Downlink channel recovery method for asymmetric large-scale MIMO system under UPA

Abstract

The invention provides a downlink channel recovery method for an asymmetric large-scale MIMO system under UPA. The method comprises the following steps: S1, according to the distribution of an asymmetric array, carrying out zero padding on a received signal, defining a mixed angle and constructing a virtual array; s2, designing an antenna selection algorithm on the premise of ensuring the UPA first row array aperture, and constructing a maximum aperture virtual array with as few antennas as possible; s3, respectively executing improved Newton orthogonal matching pursuit on receiving signals of the virtual array and the first row of arrays, and obtaining all possible path gains, azimuth angles and elevation angles until the gains, azimuth angles and elevation angles respectively meet threshold values; s4, combining and pairing the estimated path gain, elevation angle and azimuth angle according to related criteria; and S5, recovering a downlink channel by using the obtained path gain and angle parameter pair and the downlink steering vector. According to the method, part of channel information contained in the virtual array is fully utilized, and the downlink channel recovery performance is effectively improved while the number of received uplink radio frequency chains and the calculation complexity are remarkably reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communication, and particularly relates to a downlink channel recovery method for an asymmetric massive MIMO system under UPA. Background Art

[0002] Massive multiple-input multiple-output (MIMO) technology deploys a large-scale antenna array at the base station side and uses spatial degrees of freedom to achieve spatial division multiplexing of multi-user signals, which can significantly improve the system spectral efficiency, energy efficiency, and link reliability, and has become one of the key enabling technologies for 5G / 6G (Fifth Generation / Sixth Generation) communication systems.

[0003] In actual communication scenarios, the requirements for uplink data transmission are different from those for downlink data transmission. At the same time, the data transmission rate is highly correlated with the number of radio frequency (RF) chains. The more RF chains, the higher the achievable transmission rate; however, the higher the circuit complexity and cost that need to be addressed. Therefore, in a fully digital architecture, configuring a complete RF chain for each antenna will result in a linear increase in hardware complexity, cost, and power consumption with the number of antennas, making it difficult to support massive MIMO. For a hybrid beamforming architecture, although using a small number of RF chains to connect more antennas through analog components (such as phase shifters) can reduce the hardware cost compared to a fully digital architecture, the analog phase shifters can only achieve limited beamforming degrees of freedom (such as adjusting the phase but unable to independently control the amplitude), resulting in limited beamforming capabilities in the downlink. When multiple users need to be served simultaneously or high-precision beams are required, the shortage of RF chains will become a performance bottleneck. Especially in a large-scale uniform planar array (UPA) system in the millimeter-wave / terahertz band, the high-dimensional matrix operations and data throughput requirements of baseband signal processing severely restrict its actual deployment. In contrast, an asymmetric massive MIMO system decouples the transmit and receive RF chains, allowing fewer receive RF chains on the uplink while ensuring sufficient transmit RF chains, so as to reduce hardware complexity and cost, and has significant advantages in improving system performance, reducing costs, and enhancing signal reliability.

[0004] In an asymmetric massive MIMO system under UPA, full-dimension channel recovery is the core to ensure downlink transmission performance. However, the significant difference in the number of transmit and receive antennas at the base station end leads to a structural mismatch in the channel dimensions between the uplink and downlink, rendering the method that relies on channel reciprocity to implement the mapping of uplink and downlink channel state information (CSI) in a traditional symmetric transceiver architecture time-division duplex system ineffective. Therefore, channel recovery has become a key technology for reconstructing the complete downlink CSI in an asymmetric transceiver system.

[0005] Traditional channel recovery / estimation methods mainly focus on uniform linear arrays, while actual systems usually employ uniform planar arrays. Due to the inherent high-dimensional nature of UPA, directly extending these uniform linear array (ULA) estimation methods to UPA will result in a significant increase in computational complexity. In addition, due to the simultaneous presence of azimuth and elevation angles, when using traditional ULA estimation methods, different combinations of azimuth and elevation angles may correspond to the same received signal, leading to angle ambiguity. Moreover, in a UPA configuration, since a higher density of users needs to be served in the horizontal direction, the number of antennas in the horizontal direction is usually more than that in the vertical direction. This hinders the estimation performance of the elevation angle and further exacerbates the angle ambiguity problem. Summary of the Invention

[0006] The object of the present invention is to propose a downlink channel recovery method for an asymmetric massive MIMO system under UPA to solve the downlink channel recovery problem of the asymmetric massive MIMO system under UPA. This method solves the angle ambiguity problem and high computational complexity problem of UPA parameter estimation with as few RF chain numbers as possible, and effectively improves the downlink channel recovery performance of the asymmetric massive MIMO system under UPA.

