Three-Dimensional Coprime Array for Massive MIMO: Array Configuration Design and 2D DOA Estimation

Gong, Pan; Ahmed, Tanveer; Li, Jianfeng

doi:https://doi.org/10.1155/2020/2686257

Wireless Communications and Mobile Computing

On this page

Abstract Introduction Conclusions Data Availability Conflicts of Interest Acknowledgments References Copyright Related Articles

Special Issue

Antenna Design Techniques for 5G Mobile Communications

View this Special Issue

Research Article | Open Access

Volume 2020 | Article ID 2686257 | https://doi.org/10.1155/2020/2686257

Three-Dimensional Coprime Array for Massive MIMO: Array Configuration Design and 2D DOA Estimation

Pan Gong,¹Tanveer Ahmed,¹and Jianfeng Li¹

Guest Editor: Raed A. Abd-Alhameed

Received25 Jul 2019

Revised14 Oct 2019

Accepted28 Dec 2019

Published13 Jan 2020

Abstract

In massive multiple-input multiple-output (MIMO) systems, it is critical to obtain the accurate direction of arrival (DOA) estimation. Conventional three-dimensional array mainly focuses on the uniform array. Due to the dense arrangement of the sensors, the array aperture is limited and severe mutual coupling effects arise. In this paper, a coprime cubic array (CCA) configuration design is presented, which is composed of two uniform cubic subarrays and can extend the interelement spacing with a selection of three pairs of coprime integers. Compared with uniform cubic array (UCA), CCA achieves the larger array aperture and less MC effects. And the analytical expression of Cramer–Rao Bound (CRB) for CCA is derived which verifies that the proposed CCA geometry outperforms the conventional UCA in two-dimensional (2D) DOA estimation performance in massive MIMO systems. Meanwhile, we propose a computationally efficient 2D DOA estimation algorithm with high accuracy for CCA. Specifically, we utilize array mapping to extract two uniform arrays from the nonuniform array by exploiting the relation derived from the signal subspace and the two directional matrices. Then, we operate a reduced dimension process on the uniform arrays and convert the 2D spectrum peak searching (SPS) problem into one-dimensional (1D) one, which significantly reduces the computational complexity. In addition, we employ the polynomial root finding technique with a lower complexity instead of 1D SPS to further relieve the computational complexity. Simultaneously, with coprime property, the phase ambiguity problem is solved, which results from the large interelement spacing. Numerical simulation results demonstrate that the proposed algorithm is very computationally efficient without degradation of DOA estimation performance.

1. Introduction

Nowadays, massive multiple-input multiple-output (MIMO), known as large-scale MIMO, is considered as one of the promising technologies in the development of future wireless communication systems. Massive MIMO attracts considerable attention due to the high throughput, enhanced link reliability, and improved spectral efficiency [1, 2]. Also, massive MIMO can effectively reduce the latency and robustness to interference [3, 4], which are important factors in wireless communication systems.

Direction of arrival (DOA) estimation plays an important role in massive MIMO, since precise DOA estimation is vital for the base station (BS) to conduct downlink precoding or beamforming [5]. DOA estimation is also widely utilized in engineering applications, like radar, sonar, wireless communication, and navigation [6–9]. Many classic DOA estimation algorithms [10–13] and plenty of derived algorithms [14–16] have been proposed to solve the DOA estimation problem. Most of these existing algorithms mainly utilize the conventional uniform array configurations [17, 18] with the interelement spacing no larger than typical half-wavelength to tackle the problem of phase ambiguity [19]. However, in the massive MIMO systems, with a large number of antennas, severe mutual coupling (MC) effects arise, which significantly affect the DOA estimation accuracy.

In recent years, a new configuration called coprime array [20] has been a hot research direction and attracts much attention since it can significantly enhance the degree of freedom (DOF) [21], improve the resolution [22], and relieve MC effects [23]. This motivates many studies of DOA estimation using coprime array over the decade. In [24], a total angular search based algorithm by employing the multiple signals classification (MUSIC) algorithm was proposed. By exploiting the coprime property, the phase ambiguity problem is tackled. In [25], an ambiguity-free DOA estimation algorithm was proposed, which exploits the total information including self-information and mutual information to eliminate the ambiguity problem with high accuracy of DOA estimates. However, due to the total angular search, this algorithm results in large computational complexity. To deal with this obstacle, a partial spectral search (PSS) algorithm was introduced which only performs spectrum searching in a restricted sector and hence lowers the computational cost but has no effect on DOA estimation performance [26]. In the case of multiple sources, this algorithm will cause the target matching error when eliminating the phase ambiguity. To tackle the problem, an improved DOA estimation algorithm based on root-MUSIC was proposed in [27] for coprime linear array (CLA), which employs the relation between steering matrices and signal subspaces of two subarrays to achieve DOA estimation.

