Robust adaptive beam forming method based on null broadening and main lobe interference detection and suppression

Abstract

The invention discloses a robust adaptive beamforming method based on null broadening and main lobe interference detection and elimination, which uses time decay covariance matrix sampling to improve the real-time performance of an algorithm, reconstructs a covariance matrix to separate a desired signal and interference, uses virtual interference rotation broadening null, and uses the main lobe interference detection and elimination to detect and eliminate the main lobe interference. And main lobe interference detection based on an improved multiple signal classification algorithm is used, and interference in a main lobe is eliminated. According to the beam forming method, the output signal to interference plus noise ratio in a high dynamic environment is effectively improved, and the system performance is improved. Compared with an existing classical beam forming algorithm in a high dynamic environment, the scheme provided by the invention has obvious performance gain.

Description

Technical Field

[0001] The present invention relates to the field of beamforming technology, and more particularly to a robust adaptive beamforming method based on null broadening and mainlobe interference detection and suppression. Background Art

[0002] Beamforming technology has been widely used in systems such as communications, radar, and sonar to enhance the desired signal and suppress interference and noise in the spatial domain. By spatially directing the main lobe toward the desired signal and placing nulls in the direction of interference, beamforming ensures high-performance signal reception. However, in practical applications, traditional beamforming methods rely on precise assumptions about signal and interference characteristics, and factors such as steering vector errors and covariance matrix inaccuracies often degrade their performance. To address these challenges, robust adaptive beamforming techniques have been developed to improve system performance.

[0003] In highly dynamic environments, including satellites, drones, high-speed rail, and autonomous vehicles, beamforming robustness faces additional challenges that are even more formidable. In these environments, interference can move rapidly and unpredictably, often drifting out of the null or into the mainlobe where the desired signal resides. This can happen, for example, when drones in a swarm communicate with a base station, or when mobile users interfere with an IoT network with fixed devices. This misalignment reduces the beamformer's ability to distinguish interference from the desired signal, resulting in a degradation of the signal-to-interference-and-noise ratio (SINR). Furthermore, the constant variability of interference demands that the system adapt quickly to various conditions, something that many classic robustness methods, such as diagonal loading techniques, subspace-based techniques, and covariance matrix reconstruction, are unable to effectively handle.

[0004] To address the challenges in this scenario, many algorithms have improved classical algorithms based on null stretching. Although the covariance matrix tapering algorithm and its variants enhance interference suppression by stretching the nulls, they essentially trade null depth for null width, which greatly weakens the suppression effect. Other null stretching methods include derivative constraint methods, which flatten the beam pattern by setting high-order derivatives to zero at specific angles. However, constrained methods have the problems of increased computational requirements, limited flexibility in controlling null width, and reduced scalability. In addition, these null stretching methods usually assume a static or quasi-static interference environment, which limits their adaptability to rapid and unpredictable changes in highly dynamic scenarios. Summary of the Invention

[0005] In view of the above problems, the present invention provides a robust adaptive beamforming method based on nulling and widening and mainlobe interference detection and suppression, aiming to solve the problem of rapid interference movement in existing high-dynamic environments. This method uses time-decayed covariance matrix sampling to improve the real-time performance of the algorithm, reconstructs the covariance matrix to separate the desired signal and interference, uses virtual interference rotation to widen the nulling, and uses mainlobe interference detection and suppression based on an improved multiple signal classification (MUSIC) algorithm. This beamforming method effectively improves the output signal-to-interference-noise ratio in high-dynamic environments and enhances system performance.

[0006] To achieve the above objectives, the technical solutions of the present invention are as follows:

[0007] A robust adaptive beamforming method based on null broadening and mainlobe interference detection and suppression includes the following steps:

[0008] S1: input array antenna receiving signal;

[0009] S2: Calculate the array signal sampling covariance matrix with time attenuation

[0010] S3: Calculate the reconstruction matrix of the sampling covariance matrix;

[0011] S4: performing a virtual vector rotation operation on the reconstructed covariance matrix and calculating a beamforming weight coefficient vector after the virtual vector rotation;

[0012] S5: Use the improved MUSIC algorithm to detect and suppress mainlobe interference and update the beamforming weight coefficient vector;

[0013] S6: Output the calculation result of the weight coefficient vector.