[0007] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0008] A downlink channel recovery method for an asymmetric massive MIMO system under UPA, the asymmetric massive MIMO system under UPA includes a base station configured with an asymmetric massive MIMO transceiver with a uniform planar array and several single-antenna users; the system operates in a time-division duplex mode and can perform uplink and downlink transmissions within a coherence interval, and the signal propagation environment between the users and the base station is a far-field plane wavefront propagation environment; the downlink channel recovery method includes the following steps:

[0009] S1. Preprocessing: Zero-pad the received signal according to the distribution of the asymmetric array, and calculate the correlation matrix of the zero-padded received signal obtained from multiple snapshots; define the mixed angle as a linear combination of the elevation angle and the azimuth angle, and construct a virtual array around the mixed angle;

[0010] S2, Antenna Selection: First, ensure the aperture of the first row of the UPA antenna array. Search for virtual array elements in the correlation matrix, determine the optimal mixing angle and the corresponding virtual array, find the position of the single-array antenna corresponding to the virtual array element, until all virtual array elements are searched, construct the virtual array with the largest aperture, and determine the minimum number of required antennas and their corresponding positions;

[0011] S3, Parameter Estimation: Perform improved Newtonian orthogonal matching pursuit on the received signals of the virtual array with the largest aperture constructed in step S2 and the first row of the UPA antenna array respectively, to obtain all possible path parameters: path gain, azimuth angle, and elevation angle, until the threshold is met respectively;

[0012] S4, Path Matching: Adopt a greedy strategy to iteratively pair the angle parameter and the path gain parameter combination obtained in step S3 by minimizing the loss function;

[0013] S5, Channel Recovery: Recover the downlink channel according to the path parameter combination obtained in step S4 and the downlink steering vector.

[0014] Furthermore, the base station adopts a fully digital transceiver structure, and the number of base station antennas is N, where each antenna is connected to an independent transmit radio frequency chain, but only M antennas can be connected to the receive radio frequency chain; the user terminal adopts orthogonal pilot sequences in the uplink, and the uplink channel estimation for each user can be performed separately, and the estimation method adopts least squares estimation or linear minimum mean square error estimation, M << N, and both M and N are positive integers.

[0015] Furthermore, the zero-padding of the received signal in step S1 is specifically: zero-padding the received signal y k of the k-th user to obtain the received signal y k,S , such that the dimension of the zero-padded received signal y k,S is the same as the number of downlink antennas deployed at the base station, which is N, where the received signals at the remaining N - M antennas are zero; the correlation matrix: the correlation matrix is obtained by calculating T snapshots The defined mixing angle is: γ k,l = asinθ k,l + bcosθ k,l cosφ k,l , where γ k,l is the mixing angle of the l-th path of the k-th user, θ k,l is the elevation angle of the l-th path of the k-th user, φ k,lis the azimuth angle of the l-th path of the k-th user; the virtual array: according to the element phases of the UPA steering vector, linearly combines the elevation angle and the azimuth angle by undetermined integers a and b, takes the elements containing the mixed angle and its integer multiples in the correlation matrix as each virtual array element, and constructs a virtual array accordingly. The apertures of the virtual arrays constructed with different parameters a and b are also different.

[0016] Further, step S2 specifically includes: ensuring that the maximum physical spacing of the first-row array remains unchanged, thereby ensuring the maximum array aperture; searching in the correlation matrix R k,S for the integers a and b that make the mixed angle have the maximum aperture among all possible virtual arrays, and thus determining |a| = 1, |b| = 1 to form the optimal mixed angle; searching for each virtual array element containing this mixed angle in the correlation matrix, taking the mean value of each same virtual array element in the correlation matrix R k,S according to the Toeplitz characteristics of the block Toeplitz of the correlation matrix and its sub-matrices, and finding the single array antenna position corresponding to this virtual array element, that is, the antenna at the UPA anti-diagonal direction position, until all virtual array elements containing the mixed angle are searched. While ensuring as few RF chain numbers as possible, a virtual array with the maximum aperture is constructed.

