue decomposition, a set of dimensionless initial relative coordinate points is reconstructed. .

[0167] After obtaining the initial set of relative coordinate points Subsequently, this application also includes a scale calibration operation on the local configuration matrix. Since the angle of arrival (AoA) observation itself has scale uncertainty, the scale operator needs to be decoupled using at least two anchor nodes in the cluster that have reference position information.

[0168] First, calculate the relative distance: in the initial set of relative coordinate points. In the middle, extract two reference nodes. and Corresponding relative coordinates and And calculate its Euclidean distance. Next, the physical distance is obtained: based on the obtained reference position information of the target node. and Calculate the actual physical distance between the two nodes in the global geographic coordinate system. Then, the scale factor is determined: the scale factor is determined by calculating the ratio of the actual physical distance to the relative distance. Finally, scale-reduction mapping: using the aforementioned scale factor The initial set of relative coordinate points is scaled, i.e. This yields a local configuration matrix containing information about the actual physical scale. S1062. Extract the relative coordinates of the corresponding target nodes in the local configuration matrix, calculate the rotation and translation features between the relative coordinates of the nodes and the reference position information, and determine the coordinate transformation parameters.

[0169] Rotation and translation features refer to the geometric parameters describing the deviation between the local relative coordinate system and the global geographic coordinate system, mainly including the rotation matrix, translation vector, and the crucial scaling factor. Since the initial configuration based on manifold optimization only possesses geometric topology and lacks absolute distance information, the coordinate transformation parameters are the rotation matrix, translation vector, and scaling factor used to recover the physical spatial dimensions, calculated through anchor point constraints. During the implementation process, the first step is to... The relative coordinate column vectors corresponding to the anchor nodes are extracted. Simultaneously, the reference position information of these two nodes, measured by GPS, is obtained. Next, the scale factor is determined by calculating the ratio of the actual physical distance between the anchor nodes in the global coordinate system to their relative distance in the local coordinate system. Then, using Procrustes analysis or the Kabsch algorithm, a least-squares problem is constructed based on scale alignment to find the optimal rotation matrix that minimizes the error between the relative coordinates after scaling, rotation, and translation transformations and the reference position. Translation vector For example, suppose we extract the relative coordinates of nodes 1 and 2. and reference position First, the physical distance ratio between the two is calculated to determine the scale factor, and then the rotation angle is calculated using an algorithm. That is, the corresponding rotation matrix Translation vector This group These are the defined coordinate transformation parameters, which enable the precise mapping of UAV swarms from a local perception configuration to a global geographic coordinate system.

[0170] S1063. By performing an affine transformation on the local configuration matrix using coordinate transformation parameters, the UAV swarm is mapped to the global geographic coordinate system, thus obtaining the global position coordinates of each node in the UAV swarm.

[0171] Affine transformation is a geometric transformation that preserves linearity and parallelism. It includes operations such as rotation, translation, and scaling, and is used to map a set of points in one coordinate system to another without loss of quality. Global position coordinates refer to the precise coordinates of each node in the cluster under a unified geographic reference system, such as WGS-84, after the transformation.

[0172] During the implementation of the scheme, the coordinate transformation parameters determined in S1062 were used. For the local configuration matrix in S1061 The transformation formula is applied to the relative coordinates of each column, i.e., each node: ,in Relative coordinates For rotation matrix, The result is calculated using the translation vector. This refers to the absolute geographical location of each node. For example, for a regular node 3, its relative coordinates are... Substitute into the formula to calculate. Assuming we get This means that node 3 is located at longitude (X-axis coordinate) 150.5 and latitude (Y-axis coordinate) 220.8. The final output includes the set of coordinates of all nodes, completing the dynamic self-organizing network positioning.

[0173] This embodiment not only solves the common problem of accurate relative position but inaccurate absolute position in self-organizing network positioning, but also further corrects the overall rotation and translation errors through multi-anchor point fusion, realizing seamless connection from local perception to global positioning of UAV swarm.

[0174] Figure 4 This is a schematic diagram illustrating a specific implementation of a dynamic self-organizing network positioning system based on multi-UAV cooperation, as provided in this application embodiment. (Refer to...) Figure 4 The system may include:

[0175] 410 Acquisition Module is used to acquire the arrival angle spectrum data of the signal transmission between the node and the neighboring node in the UAV cluster, and to acquire the bit error rate distribution data generated by statistically analyzing the bit error rate of data transmission between the communication links.

