A distributed radar non-coherent direct positioning method and device for data anomaly scenarios, a terminal and a medium

Abstract

This invention provides a distributed radar noncoherent direct localization method, device, terminal, and medium for data anomaly scenarios, relating to the field of radar technology. The method includes: acquiring echo signal data received by each observation array; constructing a signal model characterizing the mapping relationship between the echo signal data and potential source locations, where the common detection area of ​​each observation array is a two-dimensional grid, the grid point coordinates correspond to the potential source locations, the signal model does not contain phase synchronization parameters between observation arrays, and is adapted to the noncoherent characteristics of the signal data; and constructing a system based on the signal model... 2,1 This invention addresses the group sparse optimization problem of norms. A pre-defined optimization algorithm is used to solve the problem, yielding a row sparse matrix. Based on this matrix, the true source localization result is determined, and the column vector elements of the row sparse matrix correspond to the potential source reflection energy estimation results at the grid points. This invention utilizes echo signal data from various observation arrays to achieve primary localization, eliminating secondary localization errors and reducing abnormal noise interference.

Description

Technical Field

[0001] This invention relates to the field of radar technology, and in particular to a distributed radar noncoherent direct positioning method, device, terminal, and medium for data anomaly scenarios. Background Technology

[0002] Currently, distributed array radar, through the collaborative work of multiple spatially distributed radar nodes, can effectively improve the detection range and data dimension, and is an important technical approach to achieve high-precision target positioning.

[0003] However, existing distributed positioning methods typically rely on high-precision time or phase synchronization between radar nodes and have high requirements for the integrity and quality of radar node data. When some radar nodes experience abnormal sampling data due to environmental interference, hardware failure, or communication interruption, the performance of traditional positioning methods will significantly degrade or even fail. For example, methods based on secondary positioning or data fusion lack sufficient robustness when node data is abnormal, and most multi-target direct positioning methods usually require known target numbers or rely on strict system synchronization conditions, limiting their applicability in scenarios where some radar nodes are abnormal.

[0004] Therefore, in response to the practical problem of abnormal data from some radar nodes in a distributed array radar system, how to maintain high positioning accuracy to achieve robust multi-target positioning is a matter of great concern to those skilled in the art. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a distributed radar noncoherent direct positioning method, device, terminal and medium for data anomaly scenarios, aiming to solve the problem of decreased positioning accuracy and low positioning robustness caused by data anomalies in some radar nodes in the distributed array radar system in the prior art.

[0006] The technical solution adopted by this invention to solve the technical problem is as follows: In a first aspect, the present invention discloses a distributed radar noncoherent direct localization method for data anomaly scenarios, wherein the method includes: Acquire echo signal data received by each observation array in a distributed radar system; A signal model is constructed to characterize the mapping relationship between the echo signal data and the potential source location; wherein, the common detection area of ​​each observation array is a two-dimensional grid, the coordinates of each grid point in the two-dimensional grid correspond to a potential source location, and the signal model does not include the phase synchronization parameters between the observation arrays, so as to adapt to the incoherent characteristics of the echo signal data; A localization problem is constructed based on the aforementioned signal model; the localization problem is based on... 2,1The group sparsity optimization problem of norms; The localization problem is solved using a preset optimization algorithm to obtain a corresponding row sparse matrix, and the localization result of the real source is determined based on the row sparse matrix; each element on the column vector of the row sparse matrix corresponds to a grid point and the reflection energy estimation result of the potential source exists.

[0007] Optionally, constructing a signal model characterizing the mapping relationship between the echo signal data and the potential source location includes: Based on the known coordinates of each observation array and the coordinates of the grid points in the two-dimensional grid, calculate the coordinate difference between the known coordinates of each observation array and the coordinates of the grid points; The azimuth angle between each observation array and the potential source is determined based on the coordinate difference, and a steering vector for calculating the direction of each potential source is constructed based on the azimuth angle. A steering matrix is ​​constructed using the steering vectors of each potential source direction, and a signal model is constructed based on the steering matrix.

[0008] Optionally, the signal model is: ; ; ; ; ; ; ; ; ; ; in, Indicates the first The first observation array received the first Snap array signal, For source signals, For noise vectors, for 3D guided matrix, The number of array elements on each observation array. The number of grid points, The number of observation arrays, For the first The first potential source and the first Azimuth angles between observation arrays For the first The observation array is for the first The steering vector of a potential information source. At the speed of light, For the first The first potential source and the first One observation array in Coordinate difference along the axis, For the first The first potential source and the first One observation array in Coordinate difference along the axis, For the first One potential source location, For the first The known coordinates of an observation array.

[0009] Optionally, the basis 2,1 The group sparse optimization problem of norm is: ; ; ; ; ; ; ; ; in, This is the estimated reflection energy when the real source exists at the g-th grid point on the m-th observation array. To estimate the unknown reflection energy variable when the real source exists at the g-th grid point on the m-th observation array, Let covariance matrix be the variance matrix. For the first The grid point and the first Azimuth angles between observation arrays For including all azimuth angles 3D matrix For along Number of grid cells along the axis, For along Number of grid cells along the axis, For mesh stacking 3D guided matrix, For the first The coordinates of each grid point For the first The known coordinates of the observation array For the first The coordinates of the first grid point and the first One observation array in Coordinate difference along the axis, For the first The coordinates of the first grid point and the first One observation array in Coordinate difference along the axis, For the first The constraint matrix of the observation array, The result is the constraint matrices of all observation arrays stacked column-wise. It is a convex function. This is the regularization parameter.

[0010] Optionally, the row sparse matrix is: .