[0017] Further, step S3 specifically includes: performing linear minimum mean square error estimation on the received signals of the first-row antennas in the horizontal direction of the UPA, respectively using them and the constructed virtual array with the maximum aperture as input signals for path detection, performing fast Fourier transform to obtain the path gain, mixed angle, and azimuth angle of the rough estimation, and then successively performing local and global Newton refinement steps and least squares to obtain the final parameter estimation until the threshold is satisfied respectively, and this threshold is determined by the number of base station uplink antennas and the transmission signal-to-noise ratio; when the improved Newtonized orthogonal matching pursuit cycle executes the Newton refinement step, the global refinement times and local refinement times are respectively default set to 3 and 1.

[0018] Further, step S4 specifically includes: adopting a greedy strategy, subtracting the recovered channel from the uplink channel and calculating the L2 norm as the loss function, and iteratively pairing to obtain the combination of path gain and angle parameters by minimizing the loss function.

[0019] Beneficial effects: A downlink channel recovery method for an asymmetric large-scale MIMO system under UPA provided by the present invention has the following advantages:

[0020] 1. The present invention does not need to increase the physical aperture of the UPA vertical direction antennas, and only needs a small number of RF chains to improve the array elevation resolution and the downlink channel recovery performance.

[0021] 2. The present invention can make full use of the correlation matrix information of the received signal, only select a small number of antennas to further reduce the computational complexity brought by the inherent high-dimensional characteristics of the UPA, and at the same time use the path matching step to solve the angle ambiguity problem. Description of the Drawings

[0022] Figure 1 Schematic diagram of a large-scale MIMO system based on a UPA asymmetric transceiver according to an embodiment of the present invention;

[0023] Figure 2 Schematic flow chart of an embodiment of the present invention. Detailed Embodiments

[0024] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the embodiments and the drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments.

[0025] Generally, the components of the embodiments of the present invention described and illustrated herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0026] Embodiment

[0027] This embodiment provides a downlink channel recovery method for an asymmetric large-scale MIMO system under a UPA. As Figure 1 shown, the asymmetric large-scale MIMO system under the UPA consists of a base station and multiple single-antenna users. The array deployed at the base station is a uniform planar array, the number of antenna elements included in the array is N, but there are only M receiving radio frequency chains, and N≠M.

[0028] The base station serves K single-antenna users simultaneously. The system operates in the TDD mode and can perform uplink and downlink transmissions within the coherence interval. Therefore, the uplink and downlink channel reciprocity holds. When considering the block fading parameter channel model, due to the channel reciprocity, the uplink and downlink channels of user k in the large-scale MIMO system based on the asymmetric transceiver can be respectively expressed as:

[0029]

[0030] where L k represents the number of signal paths of user k; g k,l is the channel complex gain of the l-th path of the k-th user and satisfies is the path power; θk,l and φ k,l represent the elevation angle and azimuth angle of the l-th path respectively. is the steering vector of the uplink channel, is the steering vector of the downlink channel:

[0031]

[0032] where d is the antenna spacing, λ is the carrier wavelength, and N (ν) represents the number of downlink vertical antennas, and N (h) represents the number of downlink horizontal antennas, and a i ∈{0, 1, 2, …, N (ν) -1} and b j ∈{0, 1, 2, …, N (h) -1} represent the antenna coefficients, represents the Kronecker product.

[0033] During the uplink pilot training, the user sends an orthogonal pilot sequence Ι. Let the length of the pilot sequence be K, then

[0034]

[0035] where, is the received signal at the base station side, ρ τ is the pilot power, is the complex Gaussian white noise, is the noise power. represents the uplink channel.

[0036] When using linear minimum mean square error channel estimation, the uplink channel estimation can be expressed as

[0037]

[0038] where, represents the uplink channel estimation of user k, represents the channel correlation matrix, represents the expectation operation.