The algorithms mentioned above [24–27] were presented for one-dimensional (1D) DOA estimation, whereas, practically, two-dimensional (2D) DOA estimation possesses more importance, and various studies have been introduced [28–30]. In [28], the PSS algorithm for coprime planar array (CPA) was presented, which can reduce the computational complexity. A reduced dimension notion to achieve 2D DOA estimation was proposed in [29]. A generalized CPA geometry with more flexible array layouts was constructed for 2D DOA estimation [30], which can attain more DOFs compared with CPA configuration. However, spectrum peak searching (SPS) is involved in these mentioned works [28–30] and hence causes expensive computational cost.

In this paper, on the basis of the 1D CLA and 2D CPA, we construct a three-dimensional (3D) array geometry called coprime cubic array (CCA). In the context of massive MIMO systems [31, 32], MC effects become important which have a vital influence on DOA estimation accuracy. The geometry of CCA possesses the extended array aperture and can effectively reduce MC effects. Especially, in the cases with small elevations, the CCA obtains superior DOA estimation performance. Simultaneously, the Cramer–Rao Bound (CRB) is derived which demonstrates that the proposed CCA geometry outperforms the conventional uniform cubic array (UCA) configuration [33–35]. In addition, we propose an array mapping and reduced dimension based on computationally efficient MUSIC (AMRD-MUSIC) algorithm to estimate 2D DOAs. Different from the conventional multiple-dimensional MUSIC (MD-MUSIC) algorithm, where 2D SPS involves a tremendous computation burden, in the proposed algorithm, we utilize array mapping to extract two uniform arrays from the nonuniform array by exploiting the relation derived from the signal subspace and the two directional matrices. Then, we operate a reduced dimension process on the uniform arrays and convert the 2D spectrum peak searching (SPS) problem into 1D one, which significantly reduces the computational complexity. In addition, we employ the polynomial root finding technique with a lower complexity instead of 1D SPS to further reduce the computational complexity. Meanwhile, according to the coprime property, the phase ambiguity problem is solved which results from large interelement spacing and the proposed algorithm obtains paired angles automatically. Numerical simulation results demonstrate that the proposed AMRD-MUSIC algorithm can significantly reduce the computational complexity cost with no degradation of DOA estimation performance.

Specifically, we summarize the main contributions of this paper.(1)In massive MIMO systems, conventional 3D array mainly focuses on the uniform array, which limits the array aperture and suffers from severe MC effects due to the dense location of sensors. We construct a CCA configuration which can achieve the larger array aperture and less MC effects compared with UCA. In addition, the analytical expression of CRB with CCA is derived and simulations demonstrate that the proposed CCA performs better DOA estimation performance than the conventional UCA configuration.(2)We propose an AMRD-MUSIC algorithm for 2D DOA estimation that can achieve almost the same DOA estimation performance as classic MD-MUSIC algorithm but with lower computational complexity. Through array mapping, we extract two uniform arrays from the nonuniform array and operate a reduced dimension on these two arrays to reduce 2D SPS into 1D one, which can effectively lower the computational complexity. To further reduce complexity, we utilize the polynomial root finding technique with a lower complexity instead of SPS. Therefore, computational complexity obtains a significant decrease.(3)The proposed algorithm can achieve the full DOFs and obtain DOA estimates with pairing automatically.

Section 2 introduces the array configuration and signal model. In Section 3 and Section 4, we present the proposed algorithm and analyse the performance of the proposed algorithm, respectively. Sections 5 and 6 provide numerical simulations and conclusions, respectively.

Notations 1. We use as the transpose, and is utilized as the conjugate transpose. signifies Khatri–Rao product and denotes Kronecker product. is expectation. is a diagonal matrix that is formed of the m-th row of the matrix. is a phase operator.

2. Array Configurations and Signal Model

2.1. Uniform Cubic Array

UCA geometry is designed with sensors, where denotes the number of sensors in the x-axis, and denote the number of sensors in the y-axis and z-axis, respectively. The interelement spacing of UCA is , which denotes the interelement spacing of x-axis, y-axis, and z-axis, respectively. is wavelength. Figure 1(a) shows an example of UCA with .

(a)

(b)

2.2. Coprime Cubic Array

For UCA configuration, the array aperture is limited by the interelement spacing no larger than conventional half-wavelength. Simultaneously, MC effects exist due to the dense location of sensors. Based on CLA and CPA, we propose a CCA configuration, which can effectively increase the array aperture and decrease the MC effects.