[0176] The 420 generation module is used to construct a spatial covariance matrix based on the angle of arrival spectrum data, map the bit error rate distribution data to the channel weight factor, and correct the covariance matrix through the channel weight factor to generate the target covariance matrix.

[0177] The 430 module is used to perform eigenvalue decomposition on the target covariance matrix to determine the noise feature space. It then uses a preset array manifold vector to perform orthogonal projection on the noise feature space to determine the relative orientation vector between each node and its neighboring nodes, and constructs the observation matrix of the node based on the relative orientation vector.

[0178] The 430 building module is also used to map the observation matrix of all nodes to a preset geometric configuration constraint space, transform it into configuration coordinate data representing the spatial geometric configuration in the UAV cluster, and construct a topological cost function including formation geometric constraints based on the configuration coordinate data;

[0179] The 420 generation module is also used to calculate the gradient of the topological cost function in the geometric configuration constraint space, and iteratively update the position coordinates of each neighboring node along the opposite direction of the gradient until the topological cost function converges to obtain the target configuration coordinates.

[0180] The 440 conversion module is used to obtain the reference position information of the target node in the UAV cluster, and use the reference position information to perform coordinate transformation on the target configuration coordinates to obtain the global position coordinates of each node in the UAV cluster, thus completing the dynamic self-organizing network positioning.

[0181] The dynamic ad hoc network positioning system based on multi-UAV cooperation in this application embodiment is used to implement the aforementioned dynamic ad hoc network positioning method based on multi-UAV cooperation. Therefore, the specific implementation of the dynamic ad hoc network positioning system based on multi-UAV cooperation can be found in the embodiment section of the dynamic ad hoc network positioning method based on multi-UAV cooperation mentioned above. The specific implementation can be referred to the description of the corresponding embodiments, and will not be repeated here.

[0182] Figure 5 A schematic diagram of the hardware structure of an electronic device provided in one embodiment of this application is shown.

[0183] The electronic device may include a processor 510 and a memory 520 storing computer program instructions.

[0184] Specifically, the processor 510 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0185] Memory 520 may include mass storage for data or instructions. For example, and not limitingly, memory 520 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 520 may include removable or non-removable (or fixed) media. Where appropriate, memory 520 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 520 is non-volatile solid-state memory.

[0186] Memory may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, and electrical, optical, or other physical / tangible memory storage devices. Therefore, typically, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to the first aspect of this disclosure.

[0187] The processor 510 reads and executes computer program instructions stored in the memory 520 to implement any of the dynamic self-organizing network positioning methods based on multi-UAV cooperation in the above embodiments.

[0188] In one example, the electronic device may also include a communication interface 530 and a bus 540. Wherein, such as Figure 5 As shown, the processor 510, memory 520, and communication interface 530 are connected through bus 540 and complete communication with each other.

[0189] The communication interface 530 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0190] Bus 540 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 540 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, any suitable bus or interconnect is contemplated herein.

[0191] The electronic device can execute the dynamic self-organizing network positioning method based on multi-UAV cooperation in the embodiments of this application, thereby realizing the dynamic self-organizing network positioning method based on multi-UAV cooperation described in conjunction with the accompanying drawings.

[0192] Furthermore, in conjunction with the dynamic ad hoc network positioning method based on multi-UAV cooperation in the above embodiments, this application embodiment can provide a computer-readable storage medium for implementation. This computer-readable storage medium stores computer program instructions; when executed by a processor, these computer program instructions implement any of the dynamic ad hoc network positioning methods based on multi-UAV cooperation in the above embodiments.