[0011] Optionally, solving the localization problem using a preset optimization algorithm includes: The localization problem is solved using a convex optimization solver, a complex-valued conjugate gradient descent algorithm, or an iterative reweighted least squares algorithm.

[0012] Optionally, the determination of the location result of the true information source based on the row sparse matrix includes: Determine the non-zero elements in the row sparse matrix, and identify the grid points corresponding to the non-zero elements as target grid points where a real information source exists; The coordinate information of the target grid point in the preset rectangular coordinate system is determined to obtain the positioning result of the real information source.

[0013] Secondly, the present invention also discloses a distributed radar noncoherent direct positioning device for data anomaly scenarios, wherein the device comprises: The data acquisition module is used to acquire echo signal data received by each observation array in the distributed radar system; The signal model construction module is used to construct a signal model that characterizes the mapping relationship between the echo signal data and the potential source location; wherein, the common detection area of ​​each observation array is a two-dimensional grid, the coordinates of each grid point in the two-dimensional grid correspond to a potential source location, and the signal model does not include phase synchronization parameters between the observation arrays, so as to adapt to the incoherent characteristics of the echo signal data; The localization problem construction module is used to construct a localization problem based on the signal model; the localization problem is based on... 2,1 The group sparsity optimization problem of norms; The localization problem solving module is used to solve the localization problem using a preset optimization algorithm to obtain the corresponding row sparse matrix; The source localization module is used to determine the localization result of the real source based on the row sparse matrix; each element of the column vector of the row sparse matrix corresponds to a grid point and has the reflection energy estimation result of a potential source.

[0014] Thirdly, the present invention discloses a terminal, comprising: a memory, a processor, and a distributed radar noncoherent direct localization program for data anomaly scenarios stored in the memory and executable on the processor, wherein the distributed radar noncoherent direct localization program for data anomaly scenarios, when executed by the processor, implements the steps of the distributed radar noncoherent direct localization method for data anomaly scenarios as described above.

[0015] Fourthly, the present invention discloses a computer-readable storage medium storing a computer program that can be executed to implement the steps of the distributed radar incoherent direct localization method for data anomaly scenarios as described above.

[0016] This invention provides a distributed radar incoherent direct localization method, apparatus, terminal, and medium for data anomaly scenarios. The distributed radar incoherent direct localization method for data anomaly scenarios includes: acquiring echo signal data received by each observation array in a distributed radar system; constructing a signal model characterizing the mapping relationship between the echo signal data and potential source locations; wherein the common detection area of ​​each observation array is a two-dimensional grid, the coordinates of each grid point in the two-dimensional grid correspond to a potential source location, and the signal model does not include phase synchronization parameters between the observation arrays to adapt to the incoherent characteristics of the echo signal data; and constructing a localization problem based on the signal model; the localization problem is based on... 2,1 The present invention addresses the group sparse optimization problem of the norm; it solves the localization problem using a preset optimization algorithm to obtain a corresponding row sparse matrix, and determines the localization result of the true source based on the row sparse matrix; each element of the column vector of the row sparse matrix corresponds to the reflection energy estimation result of a potential source at a grid point. Thus, the present invention achieves primary localization by jointly utilizing the echo signal data collected by all observation arrays, that is, by stacking and optimizing the observation matrices of all observation arrays for each grid point, thereby achieving primary localization of the true source, eliminating the need for secondary localization, removing secondary localization errors, and reducing abnormal noise interference. This allows for better handling of abnormal noise in the detection area and eliminating secondary noise error interference introduced by secondary localization. Even when some array radar nodes have abnormal sampling data and high-precision time or phase synchronization is not required, it still ensures high localization accuracy and robustness, achieving robust multi-target localization. Attached Figure Description

[0017] Figure 1 This is a flowchart of a preferred embodiment of the distributed radar noncoherent direct localization method for data anomaly scenarios in this invention; Figure 2 This is a schematic diagram of a specific distributed radar system disclosed in this invention; Figure 3 This is a schematic diagram of the localization result of a specific CVX optimization algorithm disclosed in this invention; Figure 4 This is a schematic diagram of the localization result of a specific complex conjugate gradient descent algorithm disclosed in this invention; Figure 5 This is a schematic diagram of the localization result of a specific iterative reweighted least squares algorithm disclosed in this invention; Figure 6 This is a schematic diagram illustrating the variation of the positioning accuracy of the three algorithms under the general Gaussian case as a function of the signal-to-noise ratio, as disclosed in this invention. Figure 7 This is a schematic diagram illustrating the variation of positioning accuracy of the three algorithms under general Gaussian conditions with snapshots, as disclosed in this invention. Figure 8 This is a schematic diagram illustrating the variation of the positioning accuracy of four algorithms with the signal-to-noise ratio under noise conditions with anomaly ratios of 0.1% and 0.3% as disclosed in this invention. Figure 9 This is a schematic diagram illustrating the variation of positioning accuracy of four algorithms with snapshots under specific noise conditions with anomaly ratios of 0.1% and 0.3%. Figure 10 This is a schematic diagram illustrating the variation of positioning accuracy of the two algorithms with signal-to-noise ratio under Laplace noise conditions disclosed in this invention; Figure 11 This is a schematic diagram illustrating the variation of positioning accuracy of the two algorithms under Laplace noise conditions with snapshots, as disclosed in this invention. Figure 12 This is a functional principle block diagram of a preferred embodiment of the distributed radar noncoherent direct positioning device for data anomaly scenarios in this invention. Figure 13 This is a functional principle block diagram of a preferred embodiment of the terminal in this invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer and more explicit, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0019] The demand for radar-based target detection and localization is growing for payload- and cost-sensitive platforms such as drones and smart cars. In these scenarios, reducing complexity and cost while ensuring positioning accuracy and robustness has become a key challenge for the development of distributed radar technology. Maintaining effective localization capabilities, especially when some radar nodes experience data anomalies or failures, is crucial for improving applicability in complex environments.