Definition 1 (coprime cubic array (CCA)). The CCA consists of two subarrays. The first subarray is composed of sensors with the interelement spacing of in x-axis direction, in y-axis direction, and in z-axis direction. For the second subarray, it contains elements and , , , where and are coprime integer pairs and stand for the number of sensors in x-axis, y-axis, and z-axis, respectively. Therefore, there are sensors in total. Figure 1(b) is an example of CCA geometry with and , , .
In the following part, we compare the array performance in array aperture and MC effects of UCA and CCA, respectively.

2.3. Array Aperture Comparison

In this subsection, we compare the array aperture of UCA and CCA in the same cases. For a fair comparison, we assume that the number of sensors in UCA and CCA is basically the same. According to [36], we can compute the array apertures of these arrays, which are denoted as and

For example, we set the UCA with , so the total number of sensors in UCA is One subarray of CCA is of and the other subarray is of The number of sensors in CCA is in total. We can note that the CCA has fewer number of sensors than UCA. As a result, for the two given arrays, the array aperture can be obtained as , . So we have

It can be seen that the proposed CCA configuration can effectively enlarge the array aperture with fewer sensors.

2.4. Mutual Coupling Effects Comparison

Array signal models usually neglect the MC between the sensors. In practical application, the MC effects between the closely located sensors should be considered. Without loss of generality, we employ a coupling matrix , where denotes the total number of sensors in the array. In general, the expression of is very complicated. According to [37], for given UCA and CCA configurations, can be denoted approximately by a B-banded Toeplitz matrix as follows:where the value of is inversely proportional to sensor distance and satisfies is a constant and denotes the positions of sensors.

According to [37], the coupling leakage can be exploited to evaluate the total MCwhere is the amount of MC. It is indicated that the smaller is, the less MC is.

We compare the coupling leakages of the given UCA of and CCA of and The total number of sensors of UCA and CCA is and , respectively. Hence, we have and . So the coupling coefficients can be computed by the following:where and denotes the distance between the random two sensors. Then, we compute the coupling leakages of UCA and CCA as and , respectively. It can see that the proposed CCA can effectively decrease the MC effects.

Table 1 shows the comparison of array aperture and coupling leakage for UCA and CCA in the same conditions. From Table 1, we can see that the proposed CCA configuration has the least sensors but outperforms the UCA in array aperture and MC effects. In addition, the CCA achieves better DOA estimation performance that is shown in Section 5. Therefore, in the following part, we employ the CCA into consideration.

2.5. Signal Model

Assume that there are far-field uncorrelated narrowband incident signals impinging on the CCA from , where and denote the DOA (the elevation angle and azimuth angle) of the k-th target [19]. And we define , , , and . In the following, we take the subarray that has sensors to derive the proposed algorithm in the following part.

We denote the received signal as follows [28]:where represents the signal matrix and denotes the number of snapshots. is the additive white Gaussian noise matrix, whose variance and mean are and zero, respectively. denotes directional matrix of the i-th subarray and is presented by the following:where and correspond to steering vectors of x-axis, y-axis, and z-axis of the subarray .

Denote the directional matrix in (7) as follows:where denotes directional matrix of x-axis of the subarray i, and stand for the matrices of y-axis and z-axis, respectively, denotes a diagonal matrix which is formed of the n-th row of the matrix for the subarray i, and denotes a diagonal matrix which is formed of the m-th row of the matrix for the subarray i.

3. 2D DOA Estimation Algorithm

Conventional MD-MUSIC algorithm can obtain DOA estimates with high accuracy by selecting a refined search step, which causes expensive computational complexity. To tackle the problems, an AMRD-MUSIC algorithm is proposed.

3.1. MD-MUSIC Spectrum

We concatenate the array measurements and as follows:where and . and Then, the covariance matrix of the total array can be computed by the following: where refers to noise subspace and signifies signal subspace. is a diagonal matrix with the K largest eigenvalues of and contains the remaining eigenvalues. Here, the symbol of ( ^ ) means the estimation of the covariance matrix .

Then, according to 1D MUSIC spectrum [12], the MD-MUSIC spectrum can be solved as follows:

3.2. Array Mapping

In this subsection, we utilize array mapping to extract two uniform arrays from the nonuniform array by exploiting the relation derived from the signal subspace and the two directional submatrices.

As the signal subspace spans the same space with the steering matrix doing, we havewhere represents a nonsingular matrix. Then, the signal subspace can be decomposed into two parts:where and .

Then, the transformation matrices can be defined as follows [38]:where denotes the pseudoinverse operation. From (15), we have the following:and the steering vectors of and also satisfy this relation in (16), so we have the following:

By utilizing (17), we transform (13) into the following: where and .

3.3. Reduced Dimension Processing

In this part, we employ the double reduced dimension processing on (16) and (17), respectively, to transform the 2D SPS into 1D one. Therefore, the computational complexity can be significantly reduced.