[0014] Preferably, the array antenna receiving signal in step S1 is specifically:

[0015] The signal received by the array as a whole can be expressed as

[0016] x=[x1,x2,…,x M ] T

[0017] Where x m Is the signal received by the mth array element. To facilitate subsequent analysis and explanation, from the perspective of signal synthesis, the overall array received signal x can be expressed as

[0018]

[0019] Where x0, x i and x n are the expected signal, interference and zero-mean Gaussian white noise vectors of size M×1 respectively; s0 and si are the complex envelopes corresponding to the desired signal and interference respectively; is the spatial guidance vector, where θ i is the incident angle of the corresponding signal, and λ is the wavelength of the incident signal.

[0020] The adaptive beamformer output data is

[0021] y=w H x

[0022] Where w=[w1,w2,…,w M ] T is the adaptive array weight vector derived by the beamformer.

[0023] Preferably, the time-attenuated array signal sampling covariance matrix in step S2 is specifically:

[0024] The sampling covariance matrix of the time-decayed array signal can be expressed as

[0025]

[0026] Where α∈(1,+∞) represents the time decay factor, that is, under the influence of the exponential factor, older data has a smaller weight, while newer data has a larger weight; is the inverse of the sum of the weights.

[0027] Preferably, the reconstruction matrix of the sampling covariance matrix in step S3 is specifically:

[0028] The reconstruction matrix of the time-decayed array signal sampling covariance matrix can be expressed as

[0029]

[0030] Where Θ represents the approximate estimated range of the direction angle of the desired signal, and represents the complement of Θ in the entire spatial domain.

[0031] Preferably, the virtual vector rotation in step S4 is specifically as follows:

[0032] S4.1: Calculate the covariance matrix of the virtual vector after rotation;

[0033] S4.2: Calculate the beamforming weight coefficient vector after the virtual vector is rotated.

[0034] Preferably, in step S4.1, the covariance matrix after the virtual vector rotation of the reconstructed matrix can be expressed as

[0035]

[0036] Where B = diag{1, exp(j2πd / λsinΔθ),…, exp(j(M-1)2πd / λsinΔθ)} is the reconstruction matrix The rotation matrix can achieve the effect of null-stretching.

[0037] Preferably, in step S4.2, the beamforming weight coefficient vector after the virtual vector is rotated can be expressed as

[0038]

[0039] Preferably, the detection and suppression of main lobe interference in step S5 is specifically as follows:

[0040] S5.1: Mainlobe interference detection using the improved MUSIC algorithm;

[0041] S5.2: Main lobe interference suppression and beamforming weight coefficient vector update are performed according to the detection results.

[0042] Preferably, the main lobe interference detection in step S5.1 is specifically as follows:

[0043] Select K / 2 snapshots from the collected K snapshots, and record the selected snapshot data set as k1, and the remaining snapshot data set as κ2 (assuming K is an even number). Then use the data of κ1 and κ2 to calculate two sampling covariance matrices, which are recorded as and The calculation process is:

[0044]

[0045] By canceling the two matrices, we can get the cancellation matrix that eliminates the desired signal in a fixed direction.

[0046]

[0047] Cancellation Matrix Perform singular value decomposition (SVD) decomposition:

[0048]

[0049] Where U and V are the left and right singular matrices respectively; ∑=diag{λ1,λ2,…,λ M} is a diagonal matrix consisting of singular values, and λ1≥λ2≥…≥λ M , decompose U into mutually orthogonal interference subspace and noise subspace [U N ,U N ], apply the MUSIC algorithm to obtain the spatial power spectrum

[0050]

[0051] In θ∈[θ0-θ w ,θ0+θ w ] range to search for the angle that maximizes the MUSIC power spectrum, which is used as the direction of the main lobe interference. The search result of the main lobe interference direction is

[0052]

[0053] Set a threshold θ th ,|θ ml -θ0|≤θ th When the main lobe interference is absent,