[0020] Currently, distributed array radar, through the collaborative work of multiple spatially distributed radar nodes, can effectively improve the detection range and data dimension, and is an important technical approach to achieve high-precision target positioning.

[0021] However, existing distributed positioning methods typically rely on high-precision time or phase synchronization between radar nodes and have high requirements for the integrity and quality of radar node data. When some radar nodes experience abnormal sampling data due to environmental interference, hardware failure, or communication interruption, the performance of traditional positioning methods will significantly degrade or even fail. For example, methods based on secondary positioning or data fusion lack sufficient robustness when node data is abnormal, and most multi-target direct positioning methods usually require known target numbers or rely on strict system synchronization conditions, limiting their applicability in scenarios where some radar nodes are abnormal.

[0022] Therefore, addressing the practical problem of data anomalies at some radar nodes in a distributed array radar system, and maintaining high positioning accuracy to achieve robust multi-target positioning, is a crucial concern for those skilled in the art. To this end, this application provides a distributed radar noncoherent direct positioning scheme for data anomaly scenarios. This scheme achieves robust multi-target direct positioning without relying on strict synchronization conditions or prior information about the number of targets, improving the reliability and adaptability of distributed radar in complex application environments. Thus, even when data anomalies occur at some radar nodes in a distributed array radar system, it still ensures high positioning accuracy and robustness, achieving robust multi-target positioning.

[0023] It should be noted that distributed array radar target direct localization technology is a cutting-edge direction in the field of modern radar signal processing. It avoids the problems of error propagation and accumulation, as well as the time delay matching and data correlation ambiguity issues in traditional direction finding and cross-localization methods by working together with multiple spatially distributed receiving units and the incoherence of signals, thereby improving positioning accuracy and robustness.

[0024] Please see Figure 1 , Figure 1 This is a flowchart of the distributed radar incoherent direct localization method for data anomaly scenarios in this invention. For example... Figure 1 As shown in the embodiment of the present invention, the distributed radar noncoherent direct localization method for data anomaly scenarios includes: Step S11: Obtain echo signal data received by each observation array in the distributed radar system.

[0025] Among them, see Figure 2 As shown, the distributed radar system comprises multiple observation arrays. Each array is a small-sized radar node that can be mounted on a small unmanned aerial vehicle (UAV). This allows the UAV to be controlled so that each radar node's beam covers the same region of interest. By jointly processing the received signals from the radar nodes, the positions of multiple targets within that region of interest can be determined. Furthermore, in this distributed radar system, the positions of each radar node are known, but high-precision time or phase synchronization between the signals does not require. It should be noted that the phase or delay relationship between the received array signals from each radar node is unknown.

[0026] In this embodiment, echo signal data received by each observation array is acquired. This echo signal data provides observation values ​​and a data foundation for the subsequent construction of signal models. The echo signal data contains the spatial phase characteristics (manifested through phase difference or time delay) and noise characteristics of the real source. Although the phases of the radar nodes in each array are not synchronized (i.e., incoherent), the echo signal within each node contains the angular information of the real source relative to that radar node. In other words, the echo signal data is an observation vector containing target azimuth information. For example, the echo signal data represents the complex sampling sequence of the target reflection signal on each array element under incoherent conditions.

[0027] Step S12: Construct a signal model characterizing the mapping relationship between the echo signal data and the potential source location; wherein, the common detection area of ​​each observation array is a two-dimensional grid, the coordinates of each grid point in the two-dimensional grid correspond to a potential source location, and the signal model does not include phase synchronization parameters between the observation arrays, in order to adapt to the incoherent characteristics of the echo signal data.

[0028] In this embodiment, after acquiring the echo signal data, a signal model (i.e., a radar echo signal model) is constructed. This model can be based on the geometric topology of the distributed radar system, representing the mapping relationship between the echo signal data and the potential source locations. For example, the positions and antenna array configurations of each array radar node in the distributed radar system are determined, and a signal model representing the mapping relationship between the echo signal data and the potential source locations is constructed based on the positions of each array radar node and the antenna array configuration.

[0029] It should be noted that, to ensure no loss of generality among all observation arrays, the common detection area of ​​all observation arrays in the Cartesian coordinate system is considered as a square, and this area is divided into... 1 grid, of which For along Number of grid cells along the axis, For along The number of grid points along the axis. The position of each grid point can be determined by... The determined ones, among which , .

[0030] Specifically, based on the known coordinates of each observation array and the coordinates of the grid points within the two-dimensional grid, the coordinate difference between the known coordinates of each observation array and the coordinates of the grid points is calculated. The azimuth angle between each observation array and the potential signal source is determined based on the coordinate difference, and a steering vector for each potential signal source direction is constructed based on the azimuth angle. A steering matrix is ​​constructed using the steering vectors for each potential signal source direction, and a signal model is built based on the steering matrix. It can be understood that the steering matrix is ​​determined by the known coordinates of the observation array and the coordinates of the grid points, and each column of the steering matrix corresponds to a potential signal source location (grid point).

[0031] The expression for the signal model can be: ; ; ; ; ; ; ; ; ; ; in, Indicates the first The first observation array received the first Snap array signal, For source signals, For noise vectors, for 3D guided matrix, The number of array elements on each observation array. The number of grid points, The number of observation arrays, For the first The first potential source and the first Azimuth angles between observation arrays For the first The observation array is for the first The steering vector of a potential information source. At the speed of light, For the first The first potential source and the first One observation array in Coordinate difference along the axis, For the first The first potential source and the first One observation array in Coordinate difference along the axis, For the first One potential source location, For the first The known coordinates of an observation array.

