Method for jointly estimating gain-phase error and direction of arrival (DOA) based on unmanned aerial vehicle (UAV) array

Abstract

A method for jointly estimating gain-phase error and direction of arrival (DOA) based on an unmanned aerial vehicle (UAV) array includes: equipping each UAV with an antenna, and forming a receive array through a swarm of multiple UAVs to receive source signals; when an observation baseline of the swarm remains unchanged, changing array manifold through movement of the UAVs, and re-sensing the source signals; for each sensed source signals, calculating a covariance matrix, and obtaining a corresponding noise subspace through eigenvalue decomposition; and constructing a quadratic optimization problem based on the noise subspace and array steering vector, constructing a cost function, and implementing joint estimation of the gain-phase error and the DOA through spectrum peak search. The method can jointly estimate the DOA and gain-phase error and calibrate the gain-phase error, thereby improving accuracy of passive positioning.

Description

CROSS REFERENCE TO THE RELATED APPLICATIONS

This application is based upon and claims priority to Chinese Patent Application No. 202111398473.8, filed on Nov. 24, 2021, the entire contents of which are incorporated herein by reference.

TECHNICAL FIELD

The present disclosure relates to the technical field of array signal processing, and specifically, to a method for jointly estimating a gain-phase error and a direction of arrival (DOA) based on an unmanned aerial vehicle (UAV) array.

BACKGROUND

As an important branch of modern signal processing, array signal processing has seen a rapid development in recent decades, and has been widely used in military and civilian fields such as radar, sonar, and wireless communications. DOA estimation is one of the key technologies of array signal processing, and is mainly used to estimate a position of a spatial source. The DOA estimation technology is developing rapidly, but related theories and technologies are still being improved.

Ubiquitous array errors make it difficult to apply a spatial spectrum estimation technology to practical projects. Normally, almost all DOA estimation algorithms are based on an accurately known array manifold. To obtain a good algorithm estimation effect, it is necessary to ensure that an actually used array is totally consistent with a standard array model in theoretical research. However, in practical applications, device or actual environment factors may cause errors in the array. When an ideal array manifold is used for estimating actual DOA, it is inevitable that an obtained estimation result has a relatively large error or the estimation result is totally invalid. Most of the array errors can ultimately be attributed to array gain-phase errors. Therefore, it is essential to study a DOA estimation algorithm in the presence of array gain-phase errors for application of the spatial spectrum estimation technology.

SUMMARY

In view of the defects in the prior art, the present disclosure provides a method for jointly estimating a gain-phase error and a DOA based on a UAV array, to implement position change of an array based on a UAV swarm, jointly estimate a DOA and a gain-phase error, and calibrate the gain-phase error, thereby improving accuracy of passive positioning.

To achieve the above objective, the present disclosure adopts the following technical solution:

A method for jointly estimating a gain-phase error and a DOA based on a UAV array includes the following steps:

S1: equipping each UAV with an antenna, and forming a receive array through a swarm of multiple UAVs to receive source signals;

S2: when an observation baseline of the swarm remains unchanged, changing array manifold through movement of the UAVs, and re-sensing the source signals;

S3: for each sensed source signals, calculating a covariance matrix, and obtaining a corresponding noise subspace through eigenvalue decomposition; and

S4: constructing a quadratic optimization problem based on the noise subspace and array steering vector, constructing a cost function, and implementing joint estimation of a gain-phase error and a DOA through spectrum peak search.

In order to optimize the technical solution, the following specific measures are also used.

Further, in S1, a process of equipping each UAV with the antenna, and forming the receive array through the swarm of multiple UAVs to receive the source signals includes the following steps:

S11: enabling M UAVs to be evenly arranged, and equipping each UAV with the antenna, where a distance between array elements is a unit distance d=λ/2, λ represents a wavelength, and M is a positive integer greater than 2; and

S12: assuming that K parallel plane waves are incident from a direction θ_k, where k=1, 2, . . . , K, K is a positive integer greater than 2, and when the array has a gain-phase error, a signal received by the array in an initial state is expressed as:x₁(t)=CA₁s(t)+n(t)

where C=diag(c)=diag(c₁, c₂, . . . , c_M) is a gain-phase error diagonal matrix, where diag(c) is a diagonal matrix formed by elements in a vector c; s(t)=[s₁(t), s₂(t), . . . , s_K(t)]^T is a signal vector, n(t) is additive white Gaussian noise, A₁=[a₁(θ₁), a₁(θ₂), . . . , a₁(θ_K)] is a direction matrix, and a₁(θ_k) is an array steering vector in the direction θ_k and is expressed as:a₁(θk)=[e^{−jπd11sin θk},e^{−jπd12sin θk}, . . . ,e^{−jπd1Msin θk}]^T

where d₁₁, d₁₂, . . . , d_1M represent respective position information of the M UAVs in the initial state.