[0054] Preferably, the main lobe interference suppression in step S5.2 is specifically:

[0055] Defining the projection matrix

[0056] B mlb =Ia(θ ml )(a(θ ml ) H a(θ ml )) -1 a(θ ml ) H

[0057] According to the search results of the main lobe interference direction: if there is no main lobe interference, the beamforming weight coefficient vector is not updated, that is,

[0058] w proposed =w VIR

[0059] If there is main lobe interference, the beamforming weight coefficient vector is updated, that is,

[0060]

[0061] Preferably, the weight coefficient vector calculation result is output in step S6, specifically: w proposed as the final output result. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 Schematic diagram of the method of the present invention.

[0063] Figure 2 This is a schematic diagram showing a performance comparison of the output signal-to-interference-plus-noise ratio (SINR) of different algorithms provided in the embodiment when there is an error in the signal direction as a function of the input SNR.

[0064] Figure 3 This is a schematic diagram showing a performance comparison of different algorithms provided in the embodiments, showing how the output signal-to-interference-and-noise ratio varies with the number of snapshots when there is an error in the signal direction.

[0065] Figure 4 This is a schematic diagram showing the performance comparison of the output signal-to-interference-plus-noise ratio (SINR) of different algorithms provided in the embodiment when there is an error in the steering vector as a function of the input SNR.

[0066] Figure 5 This is a schematic diagram showing a performance comparison of different algorithms provided in the embodiments, where the output signal-to-interference-and-noise ratio varies with the number of snapshots when there is an error in the steering vector. DETAILED DESCRIPTION

[0067] The accompanying drawings are for illustrative purposes only and are not to be construed as limiting this patent;

[0068] In order to better illustrate this embodiment, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product size;

[0069] It is understandable to those skilled in the art that some well-known structures and descriptions thereof may be omitted in the drawings.

[0070] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0071] Example 1

[0072] This example provides a robust adaptive beamforming method based on null broadening and mainlobe interference detection and suppression. Figure 1 As shown, the following steps are included:

[0073] S1: input array antenna receiving signal;

[0074] S2: Calculate the array signal sampling covariance matrix with time attenuation

[0075] S3: Calculate the reconstruction matrix of the sampling covariance matrix;

[0076] S4: performing a virtual vector rotation operation on the reconstructed covariance matrix and calculating a beamforming weight coefficient vector after the virtual vector rotation;

[0077] S5: Use the improved MUSIC algorithm to detect and suppress mainlobe interference and update the beamforming weight coefficient vector;

[0078] S6: Output weight coefficient vector calculation result;

[0079] Example 2

[0080] This embodiment, based on the first embodiment, further discloses the following contents:

[0081] The array antenna receives the signal in step S1, specifically:

[0082] The signal received by the array as a whole can be expressed as

[0083] x=[x1,x2,…,x M ] T

[0084] Where x m Is the signal received by the mth array element. To facilitate subsequent analysis and explanation, from the perspective of signal synthesis, the overall array received signal x can be expressed as

[0085]

[0086] Where x0, x i and x n are the expected signal, interference and zero-mean Gaussian white noise vectors of size M×1 respectively; s0 and s i are the complex envelopes corresponding to the desired signal and interference respectively; is the spatial guidance vector, where θ i is the angle of the corresponding signal, and λ is the wavelength of the incident signal. The power of the desired signal, interference, and noise are respectively and

[0087] The adaptive beamformer output data is

[0088] y=w H x

[0089] Where w=[w1,w2,…,w M ] T is the adaptive array weight vector derived by the beamformer.

[0090] The sampling covariance matrix of the time-attenuated array signal in step S2 is specifically:

[0091] The sampling covariance matrix of the time-decayed array signal can be expressed as

[0092]

[0093] Where α∈(1,+∞) represents the time decay factor, that is, under the influence of the exponential factor, older data has a smaller weight, while newer data has a larger weight; is the inverse of the sum of the weights.