[0032] It should be noted that in real-world environments, there is interference from various outliers mixed with noise. This refers to anomaly noise models, such as Gaussian mixture noise and Laplace noise. Gaussian mixture noise is a composite noise formed by superimposing multiple Gaussian noise components with different means and variances according to certain weights. A two-component Gaussian mixture model is used to generate outlier noise, and its probability density function is: ;in, Indicates the first The proportion of each component, Indicates the first The variance of each component, and ,and The second component corresponds to outliers. Specifically, , , , , , , , , , .

[0033] Furthermore, Laplace noise is a continuous probability distribution with a peak at the mean and a steeper tail than Gaussian noise, making it commonly used for modeling noise with bursty characteristics. Its probability density function is defined as: .

[0034] Step S13: Construct a localization problem based on the signal model; the localization problem is based on... 2,1 The group sparsity optimization problem of norm.

[0035] In this embodiment, for scenarios where some array radar node sampling data is abnormal and multi-target incoherent localization is required, a corresponding localization problem is established based on a signal model. This localization problem is based on... 2,1 The problem of group sparsity optimization based on norms. It is understandable that this involves constructing a system based on a signal model. 2,1 The group sparsity optimization problem of norm.

[0036] Specifically, when all noise applied to a distributed radar system is considered as a variable to be estimated in terms of reflected energy, based on... 2,1 The group sparse optimization problem of norm is: ; ; ; ; ; ; ; ; in, This is the estimated reflection energy when the real source exists at the g-th grid point on the m-th observation array. To estimate the unknown reflection energy variable when the real source exists at the g-th grid point on the m-th observation array, Let covariance matrix be the variance matrix. For the first The grid point and the first Azimuth angles between observation arrays For including all azimuth angles 3D matrix For along Number of grid cells along the axis, For along Number of grid cells along the axis, For mesh stacking 3D guided matrix, For the first The coordinates of each grid point For the first The known coordinates of the observation array For the first The coordinates of the first grid point and the first One observation array in Coordinate difference along the axis, For the first The coordinates of the first grid point and the first One observation array in Coordinate difference along the axis, For the first The constraint matrix of the observation array, The result is the constraint matrices of all observation arrays stacked column-wise. It is a convex function. This is the regularization parameter.

[0037] It should be noted that this can be achieved by examining all and Azimuth of corresponding grid points Stacking is used to construct a system containing all azimuth angles. 3D matrix In a pre-defined Cartesian coordinate system, the two-dimensional grid is converted into an azimuth domain, since the azimuth set corresponding to each grid is unique. Furthermore, when... hour, .

[0038] Step S14: Solve the localization problem using a preset optimization algorithm to obtain the corresponding row sparse matrix, and determine the localization result of the real source based on the row sparse matrix; each element on the column vector of the row sparse matrix corresponds to a grid point and the reflection energy estimation result of the potential source exists.

[0039] In this embodiment, a preset optimization algorithm is used to solve the localization problem. Specifically, a convex optimization solver, a complex-valued conjugate gradient descent algorithm, or an iterative reweighted least squares algorithm is used to solve the localization problem, i.e., the solution is based on... 2,1 The group sparse optimization problem of the norm yields a row sparse matrix. Each element in the column vector of the row sparse matrix corresponds to a grid point containing the reflection energy estimation result of a potential signal source. Based on this row sparse matrix, the location of the true signal source is determined. Specifically, the non-zero elements in the row sparse matrix are identified, and the grid points corresponding to these non-zero elements are designated as target grid points containing the true signal source. The coordinates of these target grid points in a predefined Cartesian coordinate system are then determined to obtain the location of the true signal source. In essence, the row in the row sparse matrix with a non-zero value indicates the grid point where the true signal source is located. For example, the index of a non-zero row in the row sparse estimation matrix corresponds to the grid number of the detection area. This grid number is then mapped to the Cartesian coordinates of the grid points in the detection area to obtain the location of the true signal source.

[0040] The row sparse matrix is: .

[0041] It should be noted that each real source has the same sparsity across the entire two-dimensional grid as each observation array. Therefore, the observation matrices of all observation arrays for each grid point can be stacked together for optimization, thereby achieving localization of the real source in one step. This can be achieved through a dimension... matrix ,in, The total number of grid points. The number of observation arrays, Let be the number of elements in each observation array. And for each observation array, the dimension is... Unknown reflection energy variable to be estimated Stacked by row , where row vector It is a matrix The Okay, satisfied. Each row vector All elements in the array are at the same grid point. Related, among which Therefore, these elements all have the same two-dimensional support area.

[0042] in, It is a convex function, that is It is a convex function, which can be approximated by a concave function. The norm is used to obtain row sparse solutions, and the parameters can be changed. ,get and The optimal trade-off curve between them.

[0043] In the first specific implementation, a convex optimization solver is used to solve the localization problem. It is understood that the above is based on... 2,1 The objective function of the norm-based group sparse optimization problem is a convex function. Therefore, the CVX tool (i.e., the convex optimization solver) can be used to solve the group sparse optimization problem. As an unconstrained optimization problem, the group sparse optimization problem can be further expressed as: ; ; Therefore in When the minimum value is obtained, the row sparse matrix is The column vectors of this row of sparse matrix. On Each element corresponds to Each grid point, with non-zero elements corresponding to the target grid points where a real information source exists, directly reflects the location of the real information source, thus avoiding secondary error interference caused by secondary positioning.