[0193] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0194] The functional blocks shown in the above-described structural diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0195] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0196] The aspects of this application have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by dedicated hardware performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0197] The foregoing provides a detailed description of a dynamic ad hoc network positioning method and system based on multi-UAV cooperation provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A dynamic self-organizing network positioning method based on multi-UAV cooperation, characterized in that, The method includes: Acquire the arrival angle spectrum data of the signal transmission between the nodes and neighboring nodes in the UAV cluster, and acquire the bit error rate distribution data generated by statistically analyzing the bit error rate of data transmission between the communication links; A spatial covariance matrix is ​​constructed based on the arrival angle spectrum data, and the bit error rate distribution data is mapped to a channel weight factor. The covariance matrix is ​​then corrected using the channel weight factor to generate a target covariance matrix. The target covariance matrix is ​​subjected to eigenvalue decomposition to determine the noise feature space. Orthogonal projection is performed on the noise feature space using a preset array manifold vector to determine the relative orientation vector between each node and its neighboring nodes. The observation matrix of the node is then constructed based on the relative orientation vector. The observation matrices of all nodes are mapped to a preset geometric configuration constraint space, which is then transformed into configuration coordinate data representing the spatial geometric configuration of the UAV cluster. A topological cost function including formation geometric constraints is constructed based on the configuration coordinate data. Calculate the gradient of the topological cost function within the geometric configuration constraint space, and iteratively update the position coordinates of each neighboring node along the opposite direction of the gradient until the topological cost function converges to obtain the target configuration coordinates; The reference position information of the target node in the UAV cluster is obtained, and the target configuration coordinates are transformed using the reference position information to obtain the global position coordinates of each node in the UAV cluster, thus completing the dynamic self-organizing network positioning.

2. The method according to claim 1, characterized in that, The method further includes: The array antenna unit mounted on the node is used to collect radio frequency sampling signals transmitted by neighboring nodes, and the arrival angle spectrum data is obtained by calculating the spatial phase difference of the radio frequency sampling signals. The total number of transmitted bits and the total number of error bits between the statistical node and its neighboring nodes during the data exchange process are counted, and the ratio of the total number of error bits to the total number of transmitted bits is calculated to obtain the bit error rate distribution data. Extract the node number information and acquisition time information that are commonly associated from the arrival angle spectrum data and the bit error rate distribution data, and establish the spatiotemporal association between the arrival angle spectrum data and the bit error rate distribution data based on the node number information and the acquisition time information.

3. The method according to claim 1, characterized in that, The process of constructing a spatial covariance matrix based on the arrival angle spectrum data, mapping the bit error rate distribution data to a channel weight factor, and correcting the covariance matrix using the channel weight factor to generate a target covariance matrix includes: The spatial covariance matrix is ​​obtained by multiplying the vector dimensions of all the arrival angle spectrum data by conjugate transpose and taking the statistical mean. The channel weight factor is obtained by numerically transforming the bit error rate distribution data using a preset mapping function. The spatial covariance matrix is ​​weighted element-wise according to the channel weight factor to adjust the distribution intensity of each feature component in the spatial covariance matrix and generate the target covariance matrix.

4. The method according to claim 1, characterized in that, The process of performing eigenvalue decomposition on the target covariance matrix to determine the noise feature space, performing orthogonal projection on the noise feature space using a preset array manifold vector to determine the relative orientation vector between each node and its neighboring nodes, and constructing the observation matrix of the node based on the relative orientation vector includes: The target covariance matrix is ​​subjected to eigenvalue decomposition, and a preset number of feature vectors are extracted in order of eigenvalue values ​​from low to high to generate noise feature vectors. A noise feature space is constructed using the noise feature vectors, and a preset array manifold vector is orthogonally projected onto the noise feature space to generate spatial pseudospectral data representing the distribution of signal energy with spatial angle. Extract the numerical maxima of the spatial pseudospectral data within a preset angle search area, and determine the relative orientation vector of each neighboring node belonging to the same node relative to the node. According to the preset node numbering order, the relative orientation vectors belonging to the same node are combined to obtain the observation matrix of each node.

5. The method according to claim 1, characterized in that, The process involves mapping the observation matrices of all nodes to a preset geometric configuration constraint space, transforming them into configuration coordinate data representing the spatial geometric configuration of the UAV cluster, and constructing a topology cost function including formation geometric constraints based on the configuration coordinate data, including: The observation matrices corresponding to all nodes are subjected to standard orthogonalization to extract orthogonal basis data; The orthogonal basis data is mapped to the geometric configuration constraint space, and the configuration coordinate data representing the geometric distribution of the cluster is obtained by determining the subspace configuration point corresponding to each node in the geometric configuration constraint space. Based on the relative positional relationship between each node in the formation geometric constraints, formation topology data is constructed, and the formation topology data is transformed into reference configuration data located in the geometric configuration constraint space through orthogonal mapping; The projection deviation values ​​of the configuration coordinate data and the reference configuration data in the geometric configuration constraint space are calculated, the projection deviation values ​​are squared and accumulated to generate error scalar data, and the error scalar data is determined as the topological cost function.