As , then we havewhere and are both identity matrices.

Then, we define the function as follows:which transforms the 2D SPS to find the minimum value of (22). According to (20) and (21), we have the following:where , , and . Note that (23) is an optimization problem. Based on , where , we can eliminate the trivial solution . This optimization problem can be reconstructed as follows:

Then, we can construct the cost function as follows:where is a constant. Then, we have the following:

According to (26), we have , where is a constant. Combining (24), we have the following:

Then, can be denoted as follows:

Based on (24) and (28), we can obtain the following:

According to (29), the 2D SPS is transformed into 1D one by utilizing a reduced dimension process. To further reduce complexity, we employ the polynomial rooting technique with the lower computational complexity to obtain the final DOA estimates instead of the 1D SPS process.

3.4. Polynomial Root Finding Process

Define and denote as follows:

Then, we rewrite as follows:

Then, we employ polynomial root finding technique instead of 1D SPS.

If cannot match with one source signal and ifthe matrix is invertible [38].

In order to achieve the polynomial (32), satisfies

Then, we can obtain the final by finding K roots which are closest to the unit circle of the polynomial ,

In order to obtain the estimation of or , the similar reduced dimensional polynomial root finding technique can be employed again. We can reform the whole steering matrix as , which contains the same elements as the matrix and only differs in elements positions. Then, a reconstructed received signal can be obtained. Then, we operate the proposed array mapping, reduced dimension, and polynomial rooting technique again to attain the estimate of

Similarly, we can achieve . Phase ambiguity problem, resulting from the larger element spacing than half-wavelength, will be solved by coprime property in the following section.

3.5. Ambiguity Elimination

Firstly, the phase ambiguity problem is illustrated. Then, the coprime property is exploited to tackle the problem.

Assume that only one signal located with impinges on the CCA subarray and the subarray consists of sensors. is assumed as one of the ambiguous DOAs of the real angle .

According to [38],where the mod operation depends on the period of the exponential function . So it can be noted that the phase difference between the real and ambiguous DOAs is [39] and are integers. And denote the element spacing among three axes, respectively. , .

Then, we employ the obtained estimates to recover all estimates [28],where , and are defined and they denote the ambiguous values.

As and , we have , , . Simultaneously, and demand to satisfy each estimate.

As the interelement spacing of conventional UCA is limited to half-wavelength, , , can only be the value of zero, which means no ambiguity problem exists. However, in CCA, due to the interelement spacing larger than half-wavelength, there are other pairs to make (37)–(39) hold besides the infeasible , , and pairs, whereas only one pair is the candidate related to the actual DOA. Coprime property is utilized to solve the phase ambiguity problem and achieve true DOA.

Assume as one of the ambiguous DOAs that appears in one of the subarrays. Simultaneously, define as the corresponding actual DOA. There exists the following:

It exists that , , , [38], so we can obtain

Since , , and are coprime integers pairs, respectively, (41) can hold only in the case of , , . It is indicated that the ambiguity problem is solved and actual DOA estimates can be achieved.

In practice, we only need two variables of to achieve DOAs estimates, where , are the estimates of .

The final estimate of can be calculated by the following:

3.6. Detailed Steps of the Proposed Algorithm

The steps are provided in detail as follows: Step 1: compute the covariance of the received signal data to obtain the MD-MUSIC spectrum function (12). Step 2: according to the signal subspace, construct and by (15), with which we can obtain the reconstructed spectrum function according to (19). Step 3: by array mapping, we extract the two uniform arrays (19) from the nonuniform array (12) according to the relation derived from two directional matrices. Step 4: according to (20) and (21), transform the 2D SPS process into 1D SPS. Step 5: according to (32), convert 1D PSS to polynomial root finding process and obtain the K roots of . Similarly, we can achieve and . Step 6: by (41), all the ambiguous DOA estimates are recovered, and by searching for the nearest ones, the actual estimations of , and can be achieved. Step 7: according to (43), the final DOAs estimates can be obtained.

4. Performance Analysis of the Algorithm

4.1. DOF Comparison

The proposed AMRD-MUSIC algorithm can obtain DOFs, which indicates that the proposed algorithm can achieve signals by employing antennas.

The algorithms in [24, 26] process the two subarrays separately so that they have a great loss in DOFs since they only exploit the autoinformation. The number of achievable DOFs by using the method in [24, 26] can be denoted as follows:where .