[0044] In the second specific implementation, the complex-valued conjugate gradient descent algorithm is used to solve the localization problem. The complex-valued conjugate gradient descent algorithm includes adjustments to the variables... and The update steps, namely, for those based on 2,1 Group sparse optimization problem of norm: ; The objective function gradient can be solved using the conjugate gradient descent method, and its gradient is: ;

[0045] ;

[0046] ;

[0047] ; in, For each row of the matrix, .

[0048] The specific process of solving the localization problem using the complex-valued conjugate gradient descent algorithm is as follows: Step 1: Initialization ; ; .

[0049] Step 2, Iteration , .

[0050] Step 3: Use a linear backtracking method to determine the optimal value for each iteration. Set the convergence factor , Initial step size Check in this iteration Are the conditions met? ; If the condition is met, exit the loop; If the condition is not met, then reduce the convergence step size: Repeat the loop until the condition is met to obtain the optimal step size for this iteration.

[0051] Step 4: Set the maximum number of iterations. When the conditions are met or Stop the iteration if the condition is met; otherwise, proceed to step five.

[0052] Step 5 Then return to step two.

[0053] In the third specific implementation, the localization problem is solved using an iterative reweighted least squares algorithm, wherein the iterative reweighted least squares algorithm includes adjustments to the variables. and The update steps are as follows. In this embodiment, a block continuous upper bound minimization method is used to construct an off-grid iterative reweighting framework.

[0054] Minimize based on block continuum upper bound The update formula is as follows: ; in, Indicates the optimization variable corresponding to the first iteration Indicates the optimization variable corresponding to the first iteration Represents the cost function exist The upper bound function at point A satisfies: ; ; in, This represents the cost function of the problem being addressed.

[0055] Next, it is updated adaptively using the following formula: ; in, It includes both signal-to-noise ratio information and source number information, and can achieve a stable balance between the likelihood term and the prior term. This characteristic can be observed from the... Further verification was performed during the derivation process. New This only corresponds to the initial stage of a new round of iterations that minimize the continuous upper bound of the block. Through alternating iterations... and , The change is extremely small, or satisfies ( The iteration program can terminate when the maximum number of iterations is reached.

[0056] for The iterative process, , The upper bound function of can be given by the following inequality, namely: ; in, The original cost function at the t-th iteration The upper bound function is:

[0057] ; Its upper bound function is also: ; ; ; Let be the row number of the matrix. .

[0058] At this time based on 2,1 The group sparse optimization problem of norm can be categorized as follows:

[0059] ; The essence of this optimization problem is a standard iterative reweighting problem. and Decreasing the value will affect the matrix in the next iteration. The The row gains greater weight, thereby forcing and Approaching 0, on the contrary and The The number of rows has increased.

[0060] After multiple iterations It will become a row sparse matrix. Able to reduce the signal-to-noise ratio for the number of signal sources This improves the sensitivity of the signal-to-noise ratio and the number of signal sources, thus enabling greater stability in scenarios with different signal-to-noise ratios and number of sources.

[0061] Specifically, The update formula is as follows:

[0062] ; in, It is derived using the matrix inversion lemma: ; because and The computational complexities are respectively and Therefore, update The computational complexity is approximately .

[0063] The specific process of solving the localization problem using the iterative reweighted least squares algorithm is as follows: Step 1: Initialization ; ; ; Step 2: Update .

[0064] Step 3: According to the formula renew .

[0065] Step 4, the maximum number of iterations is: When the conditions are met or Stop the iteration if the condition is met; otherwise, proceed to step five.

[0066] Step 5 Then return to step two.

[0067] As can be seen, in this embodiment of the invention, primary localization is achieved by jointly utilizing the echo signal data collected by all observation arrays. This involves stacking and optimizing the observation matrices of all observation arrays for each grid point, thereby achieving primary localization of the true signal source. This eliminates the need for secondary localization, removes secondary localization errors, and reduces abnormal noise interference. It can better address the presence of abnormal noise in the detection area and eliminate the secondary noise error interference introduced by secondary localization. Even when some array radar nodes have abnormal sampling data and high-precision time or phase synchronization is not required, high positioning accuracy and robustness are still guaranteed, achieving robust multi-target localization.

[0068] It should be noted that the distributed radar incoherent direct positioning technology solution in this application refers to the technology that uses received signals to determine the location of a target object. Data from one array radar ,in To estimate Coordinates of a real source ,in In this process, noise and other unknown variables are stacked and the unique row sparsity constraint of the 2,1 norm is utilized. The dual characteristics of column stability enable it to eliminate rows corresponding to outliers through sparsity constraints, thereby avoiding interference from extreme outliers. In general noisy environments, it can improve positioning accuracy. Furthermore, as the proportion of outliers in the noise increases, the positioning scheme based on the 2,1 norm array sparsity framework of this application utilizes the idea of ​​array sparsity and simultaneously uses all echo signal data collected by the distributed radar system to achieve direction of arrival estimation. This achieves single-stage positioning while maintaining better positioning accuracy, overcoming the secondary error problem and outlier noise problem caused by secondary positioning. Therefore, even when the sampling data of some array radar nodes is abnormal and high-precision time or phase synchronization is not required for distributed radar systems, it can still guarantee high positioning accuracy and robustness, achieving robust multi-target positioning. It is suitable for distributed detection and monitoring applications with high requirements for reliability and environmental adaptability.

[0069] In simulation testing, the Cramer-Rao lower bound can be derived as an important comparison parameter. Specifically, for near-field sources, the Cramer-Rao lower bound is obtained by extending the direction of arrival of the interfered source signal to the distance difference between the measured target and the real target, rather than being derived from the original data samples. The process of deriving the Cramer-Rao lower bound for near-field sources is as follows: when yes When considering the probability density function, this probability density function depends on a vector containing real-number unknown parameters. Under typical Gaussian white noise conditions, the observation array receives data with unknown parameters. The deterministic signal is: ; The signal pairs are clearly marked. Dependence.