[0094] The reconstruction matrix of the sampling covariance matrix in step S3 is specifically:

[0095] The reconstruction matrix of the time-decayed array signal sampling covariance matrix can be expressed as

[0096]

[0097] Where Θ represents the approximate estimated range of the direction angle of the desired signal, and represents the complement of Θ in the entire spatial domain.

[0098] The virtual vector rotation in step S4 is specifically as follows:

[0099] S4.1: Calculate the covariance matrix of the virtual vector after rotation;

[0100] S4.2: Calculate the beamforming weight coefficient vector after the virtual vector is rotated;

[0101] In step S4.1, the covariance matrix after the virtual vector rotation of the reconstruction matrix can be expressed as

[0102]

[0103] Where B = diag{1, exp(j2πd / λsinΔθ),…, exp(j(M-1)2πd / λsinΔθ)} is the reconstruction matrix The rotation matrix can achieve the effect of null-stretching.

[0104] In step S4.2, the beamforming weight coefficient vector after the virtual vector rotation can be expressed as

[0105]

[0106] Step S5 of detecting and suppressing main lobe interference is specifically as follows:

[0107] S5.1: Mainlobe interference detection using the improved MUSIC algorithm;

[0108] S5.2: Main lobe interference suppression and beamforming weight coefficient vector update based on the detection results;

[0109] The main lobe interference detection in step S5.1 is specifically as follows:

[0110] Select K / 2 snapshots from the collected K snapshots, and record the selected snapshot data set as κ1, and the remaining snapshot data set as κ2 (assuming K is an even number). Then use the data of κ1 and κ2 to calculate two sampling covariance matrices, which are recorded as and The calculation process is:

[0111]

[0112] By canceling the two matrices, we can get the cancellation matrix that eliminates the desired signal in a fixed direction.

[0113]

[0114] Cancellation Matrix Perform singular value decomposition (SVD) decomposition:

[0115]

[0116] Where U and V are the left and right singular matrices respectively; ∑=diag{λ1,λ2,…,λ M} is a diagonal matrix consisting of singular values, and λ1≥λ2≥…≥λ M , decompose U into mutually orthogonal interference subspace and noise subspace [U N ,U N ], apply the MUSIC algorithm to obtain the spatial power spectrum

[0117]

[0118] Set the main lobe width to In θ∈[θ0-θ w ,θ0+θ w ] range to search for the angle that maximizes the MUSIC power spectrum, which is used as the direction of the main lobe interference. The search result of the main lobe interference direction is

[0119]

[0120] Set a threshold θ th ,|θ ml -θ0|≤θ th When the main lobe interference is absent,

[0121] The main lobe interference suppression in step S5.2 is specifically as follows:

[0122] Defining the projection matrix

[0123] B mlb =Ia(θ ml )(a(θ ml ) H a(θ ml )) -1 a(θ ml ) H

[0124] According to the search results of the main lobe interference direction: if there is no main lobe interference, the beamforming weight coefficient vector is not updated, that is,

[0125] w proposed =w VIR

[0126] If there is main lobe interference, the beamforming weight coefficient vector is updated, that is,

[0127]

[0128] The calculation result of the weight coefficient vector is output in step S6, specifically: w proposed as the final output result.

[0129] Example 3

[0130] This embodiment provides the following specific embodiments based on Embodiment 1 and Embodiment 2:

[0131] The simulation parameters are set as follows: Assume that the array is a uniform linear array with 10 elements and an element spacing of half the wavelength of the incident signal. The incident desired signal and interference signal are randomly generated Gaussian distribution signals s i (t), where the signal-to-noise ratio is 30 dB, the incident angle of the desired signal is 0°, the incident angle of the interference signal is [-70°, -38°, 6°, 65°], the interference moving speed is 0.1° per snapshot, and the decision threshold θ th 0.1θ w , snapshot data set k1 = [-K, -K+1, …, -K / 2+1], snapshot data set k2 = [-K / 2, -K / 2+1, …, -1], and the number of Monte Carlo simulations is 1000.