Further, in step S3, a process of calculating, for each sensed source signals, the covariance matrix, and obtaining the corresponding noise subspace through eigenvalue decomposition includes the following steps:

S21: when the baseline remains unchanged, changing a position of the corresponding array element through movement of the UAVs, to form a new array, and sensing a corresponding source signal for the newly formed array;

S22: for the i-th sensed source signal x_i(t), calculating a covariance matrix according to the following formula:

$R_i=E\left[x_i\left(t\right)x_i^H\left(t\right)\right]=\frac{1}{L}\sum _{l=1}^L{x}_i\left(t_l\right)x_i^H\left(t_l\right)$

where L represents the number of data snapshots; E(·) represents expectation; (·)^H represents a conjugate transpose operation; and t_l represents the l-th snapshot; and

S23: performing eigen decomposition on R_i to implement eigenvalue decomposition for the covariance matrix, which is expressed as:R_i=U_SiD_SiU_Si^H+U_NiD_NiU_Ni^H

where D_Si represents a K×K-dimension diagonal array, of which diagonal elements include K larger eigenvalues obtained through the eigenvalue decomposition, and D_Ni is a diagonal array of M−K smaller eigenvalues; U_Si is a matrix of eigenvectors corresponding to the K larger eigenvalues, and U_Ni is a matrix of eigenvectors corresponding to other eigenvalues; U_Si and U_Ni are called signal subspace and noise subspace, respectively; and i=1, 2, . . . , p, where p is the total number of observations.

Further, in step S4, a process of constructing the quadratic optimization problem based on the noise subspace and the array steering vector, constructing the cost function, and implementing joint estimation of the gain-phase error and the DOA through spectrum peak search includes the following steps:

S41: constructing the quadratic optimization problem:min_θc^HQ(θ)c, s.t. e₁^Hc=1

where c=[c₁, c₂, . . . , c_M] represents the gain-phase error, e₁=[1, 0, . . . , 0]^T, (·)^T represents a transpose operation, and θ represents a to-be-estimated DOA parameter;

$Q\left(\theta \right)={\left[\begin{array}{c}{U}_{N_1}^H\mathrm{diag}\left(a_1\left(\theta \right)\right)\\ ⋮\\ U_{N_P}^H\mathrm{diag}\left(a_P\left(\theta \right)\right)\end{array}\right]}^H\left[\begin{array}{c}{U}_{N_1}^H\mathrm{diag}\left(a_1\left(\theta \right)\right)\\ ⋮\\ U_{N_P}^H\mathrm{diag}\left(a_P\left(\theta \right)\right)\end{array}\right]$

where a_i(θ) represents a steering vector at the i-th observation, diag(a_i(θ)) represents a diagonal matrix formed by elements in a vector a_i(θ), and i=1, 2, . . . , p;

S42: constructing the cost function:L(θ,c)=c^HQ(θ)c−ε(e₁^Hc−1)

where ε is a constant;

S43: finding a partial derivative of L(θ,c):∂L(θ,c)/∂c=2Q(θ)c−εe₁=0

where c=ξQ⁻¹ (θ)e₁, and ξ is a constant;

S44: obtaining estimated angle and gain-phase error values:

$\hat{c}=\frac{Q^{-1}\left(\theta \right)e_1}{e_1^H{Q}^{-1}\left(\theta \right)e_1};\mathrm{and}$

S45: substituting the expression of ĉ into min_θ c^HQ(θ)c, and calculating an estimated DOA value:

$\hat{\theta }=\arg \underset{\theta }{\min }\frac{1}{e_1^H{Q}^{-1}\left(\theta \right)e_1}=\arg \underset{\theta }{\max }{e}_1^H{Q}^{-1}\left(\theta \right)e_1.$

Further, the method further includes the following steps:

using a root mean square error (RMSE) as a performance estimation indicator to evaluate validity of the estimation results; and calculating the corresponding RMSE according to the following formulas:

$\begin{array}{c}\mathrm{RMS}{E}_{\theta }=\sqrt{\frac{1}{NK}\sum _{n=1}^N\sum _{k=1}^K{\left({\theta }_k-{\hat{\theta }}_{k,n}\right)}^2}\\ \mathrm{RMS}{E}_c=\sqrt{\frac{1}{MN}\sum _{n=1}^N\sum _{m=1}^M{\left(c_m-{\hat{c}}_{m,n}\right)}^2}\end{array}$

where N represents the number of Monte Carlo simulations, θ_k represents a real incident angle of the k-th signal, {circumflex over (θ)}_k,n represents an estimated angle value of the k-th signal in the n-th simulation experiment, c_m represents a real value of the m-th gain-phase error coefficient, and ĉ_m,n represents an estimated value of the m-th gain-phase error coefficient in the n-th simulation experiment.