[0070] Furthermore, the likelihood function is: ; Given that the source signal is incoherent, each snapshot is independent, and the likelihood function is the product of all snapshots. For The Fischer information matrix can be: ; Due to the rapid capture of the source signal, the noise and each observation array (i.e., the observation UAV) are independently and identically distributed. Therefore, the Fischer matrix of the array is: .

[0071] In the near-field case, the location of the source can be represented as... ( The unknown parameter vector is: Extending the Cramer-Rao lower bound of angles to the Cramer-Rao lower bound of positions, we know that: ; ; ; ; ; The expansion of the Fischer matrix is: ; Angular dimension Extending to the location distance dimension: ; We can obtain: ; From the main diagonal of the Fischer matrix, we can obtain: The Cramer-Rhodes lower bound of the distance error in each direction is used to obtain the lower bound of the error between the simulated position and the actual position.

[0072] As can be seen, the distributed radar incoherent direct localization method proposed in this application outperforms existing general group sparse localization algorithms under various noise conditions, including general Gaussian noise, mixed Gaussian noise, and Laplace heavy tail noise. Especially in scenarios with a high proportion of anomalous noise, the localization scheme of this application demonstrates greater stability and effectiveness, and significantly suppresses outlier noise in the environment.

[0073] Furthermore, the CVX algorithm (i.e., convex optimization solver) under the cost function, the complex-valued conjugate gradient descent algorithm, the iterative reweighting algorithm, and the general group sparse localization algorithm are used to solve the optimization problem and locate the target position of the true signal source. During this process, all potential signal source positions are traversed to find the target position of the true signal source. The localization errors of the four algorithms are compared. To compare the four different estimation methods without loss of generality, 100 Monte Carlo experiments are conducted under both fixed signal-to-noise ratios and different frequency snapshots, and under the same frequency snapshots and different signal-to-noise ratios. The mean square error (RMSE) is used for evaluation; a smaller RMSE indicates higher localization accuracy, and vice versa. The mean square error is: ; in, For the number of experiments, The number of real information sources. To locate the target and the true source target in Error in direction, To locate the target and the true source target in Error in direction.

[0074] In the simulation test, the distributed radar positioning system includes three observation arrays. See above Figure 2 As shown, these observation arrays are mounted on corresponding observation drones, and each observation drone is located at... , and Set up two signal sources, one of which is located at... The second is The coordinate units are all meters, and none of them are on the grid points.

[0075] In the pre-defined Cartesian coordinate system, the azimuth angle of the received signals from each observation array is... Take measurements: For the first signal source, the azimuth angles corresponding to the three subarrays are 111.83°, 179.92° and -48.81° respectively; For the second signal source, the azimuth angles corresponding to the three subarrays are 108.37°, -146.19° and -51.33°, respectively.

[0076] For the same signal source, the azimuth angles observed by any two different observation arrays are different, and the difference is greater than 50°. Based on this criterion, the two signal sources mentioned above can be classified as near-field signal sources. , Axis coverage Within a square region of interest, the location of the signal source is detected. With this as the core, the search scope is limited to axis to , axis to A square area (e.g., with a side length of 2 meters) is defined, and the search step size is simultaneously adjusted to 0.1 meters. At the remaining points, a grid is constructed with an accuracy of 1 meter.

[0077] First, the input signal-to-noise ratio is set to 0dB, and the number of snapshots is fixed at 100. Under the aforementioned cost function, the localization problem is solved using a convex optimization solver, complex-valued conjugate gradient descent algorithm, or iterative reweighted least squares algorithm to obtain the near-field source localization results. See [link to relevant documentation] Figure 3 , Figure 4 and Figure 5The figures show the localization results of the convex optimization solver, the complex-valued conjugate gradient descent algorithm, and the iterative reweighted least squares algorithm under a specified mesh refinement. The two prominent peaks in the figures correspond to the locations of the two detected signal sources. The localization results show that the signal source locations detected by the proposed scheme have minimal deviation from the actual coordinates of the signal sources, indicating that the convex optimization solver, the complex-valued conjugate gradient descent algorithm, or the iterative reweighted least squares algorithm are highly effective in near-field source localization.

[0078] Under typical Gaussian noise conditions, the source signal snapshot was fixed at 100, and the input signal-to-noise ratio (SNR) was varied. The input SNR was also fixed at 15 dB, and the source signal snapshot was varied. (See...) Figure 6 and Figure 7 As shown, the root mean square error (RMSE) and the derived Cramer-Rao lower bound are presented under two conditions. The results clearly show that increasing the input signal-to-noise ratio (SNR) or the number of snapshots in the source signal both improve positioning accuracy. Furthermore, the positioning scheme proposed in this application, which uses a convex optimization solver, complex-valued conjugate gradient descent algorithm, or iterative reweighted least squares algorithm to solve the optimization problem, has a smaller RMS error and is closer to the corresponding Cramer-Rao lower bound compared to general sparse positioning algorithms, resulting in more accurate positioning results.

[0079] Furthermore, to quantify the impact of outliers mixed with Gaussian noise, the positioning accuracy was tested under five gradient conditions with outlier noise values ​​of 0.1, 0.3, 0.5, 0.7, and 0.9. In the positioning scheme using the CVX algorithm to solve the optimization problem, the root mean square error of the positioning location was analyzed under different outlier probabilities (c), showing the trend of changes in the actual signal-to-noise ratio and the proportion of outliers. Specifically, as the probability of outliers increases, the root mean square error of the actual positioning target increases, and the positioning accuracy decreases.