Thus, the proposed algorithms can detect many more signals than conventional algorithms with the same sensors. From (44) and (45), we can note the relationship of DOFs of different algorithms as follows:

Figure 2 depicts the maximum number of achievable DOFs to illustrate the signals explicitly without ambiguous DOA estimations, where the number of snapshots is set to be with noise-free, the total number of signals is , and the total number of sensors is . The elevation angles and azimuth angles are set to be and respectively. From Figure 2, we can obviously see that the 28 signals can be detected uniquely due to the employment of total correlation matrix information including autocorrelation matrix and mutual correlation matrix, where no ambiguous angles arise; however, the methods in [24, 26] fail to detect 28 signals and they can detect 11 signals at most in this scenario.

4.2. Complexity Analysis

In this subsection, we compute the complexity of the classical MD-MUSIC algorithm, multiple-dimensional ESPRIT (MD-ESPRIT) algorithm and the proposed AMRD-MUSIC algorithm for the same array configuration of CCA with the same sensors.

For the MD-MUSIC algorithm, the total complexity is composed of three parts, including computing covariance, eigenvalues decomposition, and spectrum peak searching. The complexity of these three parts is presented as , and , respectively, where L represents the number of snapshots and denotes the searching step. So the total computational complexity can be denoted as .

And we compute the complexity of the MD-ESPRIT algorithm. For the two subarrays, complexity consists of three parts, computing the covariance as , eigenvalue decomposition as , computing pseudoinverse and eigenvalue decomposition after reconstruction as . So the total complexity of these operations is denoted as .

In the proposed AMRD-MUSIC algorithm, the complexity can be denoted as , and the complexity of the other subarray is approximately similar to it.

We compare the complexity of the conventional MD-MUSIC algorithm, the MD-ESPRIT algorithm, and the proposed AMRD-MUSIC algorithm with the same CCA configuration. For clarity, we list the computational complexity of these algorithms with CCA in Table 2 with , and the searching step is set to be 0.02.

We compare the computational complexity versus the number of snapshots and sensors respectively of three algorithms including MD-MUSIC, MD-ESPRIT, and ARMD-MUSIC. Figures 3 and 4 show the complexity comparison with the increasing number of snapshots and sensors, respectively. It can be seen that the proposed AMRD-MUSIC algorithm achieves the lowest computational complexity than other algorithms. Figure 5 shows the complexity comparison versus the searching step. We can see that the complexity of MD-MUSIC reduces with the searching step increasing. However, the proposed AMRD-MUSIC algorithm outperforms the other algorithms.

4.3. Advantages of the Proposed Algorithm

In summary, the advantages of the proposed algorithm are listed as follows:(1)The proposed algorithm achieves significant complexity decrease by reducing 2D SPS into 1D one and utilizing polynomial root finding technique instead of 1D SPS.(2)The proposed algorithm possesses approximately the same DOA estimation performance as the MD-MUSIC algorithm and behaves better in its estimation than the classic MD-ESPRIT algorithm.(3)The proposed algorithm can obtain full DOFs due to the utilization of the total matrix information.(4)The proposed algorithm can achieve DOA angles with pairing automatically.

4.4. Cramer–Rao Bound

We provide the Cramer–Rao Bound (CRB) [40–42] analytical derivation of DOA estimation for CCA. The signal covariance can be obtained by the following:and we define the matrix parameter vector .

Then, we can denote the -th element of the Fisher information matrix (FIM) as follows [42]:whereand we denote ; then we have the following:

By utilizing (52), we can calculate the following:where

The matrix is positive definite since the in (50) is positive definite.

By lettingthe CRB can be obtained as follows:where .

5. Simulation Results

Numerical simulations are provided to evaluate the performance of the proposed array configuration of CCA and algorithm in this part. For the CCA geometry, it consists of two uniform cubic subarrays of and sensors, where , , , , , and . The total number of these two subarrays is computed as . The UCA consists of sensors. Assume signals impinging on the arrays from and . From Figure 6, we can clearly see that the AMRD-MUSIC algorithm is able to detect the DOA targets successfully.

The root mean square error (RMSE), as the performance metric of the DOA estimation, is defined as follows:where is the number of sources. are estimates of of the p-th Monte Carlo trial, respectively. In this paper, we set .

5.1. DOA Estimation Performance Comparison of Different Array Configurations with the Same Algorithm

We employ simulations to illustrate the DOA estimation performance of UCA and CCA configurations by using the same algorithm, where.

Figures 7 and 8 exhibit the RMSE performance comparison between CCA and UCA by utilizing the same algorithm. It captures explicitly that the proposed AMRD-MUSIC algorithm with CCA possesses better DOA estimation performance because of the extended array aperture and less MC effects. And CRB comparison shows that CCA can obtain the lower bound of theoretical DOA estimation performance compared with UCA geometry.