The present disclosure has the following beneficial effects:

Compared with the prior art, the present disclosure breaks through the limitations of DOA estimation on a gain-phase error in the prior art, and can obtain an accurate estimated angle value, and achieve more accurate positioning. In the presence of a gain-phase error, the present disclosure can estimate and calibrate the gain-phase error value without requiring auxiliary calibration sources, auxiliary calibration array elements, or iterative solutions, and can achieve high-resolution estimation.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a flowchart of a method for jointly estimating a gain-phase error and a DOA based on a UAV array according to the present disclosure.

FIG. 2 is a scenario diagram of a receive array formed by a UAV swarm according to the present disclosure.

FIG. 3 is a comparison diagram of spectrum peaks with a gain-phase error uncalibrated and calibrated by using the method of the present disclosure.

FIG. 4 is a comparison diagram of DOA estimation performance of the present disclosure under different signal-to-noise ratios (SNRs).

FIG. 5 is a comparison diagram of estimation performance of a real part and an imaginary part of a gain-phase error in the present disclosure in under different SNRs.

DETAILED DESCRIPTION OF THE EMBODIMENTS

The present disclosure is described in further detail below with reference to the accompanying drawings.

It should be noted that, as used herein, terms such as “upper”, “lower”, “left”, “right”, “front” and “back” are merely employed for ease of a description, and not intended to limit the implementable scope of the present disclosure, and a change or adjustment of its relative relation shall also be deemed as falling within the implementable scope of the present disclosure without a substantial alteration of a technical content.

For ease of description, the meanings of symbols in the embodiments are as follows: E(·) represents expectation, (·)^H represents a conjugate transpose operation, (·)^T represents a transpose operation, and diag(a) represents a diagonal matrix formed by elements in a vector a.

FIG. 1 is a flowchart of a method for jointly estimating a gain-phase error and a DOA based on a UAV array according to the present disclosure. In this method, multiple UAVs are used as platforms to carry array elements, and multiple array elements form an array to receive signals. The method includes the following steps:

S1: equipping each UAV with an antenna, and forming a receive array through a swarm of multiple UAVs to receive source signals;

S2: when an observation baseline of the swarm remains unchanged, changing array manifold through movement of the UAVs, and re-sensing the source signals;

S3: for each sensed source signals, calculating a covariance matrix, and obtaining a corresponding noise subspace through eigenvalue decomposition; and

S4: constructing a quadratic optimization problem based on the noise subspace and array steering vector, constructing a cost function, and implementing joint estimation of a gain-phase error and a DOA through spectrum peak search.

Specific implementation steps are as follows:

Step 1: Receiving a Signal

Enable M UAVs to be evenly arranged, and equip each UAV with an antenna, where a distance between array elements is a unit distance d=λ/2, and λ represents a wavelength. Assume that K parallel plane waves are incident from θ_k (k=1, 2, . . . , K). When the array has a gain-phase error, a signal received by the array may be expressed as:x₁(t)=CA₁s(t)+n(t)

where C=diag(c)=diag(c₁, c₂, . . . , c_M) is a gain-phase error diagonal matrix, s(t)=[s₁(t), s₂(t), . . . , s_K(t)]^T is a signal vector, n(t) is additive white Gaussian noise, A₁=[a₁(θ₁), a₁(θ₂), . . . , a₁ (θ_K)] is a direction matrix, and a₁(θ_k) is an array steering vector in the direction θ_k and is expressed as:a₁(θk)=[e^{−jπd11sin θk},e^{−jπd12sin θk}, . . . ,e^{−jπd1Msin θk}]^T

where d₁₁, d₁₂, . . . , d_1M represents current positions of the UAVs.

According to a data model, information of the received signal may be obtained. Calculate a covariance matrix, which may be expressed as:

$R_1=E\left[x_1\left(t\right)x_1^H\left(t\right)\right]=\frac{1}{L}\sum _{l=1}^L{x}_1\left(t_l\right)x_1^H\left(t_l\right)$

where L represents the number of data snapshots. Perform eigen decomposition on R₁ to implement eigenvalue decomposition for the covariance matrix, which may be expressed as:R₁=U_S1D_S1U_S1^H+U_N1D_N1U_N1^H

where D_S1 represents a K×K-dimension diagonal array, of which diagonal elements include K larger eigenvalues obtained by the eigenvalue decomposition, and D_N1 is a diagonal array of M−K smaller eigenvalues; U_S1 is a matrix of eigenvectors corresponding to the K larger eigenvalues, and U_N1 is a matrix of eigenvectors corresponding to other eigenvalues; U_S1 and U_N1 are called signal subspace and noise subspace, respectively.