[0080] In the proposed technical solution, the location of the signal source is calculated jointly using the 2,1 norm of data collected from all observation arrays; therefore, its performance consistently outperforms general group sparse localization algorithms. See [link to relevant documentation]. Figure 8 and Figure 9As shown, the localization accuracy of the localization schemes using convex optimization solvers, complex-valued conjugate gradient descent algorithms, or iterative reweighted least squares algorithms to solve the optimization problem was compared with that of the general group sparse localization algorithm when the outlier probabilities were 0.1 and 0.3, respectively, as a function of signal-to-noise ratio (SNR) and frequency snapshot. Similarly, experiments were conducted with a fixed frequency snapshot of 100, varying the SNR, and with a fixed SNR of 15 dB, varying the frequency snapshot. Notably, under low input SNR conditions, the localization error of the general group sparse localization algorithm increased significantly when the proportion of noise outliers was large, and it lost its statistical characteristics. The general group sparse localization algorithm is easily affected by the outliers involved in the noise fusion process. Therefore, the root mean square error (RMSE) curves for outliers of 0.5, 0.7, and 0.9 are not presented. As shown in the figure, the localization scheme of this application, which uses a convex optimization solver, complex-valued conjugate gradient descent algorithm, or iterative reweighted least squares algorithm to solve the optimization problem, shows better localization accuracy than the general group sparse localization algorithm in both cases, and is closer to the system Cramer-Rao lower bound.

[0081] In the case of Laplace noise, see Figure 10 and Figure 11 The paper demonstrates the changes in localization accuracy under various conditions: a fixed frequency snapshot of 100, varying the signal-to-noise ratio (SNR) compared to a fixed SNR of 15 dB, and varying the frequency snapshot. The results show the effects of different localization schemes using a convex optimization solver, complex-valued conjugate gradient descent algorithm, or iterative reweighted least squares algorithm, as well as a general grouped sparse localization algorithm. Similarly, the localization scheme using a convex optimization solver, complex-valued conjugate gradient descent algorithm, or iterative reweighted least squares algorithm exhibits better localization accuracy than the general grouped sparse localization algorithm in both cases, and is closer to the system's Cramer-Rao lower bound.

[0082] In the single localization process, the complex conjugate gradient descent algorithm takes the longest time, the iterative reweighted least squares algorithm is in the middle, and the CVX algorithm is the fastest. The localization accuracy of the three algorithms in different environments is better than the general group sparse localization algorithm, and the iterative reweighted least squares algorithm has the highest accuracy, the CVX algorithm is in the middle, and the complex conjugate gradient descent algorithm is the worst.

[0083] In summary, outlier noise can degrade the overall performance of general sparse localization algorithms. However, the localization scheme proposed in this application, which uses a convex optimization solver, complex-valued conjugate gradient descent algorithm, or iterative reweighted least squares algorithm to solve the optimization problem, can achieve good estimation results in the same scenario.

[0084] In one embodiment, such as Figure 12 As shown, based on the above-mentioned distributed radar incoherent direct localization method for data anomaly scenarios, the present invention also provides a distributed radar incoherent direct localization device for data anomaly scenarios, comprising: Data acquisition module 11 is used to acquire echo signal data received by each observation array in the distributed radar system; The signal model construction module 12 is used to construct a signal model that characterizes the mapping relationship between the echo signal data and the potential source location; wherein, the common detection area of ​​each observation array is a two-dimensional grid, the coordinates of each grid point in the two-dimensional grid correspond to a potential source location, and the signal model does not include the phase synchronization parameters between the observation arrays, so as to adapt to the incoherent characteristics of the echo signal data; Location problem construction module 13 is used to construct a location problem based on the signal model; the location problem is based on 2,1 The group sparsity optimization problem of norms; The localization problem solving module 14 is used to solve the localization problem using a preset optimization algorithm to obtain the corresponding row sparse matrix; The source localization module 15 is used to determine the localization result of the real source based on the row sparse matrix; each element of the column vector of the row sparse matrix corresponds to a grid point and has the reflection energy estimation result of a potential source.

[0085] Furthermore, it is worth noting that the working process of the distributed radar noncoherent direct positioning device for data anomaly scenarios provided in this embodiment is the same as the working process of the distributed radar noncoherent direct positioning method for data anomaly scenarios described above. Therefore, it will not be repeated here. For details, please refer to the working process of the distributed radar noncoherent direct positioning method for data anomaly scenarios described above.

[0086] Figure 13 A schematic diagram of the structure of a terminal provided in an embodiment of this application. The terminal may include: The memory 501, the processor 502, and the computer program stored on the memory 501 and capable of running on the processor 502.

[0087] When the processor 502 executes the program, it implements the distributed radar noncoherent direct localization method for data anomaly scenarios provided in the above embodiments.

[0088] Furthermore, the terminal also includes: Communication interface 503 is used for communication between memory 501 and processor 502.

[0089] The memory 501 is used to store computer programs that can run on the processor 502.

[0090] Memory 501 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0091] If the memory 501, processor 502, and communication interface 503 are implemented independently, they can be interconnected via a bus to communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, only one line is used in the diagram, but this does not imply that there is only one bus or one type of bus.

[0092] Optionally, in a specific implementation, if the memory 501, processor 502, and communication interface 503 are integrated on a single chip, then the memory 501, processor 502, and communication interface 503 can communicate with each other through an internal interface.

[0093] Processor 502 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0094] This embodiment also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described distributed radar incoherent direct localization method for data anomaly scenarios.

[0095] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein.