[0039] According to the statistical characteristics of the received signal of user k, the correlation matrix R can be obtained k

[0040]

[0041] where,

[0042] The downlink channel recovery method for the asymmetric large-scale MIMO system under UPA proposed in the present invention is carried out based on this channel correlation matrix. As Figure 2As shown in the figure, a downlink channel recovery method for an asymmetric large-scale MIMO system under UPA provided by an embodiment of the present invention includes the following steps:

[0043] S1. Preprocessing: Padding zeros to the received signal to make the dimension of the vector consistent with the number of downlink antennas deployed at the base station, which is N. Among them, the received signals at the remaining N - M downlink antennas are zero.

[0044]

[0045] Among them, represents the set cardinality of.

[0046] According to the statistical characteristics of the received signal after zero-padding,

[0047]

[0048] S2. Antenna selection: From the correlation matrix R k,S it can be obtained that the maximum aperture of the virtual array formed by the mixing angles is Cλ, and it satisfies

[0049] Ca ≤ (N (ν) -1),(10)

[0050] Cb ≤ (N (h) -1),(11)

[0051] Therefore, and |a| ≥ 1, |b| ≥ 1. And when |a| = 1, |b| = 1, the mixing angle formed has the maximum aperture among all possible virtual arrays. Search each virtual array element. According to the block Toeplitz property of the correlation matrix and the Toeplitz property of its sub-matrices

[0052]

[0053] Among them, R k,S contains N (ν) × N (ν) sub-matrices, and each sub-matrix R k,S (i,j)(1 ≤ i, j ≤ N (ν) ) contains N (h) × N (h) elements. Take the average value of the values at the positions of the same virtual array elements in R k,S and find the single array antenna position corresponding to this virtual array element until all virtual array elements are searched. While ensuring as few RF chains as possible, construct a virtual array with the maximum aperture, that is:

[0054] S3. Parameter Estimation: The maximum aperture virtual array constructed and the received signals from only one ULA in the horizontal direction of the UPA are respectively expressed as:

[0055]

[0056] Among them, denotes a vector composed of the first M2 elements, where M2 is the number of ULA antennas in the horizontal direction.

[0057] i) Detection gain and spatial angle

[0058]

[0059] Among them, g i denotes the i-th element of g, ζ denotes the oversampling factor, and F denotes the N a ζ-dimensional FFT matrix. M a and N a respectively denote the number of non-zero elements of y r . The updated residual vector is:

[0060]

[0061] Among them, a B (·) respectively denote a γ (·) and a S (·)

[0062]

[0063]

[0064] Among them,

[0065] ii) Newton refinement gain and spatial angle

[0066] Using the Newton method to refine w max and g max , the objective function is:

[0067]

[0068] Refining w max and g max , we get:

[0069]

[0070] Among them, Re{·} denotes taking the real part of the complex value,

[0071]

[0072] S4. Path Matching: Adopt a greedy strategy to minimize the loss function Iteratively pair the path gain, elevation angle, and azimuth angle parameter combinations refined by Newton

[0073]

[0074] S5. Channel Restoration: According to the path parameter combination and the downlink steering vector a D (·) restore the downlink channel:

[0075]

[0076] where is the number of matched paths.

Claims

1. A downlink channel recovery method for an asymmetric large-scale MIMO system under UPA, characterized in that The UPA-based asymmetric massive MIMO system includes a base station equipped with an asymmetric massive MIMO transceiver configured with a uniform planar array and several single-antenna users; the system operates in a time-division duplex mode and can perform uplink and downlink transmissions within a coherence interval, and the signal propagation environment between the users and the base station is a far-field plane wavefront propagation environment; the downlink channel recovery method includes the following steps: S1. Preprocessing: Zero-pad the received signal according to the distribution of the asymmetric array, and calculate the correlation matrix of the zero-padded received signal obtained from multiple snapshots; define the mixing angle as a linear combination of the elevation angle and the azimuth angle, and construct a virtual array around the mixing angle; S2. Antenna selection: First, ensure the aperture of the first row antenna array of the UPA. Search for the virtual array elements in the correlation matrix and determine the optimal mixing angle and the corresponding virtual array. Find the positions of the single-array antennas corresponding to the virtual array elements until all virtual array elements are searched. Construct the virtual array with the largest aperture and determine the required minimum number of antennas and the corresponding positions; S3. Parameter estimation: Perform the improved Newtonized orthogonal matching pursuit on the received signals of the virtual array with the largest aperture constructed in step S2 and the first row antenna array of the UPA respectively to obtain all possible path parameters: path gain, azimuth angle, and elevation angle, until the threshold thresholds are satisfied respectively; S4. Path matching: Adopt a greedy strategy to iteratively pair the angle parameter and the path gain parameter combination obtained in step S3 by minimizing the loss function; S5. Channel recovery: Recover the downlink channel according to the path parameter combination obtained in step S4 and the downlink steering vector.