[0132] The output performance of the proposed new method (hereinafter referred to as HDSBF) is compared with the classic beamforming algorithms INCMR, INCMR-CMT, F-CMT, and AN-CMT in high dynamic scenarios. The details are as follows: Figure 2 、 Figure 3 Test the algorithm performance with random errors in the signal direction, Figure 4 、 Figure 5 Test the algorithm performance with random errors in the steering vector.

[0133] like Figure 2 As shown in Figure 2, under the same conditions, that is, when the random errors of the desired signal and interference directions follow a uniform distribution of [-1°, 1°] and the number of snapshots K = 30 is kept constant, the performance of the above methods is compared as the output signal-to-noise ratio changes with the input signal-to-noise ratio. The output signal-to-noise ratio performance of the proposed method is better than that of the classic beamforming algorithm, with a performance improvement of about 0.5-2dB. Figure 3 As shown in the figure, under the same conditions, that is, when the random errors in the desired signal and interference directions follow a uniform distribution of [-1°, 1°] and the input signal-to-noise ratio is kept constant at 10dB, the performance of the above methods is compared as the output signal-to-interference-noise ratio changes with the number of snapshots. The output signal-to-interference-noise ratio performance of the proposed method is better than that of the classic beamforming algorithm, and the performance is improved by about 0.5-3dB at low snapshot numbers.

[0134] like Figure 4As shown in Figure 1, under the same conditions, that is, when the random error of the steering vector amplitude in the desired signal and interference directions follows a uniform distribution of [0, 0.1], and the random error of the phase follows a uniform distribution of [0, 2π], and the number of snapshots K = 30 remains unchanged, the performance of the above methods is compared with the change of the output signal-to-noise ratio with the input signal-to-noise ratio. The output signal-to-noise ratio performance of the proposed method is better than that of the classic beamforming algorithm, with a performance improvement of about 0.5-1dB. Figure 5 As shown in the figure, under the same conditions, that is, the random errors in the amplitude of the steering vectors in the desired signal and interference directions follow a uniform distribution in the range [0, 0.1], the random errors in the phase follow a uniform distribution in the range [0, 2π], and the input signal-to-noise ratio is kept constant at 10 dB, the performance of the above methods is compared as the output signal-to-interference-noise ratio changes with the number of snapshots. The output signal-to-interference-noise ratio performance of the proposed method is better than that of the classic beamforming algorithm, and the performance is improved by about 0.5-2.5 dB at low snapshot numbers.

[0135] The above examples show that the beamforming scheme proposed in the present invention can obtain processing gain brought by time attenuation, null stretching and mainlobe interference suppression in high dynamic environment scenarios, thereby bringing better output signal-to-interference-noise ratio performance.

[0136] The same or similar reference numerals correspond to the same or similar components;

[0137] The positional relationships described in the drawings are for illustrative purposes only and should not be construed as limiting this patent.

[0138] Obviously, the above examples of the present invention are merely illustrative examples for the purpose of clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the claims of the present invention.

Claims

1. An adaptive beamforming method based on null broadening and mainlobe interference detection and suppression, characterized in that: The following steps are involved: S1: input array antenna receiving signal; S2: Calculate the array signal sampling covariance matrix with time attenuation S3: Calculate the reconstruction matrix of the sampling covariance matrix; S4: performing a virtual vector rotation operation on the reconstructed covariance matrix and calculating a beamforming weight coefficient vector after the virtual vector rotation; S5: Use the improved MUSIC algorithm to detect and suppress mainlobe interference and update the beamforming weight coefficient vector; S6: Output the calculation result of the weight coefficient vector.

2. The adaptive beamforming method based on null broadening and mainlobe interference detection and suppression according to claim 1, characterized in that: In step S1, the signal received by the array as a whole is specifically: The signal received by the array as a whole can be expressed as x = [x1, x2, ..., x M ] T Where x m Is the signal received by the mth array element. To facilitate subsequent analysis and explanation, from the perspective of signal synthesis, the overall array received signal x can be expressed as Where x0, x i and x n are the expected signal, interference and zero-mean Gaussian white noise vectors of size M×1 respectively; s0 and s i are the complex envelopes corresponding to the desired signal and interference respectively; is the spatial guidance vector, where θ i is the incident angle of the corresponding signal, and λ is the wavelength of the incident signal. The adaptive beamformer output data is y=w H x Where w=[w1,w2,…,w M ] T is the adaptive array weight vector derived by the beamformer.