Step 2: Obtaining Multiple Noise Subspaces

When the baseline remains unchanged (that is, an angle of incidence of signals remains unchanged), change a position of the corresponding array element through movement of the UAVs, to form a new array. In this case, re-receive a signal and perform the same processing, to obtain a noise subspace U_N2. Control the UAVs to go on moving, to form a new array, and obtain a noise subspace U_Np at the p-th observation, assuming that a total of P observations are made.

Step 3: Jointly Estimating a DOA and a Gain-Phase Error

When there is a gain-phase error, the MUSIC function changes to:U_N1^H diag(a₁(θ_k))c=0U_N2^H diag(a₂(θ_k)c=0U_NP^H diag(a_P(θ_k))c=0

where c=[c₁, c₂, . . . , c_M] is the gain-phase error. Let

$Q\left(\theta \right)={\left[\begin{array}{c}{U}_{N_1}^H\mathrm{diag}\left(a_1\left(\theta \right)\right)\\ \dots \\ U_{N_P}^H\mathrm{diag}\left(a_P\left(\theta \right)\right)\end{array}\right]}^H\left[\begin{array}{c}{U}_{N_1}^H\mathrm{diag}\left(a_1\left(\theta \right)\right)\\ \dots \\ U_{N_P}^H\mathrm{diag}\left(a_P\left(\theta \right)\right)\end{array}\right]$ and construct the quadratic optimization problem:min_θc^HQ(θ)c, s.t. e₁^Hc=1

where e₁=[1, 0, . . . , 0]^T. Construct the cost function:L(θ,c)=c^HQ(θ)c−ε(e₁^Hc−1).

Find a partial derivative of L(θ,c): ∂L(θ,c)/∂c=2Q(θ)c−εe₁=0, and c=ξQ⁻¹(θ)e₁, where ξ is a constant. From e₁^Hc=1, ξ=1/e^HQ⁻¹(θ)e₁ may be obtained. Therefore, an estimated value of c may be obtained:

$\hat{c}=\frac{Q^{-1}\left(\theta \right)e_1}{e_1^H{Q}^{-1}\left(\theta \right)e_1}$

Substitute the expression of ĉ into min_θ c^HQ(θ)c. An estimated value of DOA may be expressed as:

$\hat{\theta }=\arg \underset{\theta }{\min }\frac{1}{e_1^H{Q}^{-1}\left(\theta \right)e_1}=\arg \underset{\theta }{\max }{e}_1^H{Q}^{-1}\left(\theta \right)e_1.$

To verify the effectiveness of the algorithm of the present disclosure, MATLAB simulation analysis is made, where an RMSE is used as a performance estimation indicator, which is defined as:

$\begin{array}{c}\mathrm{RMS}{E}_{\theta }=\sqrt{\frac{1}{NK}\sum _{n=1}^N\sum _{k=1}^K{\left({\theta }_k-{\hat{\theta }}_{k,n}\right)}^2}\\ \mathrm{RMS}{E}_c=\sqrt{\frac{1}{MN}\sum _{n=1}^N\sum _{m=1}^M{\left(c_m-{\hat{c}}_{m,n}\right)}^2}\end{array}$

where N represents the number of Monte Carlo simulations, θ_k represents a real incident angle of the k-th signal, {circumflex over (θ)}_k,n represents an estimated angle value of the k-th signal in the n-th simulation experiment, c_m represents a real value of the m-th gain-phase error coefficient, and ĉ_m,n represents an estimated value of the m-th gain-phase error coefficient in the n-th simulation experiment.

FIG. 2 is a scenario diagram of the present disclosure. The number of UAVs selected for the simulation is M=6, P=2, d₁=λ/2, and d₂=0.35λ, where λ represents a wavelength.

FIG. 3 is a comparison diagram of spectrum peaks before and after a gain-phase error is calibrated by using the method of the present disclosure. In the simulation, it is assumed that an SNR is 20 dB and the number of snapshots is J=500. It can be seen from the spectrum peak diagram that the gain-phase error affects positions and amplitudes of the spatial spectrum peaks. When there is a gain-phase error, estimation performance of the MUSIC algorithm decreases. If the error is large, the algorithm may even fail.