6. The method according to claim 5, characterized in that, The calculation of the gradient of the topological cost function within the geometric constraint space, and the iterative update of the position coordinates of each neighboring node along the opposite direction of the gradient until the topological cost function converges to obtain the target configuration coordinates, includes: By performing partial derivative calculations on the topological cost function, the initial gradient data of the configuration coordinate data at the current position is determined; The initial gradient data is projected onto the local constraint surface of the geometric constraint space to obtain the tangential projection component located in the local constraint surface, and the tangential projection component is determined as the changing gradient to serve as the iterative search direction. The iterative search direction is multiplied by a preset step size coefficient to obtain the position displacement vector of each neighbor node in the local constraint surface. The position displacement vector is then mapped to the geometric constraint space using a preset nonlinear mapping operator to update the iterative position coordinates of each neighbor node in the geometric constraint space. Using the updated iterative position coordinates as input, the iterative process from calculating the initial gradient data to updating the iterative position coordinates is executed repeatedly until the numerical change of the topological cost function between two adjacent iterations is less than a preset convergence threshold, thereby generating the target configuration coordinates.

7. The method according to claim 1, characterized in that, The process of obtaining the reference position information of the target node in the UAV cluster, and using the reference position information to perform coordinate transformation on the target configuration coordinates to obtain the global position coordinates of each node in the UAV cluster, thereby completing dynamic ad hoc network positioning, includes: Perform a geometric space reconstruction mapping on the target configuration coordinates to obtain a local configuration matrix representing the relative distribution relationship of nodes in the UAV cluster; Extract the relative coordinates of the nodes corresponding to the target node in the local configuration matrix, and calculate the rotation and translation features between the relative coordinates of the nodes and the reference position information to determine the coordinate transformation parameters; By performing an affine transformation on the local configuration matrix using the coordinate transformation parameters, the UAV cluster is mapped to a global geographic coordinate system, thereby obtaining the global position coordinates of each node in the UAV cluster.

8. A dynamic self-organizing network positioning system based on multi-UAV cooperation, characterized in that, include: The acquisition module is used to acquire the arrival angle spectrum data of the signal transmission between the nodes and neighboring nodes in the UAV cluster, and to acquire the bit error rate distribution data generated by statistically analyzing the bit error rate of data transmission between the communication links. The generation module is used to construct a spatial covariance matrix based on the arrival angle spectrum data, map the bit error rate distribution data into a channel weight factor, and modify the covariance matrix through the channel weight factor to generate a target covariance matrix. The construction module is used to perform eigenvalue decomposition on the target covariance matrix to determine the noise feature space, perform orthogonal projection on the noise feature space using a preset array manifold vector, determine the relative orientation vector between each node and its neighboring nodes, and construct the observation matrix of the node based on the relative orientation vector. The construction module is also used to map the observation matrix of all nodes to a preset geometric configuration constraint space, convert it into configuration coordinate data representing the spatial geometric configuration in the UAV cluster, and construct a topological cost function including formation geometric constraints based on the configuration coordinate data; The generation module is also used to calculate the gradient of the topological cost function in the geometric configuration constraint space, and iteratively update the position coordinates of each neighboring node along the opposite direction of the gradient until the topological cost function converges to obtain the target configuration coordinates. The conversion module is used to obtain the reference position information of the target node in the UAV cluster, and use the reference position information to perform coordinate transformation on the target configuration coordinates to obtain the global position coordinates of each node in the UAV cluster, thereby completing dynamic self-organizing network positioning.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the dynamic self-organizing network positioning method based on multi-UAV cooperation as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the implementation of the dynamic self-organizing network positioning method based on multi-UAV cooperation as described in any one of claims 1 to 7.

Images