3. The adaptive beamforming method based on null broadening and mainlobe interference detection and suppression according to claim 2, characterized in that: The sampling covariance matrix of the time-attenuated array signal in step S2 is specifically: The sampling covariance matrix of the time-decayed array signal can be expressed as Where α∈(1,+∞) represents the time decay factor, that is, under the influence of the exponential factor, older data has a smaller weight, while newer data has a larger weight; is the inverse of the sum of the weights.

4. The adaptive beamforming method based on null stretching and mainlobe interference detection and suppression according to claim 3, wherein: The reconstruction matrix of the sampling covariance matrix of S3 is: The reconstruction matrix of the time-decayed array signal sampling covariance matrix can be expressed as Where Θ represents the approximate estimated range of the direction angle of the desired signal, and represents the complement of Θ in the entire spatial domain.

5. The adaptive beamforming method based on null broadening and mainlobe interference detection and suppression according to claim 4, characterized in that: In step S2.2, the power of each antenna signal is normalized, specifically: S4.1: Calculate the covariance matrix of the virtual vector after rotation; S4.2: Calculate the beamforming weight coefficient vector after the virtual vector is rotated.

6. The adaptive beamforming method based on null broadening and mainlobe interference detection and suppression according to claim 5, characterized in that: In step S4.1, the covariance matrix after the virtual vector rotation of the reconstructed matrix is ​​specifically: Where B = diag{1, exp(j2πd / λsinΔθ),…, exp(j(M-1)2πd / λsinΔθ)} is the reconstruction matrix The rotation matrix can achieve the effect of null-stretching. In step S4.2, the beamforming weight coefficient vector after the virtual vector is rotated is specifically:

7. The adaptive beamforming method based on null broadening and mainlobe interference detection and suppression according to claim 6, characterized in that: Step S5 of detecting and suppressing main lobe interference is specifically as follows: S5.1: Mainlobe interference detection using the improved MUSIC algorithm; S5.2: Main lobe interference suppression and beamforming weight coefficient vector update are performed according to the detection results.

8. The adaptive beamforming method based on null broadening and mainlobe interference detection and suppression according to claim 7, characterized in that: The main lobe interference detection in step S5.1 is specifically as follows: Select K / 2 snapshots from the K snapshots collected, and record the selected snapshot data set as κ1, and the remaining snapshot data set as κ2 (assuming K is an even number). Then use the data of k1 and k2 to calculate two sampling covariance matrices, which are recorded as and The calculation process is: By canceling the two matrices, we can get the cancellation matrix that eliminates the desired signal in a fixed direction. Cancellation Matrix Perform singular value decomposition (SVD) decomposition: Where U and V are the left and right singular matrices respectively; Σ=diag{λ1,λ2,…,λ M } is a diagonal matrix consisting of singular values, and λ1≥λ2≥…≥λ M , decompose U into mutually orthogonal interference subspace and noise subspace [U N ,U N ], apply the MUSIC algorithm to obtain the spatial power spectrum In θ∈[θ0-θ w ,θ0+θ w ] range to search for the angle that maximizes the MUSIC power spectrum, which is used as the direction of the main lobe interference. The search result of the main lobe interference direction is Set a threshold θ th ,|θ ml -θ0|≤θ th When the main lobe interference is absent, The main lobe interference suppression in step S5.2 is specifically as follows: Defining the projection matrix B mlb =Ia(θ ml )(a(θ ml ) H a(θ ml )) -1 a(θ ml ) H According to the search results of the main lobe interference direction: if there is no main lobe interference, the beamforming weight coefficient vector is not updated, that is, W proposed =W VIR If there is main lobe interference, the beamforming weight coefficient vector is updated, that is, 9. The adaptive beamforming method based on null broadening and mainlobe interference detection and suppression according to claim 8, characterized in that: The calculation result of the weight coefficient vector is output in step S6, specifically: Will w proposed as the final output result.

Images