[0096] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0097] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus or device (such as a computer-based system, a processor-included system or other system that can read and execute instructions from and from an instruction execution system, apparatus or device).

[0098] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0099] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A distributed radar incoherent direct localization method for data anomaly scenarios, characterized in that, The method includes: Acquire echo signal data received by each observation array in a distributed radar system; A signal model is constructed to characterize the mapping relationship between the echo signal data and the potential source location; wherein, the common detection area of ​​each observation array is a two-dimensional grid, the coordinates of each grid point in the two-dimensional grid correspond to a potential source location, and the signal model does not include the phase synchronization parameters between the observation arrays, so as to adapt to the incoherent characteristics of the echo signal data; A localization problem is constructed based on the aforementioned signal model; the localization problem is based on... 2,1 The group sparsity optimization problem of norms; The localization problem is solved using a preset optimization algorithm to obtain a corresponding row sparse matrix, and the localization result of the real source is determined based on the row sparse matrix; each element on the column vector of the row sparse matrix corresponds to a grid point and the reflection energy estimation result of the potential source exists.

2. The distributed radar incoherent direct localization method for data anomaly scenarios according to claim 1, characterized in that, The construction of the signal model characterizing the mapping relationship between the echo signal data and the potential source location includes: Based on the known coordinates of each observation array and the coordinates of the grid points in the two-dimensional grid, calculate the coordinate difference between the known coordinates of each observation array and the coordinates of the grid points; The azimuth angle between each observation array and the potential source is determined based on the coordinate difference, and a steering vector for calculating the direction of each potential source is constructed based on the azimuth angle. A steering matrix is ​​constructed using the steering vectors of each potential source direction, and a signal model is constructed based on the steering matrix.

3. The distributed radar incoherent direct localization method for data anomaly scenarios according to claim 1, characterized in that, The signal model is as follows: ； ； ； ； ； ； ； ； ； ； in, Indicates the first The first observation array received the first Snap array signal, For source signals, For noise vectors, for 3D guided matrix, The number of array elements on each observation array. The number of grid points, The number of observation arrays, For the first The first potential source and the first Azimuth angles between observation arrays For the first The observation array is for the first The steering vector of a potential information source. At the speed of light, For the first The first potential source and the first One observation array in Coordinate difference along the axis, For the first The first potential source and the first One observation array in Coordinate difference along the axis, For the first One potential source location, For the first The known coordinates of an observation array.

4. The distributed radar incoherent direct localization method for data anomaly scenarios according to claim 3, characterized in that, The basis 2,1 The group sparse optimization problem of norm is: ； ； ； ； ； ； ； ； in, This is the estimated reflection energy when the real source exists at the g-th grid point on the m-th observation array. To estimate the unknown reflection energy variable when the real source exists at the g-th grid point on the m-th observation array, Let covariance matrix be the variance matrix. For the first The grid point and the first Azimuth angles between observation arrays For including all azimuth angles 3D matrix For along Number of grid cells along the axis, For along Number of grid cells along the axis, For mesh stacking 3D guided matrix, For the first The coordinates of each grid point For the first The known coordinates of the observation array For the first The coordinates of the first grid point and the first One observation array in Coordinate difference along the axis, For the first The coordinates of the first grid point and the first One observation array in Coordinate difference along the axis, For the first The constraint matrix of the observation array, The result is the constraint matrices of all observation arrays stacked column-wise. It is a convex function. This is the regularization parameter.

5. The distributed radar incoherent direct localization method for data anomaly scenarios according to claim 4, characterized in that, The row sparse matrix is: 。 6. The distributed radar incoherent direct localization method for data anomaly scenarios according to any one of claims 1 to 5, characterized in that, The step of solving the localization problem using a preset optimization algorithm includes: The localization problem is solved using a convex optimization solver, a complex-valued conjugate gradient descent algorithm, or an iterative reweighted least squares algorithm.

7. The distributed radar incoherent direct localization method for data anomaly scenarios according to claim 6, characterized in that, The location result of determining the true information source based on the row sparse matrix includes: Determine the non-zero elements in the row sparse matrix, and identify the grid points corresponding to the non-zero elements as target grid points where a real information source exists; The coordinate information of the target grid point in the preset rectangular coordinate system is determined to obtain the positioning result of the real information source.

8. A distributed radar incoherent direct positioning device for data anomaly scenarios, characterized in that, The device includes: The data acquisition module is used to acquire echo signal data received by each observation array in the distributed radar system; The signal model construction module is used to construct a signal model that characterizes the mapping relationship between the echo signal data and the potential source location; wherein, the common detection area of ​​each observation array is a two-dimensional grid, the coordinates of each grid point in the two-dimensional grid correspond to a potential source location, and the signal model does not include phase synchronization parameters between the observation arrays, so as to adapt to the incoherent characteristics of the echo signal data; The localization problem construction module is used to construct a localization problem based on the signal model; the localization problem is based on... 2,1 The group sparsity optimization problem of norms; The localization problem solving module is used to solve the localization problem using a preset optimization algorithm to obtain the corresponding row sparse matrix; The source localization module is used to determine the localization result of the real source based on the row sparse matrix; each element of the column vector of the row sparse matrix corresponds to a grid point and has the reflection energy estimation result of a potential source.

9. A terminal, characterized in that, include: The system includes a memory, a processor, and a distributed radar incoherent direct localization program for data anomaly scenarios stored in the memory and executable on the processor. When executed by the processor, the distributed radar incoherent direct localization program for data anomaly scenarios implements the steps of the distributed radar incoherent direct localization method for data anomaly scenarios as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that can be executed to implement the steps of the distributed radar incoherent direct localization method for data anomaly scenarios as described in any one of claims 1 to 7.

Images