5.2. DOA Estimation Performance Comparison of Different Algorithms with the Same Array

In this subsection, we compare the RMSE performance of different algorithms with the same array of CCA, including the proposed AMRD-MUSIC algorithm, conventional MD-MUSIC algorithm, and MD-ESPRIT algorithm.

Figures 9 and 10 present the 2D angle estimation comparisons versus SNR and , respectively. The two figures clearly show that the proposed AMRD-MUSIC algorithm has approximately the same DOA estimation performance as the MD-MUSIC algorithm and outperforms the classic MD-ESPRIT algorithm owing to the utilization of total information including autocorrelation matrix and mutual correlation matrix, while the proposed algorithm has a lower computational complexity as shown in Figure 2.

5.3. Comparison of DOA Estimation Performance versus the Number of Sensors

In this subsection, we employ the proposed approach to a CCA geometry which consists of two subarrays including and sensors, where . The RMSE performance versus SNR and snapshots are displayed in Figures 11 and 12, respectively. It is obviously shown that the RMSE estimation performance of the proposed AMRD-MUSIC algorithm is improving with the increase of SNR or snapshots due to the diversity gain.

6. Conclusions

In this paper, a CCA geometry in massive MIMO systems is constructed, which consists of two uniform cubic subarrays and obtains the extension of interelement spacing by selecting three pairs of coprime integers. Meanwhile, CRB for CCA is derived and the proposed CCA geometry is verified to outperform the conventional UCA geometry in 2D DOA estimation performance in massive MIMO systems. In addition, we propose a computationally efficient 2D DOA estimation algorithm via array mapping and reduced dimension for CCA. In the proposed algorithm, we utilize the array mapping and map the nonuniform array into two uniform arrays and operate the reduced dimension process on the two uniform arrays to transform the 2D SPS problem into 1D one. In addition, the computational complexity gets further reduced by converting the 1D SPS problem into a polynomial root finding problem. Numerical simulation results verify the effectiveness of the proposed algorithm that can reduce the computational complexity cost with no degradation of DOA estimation performance.

Data Availability

No data were used to support this study.

Conflicts of Interest

The authors declare that they have no conflicts of interest.

Acknowledgments

This work was supported by the China NSF Grants (61971217, 61371169, 61601167), Shandong Province NSF Grants (ZR2016FM43), Key Laboratory of Dynamic Cognitive System of Electromagnetic Spectrum Space (Nanjing University of Aeronautics and Astronautics), and Ministry of Industry and Information Technology (KF20181915).

References

H. Q. Ngo, E. G. Larsson, and T. L. Marzetta, “Energy and spectral efficiency of very large multiuser MIMO systems,” IEEE Transactions on Communications, vol. 61, no. 4, pp. 1436–1449, 2013.
View at: Publisher Site | Google Scholar
Y. Huang, S. He, S. Jin, and W. Chen, “Decentralized energy-efficient coordinated beamforming for multicell systems,” IEEE Transactions on Vehicular Technology, vol. 63, no. 9, pp. 4302–4314, 2014.
View at: Publisher Site | Google Scholar
X. Gao, L. Dai, S. Han, C.-L. I, and R. W. Heath, “Energy-efficient hybrid analog and digital precoding for mmwave MIMO systems with large antenna arrays,” IEEE Journal on Selected Areas in Communications, vol. 34, no. 4, pp. 998–1009, 2016.
View at: Publisher Site | Google Scholar
J. Zuo, J. Zhang, C. Yuen, W. Jiang, and W. Luo, “Energy-efficient downlink transmission for multicell massive DAS with pilot contamination,” IEEE Transactions on Vehicular Technology, vol. 66, no. 2, pp. 1209–1221, 2017.
View at: Publisher Site | Google Scholar
E. G. Larsson, O. Edfors, F. Tufvesson, and T. L. Marzetta, “Massive MIMO for next generation wireless systems,” IEEE Communications Magazine, vol. 52, no. 2, pp. 186–195, 2014.
View at: Publisher Site | Google Scholar
Y. Z. Su, X. W. Liu, D. Y. Li, X. B. Wang, and Y. H. Zhang, “Research on impact localization in composite materials using array signal processing and lamb wave,” Advanced Materials Research, vol. 304, pp. 65–72, 2011.
View at: Publisher Site | Google Scholar
X. Zhang, L. Xu, L. Xu, and D. Xu, “Direction of departure (DOD) and direction of arrival (DOA) estimation in MIMO radar with reduced-dimension MUSIC,” IEEE Communications Letters, vol. 14, no. 12, pp. 1161–1163, 2010.
View at: Publisher Site | Google Scholar
J. Wang, J. Chen, and D. Cabric, “Cramer-rao bounds for joint RSS/DoA-Based primary-user localization in cognitive radio networks,” IEEE Transactions on Wireless Communications, vol. 12, no. 3, pp. 1363–1375, 2013.
View at: Publisher Site | Google Scholar
X. Wang, W. Wang, X. Li, and J. Liu, “Real-valued covariance vector sparsity-inducing DOA estimation for monostatic MIMO radar,” Sensors, vol. 15, no. 11, pp. 28271–28286, 2015.
View at: Publisher Site | Google Scholar
S. Marcos, A. Marsal, and M. Benidir, “The propagator method for source bearing estimation,” Signal Processing, vol. 42, no. 2, pp. 121–138, 1995.
View at: Publisher Site | Google Scholar
J. Capon, “High-resolution frequency-wavenumber spectrum analysis,” Proceedings of the IEEE, vol. 57, no. 8, pp. 1408–1418, 1969.
View at: Publisher Site | Google Scholar
R. Schmidt, “Multiple emitter location and signal parameter estimation,” IEEE Transactions on Antennas and Propagation, vol. 34, no. 3, pp. 276–280, 1986.
View at: Publisher Site | Google Scholar
R. Roy and T. Kailath, “ESPRIT-estimation of signal parameters via rotational invariance techniques,” IEEE Transactions on Acoustics Speech and Signal Pro-cessing, vol. 37, no. 7, pp. 984–995, 2002.
View at: Google Scholar
M. Pesavento, A. B. Gershman, and M. Haardt, “Unitary root-MUSIC with a real-valued eigen-decomposition: a theoretical and experimental performance study,” IEEE Transactions on Signal Processing, vol. 48, no. 5, pp. 1306–1314, 2000.
View at: Publisher Site | Google Scholar
X. Zhang, H. Wu, J. Li, and D. Xu, “Computationally efficient DOD and DOA estimation for bistatic MIMO radar with propagator method,” International Journal of Electronics, vol. 99, no. 9, pp. 1207–1221, 2012.
View at: Publisher Site | Google Scholar
F. Gao and A. B. Gershman, “A generalized ESPRIT approach to direction-of-arrival estimation,” IEEE Signal Processing Letters, vol. 12, no. 3, pp. 254–257, 2005.
View at: Publisher Site | Google Scholar
H. Chen, C. Hou, Q. Wang, L. Huang, W. Yan, and L. Pu, “Improved azimuth/elevation angle estimation algorithm for three-parallel uniform linear arrays,” IEEE Antennas and Wireless Propagation Letters, vol. 14, pp. 329–332, 2015.
View at: Publisher Site | Google Scholar
C. P. Huang and M. D. Zoltowski, “Eigenstructure techniques for 2-D angle estimation with uniform circular array,” IEEE Transactions on Signal Processing, vol. 42, no. 9, pp. 2395–2407, 1994.
View at: Publisher Site | Google Scholar
X. Zhang, W. Zheng, W. Chen, and Z. Shi, “Two-dimensional DOA estimation for generalized coprime planar arrays: a fast-convergence trilinear decomposition approach,” Multidimensional Systems and Signal Processing, vol. 30, no. 1, pp. 239–256, 2019.
View at: Publisher Site | Google Scholar
P. Pal and P. P. Vaidyanathan, “Sparse sensing with coprime samplers and arrays,” IEEE Transactions on Signal Processing, vol. 59, no. 2, pp. 573–586, 2011.
View at: Google Scholar
S. Qin, Y. D. Zhang, and M. G. Amin, “Generalized coprime array configurations for direction-of-arrival estimation,” IEEE Transactions on Signal Processing, vol. 63, no. 6, pp. 1377–1390, 2015.
View at: Publisher Site | Google Scholar
P. Pal and P. P. Vaidyanathan, “Coprime sampling and the music algorithm,” Digital Signal Processing Workshop and IEEE Signal Processing Education Workshop, pp. 289–294, 2011.
View at: Publisher Site | Google Scholar
Z. Ye and C. Liu, “2D DOA estimation in the presence of mutual coupling,” IEEE Transactions on Antennas and Propagation, vol. 56, no. 10, pp. 3150–3158, 2008.
View at: Publisher Site | Google Scholar
C. Zhou, Z. Shi, Y. Gu, and X. Shen, “DECOM: DOA estimation with combined MUSIC for coprime array,” IEEE International Conference on Wireless Communication and Signal Processing (WCSP), pp. 1–5, 2013.
View at: Publisher Site | Google Scholar
W. Zheng, X. F. Zhang, P. Gong, and H. Zhai, “DOA estimation for coprime linear arrays: an ambiguity-free method involving full DOFs,” IEEE Communication Letters, vol. 22, no. 3, pp. 562–565, 2018.
View at: Publisher Site | Google Scholar
F. Sun, B. Gao, and P. Lan, “Partial spectral search based DOA estimation method for co-prime linear arrays,” Electronics Letters, vol. 51, no. 24, pp. 2053–2055, 2015.
View at: Publisher Site | Google Scholar
D. Zhang, Y. Zhang, G. Zheng, C. Feng, and J. Tang, “Improved DOA estimation algorithm for co-prime linear arrays using root-MUSIC algorithm,” Electronics Letters, vol. 53, no. 18, pp. 1277–1279, 2017.
View at: Publisher Site | Google Scholar
Q. Wu, F. Sun, P. Lan, G. Ding, and X. Zhang, “Two-dimensional direction-of-arrival estimation for co-prime planar arrays: a partial spectral search approach,” IEEE Sensors Journal, vol. 16, no. 14, pp. 5660–5670, 2016.
View at: Publisher Site | Google Scholar
Z. Wang, Z. Xiaofei, and S. Zhan, “Two-dimensional direction of arrival estimation for coprime planar arrays via a computationally efficient one-dimensional partial spectral search approach,” IET Radars and Sonar Navigation, vol. 11, no. 10, pp. 1581–1588, 2017.
View at: Publisher Site | Google Scholar
W. Zheng, X. Zhang, and H. Zhai, “Generalized coprime planar array geometry for 2-D DOA estimation,” IEEE Communications Letters, vol. 21, no. 5, pp. 1075–1078, 2017.
View at: Publisher Site | Google Scholar
L. Xu, J. Li, and P. Stoica, “Target detection and parameter estimation for MIMO radar systems,” IEEE Transactions on Aerospace and Electronic Systems, vol. 44, no. 3, pp. 927–939, 2008.
View at: Publisher Site | Google Scholar
L. Wan, G. Han, J. Jiang, J. J. P. C. Rodrigues, N. Feng, and T. Zhu, “DOA estimation for coherently distributed sources considering circular and noncircular signals in massive MIMO systems,” IEEE Systems Journal, vol. 11, no. 1, pp. 41–49, 2017.
View at: Publisher Site | Google Scholar
D. Nion and N. D. Sidiropoulos, “Adaptive algorithms to track the PARAFAC decomposition of a third order tensor,” IEEE Transactions on Signal Processing, vol. 57, no. 6, pp. 2299–2310, 2009.
View at: Publisher Site | Google Scholar
Y. Doi, K. Ichuge, H. Arai et al., “3-D array configuration using multiple regular tetrahedra for high-resolution 2-D DOA estimation,” IEEE. Signal Processing Conference, pp. 1955–1959, 2014, http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=6952711&isnumber=6951911.
View at: Google Scholar
H. Moriya, K. Ichige, H. Arai, T. Hayashi, H. Matsuno, and M. Nakano, “Improving elevation estimation accuracy in DOA: how planar arrays can be modified into 3-D configuration,,” IEICE Transactions on Fundamentals, vol. E95.A, no. 10, pp. 1658–1666, 2012.
View at: Google Scholar
D. Bernard, Principles of Aperture and Array System Design, Wiley, Hoboken, NJ, USA, 1976.
J. Liu, Y. Zhang, Y. Lu et al., “Augmented nested arrays with enhanced DOF and reduced mutual coupling,” IEEE Transactions on Signal Processing, vol. 65, no. 21, pp. 5549–5563, 2017.
View at: Publisher Site | Google Scholar
D. Zhang, Y. Zhang, G. Zheng, B. Deng, C. Feng, and J. Tang, “Two-dimensional direction of arrival estimation for coprime planar arrays via polynomial root finding technique,” IEEE Access, vol. 6, pp. 199540–219549, 2018.
View at: Publisher Site | Google Scholar
D. Tse and P. Viswanath, Fundamentals of Wireless Communication, Cambridge University Press, Cambridge, UK, 2005.
P. Stoica and A. Nehorai, “MUSIC, maximum likelihood, and Cramer-Rao bound,” IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. 37, no. 5, pp. 720–741, 1989.
View at: Publisher Site | Google Scholar
P. Stoica and A. Nehorai, “MUSIC, maximum likelihood, and Cramer-Rao bound: further results and comparisons,” IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. 38, no. 12, pp. 2140–2150, 1990.
View at: Publisher Site | Google Scholar
P. Stoica and A. Nehorai, “Performance study of conditional and unconditional direction-of-arrival estimation,” IEEE Transactions on Acoustics, Speech, and Signal Processing, vol. 38, no. 10, pp. 1783–1795, 1990.
View at: Publisher Site | Google Scholar

Copyright

Copyright © 2020 Pan Gong et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

PDF Download Citation

Download other formats

Order printed copies

Views

937

Downloads

1297

Citations