2. The downlink channel recovery method for an asymmetric large-scale MIMO system under UPA according to claim 1, wherein The base station adopts a full-digital transceiver structure. The number of base station antennas is N, and each antenna is connected to an independent transmit radio frequency chain, but only M antennas can be connected to the receive radio frequency chain; the user side adopts orthogonal pilot sequences in the uplink, and the uplink channel estimation for each user can be performed separately, and the estimation method adopts least squares estimation or linear minimum mean square error estimation, M << N, and both M and N are positive integers.

3. The downlink channel recovery method for an asymmetric large-scale MIMO system under UPA according to claim 1, characterized in that The zero-padding of the received signal described in step S1 is specifically as follows: For the received signal y of the k-th user k zero-padding is performed to obtain the received signal y k,S such that the dimension of the zero-padded received signal y k,S is the same as the number of downlink antennas deployed at the base station, which is N, where the received signals at the remaining N - M antennas are zero; The correlation matrix: The correlation matrix is obtained by calculating T snapshots represents the vector of the received signal at the base station after zero-padding for the t-th snapshot, ∑ represents the summation operation, and (·) H represents the conjugate transpose operation; The mixing angle is defined as: γ k,l = asinθ k,l + bcosθ k,l cosφ k,l , where γ k,l is the mixing angle of the l-th path of the k-th user, θ k,l is the elevation angle of the l-th path of the k-th user, φ k,l is the azimuth angle of the l-th path of the k-th user, and a and b are undetermined integers; The virtual array: According to the element phases of the UPA steering vector, the elevation angle and the azimuth angle are linearly combined according to the undetermined integers a and b, and the elements in the correlation matrix that contain the mixing angle and its integer multiples are used as each virtual array element, and a virtual array is constructed accordingly. The virtual array apertures constructed according to different parameters a and b are also different.

4. The downlink channel recovery method for an asymmetric large-scale MIMO system under UPA according to claim 1, characterized in that The specific steps of step S2 include: ensuring that the maximum physical spacing of the first row array remains unchanged to guarantee the maximum array aperture; searching in the correlation matrix R k,S for integers a and b such that the mixing angle has the maximum aperture among all possible virtual arrays, and thereby determining that |a| = 1 and |b| = 1 to form the optimal mixing angle; searching for each virtual array element containing this mixing angle in the correlation matrix, and taking the average value of each identical virtual array element in the correlation matrix R k,S according to the block Toeplitz property of the correlation matrix and the Toeplitz property of its sub-matrices, and finding the single array antenna position corresponding to this virtual array element, that is, the antenna located in the UPA anti-diagonal direction position, until all virtual array elements containing the mixing angle are searched. While ensuring as few RF chain numbers as possible, a virtual array with the maximum aperture is constructed 5. The downlink channel recovery method for an asymmetric large-scale MIMO system under UPA according to claim 1, characterized in that The step S3 specifically includes: performing linear minimum mean square error estimation on the received signals of the first row of antennas in the horizontal direction of the UPA, and using the constructed virtual array with the largest aperture and the received signals as input signals for path detection respectively. Perform fast Fourier transform to obtain the coarsely estimated path gain, mixing angle, and azimuth angle, and then sequentially perform local and global Newton refinement steps and least squares to obtain the final parameter estimation until the threshold thresholds are satisfied respectively, and the threshold thresholds are determined by the number of base station uplink antennas and the transmit signal-to-noise ratio; when the improved Newtonized orthogonal matching pursuit cycles, the number of global refinement times and the number of local refinement times are default set to 3 and 1 respectively when performing the Newton refinement steps.

6. The downlink channel recovery method for an asymmetric large-scale MIMO system under UPA according to claim 1, characterized in that The step 4 specifically includes: adopting a greedy strategy, subtracting the recovered channel from the uplink channel and calculating the L2 norm as the loss function, and iteratively pairing the path gain and the angle parameter combination by minimizing the loss function.

Images