FIG. 4 is a comparison diagram of DOA estimation performance of the method proposed in the present disclosure and an uncalibrated method under different SNRs. It can be seen from the simulation results that the method of the present disclosure achieves high estimation accuracy. In addition, the estimation accuracy of the method can be further improved by appropriately increasing the SNR.

FIG. 5 is a diagram of gain-phase error estimation performance of the method in the present disclosure under different SNRs. It can be seen from the simulation results that the method of the present disclosure can accurately estimate the real part and the imaginary part.

The present disclosure provides a method for jointly estimating a gain-phase error and a DOA based on an UAV array, relating to the technical field of array signal processing. According to the present disclosure, each UAV is equipped with an antenna, and a swarm of multiple UAVs form a receive array. By changing positions of the individual UAVs in the swarm, positions of the corresponding array elements are also changed, thereby changing an array manifold. For signals sensed by the array in multiple times, covariance matrices are calculated, and eigenvalue decomposition is performed to obtain multiple signal noise subspaces. A quadratic optimization problem is constructed based on the noise subspaces and an array steering vector, a cost function is constructed, and finally, a DOA is determined and a gain-phase error is estimated through spectrum peak search. The present disclosure breaks through the limitation that DOA estimation accuracy in the traditional cooperative sensing of a UAV swarm is limited by a gain-phase error between UAVs, and has important application value because it can implement joint estimation of a DOA and a gain-phase error accurately without requiring auxiliary signal sources, array elements, or iterative solutions.

What is described above is merely the preferred implementation of the present disclosure, the scope of protection of the present disclosure is not limited to the above examples, and all technical solutions following the idea of the present disclosure fall within the scope of protection of the present disclosure. It should be noted that several modifications and adaptations made by those of ordinary skill in the art without departing from the principle of the present disclosure should fall within the scope of protection of the present disclosure.

Claims

1. A method for jointly estimating a gain-phase error for a receive array and a direction of arrival (DOA) of source signals based on an unmanned aerial vehicle (UAV) array, wherein the method comprises the following steps: S1: equipping each UAV with an antenna, and forming the receive array through a swarm of multiple UAVs to receive the source signals;S2: when an observation baseline of the swarm remains unchanged, changing array manifold through movement of the UAVs, and sensing the source signals with the moved UAVs;S3: for each sensed source signal, calculating a covariance matrix, and obtaining a corresponding noise subspace through eigenvalue decomposition;S4: constructing a quadratic optimization problem based on the noise subspaces and an array steering vector, constructing a cost function, and implementing joint estimation of the gain-phase error and the DOA through spectrum peak search;wherein in step S1, a process of equipping each UAV with the antenna, and forming the receive array through the swarm of multiple UAVs to receive the source signals comprises the following steps: S11: enabling M UAVs to be evenly arranged in an initial state, and equipping each UAV with the antenna, wherein a distance between array elements of the receive array is a unit distance d=λ/2, λ represents a wavelength, and M is a positive integer greater than 2; andS12: assuming that K parallel plane waves having the wavelength are incident from a direction θk, wherein k=1, 2, . . . , K, K is a positive integer greater than 2, a signal received by the receive array in the initial state is expressed as: x1(t)=CA1s(t)+n(t) wherein C=diag(c)=diag(c1, c2, . . . , cM) is a gain-phase error diagonal matrix, wherein diag(c) is a diagonal matrix formed by elements in a vector c; s(t)=[s1(t), s2(t), . . . , sK(t)]T is a signal vector, n(t) is additive white Gaussian noise, A1=[a1(θ1), a1(θ2), . . . , a1(θK)] is a direction matrix, and a1(θk) is an array steering vector in the direction θk and is expressed as: a1(θk)=[e−jπd11sin θk,e−jπd12sin θk, . . . ,e−jπd1Msin θk]T  wherein d11, d12, . . . , d1M represent respective position information of the M UAVs in the initial state;wherein step S2 comprises: S21: when the baseline remains unchanged, changing a position of the corresponding array element through movement of the UAVs, to form a new array, and sensing a corresponding source signal for the new array;wherein in step S3, a process of calculating, for each sensed source signals, the covariance matrix, and obtaining the corresponding noise subspace through eigenvalue decomposition comprises the following steps: S22: for an i-th sensed source signal xi(t), calculating a covariance matrix according to the following formula:

2. The method for jointly estimating the gain-phase error and the DOA based on the UAV array according to claim 1, wherein the method further comprises the following step: using a root mean square error (RMSE) as a performance estimation indicator to evaluate validity of estimation results; and calculating a corresponding RMSE according to the following formulas: